Deliverable 4.1 Safeland

Grant Agreement No.
: 226479
SafeLand
Living with landslide risk in Europe: Assessment,
effects of global change, and risk management strategies
7th Framework Programme

Cooperation Theme 6 Environment (including climate change)
Sub-Activity 6.1.3 Natural Hazards
Deliverable 4.1
Review of Techniques for Landslide Detection, Fast Characterization, Rapid
Mapping and Long-Term Monitoring
Work Package 4.2 Remote Sensing Technologies for Landslide Detection,
Monitoring and Rapid Mapping
Deliverable Leader: Unil
Revision: 2 Approved
February 2012
Rev.
0
1
2
Deliverable Responsible
UNIL
UNIL
UNIL (updated version)
Controlled by
Specialized research institutes
ICG-NGI
Date
Summer 2010
October 2010
February 2012
SafeLand deliverable 4.1
Rev. No: 2
Date: 2012-02-15
SUMMARY
One of the main aims of the European project SafeLand in the 7th Framework Program is to
develop innovative mapping and monitoring methods in order to improve the methods for
regional landslide risk assessment and design of early warning systems. More specifically,
Area 4 of SafeLand addresses the technical and practical issues related to monitoring and
early warning for landslides. During the last decade, different monitoring and remote sensing
techniques have undergone rapid development. In order to summarize these scientific and
technical advances, the University of Lausanne, in close collaboration with 18 European
institutions, has produced the project Deliverable 4.1: Review of Techniques for Landslide
Detection, Fast Characterization, Rapid Mapping and Long-Term Monitoring1.
The core of this deliverable consists of two main chapters:
(a)
Chapter two summarizes, from a theoretical point of view, the different ground-based
and remote sensing techniques currently used to investigate, map and monitor
landslides. Each technique described in the deliverable has benefited from the
expertise of specialized research groups.
(b)
Chapter three shows the main applications of the remote sensing techniques to
landslides through the synthesis of seventeen case studies.
This review aims to represent a common reference for the different deliverables of SafeLand
Area 4. In addition to being a state-of-the art overview, this deliverable provides helpful and
extensive support for non-specialists and students interested in the application of new
techniques to different mass movements such as landslides, rockfall, etc.
To quote this document:
SafeLand deliverable 4.1, 2010. Review of Techniques for Landslide Detection, Fast Characterization, Rapid
Mapping and Long-Term Monitoring. Edited for the SafeLand European project by Michoud C., Abelln A.,
Derron M.-H. and Jaboyedoff M. Available at http://www.safeland-fp7.eu
Grant Agreement No.: 226479

SafeLand - FP7
Page 3 of 401
Rev. No: 2
Date: 2012-02-15
Note about contributors

The following organisations contributed to the work described in this deliverable:
Lead partner responsible for the deliverable:

Institut de Gomatique et dAnalyse du Risque, Universit de Lausanne (UNIL)
Deliverable prepared by:
Clment Michoud, Antonio Abelln, Marc-Henri Derron, Michel Jaboyedoff
Partner responsible for quality control:

Revision 0 Chapter 2 Reviewers:
All sections of the chapter 2, which summarizes the different techniques and methods,
were written and reviewed by at least two specialized experts of independent

institutes. The reviews guarantee the quality of this chapter.
Revision 1 - Deliverable General Reviewer:
Farrokh Nadim (ICG-NGI)
Main contributors:
knes/Tafjord Early Warning Centre (IKS): L.H. Blikra, L. Kristensen
AMRA Scarl: G. Fornaro
Bureau de Recherches Gologiques et Minires (BRGM): G. Grandjean, M. de Michele, D.
Raucoules
Centre National de la Recherche Scientifique, Institut de Physique du Globe de Strasbourg
(CNRS- IPGS): J.-P. Malet, A. Tonnellier
Centre National de la Recherche Scientifique, Laboratoire de Godynamique Interne et de
Tectonophysique de Grenoble (CNRS-LGIT): D. Jongmans
Centro Servizi di Geoingegneria S.r.l. (CSG): E. Abbiate, M. Lovisolo
Geological Survey of Austria (GSA): I. Baron, R. Supper
Geological Survey of Slovenia (GeoZS): M. Carman, M. Jemec, S. Kumelj
International Center for Geohazards, Geological Survey of Norway (ICG-NGU): R.
Hermanns, J.-S. LHeureux, T. Oppikofer
International Center for Geohazards, NORSAR (ICG-NORSAR): I. Lecomte, M. Roth
International Center for Geohazards, Norwegian Geotechnical Institute (ICG-NGI): S.
Bazin, A. A. Pfaffhuber
International Center for Geohazards, University of Oslo (ICG-UiO): T. Eiken
International Institute for Geo-information Sciences and Earth Observation (ITC): N. Kerle ,
A. Schtumpf
Joint Research Center (JRC): M. van d. Eeckhaut, J. Hervs
Universit degli studi di Firenze (UNIFI): N. Casagli, F. Catani, F. Cigna, V. Tofani
Universit di Salerno (UNISA) : L. Cascini, D. Peduto
Universitat Politcnica de Catalunya (UPC): J. Gili
Universit de Lausanne (UNIL): A. Abelln, M.-H. Derron, F. Humair, M. Jaboyedoff, C.
Michoud

SafeLand - FP7
Page 4 of 401
Rev. No: 2
Date: 2012-02-15
Acknowledgements
On behalf of the SafeLand European project, the editors of the deliverable 4.1 would like to
thanks all authors and reviewers for their excellent and motivated contributions in the
conception and the redaction of this deliverable. The editors show their deep grateful to:
E. Abbiatte, S.E. Amran, I. Baron, J.I. Barredo S. Bazin, L.H. Blikra, J. Calvet, M. Carman,
N. Casagli, L. Cascini, F. Catani, F. Cigna, M. van den Eeckhaut, T. Eiken, G. Fornaro,
J. Gili, G. Grandjean, R. Hermanns, J. Hervs, P. Horton, F. Humair, M. Jemec, D. Jongmans,
N. Kerle, L. Kristensen, S. Kumelj, I. Lecomte, M. Lovisolo, J.-S. L'Heureux, A. Loye,
G. Luzi, J.-P. Malet, R. Metzger, M. de Michele, F. Nadim, T. Oppikofer, A. Pedrazzini,
D. Peduto, A.A. Pfaffhuber, J. Poesen, D. Raucoules, J.S. Rmming, P.I. Rosin, M. Roth,
A. Schtumpf, R. Supper, V. Tofani, A. Tonnellier, L. Vandekerckhove, J.M. Vilaplana.

SafeLand - FP7
Page 5 of 401
Rev. No: 2
Date: 2012-02-15
List of acronyms
2SM
A-DInSAR
AEM
ALOS
ALS
ANN
ASAR
ASTER
ATM
Bn
Bt
CCD
CMOS
CMP
CNES
CPT
CVA
DB
DDV
DEM
DGPS
DIC
DInSAR
DMC
DN
DOS
DSM
DTM
EM
EO
ERS
ERT
ESA
GB-InSAR
GCP
Two strata model

Advanced Differential Interferometric Synthetic Aperture Radar
Airborne electromagnetics
Advanced Land Observing Satellite
Aerial Laser Scanner
Artificial Neural Networks
Advanced Synthetic Aperture Radar
Advanced Spaceborne Thermal Emission and Reflection Radiometer
Airborne Thematic Mapper
Normal Baseline
Temporal Baseline
Charged Coupled Device
Complementary Metal Oxide Semiconductor
Common Mid-Point
Centre National d'tudes Spatiales
Cone Penetration Test
Change Vector Analysis
Dark-Bright Set
Dense Dark Vegetation
Digital Elevation Model
Differential Global Positioning System
Digital Image Correlation
Differential Interferometric Synthetic Aperture Radar
Disaster Monitoring Constellation
Digital Number
Dark Object Subtraction
Digital Surface Model
Digital Terrain Model
Electromagnetics
Exterior Orientation
European Remote Sensing
Electrical Resistivity Tomography
European Space Agency
Ground Based Interferometric Synthetic Aperture Radar
Ground Control Point

SafeLand - FP7
Page 7 of 401
GEOBIA
GEOSS
GIS
GLCM
GLOF
GMES
GNSS
GPR
GPS
GSD
HEM
HHS
HIS
HM
HRDEM
I
IGRF
HIS
InSAR
INSPIRE
IO
IP
IWLSR
JAXA
LIDAR
LOS
MAD
MASW
MDDV
MLC
MM
MNCR
MPGC
MS
MTF
NASA
NDVI
NIR
OBIA
Rev. No: 2
Date: 2012-02-15
Geographic Object-Based Image Analysis

Global Earth Observation System of Systems
Geographic Information System
Grey Level Co-Occurrence Matrices
Glacial-Lake Outburst Flow
Global Monitoring for Environment and Security
Global Navigation Satellite System
Ground Penetrating Radar
Global Positioning System
Ground Sampling Distance
Helicopter-borne electromagnetics
Homogenous Half-Space model
Hue-Saturation-Intensity
Histogram Matching
High Resoluton Digital Elevation Model
Intensity (of the laser signal)
International Geomagnetic Reference Field
Intensity-Hue-Saturation
Interferometry Synthetic Aperture Radar
Infrastructure for Spatial Information in Europe
Interior Orientation
Induced Polarization
Iteratively Weighted Least Squares Regression
Japan Aerospace Exploration Agency
Light Detection and Ranging
Line of Sight
Multivariate Alteration Detection
Multichannel Analysis of Surface Waves
Modified Dense Dark Vegetation
Maximum Likelihood Classification
Modified Minnaert
Manual No Change Regression
Multi-Photo Geometrically Constrained
Multispectral
Modulation Transfer Function
National Aeronautics and Space Administration
Normalized Difference Vegetation Index
Near Infrared
Object-Based Image Analysis

SafeLand - FP7
Page 8 of 401
OOA
PAN
PCA
PIF
PIFR
PIV
PSI
PS-InSAR
RADAR
RFM
RMSE
RPC
RT
SAM
SAR
SAT-PP
SBAS
SLR
SP
SPOT
SR
SRTM
SVM
TDI
TEM
TIN
TLS
UAV
UPN
UTM
VES
VHR
WGS
XYZ
Rev. No: 2
Date: 2012-02-15
Object-Oriented Analysis
Panchromatic
Principal Component Analysis
Pseudoinvariant Feature
Pseudoinvariant Feature Regression
Particle Image Velocimetry
Persistent Scatterer Interferometry
Permanent Scatterer InSAR
Radio Detection And Ranging
Rational Function Model
Root Mean Square Error
Rational Polynomial Coefficients
Roto-Translation
Spectral Angle Mapper
Synthetic Aperture Radar
Satellite Image Precision Processing
Small Baseline Subset
Single Lens Reflex
Self Potential
Satellite Pour l'Observation de la Terre
Simple Regression
Shuttle Radar Topographic Mapping
Support Vector Machines
transfer delay integration
Transient Electromagnetics
Triangle irregular Network
Terrestrial Laser Scanner
Unmanned Aerial Vehicle
Unchanged Pixel Normalization
Universal Transverse Mercator
Vertical Electric Sounding
Very High Resolution
World Geodetic System
3D coordinates

SafeLand - FP7
Page 9 of 401
Rev. No: 2
Date: 2012-02-15
CONTENTS
1.
INTRODUCTION .......................................................................................................................... 17
2.
DATA AND TECHNIQUES .......................................................................................................... 23
PART A - PASSIVE OPTICAL SENSORS ..................................................................................................... 25

1. GENERAL PRE-PROCESSING STEPS .......................................................................................................... 26
1.1.
Digitizing ...................................................................................................................................... 26
1.2.
Geometric correction ..................................................................................................................... 26
1.3.
Radiometric correction .................................................................................................................. 28
1.4.
Filtering and Texture ..................................................................................................................... 32
1.5.
Orthorectification .......................................................................................................................... 33
1.6.
Data fusion .................................................................................................................................... 34
2. GROUND-BASED IMAGING....................................................................................................................... 35
2.1.
Sensor and platform development ................................................................................................. 35
2.2.
Visual interpretation ...................................................................................................................... 36
2.3.
Ground-Based Digital Photogrammetry ........................................................................................ 36
2.4.
Digital Image Analysis of Ground-Based Photographs ................................................................ 37
2.5.
Analysis of videos ......................................................................................................................... 37
3. AERIAL IMAGING ..................................................................................................................................... 39
3.1.
Analogue vs. digital camera systems ............................................................................................ 39
3.2.
Panchromatic vs. Natural Color and Near Infrared ....................................................................... 40
3.3.
New sensor platforms .................................................................................................................... 41
3.4.
Visual interpretation ...................................................................................................................... 42
3.5.
Automated techniques of image interpretation .............................................................................. 43
4. SATELLITE IMAGING ................................................................................................................................ 44
4.1.
Sensors and platform development ............................................................................................... 44
4.2.
Visual Interpretation...................................................................................................................... 48
4.3.
Stereoprocessing ........................................................................................................................... 48
4.4.
Supervised classification techniques ............................................................................................. 52
4.5.
Unsupervised classification and clustering techniques.................................................................. 53
4.6.
Change detection ........................................................................................................................... 54
4.7.
Digital Image Correlation (DIC) ................................................................................................... 56
4.8.
Object-oriented concepts ............................................................................................................... 56
5. APPENDIX ................................................................................................................................................ 63
PART B - ACTIVE OPTICAL SENSORS ....................................................................................................... 65
1. BASIC PRINCIPLES OF LASER RANGING ....................................................................................... 66
1.1.
Distance calculation ...................................................................................................................... 66
2. ELECTRONIC DISTANCE METERS................................................................................................... 68
2.1.
General overview of the sensor ..................................................................................................... 68
2.2.
Method .......................................................................................................................................... 71
2.3.
Main commercial systems ............................................................................................................. 72
2.4.
Output............................................................................................................................................ 73
2.5.
Main applications .......................................................................................................................... 74
3. TERRESTRIAL LASER SCANNER ..................................................................................................... 76
3.1.
3.2.
Method .......................................................................................................................................... 79
3.3.
Main commercial systems ............................................................................................................. 83
SafeLand - FP7
Page 11 of 401
Rev. No: 2
Date: 2012-02-15
3.4.
Output............................................................................................................................................ 86
3.5.
4. AERIAL LASER SCANNER ........................................................................................................................ 91
4.1.
4.2.
Method .......................................................................................................................................... 94
4.3.
Output............................................................................................................................................ 97
4.4.
PART C - ACTIVE MICROWAVE SENSORS ............................................................................................ 103
1. RADAR WAVE PRINCIPLES...................................................................................................................... 105
1.1.
Introduction ................................................................................................................................. 105
1.2.
Range and Phase Shift ................................................................................................................. 106
2. INTERFEROMETRIC RADAR DISTANCE-METER ....................................................................................... 107
3. DIFFERENTIAL SAR INTERFEROMETRY ................................................................................................. 108
3.1.
Synthetic Aperture Radar ............................................................................................................ 109
3.2.
SAR Interferometry (InSAR) ...................................................................................................... 115
3.3.
Differential SAR Interferometry (DInSAR) ................................................................................ 125
4. MULTI-TEMPORAL REPEAT-PASS INSAR (ADVANCED DINSAR) .......................................................... 131
4.1.
Introduction ................................................................................................................................. 131
4.2.
PS-based A-DInSAR techniques or PSI (full resolution / dominant scatterering) ...................... 134
4.3.
Interferogram stacking techniques (low resolution / distributed scatterering) ............................ 146
4.4.
Final remarks............................................................................................................................... 150
5. GROUND-BASED INSAR........................................................................................................................ 151
5.1.
Introduction ................................................................................................................................. 151
5.2.
The GBInSAR instrumentation ................................................................................................... 152
5.3.
Characteristics of application ...................................................................................................... 154
6. POLARIMETRY FOR SOIL MOISTURE ESTIMATION ................................................................................... 158
6.1.
Introduction ................................................................................................................................. 158
6.2.
Soil moisture estimation .............................................................................................................. 159
PART D - GEOPHYSICAL INVESTIGATIONS ......................................................................................... 165
1. GROUND-BASED GEOPHYSICAL INVESTIGATIONS ................................................................. 167
1.1.
Introduction ................................................................................................................................. 167
1.2.
Methods ....................................................................................................................................... 171
1.3.
Discussion ................................................................................................................................... 193
1.4.
Conclusions ................................................................................................................................. 195
2. ACOUSTIC AND MICRO-SEISMIC MONITORING ........................................................................ 195
2.1.
Introduction ................................................................................................................................. 195
2.2.
The pioneering period identification of AE signals and limitations of the tecnnique .............. 200
2.3.
The recent period micro-seismic monitoring and characterization of AE signals .................... 203
2.4.
Conclusions ................................................................................................................................. 212
3. INTRODUCTION TO OFFSHORE METHODS ................................................................................. 213
3.1.
Multi-beam systems .................................................................................................................... 213
3.2.
2D and 3D high resolution seismic ............................................................................................. 216
3.3.
Summary and conclusions ........................................................................................................... 218
4. STAKES OF THE AERIAL GEOPHYSIC .......................................................................................... 218
PART E - GEOTECHNICAL GROUND-BASED MONITORING SYSTEMS ......................................... 221
1. INTRODUCTION ...................................................................................................................................... 222
2. EXTENSOMETERS ............................................................................................................................ 222
2.1.
Probe extensometers.................................................................................................................... 222
2.2.
Fixed borehole extensometers ..................................................................................................... 222
2.3.
Wire extensometers ..................................................................................................................... 223
3. INCLINOMETERS .............................................................................................................................. 224
3.1.
Probe inclinometers ..................................................................................................................... 224
3.2.
Automatic probe inclinometer system ......................................................................................... 227
3.3.
In-place inclinometers ................................................................................................................. 228

SafeLand - FP7
Page 12 of 401
Rev. No: 2
Date: 2012-02-15
4.
PIEZOMETERS ................................................................................................................................... 229

4.1.
Observation wells ........................................................................................................................ 229
4.2.
Open standpipe piezometers........................................................................................................ 230
4.3.
Twin-tube hydraulic piezometers ................................................................................................ 232
4.4.
Pneumatic, vibrating wire and electrical resistance piezometers ................................................ 233
5. CONTACT EARTH PRESSURE CELLS ............................................................................................ 235
6. MULTIPARAMETRIC IN PLACE SYSTEMS ................................................................................... 236
6.1.
DMS system ................................................................................................................................ 236
PART F - OTHERS TECHNIQUES ............................................................................................................... 239
1. GLOBAL NAVIGATION SATELLITES SYSTEMS (GNSS, GPS) ............................................................. 240
1.1.
Introduction ................................................................................................................................. 240
1.2.
Basic principle............................................................................................................................. 240
1.3.
Field procedures .......................................................................................................................... 242
1.4.
Some considerations on the use of GNSS in landslide monitoring and its precision .................. 244
1.5.
Examples ..................................................................................................................................... 246
1.6.
Trends.......................................................................................................................................... 246
2. INTRODUCTION TO CORE LOGGING ....................................................................................................... 247
3.
APPLICATION TO LANDSLIDES: CASE STUDIES ................................................................ 253

1. Monitoring Tessina landslide, Italy, from optical remotely sensed imagery. ........................................ 255
2. Landslide mapping by textural analysis of ATM data: a case study from the LOS Vlez district, south
eastern Spain.............................................................................................................................................. 261
3. Using TLS for rockslide monitoring. Case study of the Eiger collapse, Swiss Alps. ............................ 267
4. Structural analysis using color coded information derived from 3D point clouds................................. 275
5. Using TLS for rockfall detection and prediction. .................................................................................. 277
6. Susceptibility assessment at regional scale based on areal LIDAR datasets. Application to the canton of
Vaud, western Switzerland. ....................................................................................................................... 281
7. The application of LIDAR-derived images for landslide inventory and susceptibility mapping in the
Flemish Ardennes, Belgium. ..................................................................................................................... 287
8. Characterization and analyses of surface deformation, mass movements and alpine region lifting with
PSInSAR techniques/methods. .................................................................................................................. 293
9. Landslide detection and mapping at the basin scale by means of radar-interpretation. ......................... 297
10. Integration of PSInSAR data and ground-based instrument measurements for landslide characterization
and monitoring at the local scale. .............................................................................................................. 303
11. A new approach to the use of DInSAR data in landslide studies at different scales: the case study of
National Basin Authority of Liri-Garigliano and Volturno rivers. ............................................................ 307
12. Microseismic monitoring at the unstable rock slope at knes, Norway.............................................. 313
13. Landslide of Rindberg / Sibratsgfll investigated with airborne geophysical survey. ......................... 319
14. Site-specific geotechnical investigations for landslide characterization and detection. Application at the
knes rockslide, western Norway. ............................................................................................................ 323
15. From early warning to site-specific kinematic analysis: the case of Bagnaschino landslide. .............. 329
16. Detecting and mapping landslides in Salazie area, La Runion, using remote sensing techniques. .... 335
17. Vallcebre landslide, Spain, monitoring, from wire extensometers to GB-InSAR. .............................. 339
4.
DISCUSSION.............................................................................................................................. 347
4.1.
Advantages and limitations of each method ................................................................................ 347
4.2.
Appropriate techniques according to landslide situation ............................................................. 352
4.3.
Innovative techniques and further developments ........................................................................ 353
5.
CONCLUSIONS ......................................................................................................................... 359
6.
REFERENCES ........................................................................................................................... 363

6.1.
General references within the deliverable ................................................................................... 363
6.2.
Specific references for the chapter 2A ........................................................................................ 363

SafeLand - FP7
Page 13 of 401
6.3.
6.4.
6.5.
6.6.
6.7.
7.
Rev. No: 2
Date: 2012-02-15
Specific references for the chapter 2B......................................................................................... 374

Specific references for the chapter 2C......................................................................................... 377
Specific references for the chapter 2D ........................................................................................ 383
Specific references for the chapter 2E ......................................................................................... 392
Specific references for the chapter 2F ......................................................................................... 392
GLOSSARY ................................................................................................................................ 397

SafeLand - FP7
Page 14 of 401
1. Introduction
Rev. No: 2
Date: 2012-02-15
INTRODUCTION

SafeLand - FP7
Page 15 of 401
1. Introduction
Rev. No: 2
Date: 2012-02-15
1. INTRODUCTION
European countries are exposed to numerous geohazards, such as landslides, debris flows and
rockfalls, which endangered inhabitants and infrastructures. One of the main aims of the
European project SafeLand in the 7th Framework Programme is to develop innovative
mapping and monitoring methods in order to improve the methods for regional landslide risk
assessment and design of early warning systems. More specifically, Area 4 of SafeLand
addresses the technical and practical issues related to monitoring and early warning for
landslides. During the last decade, different monitoring and remote sensing techniques have
undergone rapid development. In order to summarize these scientific and technical advances,
the University of Lausanne, in close collaboration with 18 European institutions, has
produced the project Deliverable 4.1: Review of Techniques for Landslide Detection, Fast
Characterization, Rapid Mapping and Long-Term Monitoring.
Initially, only remote sensing methods were going to be taken into account; but afterward, the
co-authors decided to also include ground-based techniques, which were not described in the
other deliverables of SafeLand project. The core of this deliverable consists of two main
chapters (2 and 3), which aim to develop the basic technical knowledge for (a) landslide
detection (new landslides recognition from space or airborne imagery), (b) fast
characterization (retrieving information on failure mechanism and volume involved), (c) rapid
mapping (fast semi-automatic image processing for changes detection and/or target detection;
hotspot mapping), and (d) long-term monitoring (processing data for retrieving deformation
patterns and time-series).
Figure 1 : structure of the major chapter of the deliverable exploring the state of the art and the theory of remote
sensing and ground based techniques applied to landslides detection, fast characterization, rapid mapping and
long-term monitoring.
SafeLand - FP7
Page 17 of 401
Deliverable 4.1
Review of Techniques for Landslide Detection, Fast Characterization, Rapid Mapping and Long-Term Monitoring.
Chapter 2 summarizes the different techniques and methods from a theoretical point of view.
These include: (1) optical images from ground based and space borne sensors, (2) airborne
and terrestrial laser scanning, (3) ground-based and space borne radar interferometry, (4)
ground based, airborne and offshore geophysical investigations, (5) geotechnical groundbased monitoring systems, and (6) global navigation satellite systems. Each technique
described in the deliverable has benefited of the expertise of specialized research groups. The
structure of this chapter is illustrated in Figure 1.
Chapter 3 shows the main applications of these techniques to landslides, through the synthesis
of seventeen case studies. To this end, each partner provided different examples which
summarize the state-of-the-art of a given technique for different hazards in several situations.
As an example, Figure 2 shows a combination of Ground-Based Radar Interferometry and
Terrestrial Laser Scanning for huge rockslide characterization and monitoring.
Figure 2 : example of a combination of remote sensing method applied to huge rockslide characterization and
monitoring, intergrating 2011 GB-InSAR data of the entire instable area (up: green: stable > red: 8 mm
displacements in two weeks) with a comparison of 2006 and 2011 TLS point clouds (down) of the frontal part of
the rockslide (Aerial photo Swisstopo; GBInSAR and TLS data IGAR-UNIL).

SafeLand - FP7
Page 18 of 401
1. Introduction
Rev. No: 2
Date: 2012-02-15
The state of the art overview demonstrates the fast evolution of techniques having a strong
impact on landslide mapping and monitoring. For example, thanks to high resolution satellite
imagery or aerial laser scanning, the way we are currently characterizing and monitoring
landslides has drastically changed. In addition, A-DInSAR gives now reliable information to
monitor slope movements at a regional scale.
In addition to being a state-of-the art overview, this deliverable provides helpful and extensive
support for non-specialists and students interested in the application of new monitoring
techniques to different types of mass movement. It also aims to be a common reference for
the different deliverables of SafeLand Area 4, specifically for the Deliverable 4.3, Creation
and updating of landslide inventory maps, landslide deformation maps and hazards maps as
input for quantitative risk assessment using remote sensing technologies, the Deliverable 4.4,
Guidelines for the selection of appropriate remote sensing technologies for monitoring
different types of landslides, and the Deliverable 4.8, Guidelines for monitoring and earlywarning systems in Europe Design and required technology.

SafeLand - FP7
Page 19 of 401
2. Data and Techniques
Rev. No: 2
Date: 2012-02-15
DATA AND TECHNIQUES

SafeLand - FP7
Page 21 of 401
Rev. No: 2
Date: 2012-02-15
2. DATA AND TECHNIQUES
PART A: Passive Optical Sensors
p. 25
PART B: Active Optical Sensors
p. 65
PART C: Active Microwave Sensors
p. 103
PART D: Geophysical Investigations
p. 165
PART E: Geotechnical Ground-Based Monitoring Systems
p. 221
PART F: Other Techniques
p. 239
Table 1 (below) : Summary of the main characteristics of the techniques discussed in this deliverable. To go
further, please read the appropriate section.

SafeLand - FP7
Page 23 of 401
Deliverable 4.1

SafeLand - FP7
Page 24 of 401

Part A Passive Optical Sensors
Rev. No: 2
Date: 2012-02-15
PART A
PASSIVE OPTICAL SENSORS
SECTION 1: General Pre-Processing Step
p. 26
SECTION 2: Ground-Based Imaging
p. 35
SECTION 3: Aerial Imaging
p. 39
SECTION 4: Satellites Imaging
p. 44
Sections 1, 2, 3 and 4
- Authors: A. Stumpf1,2, N. Kerle1, J.-P. Malet2
- Reviewers: J. Hervs3, M. Van d. Eeckhaut3
1
. School for Disaster Geo-information Management, International Institute of Geo-Information Sciences and Earth Observation
(ITC), United Nation University, Enschede, The Netherlands. stumpf24883@itc.nl
2
. Centre National de la Recherche Scientifique (CNRS), Institut de Physique du Globe de Strasbourg (IPGS), France.
3
. Joint Research Center, Ispra, Italia.
Stumpf A., Kerle N., Malet J.-P., 2010. Passive Optical Sensors. In the Deliverable 4.1 of the European project SAFELAND: Review of
Techniques for Landslide Detection, Fast Characterization, Rapid Mapping and Long-Term Monitoring. Edited in 2010 by Michoud
C., Abelln A., Derron M.-H. and Jaboyedoff M. Available at http://www.safeland-fp7.eu

SafeLand - FP7
Page 25 of 401
Deliverable 4.1
1. GENERAL PRE-PROCESSING STEPS

1.1. DIGITIZING
Analogue recording systems are getting gradually replaced by digital systems, whereas large
archives of analogue photographs, especially aerial photographs, represent a valuable source
of historical information. Furthermore, the geometric resolution of analogue photography is
still slightly higher than geometric resolution opto-electronic devices. Thus, analogue
photography is expected to still play an important role for the most accurate photogrammetric
tasks in the future [Kraus, 2007]. Scanning is the first necessary step to integrate them into
modern Geographic Information Systems (GIS) and digital image analysis. Some low cost,
flatbed scanners offer a satisfactory spatial and radiometric resolution for digitizing large
scale aerial photographs, but introduce geometric distortions into the scanned image. Highend scanners, designed for digital photogrammetry, (Zeiss/Intergraph PS, Leica/Helava DSW,
ISMs, Vexcel, and Wehrli) are characterized by a high geometric accuracy (2-4 m), a high
geometric resolution (4-12.5 m) and a high radiometric accuracy (1-2 grey values). In
general newer more mature instruments (DSW 500, SCAI, RM-2, and UltraScan 5000) lead
to better results, while quality is also highly dependent on proper calibration, maintenance,
environmental conditions and high performance software tools. As eventual errors introduced
during the scanning process will substantially influence the results of consequent image
analysis the user should be aware of those aspects [Baltsavias, 1999].
1.2. GEOMETRIC CORRECTION
An appropriate geometric correction of images involves the modeling of systematic (scan
skew, mirror-scan velocity variance, panoramic distortion, platform velocity, earth rotation,
perspective) and non-systematic (altitude, variance in altitude and angles) errors to derive a
relationship between the image and a global coordinate system [Jensen, 2005].
Due to the introduction of off-nadir views with very high resolutions an accurate modeling of
geometric distortions gains even more importance today. If this step is not performed properly
their digital interpretation or fusion with other datasets will fail or lead to bad or misleading
results. Systematic distortions, especially those related to the instrumentation are generally
well known by the operators of satellite and airborne photogrammetry and corrected at ground
receiving stations or by image vendors [Toutin, 2004].
In ground-based and low-altitude airborne photography non-metric off-the shelf digital
cameras are recently more frequently in use. The use of such systems for photogrammetric
applications renewed the interest in proper calibration methods of the camera what usually is
done on 2D/3D test fields or during the image acquisition itself, latter often referred as on-thejob or self-calibration [Luhmann et al., 2006]. Analytical self-calibration of analogue and
digital cameras already reached maturity. Very commonly bundle-block adjustment is used to
determine the exterior orientation of multiple camera views. Thereby, interior and exterior
orientation of the acquired images can be obtained at the same time. However, where the
network geometry in the image is weak, previous calibration on a test field might be favorable
to first determine interior orientation parameters and account for lens distortion. Nowadays,
this can be a fully automatic procedure to employ consumer grade cameras on low and
medium accuracy levels [Ergun et al., 2010].

SafeLand - FP7
Page 26 of 401

Rev. No: 2
Date: 2012-02-15
Distortions in images taken by metric cameras can originate from variations in the platform
altitude, platform velocity, and platform orientation, uncertainties in calibration parameters,
panoramic views, earth rotation, curvature and relief. Finally deformation will also occur
when the large corrected images are projected on the tangent plane of the reference ellipsoid
that is approximation of the geoid.
All those possible distortions require geometric corrections using models and mathematical
functions such as, either 2D/3D polynomial models, 3D rational functions, rigorous 2D/3D
physical models and deterministic models. A simple affine polynomial model requires at least
3 well-distributed ground control points (GCPs). More complex models will need at least 5
GCPs. Similarly, in close range photogrammetric applications algorithms that perform a
relative orientation of multiple images will need five [Stewnius et al., 2006] or at least three
[Kalantari et al., 2009] matching points. Regardless of the approach, more than the minimum
number of GCPs should be employed in order to obtain an optimum solution. The stepwise
removal of distortions with 2D/3D physical models is generally applied at ground receiving
station and resulting products are sold at different processing steps (e.g. raw, georeferenced,
geocoded). Whatever mathematical models are used, four basic processing steps can be
distinguished (Figure 3):
acquisition of image(s) and pre-processing of metadata
acquisition of the ground points (control/check/pass)
computation of correction model parameters
image(s) rectification
Slight differences occur using empirical models where metadata is useless and 2D empirical
models where the Z-elevation coordinates for GCPs and Digital Elevation Models (DEMs)
are needless [Toutin, 2004]. Further issues concerning orthorectification such as adequate
resampling methods and topographic information are treated below.

SafeLand - FP7
Page 27 of 401
Deliverable 4.1
Figure 3 : Typical workflow for geometric correction [Toutin, 2004]
A good overview of geometric correction models mainly for satellite and airborne imagery is
given by Toutin [2004]. An evaluation of typically achieved accuracies was more recently
published by Sertel et al. [2007]. Due to ongoing frequent launches of high-resolution
satellites Rational Function Models (RFM) currently receive greater attention. RFMs are
usually derived from satellite onboard instruments including GPS and improvised star sensors
[Sadasiva Rao et al., 2006]. They are used to determine Rational Polynomial Coefficients
(RPCs) for the generation of DEMs. Their accuracy can be increased considering additional
GCPs [Hu and Tao, 2002], whereas there is no complete agreement if in some cases less than
six GCPs are enough [Martha et al., 2010a]. RFMs proved to lead to highly accurate results
independently from the sensor and are an important alternative where satellite agencies may
not like to release the complex camera model and metadata [Nagasubramanian et al., 2008].
1.3. RADIOMETRIC CORRECTION
Digital sensors record the intensity of electromagnetic radiation from Earths Surface as a
digital number (DN). Its value range is dependent on the bit-depth of the acquired product
(typically 8 or 16 bit) and does not necessarily represent the bit-depth that was originally
recorded by the sensor (typically from 8 to 11 bit). The intensity received at the sensor is not
only dependent on the spectral characteristics of the earth surface but on a number of further
parameters such as atmospheric conditions, viewing geometry, sun angle, and so on. Where
the actual reflectance of a ground target is of interest and/or a comparison of spectral
characteristics between different images is intended previous radiometric correction is
essential [cf. Song et al., 2001]. Absolute and relative radiometric corrections are two
basically different approaches to perform those tasks, respectively.
A product level including corrections of sensor inherent radiometric errors (e.g. destriping of
Landsat TM images) with sensor specific parameters may be ordered readily preprocessed
from corresponding data providers or space agencies. However, those products in general
represent the incoming radiance at the sensor and usually do not include a radiometric
SafeLand - FP7
Page 28 of 401

Rev. No: 2
Date: 2012-02-15
correction of scene specific atmospheric effects. Numerous approaches have been developed
for the correction of atmospheric effects to gain surface reflectance and might be grouped into
the following main categories [Clark et al., 2002; Gao et al., 2009]:
Ground calibration methods [e.g. Ferrier, 1995; Xu and Huang, 2008]
Empirical atmospheric correction with scene derived parameters

-
Dark Object Subtraction (DOS) [Chavez, 1988; 1996; Song et al., 2001]
Dense Dark Vegetation (DDV) [Kaufman et al., 1997]
Modified Dense Dark Vegetation (MDDV) [Song et al., 2001]
Path Radiance [Wen et al., 1999]
Radiative transfer models (A good overview was given recently by Gao et al. [2009]
and some further examples are listed here)
-
EXACT [Popp, 1995]
SMAC [Rahman and Dedieu, 1994]
6S [Vermote et al., 1997]
Durchblick [Holzer-Popp et al., 2001]
Hybrid approaches [e.g. Clark et al., 2002]
Based on radiative transfer models absolute radiometric correction methods convert the DNs
within a satellite or airborne image into the corresponding reflectance of the surface. The
inputs for the models are typically atmospheric and sensor parameters for the acquisition date
[Richter, 1990] topographic information [Richter, 1998] or DNs of dark image objects in one
or multiple sensor channels [e.g. Chavez, 1988; Kaufman, 1989; Kaufman et al., 1997]. While
the necessary sensor parameters are mostly included in the header files of satellite images and
global DEMs are now commonly available, the provision of accurate atmospheric parameters
for the acquisition data is often difficult. Profound assessment of absolute radiometric
correction methods with respect to the quality of classification and change detection
[Schroeder et al., 2006; Song et al., 2001] reveals that the more complex methods do not
necessarily lead to more accurate results.
As another category we might consider Topographic Normalization methods. Especially in
mountainous areas there is a strong influence of the topography on the signal recorded by
spaceborne optical sensors. The signal from equal surfaces may vary considerably according
to their position on slopes oriented away or towards the sun. The effects influence increases
towards low sun angles and causes problems for many subsequent steps of image analysis. A
comprehensive evaluation of different techniques was recently provide by Richter et al.
[2009]. They concluded that visually best results are achieved with the modified Minnaert
(MM) method, although sometimes the C approach yields better results for the visible bands

SafeLand - FP7
Page 29 of 401
Deliverable 4.1
in the blue to red spectral region. However, further research should evaluate on a global basis
which method performs best in which situation.
During relative radiometric correction (Normalization) the radiometric properties of a subject
image are adjusted to match the radiometric properties of a reference image. Therefore a
mathematical relationship based on image statistics (i.e. standard deviation, minimummaximum) the histograms or most commonly on a linear regression function, is established
between the two images.
The normalized image then should appear to have been acquired under the same solar and
atmospheric conditions as the reference image. Typically the image with the most favorable
radiometric properties (e.g. low atmospheric water content) is chosen as a reference. Resulting
values of the subject image are unitless and only comparable relatively within a set of similar
corrected images if the reference did not undergo absolute radiometric correction before. A
great variety of relative correction techniques has already been developed and tested for the
first generation of medium to high resolution satellites, especially from the Landsat program
[c.f. Janzen et al., 2006; Over et al., 2003; Yang and Lo, 2000]:
Simple Regression (SR), [Jensen, 1983]
Histogram Matching (HM)
Pseudoinvariant Feature (PIF), [Schott et al., 1988]
Pseudoinvariant Feature Regression (PIFR), [Du et al., 2002; Du et al., 2001; Paolini
et al., 2006]
Dark-Bright Set (DB), [Hall et al., 1991]
No Change Set (NC), [Elvidge et al., 1995]
Other Statistical methods [Yuan and Elvidge, 1996]
Manual No Change Regression (MNCR), [Over et al., 2003]
Unchanged Pixel Normalization (UPN), [Yaallah and Saradjian, 2005]
Ridge methods [Chen et al., 2005; Song et al., 2001]
Artificial Neural Network [Velloso et al., 2002]
Iteratively weighted least squares regression (IWLSR), [Zhang et al., 2008b]
Ordinal conversion [Nelson et al., 2005]
Multivariate alteration detection (MAD), [Canty et al., 2004]

SafeLand - FP7
Page 30 of 401

Rev. No: 2
Date: 2012-02-15
Most commonly used linear regression methods in general require the manual or automated
selection of unchanged targets in the images to be normalized. Eckhardt [1990] gave a list of
criteria such as similar altitude or minimal amount of vegetation for choosing the targets.
Hong [2007] presented a comprehensive review of common radiometric normalization
methods for Landsat and stated a number of particular advantages and drawbacks.
Simple Regression (SR) works well if changes between the scenes are not too big and
geometric coregistration is sufficiently accurate. Histogram matching (HM) is useful for
scenes with slightly different sun angles or atmospheric effects. It is not dependent on an
accurate spatial coregistration of the images. When the matching is performed band to band
the internal coherence of the image bands is distorted. HM shows weaknesses if surface
changes and/or cloud cover are too strong. It produces favorable results for an optical
comparison but due to its non-linear transformation it affects the relative distribution of
spectral change. Yang and Lo [2000] noted that in general visually and statistically robust
methods (such as the HM, and NC methods) tend to reduce the magnitude of spectral change.
On the other hand, the DB and PIFR methods cut down only a moderate degree of spectral
change and favor better results for change detection. Both methods are not dependent on
geometric coregistration of the scenes but require intervention for the selection of unchanged
targets (PIFR) or appropriate threshold values (DB) to define dark and bright pixel sets.
Finding a suitable number of unchanged objects may represent problem if the time step
between the two scenes is too large or the sensor parameters strongly differ. The number of
common unchanged targets further decreases when more than two images should be
considered [Paolini et al., 2006].
If land and water is present in the images and changes in solar illumination geometries,
phenological conditions and landcover are not too large No Change Set (NC) is a valuable
method. The method is computationally efficient and does not comprise the need of
identifying dark and bright pixels. It reduces cloud, shadow, and snow effects, uses a large
percentage of the total number of image pixels and distributes normalization error among
different land-cover types [Yuan and Elvidge, 1996].
A number of problems occur when applying most of those image processing methods on high
to very high resolution images. The higher resolution complicates the selection of unchanged
features, only one band is recorded in the near-infrared and thermal spectra are typically not
recorded with VHR sensors [Hong, 2007; Hong and Zhang, 2008].
In general there is no best universal solution for radiometric correction. Especially for relative
normalization the most suitable method is rather dependent on the satellite data, the specific
scene and the purpose of its application. However, a few general statements might be
possible. For a simple visual comparison, histogram matching (HM) which is implemented in
several image-processing software tools might be an efficient choice. For change detection
Multivariate Alteration Detection (MAD), and Pseudoinvariant Feature Regression (PIFR)
offer efficient automated approaches without any need for additional atmospheric data or
manual selection of image samples.
Newer approaches such as Iteratively weighted least squares regression (IWLSR), [Zhang et
al., 2008b] and Ordinal Conversion [Nelson et al., 2005] demonstrate promising results and

SafeLand - FP7
Page 31 of 401
Deliverable 4.1
wait for evaluation on data from different sensors and with respect to different applications.
Considering the wide variety of approaches analysts must be aware of the particular strength
and weaknesses of existing procedures to choose the most efficient approach.
While for classification and post-classification change detection radiometric correction might
be even dispensable [Song et al., 2001], there is a general agreement that any correction
method enhances the results of direct change detection and other applications.
Comparison of the spectral signatures of forest patterns on Landsat TM images after absolute
and relative radiometric correction techniques [Janzen et al., 2006] demonstrated a better
performance of relative correction methods. Thus, especially when relative differences in the
spectral properties are rather important than absolute values, relative radiometric
normalization is an attractive alternative. It avoids the difficulties in obtaining accurate
atmospheric and sensor parameters, is usually less computationally intensive and easier to
apply than absolute radiometric correction.
1.4. FILTERING AND TEXTURE
Image enhancement techniques most commonly rely on operations within a template of a
certain extent, moving over the image and redefining pixel values based on neighborhood
relations. This can be done directly on the image domain or on a Fourier transformed
representation of the image, whereas the latter approach is especially favorable if a large
template should be used and/or the noise in the image has a periodical pattern.
Low-pass filters such as median, mode or mean filters are commonly used to remove artifacts
from an image but lead to a loss of local detail. Depending on the size of the smoothening
window the resulting image has a typically blurred appearance but high-frequency noise has
been removed.
High-pass filters are useful to emphasize local variations of texture such as abrupt transitions
in brightness. The result is an image representation with enhanced edges and suppressed low
frequency detail within homogenous areas. High-pass filtering can be achieved by subtracting
a low pass filtered image from the original. Other methods such as Sobel, Prewitt, Roberts or
Laplacian filters involve the use of a kernel. The distribution of grey values within the spatial
domain of the kernel is used to quantify the local contrast of each pixel relative to its
neighbors. Pixels with a high contrast in the local neighborhood obtain higher values and
thereby edges in the image are enhanced. Such operators are also known edge detection filters
and have been demonstrated as useful to detect to highlight hummocky main bodies,
accumulation zones, crowns and back scarps of landslides [Eyers et al., 1998; Mason et al.,
1998]. The noise level in an image is typically higher in the high spatial frequencies, what
will be emphasized and should be considered for the application of a high-pass filter.
Low- and high-pass filters are also commonly involved in the pre-processing of image
matching [e.g. Honda and Nagai, 2002] and image correlation procedures for displacement
measurement [e.g. Leprince et al., 2007b]. For a comprehensive overview of filtering
techniques for image registration or image enhancement the interested reader is referred to
Zitov and Flusser [2003] or Schowengerdt [2007], respectively.

SafeLand - FP7
Page 32 of 401

Rev. No: 2
Date: 2012-02-15
Hervs and Rosin [1996] were among the first who combined a number of more complex
Grey-Level-Co-Occurrence-Matrices (GLCM after Haralick et al., 1973) to quantify
differences between landslides and spectrally and textural similar features. Hervs et al.
[1996] noted that the down slope movements also affect the directional components of
texture. Whitworth et al. [2005] demonstrated the utility of GLCM texture measures to
quantify the hummocky surfaces roughness caused by landslides in Jurassic clays.
Geometrically selective filters like a rectangularity filter [Hervs et al., 2003] have been
reported as useful to remove undesired remnant scene noise like buildings and agricultural
crop patterns after performing change detection and thresholding on VHR imagery for
landslide mapping purposes.
1.5. ORTHORECTIFICATION
Especially VHR imagery from off-nadir viewing sensors needs to be orthorectified to account
for terrain-related distortions. Ground control points, an appropriate sensor model and a DEM
with sufficiently high resolution and accuracy are incorporated in this process. Stereo-pairs
and triplet satellite images are now more commonly available at better prices and can be used
for the derivation of DEMs. Alternatives are provided through costless and globally available
datasets from the SRTM [Jarvis et al., 2008] or ASTER [ASTER-GDEM-VALIDATIONTEAM, 2009] however, with a rather low resolution and accuracy. DEMs with very high
resolution and high accuracy can also be obtained with SAR or LIDAR but are still rather cost
intensive and may represent an alternative in special cases.
Orthorectification is implemented in a user friendly way in most commercial and many Open
Source software applications and can be considered as mature techniques. However, image
coregistration and orthorectification methods still undergo further development for automated
registration with increased accuracy. Recently proposed methods include the more accurate
automatic extraction of GCPs from reference images [Gianinetto and Scaioni, 2008] or their
stepwise refinement through image correlation and statistical correlation with DEMs
[Leprince et al., 2007a]. Once the image is registered with respect to a DEM, Nearest
Neighbor, Bilinear Interpolation and Cubic Convolution are the most commonly used
resampling methods to point image pixels from 2D coordinates to 3D coordinates on the
surface. The Nearest Neighbor algorithm is the fastest choice and simply assigns the value of
the nearest pixel in the image to the new coordinate system. It has the advantage that the
original pixel value will remain for the orthorectified image. Bilinear Interpolation and Cubic
Convolution techniques combine a greater number of nearby cells to compute the value of the
transformed cell. The Cubic Convolution method avoids the sometimes jagged appearance
that arises from the use of the Nearest Neighbor method and gives a somewhat sharper image
than the Bilinear Interpolation method. In exchange the original pixel values are changed and
the method is rather computational intensive. Especially for further use of correlation
techniques aliasing that is often introduced by the mentioned kernel methods should be
avoided. Sinc functions and Modulation Transfer Function (MTF) resampling are further
more sophisticated resampling methods. Both also make use of a kernel window, whereas,
MTF includes the empirical modeling of the optical and electronic properties of the specific
sensor. Leprince et al. [2007a] further introduced resampling with Inverse Transformation
Matrices to minimize aliasing effects in consequent image correlation.

SafeLand - FP7
Page 33 of 401
Deliverable 4.1
1.6. DATA FUSION

Hundreds of different data fusion methods are mentioned in scientific literature and/or are
available in image processing software packages. According to their position in image
processing workflows they can be categorized as signal level, pixel level, feature level, and
decision level fusion methods [Dong et al., 2009] (Figure 4).
Figure 4 : Conceptual overview of data fusion at different processing levels [Dong et al., 2009]
Standard fusion algorithms such as Principal Component Analysis (PCA), Intensity-HueSaturation (IHS) or the Brovey-Transformation are very common methods for fusion of
multispectral channels with a panchromatic image of higher resolution (pansharpening). They
are implemented in image processing software to obtain more details for visual interpretation
or to reduce the number and degree of correlation in hyperspectral or multitemporal datasets.
IHS can be used for pansharpening of multispectral imagery, whereas the resulting Intensitychannel is simply replaced by the corresponding panchromatic image. The main limitation of
this method is that the number of input bands is restricted to three. More bands can be
transformed with PCA to separate color from intensity information in the multispectral image.
The success of these pixel-based approaches largely depends on a very exact image
coregistration and they tend to distort the original spectral signatures. Newer approaches
include Wavelet Fusion and fusion methods that weigh the influence of the panchromatic

SafeLand - FP7
Page 34 of 401

Rev. No: 2
Date: 2012-02-15
image based on dependency of local texture. It has been demonstrated that the latter method
can produce results with smaller deviation in the grey values [Hirschmugl et al., 2005].
In recent years especially the use of wavelets - now already implemented in standard image
processing software - and artificial neural networks gained greater popularity. The latter is
part of a large number of fusion methods functioning on feature or decision level
incorporating multiple datasets in a probabilistic [Bendjebbour et al., 2001] or rather
deterministic [Martha et al., 2009a] framework.
The final aim of data fusion is to integrate complementary and redundant information to
increase the information content in a particular scene and thereby increase the success of
identification and classification of image objects such as landslides [Chang et al., 2007], the
accuracy of landcover classifications [e.g. Sarkar et al., 2005], the reliability of change
detection analysis or the success of military operations [Dong et al., 2009]. A variety of the
mentioned techniques useful for pansharpening has been evaluated with respect to their effect
on change detection with VHR satellite imagery by Bovolo et al. [2010]. They concluded that
the Minimum Mean Square Error [Garzelli et al., 2008] is the most reliable method for this
purpose.
2. GROUND-BASED IMAGING
2.1. SENSOR AND PLATFORM DEVELOPMENT
For most practical applications of ground-based imaging digital cameras replaced analogue
systems and cost-efficient, non-metric, single lens reflex (SLR) cameras are nowadays a
commonly used, cost-efficient alternative to rather expensive metric systems. Due to the
exploitation of mass consumer market the sensor resolution is increasing rapidly, whereas of 6
to 12 Mega-pixel cameras (e.g. Figure 5, most left) being readily affordable (100-300 EUR)
for both the layperson and the scientist. Costs for commercial SLR cameras commonly used
for photogrammetric measurements are in the range of 1,000-2,000 EUR, whereas the top-end
high resolution cameras with up to 160 Mega-pixel resolutions (> 20,000 EUR) still wait for
scientific applications. Thereby, it is worth noting that especially scientific measurements can
benefit from higher resolutions but that radiometric and geometric quality of the acquired
imagery are at least of equal importance. Although, in some cases vendor-provided camera
parameter might be sufficient for the photogrammetric interpretation of single views, the
correction of geometric distortions is crucial for tasks such as the generation of highly
accurate Digital Surface Models (DSM) [Chandler et al., 2005].
Figure 5 : Typical sensors for terrestrial photogrammetric surveys in historical order (left to right):
Phototheodolite Zeiss Photheo, Linhof Metrika, Rolleiflex 6006, and Nikon D300

SafeLand - FP7
Page 35 of 401
Deliverable 4.1
As many geomorphological processes (e.g. debris flows) exceed the maximum frame rate of
solid state digital cameras video cameras (20-30 Hz) and high-speed cameras (> 100 Hz)
became valuable components for the in-situ and laboratory observations [e.g. Arattano and
Marchi, 2000; Imaizumi et al., 2005; Zakeri et al., 2008]. Recent advances in direct
georeferencing, imaging sensor technology and easy access to inexpensive telecommunication
enhanced the implementation of mobile mapping systems. Although most systems do not
target top-level accuracy they offer quick and flexible disaster response for natural hazards
and systems for real-time monitoring of debris flows are under development [Yin et al.,
2007].
The mentioned sensors systems operate in general in the visible spectrum of light. Most CCD
based cameras are partly sensible in the NIR and special systems for night vision (especially
for military operations) and laboratory measurements are available. While NIR measurements
are useful for studies of rock fractures in the laboratory [Brady and Rowell, 1986] no example
for ground-based NIR imaging of landslides was found in the literature.
2.2. VISUAL INTERPRETATION
Analogue photographs are valuable testimonies for historical evolution of landslides and other
landscape features. Digital photography is often the first source of information during and
short after a given event and helpful to reconstruct the failure process. Photos remain as a
valuable source of information especially where traces of landslide processes are removed
quickly due to natural processes or human efforts. Repeated terrestrial photography can be
used to asses displacement directly through observations of surfaces [Schmidt and Nsser,
2009] or indirectly through associated changes of landscape elements [e.g. Coe et al., 2009].
Multiple images can be used to assess displacements by identifying the same objects in
multiple images [Arattano and Marchi, 2000]. Information about the granulometric
composition of deposits can be derived from manually assessment or with computer-aided
visual interpretation [Genevois et al., 2001]. Similarly, structural characteristics of rock
masses such as volumetric fracture intensity can be derived from terrestrial photos and reveal
information on rock slope stability [Crosta, 1997].
2.3. GROUND-BASED DIGITAL PHOTOGRAMMETRY
Photogrammetric methods have been extended with great success to the generation of high
resolution DSMs. Created at one single time step, such models reveal detailed information on
the structure of a particular slope. DSMs generated form terrestrial photographs of rock cuts
have been analyzed in detail to determine rock mass discontinuities [Roncella et al., 2005;
Sturzenegger and Stead, 2009] and DSMs of river channels have been created noting their
usefulness to asses changes of the river bed and model flows in the channel [Chandler et al.,
2002].
In principle any ground deformation process of sufficient magnitude can be revealed
analyzing multiple photogrammetric DEMs of the same location. This approach has been
followed for studies of rock-glaciers [Ladstdter and Kaufmann, 2005], coastal cliff erosion
[Lim et al., 2010], lava flows [Robson and James, 2007] and the monitoring of landslides
[Cardenal et al., 2008]. DSMs from present consumer-grade SLR cameras can achieve
sufficient accuracy when proper methods are used (convergent networks with self-

SafeLand - FP7
Page 36 of 401

Rev. No: 2
Date: 2012-02-15
calibration), even in cases of pre-calibrated cameras or when there is a lack of externally

surveyed control points. Therefore they are an interesting alternative to more accurate but also
more expensive LIDAR scanning campaigns.
2.4. DIGITAL IMAGE ANALYSIS OF GROUND-BASED PHOTOGRAPHS
Concepts for a real-time warning system incorporating terrestrial photography have been
proposed to monitor volcanic activity [Honda and Nagai, 2002] and landslides
[Kandawasvika, 2009]. Their potential use for early warning systems remains restricted to
processes where daytime observations offer sufficient information, while at the same time
valuable information for process understanding can be gained from continuous observations.
Recently a system for the automated retrieval of displacement fields via digital image
correlation of daily to hourly ground-based image acquisitions has been proposed and results
have been demonstrated to be in good agreement with point-wise GPS measurements
[Travelletti et al., 2009].
The applied technique can be considered as a special application of correlation techniques
also applied on video footages (3.2.5) and satellite imagery (3.4.7), whereas the main
differences are frame frequencies (lower than video but typically higher than airborne
imagery), scale (higher resolution and lower spatial coverage than airborne and satellite
imagery) and a oblique camera view (compared to airborne and satellite imagery).
2.5. ANALYSIS OF VIDEOS
As direct observations of fast moving landslides are rather rare digital and analogue video
recordings are very useful documents to distribute observations within the scientific
community but also to raise public awareness. Early collections of video material [e.g. Costa
and Williams, 1984] led to a better understanding of the dynamics of debris flows and settled
the base for many studies of their dynamic behavior. The widespread of small digital cameras
with favorable prices increased the amount of amateur footages that are often available via the
Internet. For most people without the specific scientific background those records may
constitute the only opportunity to actually see a landslide and gain understanding of related
hazards and risks.
Interpretation of videos became a common tool for the qualitative monitoring of landslide
behavior in experimental designs [e.g. Enet and Grilli, 2007; Ilstad et al., 2004; Ochiai et al.,
2004]. Field studies often integrate [e.g. Jibson, 2006] or even fully rely [e.g. Fujisawa and
Ohara, 2008] on video captures to assess failure mechanisms, down slope processes and
estimate velocities of real natural events. Video cameras may also be integrated into
monitoring systems to observe critical slope units [e.g. Mantovani et al., 2000] or channels
[e.g. Marchi et al., 2002]. However, because of restricted reliability (night, fog during rainfall
events, etc.) video captures should only be an additional element within early warning
systems. On the other hand they still reveal a lot of information for comparison with other
monitoring instruments and validation data for numerical models.
Considerable research has been carried out towards an automated evaluation of surface
velocities of debris flows [e.g. Arattano and Marchi, 2000; Genevois et al., 2001] from video
footages.

SafeLand - FP7
Page 37 of 401
Deliverable 4.1
The determination of surface velocities is general referred as image velocimetry. Five

different approaches, applicable on textured surfaces, can be distinguished [Itakura et al.,
2005].
Spatial filtering is based on a grating lines filter placed between a moving object and a photo
detector. The periodical variation of the light intensity passing through the filter is directly
related to the velocity of the object perpendicular to the filter grate. A setup of two photo
detectors with filter lines perpendicular to one another can be used to obtain x and y velocity
components of the surface on a moving object [Aizu and Asakura, 1987].
The Spatio-Temporal Derivative Space Method [Ando, 1986] is based on luminance vectors
of a pixel and its neighborhood with time. Assuming that luminance of pixel within a defined
neighborhood region would not change without any movement, x and y components of the
movement are calculated in order to reduce difference in vector space. A crucial step of this
method is the definition of the size of the neighborhood area. If the movement of the surface
is nearly uniform the errors but also the resolution of measurements decrease with a larger
chosen neighborhood areas. However, many surfaces (e.g. debris flows) do not demonstrate a
uniform movement and an optimal size (e.g. order of the moving particles) exists. An iterative
increasing measurement area can be used to stepwise reduce an intrinsic error term and define
the optimal scale of the observation [Inaba et al., 2000].
The encoding of digital videos attempts to exploit redundancies between consecutive or
temporally close video images. Most commonly this is done for compression purposes.
Instead of saving both consecutive images in full size only parts of the images where changes
occur are stored in the second image, while the rest of the information is obtained from
unchanged areas of the first image. Motion is estimated by block matching algorithms and
encoded as motion vectors. The latter reproduce the changes and are a crude approximation to
optical flow, although heavily corrupted by noise [Coimbra and Davies, 2005]. Several
methods, e.g. motion segmentation [Avrithis et al., 1999] or the use of vector fields with
associated confidence maps have been developed for the analyses of MPEG motion vectors
and were successfully applied to surveillance camera captures or images of debris flows
[Koyama et al., 2000].
Cross-correlation techniques such as correlation-based optical flow and classical particle
image velocimetry (PIV) [e.g. White et al., 2003] are most frequently applied to measure
velocity fields on multiple images of fluids and surfaces. Thereby the content of search
windows in at least two consecutive images is cross-correlated to determine the average
spatial shift maximizing the cross-correlation function. Commonly a two-dimensional discrete
Fourier transform is used to facilitate the evaluation of the cross-correlation function [Qunot
et al., 1998]. A wide range of different techniques is readily available, while most of them
rely on three different stages of image processing [Barron et al., 1994].
Despite their differences, many of these techniques can perform:
1.
Filtering to extract signal structure of interest and enhance the signal-to-noise ratio
2.
Extraction of local correlation surfaces

SafeLand - FP7
Page 38 of 401

Rev. No: 2
Date: 2012-02-15
3.
Integration of measurements to produce a 2-D flow field, often including smoothening
of derived motion vectors
If the observed surface is a plane parallel to the image plane the derivation of the motion
vectors is a straightforward task [e.g. Take et al., 2004]. However, many natural surfaces
expose irregular surfaces and there is often no unrestricted choice of view angles in a realworld situation. Thus, many practical applications will include a fourth step if metric vectors
in the object domain are desired. If the object-domain resembles a plane oblique to the image
plane it is possible to approximate the position of points in 3D space (and consequently
vectors) by introducing an interface plane between object and image space [Arattano and
Grattoni, 2000].
In general natural surfaces expose more complex geometries than that of a plane and the
derivation of 3D motion vectors will require at least stereoviews of the object. In such cases a
single view images can reveal 2D motion vectors in the object space if additional information
about the surface geometry is available. Related techniques are more commonly in use for the
processing of multitemporal photographs and are treated in sections 3.2.4 and 3.4.7.
Considerable progress has also been made in recent years in the analysis of video surveillance
for security applications. Main tasks are the real-time recognition and tracking of moving
objects but also the recognition of unusual events [e.g. Choudhary et al., 2008], what might be
of particular interest for additional components of early warning systems (to see further,
please read the deliverable 4.7 of the SafeLand project).
3. AERIAL IMAGING
3.1. ANALOGUE VS. DIGITAL CAMERA SYSTEMS
Digital systems are cost saving (no film, no photo lab and better automation possibility), the
product derivation is time saving (no film development, no scanning and possible automation
of the digital workflow), and resulting images have a higher quality in terms of radiometric
dynamics, signal to noise ratio and reproducible color and in-flight image control [Reulke,
2003]. Analogue film cameras have a standard square format of 23cm, whereas the market for
digital cameras is subdivided in large, medium and small format cameras. Among those
digital systems the development of large-format area arrays cameras stagnates and is replaced
by systems that integrate multiple medium and small format cameras (with individual lenses
and CCD arrays). Many already combine more than 86 megapixels of array area and can be
considered as the most advanced digital systems for the aerial image acquisition over large
areas. The Intergraph Digital Mapping Camera DMC, for example, achieves a Ground
Sampling Distance (GSD) of 20 cm if a flight height of 2000 m is assumed. The German
Society of Photogrammetry, Remote Sensing and Geoinformation is currently running a large
initiative for the evaluation of new systems [Cramer et al., 2009]. While the geometric
resolution of digital cameras still lags slightly behind the resolution of similar analogue
systems the location accuracy is quite similar and they achieve a higher radiometric image
quality. For the generation of DSMs this leads to superior results from the digital systems
[Cramer, 2009]. In many cases analogue cameras purchased before the raise of CCD/CMOS
will continue to be used operationally, so that at present and in the near future we can still
find products from both instrument types on the market. At the same time the development of

SafeLand - FP7
Page 39 of 401
Deliverable 4.1
analogue films is still advancing and leads to grainless images at ten microns, while the
sensitivity increases by a factor of five, thereby reducing necessary illumination time
significantly [Brandes, 2003]. A recent review [Honkavaara et al., 2009] revealed that highquality digital systems, despite their already common operational use, still need further
enhancement in terms of overlapping spectral channels, incomplete camera parameters or onflight atmospheric correction. Some of these issues have already been recognized or partly
solved by the manufactures.
3.2. PANCHROMATIC VS. NATURAL COLOR AND NEAR INFRARED
The traditionally most used film for aerial photography is panchromatic (sensitive in the range
from 400 nm-780 nm) black and white film. The film is available in a wide range of
resolutions, gradations and sensitivities. Its geometric resolution is usually higher than that of
color films and with an appropriate set of filters it can also be used during hazy or misty
weather. Color and color-infrared are at present more and more commonly used film types in
aerial surveys. Corresponding films comprise three photo sensitive layers to generate a color
composite which is built out of the colors blue, yellow-green and red of the visible spectrum.
In case of color infrared usually the blue layer is substituted by a layer sensible in the near
infrared spectrum. Black and white infrared film is alternatively used for vegetation surveys
where an increased contrast of different vegetation types or water bodies is required. A further
advantage of infrared aerial photography is that in this part of the spectrum atmospheric
interferences are lower.
Instead of representing light by the intensity of a chemical reaction on the film material digital
systems convert light into an electric signal. Two sensor types, charge-coupled device (CCD)
and complimentary metal-oxide semiconductor (CMOS) are currently in use in commercial
systems. The mass production of CCD arrays has so far reached a higher maturity. CMOS
arrays are in general cheaper to produce and more susceptible to noise but have the advantage
that each cell can be read individually with a lower consumption of power. Multispectral
images can be derived by charging parts of the array differently, by filtering of the incoming
light or by splitting the incoming light into desired spectral components. The necessary tradeoff between spectral and spatial resolution remains similar as for analogue images.
For the derivation of DEMs panchromatic pictures with higher spatial resolution are
favorable, while for landcover classification and the detection of landcover changes
multispectral information is mostly indispensable.
Current spaceborne sensor development suggests that commercial systems with 25 cm GSD
will be implemented by 2012. Due to US regulations it is still under discussion if the full
resolution imagery or only a downsampled version (50 cm) will be available for general sale
(http://spatialnews.geocomm.com/dailynews/2009/jan/05/news2.html). This will probably
focus the application of airborne mapping systems on to the local scale, where they still can
achieve higher resolutions and on the use of multiple camera views for a very detailed
photogrammetric reconstruction of surfaces. Spaceborne imagery can be acquired within less
than 24 h for any area of the world and the data product can be delivered in less than 25 h.
The operational revisit time of spaceborne sensors is higher and in certain cases the delivery
time (minimized to 7 h after order) might be even less than the time required only to get the
tower clearance for an aircraft to start mapping (not including mapping and delivery time)

SafeLand - FP7
Page 40 of 401

Rev. No: 2
Date: 2012-02-15
[Eckardt et al., 2009]. For this reason the methodological review of analysis techniques will
concentrate on very high resolution and mainly local scale applications.
3.3. NEW SENSOR PLATFORMS
Recently a number of Unmanned Aerial Vehicles (UAV, Figure 6), often equipped with lowcost non-metric camera systems, have been used for applications such as vegetation mapping,
monitoring of crops or landslides and photogrammetric derivation of DSMs. They can be
deployed with great flexibility and were demonstrated to be especially useful for the
acquisition of sub-decimeter resolution imagery (Figure 7) on local scale.
Figure 6 : UAVs for cost-efficient acquisition of VHR imagery. Upper: [Laliberte and Rango [2009], Lower:
Niethammer et al. [forthcoming]
There they can supply imagery with such a resolution very cost effectively and with a
temporal resolution defined in accordance with the specific task. UAVs can be equipped with
GPS receiver to obtain a first rough approximation of the sensors exterior orientation (EO),
also referred as direct georeferencing. Consequently, via integrated geo-referencing,
remaining systematic errors of the direct georeferencing solution can be corrected with the aid
of additional image observations [Cramer, 2001]. A calibration of the camera previous to
image acquisition can define the interior orientation (IO) of the camera. EO, IO and a
topographic model of the underlying surface are combined to iteratively enhance the
parameters for the EO, which is used for orthorectification when a certain degree of accuracy
is reached. An RMSE (XY) of less than 0.5 m can be achieved with this approach [Laliberte
et al., 2008]. Similar accuracies can be achieved if instead of camera calibration or onboard
GPS receiver a dense grid of GCPs is distributed on the observed object and measured with
DGPS [Niethammer et al., 2009].

SafeLand - FP7
Page 41 of 401
Deliverable 4.1
Figure 7 : Displacement, surface fractures and soil moisture patterns as revealed from a UAV acquiered image.
Super Sauze mudslide, France [Niethammer et al., 2009].

The visual interpretation of aerial photography has a long history comprising applications
such as military operations, agriculture, forestry and geology or the creation of topographic
maps. Landslide inventories are traditionally prepared by a combination of visual
interpretation of aerial photographs and field work. Despite their time-consuming and labor
intensive nature, those are still the most common methods in many scientific studies [Hovius
et al., 1997; Korup et al., 2004] and the inventorization carried out by administrative bodies
[Hervs and Bobrowsky, 2009].
The human interpreter thereby relies on his personal experience and can take into account
color, tone, shape, size, pattern, texture, shadows and contextual information from single or
multiple images. Stereoscopy helps to simulate the third dimension, whereas the interpretation
benefits greatly in many applications. As the birds-eye view is a very unfamiliar perspective
for most people the success of interpretation depends largely on the experience of the
interpreter. Even still if experts are assigned with the image interpretation resulting landslide
inventories comprise a large degree of subjectivity [Galli et al., 2008; Wills and McCrink,
2002]. False color representations may help to enhance the contrast between background and
the area of interest. However, the human eye is in general less sensitive to spectral differences
than modern sensors and not all information contained in the image can be used at once by the
interpreter. Considering furthermore the time necessary for the visual inspection of large
areas, current and future development goes strictly toward automated and semi-automated
methods for image interpretation.

SafeLand - FP7
Page 42 of 401

Rev. No: 2
Date: 2012-02-15
3.5. AUTOMATED TECHNIQUES OF IMAGE INTERPRETATION

Despite certain differences in acquisition and pre-processing of aerial photographs most
analysis techniques apply similar for them as for satellite imagery. Thus, a generic overview
of image analysis methods will be given below under 3.4. However, the generation and of
multitemporal DTMs is a field where aerial photographs are still the dominant source of
stereo data and it is worth adding another paragraph overviewing studies that involved aerophotogrammetry for the study of landslides.
Comparison of multitemporal DTMs (Figure 8) is typically carried out by a simple
subtraction, whereas the subtraction between two epochs cannot provide the overall volume
of the displaced mass but, this difference allows an estimation of the volume of the uplifted
and subsided parts of the terrain between the two epochs [Weber and Herrmann, 2000].
However, average displacement velocities can be derived if the thickness of the landslide
body is known from geophysical measurements [Brckl et al., 2006].
Photogrammetric elaborated DTMs can be compared in this way also with DTMs from other
sources such as airborne LIDAR scans. An important prerequisite is in any case an accurate
alignment of the DTMs through the identification of homologous points on non moving
terrain parts [Bitelli et al., 2004]. Furthermore, effects of vegetation and changes of vegetation
have to be considered. Several filtering methods are available to remove vegetation from
LIDAR point clouds, whereas aerotriangulation is typically restricted to ground points. DTM
subtraction implies that individual DTM errors can be combined as independent random
variables. With the assumption of normal distribution, this combined error can be used as
standard deviation [Dewitte et al., 2008]. Repeated photogrammetric surveys were conducted
to monitor the eruption of Stromboli Volcano and the analysis of the multi-temporal DTMs
allowed the quantification of erosion/accumulation phenomena to determine a mass balance,
useful for the implementation of a stability analysis [Baldi et al., 2005]. A set of six historical
aerial photographs was used to create DTMs and evaluate the deformation of the La Clapire
landslide (France). It was noted that due to the frequency of aerial surveys this method is
mostly only suitable for slow moving landslides but also the potential of these data for slip
surface characterization was pointed out [Casson et al., 2005]. The activity of the Tessina
landslide (Italy) similarly has been documented with a set of multiple DTMs constructed from
historical aerial photographs [van Westen and Getahun, 2003]. Thereby, it was possible to
reconstruct activity and involved volumes through several decades.

SafeLand - FP7
Page 43 of 401
Deliverable 4.1
Figure 8 : Map of vertical ground displacements within a landslide in western Belgium inferred from the
comparison between DTMs from 1973 and 1996. DTMs were generated from photogrammetrically ground spot
heights. [Dewitte et al., 2008]
4. SATELLITE IMAGING
4.1. SENSORS AND PLATFORM DEVELOPMENT
More than 150 earth observation satellites are currently in orbit, whereas the majority carry
passive sensors, measuring electromagnetic radiation from the Earths surface or atmosphere
[Tatem et al., 2008]. Over 250 launches of private and public satellites are expected between
2009- 2018 [Euroconsult, 2009].
In recent years four main innovations contributing to enhance the value of datasets provided
by passive optical satellite sensors can be summarized.
1.
As mentioned frequently throughout this document the achieved Ground Sampling
Distance (GSD) of passive optical sensors reduced considerably during recent years. This is
achieved through transfer delay and integration sensors (TDI, e.g. IKONOS, Quickbird),
whereas the view of the sensor changes according to the speed of the satellite and the same
surface is scanned several times, or staggered CCD lines shifted half a pixel against each
other (e.g. Orbview-3, Spot-5) [Jacobsen, 2006].
2.
Previously used across-track acquisition mode for stereo imagery has been largely
replaced through satellites with along-track stereo imaging capabilities (ALOS Prism,
Cartosat-1, ASTER, and SPOT-5). This technique allows the acquisition of stereo imagery
with minimal time delay and decorrelation between the stereopairs.
3.
The development of innovative three mirror anastigmatic (TMA) telescopes enabled
the construction of a series of mini-satellites (e.g. RapidEye) mainly manufactured by the
British company SSTL. Those systems can provide high-resolution imagery at relatively low
costs and satellite constellations like the Disaster Monitoring Constellation (DMC) or
RapidEye enable daily revisit times.

SafeLand - FP7
Page 44 of 401

Rev. No: 2
Date: 2012-02-15
4.
Short revisit times, international agreements (International Charter Space & Major
Disasters, GMES SAFER), effort of private companies (e.g. Google) and web-based
communities (e.g. OpenStreetMap) enable access to VHR satellite imagery up to one day
after major disasters (compare Figure 9, Figure 10) and consequent fast assessment of
damages and critical points on the ground.
The highest resolutions commercially available are currently provided by the companies
Geoeye and Digital Globe. After OrbView-3 failed in March 2007 Geoeye is at the moment
operating with the Ikonos (1 m PAN , 4 m MS ) and Geoeye-1 satellites (0.41 m PAN, 1.65m
MS). The launch of Geoeye-2 with a GSD of 0.25 m is planned for the years 2011-2012,
whereas, due to current U.S. regulations, only a resampled version (0.5 m) of the data will be
available for public market.
Figure 9 : Satellite imagery of an area affected by landslides during the Haiti earthquake of 12/01/2010. The
first images were acquired 17.5h after the event and published 12h later, Source: Google Earth.
Since 2001 Digital Globe operates the Quickbird (0.61 m PAN, 2.44 m MS) satellite and
recently extended its fleet with the WorldView-1 (since 2007, 0.5m PAN) and WorldView-2
(since 2009, 0.5 m PAN, 2 m 8-band MS). Similar high resolutions are achieved by EROS-B
(0.7 m PAN), Cartosat-2 (0.82 m PAN), Kompsat-2 (1 m PAN, 4 m MS) the Russian Resurs
DK-1 [Petrie, 2008]. SSTL recently announced plans for the construction of a mini-satellite
named SSTL ART (0.6 m PAN, 2.4 m MS) to provide submeter resolution at less than 0.2
EUR per km2. The launch of the French satellites Pleiades-1 and 2 (0.7 m PAN, 2 m MS) is
scheduled for 2010 and 2011, respectively. EROS-C (0.7 m PAN, 2.8 m MS) is expected to
be launched already this year and KOMPSAT-3s (0.7 m PAN 2.8 m MS) launch is planned
for 2011.

SafeLand - FP7
Page 45 of 401
Deliverable 4.1
Figure 10 : Multitemporal imagery from RapidEye capturing surfaces changes during the Chilean earthquake
2010. Left: Forested region with steep slope in proximity to Constitucin (2010-01-22), Middle: Same scene
after the Chilean earthquake 2010-02-27 (hours after the event) showing exposures of bare soil due to
landsliding,
Right:
Results
change
detection.
Source:
http://www.rapideye.de/upload/documents/Press_Releases/ChangeDetection_Chile-Earth-Quake_2010-0304.pdf
Whereas the area of resolutions between 2.5 and 10 m has been formerly rather sparsely
occupied by SPOT-5 and EROS-A in recent years a number of larger Asian satellites and
SSLT based mini-satellites filled that segment. Avoiding an extensive listing of satellites only
a few systems will be explained here. The Indian Cartosat-1 (2.5 m PAN) offers geometric
and radiometric characteristics making it an interesting option especially for DSM generation.
It is complemented by the series of Indian satellites equipped with multispectral sensors
(LISS III 5.8 m, LISS IV 23.5, AWiFIS 56 m). The Taiwanese Formosat-2 (2 m PAN, 8 m
MS) imagery is available with worldwide coverage and offers similar imaging characteristics
as SPOT-5. Meanwhile Thailands THEOS satellite (2 m PAN, 15 m MS) is still waiting for
launch. A comprehensive satellite system including panchromatic (PRISM 2.5 m)
multispectral (AVNIR 10 m) and L-band radar (PALSAR) was launched in 2006 by the
Japanese JAXA. A number of mini-satellites including for example the RapidEye (Figure 10)
constellation and the UK-DMC 2 are largely based on the SSTLs platform 150 (2.5 m PAN, 5
m MS) and 100 (22 m MS), respectively As part of the SPOT Continuity Program, SPOT 6
and SPOT 7 are currently under development planned to be launched in 2012 and 2014,
respectively. The panchromatic channel of those systems is planned to provide ground
resolution of 2m. Similarly NASA is planning to continue the LANDSAT program with a
Data Continuity Mission (LANDSAT DCM) by December 2012.
A comprehensive overview of potentially interesting, active and planned satellite missions is
provided in Table 1. A more complete overview of active and planned sensors including a
large number of further SSTL-based microsatellites and platforms with dual use for military
and commercial applications is provided in the Appendix of this document.
Table 2 : Current and planned satellite missions for passive remote sensing of the earth surface with a minimal
resolution of 30 m. Note that the table was compiled from a great variety of online sources and future launch
dates might be a subject of frequent change.
Satellite
Landsat 7
Channels
panchromatic
7-bands multispectral

SafeLand - FP7
Resolution
[m]
15
30
Launch
15-Apr-99
Mission duration
end around 2012
Page 46 of 401

Landsat DCM
CBERS-2
CBERS-2B
CBERS-3
CBERS-4
Terra Aster
RESOURCESATRESOURCESATRESOURCESAT-
ALOS
Cartosat 1
SPOT 5
SPOT 4
SPOT 6
SPOT 7
Formosat 2
Kompsat 2
Kompsat 3
Ikonos
Cartosat 2
Cartosat 2A
EROS A
EROS B
EROS C
Quickbird
panchromatic
panchromatic
panchromatic
panchromatic
12-band multispectral
panchromatic
along-track stereoview
3band visible + NIR
3band visible + NIR
panchromatic
34 bands visible +
hyperspectral(200
bpanchromatic
d)
4band visible + NIR
along-track triplet view
along-track stereo
panchromatic
4band visible + NIR
along-track stereoview
panchromatic
4band visible + NIR
panchromatic
multispectral
panchromatic
4band visible + NIR
panchromatic
4band visible + NIR
panchromatic
4band visible + NIR
panchromatic
4band visible + NIR
panchromatic
panchromatic
panchromatic
panchromatic
panchromatic
4band multispectral
panchromatic

SafeLand - FP7
Rev. No: 2
Date: 2012-02-15
15
30
2.7
20 - 260
2.7
20 - 260
5 (10)
20 - 80
5 (10)
20 - 80
15 - 90
15
5.8
23.5 - 56
5.8
23.5 - 56
2.5
5.8
23.5 - 80
25???
2.5
10
2.5
2.5
2.5(5)
10
5
10
20
2m
8m
2
8
1
4
submeter
<4
1
4
1
1
1.9
0.82
0.7
2.8
0.61
2012
???
21-Oct-03
planned for 2 years
19-Oct-07
planned for 2 years
postponed to
planned for 2 years
2011
planned for 2 years
18-Dec-99
planned for 6 years
17-Oct-03
planned for 5 years
late 2010
planned for 5 years
mid 2011
planned for 5 years
23-Oct-06
planned for 5 years
5-May-05
4-May-02
planned for 5 years

planned for 5 years
24-Mar-98
planned for 5 years
2012
2012
21-May-04
10 years?
10 years?
planned for 5 years
28-Jul-06
planned for 4 years
2010
planned for 4 years
24-Sep-99
planned for 8 years
10-Jan-07
28-Apr-08
5-Dec-00
25-Apr-06
Sep-2010
planned for 5 years

planned for 5 years
planned for 10 years
18-Oct-01
planned for 5 years
Page 47 of 401
Deliverable 4.1
World View 1
World View 2
Pleiades 1
Pleiades 2
Geoeye 1
GeoEye-2
4band visible + NIR

panchromatic
panchromatic
8band multispectral
panchromatic
4band visible + NIR
panchromatic
4band visible + NIR
panchromatic
4band color
panchromatic
4band color
2.44
0.5
0.46
1.84
0.50
2.00
0.5
2
0.5
2
0.25 (0.5)
18-Sep-07
8-Oct-09
planned for 7.25 year

planned for 7.25 year
Oct-2010
2011
6-Sep-08
7-10 years
late 2012
7-10 years

VHR spaceborne remote sensing products can be expected to gradually replace aerial
photographs for the purpose of landslide mapping and monitoring. Earlier attempts to take
advantage of satellite imagery for landslide mapping still suffered from the rather coarse
resolution of Landsat imagery [Petley et al., 2002]. It has been demonstrated that the
integrated use of optical and radar data can partly compensate coarse resolution and is helpful
to characterize in high and low relief areas [Singhroy et al., 1998]. New sensors with higher
resolutions can already provide sufficient ground resolution for comprehensive inventorying
of large landslides and to some extent also mud- and debris flows [Haeberlin et al., 2005].
Several studies have taken advantage of VHR satellite imagery for visually interpreting the
imagery. It has been suggested that visual interpretation of spaceborne imagery is suitable for
the creation of inventories [e.g. Proske et al., 2008], the planning of ground campaigns [Owen
et al., 2008] and in certain cases also for the detection and assessment of fissures and tension
cracks [Youssef et al., 2009]. There are already VHR satellite missions operating with ground
resolutions up to 0.41 m (Geoeye-1) and up to eight spectral bands WorldView-2. Further
satellites with a ground resolution of up to 0.25 m (Geoeye-2) are under construction.
Keeping this in mind while revising Table 1 it can be concluded that spaceborne imagery can
already reach similar, if not superior quality compared to aerial photography. However,
problems of visual interpretation remain the same as stated in 3.3.4.
4.3. STEREOPROCESSING
One possible way to observe morphology or morphological changes of the earths surface is
the generation of Digital Surface Models (DSMs) from stereopairs images at one or
subsequent epochs, respectively. In digital photogrammetry automatic matching procedures
are employed for the 3D point measurement over wide areas. This implies that the same
object (pixel) is visible and can be identified in two images from different perspectives.
Several problems such as decorrelation within the two images, low resolution and low
performance of matching algorithm have been solved in recent years. Difficulties with
variable atmospheric conditions at different times of across-track image acquisition motivated
the implementation of a number of along-track mapping platforms, whereas the first of their
kind were the Aster sensor on board of the Terra satellite (1999) and Spot 5 (2002). Aster
DSMs have already been used in several geomorphological [e.g. Bubenzer and Bolten, 2008;
SafeLand - FP7
Page 48 of 401

Rev. No: 2
Date: 2012-02-15
Kb, 2005] and risk assessment studies [Fourniadis et al., 2007]. A global DSM derived
from the multiple stereopairs is available for free, however, with a rather high mean RMSE of
9.35 m and a rather low resolution of 30 m [ASTER-GDEM-VALIDATION-TEAM, 2009].
Costs of high resolution satellite images with stereo capabilities are decreasing and accurate
DSMs are becoming more attractive for a wider range of applications. An increased number
of sensors with along-track stereo-acquisition capabilities give a strong advantage in
comparison with former multi-date across-track stereo-data acquisition because the
radiometric image variations are less and the correlation success rate in any image matching
process increases.
In modern VHR satellites two different approaches are implemented to realize stereo-imaging
capabilities. Several satellites have one pointable lens and change the viewing angle to
capture stereo images of the same area along the same track (e.g. Ikonos, WorldView, and
Quickbird). As stereo imaging is reducing the imaging capacity, their price is considerably
higher than for one single scene (e.g. Quickbird 2.3 times higher price). During the time
needed for the turn-around of the lens to another view direction no other image can be
acquired. Thus, the required time restricts the imaging capacity (for Quickbird by a factor of
9). Newer pointable sensors such as the one on board of WorldView-2 are therefore designed
to change the viewing direction much faster [Byksalih and Jacobsen, 2007]. Another class
of satellites is equipped with multiple fixed cameras (two or three), which record images of
the same from different viewing angles due to the forward movement of the platform along
the track (e.g. Cartosat-1 with forward and backward, ALOS/PRISM with forward, nadir and
backward) [Wolff and Gruen, 2008]. To fully exploit both new imagery types Rational
Function Models (RFPs) [e.g. Zhang, 2005] and physical models [e.g. Toutin, 2006] became
main lines of research in recent years with considerable advances in terms of more stable and
accurate algorithms.

SafeLand - FP7
Page 49 of 401
Deliverable 4.1
Table 3 : Minimum size of objects to be recognized in various types of RS imagery (after Mantovani et al. [1996]
and Wang [2004], modified by Zhang et al. [2005])
As a rule of thumb the resolution of the photogrammetrically derived DSM should have a
resolution at two times lower (more for better accuracy) than the input data. It has been
demonstrated that under favorable atmospheric conditions, DSMs with a pixel-spacing of 2m
and 3m can be derived from Quickbird and Ikonos stereopairs, respectively [Byksalih and
Jacobsen, 2007]. An accuracy of vertical RMSE of 5 m and 1.5 m has been reported for
DSMs (5 m resolution) generated from stereopairs from WorldView-1 using no and one
accurate GCP, respectively [Cheng and Chaapel, 2008]. More recently Martha et al. [2010b]
demonstrated that in some cases GCPs are dispensable to obtain accurate landslide volume
estimations from multitemporal DSMs (Figure 11).
In general the DSM quality is largely dependent upon image resolution, landcover type and
the adopted image matching algorithm. A comparative study of different software packages
with Quickbird and Ikonos stereo imagery at different test sites revealed that a Multi-Photo
Geometrically Constrained (MPGC) as implemented in SAT-PP leads to superior results with
a heighting accuracy of close-to two pixels [Poon, 2007]. Expressed in pixels the DSMs
generated from ALOS PRISM [Maruya and Ohyama, 2008; Wolff and Gruen, 2008] achieve
very similar accuracies. Due to a unique design dedicated in particular to stereo imaging,
DSMs generated from the Indian Cartosat-1 (GSD 2.5 m) stand out by high accuracy. It has
been demonstrated that subpixel accuracy can be reached for a 10m spaced DSM [Martha et
al., 2010a]. Sacrificing some accuracy also DSMs with finer resolution (up to 2.5 m) can be
derived from the stereo pairs [Gianinetto, 2008].

SafeLand - FP7
Page 50 of 401

Rev. No: 2
Date: 2012-02-15
Considerable differences in accuracy also occur in dependence of the sun angle elevation and
the local valley orientation [Martha et al., 2010a]. The greater variety of available input data
and processing algorithms, a common standard for assessing the final products and
communicate it to the end user becomes desirable [Poon, 2007].
Figure 11 : Volumetric analysis of the Salna landslide, Indian Himalayas. Elevation difference due to
landsliding with lowering and rising parts of the surface [Martha et al., 2010b]
DSMs are an important input for orthorectification of any kind of imagery of the earth
surface. Higher resolution and accuracy of the DSM helps to reduce distortion and other
topographic effects and thereby reduces possible errors during later processing steps such a
change detection or image correlation.
Applications of monotemporal DEMs are, for instance, landform classifications [Schneevoigt
et al., 2008], structural geology [Murphy and Burgess, 2006] and ecosystem studies
[Tenenbaum et al., 2006]. In most hazard assessment procedures, such as for floods and
landslides, topography is an important factor and photogrammetric DEMs constitute a costeffective alternative where suitable ground-based information is not available [e.g. Nichol et
al., 2006; Roessner et al., 2000]. More recently Martha el al. [2009a] showed that
incorporation of post-failure DSMs derived from satellite stereo views are suitable to identify
the type of failure. Radar interferometric techniques show particular strength quantifying
small or subtle surface deformation and laser scans are a highly accurate source to quantify
surface deformation and displaced volumes. For sudden large scale events they are often not
suitable and photogrammetric techniques become more attractive. Several studies
demonstrated the usefulness of multitemporal DSMs from satellite captured stereopairs to
detect and quantify ground deformation processes [Martha et al., 2009b; Tsutsui et al., 2007].
While most photogrammetric investigations of landslide volumes and ground deformation
SafeLand - FP7
Page 51 of 401
Deliverable 4.1
still rely on the aerial photographs that provide higher resolution [Dewitte et al., 2008;
Walstra et al., 2007] it can be expected that the new VHR sensors with submeter GSD will
gradually replace aerial photography in the future.
4.4. SUPERVISED CLASSIFICATION TECHNIQUES
Despite a great variety of possible methods supervised classification typically involves the six
following processing steps.
1.
Deciding the number and nature of targeted classes
2.
Choose a representative set of training pixels for all desired classes
3.
Training of a chosen classifier algorithm with the training data
4.
Apply the parameters estimated by the classifier to the whole image
5.
Produce tabular summaries of the obtained classification results
6.
Accuracy assessment using test data
A practical minimum for the number of training pixels is ten times the number of features
(typically the number of bands) employed for the classification. For satellite images with a
low number of spectral bands this number of samples can mostly be provided relatively easy
while the task gets more difficult using hyper spectral data.
The most common supervised approach for image classification is the Maximum Likelihood
Classification (MLC). Based on a set of samples, probability functions are derived in
multivariate feature space and used to calculate the probability of each pixel in the image to
belong to one of the given classes. Finally, the class with the highest probability is assigned to
the pixel. Additionally prior probabilities for the occurrence of a class can be integrated in the
classification and the definition of accuracy thresholds is helpful to avoid misclassifications of
pixels with low preferences toward a certain class membership. Borghuis et al. [2007] used a
supervised MLC for the detection of typhoon triggered landslides in Taiwan.
If the amount of training data and/or computational capacity is limited the Minimum Distance
Classification is an appropriate alternative. Thereby, class memberships are defined by the
shortest (mostly Euclidean) distance to the centre of a given class. A more sophisticated
extension of the approach is the Mahalanobis Classifier that accounts for the scaling of the
coordinate axes, correlations between the different features and can provide curved as well as
linear decision boundaries. However, the computational expense grows quadratic with the
number of features. The k-nearest neighbor algorithm defines the class membership of a pixel
based on the class membership of the k nearest neighbors within the feature space. K defines
the number of considered neighbors and is usually chosen to be an odd number between 1 and
5 or larger to reduce the effect of noise within training data. Parallelepid Classification can be
considered as a very simple classification method. A bounding box is constructed around a set
of samples in multivariate space. Classes are consequently assigned according to the position
of the unclassified pixel within those bounding boxes. However, this approach comprises

SafeLand - FP7
Page 52 of 401

Rev. No: 2
Date: 2012-02-15
several drawbacks in the classification of pixels that lay outside or in the overlap of the
bounding boxes.
It has been recognized that pixel-based approaches lack to consider contextual information
and Markov Random Fields have been successfully introduced to a large number of image
classifications tasks [Serpico and Moser, 2007]. The underlying assumption is that the class
value of a pixel is to some degree conditional independent to its neighborhood. The
probability density function among a local neighborhood is determined and used to remove
outliers.
Experimental results on Formosat-2 imagery indicate that this technique can increase the
accuracy of landslide detection [Chang, 2006; Hsieh, 2006].
Another large group of supervised classifiers are non parametric models such as Linear
Discriminators, Support Vector Machines (SVM), Artificial Neural Networks (ANN) and
Random Forests. Many of them have been applied for landcover classification [e.g. Gislason
et al., 2006; Kavzoglu and Colkesen, 2009]. Danneels et al. [2007] compared MLC and ANN
to detect landslides on ASTER images, whereas ANN provided the better results. The
obtained map, stating the probability of each pixel to belong to a landslide, was refined by
applying threshold segmentation on the probability map. Consequently, a larger number of
false positives were eliminated through expert-knowledge based on decision rules including
for example thresholds of slope. Marcelino et al. [2003] evaluated several simple supervised
classifiers (MAXVER, Bhattacharyya, ISOSEG) combined with a variety of image fusion
techniques for landslide mapping with Landsat TM imagery in southern Brazil. From visual
assessment of the results they concluded that Wavelet Transform and the MAXVER classifier
led to the best results. The study was later continued demonstrating positive results for
general landcover classification but, due to regeneration of vegetation there was strong
confusion between landslide and pasture areas [Marcelino et al., 2009].
Image texture can alternatively be used to pixel-based information in supervised
classification. For instance, Hervs and Rosin [1996] used a supervised semi-automated
texture discrimination technique based on the image texture spectrum, which quantifies
texture in terms of frequency distribution of local pixel intensity variation, to discriminate
landslides with hummocky surface or multiple scarps on high-resolution Deadalus ATM
imagery of a semi-arid area in Spain. This supervised technique proved superior to
unsupervised texture measures derived from grey-level co-occurrence matrix statistics.
Incorporated into a classification scheme the additional information from texture helped to
distinguish the vegetated surfaces of the landslides from surrounding pasture and enhanced
the accuracy. Several other studies tested a number of different GLCM texture measures to
distinguish between landslide-affected areas and stable parts [Carr and Rathje, 2008;
Fernndez et al., 2008; Whitworth et al., 2002; 2005; Whitworth et al., 2006]. They generally
agreed that texture is able to enhance class seperability as well as the accuracy of the
classification and highlight GLCM entropy as an especially useful measure.
4.5. UNSUPERVISED CLASSIFICATION AND CLUSTERING TECHNIQUES
Unsupervised classification and clustering techniques can be used to obtain a first overview of
the inherent structure of a given dataset. On the one hand this technique does not comprise the

SafeLand - FP7
Page 53 of 401
Deliverable 4.1
need of sample, but on the other hand, obtained classes must usually be interpreted and
labeled by the user. Typically the user defines the number of targeted classes and a clustering
algorithm assigns pixels to the different classes according to their position in a multivariate
space. Many different algorithms using different similarity measures have been proposed.
Those used commonly are based on simple distance measures in multivariate space [Richards
and Jia, 2006]. Borghuis et al. [2007] employed unsupervised classification using SPOT-5
imagery combined with a slope mask what produced a 63% area concordance with the
manually mapped landslides. The performance was better than with other supervised
classification methods. Unsupervised classification was also used by Dymond et al. [2006],
taking into account bright areas on slopes only steeper than 5. However only landslides
greater than 10,000 m2 were checked and no independent data set was used for verification.
Testing eight different methods for landslide mapping with SPOT-5 imagery Joyce et al.
[2008] concluded that the spectral angle mapper (SAM), supervised classification and NDVI
thresholding were the most accurate semi-automated techniques compared with the results
achieved with parallelepiped, minimum distance to means, principal components, and
multitemporal image differencing.
4.6. CHANGE DETECTION
Joyce et al. [2009] noted that if multi-temporal imagery of a given area can be provided
change detection is certainly the most promising approach for landslide mapping and
monitoring. Change detection has a long known history and most methods have been
developed for the evaluation of changes in land-use and land-cover. A great number of
techniques is available and can be grouped into algebra, transformation, classification,
advanced models, GIS approaches, visual analysis and other approaches [Lu et al., 2004].
Despite the great variety of proposed approaches, change-detection methods usually comprise
a modeling phase and a subtraction phase. Thereby the modeling phase is the application of
algorithms to infer meaning from spectral data. Subtraction compares data from different
dates via image algebra or other methods [Kennedy et al., 2009]. While there is no general
agreement on the best available method Differencing, PCA, Change Vector Analysis (CVA)
and Post-Classification Comparison are at present most frequently used [Lu et al., 2004].

SafeLand - FP7
Page 54 of 401

Rev. No: 2
Date: 2012-02-15
Figure 12 : SPOT imagery showing the evolution of landslide from before (a) and after two typhoons (b,c),
[Rau et al., 2007]
For landslide mapping Cheng et al. [2004] carried out change detection by differencing the
band ratios of IR and red channel of SPOT images. A change-threshold was manually defined
from the histogram of the difference image and a slope mask of 22 was applied to filter out
changes not related to landslides. Nichol and Wong [2005] tested several change detection
methods to detect landslides, including subtraction of single bands and band ratios and postclassification comparisons using the Neural Network classifier and the Maximum Likelihood
classifier (MLC). Best results were obtained using MLC and led to detection rates of 67% in
grassland areas and 71% in woodland areas. Rau et al. [2007] examined a series of optical
satellite imagery with a temporal resolution of about 19 days for a period of 2.5 years (Figure
12). They combined CVA and NDVI for change detection with empirical thresholding and
were able to monitor seasonal dynamics and landslide triggered by three major typhoons.
Hervs et al. [2003] combined image differencing with an automated thresholding technique
to differentiate between change and no change areas. The technique was applied on simulated
IKONOS imagery of the Tessina landslide (Italy). Discrimination of landslide active and
stable areas was further accomplished by applying a rectangular blob filtering technique on
the thresholded difference image. The method was also demonstrated useful for long-term
landslide monitoring, especially when datasets with even better spatial and temporal
resolution are getting available.

SafeLand - FP7
Page 55 of 401
Deliverable 4.1
4.7. DIGITAL IMAGE CORRELATION (DIC)

Measuring the offset of pixels in at least two consequent images of an area can help to
coregister the images and to gain information on the direction and magnitude of horizontal
ground displacements. Different implementations of this concept have been proposed in the
past [Crippen, 1992; Puymbroeck et al., 2000; Strang and Nguyen, 1996; Yamaguchi et al.,
2003] to measure for example ground displacement along ruptured faults or due to slow
moving landslides. The increased availability of high resolution earth observation products
and a number of sophisticated algorithms readily implemented in software tools (Correli [Hild
et al., 2002], MIC-MAC [Pierrot-Deseilligny and Paparoditis, 2006] or MEDICIS [CNES],
PICMATCH [JPL]) contribute to a recently more frequent use of this concept for observation
of terrestrial and also extraterrestrial surfaces [e.g. Tokumaru and Dimotakis, 1995].
Delacourt et al. [2007] summarized the prerequisites for the application of DIC as follows:
The correlated images have to share a common (ground or image) geometry, which is
obtained either by orthorectifying both images (the correlation is performed in the ground
geometry) or by resampling a secondary image in the geometry of a reference image
(correlation performed in the image geometry). A DEM is necessary in both cases and its
required accuracy is inversely related to the difference in viewing angles of the correlated
images. Ideally, two different DEMs, contemporary to each correlated image, should be
used.
Kb [2002; 2004], Berthier et al. [2005] and Scherler et al. [2008] applied DIC on aerial
photographs and imagery from ASTER and SPOT and obtained surface velocity fields with
accuracies in the range of 1 to 1/20 of the respective pixel sizes.
Yamaguchi et al. [2003] were among the first to used DIC techniques for landslide
monitoring and successfully detected displacement within correlating SPOT imagery.
However, they did not provide information on the accuracy of the obtained displacement
vectors. Delacourt et al. [2004] used a combination of aerial photographs and Quickbird
imagery to determine the displacement field of the La Clapire landslide. They conclude that
practically, measurement of displacement reaching from 0.1 to hundreds of pixels can be
performed. With VHR imagery (GSD 1 m) motion as low as 0.1 m can be detected, whereas
the accuracy strongly decreases if imagery from different sensors is compared. The study has
later been extended using a series of aerial photographs for DIC and the generation of
corresponding DEMs at each time step. Obtained displacements were found to be in excellent
accordance with ground measurements from electronic distance-meters [Casson et al., 2005].
More recently a user friendly software tool has been developed and its applicability for
operational use demonstrated [Leprince, 2008]. Optical correlation can be applied with sub
pixel accuracy over few years where vegetation causes a loss of coherence for DInSAR
techniques (cf. section 2.C.3). The increase in spatial resolution of the new satellites will
allow a better accuracy in the computation of lateral displacements using correlation
techniques once a comprehensive archive will be acquired [Delacourt et al., 2009].
4.8. OBJECT-ORIENTED CONCEPTS
Due to the increasing availability of VHR remote sensing datasets and the raise of
professional software tools Object-Based Image Analysis (OBIA) has been introduced in a

SafeLand - FP7
Page 56 of 401

Rev. No: 2
Date: 2012-02-15
wide range of scientific disciplines, especially during the last decade. OBIA, also referred as
GEOBIA (Geographic Object-Based Image Analysis) or OOA (Object-Oriented Analysis),
builds on long known methods for segmentation, edge-detection, feature extraction and image
classification. In a first step a segmentation algorithm divides the image into meaningful
homogenous areas (Figure 15, Figure 16). Consequent classification labels the derived objects
based on rules from expert knowledge and/or training-data. It is frequently pronounced that
the Object-Based approach can overcome several restrictions of traditional pixel-based image
analysis methods [Blaschke, 2010].
Due to greater detail in VHR resolution images statistical seperability between classes
gets reduced, what leads to a salt-and-pepper appearance of the classification results
Objects have additional spectral information compared to single pixels (means,

medians, minimum, maximum, variance, mean ratios in different bands, etc.)
Segments reveal additional spatial information (shape, size, distances, neighborhood,

topologies etc.)
The analysis of geographic objects is in better agreement with the human perception
and the composition of the earth surface
OBIA is more suitable for multi-scale analysis approaches
Facilitated mining, -querying and -modeling of remote sensing data with richer
semantics, and greater integration with vector-based GIS.
The integration of additional datasets as object features is relative straight-forward.
With respect to the implemented algorithms, image segmentation can be divided into four
categories, which are point-based, edge-based, region-based and combined approaches.
Deeper insights into the technical implementation of these segmentation concepts can be
gained from works such as Haralick and Shapiro [1985] and Pal and Pal [1993]. A crucial
step of OBIA states the choice of the segmentation method and the corresponding parameters.
Considering the Shannon sampling theorem, an object-size should be of the order of ten
pixels to ensure that it will be completely independent of its random position and its
orientation relative to the sampling scheme. Region-growing algorithms have become a
widely used segmentation method and similar to other segmentation methods, require a
number of parameters to determine the maximum degree of heterogeneity. Criteria for
heterogeneity can be defined through color (grey values in different channels), shape, texture,
class memberships and others. In any case, a suitable set of parameters constraining the
degree of heterogeneity within single objects and/or among all objects has to be determined.
Allowing a larger degree of heterogeneity in general results in larger objects. Consequently,
the choice of the parameters is directly related to the resolution of the image and the desired
scale of the analysis. The choice of the maximum heterogeneity and scale of the analysis can
be adjusted to the heterogeneity and size of a group of target objects. At the same time, we
can observe that in many cases significant objects emerge at different analysis within the
same image [Blaschke, 2010], what complicates the task of finding an appropriate scale. This
is especially the case if the process of interest manifests on different scales or if little prior

SafeLand - FP7
Page 57 of 401
Deliverable 4.1
information about the scale of a respective process is at hand. In many works the evaluation
of the best segmentation method and parameters is still done by simple trial and error and
visual inspection of the obtained results (subjective evaluation), whereas several studies
already investigated methods for an objective evaluation and optimization of the
segmentation[Zhang et al., 2008a].
Figure 13 : Conceptual overview of methods for segmentation optimization [Zhang et al., 2008a]
System-level evaluation (Figure 13) can provide a statement on the most suitable
segmentation methods for a specific set of consequent processing steps. Analytical methods
evaluate the performance of an adopted algorithm in terms of computational efficiency but are
not appropriate to judge the obtained results. Supervised methods comprise the need of
defined objects of interest as delineated on the ground or by a human interpreter.
Discrepancies between geometry and/or feature values of manual and automated delineated
objects are then used to determine the segmentation with the best fit. However, providing the
necessary samples is time consuming and especially if the characteristics of the objects of
interest vary strongly in shape, size, color and texture there might be no unique solution.
Furthermore, it can be argued whether subjective delineated objects are really a good
reference [Zhang et al., 2008a] Within unsupervised evaluation methods a great number of
criteria evaluating intra-region uniformity, inter-region disparity and semantic cues exist. The
resulting best segmentation varies considerably on the chosen criteria and the performance of
each criteria differs largely from image to image[Zhang et al., 2008a].
We can conclude that for relative homogenous objects like houses or crops it is possible to
estimate one best segmentation method and one best set of related parameters using
supervised and unsupervised empirical methods. However, dealing with objects such as
landslides or vegetation patches, which differ considerably in shape, size, texture and color,
finding the optimal segmentation becomes an ill-posed problem and proposed approaches
appear not suitable.
A trial and error evaluation by a human interpreter is one possible choice but might not be
satisfactory due to the large amount of subjectivity and, as it is time consuming, it reduces the
benefits of the automated image processing. An evaluation on the system-level is possible if
sufficient samples for all objects of interest can be provided [e.g. Laliberte and Rango, 2009].

SafeLand - FP7
Page 58 of 401

Rev. No: 2
Date: 2012-02-15
A strictly different solution for the problem of finding the optimal segmentation method and
parameters was recently conceptualized by Baatz et al. [2008] as illustrated in Figure 14.
Figure 14 : The conceptual workflow of object-oriented image analysis [Baatz et al., 2008]
Instead of basing the classification process on one static image segmentation the process starts
with object primitives that are altered during iterative steps. This includes for example the
definition of subsets after an initial segmentation and the local refinement of their borders. It
also includes the possibility to look at objects on different hierarchical levels, whereas statistic
measures can be employed to evaluate relationships between different levels (e.g. number or
standard deviations of sub-objects) or to the local neighborhood within each level (e.g.
difference to neighborhood objects). In each iterative step further object characteristics,
knowledge and semantics can be integrated approaching the form of an object of interest.
During each step statistic measures are stored in variables. Absolute and relative values as
well as their evolution during processing steps (e.g. persistency of borders) are assumed to be
an important tool for auto-adaptive strategies to provide robust results relatively independent
from the initial segmentation parameters and over any expected variability of object
characteristics [Baatz et al., 2008].
A further important advantage of the OBIA approach is the possibility to incorporate
additional datasets as object features. According to the classification of Dong et al. [2009] this
falls into the category of feature level data fusion. Thereby the original datasets undergo no
physical change but are present as object features to enhance boundaries and classification
efficiency. This includes for example the integration of multitemporal data [Park and Chi,
2008], DEM derivates [Barlow et al., 2006] or thematic data such as river network [Martha et
al., 2009a]. As most classification techniques build on the assumption that given object
classes can be recognized as homogenous clusters in feature space it is recommendable to

SafeLand - FP7
Page 59 of 401
Deliverable 4.1
differentiate a higher number of classes (e.g. active vs. inactive landslides, debris flows vs.
deep seated rock slides) as the number of integrated additional features (from the same or
different datasets) increases. If this is not considered clusters may include many outliers and
get very dispersed among feature space [e.g. Moine et al., 2009].
Beyond the potential benefits for landslide mapping and monitoring (further outlined in D
4.3) valuable data for hazard and risk assessment comes from neighboring disciplines with
OBIA approaches. It has been demonstrated that the concept is suitable for the detection of
geological lineaments [Mavrantza and Argialas, 2008] or fault lines [Marpu et al., 2008],
geomorphological mapping [van Asselen and Seijmonsbergen, 2006], the quantification of
landcover changes [e.g. Riedel et al., 2008] or the indirect evaluation of social vulnerability
[Ebert et al., 2009]. A comprehensive review of case studies and current developments in the
application of OBIA concepts was recently provided by Blaschke [2010]. From there it can be
concluded that the concept is so far most frequently used in landcover and land use
classifications with especially good results in urban environments.
Figure 15 : Result of image segmentation on post-landslide image at Messina, Italy 2009 (Quickbird, 0.41 m
pansharpened). a, The white arrows in the input image indicate the flow path of several debris flows, b, Results
after image segmentation The resulting image objects have additional features such as shape, size and texture,
which can be used to enhance the classification.

SafeLand - FP7
Page 60 of 401

Rev. No: 2
Date: 2012-02-15
Figure 16 : Result of image segmentation on post-landslide image at Wenchuan, China 2008 (ALOS, 10 m
multispectral). a, The orange arrows in the input image indicate the position and direction path of several rock
slides, b, Results after image segmentation The resulting image objects have additional features such as shape,
size and texture, which can be used to enhance the classification.
OBIA is still a very young and emerging discipline what comprises a number of general and
landslide specific challenges. As mentioned above a meaningful integration of additional data
(e.g. established landslide inventories, DEMs, multi-temporal images), parameters (e.g.
hierarchies between scars, deposits, toes, reactivation of deposits) and methods (e.g. change
detection) is needed. New methods and tools for accuracy assessment (e.g. thematic and
geometric comparison with expert inventories) dealing with object specific characteristics
such as fuzzy thematic information and complex geometries [Schpfer et al., 2008] should be
tested. In the pan-European context of SAFELAND further landslide specific case studies can
investigate the adaptability of OBIA workflows to different environments. First steps should
include a comprehensive statistical evaluation of assumable landslide specific object-features
on a broad set of images of different processes from different regions. Any kind of semiautomated or fully automated system will need a comprehensive amount of training and test
data, what should be considered in the implementation of a European landslide inventory.
A database of well documented samples or samples interactively provided by an expert are
useful to train stochastic models for image information mining such as described by Datcu et
al. [2003], (Figure 17). Similar to the hierarchical order of objects and sub-objects in OBIA
the system uses a multi-resolution image pyramid and furthermore, an initially unsupervised
clustering of the image content based in multidimensional feature space (grey values and
texture). A set of class primitives is provided to a user that labels the object and thereby
provides the training data to the system. A Bayesian network is fed by that training data and
labels the unsupervised part of the image based on similarities, or more specific on their
proximity in feature space.

SafeLand - FP7
Page 61 of 401
Deliverable 4.1
Figure 17 : Conceptual representation of Bayesian model for image mining [Datcu et al., 2003]
The formalization and implementation of such models in combination with objective methods
for the determination of segmentation and classification parameters is a promising approach
to take full advantage of the increasing availability of remote-sensing data from supranational
initiatives (GEOSS, GMES) on an operational basis. Considering environmental changes (e.g.
climate change, urbanization) the latter is essential for a frequent updating in hot spot areas
and quick response after major events. In a way OBIA also converges with current efforts of
object-oriented modeling of a European spatial data infrastructure (INSPIRE;
http://inspire.jrc.ec.europa.eu). This may simplify the interoperability with ground-based
monitoring systems, the construction of comprehensive knowledge bases and the
communication of results in the future.

SafeLand - FP7
Page 62 of 401

Rev. No: 2
Date: 2012-02-15
5. APPENDIX
Appendix 1: Main Medium Resolution Satellites Launched to 2007, Planned/Anticipated:

2007-2016, [Euroconsult, 2008].
Appendix 2: Governmental remote sensing programs of India (ISRO) and South Korea
(KARI),[Euroconsult, 2008].

SafeLand - FP7
Page 63 of 401
Deliverable 4.1
Appendix 3: Roadmap for European defense Earth observation satellites [Euroconsult, 2008].
Appendix 4: Commercially operated optical satellites launched to 2007, Planned/Anticipated:

2007-2016,[Euroconsult, 2008].

SafeLand - FP7
Page 64 of 401

Part B Active Optical Sensors
Rev. No: 2
Date: 2012-02-15
PART B
ACTIVE OPTICAL SENSORS
SECTION 1: Basic Principles of Laser Ranging
p. 66
SECTION 2: Electronic Distance Meter (EDM)
p. 68
SECTION 3: Terrestrial Laser Scanner (TLS)
p. 76
SECTION 4: Aerial Laser Scanner (ALS)
p. 91
Sections 1, 2, 3 and 4
- Authors: A. Abelln1, M. Jaboyedoff1, C. Michoud1, M.-H. Derron1
- Reviewer: T. Oppikofer2,3
1
. Institute of Geomatics and Risk Analysis (IGAR). Universit de Lausanne, Switzerland. antonio.abellanfernadez@unil.ch
. International Center for Geohazards (ICG), Oslo, Norway.
3
. Geological Survey of Norway (NGU), Trondheim, Norway.
2
Abelln A., Jaboyedoff M., Michoud C., Derron M.H., Oppikofer, T., 2010. Active Optical Sensors (LASERS). In the Deliverable 4.1 of the
European project SAFELAND: Review of Techniques for Landslide Detection, Fast Characterization, Rapid Mapping and LongTerm Monitoring. Edited in 2010 by Michoud C., Abelln A., Derron M.-H. and Jaboyedoff M. Available at http://www.safelandfp7.eu

SafeLand - FP7
Page 65 of 401
Deliverable 4.1
This part of Deliverable 4.1 (Review of methodologies for landslide detection, fast
characterization, rapid mapping and long-term monitoring) is focused in the description of
Active Optical Sensors, also referred to as LASERs (Light Amplification by Stimulated
Emission of Radiation). A laser is a device that produces and emits a beam (or a pulse series)
of highly collimated, coherent, monochromatic, directional and in phase electromagnetic
radiation.
This part is divided into four sections. Section 1 describes the basic principles of laser ranging
and scanning. Sections 2, 3 and 4 describe the different techniques, Electronic Distance Meter
(EDM), Terrestrial Laser Scanner (TL) and Aerial Laser Scanner (ALS), respectively. These
three techniques are complementary: whereas EDM allow for a high accuracy (mm scale)
acquisition of single points, scanning systems (TLS and ALS) allow for the acquisition of
million points with centimetre-level accuracy.
Basic principles of the laser technique and methods for ranging, scanning, filtering and/or
aligning of the laser datasets will be described in the following sections. Furthermore, a
description of the main applications (distance calculation, monitoring, landslide displacement
computation, structural analysis, etc.) for landslides will also be discussed in this report.
Finally, some practical case studies on the use of this technique are provided in the Chapter 4
of this Deliverable.
Direct visibility is required between the electronic device and the surveyed point. Distance
calculation was initially carried out at specific points using reflectors (referred to as prisms or
targets), which return the laser beam into the direction of its source. Currently, distance
measurements can be carried out directly over the terrain surface (i.e., without artificial
reflectors), which is highly recommended in hazardous areas.
Two interesting books have been recently published: Topographic laser ranging and
scanning, by Shan and Toth (2008) and Laser Scanning for the Environmental Sciences,
by Heritage and Large (2009). The former reflects the state-of-the-art of all the physical
aspects of the sensor (components, calibration, waveform analysis, quality control of LIDAR
data, filtering and feature extraction techniques, etc). The latter provides a number of
applications of the LIDAR technique to the environmental and geosciences, being of great
value for those researches in the field of geomorphology, geostatistics, remote sensing and
GIS.
1. BASIC PRINCIPLES OF LASER RANGING

1.1. DISTANCE CALCULATION
Two different methods for distance (also called range) calculation are possible (Wehr and
Lohr, 1999): phase and pulse method. The former allows more accurate range determination
but suffers from a limited range (Petrie and Toth, 2008). Alternatively, the latter allows a
greater range, being this circumstance the main reason for which most of the ALS and TLS
used for Earth surface observations usually employ pulse laser method (Wehr and Lohr, 1999;
Baltsavias, 1999).

SafeLand - FP7
Page 66 of 401

Rev. No: 2
Date: 2012-02-15
1.1.1. Pulse method

Distance measurement (range: , fig. 1) is based on the Time-Of-Flight (TOF) of the laser
pulse, as already stated in equation 1:
= c (TOF/2)
[equation 1, Petrie and Toth, 2008]

Where c = speed of light
Figure 18: Basic operation of a laser rangefinder (from Petrie and Toth, 2008).
The range measurement is performed on an area equal to the spot dimension (SD; also called
beamwidth). Although the laser signal is highly collimated, the spot dimension increases with
distance due to the laser beam divergence (equation 2):
SD= tan + a
[equation 2, Petrie and Toth, 2008]
Where = angle of divergence; a = initial beam size;

Equation in any consistent units of SD and .
Figure 19 shows the shape of the waveform of the returned pulse. The recording of the last
pulse can be performed in order to obtain the return signal from the ground surface (in place
of vegetation).
Figure 19: View of the complete waveform of the returned pulse. The first and last pulse can be seen in this
picture (from Petrie and Toth, 2008; original figure by Brenner, 2006).

SafeLand - FP7
Page 67 of 401
Deliverable 4.1
1.1.2. Phase comparison method

Range calculation can also be performed using laser radiation emitted through a continuous beam, instead
of a series of pulses. Range is calculated in a two steps process (Figure 20):
o
Firstly, a rough distance value is calculated by measuring the total number of wavelengths. In order
to solve the ambiguity problem, a multi-frequency phase measuring system is normally
employed: the wavelength of the laser signal is rapidly changed n times, allowing for the univocal
calculation of the number of waves.
Secondly, a more precise calculation is obtained by comparing the differences in phase between
the transmitted and received signals of the laser beam.
Although its high accuracy (submillimetre level), this method for distance calculation suffers for a limited
maximum range, typically below 100 m. Hence, its practical applications in landslide detection,
characterization and monitoring are limited.
Figure 20: Fundamentals of range measurement through phase comparison method.
2. ELECTRONIC DISTANCE METERS

2.1. GENERAL OVERVIEW OF THE SENSOR
An Electronic Distance Meter (EDM) is an instrument that sends and receives a laser beam,
allowing for distance calculation depending on the properties of the returned signal (time of
flight or phase difference, as explained above). EDM were developed during the late 1950s
and early 1960s and started to be used by surveyors for distance measurements (Dallaire,
1974) and deformation monitoring.
Main components of a laser device are the laser source material, an energy source and a
mirror. Lasers are usually classified according to its source material: gas lasers, solid-state
lasers and semiconductor lasers. The study of the different source materials, physics of the
laser beam, energy source, etc. are beyond the scope of this report, which focuses on the
practical use of laser sensor for landslide studies. Nevertheless, the book by Petrie and Toth
(2008) provides a detailed overview of the main principles of laser components.
EDM are specifically designed for distance calculation. When using EDM in combination
with other devices able for angle calculation (e.g. theodolites), 3D coordinates (x, y, z) of
single points can be obtained. These two types of measurements (distance and angle) are
SafeLand - FP7
Page 68 of 401

Rev. No: 2
Date: 2012-02-15
usually conducted using a Total Station. First total stations were developed in the 1980's. A
modern device is shown in Figure 21.
Figure 21: The TOPCON GPT-8200 Series Reflectorless - Robotic Total Station (picture from topcon.com)
The main advantages of this sensor are the precision of the measurements (millimetre level),
the low cost and the high potential of automation.
Following devices are commonly used in combination with EDMs and/or total stations:
o Tripod: this object is normally used to stabilize and positioning the surveying
instrument. Alternatively, EDM may be placed over a permanent base, e.g. concrete
pillars.
o Reflector (Figure 22): devices designed for the reflection of the laser signal back to its
source (i.e., the EDM or total station) with a minimum loss of energy.

SafeLand - FP7
Page 69 of 401
Deliverable 4.1
Figure 22: : frontal and lateral view of a prism reflector.
2.1.1. Accuracy
Lasers allow for very accurate measurements in one dimension (range or distance). The
typical accuracy of the EDM is 1.5mm within maximum distances of about 800-1000m
(Manetti et al., 2007). Electro-optical distance meters based on infra-red light allow for the
acquisition of single points with an accuracy (1) of up to [1mm + 10-6 * distance] (Settles
et al., 2008).
It has to be mentioned that the instrumental accuracy is affected by unfavourable conditions,
as follows:
o high reflecting surfaces;
o very rough surfaces;
o weather conditions;
o very bright ambient conditions;
o excessive range.
EDM accuracies also depend on the operating mode. When using reflectors, range
measurements are more accurate than those measurements directly reflected on the ground
surface, due to the higher intensity of the returned signal
Laser precisions required for long term deformation (e.g. monthly or yearly changes) are
lesser than that required for short-term measurements (e.g. surveying and/or daily changes).
Changes in the geometry of the slope are usually plotted using a 3D vector or alternatively
computing the horizontal and vertical angles.

SafeLand - FP7
Page 70 of 401

Rev. No: 2
Date: 2012-02-15
2.1.2. Resolution
Spatial resolution (also referred to as point density) of this sensor is on demand, i.e. from few
points per m2 up to few points by km2. These sensors allow for a high accurate acquisition of
single points of the terrain surface but are not designed for a wide coverage of the whole
slope.
Typical point density for mapping purposes is around 50-100 points per km2 (the higher the
topographical variability of the area, the higher the point density should be).
Typical point density for monitoring purposes is from 20 to 30 control points per analyzed
slope (the higher the number of points and the accuracy of the sensor, the more reliable the
results are).
2.1.3. Maximum range
Maximum range depends on the type of instrument and reflectivity of the surveyed point.
Very high distances (up 5.000 meters) can be acquired when using artificial reflectors (e.g.
prisms). In opposition, maximum range decreases its value up to 5 times when directly
reflecting the laser over the terrain surface (max. range up to 1000 m).
2.2. METHOD
2.2.1. Data acquisition
Best practices for distance calculation (laser ranging) are as follows (USACE, 2002):
o A tripod and theodolite must be accurately plumbed and levelled over a fix base (e.g. a
tripod or a concrete foundation). Concrete pillars with centring plates guarantee an
increased accuracy (Malet et al., 2002) by preventing eccentricity errors and by
facilitating positioning and trimming of the apparatus.
o A "warm up" period prior to the use of the total station is recommended.
o In case of using reflectors, they must be must be placed perpendicularly to the total
station.
o In order to assure the repeatability of the measurement, the measurement should be
conducted at least 3 times (five readings are recommended).
o Measurements should be corrected taken into account the meteorological influence.
Barometric pressure and temperature should be measured in each station in a shadow
emplacement, and not exposed to direct wind. Formulas for atmospheric correction are
explained in chapter 5 of USACE (2002) manual.
o In order to maintain certain coherence in the measurements, the same total station and
reflector shall be used during the whole monitoring period.
o When using a total station mounted over a tripod, the horizontality of the system
should be checked few times during the day (the instrument suffer a certain drift out of
the level as the day progresses).

SafeLand - FP7
Page 71 of 401
Deliverable 4.1
o Line Of Sight (LOS) is the imaginary line from the sensor to the surveyed point.
Direct visibility is mandatory to perform distance measurements.
o Finally and needless to say, the stable control points (fixed stations) must be outside of
the expected moving area of the landslide. Otherwise, displacement measurement will
be underestimated due to a global drifting. A movement of a control point can
however be detected, when always the same instrument position is used.
2.2.2. 3D coordinate calculation
3D coordinates of the terrain are usually obtained through triangulation or trilateration
methods:
o A triangulation system consists of a series of overlapping triangles in which a given
side and two angles are measured. Using basic trigonometric formulas (Figure 23), it
is easy to calculate the length of the remaining sides and the angles of the remaining
vertex.
o Trilateration systems consist of a series of overlapping triangles in which the length of
all the sides and only few angles are measured to obtain azimuth.
Triangulation and trilateration techniques are used for accurate detection of terrain
deformation. A more detailed description of both methods is given in different surveying
books (e.g. Chandra, 2005).
Figure 23: Basic trigonometric formula for triangulation method.
2.3. MAIN COMMERCIAL SYSTEMS

2.3.1. Electronic Distance Meters
There are two types of EDM operating modes:
o Reflector-EDM, which requires a reflector device (e.g. prism) and hence, physical
contact with the surveyed point by a second operator. This operating mode allow for
the acquisition of points at very long distances (typically up to 5000 m).

SafeLand - FP7
Page 72 of 401

Rev. No: 2
Date: 2012-02-15
o Reflectorless-EDM, which does not require any reflector device, i.e., the instrumental
will acquire the coordinates of the natural features (rocks, soil, vegetation, etc).
Maximum distances are more reduced than using a reflector-EDM. Nevertheless, its
operability is higher: only one person is needed for the fieldwork campaign. This
operating mode allow for the acquisition of points at medium distances (typically up
to 1000 m).
2.3.2. Total stations
Total stations allow for the acquisition of:
o Horizontal angle,
o Vertical angle,
o Distance,
As a combination of both measurements, the XYZ coordinates of a point is obtained. These
coordinates are relative to the internal coordinate system of the total station. Total stations
allow for the electrical (instead of optical) acquisition of angles and distances. Digital
measurements eliminate the errors due to optical measurements. Total stations also
incorporate a collector device, which allow the downloading and processed of the recorded
datasets using an external PC.
There are three types of total station operating modes (USACE, 2007):
o Reflector-total station, which requires a reflector device (e.g. prism) and hence,
physical contact with a second operator with the surveyed point. This operating mode
allow for the acquisition of points at very long distances (typically up to 5000 m).
o Reflectorless-total station, which does not require any reflector device, i.e., the
instrumental will acquire the coordinates of the natural features (rocks, soil,
vegetation, etc). Maximum ranges are lower than using reflector-total stations.
Nevertheless, its operability is higher: only one person is needed for the fieldwork
campaign. This operating mode allow for the acquisition of points at medium
distances (typically up to 1000 m).
o Robotic-total station, which allow the automatic tracking of specific control points
previously defined. No operator is needed for the automatic acquisition of the control
points coordinates. The maximum range depends on the type of instrumental and on
the operating mode: reflector or reflectorless total station (up to 5000 and 1000 meters,
respectively).
2.4. OUTPUT
The measurements obtained with an EDM or Total Stations are distances and 3D coordinates.
2.4.1. Distances
Distances from a well known positioning point (e.g. concrete pillar) are usually acquired with
EDMs. As practical applications for landslides, distance calculation allow for the detection of
SafeLand - FP7
Page 73 of 401
Deliverable 4.1
the terrain displacements along time (see 1.5.2). Similarly, an automatic measurement of
distances along time allow for the monitoring of the displacements over single points of the
slope (see 1.5.3).
2.4.2. 3D coordinates
3D coordinates of the terrain surface (X, Y, Z) are the main output of surveying campaigns. A
set of 3D points is commonly used in conventional topography to generate maps (see 1.5.1).
2.5. MAIN APPLICATIONS
2.5.1. Digital Elevation Models and topographic maps
Digital Elevation Models (DEMs) and topographic maps are 2D representations of the 3D
geometry of the terrain. These representations are usually performed as follows:
a) Acquisition of 3D coordinates of the terrain surface. The higher the slope variability,
the higher the density of points. Similarly, the larger the scale of the map (e.g.
1:25.000, 1:10.000, 1:500, etc.), the higher the density of points.
b) Interpolation: creation of a Triangular Irregular Network (TIN) or a square grid in the
XY plane with the altitude Z as attribute for each TIN node resp. grid value.
c) Creation of contour lines or altitude-scaled colouring for the Z representation.
DEMs and topographic maps are the basis for further studies on landslide detection, fast
characterization and monitoring.
2.5.2. Landslide displacements
The landslide field of displacements can be measured either using point based (e.g. total
stations, DGPS, etc.) or area based techniques (photogrammetry, radar techniques, laser
scanning, etc.). An example of the former is depicted in Figure 24. Four essential aspects
must be borne in mind when studying landslide displacements (Malet et al., 2002): the degree
of accuracy required, the geomorphodynamic characteristics of the moving mass, the various
site constraints and, lastly, the resources available for the operation. For example, the
measurement of single points using EDM can be conducted using two different techniques:
o Using artificial reflectors: this technique is recommended when very accurate
measurements are required, e.g. for real time monitoring of millimetre or centimetre
level of accuracy.
o Using natural features: reflectorless measurements are recommended for very active
rockfall areas, in following cases: (a) when operators safety can not be assured and
(b) when reflector can be destroyed due to rock impacts. An example of this technique
is described in Miko et al. (2005).
The repetition of EDM measurements over fixed points allow for real-time monitoring and
the implementation of early warning systems (see next section).

SafeLand - FP7
Page 74 of 401

Rev. No: 2
Date: 2012-02-15
Figure 24: Sketch of the displacements measured at certain point of the landslide.
2.5.3. Real-time monitoring

Real-time monitoring of single points of the slope (natural features and/reflectors) can be
conducted using a robotic total station. Millimetric accuracies can be obtained using these
techniques. As an output of the system, displacements vs. time curve series are obtained
(Figure 25). The prediction of the date of failure can be inferred plotting inverse velocity
against time (Saito, 1969; Fukuzono, 1985). Some examples of rockfall prediction are
discussed in Crosta and Agliardi (2003).
Finally, it is important to mention that a high degree of autonomy and automation can be
obtained using EDMs: measurements can be transmitted via satellite to a control station,
which may be handily accessed via a standard internet connection.

SafeLand - FP7
Page 75 of 401
Deliverable 4.1
Figure 25: Displacements vs. time for three diferrent control points measured with a robotic total station. The
influence of the water content (intense rainfall) in the aceleration of the landslide can be observed in the
graphic.
3. TERRESTRIAL LASER SCANNER

The acronym LIDAR stands for LIght Detection and Ranging. When mounted over a groundbased platform, this instrument is also known as a Terrestrial Laser Scanner (TLS) or groundbased LIDAR. This new technology acquires dense 3D information of the terrain with high
accuracy, high resolution (mm to cm order) and very high data acquisition speeds. TLSs
appeared at the end of the 1990s and are currently used in a widespread variety of Earth
sciences applications, such as landslide characterization and monitoring. The number of
publications on this instrumental in landslide characterization and monitoring has increased
considerably in the last 5 years (e.g. Slob and Hack, 2004; Bauer et al., 2005; Rosser et al.,
2005; Lim et al., 2006; Jaboyedoff et al., 2007; Oppikofer et al, 2008; Sturzenegger and
Stead, 2009). Technical information on standards, specifications and quality control of
different TLS is provided in Hiremagalur et al. (2007). Finally, useful information in the use
of TLS for rock slope characterization is discussed in Kemeny and Turner (2008).
This instrument (Figure 26) consists of:
o a transmitter/receiver of infrared laser pulses for distance calculation (see section 1)
o a scanning device, formed by an internal system of rotating mirrors and/or a rotating
instrument.
In opposition with the EDM described in the previous chapter, the combination of laser
ranging and scanning devices allow for a very fast acquisition speed, typically thousands of
points per second. As a result, millions of points can be obtained in each data acquisition.
Range measurement can be undertaken using either the first or last pulse of the return signal.
An overview of laser ranging, profiling and scanning are already discussed by different
authors (e.g. Lichti and Jamtsho, 2006; Teza et al., 2007; Petrie and Toth, 2008). Therefore,
this section of the Deliverable will focus in basic principles of the TLS technique and method.
As regards practical applications for landslides, laser systems can be used for the acquisition
of large volumes of 3D coordinates of the terrain, with a high accuracy and at an extremely
fast recording rate.
Figure 26: TLS ILRIS-3D (Optech) at different locations.

SafeLand - FP7
Page 76 of 401

Rev. No: 2
Date: 2012-02-15
3.1.1. Accuracy
Technical characteristics supplied by the different manufacturers show point accuracy ranging
from mm to cm values, mainly depending on the range and the TLS model (i.e. ~0.7 cm at
100 m for the ILRIS 3D model). A review of the different commercial systems is discussed in
section 2.3. The accuracy of the measurement is a main function of the range, the reflectivity
of the material, the complexity of the scanned surface and the angle of incidence, as follows:
o Range: the higher the range, the lower the accuracy is (see technical specifications for
each instrument)
o Reflectivity of the material: the lower the reflectivity, the lower the accuracy is
(Voegtle et al., 2008)
o Surface complexity: the higher the complexity of the material, the lower the accuracy
is (Abelln et al., 2009).
o Incident angle: the higher the obliquity of the incidence angle, the lower the accuracy
is (Lichti, 2007, Ingensand et al., 2006).
For practical applications, the accuracy of a given TLS dataset can be estimated through the
comparison of two different point clouds acquired consecutively. As practical guidance, the
quality of the measurement can be improved by:
o Perform data acquisition as closest and perpendicular to the slope as possible.
o Filtering the low reflective points and the areas similarly oriented to the LOS and by
the deletion of erroneous points, i.e. vegetation, wires, birds, unexpected points, etc.
o For monitoring purposes, a good practice (although not mandatory) is to be consistent
with the location of each station during the different periods of time.
Finally, it has to be mentioned that error in comparison of the TLS datasets is not only
influenced by the instrumental error, but also by alignment and modelling errors (Teza et al.,
2007). The alignment error mainly depends on the alignment method and point density. The
modelling error is also influenced by the density of points, as was tested by Abelln et al.
(2009) through the comparison of two point clouds acquired consecutively. Figure 27 shows
the percentiles of the error as a function of the density of points in a basaltic cliff (Abelln et
al., 2009). As expected, the smaller the density of points, the lower the accuracy of the
comparison. Nevertheless, the values of the 75th, 90th and 95th percentiles show similar values
of the error for point spacing below 4.7 cm (e.g. 2.3cm, see Figure 27), meaning that no
significant improvements are obtained by decreasing the point spacing. As a result, a more
efficient density of point from a computational point of view can be obtained.

SafeLand - FP7
Page 77 of 401
Deliverable 4.1
Figure 27: Assessment of the quality of the rock face modelling: variance of the error for different densities of
points (extracted from Abelln, 2009).
3.1.2. Resolution
Laser scanners resolution is a parameter that determines the level of detail that can be
observed from a scanned point cloud. This parameter can be divided into range and angular
(or spatial) resolution:
o Range resolution is the ability of a rangefinder to resolve two objects on the same line
of sight and is governed by pulse length for a pulsed system (Kamerman, 1993).
o Angular resolution is the ability to resolve two objects on adjacent sight lines. This
parameter is governed by two parameters: sampling interval (also called point spacing,
which is user defined) and laser spot dimension (which is a fixed, but range depending
value, see section 1).
The density of points is user defined as an angular spacing between adjacent sight lines. In
practice, the technical operator should borne in mind that sampling interval is range
dependent: the areas located farther away from the TLS suffer for a low sampling interval
compared to the closer ones.
The estimation of the TLS resolution is usually misunderstood and commonly interpreted as
equal to the point spacing, obviating the influence of the laser beam width. This influence is
of great value when sampling interval is in the same order of magnitude or even lower than
the laser beamwidth. In these cases, fine details are effectively blurred, as discussed by Lichti
and Jamtsho (2006). In order to find a proper value for the estimation of the resolution value,
these authors proposed the used of the Effective Instantaneous Field of View (EIFOV). The

SafeLand - FP7
Page 78 of 401

Rev. No: 2
Date: 2012-02-15
optimal sampling interval proposed by Lichti and Jamtsho (2006) is 86% of the laser
beamwidth.
Table 4 provides a rough guide for the definition of the sampling interval (or point spacing)
during a fieldwork campaign using a commercial TLS (ILRIS 3D model). The parameters of
the laser beamwidth for this instrumental are as follows: = 0.00974; a= 12 mm (Optech,
2009).
Table 4: Optimal point spacing for the TLS ILRIS 3D model. This parameter was calculated as 86% of the spot
dimension (Lichti and Jamtsho, 2006).
Range (m)
50
100
150
200
250
300
350
400
450
500
550
600
Optimal sampling interval (cm)
1.8
2.5
3.2
4.0
4.7
5.4
6.1
6.9
7.6
8.3
9.1
9.8

Maximum range of the different TLS equipments depends on the type of instrumental (see
section 2.3). As guidance, the maximum range provided by the manufacturer for highly
reflective materials is almost double to the maximum range achievable in real slopes, which
are characterized by a 20% reflectivity. As an example, tests with a TLS ILRIS 3D at different
geological sites and obtained a maximum range of around 600-900 meters for dry surfaces.
3.1.4. Intensity
The intensity is defined as the amount of reflected signal with respect to the emitted one. Its
value is usually normalized on a scale ranging from 0 to 255. This parameter is usually
recorded in addition to the 3D coordinates.
Although this parameter may give an idea of the object material, automatic classifications
suffer from certain ambiguity: same value of the intensity may be recorded for different
materials at a different range. This is due to the fact that the intensity parameter mainly
depends on the range, angle of incidence, soil moisture, and object material.
3.1.5. Colour
In addition to the intensity of the returned signal, some TLS devices are able to obtain the
colour of the object. The colour (R, G, B) for each scanned point are captured with
independent detector. Alternatively, colour information can be recorded through an internal or
external digital camera (CMOS Sensor), with the result that each pixel of the photograph may
be attributed to each point of the scanned surface.
An example of spectral classification of different material based on colour attributes is
discussed in Lichti (2005).
3.2. METHOD
The laser beam is directly reflected on the land surface, obviating the need for intermediate
reflectors. The origin of the Cartesian Coordinate System (P0 =0,0,0) is usually set at the
centre of the TLS instrument. Coordinates of each point are acquired in a polar system (:
SafeLand - FP7
Page 79 of 401
Deliverable 4.1
range; : horizontal angle; : vertical angle). The binary file acquired by the TLS can be
transformed (parsed) into a Cartesian system (x, y, z) or other file formats used in commercial
software. As a result, a point cloud consisting of millions of x,y,z points can be obtained.
A thorough planning of the scanning campaign prior to fieldwork is highly recommended by
taking into consideration following characteristics:
o Point density: the number of points by a given area is user defined. Therefore, angular
resolution should be considered in advance, taking into account the necessary number
of points for a correct representation of the scanned slope.
o Location of the stations in order to minimize the shadow areas: when the laser beam is
parallel to the orientation of the geological surface, data acquisition may suffer from
occlusion problems (e.g. Lichti, 2007; Sturzenegger et al., 2007; Lato et al., 2009).
o Number of scans: the minimum number of scans necessary for a complete coverage of
the slope should be studied in advance, considering the Field of View (FoV) of the
different instruments (e.g. FoV of the ILRIS 3D model = 40x40)
o Overlapping areas: a minimum of 20 or 30% overlapping area is recommended for the
further coregistration of the point clouds acquired from the same station.
Data acquisition performed from the same point of view minimizes the geometry variations
during two epochs, making further comparison of TLS datasets more accurate. When
acquiring TLS datasets for monitoring purposes, it is recommended (but not strictly
necessary) to perform fieldwork campaign using constant station(s).
3.2.2. 3D coordinate calculation
Terrestrial Laser Scanners acquire the Coordinates of each point in a polar system (: range;
: horizontal angle; : vertical angle). These coordinates are relative to the internal coordinate
system of the total station or the laser scanner. The acquisition of the 3D coordinates in a
Cartesian system can be easily conducted using following equation (equation 3):
[x y z] t = [cos cos , cos sin , sin ] t
3.2.3. Filtering
The scattering of the TLS measurements around its true value should be taken into account
when dealing with TLS datasets. A previous manual cleaning of the TLS datasets is necessary
in most of the cases in order to improve the overall quality of the point cloud. The acquired
datasets that do not belong to the surface terrain (e.g. vegetation, wires, mobile objects, etc)
must be deleted before the processing of the point cloud.
Secondly, some automatic algorithms to filter the erroneous points may be applied, mainly
detecting which points belong to vegetation. This filtering is usually carried out either looking
to differences in geometry or to differences in Intensity. Although automatic filtering of trees
is a relatively simple task, the filtering of small plants and bushes is usually a manual, time
consuming task.

SafeLand - FP7
Page 80 of 401

Rev. No: 2
Date: 2012-02-15
In some cases, noise reduction tasks can be also carried out. For instance, filtering or
interpolation averaging can reduce the instrumental error (e.g. Bornaz and Rinaudo, 2004;
Lindenbergh and Pfeifer, 2005). In order to reduce the error in data comparison for small
scale deformation detection, a Nearest Neighbour (NN) averaging technique can be applied
(e.g. Abelln et al., 2009).
3.2.4. Registration of the point cloud
Registration is the process towards a set of point clouds (e.g. all the scans that describe an
object) is aligned into another system of coordinates. More specifically, following terms are
commonly employed:
o The term merging, alignment or unification is usually applied when the scans of
a single fieldwork campaign are referenced into a common file using an internal
coordinate system.
o The term georeferencing is usually employed when the scans are registered into a
well known, external coordinate system, e.g. UTM or national grids. Georeferencing is
usually performed using an external, already referenced point cloud (such as an
already georeferenced ALS point cloud) or using ground control points previously
acquired using a total station or a DGPS.
Registration of the point cloud is usually a affine transformation that consists in the
combination of a translation and a rotation. In 3D this affine transformation can be described
by a 4x4 matrix with 3x3 terms for the rotation, r11 to r33, 3 terms for the translation, t14 to t34,
and a scaling factor in the fourth line (equal to 1 for affine transformations without scaling)
(Stephens, 2000; Lichti et al., 2002; Oppikofer et al., 2009).
Registration of the point cloud usually consists in a two steps process:

o Step 1: A preliminary, rough registration by a visual identification of homologous
points (also referred to as N point pairs, e.g. edges of blocks, intersection of
discontinuities, reflecting targets, etc.) in the overlapping area of the different scans.
The higher the number (and accuracy) of the selected points, the better the alignment
is. Reasonably good results can be obtained using 3 to 5 homologous points equally
distributed over the 3D scene. In opposition, the use of collinear points may not give a
univocal solution and should be avoided.
o Step 2: an optimization of the alignment can be carried out using an Iterative Closest
Points (ICP) procedure (Besl and McKay, 1992; Chen and Medioni, 1992) once
reference systems of both scans resemble each other (see step 1). Through this
SafeLand - FP7
Page 81 of 401
Deliverable 4.1
algorithm, the differences between points are iteratively reduced by a minimization of

a mean square cost function. This iterative process can be improved by progressively
reducing the search distance parameter from few metres up to a few centimetres.
3.2.5. Comparison of sequential TLS datasets
3D temporal variations of the terrain can be detected by comparing sequential datasets in
accordance with the methodology described in Rosser et al. (2005) and Lim et al. (2006). A
brief description is as follows:
a) Acquisition of the reference point cloud (R0, see previous section)
b) Construction of the Surface of reference (S0)
c) Acquisition of data point clouds D1, D2,.. Dn
d) Alignment of these datasets with S0 (explained below)
e) Comparison between S0 and successive Di
f) Calculation of the differences for each period of comparison.
As stated above, the registration is usually carried out in two stages: (a) a preliminary
registration by a visual identification of homologous points and (b) optimization of the
alignment using ICP procedure. Nevertheless, a lack of convergence may be possible when the
geometries between two epochs are very different. In order to improve the alignment
procedure, the section of the point cloud that changed between the two epochs should be
automatically selected and ignored for the alignment. The single point distances between the
S0 and the successive Di can be computed using commercial software (e.g. data vs. reference
comparison in the ImInspect module of PolyWorks). The direction of comparison can be
defined as the shortest distance between different datasets or following an user-defined
direction. Differences (Difi) between two epochs can be calculated for each point as shown in
equation 3:
Difi = Distance [SiP0]- Distance[S0P0]
[equation 3]
Where P0 is the centre of origin of coordinates (0,0,0). Part of the value of Difi is due to
systematic (instrumental and methodological) errors. The other part is due to real changes in
certain parts of the slope. As regards the sign criteria, we use positive values when the time of
flight of the laser signal for Di is higher than that of S0. As a result, positive values correspond
to a lack of material at a given point, i.e. detachment of the material. Likewise, negative values
correspond to an increase in material, i.e. scree deposits, or a displacement towards the origin
of coordinates. This negative displacement may also reflect the pre-failure deformation in a
part of the slope (i.e. Abelln et al., 2010). The calculation of the volume of the main rockfalls
can be performed using commercial software (e.g. the surface to plane command of
IMInspect module, PolyWorks).

SafeLand - FP7
Page 82 of 401

Rev. No: 2
Date: 2012-02-15
3.3. MAIN COMMERCIAL SYSTEMS

Terrestrial Laser Scanners can be classified according to its measuring principle (phase vs.
pulse measurement) or to its maximum range (short, medium or long range). For practical
applications in landslides, a classification of the different commercial systems based in its
maximum range is commonly used. Practical applications of the short range TLS are limited
by its maximum range (below 100m). In opposition, medium (max. range from 100 to 350 m)
and long range (max. range up to 1200m) TLSs are commonly preferred in landslides
applications.
A description of the TLS devices usually employed in landslide investigations is provided is
subsequent sections. The main characteristics of each device are summarized in Table 5.
Table 5: Main characteristicts of the TLS most commonly used for landslide applications (summarized and
updated from Kemeny and Turner, 2008). The values of thes parameters are provided from the manufactures
and independent validation should be performed.
Manufacturer
Optech
Riegl
Trimble
I-Site
Mdl
Model
ILRIS-3D
LMS-Z620
GX 3D
4400 LR
Quarryman pro
700 m
750 m
350 m
300 m
400 m
7mm at 100m
10 mm at 100 m
7 mm at 100 m
20 mm at 50 m
50 mm at 50 m
2000 points / second
11000 points /
second
5000 points /
second
4400 points /
second
250 points /
second
0.00974
0.15 mrad
3mm at 50m
140 mm at 50 m
173 mm at 50 m
Maximum range in
natural slopes (20%
reflectivity)
Accuracy
Scanning rate
Beam divergence
angle
3.3.1. Optech
Model: ILRIS-3D (intelligent Laser Range Imaging System)
Maximum range:
o on artificial targets (90% reflectivity): 1500m
o on natural slopes (20% reflectivity): 700m
Accuracy: 7mm at 100m
Scanning rate: 2500 points / second
Beam divergence angle: 0.00974
Wavelength: 1550nm
Digital Camera (internal): 6.6 Megapixels

SafeLand - FP7
Page 83 of 401
Deliverable 4.1
Field of View (vertical x horizontal angle): 40 x 40

Optional components:
o ILRIS-36D: motorized pan and tilt base (increases Field of View: for 360x360)
o ILRIS-3D_ER: Enhanced range (increases the maximum range up to 40%)
o ILRIS-3D_MC: Motion compensation (allows its use in a moving platform, used in
combination with a GPS and Inertial System; lower accuracy)
o ILRIS-3D_HD: High density scanning (increases the laser repetition rate, allowing for
a faster data acquisition: 8000 points/second)
o External GPS.
o External Digital Camera
Software:
o Polyworks (Innovmetric): viewing, aligning, meshing and processing of the point
clouds. Feature extraction, 2D and 3D measurements, etc.
o Parser (Optech Inc): generation of the point cloud from the binary file produced by the
TLS
Figure 28: TLS ILRIS 3D (Picture extracted from: www.Optech.ca)
3.3.2. Riegl
Although 5 different models of TLS manufactured by Riegl are currently available, only the
medium and long range TLS will be summarized in this report. Given to the fact that
landslide applications of short range TLS (maximum range below 200 m) are limited, the
short range models were excluded from this review.
Models: LMS-Z420i; LMS-Z620 and LPM-321
SafeLand - FP7
Page 84 of 401

Rev. No: 2
Date: 2012-02-15
Maximum range:
o in artificial targets (80% reflectivity): 1000m (1), 2000m (2) and 6000m (3)
o in natural slopes (20% reflectivity): 350m (1), 750m (2) and 1500m (3)
Accuracy at 100m: 10mm (1) 10mm (2) and 25mm (3)
Scanning rate (points / second): 11000 (1), (2) and 1000 (3)
Beam divergence angle: 0.15mRad
Wavelength: near infrared
Digital Camera (internal): 10 Megapixels (1), (2), (3)
Field of View (vertical x horizontal angle): 80 x 360 (1), (2), 150 x 360 (3),
Optional components:
o External GPS
o External Digital Camera
o Internal synchronizer for GPS/INS
Software:
o Riscan Pro: viewing, aligning, meshing and processing of the point cloud. Feature
extraction, 2D and 3D measurements, etc.
Notes:
(1)
LMS-Z420i; (2) LMS-Z620; (3) LPM-321

SafeLand - FP7
Page 85 of 401
Deliverable 4.1
Figure 29: TLS LMS-Z620 (extracted from www.Rieglusa.com)
3.3.3. Trimble
Models: Trimble GX 3D
Maximum range:
o in artificial targets: unknown
o 20% reflectivity: 350m
Accuracy: 7mm at 100m
Scanning rate: 5000 points / second
Beam divergence angle: 3 mm at 50 m
Wavelength: 532 nm
Digital Camera (internal): Yes
Field of View (vertical x horizontal angle): 60x360 degrees
Figure 30: Trimble GX 3D (extracted from www.trimble.com)
3.4. OUTPUT
3.4.1. 3D point cloud
Laser ranging and scanning systems generate a huge quantity of information: the 3D
coordinates of million points of the terrain surface are obtained as a result of each data
acquisition. These TLS datasets are also referred to as point cloud or scans.
The simple visualization of the unprocessed point cloud may be a valuable tool for the
understanding of the 3D geometry of the scanned terrain.
SafeLand - FP7
Page 86 of 401

Rev. No: 2
Date: 2012-02-15
Figure 31: 3D point cloud of a rock slope. Discontinuity planes and vegetation are visible in this figure. The
gray shades represent the intensity of the returning signal with low intensity (dark gray) on vegetated areas and
high intensity (white) on rock surfaces perpendicular to the line-of-sight.
3.4.2. High resolution DEM

High Resolution Digital Elevation Models (HRDEMs) are accurate representations of the 3D
geometry of the terrain performed with a great quantity of 3D information, e.g. obtained
through a TLS. HRDEMs are usually performed as follows: (a) acquisition of 3D coordinates
of the terrain surface; (b) Interpolation: creation of a Triangular Irregular Network (TIN) and
(c) generation of a square mesh.
HRDEMs are the basis for further studies such as landslide mapping, detection, modelling,
fast characterization and monitoring.
During last five years, the use of TLS on different natural and artificial slopes is considerably
improving our understanding of landslide phenomena. Different applications are undergoing
rapid development, mainly in the following aspects: use of high-resolution, high-accuracy
DEM for 3D mapping and/or modelling; characterization of discontinuities; landslide
displacement detection; rockfall detection and prediction.
3.5.1. Mapping
Landslide mapping (delimitation of the moving areas, geometry of the slide, mapping of
landslide features such as scars, etc) can be directly carried out using the point cloud or in a

SafeLand - FP7
Page 87 of 401
Deliverable 4.1
LIDAR derived HRDEM. Given the high accuracy and resolution of the datasets, landslide
mapping can be carried out at an unprecedented detail.
3.5.2. Modelling
HRDEM are the basis for further treatments such as:
-
Kinematic modelling: rockfall trajectories, the modelling of debris-flow paths, etc.
Static modelling: safety factor, limit equilibrium, etc.
An example of rockfall and debris flow runoff calculation using a LIDAR derived HRDEM is
provided in the Chapter 4 of this deliverable (case studies). As discussed in Abelln et al.
(2006), a more accurate information on the detachment location, joint orientation, volume of
detached blocks and characterization of rockfall trajectories and energies are obtained using
TLS derived HRDEM.
Figure 32: Debris flow modelling (Horton et al., 2008).
3.5.3. Structural analysis

Structural analysis of the main families of discontinuities can be derived either from a DEM
or directly from the TLS point cloud. Different techniques for feature extraction can be
applied, either manually (best fitting plane of a selected section of the point cloud) or
automatically (e.g. calculation of the orientation of the TINs, calculation of the orientation of
square meshes, use of the covariance matrix, etc.).
Two commercial softwares are currently focused in the calculation of the discontinuities
orientation based on TLS datasets: Coltop 3D (Jaboyedoff et al., 2007) and Split Fx (Kemeny
et al., 2006). As an example of the former, the slope aspect and the slope angle can be
visualized in a unique colour in fig 16. Using this colour-coded information, the orientation of
the discontinuities can then be obtained in a univocal way and plotted in stereographic
projection.

SafeLand - FP7
Page 88 of 401

Rev. No: 2
Date: 2012-02-15
Figure 33: Colour-coded representation of the slope aspect and slope angle. Application to a real rockslope
(modified from Coltop 3D.com and Jaboyedoff et al., 2007.
3.5.4. Landslide displacements

Multi-temporal comparison of TLS datasets allow for accurate computation of landslide
displacements. This deformation can be computed using different methods:
o Calculating displacement vectors at different parts of the slope by identifying identical
points on the data and the reference point clouds.
o Point cloud comparisons: computation of the distance between two adjacent points in
the data and the reference point clouds (e.g. shortest distance or following a
comparison vector) using a point-to-point or point-to-surface ICP algorithm.
o Rigid body transformation: computing the roto-translation matrices between the data
and the reference point clouds associated to the displacement of certain parts of the
slope (Montserrat and Crosetto, 2008; Oppikofer et al., 2009).

SafeLand - FP7
Page 89 of 401
Deliverable 4.1
Figure 34: landslide precursory displacements at Val Canaria, Ticino, Swiss Alps. Figure extracted from
Pedrazzini et al., 2010.
3.5.5. Rockfall detection and prediction

TLS sequential datasets can be compared to detect 3D temporal variations of the terrain. Using this
method, it is possible to detect sudden changes in the morphology of the slope, allowing the
characterization of the morphology, magnitude, frequency and location of the detachment area of the
rockfalls occurred during the monitoring period (e.g. Rosser et al., 2005; Lim et al., 2006; Abelln et al.,
2009). Similarly, this technique is able for the detection of rockfall precursory indicators such as
minor rockfall events leading to grater failures (Rosser et al., 2007) and small-scale deformations
(Abelln et al., 2009; Abelln et al., 2010).

SafeLand - FP7
Page 90 of 401

Rev. No: 2
Date: 2012-02-15
Figure 35: (a) Basaltic cliff at Castellfollit de la Roca. The frame corresponds to Fig. 18b (b) Comparison of
TLS models showing a 50 m3 rockfall event. Colour scale from 0 (blue) to 1.5 m (red). Figure extracted from
Abelln et al., 2009.
An example on the application of TLS for rockfall detection and prediction is provided in the
chapter 4 of this deliverable (see case studies section).
4. AERIAL LASER SCANNER

Laser scanner instruments mounted on an airborne-based platform are known as AerialLIDAR or Aerial Laser Scanner (ALS). This instrument allows for the acquisition of the 3D
coordinates of millions of points belonging to the terrain surface with decimetre-level
accuracy. This instrumental is widely used for topographical purposes, such as the generation
of High Resolution Digital Elevation Models (HRDEM).
The use of HRDEM for landslide hazard mapping at regional scale started to be used in the
beginning of this century. The widely extended use of the Geographic Information Systems
(GIS) during the 1990ies allowed for the fast incorporation of such a detailed maps into real
landslide applications.
High resolution datasets obtained from ALS has allowed the extension of local scale
techniques for landslide detection, mapping, monitoring and characterization over wider
areas, e.g. regional scale. Nowadays, ALS derived HRDEM are widely used in the scientific
community, providing a better understanding of landslide detection, mapping,
characterization and monitoring.

SafeLand - FP7
Page 91 of 401
Deliverable 4.1

ALS is a technological evolution of the Aerial Laser Profiles, which consists basically in a
laser rangefinder mounted on an airborne platform. Early airborne laser profiles instrumentals
appeared in the middle of the 60s decade.
ALS basically consists of three components (Figure 36):
a) A laser scanner (see section 3)
b) A positioning system (differential GPS)
c) An Inertial Measurement Unit (IMU).
In addition, an imaging device (e.g. video camera) usually forms part of the ALS system.
Figure 36: main components of an ALS (extracted from Petrie and Toth, 2009)
4.1.1. Accuracy
Although there is a lack of commonly accepted standards on ALS accuracy determination and
calibration procedures (Habib and Van Rens, 2008), accuracy of ALS systems usually range
from 10 to 30 cm. This variation of the values is due to the great number of parameters that
influence in the final accuracy (Baltsavias, 1999). Indeed, this parameter is a combination of
the accuracy of the (a) laser rangefinder, (b) scanning device, (c) Inertial System (IMU), (d)
GPS and (e) time offsets. Furthermore, the accuracy depends on external parameters, such as
(a) range (i.e. flying height), (b) position of the laser beam, (c) direction of the laser beam, (d)
slope angle (the error of the ALS datasets is higher in steep slopes than in flat surfaces) and
(e) existence of vegetation.
It is important to mention that the accuracy specifications stated by manufacturers and service
providers should be checked by the final user. This verification can be carried out through: (a)
the comparison of the statistical error in overlapping areas of the point cloud (overlapping
strips) or (b) the comparison of the point cloud and ground control points. Guidelines for the
estimation of the vertical accuracy (ASPRS LIDAR Committee, 2004) states that the 95th
percentile method of the error shall be employed to determine accuracy instead of the
commonly used Root Mean Square Error (RMSE) or standard deviation (1 sigma). Accuracy
requirements for ALS datasets depends on the final purpose, i.e. monitoring requires better
accuracy than regional mapping.
SafeLand - FP7
Page 92 of 401

Rev. No: 2
Date: 2012-02-15
4.1.2. Resolution
The resolution of a point cloud is usually refered to as the size of the smallest feature that can
be measured. It is usually expressed as the number of points per area (e.g. points/m2),
although laser beamwidth is also affecting this parameter (see section 3.1.2 of Active Optical
Sensors for a complete explanation).
ALS systems allow for the acquisition of a high density of information (2-4 points per m2)
over wide areas. Nevertheless, this parameter is controlled by the flying height and the ratio
between flying velocity and scanning speed. Basic formulas for the calculation of this
parameter are described in Shan and Toth, 2009 (Table 6). ALS resolution may reach values
of 1 point per 5 or 10 m2).
Table 6: Basic formulas for the calculation of the resolution (extracted from Shan and Toth, 2009).

The maximum range of an ALS system is mainly limited to the maximum range of the laser
range finder. Normal flying heights range from 1 to 3 km, although higher flying heights are
possible (up to 6 km). Relationships between maximum range (Rmax), Swath width (SW) and
flying height (H) is provided in Table 7 (Shan and Toth, 2009).
Table 7: Basic formulas that correlate maximum range (Rmax), Swath width (SW) and flying height (H).
(extracted from Shan and Toth, 2009).
4.1.4. Intensity
The intensity is defined as the amount of reflected signal with respect to the emitted one. Its
value is usually normalized on a scale ranging from 0 to 255. This parameter is usually
recorded in addition to the 3D coordinates.
SafeLand - FP7
Page 93 of 401
Deliverable 4.1
4.2. METHOD
Figure 37 provides an overview of the processing steps, from campaign planning to the
delivery of a final product, typically a HRDEM. The customer usually defines the following
parameters: extension of the surveying area, density of points and survey accuracy.

SafeLand - FP7
Page 94 of 401

Rev. No: 2
Date: 2012-02-15
Figure 37: Overview of the processing steps, from campaign planning to the delivery of a HRDEM (extracted
from Shan and Toth, 2009)

SafeLand - FP7
Page 95 of 401
Deliverable 4.1

The laser beam is directly reflected on a portion of the land surface equal to the laser
beamwidth at a given height. Coordinates of each point are usually acquired in a polar system
(: range; : horizontal angle; : vertical angle) and has to be synchronized with the Inertial
system and the variable position obtained through the GPS. As a result, a point cloud
consisting of millions of x,y,z points is obtained.
Given that groups dealing with ALS datasets for landslide studies are usually final users, only
a few comments on data acquisition procedures will be given. A more detailed explanation
is discussed in different papers and books (e.g. Baltsavias, 1999; Heritage and Large, 2008;
Shan and Toth, 2009; Vosselman and Maas, 2010). The scanning campaign should be defined
by taking into consideration the required point density, accuracy and scanned area. These
parameters are a main function of the height of flight.
4.2.2. Filtering of erroneous points
Before the generation of the DEM, the acquired points that do not belong to the point cloud
(e.g. vegetation, buildings, wires, etc.) should be deleted (Figure 38).
Figure 38: Classification of the point cloud into points belonging to the terrain surface and points belonging to
vegetation.
Different semi-automatic procedures for the detection of these points can be used, based on
geometrical aspects (terrain points are more continuous than the points belonging to
vegetation), in the intensity of the returned signal (the intensity of the laser signal after
reflecting in the vegetation is lower than after reflecting in the terrain surface) or in the
detection of multiple returns (first pulses are usually related with vegetation cover, last pulse
usually corresponds to the terrain surface, see Figure 39). Although these semi-automatic
filters may help in the deletion of non terrain points, a visual inspection and a final manual
filtering is necessary. Dense vegetation cover can significantly decrease the density of ground
points.

SafeLand - FP7
Page 96 of 401

Rev. No: 2
Date: 2012-02-15
Figure 39: Detection of multiple returns from ALS: (a) Schematic illustration showing seven laser pulses over a
vegetated area; (b) profile (transect) obtained from the point cloud data using different returns of the laser
pulse. Extracted from Harding (2009). Graphics courtesy of Bob McGaughey, USDA Forest Service PNW
Research Station.)
4.2.3. GPS/IMU integration and registration

GPS provides position and velocity of the aircraft, while the IMU system provides attitude
and inclination of the system with respect to the horizon. The integration of these two
components is a key step in the processing of the datasets: the final accuracy is quite sensitive
to this processing step.
Registration of a set of point clouds (e.g. all the scans that describe the terrain surface) is
the transformation from a local reference system into a real world coordinate system, e.g.
UTM. Georeferencing is usually performed using the DGPS and calibrated using ground
control points.
4.3. OUTPUT
4.3.1. ALS point cloud
The 3D coordinates of million points of the terrain surface are obtained after each flight. The
simple visualization of the unprocessed point cloud may be a valuable tool for the
understanding of the 3D geometry of the scanned terrain, e.g. landslide features (scarps,
topographic roughness and waviness, etc). The inspection of the ALS point cloud has the
advantage over HRDEMs that it allows to retrieve ground information where the actual data
points are. Furthermore, visual inspection avoids interpolation artefacts from the DEM
creation.
4.3.2. High resolution DEM (HRDEM)
After data acquisition and filtering of erroneous points (e.g. wires, vegetation, birds, etc), a
high resolution DEM is usually obtained. This product is obtained by rasterizing and
interpolation the ALS point cloud. Cell size is user defined, but at least a few points per cell

SafeLand - FP7
Page 97 of 401
Deliverable 4.1
are recommended. Assuming homogeneous point coverage over the study area, the minimum
recommended HRDEM cell size, d, is given by Equation 4:
[Equation 4]
where A is the surface covered by the ALS point cloud and n is the total number of ground
points in the dataset.
HRDEM are high accuracy representations of the topographical surface obtained with a great
density of information. The visualization of the HRDEMs can be highly improved by using:
o High resolution orthophotos: colour images obtained from an aerial sensor can be
wrapped to the HRDM, allowing for a realistic visualization of the slope
o Hillshade maps: View-shaded representation of the topography taking into account the
slope orientation (slope aspect and slope angle). Hillshade maps are generated using
the 3D information provided by the HRDEM and a user-defined light source (in
general: sun elevation angle of 45; sun azimuth of 315). An example of hillshade
map is shown in Figure 40.
Figure 40: HRDEM and picture of the west face of the Grand Muveran Mountain (Switzerland). Picture from
Jaboyedoff et al. (2007). Source of the ALS dataset: MNT MO/MNS.

Main applications of ALS derived HRDEM are focused on detection, mapping,
characterization, monitoring and modelling of mass movements such as landslides, debris
flows, rockfalls and deep-seated gravitational slope deformations (DSGSD).

SafeLand - FP7
Page 98 of 401

Rev. No: 2
Date: 2012-02-15
Table 8: Main applications of ALS for landslide studies.
Mapping
Deep seated
Landslide
Shallow
landslide
Rockfall
Debris flow
DSGSD
Characterization
Monitoring
Modelling
-
-
Symbols: : feasible application using ALS-: non feasible application using ALS
4.4.1. Mapping
Slope movements have distinct morphologic features (i.e. the head scarp, the main sliding
body and the deposits) that can be recognized not only in the field, but also on remote sensing
data, such as ALS derived DEMs (Chigira et al., 2004; McKean & Roering, 2004; Derron et
al., 2005; Carter et al., 2007; Schulz, 2007; Dewitte et al., 2008). Landslide mapping can be
directly carried out on the hillshade of the HRDEM using classical mapping techniques based
on expert criteria. Alternatively, semi-automatic techniques may give an insight into the
delimitation of the moving area (see McKean and Roering, 2004). A compilation (from
Oppikofer, 2009) is shown as follows (table 6):
Table 9 :Different applications of ALS derived HRDEM for landslide mapping (modified from Oppikofer, 2009).
Chigira et al. (2004)
Shallow landslides detection in vegetated areas. These movements were not detected by
using aerial photographs or field survey.
McKean and Roering

(2004)
ALS derived HREM to automatically delimit a landslide in New Zealand using the surface
roughness of the topography
Derron et al. (2005a;

2005b)
Large rockslides investigation in western Norway (knes and Oppstadhornet rockslides,

respectively) by means of ALS derived DEM: delimitation of the main rockslide features
(scarps, lateral faults, fracture orientations, scree slopes, etc).
Van Den Eeckhaut

et al. (2005)
Use of LIDAR derived hillshade maps and expert knowledge in mapping deep-seated
landslides. This work compares landslide inventories using HRDEM compared to
conventional mapping (fieldwork campaigns).
Schulz (2007)
ALS derived HRDEM to map 173 shallow and deep-seated landslides in the coastal cliffs
near Seattle (Washington, USA).
4.4.2. Characterization
Landslide characterization (i.e. landslide magnitudes, conditioning factors such as slope
angle, discontinuity orientations, etc.) can be carried out either on the raw ALS point cloud or
on a HRDEM. To this end, the use of GIS tools is widely used. Some examples are discussed
as follows:
o The steepness of the slope is one of the most important conditioning factors for slope
stability and can be easily derived from a HRDEM (e.g. Rouiller et al., 1998;
SafeLand - FP7
Page 99 of 401
Deliverable 4.1
Baillifard et al., 2003; Guzzetti et al., 2003; Jaboyedoff & Labiouse, 2003; Baillifard
et al., 2004; Frattini et al., 2008; Loye et al., 2009). As a result, landslide prone areas
can be detected and used to produce regional susceptibility maps.
o As discussed by many authors (e.g. Hoek and Bray, 1981; Bieniwsky, 1993; Romana,
1993; Hoek, 2000), discontinuity orientation and its relationship with the geometry of
the slope is a key factor in the stability of a slope. The characterization of the
discontinuity orientation using an ALS point cloud in combination with simple
kinematic feasibility tests allows for the accurate detection of rockfall prone areas.
Some examples are discussed by: Gokceoglu et al., (2000); Meentemeyer & Moody
(2000); Gnther (2003), Jaboyedoff et al. (2004), Jaboyedoff et al., (2007), Lan et al.,
2010.
4.4.3. Landslide monitoring
ALS time series datasets allow for the comparison of the topographic relief at different
epochs. Due to its wide coverage and high accuracy, ALS is a valuable tool for the monitoring
of unstable areas, either for the detection and volume calculation of already fallen areas as for
the detection of movements of decimetre level.
Three methods are discussed in Travelletti et al. (2008) for the monitoring of LIDAR derived
DEMs: (a) the benchmark method, consisting in the computation of local displacements and
rotations at selected fixed placements (e.g. blocks, tree trunks, etc.); (b) the point cloud
comparison (data to reference, see section 3.2.5), either using a shortest distance method or
computing the differences through a given vector and (c) the correlation of shaded relief
images, based on automatic identification of common features between successive images
(see Delacourt et al., 2004).
As an example, Dewitte et al. (2008) shows a DEM comparison (extracted from
photogrammetry and ALS) to monitor over a 50 years time period a deep-seated rotational
landslide located in the Ardennes (Belgium). Finally, Figure 41 computes the vertical
differences between the pre and postfailure surfaces in a landslide in Val Canaria,
Switzerland (extracted from Pedrazzini et al., 2010).

SafeLand - FP7
Page 100 of 401

Rev. No: 2
Date: 2012-02-15
Figure 41: Comparison of pre and post failure surfaces using aerial and terrestrial laser scanning datasets.
Source area and scree deposits are clearly visible in the figure. Image from Pedrazzini et al., 2010 (with
permission from the author).
4.4.4. Modelling
The modelling of a landslide (i.e. run-out distances of a debris flow, energy and trajectories
calculation, etc.) can be greatly improved using an accurate 3D representation of the terrain.
Needless to say, each mass movement process requires different modelling approaches, as
follows:
a) Several authors (e.g. Hoek, 2000; Agliardi and Crosta 2003; Gianin, 2004) have
pointed out how the slope is one of most important parameters in rockfall simulation.
Currently, different commercial and non commercial software allow for the numerical
calculation of rockfall trajectories, energies, maximum run-out, etc. (e.g. RocFall,
RockFall analyst, CRISP, Conefall, Rotomap, Rockyfor3D, etc.). A complete example
of high resolution 3D modelling of rockfalls is discussed in Agliardi and Crosta
(2003). Complementarily, rockfall run-out area can be modelled using the
generalization of the shadow angle theory in a GIS environment (Dorren and
Seijmonsbergen, 2003; Jaboyedoff et al., 2008).
b) Debris flow modelling is usually based on two types of algorithms (Horton et al.,
2008): flow direction and run-out distance calculation. The reliability of the results
generally increases with the accuracy and the resolution of the HRDEM. An example
of debris flow modelling is shown in Figure 42.

SafeLand - FP7
Page 101 of 401
Deliverable 4.1
Figure 42: Illustration of the model Flow-R, with sources detection, runout energy and probability. Figure from
Jaboyedoff et al. (2010).

SafeLand - FP7
Page 102 of 401

Part C Active Microwave Sensors
Rev. No: 2
Date: 2012-02-15
PART C
ACTIVE MICROWAVE SENSORS
SECTION 1: Radar Wave Principles
p. 105
SECTION 2: Interferometric Radar Distance-Meter
p. 107
SECTION 3: Differential SAR Interferometry
p. 125
SECTION 4: Multi-Temporal Repeat-Pass InSAR
p. 131
SECTION 5: Ground-Based InSAR
p. 151
SECTION 6: Polarimetry for Soil Moisture Estimation
p. 158
Sections 1, 2 and 3
- Authors: C. Michoud1, M.-H. Derron1, A. Abelln1, M. Jaboyedoff1
- Reviewer: G. Fornaro2
1
. Institute of Geomatics and Risk Analysis (IGAR). Universit de Lausanne, Switzerland. clement.michoud@unil.ch
. AMRA Scarl, Naples, Italia.
Michoud C., Derron M.-H. Abelln A., Jaboyedoff M., Fornaro G., 2010. Radar Wave Principles, Interferometric Radar Distance-Meter and
Differential SAR Interferometry. In the Deliverable 4.1 of the European project SAFELAND: Review of Techniques for Landslide
Detection, Fast Characterization, Rapid Mapping and Long-Term Monitoring. Edited in 2010 by Michoud C., Abelln A., Derron
M.-H. and Jaboyedoff M. Available at http://www.safeland-fp7.eu
Section 4
- Authors: V. Tofani1, G. Fornaro2, F. Cigna1, F. Catani1
- Reviewer: C. Michoud3, M.-H. Derron3
1
. Earth Sciences Department, Universit degli Studi di Firenze, Italia. veronica.tofani@unifi.it

. AMRA Scarl, Naples, Italia.
3
. Institute of Geomatics and Risk Analysis (IGAR). Universit de Lausanne, Switzerland.
2
Tofani V., Fornaro G., Cigna F., Catani F., 2010. Multi-Temporal Repeat-Pass InSAR. In the Deliverable 4.1 of the European project
SAFELAND: Review of Techniques for Landslide Detection, Fast Characterization, Rapid Mapping and Long-Term Monitoring.
Edited in 2010 by Michoud C., Abelln A., Derron M.-H. and Jaboyedoff M. Available at http://www.safeland-fp7.eu

SafeLand - FP7
Page 103 of 401
Deliverable 4.1
Section 5
- Authors: V. Tofani1, J. Gili2, G. Luzi1*, F. Catani1
- Reviewer: M.-H. Derron3
1
. Earth Sciences Department, Universit degli Studi di Firenze, Italia. veronica.tofani@unifi.it

. Moved after his contribution.
. Department of Geotechnical Engineering and Geosciences, Universitat Politcnica de Catalunya, Barcelona, Spain.
3
1*
2
Tofani V., Gili J., Luzi G., Catani F., 2010. Ground-Based InSAR. In the Deliverable 4.1 of the European project SAFELAND: Review of
Section 6
- Author: G. Fornaro1
- Reviewer: C. Michoud2
1
. AMRA Scarl, Naples, Italia. fornaro.g@irea.cnr.it

Fornaro G., 2010. Polarimetry for Soil Moisture Estimation. In the Deliverable 4.1 of the European project SAFELAND: Review of

SafeLand - FP7
Page 104 of 401

Rev. No: 2
Date: 2012-02-15
1. RADAR WAVE PRINCIPLES

1.1. INTRODUCTION
Radar technologies have been developed since the beginning of the 20th century, first to
remotely detect objects. The acronym Radar states for Radio Detection And Ranging. Radiowaves, or microwaves, are electromagnetic waves of wavelength from 1 mm to 1 m. Then
radar methods have been used for accurate detection of movements and Earth surface
imaging. The principles Radar technologies and applications are described in many
publications, books and encyclopedias. About the radar imaging part used in environmental
and Earth sciences, Massonnet and Feigl (1998), Hanssen (2001) and Woodhouse (2006)
provide worthy introductions and overviews.
There are two kinds of Radar systems (Hanssen, 2001): bistatic radars use two different
antennas, one for the emission and one for the reception of the electromagnetic wave; On the
other hand, monostatic radars use only one antenna for both tasks. This last configuration is
the most frequently applied technology for environmental purposes. A Radar transmitting
antenna emits a wave by pulses with a predefined wavelength (Table 10) amplitude and
phase. The Radar receiving antenna estimates the range (with the two-way travel time) and
measures the amplitude and the phase of the backscattered pulses. The way the radar waves
travel and are reflected to the antenna depend on several factors such as: pressure, temperature
and partial water pressure of the environment, atmospheric layering, roughness and dielectric
properties of the target (strongly influenced by the moisture and salt content) (Hanssen, 2001;
Norland, 2006). In remote sensing, environmentalists usually describe a radar wave by its
wavelength, because it gives some information about how the waves interact with the surface
(roughness, interferences). Signal processing specialists prefer to work with frequencies as
most of the computations are done in the frequency domain (convolution, filtering).
Table 10 : Main bands traditionally used by the Standard Radar Nomenclature (Woodhouse, 2006).
Band
Wavelength
Frequency
Decimetric ( 30 cm to ~ 1m )
0.3 1 GHz
Decimetric (~ 15 to 30 cm)
1 2 GHz
Decimetric (7.5 cm to ~ 15 )
2 4 GHz
Centimetric (~ 3.75 to 7.5 cm)
4 8 GHz
Centimetric (~ 2.5 to 3.75 cm)
8 12 GHz
Compared to the optical and near infrared sensors, the radio waves are much less influenced
and attenuated by the atmospheric conditions. As shown in the Figure 43, the attenuation of
waves used by Radar systems is less than 0.1 decibel per kilometer (after Preissner, 1978). So
data can be acquired even during heavy rainfalls (even though some significant interactions
are reported for X-Band systems) or strong fog; however these are not the optimal conditions.

SafeLand - FP7
Page 105 of 401
Deliverable 4.1
Moreover, as a radar is an active sensor (sending the initial signal), data can be acquired as
well during the night than during the day.
Figure 43 : attenuation of an electromagnetic wave by the atmosphere, depending of the wavelength for a clear
weather, with shown for fog, heavy rain and drizzle. Contrary to the optical and near infrared, radio waves are
only few attenuated by the atmospheric conditions (After Preissner 1978, cited in Peckham 1991 and Woodhouse
2006).
The choice of the frequencies used by the Radar sensors is regulated by the International
Telecommunications Union. Usually, the space-borne sensors are transmitting radio waves in
the bands X, C and L (Table 10). The longer the wavelength is, the worse the accuracy is both
in terms of spatial resolution and measurement precision, but the better is the penetration
capacity and the stability of the returned signal with respect to environmental changes..
1.2. RANGE AND PHASE SHIFT
Mathematically, the phase shift corresponds to:
2
R where R 2R
4 R

SafeLand - FP7
Page 106 of 401

Rev. No: 2
Date: 2012-02-15
In this formula, is the phase shift in radians, the wavelength in meter and R the distance in
meter from the sensor to the ground target.
Figure 44 : representation of the Radar wave alond the line of sight. The wave with a wavelength of is doing
the 2-way of 2R from the sensor S to the ground target. As the sensor emitted the wave knowing the original
phase, it can record the phase shift when the wave comes back.
So we can deduce in theory the distance R knowing the phase shift:

R
2. INTERFEROMETRIC RADAR DISTANCE-METER

One direct application of the phase shift principle is the continuous monitoring of instabilities
assessing the movements of punctual ground targets by Radar distance-meters. All the
information including in this chapter comes from the case study of the Tafjorden instability in
Norway (Norland, 2006).
As described before, the phase shift is function of the wavelength and the range:

4
R
But practically, R is defined as
Co velocity of light in vacuum
Co t 2way
1
R
where t2way travel time sensor target return
2
n
n spatial and temporal index of refraction

Indeed, even if the Radar amplitudes are only few influenced by atmospherically conditions
(the attenuation of the signal is negligible in almost all used bandwidths), the phases are very
sensitive (Figure 45 and chapter 3.3.3.2). In fact, the spatial and temporal index of refraction
depends of the pressure, the temperature and of the partial water vapor pressure which can be
estimated using the relative humidity of the atmosphere (Norland, 2006).

SafeLand - FP7
Page 107 of 401
Deliverable 4.1
p pressure in mbar
p
e
5
n 77.6 3.734.10 2 where T temperature in K
T
e partial water vapor pressure in mbar
T
Figure 45 : strong correlations between the spatial and temporal index of refraction n (left) and the variation of
range measured (right) (Norland, 2006; modified).
As shown in the figures above, there is a strong correlation between atmospherically

conditions and displacement results. To separate the signal influenced by the atmosphere and
the signal influenced by ground displacements, it is useful to introduce a second target on a
stable part (Figure 46).
Figure 46 : theoretical arrangement for interferometric distance-meter (Norland, 2006).
So, assuming one reflector as fix, all variations of phase coming from the spatial and temporal
index of refraction and not from the displacement can be filtered. Then, comparing phase
results between the two reflectors by interferometry (phases differences), it is possible to
assess the displacements of the second target (Norland, 2006).
2 1
R
4
( R 2 R 1)
So, in the case study of the Tafjorden instability, the reflectors are monitored with accuracy
better than 0.1 mm at 2.9 km (Norland, 2006).
3. DIFFERENTIAL SAR INTERFEROMETRY

The main use of Differential Synthetic Aperture Radar Interferometry (DInSAR) is the
detection of small deformations or movements at the surface, analyzing the phase differences
SafeLand - FP7
Page 108 of 401

Rev. No: 2
Date: 2012-02-15
between two scenes acquired at different times. The Differential Synthetic Aperture Radar
Interferometry is a combination of three main processing steps: first the Synthetic Aperture
Radar (SAR), then the Interferometry (InSAR), and finally the Differential InSAR (DInSAR).
These steps are described hereafter.
3.1. SYNTHETIC APERTURE RADAR
3.1.1. Radar imagery
In Radar imagery, a coherent and polarized wave is emitted by an antenna and the amplitude
and the phase shift of each backscattered pulse are recorded in a complex number
(Woodhouse, 2006):
i.
z A.e
z value for each cells

where A amplitude
phase shift
The amplitude of the backscattered signal is a function of the wave initial state, the
environment crossed by the wave and the capacity of the target (reflectance, moisture) to
reflect the wave (Figure 47). During the post-processing, it is easy to distinguish the emitted
signal from the backscattered one. Moreover, airborne and satellite sensors are not looking
vertically downwards but they are side-looking, avoiding the problem of knowing from which
side of the nadir the pulse is backscattered.
Figure 47 : Complex SAR image with the return amplitudes (left) and phases (right) of Bam area in Iran,
December 2003 (Michoud et al., 2009).
In most cases, space borne Radar sensors are monostatic. So, the power of the return wave
can be determined by the following formula detailed in the ESA Radar course III (website
address in references):

SafeLand - FP7
Page 109 of 401
Deliverable 4.1
Pr Power received (W)

Pt Power emitted (W )
Pr Pt
G 2.2.
(4)3.R4
where
G Gain of the antenna

Radar wavelength (m)
radar scattering cross section (m2)
R distance antenna target (m)
But the effective area of the antenna is related to the gain by:
S
2.G
Pr Pt
S 2.
2
4. .R
The formula above gives the mean power per pixel in function of the size of the antenna. The
bigger the antenna is, the higher the mean received power will be. And the higher the mean
received power is, the better is the resolution. So one practically limit of the system is the
maximal possible size of the antenna in space. It was demonstrated that to get a metric
resolution from space borne sensor, a kilometer long antenna is needed (Prati, 2008).
To overcome this technical impossibility and obtain a reasonable ground resolution, a new
acquisition method has been developed since the 50s (Wiley, 1954): the Synthetic Aperture
Radar (SAR).
3.1.2. Synthetic Aperture Radar

Imagined in 1951 by Carl Wiley (Wiley, 1954; Hanssen, 2001), the Synthetic Aperture Radar
principle wants to sum several echoes of the same target during the post-processing to
improve the azimuthal resolution. Indeed, when the sensor transmits pulses along its track, the
same ground target can reflect several times different echoes (Figure 48). Using the Doppler
beam sharpening with the frequency information contained in the phase (Woodhouse, 2006),
it is possible to differentiate two scatterers within the beam. So, the effective antenna aperture
moving along the flight line is larger than in reality. In this way, the azimuthal resolution is
improved (Figure 49). This process of synthesis can be compared to the multi-tracks
processes in seismic geophysics.
Figure 48 : Synthetic Aperture Radar principles. The same ground scatterer is shot from different points of view
synthesizing an aperture much larger than the physical size of the real antenna, increasing the azimuth
resolution (Lauknes, 2004).

SafeLand - FP7
Page 110 of 401

Rev. No: 2
Date: 2012-02-15
Figure 49 : Real Aperture Radar (left) and Synthetic Aperture Radar (right). The azimuth resolution is much
better with SAR technology (Prati, 2008).
Once the azimuthal resolution has been improved by the SAR acquisition, it is also important
to increase the range resolution. As explained in Woodhouse (2006), the shorter the pulse is,
the better the range resolution is; but generating short pulses requires high peak power,
actually not available on a satellite. In practice, a linear frequency modulated or chirped pulse
(Figure 50) is used (Hanssen, 2001; Lauknes, 2004; Woodhouse, 2006). Thanks to a
frequency post processing using the Fourier Transform, the echoes can be separated and reattributed to their original ground target. Indeed, this technology allows to emit a chirped
pulse of 40 s with a range resolution similar with a pulse of 60 ns (Woodhouse, 2006).
However, to perform this technique, it has to be assumed that the fly is stable (reasonable in a
satellite) and that the frequency variation law (modulation) is well known.
Figure 50 : linear chirp waveform amplitude function of time (arbitrary units). Typical pulse durations for
ERS and Envisat ASAR are between 30 and 40 micro seconds. (Lauknes, 2004).
3.1.3. Terminology and Space Resolution

The spaceborne acquisition of SAR images often uses specific terms which have to be defined
or remembered in this section (to go further, see the website of the Canadian Center for
Remote-Sensing).
Ascending / Descending orbit: South-North / North-South trajectory of the platform.

SafeLand - FP7
Page 111 of 401
Deliverable 4.1
Azimuth: distance along the spaceborne flight direction. The azimuthal resolution can
be performed by Synthetic Aperture Radar techniques.
Complex signal: each pixel is registered by a complex number, keeping the

information on the amplitude and the phase of the return signal. This is the raw data
for an InSAR study.
Ephemerides: satellite position and velocity vectors at any time t.
Fringe: line of an interferogram corresponding of one color cycle.
Ground range: distance along the perpendicular of the flight direction on the grounds
surface. The range resolution can be performed emitting chirped pulses.
Line of Sight: trajectory of the chirp between the radar antenna and the ground
scatterers.
Look angle : angle between the vertical and the LOS.
Master / Slave images: reference / comparative images in interferometry.
Nadir: projection on the Earths surface of the platforms trajectory.
Near / Far range: part of the image closest to / furthest from the Nadir.
Resolution: size of the smallest object detectable by the sensor.
Slant range: distance along the perpendicular of the azimuth on the SAR image. The
slant range can be different of the ground range due to geometrical distortion, as
explained in the following chapter.
Track / Frame: Est-West / North-South limit of a SAR image.

SafeLand - FP7
Page 112 of 401

Rev. No: 2
Date: 2012-02-15
Figure 51 : geometry of SAR acquisition (Prati, 2008, modified)
3.1.4. Geometric and radiometric distortions

The satellite does not emit pulses vertically; there is always a look angle generally higher of
10 (depending of the sensor). Two types of geometric distortion (Hannsen, 2001; Squarzoni,
2003; Lauknes, 2004; Woodhouse, 2006) can be produced in rough relief areas: the
foreshortening and the layover. A third possible effect, the shadowing, is a radiometric
distortion (Figure 52).
Figure 52 : from ground range to slant range. (Hanssen, 2001, modified)

SafeLand - FP7
Page 113 of 401
Deliverable 4.1
Foreshortening may happen when a slope, with its normal vector looking towards the sensor,
is less steep than the look angle. In this case, the surface will be smaller in the slant geometry
than in reality (e.g. Figure 53). The layover may happen when the slope is steeper than the
look angle. In this case, the backscattered pulse from the top of the slope will be collected by
the sensor before the pulse reflected from the bottom of the slope (Figure 9). Then the top of
the slope will be considered as closer in the slant range than the bottom, and the slope will
look as reversed. The closer the slope is from the Nadir, the stronger the geometric distortions
will be (Woodhouse, 2006). Indeed, the look angle is smaller close to the near range than to
the far range and so more distortions are typically observed in the near range.
To correct both geometric distortions, foreshortening and layover, information on the real
topography are used. By geocoding SAR images using a DEM, it is possible to rectify these
distortions, as schematized in the Figure 54.
Figure 53 : schematic differences between ground range (left) and slant range (right), with a foreshortening
slope (Prati, 2008).
The last distortion, shadowing, is a radiometric distortion that corresponds to an absence of

signal (Figure 52). This distortion is the most restricting because the lost information is not
recoverable. The only way to obtain additional data is to shoot from different positions, e.g.
once in ascending orbit and the second time in descending orbit.
The last and facultative step of all processing is the geocoding, to transform and uniform the
pixel sizes in a ground-coordinate system (Lauknes, 2004). For this process, Ground Control
Points, DEM and precise orbit data are needed (ESA, Figure 54).

SafeLand - FP7
Page 114 of 401

Rev. No: 2
Date: 2012-02-15
Figure 54 : Geocoding of SAR image (ESA document, 2008, modified)
3.2. SAR INTERFEROMETRY (INSAR)

Interferometric Synthetic Aperture Radar (InSAR) is a method to produce images of the
interferometric phases of an area using images acquired from two different points of view.
With SAR images as in optical stereoscopy, two images of an object from two distinct points
of view allow seeing in relief and building Digital Elevation Models.
The interferograms shown as examples in the following chapters are processed using free
ESA Envisat ASAR dataset of the Bam region (southeast Iran), where a strong earthquake
occurred the 26th of December 2003.
Table 11 : raw input of Bam interferograms (free raw Envisat ASAR dataset provided by ESA).
Image
Sensor
Date
Normal Baseline
Sar1 (Slave)
ASAR Envisat
11th of June 2003
475 m
Sar2
(Master)
3rd of December 2003
0m
Sar3 (Slave)
7th of January 2004
520 m
3.2.1. Preparation of the dataset

3.2.1.1.Focusing
SAR is radar: similarly to surveillance radars, it therefore allows discriminating, i.e.,
resolving the targets in distance by measuring the delays of the transmitted pulses: the

SafeLand - FP7
Page 115 of 401
Deliverable 4.1
narrower the pulse the higher the system range resolution. As already pointed out in a
previous section, in order to limit the peak power and therefore to simplify the transmitter,
usually chirp signal are used in transmission. These classes of pulses are characterized by
large bandwidth duration products: to achieve short pulses the echoes that form the radar
image must be correlated with the transmitted replica of the chirp signal. The received signal,
i.e. the raw data, is therefore subject to a processing (range focusing) to achieve the maximum
range resolution. Furthermore, SAR exploits the movement of the sensor to synthesize an
antenna which is much larger, typically between 100 and 1000 times, the size of the real
antenna mounted on-board the satellite (about 10m for the ERS and Envisat cases).
Accordingly, the raw data must be further processed (azimuth focusing) along the azimuth
direction to synthesize the large antenna and hence to achieve the highest azimuth resolution.
Depending on the frequency and resolution, the azimuth focusing is generally more complex
than the range focusing: it is intrinsically two-dimensional and space variant. However, thank
to the research carried out in the last years, efficient focusing algorithms that make use of Fast
Fourier Transform codes have been developed. In Bamler (1992) different focusing
algorithms are compared.
It should be pointed out that to trade off between the coverage (swath width) and the
resolution, SAR sensors may operate in different modes, see Fig 13. In the classical mode,
referred to as stripmap mode, the antenna pointing is fixed: The movement of the sensor
allows collecting the information from a ground strip. In the spotlight mode, the beam is
steered backward in such a way to illuminate only a fixed portion (the ground spot): the size
of the synthesized antenna and therefore the resolution is increased at the expense of the
coverage. On the other hand by steering or better scanning the beam over different range
swaths, SAR system may increase the range coverage at the expense of the azimuth
resolution. Depending on the operational mode, the focusing operation must take into account
the characteristics of the acquisition in order to properly process the signal and hence to
achieve a full focusing. Most of the commercial packages allow focusing data for any
acquisition mode. However, in case of beam steering the acquisition characteristics can
impact also the subsequent interferometric processing steps: for instance spotlight and scansar
interferometric acquisitions must be synchronized along the azimuth in such a way to limit the
variability of the response of the scene as a function of the azimuth angle (aspect angle).
Figure 55 : different operational modes of SAR systems

SafeLand - FP7
Page 116 of 401

Rev. No: 2
Date: 2012-02-15
3.2.1.2.Co-registration
Image registration is a process that aims to obtain a precise overlap between two or more
images relative to the same object. Such images may be acquired from separated positions, by
different sensors and/or at different times. For SAR interferometry, images need to be
precisely aligned, with sub-pixel accuracy, to extract the phase interference relative to the
same ground target, which is imaged at different positions in the two images due to distortion
effects related to the different imaging geometry associated a) with a non zero orbit spatial
separation, b) with the presence of orbit crossing. Incorrect alignments of the focused images,
usually known as Single Look Complex (SLC) images, cause coherence losses in the
interference operation, which, in turn, translate into inaccuracies in the final product.
Co-registration procedures for SAR images are based on the estimation of the so called
azimuth and range warp-functions, i.e., the functions describing the transformation that maps
the slave image onto the reference master image and that are subsequently used for
resampling the slave image onto the reference master image grid. Warp functions are
generally modeled in such a way to account for a few distortion effects: bilinear expansions
are frequently used to account for rotation and scaling effects due to acquisition track
crossing. Accordingly, warping functions are described in terms of few parameters that are
estimated starting from control points. In the case of radar images, for the accuracy required
in coherent data processing, such control points cannot be generally identified starting from
features (road crossing, edges, etc.), because of the speckle noise. Ground corner reflectors
could be exploited; however in practice matching measures of small image patches are
employed to evaluate a sufficient number of tie points for the parameters estimate.
Standard image co-registration procedures based on polynomial expansion of the warp
functions have been extensively used in the past to implement interferometry at small spatial
baseline separation. However, accurate image registration of SAR data takes on a sub-pixel
basis, where the passes span an orthogonal baseline interval of the order of few kilometers
and the temporal intervals of several years (as it happen in interferometric stacking and
persistent scatterers Interferometry for ground deformation monitoring) becomes a rather
challenging step due to effects of topography and the presence of large temporal decorrelation
which impair the accurate measurements of local distortions. Even more challenging is the
registration of images for the latest high resolution SAR systems (i.e. TerraSAR-X and
COSMO/SKYMED). To overcome the limitations of standard techniques, modern coregistration techniques make use of a geometrical computation of the warp functions starting
from the orbital information and from an external (rough) DEM, (Sansosti 2006). This
procedure allows easily achieving very accurate registration almost in all cases of interest for
interferometrical application,
3.2.1.3.Filtering
In differential SAR Interferometry, angular imaging diversity is an unavoidable effect due to
the impossibility to repeat exactly the same orbit: it also plays a positive role for the
localization of the gorund scatterers: In any case it is the source of noise that adds to the
interferograms Besides the generation of image geometric distortions, which are usually
compensated via a registration step, the interferometric phase associated with the topography
plays a modulation role that causes, for a distributed scattering, the presence of incoherent
components between the two images. For a planar surface this effect is known as spectral shift
effect and it is shown schematically in the left part of Figure 14. The result is the introduction
SafeLand - FP7
Page 117 of 401
Deliverable 4.1
of decorrelation noise on the interferometric fringes which is usually referred to as

geometrical, spatial, or baseline decorrelation. Spatial decorrelation can be counteracted by
proper filtered of the co-registered SLC images aimed at zeroing the unwanted components:
the filtering is known as Common Band filtering and is schematically shown in the right part
of Figure 14. Angular decorrelation is also present in the azimuth direction due to possible
differences in the antenna beam pointing in the two acquisitions (Doppler centroid
decorrelation). Common band filtering can be therefore beneficial also along the azimuth
direction in the presence of large excursion of the Doppler Centroid. While in the azimuth
direction the common band filtering is space invariant, the range filtering should be dependent
on the local topography. In Fornaro and Monti Guarnieri 2002 it is discussed an efficient
procedure for range space variant filtering. It should be pointed out that common band
filtering should be performed only when the scattering is assumed to be distributed: this
assumption is not verified in the case of strong scatterers and hence it is generally not
implemented in permanent scatterers monitoring: Conversely it is usually performed in
approaches like the low resolution monitoring via the Small BAseline Susbet approach.
scene spectrum
on board filter
correlated contributions
common band filters
Figure 56 : the spectral shift effect (left) and the common band filtering (right)
3.2.2. Interferometric phase

Mathematically, the radar interferometry is defined by the difference of 2 phase shifts
acquired by the sensor in two distinct times. So the formula 1 is the base of this method. More
mathematical developments, which use the variables introduced in the Figure 57, are available
on the following publications: Bamler & Hartl (1998), Massonnet and Feigl (1998), Hanssen
(2001), Lauknes (2004), Catani et al. (2005), Colesanti & Wasowski (2006), Woodhouse
(2006), Ferretti et al. (2007).

SafeLand - FP7
Page 118 of 401

Rev. No: 2
Date: 2012-02-15
Figure 57 : InSAR acquisition parameters. S1 and S2 are the two antennas acquiring the image SAR 1 and SAR
2. S1 flight at the altitude H. P is a ground scatterer at the altitude h. R1 and R2 are the distances between P and
S1 or S2. B is the distance between the two sensors (baseline), Bp and Bn being resp. the parallel and normal of
B to R1. The angle between the vertical and R1 (resp. R2) is 1 (resp. 2). (Catani et al., 2005, modified)
Mathematical approach for one reflector acquired twice:
1 2

4
4
4
R1
R2
R R 2)
( 1
Due to the big distance considered, the approximation Bp R1 R2 can be admitted.

4
Bp
4 Bn tan (1 ) 4 B sin (1 )
Published with these two different forms, the formula 2 expresses the theoretical geometric
relations of a raw interferogram (Figure 58).

SafeLand - FP7
Page 119 of 401
Deliverable 4.1
Figure 58 : interferogram using the images Sar2 and Sar3. One fringe is represented by one color cycle (from
blue to yellow). (Michoud et al., 2009; processed with Erdas Imagine 9.2)
3.2.3. Interferogram flattening

The topography is not the only contribution in an interferogram; there is also the flattening
phase, which can be defined as the phase variation proportional to the slant range
displacement of the point target (Ferretti et al., 2007). Removing this orbital contribution by
an operation called interferogram flattening, it results fringes only influenced by the
topography (and possibly the ground displacement) and proportional to it (Ferretti et al.,
2007).
Figure 59 : theoretical interferogram simulating a pyramid. Left: raw interferogram Center: flattening phase
Right: fringes proportional to the topography. (Woodhouse, 2006)

SafeLand - FP7
Page 120 of 401

Rev. No: 2
Date: 2012-02-15
3.2.4. Altitude of ambiguity

The variation of the topography which creates an interferometric cycle of 2 is called altitude
of ambiguity ha (Massonnet and Feigl, 1998). In an interferogram, it corresponds to the
altitude constituted by one complete fringe (from blue to red in the Figure 58).
n 2 , n IN
H.tan 1
H.tan 1
ha
2
and for ERS ha 9400
Bn
Bn
4
2
Bn
Considering the case of a normal baseline of 200 m, the altitude of ambiguity for ERS is
47 m. A shown later, the shorter the normal baseline is, the bigger the altitude of ambiguity is
and the less accurate the altitude measurement is (Figure 64). But regarding to the geometrical
loss of coherency, the normal baselines cannot be indefinitely long (cf.). To generate DEMs
from ERS-satellite images, there is an optimum of the normal baseline at about 300-400 m
(Ferretti et al., 1997).
3.2.5. Phase unwrapping

The fact that the interferometric phase is measured modulo 2 creates an ambiguity that must
be removed by a course of phase unwrapping. This step converts each ambiguous 2 cycle to
the absolute value of the interferometric phase (Figure 60 and Figure 61).
As suggested in the Figure 60, a certain continuity of phase values recorded is required to
unwrap correctly the interferogram. Actually, several answers are acceptable for one initial
wrapped interferogram (Ferretti et al., 2007). Actually, if the variation of the topography
between two following scatterers is equal to the altitude of ambiguity (unwrapped phase 2 in
the Figure 60), this gap is not considered in this processing step and the result will be without
gap (unwrapped phase 1 in the Figure 60). Ferretti et al. (2007) emphasizes on the necessity to
use a-priori information to get the right solution.
The other key point to assure a continuity of the values is to use only pixels which have a
good coherence (cf. chapter 3.3.3), reducing the incoherent noise. Bernardino et al. (2003)
advise to take only the pixels whose coherence is greater than 0.3.

SafeLand - FP7
Page 121 of 401
Deliverable 4.1
Interferometric Phase (rad)
25
20
15
Unwrapped Phase 2
Unwrapped Phase 1
Wrapped Phase
10
Number of altitudes of ambiguity

Figure 60 : difficulties of phase unwrapping. In fact, it exists several ways to unwrap correctly an interferogram.
(Massonnet and Feigl, 1998, modified)
Figure 61 : unwrapped interferogram [Sar2|Sar3] of Bam area. (Michoud et al., 2009; processed with Erdas
Imagine 9.2)
3.2.6. DEM construction

The last step of the processing aims to construct numerically the DEM. The Figure 62 shows
the helpful parameters needed to perform the model. The platform flying always at the same
altitude, is nil (cf. Figure 57). Moreover, the approximation [S1L] = [S1P] = R1 is admitted.

SafeLand - FP7
Page 122 of 401

Rev. No: 2
Date: 2012-02-15
Figure 62 : parameters needed to calculate the altitude variations. (Michoud et al., 2009)
cos () RH
1
(H h)
cos ( ) R1
R1 cosH
( )
h H R1 cos ( )
Now, cos ( ) cos cos sin sin
and for small,
cos
sin
h H R1(cos sin )
h H H (cos sin )
cos
cos
sin
h H 1

cos
cos
h H tan
Now, deriving in the formula 6 according to , we obtain

As L and P are close, therefore we admit small and

d 4
B cos
d
B 4 cos
h H
tan
B 4 cos

SafeLand - FP7
Page 123 of 401
Deliverable 4.1
H.tan

Bn
4
Therefore, it is possible to obtain from the interferometric phase of two close scatterers the
variation of altitude. Performing this method for all the pixels in the SAR images, it is
possible to get a relative DEM of the area (Figure 63). Due to the initial ambiguity of 2, the
model cannot characterize the absolute altitudes, but only the variations of the topography. If
the exact altitude of one pixel is known, it is then possible to shift the entire area to achieve a
good DEM. To get a good DEM accuracy and to limit noise artifacts, it is advisable to process
several SAR pairs and to average the results in order to decrease the signal to noise ratio
(Ferretti et al., 2007).
Figure 63 : Topography of BAM area, provided by the processing of the interferogram [Sar2|Sar3]. The model
has a 20 m grid and a vertical accuracy of 1.5 m. (Michoud et al., 2009; processed with Surfer 8)
The Shuttle Radar Topography Mission (SRTM), achieved in February 2000 by the NASA,
scanned the Earth surface with two Radar antennas (C-band and X-band). Thanks to SRTM, it
exists now a worldwide DEM at 90 m resolution (from 60N to 56S) and an American DEM
at 30 m resolution (SRTM technical factsheet, JPL 2005).
For the data from the satellites ERS, the threshold of detection of a coherent signal regarding
to the noise is /6 (Catani et al., 2005). The following formulas show the importance of the
normal baseline in the vertical accuracy (h) of a DEM (Figure 64). The longer the normal
baseline is, the more accurate the DEM is.
As h H.tan and
Bn
6
4
H.tan(1)
h
and for ERS, h 800
24
Bn
Bn

SafeLand - FP7
Page 124 of 401

Rev. No: 2
Date: 2012-02-15
Vertical accuracy of the DEM (m) .
100
80
60
40
20
0
0
50
100
150
200
250
300
350
400
450
500
Normal Baseline (m)

Figure 64 : vertical accuracy of a DEM regarding to the normal baseline of the interferometric pair. The
function is calculated with the ERS parameters: = 5.6 cm, H = 785 km and 1 = 23. The longer the normal
baseline is, the more accurate the DEM is.
3.3. DIFFERENTIAL SAR INTERFEROMETRY (DINSAR)

The main goal of differential InSAR (DInSAR) technique is to identify and quantify small
ground deformations. Indeed, the phase difference is produced, provided by two
contributions: the topography and the deformations (Figure 65). More information is available
on Massonnet and Feigl (1998), Lauknes (2004), Catani et al. (2005), Colesanti and
Wasowski (2006), Woodhouse (2006), Ferretti et al. (2007).
way = topography + deformation
Figure 65 : Influences of a slope deformation on the phase decay. The way of the slave SAR image is longer than
the way of the master SAR image, due to the rotational landslide (Michoud et al., 2009, modified).

SafeLand - FP7
Page 125 of 401
Deliverable 4.1
3.3.1. Topographic phase

In DInSAR, the topographic phase is considered as an artifact. Indeed, the relief contribution
scrambles the signal created by the ground deformations. The base of DInSAR techniques
consists in using a ante-deformation DEM to perform a theoretical interferogram synthesizing
the phase contribution of the topography. In this way, it is then possible to remove from the 2way travel phase this topographical contribution to extract the deformation phase (Figure 66).
deformation = way - topography
There are two methods to estimate the topographic contributions (Figure 66):
Using an interferometric pair with the shortest temporal baseline as possible (e.g. from
ERS Tandem). It is assumed that there is no ground deformation between the two
acquisitions. Then the 2-way travel phase is equal to the topographic phase. But, this
solution is not the best way to process due to recommendations made in the chapter
3.2.6 and for the unavoidable effects related to the propagation of the radiation
through the atmosphere which typically varies between the two passes.
Using an external DEM produced by another method (ALS, topographic maps, etc.).
This way to process is the most powerful, because it uses one product independent of
the SAR images. It allows to prevent errors due to bad SAR raw dataset.
Figure 66 : creation of a Differential SAR Interferogram. The DEM injected provides a theoretical
interferogram only influenced by the relief. The topographical phase is removed of the 2-way travel phase to
isolate the deformation phase. (ESA, 2008, modified)
3.3.2. Deformation phase

Let consider the processing to be perfect and that all the contributions, the deformation phase
excepted, are removed. The movements detected (Figure 68) by the formula shown below is
in reality a projection of the deformations along the LOS (Figure 67) and not the 3D vector of
displacement. The first publications on landslide monitoring with DInSAR techniques were
published by Fruneau et al. (1996) and Carnec et al. (1996).

SafeLand - FP7
Page 126 of 401

Rev. No: 2
Date: 2012-02-15
Figure 67 : projection of the displacements on the LOS. The red vector dlos is the projection of the real
displacement (orange vector d) on the LOS. (Prati, 2008, modified)
Therefore, displacements perpendicular to the LOS are not detectable. In the following
development, d represents the deformation and dlos is the projection on the real displacement
on the LOS.
Rs Rm d
4
def
d
l os
def
dlo s
4
Figure 68 : unwrapped DInSAR [Sar2|Sar3] of Bam. (Michoud et al., 2009; processed with GSAR)
Finally, the accuracy of the detection depends on the Radar wavelength used and the
threshold of detection of /6 (chapter 3.2.6).

SafeLand - FP7
Page 127 of 401
Deliverable 4.1
As d , dlos d
6
4
4 6 24
Admitting a perfect processing, the accuracy expected can be subcentimetric; ERS and
Envisat, which use C-band (5.6 cm), can reach an accuracy of 2.33 mm.
Unfortunately, all expressions and formulas shown above are available only for ideal
conditions, and it is never the case on the Earth. Indeed, the SAR signal can be noised by
atmospheric disturbances, orbital inaccuracy, etc. So it is very important to know all the
contributions which can influence the radar signal, to process the data and interpret properly
the results.
3.3.3. Limitations of conventional DInSAR

Several noise sources can influence the phase of the SAR signal. It is important to know all of
theme to be able to isolate only the interesting phases (Massonnet and Feigl, 1998; Lauknes,
2004; Hanssen, 2005; Colesanti and Wasowski, 2006; Ferretti et al., 2007).
total = decorrelation + atmospheric + orbital + way
This section aims to explain those contributions.
3.3.3.1.Geometric and temporal decorrelation

Both temporal and spatial decorrelations contribute to a loss of coherence (2.3.4) between
different SAR acquisitions.
The temporal decorrelation is due to changes of geometrical and electrical properties of the
scatterer surface. Indeed, the evolutions of vegetation, morphology (erosion, rockfall, etc.),
land use (farmed crops, fallow field, etc.) and soil moisture influence a lot the reflected signal.
Now, those factors can change in time according to seasons and human activities. Therefore
Massonnet and Feigl (1998) advise to use SAR images acquired always at the same period of
the year to limit the temporal decorrelation. Usually, the best seasons are beginning of spring
and end of fall, when trees have only few leafs and without snow covering the ground.
A too big difference of point of view (too long spatial baseline) between two SAR
acquisitions creates the geometrical decorrelation. Bamler and Hartl (1998) indicate the
existence of a critical baseline above which the interferometric signal is only composed by
noise; this critical baseline is estimated at almost 1000 m for ERS dataset. It can be calculated
thanks to the following formula (sarmap, 2008):
Bn critical R tan
2 Rs
where is the wavelength, R the range, the look-angle and Rs the pixel spacing
(respectively 5.6 cm, 870 km, 23 and ~12.5 m for ERS).
3.3.3.2.Atmospheric artifacts
Even if the Radar wave amplitudes are not much attenuated by the atmosphere (Figure 1), the
phase is sufficiently influenced by atmospheric conditions to perturb results. Atmospheric
SafeLand - FP7
Page 128 of 401

Rev. No: 2
Date: 2012-02-15
conditions will change the velocity of propagation. Actually the variation in time and space of
the partial water pressure in the troposphere is the main controlling factor (Zebker et al.,
1997; Norland, 2006). Moreover, the variations of pressure and temperature can influence
slightly the signal (Norland, 2006). Massonnet and Feigl (1998) advise to use SAR images
acquired during night to limit the atmospheric artifact. As the atmosphere is colder and dryer
than during the day, the Radar signal is less perturbed.
3.3.3.3.Orbital artifacts
It is very important to know the relative position of the platforms during the acquisitions.
Indeed, as exposed in the formulas shown above, the values of the normal baselines are
required by the interferometric equations. To limit the orbital artifacts due to bad positioning,
precise orbits datafiles, called ephemerides (e.g. Delft or Doris precise orbits) are imported
during the processing (Lauknes, 2004; Ferretti et al., 2007).
3.3.4. Coherence
To have an objective idea of the accuracy and the final reliability of an interferogram, the
coherence index was created. Massonnet and Feigl (1998) define the coherence as the degree
of agreement between neighboring pixels in an interferogram. The bigger the coherence is,
the lower the artifacts are and the better the processing will be.
The coherence of an interferogram (Figure 69) processed with the complex SAR images S1
and S2 is calculated in this way:
E.S1. S2
] [
E. S1 .E. S2
where E represents the statistical expectation and S 2 the conjugated complex of S2. A
coherence of 0 means that results are only composed by random phase noise and a coherence
of 1 by very good signal with no noise (Prati, 2008). Colesanti and Wasowski (2006) explain
that in practic, the coherence is computed assuming ergodicity for the interferometric signal
and, therefore, estimating E[#] using all pixels within a (e.g. rectangular) window centred in
the image element.
Hanssen (2005) advises to consider a-priori coherence expected in an area before ordering
dataset to assess the feasibility and reliability of processed interferograms. Indeed, according
to practical observations (Massonnet and Feigl, 1998; Berardino et al., 2003; Colesanti et al.,
2004; Lauknes, 2004; Michoud et al., 2009), the coherence is better in arid or urban areas than
in wooded counties where temporal decorrelations are stronger.
The following figure aims to compile all notes written before to optimize the coherence and
the reliability of differential interferometric results.

SafeLand - FP7
Page 129 of 401
Deliverable 4.1
Figure 69 : coherences of the interferogram [Sar2|Sar3] of Bam. (Michoud et al., 2009; processed with Erdas
Imagine 9.2)
Figure 70 : optimization of the DInSAR processing. The best way to process dataset with good coherences and
reliability is to choose night scenes from the beginning of spring or end of fall (snow and leafs free) with the
shortest baselines as possible. (Michoud et al., 2009, modified)

SafeLand - FP7
Page 130 of 401

Rev. No: 2
Date: 2012-02-15
4. MULTI-TEMPORAL REPEAT-PASS INSAR (ADVANCED DINSAR)

4.1. INTRODUCTION
A remarkable improvement in the quality of DInSAR results for ground deformation mapping
and monitoring is given by the Advanced DInSAR techniques (A-DInSAR). These techniques
make use of large multi-temporal stacks of space-borne SAR images acquired over the same
area, in order to identify radar targets on which is possible to detect and measure
displacements along the LOS (Line Of Sight) direction, over time.
The basis of the A-DInSAR techniques, as well as the standard DInSAR, is the separation of
different components from the interferometric phase, INTERF: the topographic component,
TOPO, the movement component, MOV, the atmospheric contribution, APS and the noise
component, NOISE:
INTERF = TOPO + MOV + APS + NOISE
A-DInSAR techniques overcome most of the limitations of the standard single-interferogram
approach (DInSAR), such as temporal and geometric decorrelation and the atmospheric phase
delay and, unlike conventional Interferometry, they allow increasing the measurement
accuracy from centimetre up to millimetre. These techniques also overcome the standard
approach both in terms of deformation modeling capabilities and quality of measures
(Crosetto et al., 2005b) :
Temporal deformation modelling: A-DInSAR are able to provide a whole

description of the temporal behaviour of a deformation field, whereas standard
DInSAR techniques temporally sample a given deformation phenomenon with only
two acquisitions (Figure 71).
Quality of measures: A-DInSAR techniques allow the estimation of different error

sources that may affect the interferometric observations and the separation of
movement signals from the other phase components (e.g. topographic errors,
atmospheric component), achieving a good precision and reliability of deformation
measures. This component analysis cant be performed using the standard DInSAR
approach.
The output of an A-DInSAR analysis consists in a set of radar targets, for which the following
information can generally be extracted (Figure 72):
geographic coordinates (E, N) and heights
mean deformation rates (precision can be up to 0.1 mm/year; it depends mainly on the
distance from the reference point)
time series of displacements (precision can be up to 1 mm on a single measurement)
quality index (e.g. coherence)

SafeLand - FP7
Page 131 of 401
Deliverable 4.1
Figure 71 : Temporal sampling of a deformation phenomenon performed with the DInSAR and A-DInSAR
techniques (Crosetto et al., 2005b).
Figure 72 : Example of results from an Advanced DInSAR analysis: mean deformation rates measured by means
of multi-interferometric analysis of ERS 1995-2000 data (Crosetto et al., 2005b).
In general, at least 15-20 radar images acquired along the same nominal orbit (i.e. same
acquisition geometry), are necessary to perform a multi-interferometric analysis.
Nevertheless, the actual number depends on the characteristic of the area of interest and the
type of SAR sensor to be used.

SafeLand - FP7
Page 132 of 401

Rev. No: 2
Date: 2012-02-15
As mentioned above, only the pixels characterized by a low level of phase noise, are exploited
to derive the deformation estimations, even if SAR sensors perform a regular 2D sampling of
the terrain. Consequently, the ability of A-DInSAR to fully describe a deformation
phenomenon depends temporally on number of available images, and spatially on the
availability of good pixels, i.e. pixels which are characterized by a low level of phase noise.
According to the type of analysis and the number of radar data available, A-DInSAR allows
getting a radar target density even exceeding 1,000 points/km2.
In order to guarantee the measurement of ground deformations in areas with low coherence, it
is also possible to install artificial corner reflectors or active transponders (Allievi et al., 2004)
and monitor building, infrastructures and non coherent areas affected by different types of
deformation phenomena (e.g. landslides, building collapse) (Crosetto et al., 2005b).
In the last years, several A-DInSAR techniques have been developed and experimented on
multi-temporal interferometric data, such as the Permanent Scatterers (PS-InSAR) technique,
the Interferometric Point Target Analysis (IPTA), the Persistent Scatterers Pairs - Differential
SAR Interferometry (PSP-DIFSAR), the Stable Point Network (SPN), the Small BAseline
Subset (SBAS), the Coherent Pixel Technique (CPT) and the Stanford Method for Persistent
Scatterers (StaMPS). These techniques are different from each other in the interferometric
processing chain, the pixel selection criterion (amplitude- or coherence-based), the processing
approaches used to separate the different phase components (e.g. atmospheric phase) and the
deformational model of the targets (linear, non-linear). Their modelling and processing
strategies are strictly dependent on the type of application at hand (Berardino et al., 2002;
Costantini et al., 2000; Duro et al., 2003; Ferretti et al., 2000; Hooper et al., 2004; Mora et al.,
2003; Werner et al., 2003).
The temporal evolution of the deformation is generally modelled with linear functions in most
of the A-DInSAR approaches (Ferretti et al., 2000; Ferretti et al., 2001) and, in few other
cases, a more complex description of the temporal behaviour of the deformation can be used
(Berardino et al., 2002; Mora et al., 2003; Colesanti et al., 2003; Lanari et al. 2004; Hooper et
al., 2004).
Thanks to the availability of historical SAR data, it is possible to reconstruct more than 17
years of temporal evolution of past and current ground deformation, starting from 1992
(thanks to ERS1/2 archives) and up to present days (ENVISAT, RADARSAT-1/2, ALOS,
COSMO-SkyMed, TerraSAR-X).
Many applications have demonstrated the capabilities of these techniques for the detection of
slow land surface deformations (up to a few cm/year) and for the understanding of the spatial
distribution and temporal evolution of displacements, in built-up areas affected by slow or
very slow landslides (Berardino et al., 2003; Bovenga et al., 2006; Catani et al., 2005b;
Colesanti and Wasowski, 2004, 2006; Colesanti et al., 2003; Farina et al., 2003, 2006, 2008;
Hilley et al., 2004; Metternicht et al., 2005; Righini et al., 2008), tectonic motions (Brgmann
et al., 2006; Colesanti et al., 2003; Musson et al., 2004; Vilardo et al., 2009), subsidence
(Canuti et al., 2005; Colesanti et al., 2005; Colombo et al., 2003; Dixon et al., 2006; Ferretti
et al., 2000; Herrera et al. 2009; Stramondo et al., 2008), volcanic activity (Lanari et al., 2004;
Salvi et al., 2004; Tizzani et al., 2007).

SafeLand - FP7
Page 133 of 401
Deliverable 4.1
On the basis of InSAR processing approach, two different groups of A-DInSAR techniques
can be distinguished (Wasowski et al. 2007) (Table 12) :
PS-based A-DInSAR techniques or PSI (Persistent Scatterers Interferometry): these

A-DInSAR techniques study the amplitude and phase information over single isolated
objects characterized by high temporal signal stability. This approach is usually
implemented by computing differential interferograms of all the acquisitions with
respect to the same reference master image, then performing advanced phase analyses
on the pixels exhibiting stable SAR response throughout the stack. The points selected
with such a criterion are usually called Persistent Scatterers (PS) and the technique
itself is called PSI (Persistent Scatterers Interferometry).
Interferogram stacking techniques: these A-DInSAR approaches exploit more

spatially distributed information in differential interferograms obtained from
interferometric pairs with small perpendicular baseline and then infer, with a complex
averaging operation (complex multi-looking), the connected time series of phase
values due to ground deformation. These techniques mitigate the decorrelation
phenomena, maximizing the number of pixels exploited and improving the SNR
(Signal-to-Noise Ratio).
Table 12 : List of ADInSAR techniques
Persistent Scatterers Interferometry

PS-InSAR (Permanent Scatterers InSAR)
IPTA (Interferometric Point Target Analysis)
PSP (Persistent Scatterers Pair)
SPN (Stable Point Network)
StaMPS (Standford Method for Persistent Scatterers)
Interferogram Stacking Techniques

SBAS (Small Baseline Subset)
CPT (Coherent Pixel Techniques)
Most of the A-DInSAR techniques belonging to these two different classes are separately
described in the following sections.
4.2. PS-BASED A-DINSAR TECHNIQUES OR PSI (FULL RESOLUTION /

DOMINANT SCATTERERING)
As discussed in the previous section, PS-based A-DInSAR techniques are based on the
analysis of amplitude and phase information over single isolated objects (called Persistent
Scatterers, PS) characterized by a high temporal signal stability and dominating a resolution
cell on the ground.
In the PSI processing, interferograms formed with a single master scene are analyzed at single
look resolution (analyses are carried out on a pixel-by-pixel basis, with no filtering of the
interferograms) in order to maximize the Signal-to-Noise Ratio (SNR) of the resolution
elements containing a single dominant scatterer.
The first PS-based approach was developed for InSAR applications by Ferretti et al. (2000,
2001), with further enhancements by Colesanti et al. (2003), and it was called and patented by
Politecnico di Milano (POLIMI, Italy) as the Permanent Scatterers Technique (PS-InSAR).
SafeLand - FP7
Page 134 of 401

Rev. No: 2
Date: 2012-02-15
Several A-DInSAR techniques have then been developed in the last ten years after PS-InSAR
(e.g., Adam et al., 2003; Crosetto et al., 2003; Lyons and Sandwell, 2003; Werner et al., 2003)
and they are usually referred to the term Persistent Scatterers Interferometry (PSI). It is noting
that the term Permanent Scatterers is instead directly associated to the patented PS-InSAR
technique whereas, Persistent Scatterers Interferometry is usually employed to indicate all
the PS-based techniques (Crosetto and Pasquali, 2008).
4.2.1. PS-InSAR (Permanent Scatterers InSAR)

The Permanent Scatterers InSAR (PS-InSAR) is a PSI technique, developed and patented in
the late 90 by Politecnico di Milano (Italy), and now exclusively licensed to its spin-off
company TeleRilevamento Europa - TRE (Ferretti et al., 2000, 2001).
PS-InSAR allows the identification of Permanent Scatterers (PS) through the analysis of pixel
amplitude dispersion with time, over data-stack of at least 20-30 SAR acquisition.
PS are objects on the ground that return stable signals to the satellite sensor (remain coherent
over long time intervals) and are smaller than the resolution pixel cell (point-like scattering
mechanism). They typically correspond to objects on man-made structures, such as buildings,
bridges, dams, water-pipelines, antennas, as well as to stable natural reflectors (e.g. exposed
rocks) (Figure 73) (Colesanti et al., 2003; Ferretti et al., 2000, 2001).
Figure 73 : Examples of man-made structures and stable natural scatterers that return stable signals back to the
satellite (http://www.treuropa.com).
PS-InSAR allows detecting and estimating, for each PS, mean annual deformation rate
(accuracy between 0.1 and 1 mm/year) along the sensor-target direction and historical
deformation series (precision up to 1-2 mm for each single measure).
The methodology allows obtaining good phase coherence from nearly all radar scenes
regardless of geometrical baseline (perpendicular baseline for ERS up to 1.6 km), exploiting
all available radar scenes in the historical archive, estimating and removing atmospheric phase
contributions from the deformation phase signal.
The technique can be used with different approaches, for both large scale studies (e.g.
landslide inventory mapping) and high-resolution applications on limited areas (e.g. landslide
rapid mapping and monitoring, stability assessment of individual buildings or structures)
(Figure 74 and Figure 75).

SafeLand - FP7
Page 135 of 401
Deliverable 4.1
Figure 74 : Example of PS-InSAR analysis at large scale: ground deformation due to active faults in the S flank
of Mount Etna (Italy) (http://www.treuropa.com).
Figure 75 : Example of PS-InSAR analysis at local scale. Left: deep-seated gravitational slope deformation on
Motta Plet, Alpine region (Italy). Right: gravitational slope deformation threatening the village of
Valtournanche, Alpine region (Italy). The local analysis allowed the definition of the landslide boundaries and
state of activity, integrating information by geo-morphological studies on both the areas
(http://www.treuropa.com).
TRE has recently developed a new algorithm, SqueeSAR, which exploits phase signals also
for Distributed Scatterers (DS) and uses them, together with Permanent Scatterers, for
monitoring ground displacements with the same accuracy as a conventional PS-InSAR
analysis. These targets, the DS, consist of an extensive area where the back-scattered energy
is less strong than PS, but statistically homogeneous within the area (e.g. non-cultivated lands,
desert areas) (http://www.treuropa.com).
4.2.2. IPTA (Interferometric Point Target Analysis)

The Interferometric Point Target Analysis (IPTA) is a PSI technique, developed by Gamma
Remote Sensing and Consulting AG (Switzerland). As well as the other PSI approaches,
SafeLand - FP7
Page 136 of 401

Rev. No: 2
Date: 2012-02-15
IPTA exploits the temporal and spatial characteristics of interferometric signatures collected
from point targets to accurately map surface deformation histories, terrain heights and relative
atmospheric path delays. It allows using both the coherence-based and the amplitude-based
pixel selection criterion (Werner et al., 2003a; 2003b; Wegmuller et al., 2004; Strozzi et al.,
2006). Figure 76 gives an overview of its processing chain.
Figure 76 : Interferometric Point Target Analysis (IPTA) processing approach (Werner et al., 2003a; 2003b)
The phase model implemented in IPTA is the same as used in conventional Interferometry
(Figure 77): the unwrapped interferometric phase is expressed as the sum of a topographic
phase, a deformational phase, a differential path delay phase (also called atmospheric phase),
and phase noise (or decorrelation) terms.
An important aspect of IPTA is that for point targets no spatial decorrelation occurs,
permitting interpretation of the interferometric phase of pairs with long baselines, even above
the critical baseline (Werner et al., 2003a). Obviously, a reflector must also remain stable over
the time period of interest to permit analysis of the phase history. Based on these ideas one
important objective of IPTA is to achieve a more complete use of the available data. Through
the use of point targets, interferometric pairs with long baselines can be used. Consequently,
more observations are available with respect to conventional SAR Interferometry, permitting
reduction of errors resulting from the atmospheric path delay and leading to better temporal
coverage (Figure 78).

SafeLand - FP7
Page 137 of 401
Deliverable 4.1
Figure 77 : Two-dimensional regression analysis of differential interferometric phase difference of two close
points in a stack of 59 ERS interferograms (left). Linear deformation rate due to subsidence associated with the
Raymond Fault zone in south Pasadena USA (right) (Werner et al., 2003a).
Figure 78 : IPTA analysis of the Montagnon (Switzerland) landslide: PT over shaded relief, ortho-photo and
topographic map (www.gamma-rs.ch).
Most of IPTA analyses have been performed to monitor slow and temporally uniform
deformation, for which temporal and spatial sampling of the signal is very good. In the case of
high deformation rates the capability of IPTA to use pairs with large baselines have the
advantages that high phase gradients can be reduced if shorter observation intervals become
available (Figure 79). In addition, large scale corrections such as baseline errors and the large
scale component of the atmospheric distortions can be estimated independently of the areas
with high deformation gradients and interpolated or extrapolated to get relatively accurate
corrections for the entire area.
SafeLand - FP7
Page 138 of 401

Rev. No: 2
Date: 2012-02-15
Gamma Remote Sensing has recently developed an approach for the integration of ERS and
ENVISAT ASAR data into the Interferometric Point Target Analysis (IPTA). This integration
is very relevant for current time monitoring. Furthermore, the accuracy of current time
monitoring based on a smaller number of ASAR acquisitions can be improved through the
integration of additional ERS acquisitions, e.g. through a better identification of adequate
points and through a more accurate estimation of the related point heights (Wegmuller et al.,
2006).
Figure 79 : Map of linear deformation rate and selected deformation histories of rapid subsidence at the Lost
Hills oil field, California, monitored between February 2002 and February 2004, by means of IPTA technique
on RADARSAT data (Wegmuller et al., 2006).
4.2.3. PSP (Persistent Scatterer Pairs)

Persistent Scatterers Pairs - Differential SAR Interferometry (PSP-DIFSAR) is a PSI
technique, developed recently by Telespazio/e-GEOS (a Finmeccanica/Thales company)
(http://www.telespazio.it, http://www.e-geos.it).
PSP technique efficiently allows identifying Persistent Scatterers in series of full resolution
SAR images and retrieving the corresponding terrain height, displacement velocity and time
series. This approach differs from standard PSI methods because during the processing it
detects and removes atmosphere phase artefacts by exploiting their spatial correlation, without
using model-based interpolations starting from a preliminary set of measurements obtained by
radiometric or low resolution analyses. PSP does not require data calibration or pre-selection
of radiometrically stable points, thus is not affected by errors in the pre-selection or by the
density of pre-selected points. Moreover, the method is characterized by the exploitation of

SafeLand - FP7
Page 139 of 401
Deliverable 4.1
redundant information, which makes for a very good robustness to noise (Costantini et al.,
2009).
PSP-DIFSAR technique jointly exploits spatial and temporal properties of SAR data, in order
to improve the density and the accuracy of the measurements. Its results with large stacks of
SAR data have successfully shown its capabilities both for the density and the accuracy of
deformation measurements.
The PSP approach can also provide valid tools for the identification and analysis of
distributed scatterers, i.e. scatterers that do not behave as point-like objects, but nevertheless
keep a good degree of coherence over time. In fact, the distributed scatterers can be
recognized by analyzing, together with the temporal properties, the spatial correlation
properties of the images acquired at different times (Costantini et al., 2009).
First applications of PSP technique have been carried out to monitor subsidence in Bologna
and Napoli (Italy) (Costantini et al., 2000a e 2000b), ground instability in Maratea (Italy)
(Berardino et al., 2003) (Figure 80), rock-slides and deep-seated gravitational slope
deformations in Valfurva and Valdisotto areas (Italy) (Figure 81) and, most recently, the
approach has been tested with X-band COSMO-SkyMed data (Costantini et al., 2009).
Figure 80 : Ground displacement map of Maratea landslide (Italy), monitored between August 1997 and
January 2000 by means of PSP technique on ERS data and results from the GPS survey (June 1997March
2000) (Berardino et al., 2003).

SafeLand - FP7
Page 140 of 401

Rev. No: 2
Date: 2012-02-15
Figure 81 : Deformation map in the area of Valfurva and Valdisotto (Italy), monitored by means of PSPDIFSAR technique on ERS1/2 data in the framework of PREVIEW project (Righini et al., 2008).
4.2.4. SPN (Stable Point Network)

The Stable Point Network (SPN) is a PSI technique, developed by Altamira Information SL
as the result of research collaborations with CNES (Centre National dEtudes Spatiales,
French Space Agency), ESA (European Space Agency). Its processing chain was the first
advanced interferometric approach capable of merging the new ASAR data with the historical
ERS-1/2 data (Arnaud et al., 2003; Duro et al., 2005).
SPN software uses the DIAPASON (Differential Interferometric Automated process Applied
to Survey Of Nature) interferometric chain for all SAR data handling, e.g. co-registration
work and interferogram generation.
Figure 82 gives an overview of the SPN processing steps: starting from a set of multitemporal SAR images, the mean displacement rates, height errors and time series can be
extracted. The quality of the results is related with the number of acquisitions (an increase of
SAR images improves the quality of measurements) and an error of 1 mm/year for
deformation rates and 2 m for height errors can be achieved (Herrera et al., 2009b).

SafeLand - FP7
Page 141 of 401
Deliverable 4.1
Figure 82 : Stable Point Network processing chain flow chart (Herrera et al., 2009b).
SPN is very flexible and the software can work at any resolution, any frequency and with
cropped images. In particular, the algorithm can either work at full resolution (for ERS and
ENVISAT is 4 m by 20 m) selecting the measurement points by analyzing the amplitude of
the set of SLC SAR images, or at reduced resolution (e.g. 40 by 40 m) selecting ground
surface natural reflectors or Stable Point (SP) by interferometric multi-look coherence
(Herrera et al. 2009a).
The amplitude selection mode is based on analyzing the temporal stability of the
electromagnetic response to the radar signal (backscattering) for each pixel of the study area.
If the geometry of the ground surface that corresponds to a certain pixel does not vary
significantly along time, the backscatter is similar for every acquired SAR image, and the
pixel is selected as a SP. The coherence based approach, rather performs an average of the
backscattering of neighbour pixels, increasing the measurement point density, and it is
particularly useful for non-urban and vegetated areas where ground surface scatterers
(buildings, structures, metallic objects) are scarce (Herrera et al., 2009b).
In Figure 83 is reported an example of subsidence mapping using PSI data elaborated with the
SPN amplitude-based approach, whereas in Figure 84 is reported an example of mapping and
monitoring of geomorphologic processes in mountainous area by means of the SPN
coherence-based approach.
SafeLand - FP7
Page 142 of 401

Rev. No: 2
Date: 2012-02-15
A detailed description of the SPN algorithm can be found in Crosetto et al. (2008) where
some examples of urban subsidence monitoring are shown. To date the performance of the
SPN method has been mainly validated in subsidence analysis (Herrera et al., 2009a, 2009b).
Figure 83 : Total deformation estimated with amplitude-based SPN technique (full resolution) between July
1995 and December 2005 (top) and between January 2004 and December 2008 (bottom) in the area of Murcia
(SE Spain) (Herrera et al., 2009a).

SafeLand - FP7
Page 143 of 401
Deliverable 4.1
Figure 84 : Map of Formigal landslide (Central Pyrenees, Spain) showing the displacement velocity estimated
with the coherence-based SPN techniques (low resolution) in the analyzed ERS1/2 and ENVISAT descending
orbits between 1995 and 2007 (Herrera et al., 2009b).
4.2.5. StaMPS (Stanford Method for Persistent Scatterers)

The Stanford Method for Persistent Scatterers (StaMPS) is a PSI technique developed by the
Stanford University, for analyzing episodic crustal deformation in non-urban environments,
with application to volcanic settings (Hooper et al., 2004 and 2007).
Most of PSI algorithms identify an initial set of PS pixels by the analysis of their amplitude
distribution over time in a series of interferograms. These methods work best in urban areas
where man-made structures increase the likelihood of finding a non-fluctuating scatterer in
any given pixel but, the density of PS pixels in natural terrains, however, is generally too low
to obtain any reliable results (Hooper et al., 2004).
StaMPS is based primarily on interferometric phase characteristics and finds low-amplitude
pixels (Persistent Scatterers, PS) with phase stability that are not identified by the other
amplitude-based algorithms. Scatterers with stable phase characteristics independent of
amplitudes associated with man-made objects can be found, and thus the technique is also
applicable to areas where conventional InSAR fails due to complete decorrelation of the
majority of scatterers, yet a few stable scatterers are present (Hooper et al., 2004).
StaMPS algorithm involves removing the residual topographic component of flattened
interferogram phase for each PS, then unwrapping the PS phases both spatially and
temporally. The approach also uses the spatial correlation of the phases rather than a wellGrant Agreement No.: 226479
SafeLand - FP7
Page 144 of 401

Rev. No: 2
Date: 2012-02-15
defined (e.g. linear) phase history and a prior assumptions about temporal nature of
deformations, so that temporally-variable processes (e.g. volcanic deformation) can be
observed (Figure 85 and Figure 86).
Figure 85 : Mean deformation rates and example of time series achieved by means of StaMPS technique in the
area of Long Valley Caldera area, California (Hooper et al., 2004).

SafeLand - FP7
Page 145 of 401
Deliverable 4.1
Figure 86 : Mean deformation velocities for stable InSAR pixels in the area of Katla Volcano, Iceland, measured
on a data-stack of ERS images acquired between 1995 and 1998 and processed with the StaMPS technique. GPS
velocities relative to Reykjavik on the North American plate are also plotted and projected into the satellite lineof-sight (Hooper and Pedersen, 2007).
4.3. INTERFEROGRAM STACKING TECHNIQUES (LOW RESOLUTION /

DISTRIBUTED SCATTERERING)
Interferogram stacking techniques are the direct extension of the classical DInSAR technique
to the multipass case. Since the first description of the technique (Gabriel et al., 1989), the
DInSAR technique, which was originally derived with respect to two acquisitions (a single
interferogram), was applied to detect the presence of ground displacements in mainly in
volcanic areas and after earthquakes. The standard two pass DInSAR was however limited by
the presence of at least two error sources: the atmospheric propagation delay (APD) and the
inaccuracies of the external DEM used in the elimination of the topography component from
the signal interferences. Above limitations were overcame for the first time by the PS
techniques exploiting long acquisition sequences, i.e., by generating a
multitemporal/multibaseline (MT/MB) interferogram stack. Differently from the PS
technique, which is tailored to the monitoring of dominant scatterer, interferogram stacking
techniques assume the scattering to be distributed. The first interferometric stacking technique
proposed in the literature has been the Small BAseline Subset (SBAS) technique (Berardino
et al., 2002) developed by the Institute for Electromagnetic Sensing of the Environment
(IREA) of the Italian National Research Council (www.irea.cnr.it).
4.3.1. SBAS (Small Baseline Subset)

Following the lines of the classical (single-pair) DInSAR analysis, SBAS carries out an
analysis of multilook interferograms: the multilook operation, i.e. a spatial averaging which is
carried out by exploiting the hypothesis that the scattering is distributed, allows increasing the
phase signal quality. The angular decorrelation is a phenomenon which is associated with

SafeLand - FP7
Page 146 of 401

Rev. No: 2
Date: 2012-02-15
scattering chances of the phase caused even by small variations of the radar line of sight
direction in elevation (induced by orbital separation) and/or aspect (induced by doppler
centroid); temporal decorrelation is due to changes over the time of the scattering. The SBAS
technique uses only interferograms generated by choosing thresholds on the spatial and
temporal baselines, that is the spatial orbital and temporal separations, respectively, and on
the Doppler centroid difference thus limiting the effects of angular and temporal
decorrelation. After the unwrapping, which can be performed with algorithms commonly
used in SAR Interferometry (for instance the Minimum Cost Flow algorithm) the
interferograms, which embed the information of the phase variation between two acquisitions,
are inverted to retrieve the phase signal over the stack of acquisitions. The SBAS approach,
by using the Singular Value Decomposition (SVD) technique allows also handling the case in
which, due to the limitations on the baselines, the acquisitions are grouped in different
independent subsets. This latter feature allows also easily integrating data of different sensors
(f.i. ERS and ENVISAT).
After the inversion the deformation time series are separated from atmospheric contribution
and possible DEM errors (residual topographic) via the use of proper filtering based on the
deterministic and statistical characterization of the single components.
The phase unwrapping represents, as for all other techniques, the most critical step of the
SBAS approach. Advanced PhU algorithm such as that proposed in (Pepe and Lanari 2006) or
in the Enhanced Spatial Difference (ESD) approach (Fornaro et al 2009a) allows improving
the performances of this class of radar monitoring techniques.
The SBAS approach has been widely applied for the investigation of deformations associated
with several phenomena, such as those occurring in volcanic areas, fault deformation (Lanari
et al 2007, Atzori et al. 2008), mining and in general underground excavation, and landslides.
In (Casu et al. 2006) a detailed analysis of the accuracy achieved by DinSAR stacking
(particularly SBAS), is presented. Figure 87shows an example of the SBAS (low resolution)
product relevant to a very large area (Casu et al 200)

SafeLand - FP7
Page 147 of 401
Deliverable 4.1
Figure 87 : Mean deformation velocities of an area of about 600Kmx100Km obtained with the SBAS technique
in Nevada USA (Copyright IEEE).
Several techniques have been also proposed for the extension/integration of the low resolution
SBAS processing chain for the analysis of deformation at the highest resolution. Among them
it is worth to highlight the straightforward extension of SBAS approach (Lanari et al. 2004) to
the high resolution data in which, always by limiting the spatial baseline, the data set to the
highest resolution is compensated for small scale deformation and atmospheric phase
propagation delay resulting from the low resolution analysis. The limitation on the baselines
may be however, on high resolution data, inappropriate because at this scale generally the
scattering occur with respect to dominant scattererers which in most of the cases do not
exhibit angular and temporal decorrelation. Another technique, known as 4D (space-velocity)
SafeLand - FP7
Page 148 of 401

Rev. No: 2
Date: 2012-02-15
imaging or tomography, is based on both the amplitude and the phase information of the
signal (Fornaro et al 2009b, Fornaro et al 2010). This technique does not limit the baselines:
beside the possibility to locate persistent scatterers even interfering within the image pixels, it
also outperforms the classical persistent scatterers technique in terms of detection of targets at
the highest resolution that can be monitored by the radar technique (Fornaro et al 2009c).
Figure 88 shows the results of the application of the tomographic technique to the area of
Rome, see (Fornaro et al 2010).
Figure 88 : Mean deformation velocities in the city of Rome obtained via the full resolution tomographic
analisys (Copyright IEEE).
4.3.2. CPT (Coherent Pixel Techniques)

The Coherent Pixel Technique (CPT) is a DInSAR stacking technique developed by Remote
Sensing Laboratory of the Universitat Politecnica de Catalunya (Blanco et al. 2008) which is
very similar to the SBAS algorithm. In particular, as for the SBAS approach, it applies to
interferograms which have been spatially averaged (multilooked) and hence well fits with
distributed scattering such as that occurring over bare soils. Differently from the SBAS
approach, the CPT technique uses a Minimum Spanning Three (MST) strategy to connect the
acquisition stack, and therefore to generate the interferograms. On one hand this strategy can
reduce the computational effort, on the other hand it limits the degree of redundancy of the
interferogram network and therefore it may be more critical in terms of robustness with
respect to phase unwrapping errors. Another special feature of the CPT approach is related to
the implementation of a step which allows, via modelling, to retrieve the linear deformation
component and DEM error contribution prior to extraction of the non linear component. In

SafeLand - FP7
Page 149 of 401
Deliverable 4.1
this sense the technique is very similar to the ESD approach proposed in (Fornaro et al
2009a).
Figure 89 presents the results obtained by using the CPT technique for monitoring the urban
subsidence in Murcia with 39 TerraSAR-X images from 18/07/2008 to 25/11/2009 within the
framework of a project involving the Universitat Politcnica de Catalunya and the Universitat
de Alicante. The TerraSAR-X images were provided by DLR in the framework of the
scientific project GEO0389.
Figure 89 : Mean deformation velocities in the city of Murcia obtained via the (low resolution) CPT technique
applied to TerraSAR-X data (Courtesy of the RSLab of the Universitat Politecnica de Catalunya).
4.4. FINAL REMARKS

Spaceborne DInSAR technology is rather mature and many applications examples on medium
resolution data are available in the current literature. Among all the applications, spanning
from the monitoring of volcanic and seismogenetic areas to the underground extraction and
resource management, etc., the monitoring of landslides is the most challenging one mainly

SafeLand - FP7
Page 150 of 401

Rev. No: 2
Date: 2012-02-15
due to the presence of vegetation and the of slope variation. The latter causes in most of cases
the loss of visibility on one of orbit typology (ascending or descending).
The new generation high and very high resolution satellite SAR systems, such as COSMOSKYMED, TerraSAR-X and the upcoming Tandem-X represent certainly a significant
improvement for the technology in terms of data availability and coverage. Nonetheless, the
temporal sampling (related to the so-called revisiting time) and the scene viewing geometry
cannot be freely set in the case of satellites: both factors represent a rather stringent limitation
especially for the monitoring of landslides where the dynamic of the phenomenon, and both
the exposition and slope are highly variable. One of the future challenges in terms of
development of the DInSAR technology is certainly associated with the possibility to migrate
to airborne sensors. In this case operational flexibility would be strongly improved in terms
repetition and orientation of the passes. However, compensation of the effects associated with
the track instability represents a major technological gap.
5. GROUND-BASED INSAR
5.1. INTRODUCTION
Differential Interferometry based on SAR images (DInSAR) was first developed for data from
spaceborne platforms, but most of these methods can be applied to data from ground-based
microwave interferometer. Ground-based differential interferometry (GBInSAR) became
more and more popular in the last ten years, in particular for monitoring landslides and
instable slopes. A recent overview about this technique can be found in (Luzi, 2010).
It is possible to acquire SAR images through a portable SAR set up on a stable area. The
synthetization of the radar antenna is usually obtained by the linear and regular motion of a
microwave transceiver on a rail. Ground-based radar installations are usually at their best
when monitoring small objects like buildings, small urban areas or single hillsides, while
imaging from satellite radar is able to monitor very large areas. As for satellites data,
GBInSAR images acquired at different dates can be used for interferometry when the
decorrelation among the different images is kept low. Satellite observations are sometimes not
fully satisfactory because of a lengthy revisit time or of changes on observational geometry.
In ground-based interferometry, the high number of measurements, available with less than
hourly cadence, makes decorrelation of minor concern. Moreover satellite interferometry can
only imagine sites with a proper visibility and from where the component of the displacement
along the line of sight of the satellite can be observed.
Satellite and ground-based radar interferometry are derived from the same physical principles
but they are often characterized by specific problems mainly due to the difference of
geometry of the observation. In the last decade several papers have been issued about the
feasibility of monitoring buildings or structures (Tarchi et al., 1997), glaciers (Luzi et al.,
2007), volcanoes (Casagli et al., 2003, Antonello et al., 2006) and landslides (Tarchi et al.,
2003b, Leva et al. 2003), this last one appearing to be the most consolidated application. A
number of experimental results demonstrate the effectiveness of GBInSAR for remote
monitoring of terrain slopes and as an early warning system to assess the hazard of rapid
landslides.

SafeLand - FP7
Page 151 of 401
Deliverable 4.1
5.2. THE GBINSAR INSTRUMENTATION

Despite the use of the same physical principle, the satellite and ground-based approaches
differ in some aspects. In particular radar sensors of different kinds are usually employed
mainly because of technical and operational reasons. While satellite SAR systems due to the
need of a fast acquisition are based on standard pulse radar, continuous wave step frequency
(CWSF) radars are usually preferred in ground-based observations. The Joint Research Center
(JRC) has been a pioneer of this technology and in this centre the first prototype was born.
The first paper about a GBInSAR interferometry experiment dates back to 1999 (Tarchi et al.,
1999), reporting a demonstration test on dam. The used equipment was composed of a radar
sensor based on Vectorial Network Analyser (VNA), a coherent transmitting and receiving
set-up, a mechanical guide, a PC based data acquisition and a control unit.
After some years a specific system, known as GBInSAR LiSA, reached an operative state and
became available to the market by Ellegi-LiSALab company which on June 2003 obtained an
exclusive licence to commercially exploit this technology from JRC. The basic and simplest
schematic of the radiofrequency set-up used for radar measurements is shown in Figure 90
along with a simple scheme of the GBInSAR acquisition. Advanced versions of this set-up
have been realized in the following years to improve stability and frequency capabilities
(Rudolf et al., 1999 and Noferini et al., 2005). This apparatus is able to generate microwave
signals of increasing frequencies sweeping a radiofrequency band. This approach, which
apparently differs from the standard pulse radar, has actually the same physical meaning,
because a temporal pulse can be obtained after Fourier anti-transforming of the frequency data
(the so called synthetic pulse approach).
Figure 90 : A) Basic scheme of the RF section of a C band transceiver based on the Vectorial Network Analyser
VNA. B) GB SAR acquisition through a linear motion.
The rapid growth of microwave technology that occurred in the last years encouraged the
development and realization of different instruments (Pipia et al., 2007 Bernardini et al.,
2007). In the last few years an Italian company, IDS Ingegneria dei Sistemi
(http://www.idscompany.it/) has started commercializing the first off-the-shelf GBInSAR
system, the IBIS-L, which is based on a Ku-band SAR sensor, see Figure 91. Recently a
ground-based interferometer with a non-SAR approach has been designed with similar
monitoring purposes (Werner et al., 2008). Data are processed in real time. An algorithm
combines the received amplitude and phase values stored for each position and frequency
values, to return complex amplitudes (Fortuny J. and A.J. Sieber, 1994). The optimization of
focusing algorithms has been recently updated by Reale et al, 2008 and Fortuny, 2009. To

SafeLand - FP7
Page 152 of 401

Rev. No: 2
Date: 2012-02-15
reduce the effect of side lobes in range and azimuth synthesis (Mensa D.L., 1991), data are
corrected by means of a window functions (Kaiser, Hamming etc.), for range and azimuth
synthesis. The attainable spatial resolutions and ambiguities are related to radar parameters
through the relationships shown in Table 13. The accuracy of the measured phase is usually a
fraction of the operated wavelength: by using centimetre wavelengths, a millimetre accuracy
can be attained. The phase from complex images can suffer from the ambiguity due to the
impossibility of distinguishing between phases that differ by 2. Single radar images are
affected by noise and related interferometric maps must be obtained through adequate phase
stability between the pair of images: only pairs whose coherence loss can not affect the
accuracy of the interferometric maps are usable. This task is of major difficulty when the
considered time period is of the order of months.
A detailed analysis to the possible causes of decorrelation in the specific case of GBInSAR
gathering many images per day for continuous measurements has been discussed by some
researchers (Luzi et al., 2004 and Pipia et al., 2007). For campaigns carried out on landslides
moving only few centimeters per year, when the sensor is periodically installed at repeated
intervals several months apart over the observation period, a novel method has been proposed
(Noferini et al., 2005).
Table 13 : Calculated resolution available from a CWSF radar observation; B radiofrequency bandwidth, in
vacuum wavelength, f frequency step, Lx rail length, R range, c light velocity.
Range resolution
Rr
Azimuth resolution
Raz
Non ambiguous range (m )
Rna
c
2B
c
2Lx
c
2f
As mentioned, the monitoring of instable slopes through GBInSAR interferometry is a well

consolidated application largely reported in literature (Leva et al. 2003, Pieraccini et al., 2003,
Tarchi et al., 2003a). The investigation and interpretation of the patterns of movement
associated with landslides have been undertaken by using a wide range of techniques,
including the use of survey markers: extensometers, inclinometers, analogue and digital
photogrammetry, both terrestrial and aerial. In general, they suffer from serious shortcomings
in terms of spatial resolution. GBInSAR, thanks to its spatial and temporal sampling can
overcome the restrictions of the conventional point-wise measurement.

SafeLand - FP7
Page 153 of 401
Deliverable 4.1
Figure 91 : IBIS-L, the first GBInSAR system commercially available.
5.3. CHARACTERISTICS OF APPLICATION

Here are presented the main characteristics of some devices that were directly experienced by
the University of Firenze, namely LiSA-Mobile by Ellegi-Lisalab and IBIS-L by IDS. The
two systems share most of the basic features, the main difference being that IBIS-L generates
the microwave signal by using an industrialized method derived from a prototype developed
in collaboration with the Department of Electronics of the University of Firenze, while LiSAMobile is based on a Network Analyzer (NA). This theoretically allows LiSA-Mobile to
spread on a wider waveband, although for most geological applications the optimal
parameters generally vary within a relatively restricted range. A new version of LiSA-Mobile,
faster and lighter, has been produced to fulfill the need of an application in emergency
conditions, though it uses no more a Network Analyzer to generate the microwave signal.
Therefore, the main differences in the choice of a GBInSAR lie especially in the policies and
in the software adopted by the two companies. IDS favours a more commercial use of its
system and makes the customer able to carry out on his own any campaign. On the other side
Ellegi-Lisalab preferably sells a service and hence grants a major flexibility both in the radar
parameters and in the mechanical part.

SafeLand - FP7
Page 154 of 401

Rev. No: 2
Date: 2012-02-15
IBIS-L:
Target distance: 10 - 4000 m.
Central frequency: 17.2 GHz (Ku band) or 9.7 GHz (X band).
Bandwidth: 300 MHz.
Polarization: VV.
Synthetic aperture: 2 m.
Range resolution: 0.5 m.
Azimuth resolution (at 500 m): 2.2 m.
Scansion time: <5 minutes.
Opening angles: 70 (azimuth), 60 (elevation).
Minimum and maximum displacement measured: from mm/y to 1 - 2 m/day.
Notes: the system can also be installed in a static configuration (IBIS-S); in this case it
can be mounted on a single tripod, it has a scansion time of 1 msec and it can be used
to measure the vibrations of the target, though it is not able to produce 2D
displacement maps. A GBInSAR operating in X band is also produced. Aresys, a spinoff of the Politecnico di Milano, developed another software able to elaborate the data
collected by the radar by using a technique similar to the Permanent Scatterers
approach (Ferretti et al., 2000) to reduce limitations due to vegetation and atmospheric
effects.
The Figure 92 shows a displacement map of the Torgiovannetto landslide (Perugia, central
Italy) elaborated through IBIS-L system.

SafeLand - FP7
Page 155 of 401
Deliverable 4.1
Figure 92 : Displacement map of the Torgiovannetto landslide (Perugia, central Italy). A software able to
elaborate data using a technique similar to the Permanent Scatterers approach (Ferretti et al., 2000) is also
available.
LiSA-Mobile:
Target distance: 10 - 4000 m.
Central frequency: 17.2 GHz (Ku band, the new version); from 10 MHz to 20 GHz
(the version with NA).
Bandwidth: up to 400 MHz (the new version); no theoretical limit (the version with
NA).
Polarization: VV.
Synthetic aperture: 2-3 m (no theoretical limit).
Range resolution: 0.38 0.75 m (the new version); depending on the bandwidth (the
version with NA).
Azimuth resolution (at 500 m): varies depending on the synthetic aperture and on the
central frequency adopted. Using a 3 m long rail and a central frequency of 17.2 GHz
the azimuth resolution at 500 m is 1.5 m.

SafeLand - FP7
Page 156 of 401

Rev. No: 2
Date: 2012-02-15
Scansion time: varies depending on the rail length. The new system performs 3 m of
synthetic aperture in 3 minutes.
Opening angle: 50 (can vary up to 120 depending on the antennas installed).
Minimum and maximum displacement measured: from mm/y to 3 - 4 m/day (the new
version).
Notes: the new version of LiSA-Mobile, especially designed for being used in
emergency condition, has an electric consumption of less than 100 W, while the NA
version consumes about 1000 W. Both systems can be installed on a tripod to work in
static configuration. Also the rail can be mounted on two tripods to improve its prompt
installation.
Below a displacement map of the Torgiovannetto landslide monitored by using the system of
Ellegi-Lisalab is shown (Figure 93).
Figure 93 : Displacement map of the Torgiovannetto landslide (central Italy) elaborated by using the GBInSAR
system of Ellegi-Lisalab.

SafeLand - FP7
Page 157 of 401
Deliverable 4.1
It should be underlined that some features, such as the maximum displacement rates that can
be successfully measured, the minimum and maximum range distance, the limitations due to
the vegetation and so on, are only partly tied to the characteristics of the radar system as they
also depend dramatically on the conditions of the monitored target and on other parameters
that influence each other.
6. POLARIMETRY FOR SOIL MOISTURE ESTIMATION

6.1. INTRODUCTION
Polarimetry deals with the vector nature of the electromagnetic fields: radar polarimetry
concerns with the utilization of polarimetry in radar applications.
A polarimetric Synthetic Aperture Radar (SAR) system measures the electric field,
backscattered by the scene, including its polarisation state and hence the scattering matrix.
The interaction of the transmitted wave with a scattering object transforms its polarisation.
Therefore, the polarisation of the backscattered wave depends on the polarisation of the
transmitted wave as well as on the scattering properties of the imaged objects.
One special characteristic of SAR polarimetry is that it allows a discrimination of different
types of scattering mechanisms. This becomes possible because the observed polarimetric
signatures depend strongly on the actual scattering process.
In comparison to conventional single-channel SAR, the inclusion of SAR polarimetry
consequently can lead to a significant improvement in the quality of classification and
segmentation results. Certain polarimetric scattering models even provide a direct physical
interpretation of the scattering process, allowing an estimation of physical ground parameters
like soil moisture and surface roughness, as well as unsupervised classification methods with
automatic identification of different scatterer characteristics and target types.
SAR polarimetry additionally offers some limited capability for separating multiple scattering
mechanisms occurring inside the same resolution cell and can be deemed as a first step in
resolving the ambiguous scattering problem in SAR, as mentioned above. With polarimetric
decomposition techniques a received signal can be split into a sum of three scattering
contributions with orthogonal polarimetric signatures. This can be used for extracting the
corresponding target types in the image, even in the case that they are occurring
superimposed. Also, if a signal is disturbed by undesired orthogonal contributions, in this way
the relevant components can be extracted.
Radar polarimetry has a wide spectrum of applications in
Agriculture: to discriminate different agricultural fields (Lemoine et al., 1994);
Forestry: biomass estimation and discrimination of different species (Le Toan et al.,
1992; Rignot et al., 1995);
Hydrology: surface roughness and soil moisture (Lin et al., 1994; Dubois et al., 1995);
Snow moisture measurements (Shi and Donzier, 2000);

SafeLand - FP7
Page 158 of 401

Rev. No: 2
Date: 2012-02-15
Geology: characterization of geological structures (Oh et al., 1992);
Geohazard: detection of surface changes produced by landslides (Czuchlewski et al.,

2003).
Figure 94 : Example of scattering mechanism decomposition images from fully polarimetric RADARSAT-2 data
(a) and comparison with simulated results (source: Canada Centre for Remote Sensing,
http://www.ccrs.nrcan.gc.ca)
6.2. SOIL MOISTURE ESTIMATION

The backscatter coefficient of sparsely vegetated land is related to the frequency, angle of
incidence and polarization of the incident wave, as well as to the roughness of the soil and its
complex dielectric constant in Ulaby et al. (1982), Tsang et al. (1985) and Beckmann and
Spizzichino (1987). This in turn depends on the composition and density of the soil and,
above all, on its water content. Therefore, it is theoretically possible to estimate the soil
moisture data using Synthetic Aperture Radar (SAR).
More precisely, considering that at microwave frequencies the thickness of penetration into
the soil is of the order of centimeters, what is possible is to estimate the water content in the
topsoil to a depth of about 5 cm (Ulaby et al., 1982) (the exact thickness value depends on the
frequency and in turn on the water content and the intensity of the incident radiation).
However, in practice the estimation of soil moisture is complicated by the fact that the
dependence of the backscattering coefficient of dielectric constant is masked by the strong
dependence on the roughness of the soil. Therefore, methods for estimating the water content
must necessarily be based on empirical, semi-empirical or theoretical direct models, providing
the backscatter coefficient as a function of the complex dielectric constant and the parameters
describing the roughness of the soil (including, obviously, the nature of the incident wave).
Moreover, to distinguish the effect of moisture from the surface roughness at the inversion
stage, the roughness of the land must be known or otherwise provided (e.g. if you want to
SafeLand - FP7
Page 159 of 401
Deliverable 4.1
monitor developments over time in the water content of a specific small area, the roughness
can be measured in situ once and for all, and humidity can be estimated from time to time by
SAR data), or it needs to have more measurements at different frequency and/or polarization
and/or incidence angle (that is not possible with ERS data, but e.g. with ENVISAT, although
with a time of revisiting excessive, or COSMO-SkyMed).
Use of models is therefore mandatory for retrieving soil parameters. Typically used models
are divided in empirical, theoretical and semi-empirical.
Empirical models establish a simplified relation between the surface parameter and the
diffuse field. The simplicity is however paid with a limitation of the validity outside
the site of calibration.
Theoretical models are based on the evaluation of the scattered field starting from the
knowledge of the tangential electric and magnetic components on the interface
between the air and the terrain. Approximations in the evaluation of the field
components at the interface pose a limitation to the range of acceptable values of the
dielectric constant and of the roughness.
Semi-empirical models are obtained by a simplification of the theoretical models in

terms of parameter empirically estimated. This class of models has characteristics
which are intermediate with respect to the above classes.
Earlier studies using single frequency and single polarization SIR-B (Shuttle Imaging RadarB) imagery could only describe the dependence of backscattering coefficient on the surface
parameters (mainly, roughness and dielectric constant) separately. The inversion of soil
moisture information from radar backscatter became more rigorous after the availability of
polarimetric radar data. For example, Oh et al. (1992) developed a model for estimating the
rms height of the soil surface and its moisture content from simultaneous ratio measurements
of the ratios between the HV and VV polarization and HH and VV polarization where H
and V stand for horizontal and vertical polarizations, respectively, in either transmit or
receive mode of radar operation.
In Dubois et al. (1995) an empirical model that only requires measurements of HH and VV
channels at a frequency between 1.5 and 11 GHz to retrieve both rms roughness height and
soil moisture from bare soils was developed.
An algorithm based on the regression analysis of the simulated surface backscattering
coefficients by the theoretical model was developed in Shi et al. (1997) to provide an
estimation of soil moisture and surface roughness parameter from L-band SAR co-polarized
measurements over bare and short vegetated fields. This algorithm was applied to a series of
AIRSAR and SIR-C measurements obtained over the Little Washita River watershed near
Chickasha, Oklahoma. The values of the retrieved soil moisture and surface roughness
parameter for a number of bare and short-vegetated fields were compared with those sampled
on the ground in near concurrence with the AIRSAR and SIR-C measurements. A reasonable
agreement was found between the retrieved and measured values. The rms errors of the
comparison were estimated to be 3.4% and 1.9 dB for soil moisture and surface roughness
parameter, respectively.

SafeLand - FP7
Page 160 of 401

Rev. No: 2
Date: 2012-02-15
Figure 95 : Comparison in Shi et al. (1997) between the inferred and ground measured soil moisture and
surface roughness parameter from all available SIR-C (+) and AIRSAR (*) L-band images. IEEE copyright.
Based on observations indicating the potential of using correlation information among

polarimetric data for the separation of moisture and roughness, and the demonstration in
laboratory about the relationship between surface roughness and the eigenvalues of the
polarimetric coherency matrix (which themselves have a physical significance in terms of
scattering amplitude) a different approach for decoupling soil moisture and roughness over
bare soils was proposed in Hajnsek et al. (2003). The model is an extension of the small
perturbation method. The application of the technique to experimental data acquired by the ESAR airborne system collected over two test sites in Germany (Weiherbach and Elbe-Auen)
showed good agreement between the inverted values and ground measurements. The low rms
errors for the normalized roughness (i.e. the surface rms height normalized to the
wavelength/2) from 0.14 to 0.11 and for soil moisture content from 5 to 10 [vol.%] over both
test sites indicated that the structure of the data is in accordance with the predictions of the
model over a large range of surface conditions. The inversion accuracy was high enough to
point out the seasonal variation effects. It was also pointed out that the moisture content
estimation is much very sensitive to incidence angle variations. Sensitivity analysis with
respect to terrain variations have shown that 1 degree in terrain slope variation leads to an rms
error variation in the order of 0.25 [vol.%].
To visually appreciate the coverage of the estimation, a sample result from this paper
reporting the estimated soil moisture over the two test sites is shown in Figure 96.
A strong limitation for surface parameter estimation from polarimetric SAR data is the
presence of vegetation. In the case of bare surfaces, satisfactory estimation results for soil
moisture estimation can be achieved using theoretical as well as empirical or semiempirical
approaches. However, bare fields are only a special case: agricultural fields are covered by
different crop types over large periods of their yearly cycle. The presence of vegetation
increases the complexity of the scattering scenario: the waves propagate through and interact
with the vegetation layer and then interact with the underlying surface. Therefore, vegetation
and surface effects are superimposed in the measured scattering signature and the separation
of the individual contributions becomes more complex.

SafeLand - FP7
Page 161 of 401
Deliverable 4.1
This, combined with the fact that most natural surfaces are temporarily or permanently
covered by vegetation, restricts significantly the importance of radar remote sensing for a
wide spectrum of geophysical and environmental applications.
One approach for solving the problem of surface parameter inversion under vegetation cover
is to use longer wavelengths (lower frequencies). P-band is, for example, such a potential
frequency candidate with sufficient high penetration into and through vegetation layers.
However, first experimental results at P-band indicate that this approach solves only one part
of the problem.
Figure 96 : Estimated surface roughness ks, ranging from 0 to 1; not valid areas are indicated with black
Hajnsek et al. (2003). (left) Elbe-Auen test site and (right) Weiherbach test sit in Germany. IEEE copyright.
Furthermore, the fact that the effective roughness is scaled by the wavelength makes moderate
rough bare surfaces to appear very smooth at P-band, implying low backscattering
coefficients often close to the system noise floor. In this case, additive noise becomes a
significant limitation. Thus, single frequency and conventional polarimetry alone seem to be
unable to resolve satisfactorily the problem.
In Hajnsek et al. (2009), the authors discuss the potential of using quad-polarimetric data for
the separation of vegetation and ground scattering components and the estimation of moisture
of the underlying soil. Polarimetric-decomposition techniques were used to decompose the
SafeLand - FP7
Page 162 of 401

Rev. No: 2
Date: 2012-02-15
scattering signature into individual (canonical) scattering components; the inversion of

moisture of the underlying soil by using the ground-related scattering component has been
investigated. The performance of the decomposition approaches has been validated at L-band
using fully polarimetric airborne SAR data acquired in the frame of the AgriSAR campaign
by the E-SAR system of DLR over a time span of four months in summer 2006.
Key step of the procedure is the separation of volume from dihedral and surface contributions:
after this step, soil moisture is retrieved by inverting from the dihedral and surface
components as both depend on the dielectric properties of the ground. The performance of the
soil-moisture estimation has been compared against ground measurements over a whole
vegetation-growing period. The comparison of the estimated soil moisture over the whole
vegetation period demonstrated a large variance of the estimation performance of the dihedral
as well as of the surface derived soil-moisture component. During the vegetation growth for
all the three crop types, the dihedral component has the highest appearance and the best
performance when compared with the ground measurements. However, this changes when the
crop vegetation gets into the mature state. Here, the presence of the surface component
increases, leading to a good soil moisture performance. In summary, it can be stated that,
during the vegetation-growing cycle, the dominance of the scattering mechanisms are
changing, and depending on it, either the dihedral or the surface component is able to provide
reasonable soil-moisture estimates. The approach has been extended in Jagdhuber et al.
(2009) to a multiangular observation which can be obtained only by using airborne systems to
increase the success estimation rate, see Figure 97.
Figure 97 : Exemplarily results from Jagdhuber et al. (2009),(a) the land use, (b) the local incidence angle in
degrees as well as (c-e) the estimated soil moisture in vol.% is displayed for three different incidence angle
constellations (3c: master [one acquisition], 3d: master-perpendicular [two acquisitions], 3e: master-oppositeperpendicular [three acquisitions]). White color represents non-invertible pixels; red frame indicates the
overlapping zone of the three acquisitions (master, opposite, perpendicular). WT=winter triticale, WB=winter
barley, WR=winter rye, WW=winter wheat, SO=summer oat. IEEE Copyright.

SafeLand - FP7
Page 163 of 401

Part D Geophysical Investigations
Rev. No: 2
Date: 2012-02-15
PART D
GEOPHYSICAL INVESTIGATIONS
SECTION 1: Ground-Based Geophysical Investigations
p. 167
SECTION 2: Acoustic and Micro-Seismic Monitoring
p. 195
SECTION 3: Introduction to Offshore Methods
p. 213
SECTION 4: Stakes of the Airborne Geophysics
p. 218
Section 1
- Authors: I. Lecomte1,2, S. Bazin1,3, G. Grandjean4, C. Michoud5,
- Reviewer: D. Jongmans6
1

. NORSAR, Kjeller, Norway. Isabelle.lecomte@norsar.no
3
. Norwegian Geotechnical Institute (NGI), Oslo, Norway.
4
. Bureau de Recherches Gologiques et Minires (BRGM), Paris, France.
5
6
. Laboratoire de Gophysique Interne et de Tectonophysique, Universit Joseph Fourier, Grenoble, France.
2
Lecomte I., Bazin S., Grandjean G., Michoud C., 2010.Ground-Based Geophysical Investigations. In the Deliverable 4.1 of the European
project SAFELAND: Review of Techniques for Landslide Detection, Fast Characterization, Rapid Mapping and Long-Term
Monitoring. Edited in 2010 by Michoud C., Abelln A., Derron M.-H. and Jaboyedoff M. Available at http://www.safeland-fp7.eu
Section 2
- Authors: A. Tonnellier1, J.-P. Malet1
- Reviewer: M. Roth 2,3
1
. Centre National de la Recherche Scientifique (CNRS), Institut de Physique du Globe de Strasbourg (IPGS), France.
alice.tonnellier@unistra.fr
2
3
. NORSAR, Kjeller, Norway.
Tonnellier A., Malet J.-P., 2010. Acoustic and Micro-Seismic Monitoring. In the Deliverable 4.1 of the European project SAFELAND:
Review of Techniques for Landslide Detection, Fast Characterization, Rapid Mapping and Long-Term Monitoring. Edited in 2010
by Michoud C., Abelln A., Derron M.-H. and Jaboyedoff M. Available at http://www.safeland-fp7.eu

SafeLand - FP7
Page 165 of 401
Deliverable 4.1
Section 3
- Author: J.-S. LHeureux1,2, R. Hermanns1,2
- Reviewers: C. Michoud3, M.-H. Derron3
1

. Geological Survey of Norway (NGU), Trondheim, Norway. jean.lheureux@ngu.no
2
3
LHeureux J.-S., Hermanns R., 2010. Introduction to Offshore Methods. In the Deliverable 4.1 of the European project SAFELAND: Review
of Techniques for Landslide Detection, Fast Characterization, Rapid Mapping and Long-Term Monitoring. Edited in 2010 by
Michoud C., Abelln A., Derron M.-H. and Jaboyedoff M. Available at http://www.safeland-fp7.eu
Section 4
- Author: R. Supper1, I. Baron1
- Reviewer: A. A. Pfaffhuber 2,3
1
. Department of Geophysics, Geological Survey of Austria, Vienna, Austria. robert.supper@geologie.ac.at

3
. Norwegian Geotechnical Institute (NGI), Oslo, Norway.
2
Supper R., Baron I., 2010. Stakes of the Airborne Geophysics. In the Deliverable 4.1 of the European project SAFELAND: Review of

SafeLand - FP7
Page 166 of 401

Rev. No: 2
Date: 2012-02-15
1. GROUND-BASED GEOPHYSICAL INVESTIGATIONS

1.1. INTRODUCTION
Ground-based (also called near-surface) geophysical methods have significantly improved the
last 20 years, in terms of method, equipment, and software. The trend is going from 2dimensional (2D) spatial imaging to 3-dimensional (3D) spatial imaging for site assessment,
and further going to time-monitoring modes the so-called 4D approach, time being the 4th
dimension where acquisition is repeated regularly over time to analyze temporal variations
(monitoring mode). Even if geophysical techniques do not directly provide geological and
mechanical properties, but instead some physical parameters of soils and rocks (e.g., density,
acoustic/elastic parameters, resistivity, etc), ground-based geophysics often give fast and
economic information about the subsurface, pending that data are all calibrated with a
minimum of ground proofing (e.g., boreholes whenever possible) and that results are
interpreted in tight cooperation with other geoscientists, e.g., geologists, hydrogeologists,
geotechnicians, glaciologists, etc. Equipment improvement has especially contributed to the
application of geophysics for landslides as the corresponding sites are often more challenging
(e.g., steep and rough terrain, i.e., leading to difficult work conditions) than in more
conventional applications (engineering, environmental, archaeology, etc). For an excellent
and recent review of the use of geophysics for landslides, we highly recommend reading
Jongmans and Garambois (2007), which contains an extensive analysis of published examples
and which was consequently extensively used in the present report. In a slightly wider
application range, i.e., geohazards (landslides are major geohazards), Bouillon (2005) goes in
some more details though keeping it simple - through the geophysical methods used and
lists also key papers, e.g., Hack (2000) about geophysics for slope stability. For more recent
site-specific cases, Maurer et al. (2010) review the use of near-surface geophysics in the
Swiss Alps (e.g., frozen-ground issues for slope stability), and Hunter and al. (2010) present
Canadas experience in the domain, both offshore and onshore. Finally, the geophysical terms
used in the following are according to the dictionary of Sherif (2002).
Before going to field, a geophysicist must get enough information about the site of interest, in
order to select the proper methods and estimate a cost; it is indeed always good practice to
first visit the site and preferably to test a few methods. As the term of landslide refers to a
large variety of mass movements (from very slow slides in soils to rock avalanches), affecting
all soil and rock materials and with a large variety of shapes and volumes, it is crucial to
understand the key parameters needed for their assessment. Ground modifications due to a
landslide are indeed likely to generate changes of physical parameters characterizing the slide
mass.
According to the literature review of Jongmans and Garambois (2007), a geotechnical
appraisal of landslide stability has to consider three issues: (1) the definition of the 3D
geometry of the landslide, in particular failure surfaces; (2) the definition of the
hydrogeological regime; and (3) the detection and characterization of the movement. Except
in very peculiar cases, a landslide generally results in a modification of the morphology and
of the internal structure of the affected ground mass, both in terms of hydrogeological and
physical properties. Surface observations and measurements have therefore to be supported by
reconnaissance at depth with investigation of the slide mass down to the undisturbed rock or
soil. Conventional geotechnical techniques, which mainly include boreholes, penetration tests
SafeLand - FP7
Page 167 of 401
Deliverable 4.1
and trenching, allow a detailed geological description and mechanical characterization of the
material. However, these techniques only give punctual information, at a relatively high cost,
and their use is further limited by the difficulty of drilling onto steep and unstable slopes. On
the contrary, geophysics can cover rather large surfaces and depths for a lesser or similar cost.
The advantages of geophysical techniques are that (1) most are relatively quick and easy to
deploy on slopes, (2) they are non-invasive (except for borehole geophysics) while giving
information on the internal structure of the soil or rock mass, and (3) they allow a large
volume to be investigated. Their main drawbacks are: (1) the decreasing resolution (ability of
the method to detect a body of a given size) with depth, (2) the non-uniqueness of the solution
for a set of data, hence the need for calibration (e.g., boreholes) and (3) the indirect
information they yield (physical parameters instead of geological or geotechnical properties).
It is worth noting that almost all the advantages of geophysical methods correspond to
disadvantages of the geotechnical techniques, and vice-versa, outlining the complementarities
between the two investigation techniques. Geophysics can indeed be first used in a short pilot
study (exploration mode) to 1) test the feasibility of the methods and 2) determine the optimal
locations of geological/geotechnical coring, hence reducing its cost. After ground proofing,
geophysics can in a second stage be used in a more thorough manner to map the relevant
properties, both laterally and vertically. Ideally, several geophysical methods should be
combined as will be discussed later. Finally, both boreholes and geophysics can be re-used for
monitoring of temporal parameter variations, with again the former measurements for
calibration and the latter for spatial extent. However, geophysics is far to be systematically
used. The limitations of most ancient geophysical methods to adequately investigate a 3D
structure, and the problem of linking the measured geophysical parameters to geotechnical
properties have probably made many geotechnical engineers reluctant to use geophysical
methods. We will come back to that issue in a later discussion.
The geophysical methods to apply depend on their adequacy to the problem to solve and on
four controlling factors, which have to be thoroughly considered before any field experiment
(Mc Cann and Foster, 1990). The first controlling factor is the existence of a geophysical
contrast. The presence of a geological, hydrological or mechanical boundary (e.g., the limit of
the sliding mass) does not necessarily imply a variation in terms of geophysical properties.
The second issue is the characteristics of the geophysical method itself, namely the
penetration depth and the resolution. There is usually a trade-off between resolution and
penetration: the deeper the penetration, the poorer the resolution. These limits have to be
accounted for during the design of a geophysical survey. Finally, the performance of
geophysical techniques is strongly dependent on the signal-to-noise ratio. Landslide material
can be highly disturbed and consequently lead to electrical current injection difficulties or
strong seismic wave attenuation (hence the need for preliminary tests as indicated earlier).
Through specific processing, geophysical methods provide the distribution of a physical
parameter at depth at one given surface location (1D), along a line (2D) or for a grid (3D). 1D
information corresponds to a curve function of depth, while 2D and 3D information are
geophysical images (vertical sections or cubes). The relevant parameters are obtained through
an inversion process, i.e., extraction of the searched parameters from measurements assuming
a known model between data and parameters (Menke, 1989). Geophysical inversion (e.g.,
tomography) has dramatically developed during the last twenty years and has the major
advantage to give continuous information of the studied body. But inversion is often a
complex and nonlinear mathematical problem and image interpretation has to be done with a
SafeLand - FP7
Page 168 of 401

Rev. No: 2
Date: 2012-02-15
critical mind, considering the already mentioned drawbacks of geophysical techniques and
additional limits linked to the inversion process. The necessary condition an image (model)
has to fulfil is to explain the data, i.e., the forward modelling of the inverted parameters give
results close enough to the data (the so-called data fit). This is usually assessed by a misfit
error (RMS) which has to be systematically provided with the image (always ask for that!).
However, even if the RMS value is low (e.g., 5% or less), the obtained image may only be
one of many solutions explaining the data due to, e.g., the limited measurement coverage and
to errors on the data (signal-to-noise ratio). Depending on the inversion technique, different
strategies exist to address this problem of non-uniqueness: tests of inversions considering
different starting models for sensitivity analyses, introduction of a-priori information (may
come from other geophysical results or geological/geotechnical information) in the inversion
to constrain the solution, and - ideally - joint inversion of several geophysical data sets. The
second issue is the image smoothness caused by most of the inversion techniques used in
geophysical tomography, resulting in an inability to detect sharp layer interfaces. New
techniques for solving this problem are emerging, using a priori information (Wisn et al.,
2005), regularization for favouring sharp boundaries in the inversion process (Zhdanov, 2002)
or image processing tools such as crest lines extraction process in gradient images (Nguyen et
al., 2005). Finally, most of the existing images are 2D, while a landslide is a 3D phenomenon.
2D images of 3D structures may be affected by strong artefacts which are very hard to detect.
A judicious strategy to tackle this problem, i.e., when 3D acquisition is not possible, is to
perform 2D and 3D forward modelling to evaluate the robustness and reliability of the
obtained image.
After these very general considerations about geophysics and the requirements for landslide
assessment, we follow Jongmans and Garambois (2007) to list the most actual geophysical
approaches fitting for landslides, reproducing here their review table (Table 14) which relates
methods, geophysical parameters, geology, landslide types and applications through several
field examples.
Table 14 (bellow) : Synthesis of geophysical methods used for landslide investigation. Vp and Vs: P-wave and Swave seismic velocity; : electrical resistivity; V: electrical potential; : electrical permittivity; : density; :
average slope gradient. The maximum penetration depth is indicated in brackets. From Jongmans and
Garambois (2007).

SafeLand - FP7
Page 169 of 401
Deliverable 4.1

SafeLand - FP7
Page 170 of 401

Rev. No: 2
Date: 2012-02-15
1.2. METHODS
Though there is a wide range of methods in ground-based/near-surface geophysics, the scope
of the present document does not allow an extended overview and we will therefore
concentrate on a few key techniques, specifically used for landslide investigations and mainly
following Jongmans and Garambois review (2007). For more detailed information, we refer
to key comprehensive books about near-surface geophysics (e.g., Reynolds, 1997; Butler,
2005), or practical and shorter guides for a quicker/easier reading (e.g., Milsom, 2003). We
also highly recommend to check the very good web site of the Swiss Geophysical
Commission (see references) and the Code of Practice of Applied Geophysics published by
the French AGAP association (1992; see references). These two organisms do not deal only
with landslide applications and do list many more geophysical methods, giving both thorough
information about the principles and practical guidelines for good practice in the field.
To help non-geophysicist readers of the present report evaluating the use of geophysics for
landslides, we will first very briefly review the principles of each selected method, discussing
then their applications, advantages and limits with respect to landslide assessment and
monitoring. A very few examples were selected here to illustrate some applications of
geophysics for landslides and for a more detailed analysis of some landslide-related case
studies we refer again - to Jongmans and Garambois (2007). For a first general and simple
overview of geophysical methods, we reproduce below a table from the manuals of the US
Army Corps of Engineers (1995 and 2001; Table 15). This table shows that near-surface
geophysicists try to measure a wide range of ground parameters. We will see in the following
that each method studies a specific type of physical parameters, e.g., elastic properties,
resistivity, etc.

SafeLand - FP7
Page 171 of 401
Deliverable 4.1
Table 15 : Measurements and applications in an engineering context of geophysical methods (US Army Corps of
Engineers, 2001).
1.2.1. Seismic methods

Seismic methods used elastic waves propagating in solid or liquid media; note that sound
waves are elastic waves propagating in the air. Seismic waves correspond to small elastic (i.e.,
non-permanent) displacements. They can be generated by active sources, e.g., sledge hammer,
explosives, and vibrators, or by passive sources, e.g., earthquakes, rockfalls, ocean waves, etc.
Receivers are called geophones (hydrophones in water) and may contain on land up to 3
components (3C; orthogonal system with one vertical and 2 horizontal components), though
vertical geophones are most used. Seismic waves are grouped in two major categories: body
waves and surface waves (Figure 98).

SafeLand - FP7
Page 172 of 401

Rev. No: 2
Date: 2012-02-15
Figure 98 : Elastic waves used in seismic methods. Left) Body waves (adapted from Bolt, 1982): P-waves (top)
shake the ground along the propagation direction (black arrow), while S-waves (bottom) shake transverse to
that direction (black arrow; here SV-wave illustrated). Right) Surface waves (Wikipedia): two types of surface
waves are illustrated: Rayleigh waves (top) and Love waves (below). Their vibration pattern is more complex
than for body waves (see arrows on figures).
Body waves are waves passing through layers and are either compressional waves (P-wave; P
standing historically for primary because the fastest of the body waves) or shear waves (Swave; S standing historically for secondary because slower than P-waves). P-waves vibrate
along the direction of propagation while S-waves vibrate perpendicular to that direction. Swaves are often decomposed in 2 components, i.e., an SV-wave (V for vertical) with a
vibration in the vertical plane, and an SH-wave (H for horizontal) with a vibration in the
horizontal plane, though this is a simplification not really valid in complex 3D media. Ideally,
a user of seismic methods should indeed record seismic wavefields with 3C geophones to
record all wave modes but this is seldom done. The P-wave velocity is function of the
compression (bulk) modulus, K, the shear modulus, G, and the density, , while the S-wave
velocity only depends on shear modulus and density, and is lower than the P-wave velocity.
S-waves do not propagate in liquid media while P-waves do (acoustics). Examples of P- and
S-wave velocities are given in Table 16. In seismics, resolution is defined as a fraction of the
wavelength (1/4 is commonly used), which is defined by velocity divided by frequency.
Therefore low velocities (and/or high frequencies) contribute to better resolution. Note that
the velocity of both body waves may be anisotropic, i.e., even in homogeneous media the
velocity may depend of the wave direction due to, e.g., fine layering of a rock formation,
fractures, etc. Though such effect is studied and used in other geophysical applications, the
authors are not aware of much use of anisotropy in relation to landslide assessment. Dasios et
al. (1999) explain 1) that S-waves can offer better vertical resolution than compressional
waves due to their lower velocity; 2) they are little affected by changes in fluid saturation, in
contrary to P-waves, and are thus more sensitive to lithology, making them more suitable for
engineering purposes; and 3) S-waves are less absorbed than P-waves, at least in partially
saturated sands, boulder clay and gas-saturated sediments. Though the use of P-waves is still
largely dominating in near-surface geophysics, we will see successful examples using Swaves. As for optical laws, elastic body waves have different propagation modes:
transmission with refraction (bending of the waves following Snells law) at discontinuities,
reflection at the same discontinuities, and diffraction. Discontinuities generating reflection
and refraction of seismic waves are due to a change in elastic impedance, often called Z=V,
SafeLand - FP7
Page 173 of 401
Deliverable 4.1
which is the product of density and velocity V. The reflected wave generated at a
discontinuity between a layer of impedance Z1 and a layer of impedance Z2 has a strength
function of the reflection coefficient R which is (Z2-Z1)/(Z1+Z2). Note that such elastic
discontinuities may not be related to geological changes (e.g., water table crossing geological
layers) and changes in lithology may not produce impedance contrasts (Table 17). In such
cases, other geophysical data, well logs and geological information may help identifying the
structures imaged by seismic body waves.
Table 16 : P- and S-wave velocity, and density for soil and rock examples. Modified from Lavergne (1986).
Material
P-velocity Vp (m/s)
S-velocity Vs (m/s)
density (g/cm3)
soil
100-600
100-300
1.7-2.4
dry sands
200-1200
100-500
1.5-1.7
wet sands
1500-4000
400-1200
1.9-2.1
glacial maraine
1500-2700
200-700
1.9-2.7
clay
1000-2500
200-800
2.0-2.4
marl/shale
2000-4100
750-1500
2.0-2.7
gneiss
3500-7600
2200-3600
2.7-2.8
sandstone
1400-4500
1200-2800
2.1-2.4
limestone
2800-7000
2000-3300
2.4-2.7
chalk
2300-2600
1100-1300
1.8-2.3
rock salt
4000-5500
2500-3100
2.1-2.3
anhydrite
3500-5500
2200-3100
2.9-3.0
dolomite
2500-6500
1900-3600
2.5-2.9
granite
4500-6200
2500-3300
2.5-2.7
basalt
5000-6500
2800-3400
2.7-3.1
coal
2200-2700
1000-1400
1.3-1.8
ice
3400-4000
1700-1900
0.9
water
1450-1500
1.
air
315-350
0.001

SafeLand - FP7
Page 174 of 401

Rev. No: 2
Date: 2012-02-15
Table 17 : Examples of reflection coefficients for some soil and rock mass materials. Note that the reflection
coefficient is given here in absolute values. It is however a signed-value, which is positive when the elastic
impedance Z increases across the reflection boundary (Z1<Z2), while it is negative when Z decreases (Z1>Z2).
From Hack (2000).
Surface waves travel along/near-to a surface, e.g., the earths surface or at sea bottom. The
amplitudes of surface waves decrease rapidly with depth and their velocities are smaller than
S-wave velocities. In seismic exploration, all surface waves are usually referred as ground
roll, i.e., a high-energy seismic noise a geophysicist wants to get rid of when interested in
reflection and transmission of body waves. Though surface waves have proved of great use
for a long time in other domains, such as in seismology or engineering, the last decade has
seen a sudden increase in their play in near-surface application, including for landslides, as
will be described later. More specific terms exist for surface waves, e.g., Rayleigh and Love
waves as illustrated in Figure 98.
1.2.1.1. Active seismic

SEISMIC REFRACTION
Method: This method is based on the interpretation of first-arrival traveltimes, which are
picked (often manually) on seismic records, the resulting curves of traveltimes as function of
distance being called hodochrons. A strong assumption in seismic refraction is that the
velocity increases with depth. Figure 99 illustrates the path of the seismic waves of concern
with source and receivers at the surface. The elastic waves of interest (P or S depending on
the source and receiver type, though the use of P-waves is largely dominating) are
propagating in a transmission mode, either directly through the top layer or as critically
refracted, so-called headwave, recorded at the surface. The usual source for P-waves is a
sledge hammer but explosives may be needed for deeper penetration. Shear-wave vibrators
are also used but are not as portable as a sledge hammer and S-wave refraction is not that
common. To take into account dipping layers with respect to the surface, a minimum of two
source locations is needed, i.e., at a certain distance from each extremity of the recording
array, the two resulting hodochrons being interpreted together (Figure 100).

SafeLand - FP7
Page 175 of 401
Deliverable 4.1
Figure 99 : Refraction seismic. a) acquisition (from www.fugroinstruments.com) and b) corresponding seismic

wavepaths (figure taken from US Army Corps of Engineers, 1995).
Applications: Refraction seismic in its traditional form, i.e., processing and interpretation of
picked first-arrival traveltimes to map refractors and estimate a mean velocity in the layers
in between, is still widely used, especially to map clear elastic impedance contracts such as
ground-water table and sediments/bedrock. The method has proved to be applicable for
landslide investigation, as both P- and S-wave velocities are generally lower in the landslide
body than in the unaffected ground below (critically-refracted wave generated at the contact).
McCann and Forster (1990) document several case studies showing the use of seismic
refraction for locating the undisturbed bedrock below landslides. In recent studies, traveltime
data have been interpreted using delay methods (Kearey et al., 2002), like the plus-minus
technique or the Generalized Reciprocal Method (GRM), which allow the mapping of an
undulating refractor.
Figure 100 : Refraction seismic on a slope with interface non-parallel to the surface. Acquisition of a direct and
reverse recording (a) with the corresponding hodochrons (b).
Limits: Refraction seismic requires an increase of elastic impedance with depth so it does not
detect low-velocity layers. In the case studies analysed by Jongmans and Garambois (2007),
the refraction method was limited to a penetration depth from a few meters to 30 meters. This
shallow penetration depth results from the method itself, which requires a relatively long
profile (3 to 5 times the penetration depth as a rule of thumb) and from the wave attenuation
in the highly disturbed landslide material. Sledge hammer as a source is not sufficient for long
profiles and explosives are necessary, but their use is not always allowed, especially on
unstable slopes and sensitive sites (e.g., quick clays, see later).

SafeLand - FP7
Page 176 of 401

Rev. No: 2
Date: 2012-02-15
SEISMIC TOMOGAPHY
Method: Seismic tomography uses first-arrival traveltimes (like refraction) from source to
receivers in order to invert for the propagation velocity in 2D or 3D. It is therefore important
to acquire as many source/receiver combinations as possible, i.e., with sources both along the
recording array and outside on both sides, with different source-receiver distances to cover
various depths. This method is not only used for ground-located sources and geophones, but
may also be used in boreholes (cross-hole tomography) and in the ocean (air gun as sources,
and ocean-bottom seismometers or hydrophones as receivers). The principle of tomography is
as in medicine for human bodies: waves travelling across a media accumulate information
about a certain parameter and by combining different source-receiver combinations and
mapping back that parameter along certain trajectory (e.g., raypaths), an inversion procedure
will provide a 2D section or a 3D cube. In seismic or traveltime tomography, traveltime
measurements are used to map the velocity model. In a similar manner, we will see later that
in resistivity tomography, apparent resistivities measured at the earths surface are used to
map actual resistivities in the ground. In contrary to standard refraction seismic, which maps
discontinuities, tomography can compute more or less a smooth velocity model. Therefore,
the two methods should be tried together in order to better constrain the resulting model with
both clear discontinuities and smoothed velocity fields in between.
Applications: For landslide investigation, the technique was used in rock conditions
(Jongmans et al., 2000; Mric et al., 2005) and showed a significant decrease of P-wave
velocity values (division by at least a factor 2) in a slide-prone or unstable mass. Mric et al.
(2005) performed a 300-m long seismic profile across the western limit of the large
Schilienne landslide (French Alps) affecting micaschists. Out of the unstable mass, the
image showed a strong vertical variation of P-wave velocity, with values ranging from 500
m/s at the surface to 4000 m/s at 25 m depth (intact bedrock). The same profile also
pinpointed a significant lateral velocity change from 4000 m/s to 2000 m/s delineating the
landslide limit.
Limits: The main limits are the efforts (i.e., cost - money and time - and heavy logistics)
required to acquire enough sources/receivers for a good coverage of the structures, laterally
and vertically, and the rather smooth model finally obtained, though this is not a limit intrinsic
to seismic but which will also be found in, e.g., resistivity mapping. Ideally, both P- and Swave velocity should be mapped but S-sources of sufficient strength are not yet available for
difficult ground conditions, though S-wave vibrators start to work well on easier grounds (see
under seismic reflection).
SHALLOW REFLECTION SEISMIC

Method: Reflection seismic deals specifically with reflected waves generated at elastic
impedance contrasts (Figure 101). In contrary to seismic refraction and tomography, the
whole recorded signal is considered but some wave modes and noise have to be removed
prior to imaging structures with reflected waves only. Direct and refracted waves as well as
surface waves are among the seismic noise to remove, even though they may be useful for
SafeLand - FP7
Page 177 of 401
Deliverable 4.1
other methods. Reflection seismic is a key technique in oil and gas exploration which has
greatly contributed to the development of equipments, numerous processing methods and
software. The acquisition is usually carried out in a roll-along mode, i.e., the receiver lines
are moved regularly after shooting at specific locations around them. This means in practice
heavy field work though the quite recent use of streamers (geophones mounted at fixed
intervals along a towed system with no need to plant them in the ground) does facilitate the
acquisition. Though P-waves are mostly used for reflection seismic, S-waves surveys start to
emerge, especially for SH-waves (source with vibration perpendicular to the acquisition line
and horizontal geophones parallel to the vibration direction; e.g., Pugin et al., 2004).
Figure 101 : Principles of seismic reflection acquired in Common Source Gather, imaging two sliding surfaces
in a complex translational landslide (courtesy of C. Michoud).
Applications: High resolution seismic reflection has been seldom used for landslide
investigation (Bruno and Marillier, 2000; Ferrucci et al., 2000; Bichler et al., 2004). The
major interests of seismic reflection profiling are its potential for imaging the geometry of the
landslide structure, such as the internal bedding or the rupture surface(s), and the
robustness/diversity of processing tools compared to tomography. In a recent case study,
Polom et al. (2010) describes SH-wave 2D profiling on ground in the harbour of Trondheim,
Norway. The very high quality of the seismic profiles (Figure 102) with a 0.5-1.0 m
resolution obtained down to 100-150 m (bedrock) is now helping geologists and
geotechnicians detecting weak layers, already identified offshore, and assessing the risk for
submarine slides which would significantly damage the harbour infrastructures and the nearby
train station.
Limits: This method requires an even bigger effort to deploy the geophone layouts than
seismic tomography, particularly in the conditions of rugged topography, making the
acquisition time consuming and costly. The success of shallow seismic reflection requires a
good signal-to-noise ratio and the recording of high frequency waves to reach the desired
resolution. These two conditions may be difficult to fulfil on landslides where the ground is
strongly disturbed and heterogeneous, affecting the geophone-soil coupling, attenuating the
seismic waves and generating scattered waves; in such conditions 3D reflection seismic may
be necessary for a proper imaging but the cost is often too high. In addition, the processing of
these data really requires expert user and takes time.

SafeLand - FP7
Page 178 of 401

Rev. No: 2
Date: 2012-02-15
Figure 102 : SH-wave reflection profile (poststack depth migrated) with superimposed geological interpretation,
Trondheim harbour, Norway. From Polom et al. (2010).
SURFACE WAVES
Method: Surface waves may show appreciable dispersion in the presence of near-surface
velocity layering, i.e., different frequencies of the seismic signal travel with different
velocities, while dispersion of body waves is usually very small. During recent years, spectral
analysis of surface waves (SASW) has received increasing attention in the geophysical
community (Park et al., 1999; Socco and Jongmans, 2004), essentially because it offers a noninvasive means of evaluating the soil shear modulus at depth. In this approach, two or several
seismic signals are analyzed in the Fourier domain to interpret phase differences related to
Rayleigh-wave (Figure 98) dispersion effects into S-wave velocity variations in depth
(vertical sounding). Later, additional developments were performed to use a multichannel
recording system to improve signal-to-noise ratio by stacking signals without increasing
acquisition times, leading to so-called multichannel analysis of surface waves, MASW (Park
et al., 2001). The experiment consists of constructing a seismic antenna, composed of a
seismic source and several sensors spaced regularly along the seismic line, then recording the
soil particle velocity in time after the source has been activated. An integral transformation
converts the time-domain waveform data into a phase velocityfrequency domain called the
dispersion image (Figure 103). Because the resulting dispersion image is obtained from the
stack of several phase-transformed signals recorded by the sensors, aliasing artefacts are
reduced, increasing the resolution of the final image. Consequently, this improvement in
resolution makes it possible to distinguish each separate propagation mode. Inverting
dispersion properties into shear-wave velocity distributions is generally solved in the 1D
approximation (Herrmann, 2002). Another extension of the method, developed by Xia et al.
(1999), consists of obtaining a 2D S-wave velocity section through the shallowest layers (Lin
and Chang, 2004; Grandjean, 2006) by interpolating contiguous 1D S-wave velocity vertical
profiles.
SafeLand - FP7
Page 179 of 401
Deliverable 4.1
Applications: A wide range of applications have demonstrated the use of this method,
particularly for landslides studies (Grandjean and Bitri, 2006; Meric et al., 2007). When
landslides are the places of highly heterogeneous materials, an adaptation of MASW can be
proposed by extending the original method with a summation term, which sums over local
dispersion images computed for different receiver windows gathers. This leads to the
2MASW (Multifold MASW) method tested by Grandjean and Bitri (2006) on the SuperSauze landslide (Figure 103 and Figure 104).
Limits: Although MASW methods can provide robust results for imaging landslide
structures, it suffers from some limitations. First, the method is based on 1D approximation
(no lateral variations) so that 2D complex structures cannot be investigated correctly (Foti,
2000). In addition, the quality of dispersion diagrams are strongly related to the complexity of
Rayleigh wave propagation; indeed, diffractions, multiple modes and all seismic patterns
caused by sharp velocity variations lead to decrease in dispersion image quality, and thus the
quality of inverted models. Regardless of the complexity of the velocity structure, the surfacewave inversion cannot be solved directly, requiring an optimization technique to find the most
probable solution in a pool of infinite candidates. This technique can be a deterministic
approach, random, or combination of both. The final solution is however not necessarily the
exact one and it is highly recommended to display several solutions to better appreciate the
sensitivity of the inversion.
Figure 103 : a) seismic receiver-window gathers (left), dispersion image and resulting S-wave velocity model
after inversion in blue and related correlation value in gray (right) indicating the reliability of this model.

SafeLand - FP7
Page 180 of 401

Rev. No: 2
Date: 2012-02-15
Figure 104 : S-wave velocity field across the Super-Sauze landslide (France) obtained by 2D interpolation of all
S-wave velocity models estimated along the seismic section. The landslide appears in green (low velocity
values), in the centre part of the section.
1.2.1.2. Passive seismic measurements

1D AMBIENT- NOISE MEASUREMENT
Method: Passive seismic noise measurements are commonly used in earthquake engineering
for determining the geometry and S-wave velocity values of soil layers overlying bedrock
(Jongmans and Garambois, 2007). The single-station method (also called the H/V method, as
short for Horizontal-to-Vertical spectral ratio method) consists in calculating the horizontalto-vertical spectral ratio of noise records measured on 3C geophones and allows the resonance
frequency of the soft layer to be determined (Nakamura, 1989). For a single homogeneous
soft layer, this fundamental frequency is given by f = Vs/4h where Vs is the soft layer S-wave
velocity and h is the layer thickness (Figure 105). Knowing an estimate of Vs allows the
thickness of the soft layer to be calculated. The three main assumptions behind the method are
that: 1) seismic noise is composed of surface waves; 2) the structure of the soil is 1D and; 3)
the Vs contrast is large enough to generate a clear frequency peak.
Figure 105 : H/V curve obtained in an elongated alluvial valley. The fundamental frequency of the site may be
reliably estimated at 0.7 Hz. Figure from SESAME (2004).
Applications: As failure surfaces may generate S-wave velocity contrasts, the method can
theoretically directly detect landslide sliding surfaces. It was used on three landslides
affecting clayey or marly terrains in the Southern Apennines (Gallipoli et al., 2000) and in the
French Alps (Mric et al., 2007). The fundamental frequency was derived from H/V curves

SafeLand - FP7
Page 181 of 401
Deliverable 4.1
and used for deriving an estimate of the rupture surface depth. This easy-to-perform survey
opens interesting perspectives for 3D investigation.
Limits: Difficulties appear in heterogeneous soils with strong 2D or 3D effects disturbing the
propagation of surface waves (diffraction and diffusion effects) and if various frequencies can
be picked due to the presence of unexpected layers or harmonic noise.
3D PASSIVE ARRAY
Seismic arrays can be installed near landslides to record local events (Figure 106). Rockfalls
and micro-seismicity can be distinguished and classified from their signal characteristics.
Triggered and continuous data can be used for this type of seismic monitoring. An example of
a successful monitoring network installed since 2005 at knes, Norway, is presented in the
case study n12.
More complex techniques using seismic noise arrays instead of one receiver (as the H/V
method) can be used to derive consistent S-wave velocity profiles versus depth on soil
landslides (e.g., Renalier et al., 2010, in press). A dedicated chapter on acoustic and microseismic monitoring is presented in the section 2 of this part (p. 195).
Figure 106 : The Schilienne rockslide, in the French Alps, about 15 km south west of Grenoble is monitored by
a seismic network: vertical sensors (circles) and 3-components seismometers (triangles). Also shown is the
location of benchmarks of a displacement network (diamonds) within the most active zone contoured in orange.
Figure from Helmstetter and Garambois (2010)
1.2.2. Electrical methods

Electrical measurements are made at the surface of the Earth to investigate electrical
resistivity (inverse of electrical conductivity) conditions in an area. The basic measuring
equipment consists of two current electrodes and two measuring electrodes, a DC current
source, and a measuring device. The two current electrodes are inserted in the ground and
SafeLand - FP7
Page 182 of 401

Rev. No: 2
Date: 2012-02-15
between the electrodes, a DC potential is maintained that causes a DC current to flow through
the ground. The resulting potential differences are measured at the surface. Anomalous
conditions or heterogeneities within the ground, such as better or poorer conducting layers,
are inferred from the fact that they deflect the current and distort the normal potentials. For
vertical electric sounding (VES), the spacing between the electrodes is increased with regular
steps while the centre of the array is fixed. For horizontal profiling, the array of potential
electrodes and current electrodes is moved over the surface and electrical resistivity
tomography (ERT) can be performed. ERT vertical profiles along 2D lines or ERT cubes
(3D) can be obtained.
1.2.2.1. Resistivity
Method: The simplest approach to the theoretical study of current flow in the Earth is to
consider the case of a completely homogeneous isotropic earth layer of uniform resistivity.
The potential difference V across a layer of length l and resistance R, through which a
current, I, is flowing is V=RI (Ohms law). If we consider a homogeneous half space
bounded by the ground surface and let a current of strength I enter at point C1 on the ground
surface, this current will flow away radically from the point of entry (Figure 107). By
knowing the input current, the measured voltage V and the geometry of the electrode array,
the resistance R can be calculated. The property of the electrical resistance of a material is
usually expressed in terms of resistivity (m) or its inverse, the conductivity (mS/m).
As a rule, the more porous or fissured a rock and the larger its ground water salinity, the
higher is the conductivity, the lower the resistivity. There is no general correlation of the
lithology with resistivity. Nevertheless, a broad classification is possible according to which
clays and shales, sands and gravel, compact sandstones and limestones, and unaltered
crystalline rocks stand in order of increasing resistivity. Figure 108 shows the approximate
resistivity ranges of earth materials.
Figure 107 : Basic principle of electrical resistivity imaging. It is possible to determine earth resistivity by
driving a direct current I (DC signal) into the ground and measuring the resulting potentials (voltages) created
in the earth. Figure from Kndel et al. (2007).
Standard resistivity systems use electrodes planted in the ground and linked together with
electrical cables (Loke, 1999). Such an installation can be cumbersome on complex grounds
but when everything is coupled, the measurements are automatic, in various electrode
configuration modes (dipole-dipole, Wenner, Schlumberger, gradient, etc) chosen by the user
at the electronic console. This may take a couple of hours depending of the size of the array

SafeLand - FP7
Page 183 of 401
Deliverable 4.1
and the selected modes. In areas where grounded dipole measurements may be difficult
(frozen ground, etc), an alternative system is a towed one working only in dipole-dipole mode
and based on a capacitively-coupled resistivity meter. A coaxial-cable array with transmitter
and receiver sections is pulled along the ground either by a single person or attached to an allterrain vehicle. Such towed systems work best on resistive ground and for small-tointermediate depths compare to traditional electrode-based system. Data collection is faster
than with conventional DC resistivity, though a line must be repeated for increasing distances
between emitter and receiver in order to increase the penetration depth. In a grid acquisition
mode (several parallel lines) a quick surveying may give an idea of the lateral variability of
the ground structures. It could be advantageously used in an exploration mode in order to preinvestigate a site and get a first idea of the resistivity and variability of ground structures. But
conventional systems are best for reliable measurements.
Figure 108 : Electric resistivity tomography (ERT). a) Electrodes planted in the ground are linked together with
electrical cables to an electronic console (courtesy of S. Bazin). b). An alternative system can be towed (picture
from www.geometrics.com)
Figure 109 : Typical ranges of electrical resistivities of earth materials after Palacky (1987).

SafeLand - FP7
Page 184 of 401

Rev. No: 2
Date: 2012-02-15
Figure 110 : Examples of ERT profiles in rock slide. Left) Alpine site; the potential sliding plan is in dashed line
(Borle and Jaboyedoff, 2007). Right) knes site, Norway; the potential sliding plane is in dashed line and
correspond to observed GPR reflectors (Rnning et al., 2007).
Applications: For landslide investigations, a typical application of ERT is the location of the
sliding plan, identifying layers per resistivity contrast (Figure 110).We will also illustrate in
the following the use of the resistivity method for 1) quick-clay mapping in Scandinavia and
2) landslide monitoring, this in addition to the detection of sliding plans illustrated above.
Note that the use of resistivity tomography has exploded the last year, also for landslides,
thanks to significant hardware and software developments, and many other case studies can
be found (see, Jongmans and Garambois, 2007).
Figure 111 : (Top) 2D ERT profile, Buvika site, Norway, with superimposed geotechnical drilling results and
(bottom) interpretation with respect to quick/unleached clay. (from Solberg et al., 2008).
Quick-clay mapping: Marine clays in Norway and in Sweden were deposited as the ice cap
over Scandinavia retreated some 9000 years ago (Rankka et al., 2004). The clay particles were
deposited in sea water creating an open-grain skeleton with high water content. As the
isostatic upheaval progressed, the marine clays have been leached by rain water and fresh
SafeLand - FP7
Page 185 of 401
Deliverable 4.1
groundwater. Pockets/layers of quick clay above/below unleached clay often occur. The
bonds between the clay minerals decrease as the salt content decreases, causing a weaker
grain structure; eventually the clay becomes unstable (quick). If the quick clay is disturbed,
the grain structure collapses and the clay minerals float in their own pore water. Leaching
reduces the salt content in the pore water (change in electrical properties). Resistivity
measurements are thus very popular for discriminating between unleached and quick clay,
though their use should always be controlled by geotechnical drilling at a few key locations
due to large variability in salt content (Rankka et al., 2004; Solberg et al., 2008).
Landslide monitoring: Time-lapse resistivity imaging (4D ERT) can also show changes
associated with seasonal temperature variation, moisture content and ground movement
(Figure 112). Note that near-surface changes in resistivity caused by moisture content can be
masked by temperature effects so the latter need to be monitored. In addition, initial electrode
spacing can be modified by the ground movement and cause artefact in the modelled
resistivity. Hence, the geometry of the monitoring system should also be monitored
(Wilkinson et al., 2010).
Figure 112 : Time-lapse resistivity results from ALERT monitoring system over an active landslide (Chambers et
al., 2009;). a) August 2008, b) February 2009, and c) resulting differential resistivity images.
Limits: The smoothness constraint in 2D resistivity inversion needs to be stressed, as in

seismics: it tends to smear out sharp boundaries such as the basement topography. It is also
always difficult to foresee the resolution of a resistivity survey because it depends on several
variables: signal-to-noise ratio, topography, resistivity contrast of the materials, electrode
contact resistance, etc. Synthetic tests using the survey parameters/geometry are often
necessary to illustrate the lateral resolution of a survey and should be provided with the
results.
1.2.2.2. Self Potential (SP)
Method: Self-Potential surveys are conducted by measuring natural electrical potential
difference between pairs of electrodes connected to a high impedance voltmeter. These

SafeLand - FP7
Page 186 of 401

Rev. No: 2
Date: 2012-02-15
natural fields represent the ground surface electric field signature of various charging
mechanisms (electrokinetic, thermoelectric, electrochemical, cultural activity) occurring at
depth. In absence of electrochemical processes and large telluric current, electrokinetic
phenomena, describing the generation of electric fields by fluid flows, is the main source of
the recorded electric field. Groundwater and associated flows contained in any landslide body
play a major role in slope stability. The level of groundwater determines the supporting
hydrostatic pressure, which, together with hydrodynamic pressure of seepage, are factors
decreasing the landslide stability. Imaging water level and water flows within the subsurface
at a large scale, as well as their fluctuations over time is a challenging problem, which
resulted in specific research on hydrogeophysics (Rubin and Hubbard, 2005). Only few
hydrogeophysical methods were applied on landslides, except those conducted with the SelfPotential method, which is easy to deploy and monitor.
Applications: Bruno and Marillier (2000) measured an SP profile on the Boup landslide
and observed that high positive SP values (40 to 120 mV) coincide with the boundary
between the stable ground and the landslide material and interpret them as the electrical
signature of resurgent groundwater flow. Comparable large and stable-over-time positive SP
anomalies (up to 350 mV) were acquired by Mric et al. (2005) across the Schilienne
landslide. Although they noted that the shape of the SP data was highly correlated with the
displacement rate curve, the authors did not conclude whether the source of this anomaly was
electrokinecally due to a deep main water flow nearly parallel to the surface or
electrochemically due to the geological structure of the movement (fractures, lead-zinc and
quartz veins). However, large time varying negative anomalies on the edge of the landslide
were attributed to fluid flow variations within major faults and fractures.
Lapenna et al. (2003) presented two SP maps carried out at different climatic conditions on
the Giarossa landslide. They assume the positive and negative anomalies within the
landslide to be due to movements of underground water from the source zone to the
accumulation zone within the landslide body. Further, SP changes over time were explained
by the lowering of the water level inside the landslide body after the dry summer period. To
be more quantitative, they also present SP tomography (Patella, 1997) showing lateral
boundaries of the landslide as well as geological heterogeneities. Lapenna et al. (2005) also
presented an SP map of the Varco dIzzo landslide which they interpret qualitatively in
term of water infiltration and charge accumulation in different zones of the landslide. In the
future, increasing number of SP monitoring experiments using networks as well as
improvements in numerical simulations and specific signal processing techniques (Gibert and
Pessel, 2001; Sailhac and Marquis, 2001) should help understanding the acquired data and
improve hydrological information within landslides.
Limits: The SP source ambiguity and the lack of quantitative interpretation on the fluid
source (depth, extension) are the main limitations of the method, which is still poorly used on
landslides.
1.2.2.3. Induced Polarization (IP)
Method: When a current applied to electrodes is switched off, the voltage between the
potential electrodes does not drop to zero immediately (Figure 113). The ground thus acts as a
capacitor and stores electrical charges, i.e., becomes electrically polarized. The measurement

SafeLand - FP7
Page 187 of 401
Deliverable 4.1
of the decaying voltage over a certain time interval is known as time-domain Induced
Polarization (IP) surveying (Keary et al., 2002). The measured parameter is the chargeability,
M, defined as the normalized area A beneath the decay curve over a certain time interval. The
IP phenomena occur particularly in presence of metallic (e.g. pyrite, magnetite, native metal,
etc.), clay, some coals or graphite particles (US Army Corps of Engineers, 1995).
Figure 113 : The phenomenon of induced polarization. At time t0 the current is switched off and the measured
potential difference, after an initial large drop, decays gradually to zero. A represents the normalized area
under the decay curve for a time interval. Figure from Keary et al. (2002).
Applications: The IP method can be used to search for disseminated ores, clay minerals,
pollution, and groundwater. Even if IP technique can sometimes be used in landslide
investigations for specific purposes (Marescot et al., 2008), it is more confidential than
resistivity or SP methods. But the IP signal is easily acquired when running resistivity
measurements in the field so it is worth recording it as additional information.
Limits: IP measurements involve the monitoring of the decaying voltage after the current is
switched off, but it cannot distinguish between two polarization effects: the membrane
polarization (due to electrolytic flow in the pore fluid) and the electrode polarization (due to
metallic minerals present in a rock). The signal-to-noise ratio of chargeability measurements
is usually not as good as for resistivity measurements. Quantitative interpretation is
considerably more complex than for the resistivity method. Much IP interpretation is only
qualitative: simple parameters of the anomalies, such as sharpness, spatial distribution and
amplitude may be used to estimate the location, lateral extend, dip and depth of the
anomalous zone.
1.2.3. Electromagnetic methods
Method: Under electromagnetic methods, we consider here very-low frequency (from 800 Hz
to 10 kHz) electromagnetic approaches working in a diffusion model in contrary to much
higher frequency methods being in wave-propagation mode, as is the case for Ground
Penetrating Radar (see next method). The low-frequency electromagnetic methods are based
on the transmission of an electromagnetic field from a transmission coil (Hack, 2000). This
transmission field (the primary field) will cause a secondary induced field in the materials in
the sub-surface. A receiver coil receives the primary electromagnetic field together with the
secondary field. The measuring equipment allows for comparison of amplitude and phase
shift of the primary and secondary fields. The intensity of the secondary field depends on the
conductivity of the materials in the sub-surface. The form of the coils and the distance
between the coils depend on the frequencies used and the required depth of the investigation.
This also allows for two types of investigation: vertical profiling or also called depth

SafeLand - FP7
Page 188 of 401

Rev. No: 2
Date: 2012-02-15
sounding, and horizontal profiling, as in resistivity methods. In vertical profiling, the distance
between the coils is increased with regular steps and consequently the received signal is more
influenced by deeper buried materials. In horizontal profiling, the distance is kept constant but
the whole array of coils is moved and at regular distances, measurements are made. As ruleof-thumb, the penetration depth is not more than about half the spacing between the coils.
Vertical profiling or depth sounding can also be achieved by using different frequencies at the
same location. Deeper buried materials will have less influence on higher frequency
transmissions and vice versa. Low frequency EM surveys are very simple to do, fast, and the
equipment is easy to operate. However, the method will virtually never be able to determine
boundaries with enough accuracy as the resolution is low. Nonetheless the method works very
well for, e.g., determining the extent of a (thick) clay-filled discontinuity in limestone or the
presence of rock boulders in clay or sand.
Applications: As shown in Table 14, electromagnetic (EM) methods were recently used by
several authors for landslide investigation, mainly for determining the geometrical limits of
the unstable mass. Except the work of Schmutz et al. (2000) who used TEM (Transient
Electromagnetic Method) jointly with VES, EM measurements (Mric et al., 2005; Bruno and
Marillier, 2000; Mauritsh et al., 2000) were usually performed in the frequency domain with
two horizontal loops and a ground conductivity meter (Geonics EM 34 or EM31). The
method, which yields a single apparent electrical resistivity value, allows quick profiling or
mapping (Reynolds, 1997). Penetration depth depends on the coil separation (10 m, 20 m or
40 m for the EM34) and ranges from a few meters to a few tens of meters. Mric et al. (2005)
and Bruno and Marillier (2000) pointed out a significant variation of apparent resistivity at the
limit between the landslide and the stable ground. In rock landslides, Bruno and Marillier
(2000) and Mauritsch et al. (2000) interpreted electromagnetic data acquired with different
modes and coil separations, assuming a two-layer model (moving mass above stable ground).
They found a relatively good agreement between the bedrock depths derived from
electromagnetic interpretation and seismic results.
Limits: All authors who applied EM methods for landslide investigations stressed out that the
results had to be combined with other geophysical techniques for a proper interpretation.
1.2.4. Ground Penetrating Radar
Method: Ground Penetrating Radar (GPR) uses electromagnetic waves reflected on
discontinuities (dielectric contrasts) after emission from radar antennas positioned along the
earths surface. These discontinuities correspond to changes of electromagnetic impedance
= c where is the magnetic permeability and c = (1/)1/2 is the velocity of electromagnetic
waves, being the electric permittivity. It is rather similar in application and processing to
seismic reflection using elastic waves reflected at elastic impedance (Z =V) contrasts. But in
contrary to seismic where velocity V roughly increases with depth, the electromagnetic wave
velocity c decreases in general with depth with the highest velocity being in the air (0.3 m/ns;
Table 18). The frequency range in ground investigations is from 10 Mhz to 1000 MHz. As in
seismic, the higher the frequency, the better the resolution, but the lower is the penetration
depth. In addition, the penetration capacity is better in resistive materials (e.g., granite, dry
sand, ice) than in conductive layers (e.g., clay, saturated soils) (Table 18). GPR measurements
can also be carried out in boreholes and used in transmission mode for tomography as in
seismic and resistivity.

SafeLand - FP7
Page 189 of 401
Deliverable 4.1
Table 18 : Relative electric permittivity (called here dieletric constant), velocity and attenuation of
electromagnetic waves in different material (From Hack, 2000). The relative electric permittivity is the ratio
between the electric permittivity of the considered material and the electric constant 0 (electric permittivity of
vacuum).
GPR is a very popular method in near-surface geophysics, as seen during the last 10-15 years,
and is widely used by geoscientists thanks to light and portable equipments and the easy
processing. Beside the considerable development of equipment and software, GPR is popular
because of its high resolution (from a few cm to a few m, depending the soil and antennas), its
depth of penetration in resistive materials (e.g., ice on glaciers), and its sensitivity to
electromagnetic contrasts, especially due to water content. In conventional systems, both the
emitting and receiving antennas are moved keeping a fixed distance (offset) between them,
this in a profiling mode (Figure 114). Though this acquisition mode allows the quick
visualization on field of ground reflectors and diffractors, the intrinsic lack of velocity
information prevents proper depth conversion (users usually assumes a mean velocity of 0.1
m/ns to get a rough estimate of depth on field, or use diffractors - if any - to estimate an
equivalent velocity). It is therefore highly recommended to acquire a few Common Mid-Point
(CMP) gathers at key locations, a technique inherited from seismic acquisition/processing and
assuming locally horizontal reflectors. This is done by moving away emitter and receiver
from a given location, thus increasing the offset step by step. In GPR, the lower the frequency
is, the longer the antennas are (Figure 114a), which may complicate the acquisition on
difficult terrain (steep, rough, etc) as the whole antenna surface should be lying on the ground.
A more recent development is the use of towed systems which greatly facilitates the
acquisition in such cases (Figure 114b). Though such systems may not be as good as
conventional ones, their easiness and very fast use is a must, e.g., for landslide investigations.

SafeLand - FP7
Page 190 of 401

a)
Rev. No: 2
Date: 2012-02-15
b)
Figure 114 : GPR technology: a) Standard GPR system with a 25 MHz antenna (Svalbard; picture courtesy
Geological Survey of Norway); b) towed system with a 50 MHz antenna (4 m offset) (Lecomte et al., 2008).
Applications: For rockslide, GPR is particularly useful due to its high resolution (down to a
few cm) associated to a deep penetration (several tens of m) when using low frequencies
(below 50 MHz). It allows detecting interfaces, such as sliding surfaces, or groundwater table.
GPR measurements are also sensitive to fracture properties (filling, aperture). Applications of
heavy field GPR investigations in near 3D mode have recently emerged (Maurer et al., 2010;
Randa rock slide) and should be followed. The use of towed systems may also largely
contribute to more use of GPR on difficult ground (e.g., Lecomte et al., 2008). The use of
GPR for time monitoring (4D) of landslides has not yet been documented.
Limits: Severe limitations decrease the potential of GPR for landslide investigations, as
attested by the very low number of applications in this field (Jongmans and Garambois,
2007). First, GPR utilization in soil slide is limited, due to the classic presence of clay
materials and water saturated layers which absorb electromagnetic energy. This is the case
for, e.g., quick-clay site investigations in Norway and Sweden, though using GPR first may
still be good practice to delimit the clay zones. Second, the high level of diffractions in
complex 3D structures such as rock slides may yield poor imaging, though proper processing
similar to those used for reflection seismic could probably help. In comparison to seismic, a
major drawback of GPR surveys is also the lack of reliable velocity information to constrain
the scattering structures in depth as the acquisition is performed in a constant-offset mode,
i.e., with fixed distance between emitter and receiver. Without CMP information at different
locations, a proper velocity models cannot be derived and only rough estimates are obtained
by searching for diffractions patterns to determine mean velocities. Whenever possible, CMP
gathers should therefore be acquired.
1.2.5. Gravimetry
Method: Measuring microgravimetry (range: from 1 to 10 gal) allows to investigate on mass
anomalies (US Army Corps of Engineers, 1995). A picture of a microgravimeter is presented
in Figure 115.

SafeLand - FP7
Page 191 of 401
Deliverable 4.1
Figure 115 : A high precision gravimeter (picture taken from Seigel, 1995).
Applications: Regarding landslides, microgravimetric campaigns are indicated in two

situations where a sufficient density contrast may exist (Jongmans and Garambois, 2007): 1)
when the sliding surface is located at the interface between two very different materials (e.g.,
top of the bedrock for an earth slide); and 2) when the sliding mass is very disturbed
comparing to the in-place material. The light instrumentation presents big advantages in these
circumstances to study landslide geometry, led by fast campaigns requiring less effort on the
field than seismic or electric works (Jongmans and Garambois, 2007). A dense coverage can
be acquired and gravimetric anomalies generated by sufficient density contrast are detected
(at least a few tenths of g/cm3). This condition is fulfilled when the surface failure coincides
with the bedrock top or when the landslide reduces the compactness and therefore density of
the moving mass. Del Gaudio et al. (2000) underlined that together with the support of limited
other subsurface data (mechanical, geophysical, DTM), gravimetric surveys are able to
provide useful information for slope stability analysis: 1) estimation of landslide body
thickness and density contrasts between the moving and stable mass and, 2) location and
geometry of heterogeneities within the moving mass. Blha et al. (1998) also claimed that
gravimetric surveys provided an effective contribution to the description of the structures
(deformation, particular blocks, zones under tensile stress) and their dynamic control over
time (by gravimetric monitoring).
Limits: As noted by del Gaudio et al. (2000), the use of gravimetric surveys in slope stability
investigations is rather uncommon, mainly due to the long and difficult data processing and to
the strong non-uniqueness of interpretation (Reynolds, 1997). The major problem is
separating anomalies of interest from the overlapping effects of other features. In the example
of the Senerchia slump-earthflow (Southern Apennines), del Gaudio et al. (2000) performed
two microgravimetric surveys in order to evaluate the potential of gravimetry to detect
possible spatial-temporal density variations observed at surface. They showed that this
method was able to provide information on lithological heterogeneities that may control the
dynamic of landslide enlargement, if borehole measurements are available. The surveys did
not show enough sensitivity to detect any temporal density changes.
1.2.6. Borehole geophysics
Method: Though the present guidelines mostly intends to cover non-intrusive geophysical
techniques, when boreholes are first planned in a site investigation, it is highly recommended
to design them for geophysical logging as well if the cost allows it (boreholes could also be
used for various tomographies, e.g., seismic, resistivity and GPR). All physical parameters
measured by the geophysical methods mentioned earlier can thus be measured in details as
function of depth and later linked to surface-related measurements. The main advantage of

SafeLand - FP7
Page 192 of 401

Rev. No: 2
Date: 2012-02-15
borehole geophysics with regards to surface geophysics is the constant and detailed resolution
versus depth.
Applications: The literature about borehole geophysics for landslides is not extensive,
probably due to the cost of the borehole in the 1st place. But a few examples of various
geophysical logs, combined with geological and hydrological ones, can be found in Rnning
et al. (2006) and Ganerd et al. (2008) at the knes rockslide site in Norway.
Limits: Borehole measurements are obtained at a much finer vertical scale (a few cm) than
surface-based measurements and the retrieved parameters of the two approaches may be
difficult to relate to each other. A careful upscaling of the borehole data may therefore be
needed for a proper comparison, as well as calibration, with the surface-related data. In
addition to the scale issue, such measurements are very local and may not be representative of
the main structures. Ideally, the locations of boreholes should be planned according to
preliminary surface-based investigation to secure a proper sampling of the geological
structures of interest.
1.3. DISCUSSION
Jongmans and Garambois (2007) showed that geophysical prospecting is applied on various
types of landslides for slope varying from a few degrees (earth slide) to vertical (rock fall)
(Table 15). The penetration depth of the surveys may range from a few to several hundreds of
meters and the targets of the surveys are often two. The dominant one is the location of the
vertical and lateral boundaries of the slip mass (failure surface). An additional and wished
target is the mapping of the internal structure of the landslide. Four main different situations
can occur. In the first case, geophysical contrasts are due to lithological changes (layering,
tectonic contact or pre-slide weathering) and the failure surface mainly coincides with a
geological interface or layer. In the second case, geophysical contrasts are also controlled by
lithological variations, but the failure surface cuts the structure in a more complex way and
may be or not deduced from the geophysical image, depending on the landslide velocity, the
heterogeneity of the material and the resolution of the technique. Exceptionally (third
situation), the failure surface (or potential failure) is directly detected, mainly by propagation
methods (Petinelli et al., 1996; Bichler et al., 2004; Willenberg et al., 2004; Jeannin et al.,
2005). In the fourth case, the landslide develops in a globally homogeneous layer and alters
its characteristics. The geophysical contrast then arises between the slide and the unaffected
mass from the cumulative or separate action of the mechanical dislocation, the weathering and
an increase of water content (Caris and van Asch, 1991; Bruno and Marillier, 2000; Schmutz
et al., 2000; Mric et al., 2005; Lapenna et al., 2005; Lebourg et al., 2005). The second target
of geophysical prospecting is the detection of water within the slip mass, for which electrical
(Bruno and Marillier, 2000; Lebourg et al., 2005; Lapenna et al., 2005) and electromagnetic
(Caris and van Asch, 1991; Mauritsch et al., 2000) methods were most applied.
A landslide category not highlighted before is the one of debris flows (and other flows of rock
or soil) if we follow Cruden and Varnes (1996), as done in the present project. Among the
large variety of debris flows, we will briefly mention here the use of geophysics for those
generated by glacial hazards, e.g., so-called Glacial Lake Outburst Flows (GLOFs), due to
their potentially huge threat to population and infrastructures (Reynolds, 1992). The most
devastating events happened so far in the Himalaya and in Peru; to give an order, over 32 000

SafeLand - FP7
Page 193 of 401
Deliverable 4.1
people were killed by glacially-related sliding phenomena in Peru during the 20th century
(e.g., 1941: about 7000 people died in one event in Cordillera Blanca). However, global
warming with glacier melting/retreat may pose problems elsewhere (Alpes, Caucasus, etc). In
such cases, it is not possible to investigate the slide part itself as it comes brutally but the
source of such debris flows may correspond to zones where geophysics can bring lots of
information (Reynolds, 2006). Several flood scenarios are possible in glacial hazards with,
e.g., the presence of lakes at the front end of glaciers and dammed by moraines, etc. A very
recent example (summer 2010) is a large pocket of melt water, blocked by frozen channels,
detected by GPR measurements and others, and located under a glacier in French Alpes near
the Mont Blanc; if the ice roof collapses in the water pocket (happened in 1892), a huge flood
wave will rush down the valley carrying lots of boulders and soil, and destroying everything
on its way to the city of St Gervais. Mitigation of this huge risk is underway but without
detection by geophysics, such risk prevention would not have been possible. While GPR was
used here to detect the pocket, repeating such measurements over time (monitoring) would
also permit to follow the development of the water pocket. GPR is a very favourable
technique on glacier and frozen grounds (e.g., permafrost) because ice is a very resistive
material in contrary to water. For the same reasons, resistivity methods may be very useful
too. In a joint interpretation (if not inversion), the contrasts detected by GPR complete nicely
the smooth resistivity structures as imaged by electrical tomography in such cases (Lecomte
et al., 2008).
The design of a geophysical survey for landslide recognition is still a much debated question
and no unique strategy came out from Jongmans and Garambois (2007), especially
considering the wide range of landslide types around the world. In such heterogeneous
structures, the combination of different geophysical techniques anyway appeared as a
necessary condition for obtaining reliable results (e.g., geophysical investigations at knes,
Norway, as reported by Rnning et al., 2006, 2007). The choice of the techniques is clearly
guided by the expected contrasts in physical parameters. Other parameters, like the required
penetration depth, as well as the volume and the morphology of the landslide, may also have a
significant effect on the survey strategy, including for economical reasons. The major
difficulty of applying geophysical techniques to landslides is probably the complex
relationship between the measured geophysical parameters and the desired geotechnical and
hydrogeological properties, which prevents from giving a straightforward interpretation in
terms of engineering properties. In other domains, several attempts were made, e.g., in
engineering geology, to derive soil or rock properties from geophysical measurements, using
experimental relationships. In soils, correlations were developed between the small-strain
shear wave velocity (Vs) and penetration resistance from CPT, mainly in geotechnical
earthquake engineering. Recently, Ghose (2004) proposed a model-based integration of
seismic and CPT data to derive soil parameters for sandy material. In rocks, most of the
geophysical studies were aimed at characterizing the rock quality or fracturing (e.g., GPR
techniques to determine the fracture geometry). Apart from radar imaging, seismic methods
play a more and more important role in characterizing rock sites for geotechnical purpose. As
an example, a relationship between S-wave velocity and the Rock Mass Rating, which is a
geotechnical factor used for tunnel design, was recently proposed by Barton (2006) and Cha
et al. (2006). In landslide investigation, similar relationships, linking for instance geophysical
parameters to the displacement rate (Mric et al., 2005) should be studied more deeply.
Beside geotechnical parameters, assessing hydrogeological properties from geophysical data
is a necessity and outstanding results have been obtained in recent years in a new
SafeLand - FP7
Page 194 of 401

Rev. No: 2
Date: 2012-02-15
interdisciplinary field (hydrogeophysics), combining the integration of multiple sources of

data and the development of comprehensive petrophysical models. The application of these
methods allows hydrogeological parameters of the subsurface, like porosity, water content,
and hydraulic conductivity to be estimated from high-resolution fluid-sensitive geophysical
data (seismic, electrical, electromagnetic): a recent state-of-the-art can be found in Rubin and
Hubbard (2005). Both experimental relationships and quantitative approaches should be
developed in the future for landslide investigations, incorporating under-used techniques, such
as spontaneous potential and induced polarization, and including numerical modelling, data
inversion and fusion.
1.4. CONCLUSIONS
Areas affected by landslides usually exhibit dramatic spatial and temporal variations of
lithological and hydrogeological conditions. The development of 2D, and very recently 3D,
geophysical imaging techniques has been a first major advance forward for investigating the
complex structure of landslide areas. A second one will be the installation of permanent
arrays of geophysical sensors as a part of monitoring systems of landslides. Such geophysical
time-lapse surveys have recently been initiated on some landslides (Supper and Romer, 2003;
Lebourg et al., 2005), mainly with a multi-electrode electrical array. Coupled with high
resolution remote-sensing techniques (Van Westen, 2004) and SP monitoring systems for
hydrological purposes (Mric et al., 2006), these permanent geophysical imaging systems
give a new insight into the 4D deformation mechanism of a landslide. However, geophysical
techniques may suffer severe drawbacks, as listed earlier, and they need to be combined and
calibrated against geological and geotechnical data to give reliable information. The
complexity of landslides also requires using a combination of different geophysical
techniques. Geophysicists have to make an effort in the presentation and discussion of their
results. Resolution and penetration issues are not systematically discussed in an
understandable manner and the geological/geotechnical interpretation of geophysical data
should be more clearly and critically explained. Lack of communication and explanation in
early multi-disciplinary projects (not only for landslides) may partly explain the reluctance of
the engineering community to use geophysical techniques. It will be a challenge for
geophysicists in the following years to convince geologists and engineers that 3D and 4D
geophysical imaging techniques can be valuable tools for both investigating and monitoring
landslides. Finally, efforts should also be done towards more quantitative information from
geophysics in term of geotechnical parameters and hydrological properties.
2. ACOUSTIC AND MICRO-SEISMIC MONITORING

2.1. INTRODUCTION
2.1.1. Definitions: Acoustic Emission (AE) and micro-seismic monitoring
The macroscopic deformation of rocks, debris and soils, at scales ranging from laboratory
samples (cm) to hillslopes and earth crust (hm to km), is associated with local irreversible
processes (cracks/faults propagation and shearing, crack/fissure opening, material friction
along discontinuities, etc). These movements involve propagation of acoustic waves which
can be observed by remote sensing. Passive seismic monitoring during strain progression can
help our understanding of mass movement behaviour and lead to the recognition of possible
failure precursors. Although of significant potential, this observational tool has had only
SafeLand - FP7
Page 195 of 401
Deliverable 4.1
limited application in the analysis geohazards such as glaciers, volcanoes, and mass
movements such as landslides, rockfalls and debris flows.
Recording seismic noise consist in using seismic sensors (e.g. seismometers, geophones,
accelerometers, acoustic emission transducers, hydrophones) to monitor passively potential
events, also called Acoustic Emission (AE) events. The extent of the investigated area
varies from some centimetres to thousands of kilometres, depending on (1) the magnitude and
depth of the events, (2) the possible attenuation of the waves by the material, and (3) the type
of seismic devices and the design of the seismic array. In the case of natural events, seismic
noise monitoring may provide information on the characteristics of regional earthquakes to
small material instabilities observed within a local slope failure.
AE is generated by an event source which converts localised stress accumulation of elastic
energy into the release and the propagation of elastic waves that are transmitted within a
material. AE sources may characterize several types of material behaviours such as failure,
friction, shearing and sliding, crack propagation, impact, and cavitation processes. For the
monitoring of geohazards, AE is also referred to as micro-seismicity or rock noise.
To understand the failure processes and to search for precursory patterns to material failure,
the micro-seismicity tool has been extensively used at laboratory rock sample scale (Lockner,
1993) and at an intermediate scale between the laboratory scale and the large tectonic
earthquake for studies of seismicity and rock burst in mines or tunnels (Obert, 1977;
Nicholson, 1992). Most of these research activities have been conducted following the initial
ideas of Obert et al. (1941; 1942; 1945a; 1945b; 1957) who have studied sub-audible noise
associated to mining activities. Later, Goodman et al. (1965) and Cadman et al. (1967)
demonstrated from field acquisitions and laboratory experiments that rock and earth slopes
generate possible forerunners of acceleration by emitting acoustic signals. These sub-audible
noises can be detected over distances of several tens of meters in hard rocks but are attenuated
below detectable level into less than 30m in soft rocks and soils. These results suggested that
installing arrays of conventional seismic geophones on the ground surface near the most
active parts of mass movements should provide information about their geometry and
dynamic. A few application of micro-seismic monitoring for slope stability are related either
to open mines and quarries (Hardy et al., 1991; Kennedy et al., 1971). Concerning natural
mass movements, some rare experiments have been carried out mainly in rocky cliffs
(Kolesnikov et al., 2003, Amitrano et al., 2004; Willenberg et al., 2004, Eberhardt et al.,
2004; Amitrano et al., 2005). The micro-seismic monitoring technique has also been used for
the monitoring and the warning of debris flows (Arratano, 1999; Itakura et al., 2005). In the
last case, the amplitude of seismic noise appears to be correlated to the volume and velocity
of the debris flow and can be used to trigger an alarm or a video recording (Lavigne et al.,
2000). For mass movements developed in soft rocks and deforming continuously (such as
mudslides), the high attenuation of the material generally avoids the use of micro-seismic
monitoring. Some authors have overcome this difficulty by using passive or active
waveguides (Novosad et al., 1977; Dixon et al., 1996; Kousteni et al., 1999; Dixon et al.,
2003). The passive waveguide is used to provide a shorter path for transmitting the elastic
waves directly to the sensor, whereas the active waveguide generates AE signals by
deforming itself within the moving mass (as the principle of vertical inclinometers). The
system of waveguide allows using the micro-seismic monitoring technique even for very
attenuating material such as clayey soils.
SafeLand - FP7
Page 196 of 401

Rev. No: 2
Date: 2012-02-15
2.1.2. Objective and structure of the state-of-the-art

This section of deliverable D.4.1 aims at presenting a current state-of-art on passive seismic
applications carried out to monitor and characterize mass-movements. In order to illustrate
this state-of-the-art, a synthesis of the more recent applications of seismic noise monitoring,
among probably many others, is summarized in Table 19. It illustrates AE monitoring
applications for all types of mass movements (falls, slides, flows) and for several types of
objectives (e.g. observation and characterization of the AE signals in terms of amplitude,
frequency, duration and magnitude, to more advanced analyses of AE signals spatial
localization). Nowadays, most of the research activities in micro-seismic monitoring of mass
movements are to develop pattern recognition analysis tools, characterization of the sources
of the AE signals and correlation of the signals to internal and external triggers (earthquake,
rainfall).
The state-of-the-art is organized by presenting successively the different methodologies of AE
monitoring that have been introduced chronologically by different authors.
While first applications of AE monitoring aimed at detecting and identifying the temporal and
frequency domain of event sources, progress is nowadays carried out towards providing more
details on the dynamics and mechanisms of mass movements (source event localisation and
characterisation). With time, long-term monitoring (LTM) of AE aims at developing
forecasting tools based on pattern recognition techniques for mass-movement early warning
systems.
Table 19 (bellow) : AE signals monitoring used on unstable slopes during the last 20 years (non-exhaustive list).
The colours indicates the different types of geological material (successively from white to dark grey: ice, rock,
debris, and soil materials) and the numbers indicated in the 3rd column (category of study) correspond to the
hereafter sections of the state-of-the-art.

SafeLand - FP7
Page 197 of 401
Deliverable 4.1

SafeLand - FP7
Page 198 of 401


SafeLand - FP7
Rev. No: 2
Date: 2012-02-15
Page 199 of 401
Deliverable 4.1
2.2. THE PIONEERING PERIOD IDENTIFICATION OF AE SIGNALS AND

LIMITATIONS OF THE TECNNIQUE
2.2.1. Identification of AE signals
The first information that can be obtained from AE signals properly acquired is the amplitude,
the duration and the frequency content of the plotted signals considered as events. First
studies by Obert in the 1940s, and Goodman and Cadman in the 1960s highlighted the
variations of signals in the temporal and frequency domains with the load and the induced
response within a rocky slide (Figure 116).
Figure 116 : a) Oscillations of rock noise received at four different detectors at different times (Goodman et al.,
1965); b) Recording of a subaudible event generated within a small-scale landslide and received by four
piezoelectric transducers, showing the three time delays (Cadman et al., 1967).
Cadman et al. (1967) carried out flume tests in the laboratory with partially saturated sand in
order to observe the sliding mass and to record the movement noises with four piezoelectric
transducers placed within the sandy slope at known coordinates, and whose output was
recorded on a magnetic tape. These studies proved the increase of AE emission before failure
and the possibility of localising the events through non-linear resolution systems (NewtonRaphson) from the understanding of the compression wave velocity, the coordinates of the
geophones and the time delays between them.
2.2.2. Frequency domain observations: spectrum and spectrogram/sonogram

With the digitalization of the seismographs, it became easier to study the signal spectra in the
frequency domain through several transformations. Spectral representation actually allows
discriminating signal from low signal-to-noise ratio (SNR) recordings because it highlights
the frequency content of the signal and consequently helps in applying filters and discriminate
the interesting AE events from the ones induced by external disturbing noises.

SafeLand - FP7
Page 200 of 401

Rev. No: 2
Date: 2012-02-15
By the same way, for interesting AE signals that appears to be confused in the temporal
domain, the frequency content pattern can be a useful tool to discriminate the characteristics
of different events.
Rouse et al. (1991a) for instance observed four distinct types of AE signals within a muddy
landslide in South Wales, each type being characterized by different peaks in frequency
(Figure 117 left).
Figure 117 :(left) Time series and power spectrum of two identified events at Glynrhigos (South Wales) (Rouse
et al., 1991a) . (right) Seismic noise generated by a debris flow: (a) time series; (b) signals in the frequency
domain obtained using FFT; (c) spectrum obtained using Gabor transform (Huang et al., 2004).
In addition to the application of numerous transform operators (Fourier, Gabor; eg. Huang et
al., 2004; Figure 117 right), the use of spectrograms and sonograms has been developed in the
recent years. A spectrogram is an image that shows the variations of the spectrum of the
signal with time. For this reason, combination of classical seismogram analysis with
spectrogram/sonogram analysis is now a useful tool to point out with precision the located
events and to extract their characteristics in time and frequency. Some examples are detailed
in Section 3 as a useful application to discriminate different AE signals induced on landslides.
2.2.3. Limitations of the technique: identification of attenuation and noise

problems
Under stress conditions, geological materials can induce a broad range of frequency extending
from 10Hz to 500kHz. In the field, low frequencies can be contaminated by background
noises (electrical noise, wind and tide waves, human disturbance) while higher frequencies

SafeLand - FP7
Page 201 of 401
Deliverable 4.1
are affected by the high attenuations of the material. Therefore, a choice has to be made in
order to minimize the loss of information.
To prevent material attenuations, metal wave guides have been used to conduct signals from
the soil or rock body to the AE sensors (Lord et al., 1982; Hardy, 1992; Dixon, 1996 and
2003; Kousteni, 1999; Shiotani, 2006). According to Figure 118, iron and steel provide very
low attenuated responses with respect to the geological ones (Koerner et al., 1981). An
example of such an acquisition instrument is illustrated in Figure 119. The steel bar is more
sensitive to the deformation process, and also compensates the attenuations of the material.
Figure 118 : Attenuation response of different soil types, rock, coal and iron/steel (Koerner et al., 1981).
Figure 119 : Schematic behavior of the WEAD waveguide during the deformation process of rock (left: intact
state, right: AE signals generated along with rock deformation) (Shiotani, 2006).

SafeLand - FP7
Page 202 of 401

2.3. THE RECENT PERIOD MICRO-SEISMIC

CHARACTERIZATION OF AE SIGNALS
Rev. No: 2
Date: 2012-02-15
MONITORING
AND
In this part, we tried to distinguish ambient noise listening applications from micro-seismic
monitoring for event source location and characterization.
2.3.1. Low-amplitude seismic noise listening for landslide structure

characterisation
Nakamura (1989) adapted a method introduced in the early 1950s, which used microtremors
for estimating dynamic characteristics of surface layers. Microtremors are low amplitude
ambient vibrations of the ground caused by man-made or atmospheric disturbances.
Nakamura explained that the study of the H/V (horizontal on vertical components, recorded
simultaneously at a given location with a 3D seismometer placed on the ground surface)
spectral ratio of the ambient noise can be used to detect the sedimentary zones within surface
layers that could amplify seismic ground motion.
After experimental and numerical validation of the H/V method, Mric et al. (2007) applied it
to muddy landslides (Super-Sauze, Saint-Guillaume) composed of stratified soil profiles (e.g.
a soft layer on top of stiff bedrock). Mric et al. used the H/V spectral ratio method to detect
first the slip surfaces (Figure 120, Figure 121), and second to derive S-wave velocity profiles.
However, they encountered difficulties in the interpretation because of multiple resonance
frequencies probably due to deeper interfaces, heterogeneities or bad coupling with the
ground during the acquisition.
Figure 120 : Example of ambient vibration data acquired on the Super-Sauze landslide: a) Ambient vibration
data sampled at 200 Hz and recorded for 30 min, b) Spectral ratio computed using the E-W component, the N-S

SafeLand - FP7
Page 203 of 401
Deliverable 4.1
component and the H component (square root of E-W and N-S components) for point 15 of profile CA; c)
Spectral ratio computed using the E-W component, the N-S component and the H component (square root of EW and N-S components) for point 14 of profile B (Fig. 7) (Mric et al., 2007).
Figure 121 : Interpretation of H/V picked spectral frequencies on the Super-Sauze mudslide for three profiles B,
C1 and CA, assuming an average (S)-wave velocity of 260 m.s-1. Results from geotechnical and electrical data
are also shown for comparison. The interface between the sliding material and the stable bedrock was drawn
considering all the results. The presence of two peaks in some H/V ratio (Fig. 6b and 6c) resulted into two
possible interpretations, although lower frequencies are expected to be due to the presence of deeper interfaces
(former coarse mudslide), acquisition problems or local heterogeneity (Mric et al., 2007).
A critical review of microtremor H/V technique has been proposed by Mucciarelli et al.
(2001), who synthesized several case studies between 1990 and 2000. Thanks to its low-cost
and fast deployment on a site, the H/V method is convenient for landslide characterisation.
2.3.2. High-frequency seismic noise listening for landslide velocity/volume

estimation
Other studies carried out on mass movements are related to the use of seismometers or
geophones for landslide volume estimation or velocity estimation. For instance, Arattano et
al. (1999) and Marchi et al. (2002) studied the velocity and the volume of debris flows in the

SafeLand - FP7
Page 204 of 401

Rev. No: 2
Date: 2012-02-15
North East Italian Alps by installing geophones in debris flow prone torrents. Suwa et al.
(2000), Zhang et al. (2004) and Huang et al. (2007) calculated the volume and the kinetic
energy (e.g. loading) of debris flows in Japan and China. Suriach et al. (2000) determined a
relationship between seismic signals and avalanches paths and volumes and Brodsky et al.
(2003) calculated the coefficient of friction of lahars. All these results are based on the
analysis of high-frequency seismic noise using spectral analysis of the continuous seismic
recordings. Such an example of analysis is presented on Figure 122 with the detection of a
granular debris flow that occurred on 15/08/2003 in the Himalaya mountain range (Burtin et
al., 2009).
Figure 122 : Example of seismic monitoring of a large granular debris flow that occurred on 15/08/2003 in the
Himalaya mountain range. a) Vertical seismogram recorded at station H0390. The seismic signal is bandpass
filtered between 0.1 and 20 Hz. The recording starts at 2:00 pm on the 15/08/2003 (local time). b) Spectrogram
calculated from the vertical seismic recording at station H0390, using a multi-taper method with 50% overlap
between time segments of 5 min. Amplitudes are given in decibels (dB), red and blue colors stand for high and
low amplitudes, respectively. (c) 1-hr-smoothed mean high frequency seismic energy at station H0390 for the
vertical component (b) in the 2-22 Hz frequency band. Amplitudes are given in dB. (Burtin et al., 2009).

SafeLand - FP7
Page 205 of 401
Deliverable 4.1
2.3.3. Micro-seismic monitoring and correlation of the AE signal catalogues to

possible triggers
Other studies involve micro-seismic monitoring in order to detect all type of AE signals
within long periods of time and without any frequency trigger. A first application consists in
developing catalogues of landslide events to observations of possible triggers such as
earthquakes, rainfalls or explosions. The goal is also to discriminate the mass-induced AE
from the direct acoustic emission generated by these external events. For instance,
observations carried out by Rouse et al. (1991a, b), Walter et al. (2008) and Helmstetter et al.
(2009) during and after long rainfall periods on landslides demonstrated the marked change in
the number of AE signals which occurs during and after the rain. Walter et al. (2008)
observations are referred to the Heums clayey landslide in Vorarlberg (Austria); according to
Figure 123, they concluded that the events are generated from a few minutes to one day after
intense rainfalls, which let them conclude that these AE signals were generated by
precipitation-induced material failure within the landslide. Clusters of AE signals can also be
clearly observed over the period.
Figure 123 : Micro-seismic monitoring at the Heums landslide in Vorarlberg (Austria). Example of correlation
among rain intensity (red), ground water level (blue) and temporal occurrence of the recorded fracture events
(with their magnitude ML) during 7 days in September 2005. Clusters of AE signals can be clearly observed
over the period, and associated to the changes in ground water levels with some delays (10-11/09/2005) or to
rainfall (13/09/2005). (Walter et al., 2008).
2.3.4. Micro-seismic monitoring and temporal behaviour of AE signals at the

onset of large failures
Other studies have been carried out in order to characterize the temporal pattern of AE signals
at the onset of major failures.
Senfaute et al. (2009) examined the micro-seismic signal recorded 15h prior to a rockfall that
occurred at Mesnil-Val in France in a chalky cliff; the results lead to the hypothesis that
several phases of failure mechanisms contributed to the rockfall (Figure 124). These
encouraging results offered the possibility of using micro-seismic systems to possibly forecast
large failures in hard rock slopes.

SafeLand - FP7
Page 206 of 401

Rev. No: 2
Date: 2012-02-15
Figure 124 : Micro-seismic monitoring of a cliff collapse at Mesnil-Val (North-West France). a) Location of the
sensors (A1, A2, A3, A4, A5) on the cliff and photograph of the rockfall on 23/06/2002 that occurred at the
centre of the monitored zone; b) Number of micro-seismic events per day recorded only during low tides (the
number of events increase during tides) and during the 18 months of the monitoring period. (Senfaute et al.,
2009).
Correlation among large earthquakes and the dynamic response of a landslide can also be
found in Harp et al. (1995) who proposed a method to anticipate critical failures that may
occur in already unstable slopes. The same type of temporal pattern of AE events has been
observed at the Valoria landslide (Northern Apennines, Italy) in soft marls, during the failure
of the upper part of the landslide that occurred in November 2009 (Tonnellier et al., 2010);
the first results indicated the increase of AE signals until failure (Figure 125).
Figure 125 : Micro-seismic monitoring of the Valoria landslide (Northern Apennines) in November 2010. An
increase in landslide displacement rate and in the number of seismic events recorded by station IHR05 during
the day before the failure is observed. The number of events started to increase in the seismic station located
downhill and then progressively migrate uphill. (Tonnellier et al., 2010)

SafeLand - FP7
Page 207 of 401
Deliverable 4.1
2.3.5. Micro-seismic monitoring and spatial pattern of AE signals for landslide

mechanism understanding
Up to now, few studies have investigated the source location of the AE signals within a
landslide. Further information on the source should help to understand better the mechanims
that control the dynamics of the landslide. Accuracy of detection of the source location is
affected by many factors, such as the characteristics of the sensors, the geometry of the array,
the previous knowledge on the investigated slope (layering, location of slip surfaces), the
petrophysical characteristics of the material (seismic velocity model, ambient noise) and the
type of recorded AE events (Ge, 2005).
As explained in 2.3, Cadman (1967) was successful to locate the AE events between the
uphill, the middle and the downhill parts of a small-scale man-made slope in laboratory tests.
Using waveguides, Novosad et al. (1977) were able to characterize the direction of the
sources. Such pioneering results proved that spatial location of low energy AE on slopes
could be foreseen using the same type of location techniques used for high energy
earthquakes.
Near Tomsk (West Siberia, Russia), Zyatev et al. (1998) observed strongly differing
polarisation patterns of microseisms on active and inactive parts of landslides. These
polarisation patterns were confirmed thanks to numerical simulations and several processing
tests in a landslide site in the Northern Tien-Shan (Kirgizstan) with the conclusion that cracks
that develop in the vicinity of a forming slip surface often have preferred orientations due to
stress conditions in rocks (Figure 126; Kolesnikov et al., 2003). Polarisation analysis also
appears to be a useful tool for providing the orientation of the sources generated within
landslides.
Figure 126 : Detection of AE signals polarisation on a landslide in Western Russia. Horizontal projections of
possible particle trajectories in a point on the slope: a) from the original record, b) after 10-45Hz band pass
filtering, c) after 120-210Hz band pass filtering. (Kolesnikov, 2003).
After an efficient pattern recognition of the events recorded at the Schilienne rockslide in the
French Alps, Helmstetter et al. (2009) highlighted that a better location of the sources for
rockfall-induced events is still difficult because most recorded AE signals did not present
impulsive waves, which hindered the time picking of first arrivals (Figure 127), as well as the
differentiation among P and S waves because of the small distance between the sources and
the receivers. They also used the cross-correlation of signals recorded at different sensors in
order to precisely measure the time delays between the sensors. They also used the beamGrant Agreement No.: 226479
SafeLand - FP7
Page 208 of 401

Rev. No: 2
Date: 2012-02-15
forming methods proposed by Almendros et al. (1999) in order to search for the best source
point which maximises the intertraces correlation, averaged over all couples of sensors, and
after shifting the traces in time by the travel time. However, the sources of these AE signals
were not precisely located and the authors concluded that the array of sensors should be
completed.
Figure 127 : A selection of different types of signals. (a) Seismograms, (b) zoom for a window of 5 seconds
around the peak of amplitude. The maximum amplitude (in counts) is given in the upper-right corner of each
plot. (c) Spectrograms (log amplitude scale). Dots indicate the onset and end of each event (Helmstetter et al.,
2009).
Applying the nanoseismic method (Joswig, 2005), Walter et al. (2008, 2009b) analyzed
waveforms and spectrograms to detect and partly locate distinct types of events caused by the
dynamics of the Heums (Walter et al., 2008) and Super-Sauze (Walter et al., 2009b;
Tonnellier et al., 2010) landslides. Some signals correspond to collapses of brittle material
from the upper parts of the landslides; other are generated by micro-cracks propagating within
the landslide. Such events are recorded by all the stations which is a criteria to discern them
from potential noise. They concluded that more precise location of the sources should be
obtained thanks to a denser distribution of the seismic arrays, and that a detailed catalogues of
events with a quantitative description of the signals are the basis of a sound analysis.

SafeLand - FP7
Page 209 of 401
Deliverable 4.1
Figure 128 : Waveforms and sonograms of three events that were recorded during the seismic monitoring on the
Heums landslide in Vorarlberg (Austria): a) Fracture process ML=-2.0 in 160 m distance, b) fracture process
ML=-1.4 in 350 m distance and c) local earthquake ML=1.3 in 10 km distance. (Walter et al., 2008).
Figure 129 : Waveforms and sonograms of common AE events observed at the Super-Sauze and Valoria
landslides directly induced by the landslide mechanisms. (Tonnellier et al., 2010).
2.3.6. Towards mechanical description of the AE signals observed on landslides

The discrimination of the AE signals observed within a landslide provides information on its
mechanism because it helps in decomposing its evolution into several phases. Such
information can be completed by the results obtained from other monitoring devices such as
(a) ortho-photographs obtained from the use of an Unmanned-Aerial-Vehicle (UAV) to
combine evolution of the ground surface of the landslide (e.g. displacement, development of
the fissure pattern) and micro-seismic monitoring (Walter et al., 2009a), (b) displacement and
ground water level monitoring (Walter et al., 2010) and (c) 3D geometrical models of the
landslide to locate a priori the main possible sources of AE signals (Figure 130; Walter et al.,
2009b).

SafeLand - FP7
Page 210 of 401

Rev. No: 2
Date: 2012-02-15
Figure 130 : Location (in planimetry) of the main sources events detected at the Super-Sauze landslide in July
2008. Most event sources are located along crests buried below the landslide mass where friction and sliding
processes are maximized, in the most active part of the landslide where the surface displacement rates are high,
in intensively fissured areas and in the upper part near the main scarp where small amplitudes rockfall are
observed. (Walter et al., 2009b).
Using the same type of approach, de Angelis et al. (2007) demonstrated the advantage at
considering together seismograms and spectrograms in order to better identify the AE signals
induced by the travel of rapid pyroclastic flows at the Soufrire Hills Volcano, and to
characterize the phases of the signals during the propagation of the flow (Figure 131).

SafeLand - FP7
Page 211 of 401
Deliverable 4.1
Figure 131 : a) Seismic envelope, b) velocity seismogram, c) Spectrogram from the signal recorded during
pyroclastic flows at the Soufrire Hills Volcano, Montserrat (de Angelis et al., 2007).
2.4. CONCLUSIONS
Using acoustic seismic methods to characterize the structure, define the dynamics and
understand the mechanism of mass movements is not as recent as one could have thought.
This review on the application of passive seismic monitoring techniques has shown that they
can be exploited for a wide range of issues and objectives, chronologically focused on
simple seismic noise content characterization (amplitude, duration, frequency) to more
detailed analyses in terms of catalogues and accurate location of the sources. The introduction
of waveguides and above all the digitalization of seismic data increased the feasibility of
characterizing AE signals within unstable slopes using spectral approaches.
Studies have been carried out on a variety of geohazard types in a broad range of geological
contexts: soft rock landslides, rocky slopes, rock quarries, mining activities, granular and
muddy debris flows, lahars, snow avalanches, ice glaciers (Table 19). These exercises
provided successful information about the definition and the dynamics of the mass
movements. Based on these experiences, Suriach et al. (2005) and Spillman et al. (2007)
have started to investigate the possibility of automate the recognition of AE signals on
respectively snow avalanches and rockslides. On a longer term, the main issue will be to
characterize the source mechanisms, and to detect of possible seismic forerunners of mass
movement acceleration and define the potentialities of using seismic noise observations as a
predictive-tool for early-warning of all kinds of mass movement failures.
For that purpose, dense AE monitoring devices have been set up at several landslides for long
term observations (e.g. Super-Sauze and Avignonet mudslides in soft clays, Schilienne,
Aknes and Randa rockslide in hard rocks).

SafeLand - FP7
Page 212 of 401

Rev. No: 2
Date: 2012-02-15
3. INTRODUCTION TO OFFSHORE METHODS

Submarine and coastal landslides are of major concern for coastal communities and
infrastructures but also for the safe development and exploitation of seafloor resources.
Because of their unique geological settings, these landslides can be up to two orders of
magnitude bigger than those that occur on land. Hazards related to such landslides may be
catastrophic and range from destruction of offshore facilities to collapse of coastal
infrastructures and the generation of tsunamis (e.g., Longva et al. 2003, Bondevik et al. 2005,
Dan et al. 2007). During the last two decades, researchers have shown the great potential of
combining multi-beam bathymetric surveys and high resolution seismic methods for studying
landslides processes in the submarine environment and also for mapping the risk related to
such natural hazards (e.g., Locat and Lee 2002, Haflidason 2004).
3.1. MULTI-BEAM SYSTEMS

Seabed topography is conventionally mapped using vessel-based acoustic echo-sounding
devices, such as multi-beam bathymetric systems. Through digital processing techniques and
with help of differential positioning systems the data provides precise shaded relief
topographic maps of near air-photographic quality (Figure 132). Such maps are now routinely
used for detailed morphological mapping of the seabed, for mapping of landslide scars and
deposits, and also as a prerequisite to slope stability analysis. Most swath bathymetry systems
also acquire backscatter data, which can be used for mapping local changes in morphology
and also for the production of sediment maps.
Multi-beam systems use acoustic signal emitted from a series of transmitter mounted on the
hull of a vessel (Figure 133). The beams are transmitted at different angles from the same
transducer unit, creating a fan perpendicular to the vessel direction. The angle of the fan is
termed swath angle, which along with the water depth determines the width of the corridor
mapped (Table 20). As a rule of thumb, a wider swath angle and deeper water increases the
corridor mapped but reduces the resolution as the distance between the individual beams
increase. The greater the number of transmitters and higher frequencies will provide more
precise bathymetric coverage. Since multi-beam systems acquire dense sounding data both
along the ship's track and between the track lines, they can provide 100% coverage of the
seafloor.
The many multi-beam systems that are currently on the market utilise two differing
technologies to achieve bathymetry measurements across a swath of the seabed: 1) multiple
narrow-beam swath bathymetric systems (Hughes Clarke, 1996) and 2) multiple-beam
interferometric or phase discrimination systems (Green and Cunningham, 1998). Both of
these techniques have their merits; however the same end results can be achieved. Bathymetry
data from various sources can also be merged together (Figure 132).

SafeLand - FP7
Page 213 of 401
Deliverable 4.1
Figure 132 : Example of shaded relief bathymetric contour map (2 m interval) within the bay of Trondheim,
central Norway (LHeureux et al. 2010). Bathymetry data were collected using a GeoSwath 250 kHz system from
GeoAcoustics and a multi-beam echo sounder EM 2000 from Kongsberg.

SafeLand - FP7
Page 214 of 401

Rev. No: 2
Date: 2012-02-15
Figure 133 : Examples showing a multi-beam sonar swath (from SeaBeam, 1999).
Multi-beam echo sounders (MBES) collect bathymetric soundings by measuring the depth to
the sea floor in discrete angular increments or sectors across the swath (Hughes-Clarke et al.
1996). The bathymetry is calculated from the two way travel time that is corrected for the
traveling patch and measured properties of the seawater. Modern MBES systems use
frequencies spanning from a few kHz to several hundred kHz depending on the sea floor
depth (Table 20). For example, low frequency (12 kHz) systems can collect swath soundings
at full ocean depths, i.e., up to 30,000 meters. In contrast, high frequency MBES systems
(300+ kHz) are utilized for collecting swath bathymetry in depths of 200 meters or less. Table
20 shows some general characteristics of MBES available for seabed mapping, including
depth range and swath coverage. Such details must be taken into consideration when planning
a survey for a given environment.
In contrast, the term interferometry is generally used to describe multi-beam techniques that
use the phase content of the sonar signal to measure the angle of a wave front returned from a
sonar target. In general, interferometry systems have a slower ping rate then MBES system
thus they are primarily used in shallow water operations. The along line density of data in
interferometry systems is restricted by vessel speed, however, cross line data is only restricted
by the quality of the data and processor, thus small targets can be readily observed. MBES
have higher ping rates and so the along line density is not so restricted by speed but cross line
density can vary depending on the number of fixed or variable beams available.
Table 20 : Typical multi-beam echo sounder systems.
Type of system
Multi-beam
echosounders
(MBES)
Manufacturer
/ Model
Kongsberg
EM 2000
Kongsberg
EM 3002
Kongsberg /
EM 122
Kongsberg /
EM 302
Kongsberg /
EM 710

SafeLand - FP7
Frequency
(kHz)
Depth range
(m)
Swath coverage
Depth resolution
200
1-300
7 x depth or 250 m
8 cm
300
0.5-200
10 x depth or 200 m
cm
12
20-11,000
6 x depth or
30,000 m
cm
30
10-7000
5.5 x depth or 8 km
cm
70-100
3-2000
5.5 x depth or 2.5 km
cm
Page 215 of 401
Deliverable 4.1
Interferometric
systems
R2Sonic LLC /
Sonic 2024
RESON/
SeaBat8125
GeoAcoustics /
GeoSwath +
GeoAcoustics /
GeoSwath +
GeoAcoustics/
GeoSwath +
200-400
200
11 x depth
< cm
455
0.5-120
400 m
6 mm
125
200
12 x depth
6 mm
250
100
12 x depth
3 mm
500
50
12 x depth
mm
3.2. 2D AND 3D HIGH RESOLUTION SEISMIC

In combination with multi-beam bathymetric surveys, 2D high resolution seismic data is
routinely acquired for investigating submarine mass movements and other offshore
geohazards. A source, generating an acoustic signal which can vary from MHz to 10 Hz, and
receivers (hydrophones) recording the reflected sound, constitute the basic setup of seismic
acquisition systems (Figure 134). The interpretation of features mapped using a seismic data
set is based on acoustic impedance differences of subsurface material. The transmitted
acoustic energy is reflected from boundaries between mediums of different acoustic
impedances (e.g. the water-sediment interface). The interface between different materials is
displayed graphically in a seismic section and these are used to analyse the structure of a
basin. As an example, seismic data can be use to capture lateral variability in sediment
properties and to identify the presence of weak geotechnical layers on which a mass
movement may initiate (Figure 134). High resolution seismic data are also well suited for
mapping buried landslide deposits (e.g., Hermanns et al. 2006).

SafeLand - FP7
Page 216 of 401

Rev. No: 2
Date: 2012-02-15
Figure 134 : Example showing a high resolution seismic line through the 1996 landslide in Finneidfjord,
Norway (Longva et al. 2003). Note the distinct high amplitude reflection (weak layer) correlating with the initial
sliding plane.
Like seismic data that are collected for the petroleum industry, high resolution seismics use
reflected acoustic energy to image the subsurface structure and stratigraphy. The two
techniques differ in that the petroleum industry seismic data typically penetrate several
kilometres and have resolution on the order of 10s of meters, whereas high resolution seismic
data for marine geohazard purposes might penetrate several 10s of metres but resolve features
that are 10s of centimetres thick. High resolution seismics can be obtained using several
source systems such as airguns, 'boomer' and 'sparker'. The configuration can vary greatly,
depending on the source signal, distance from source to receivers, and amount of receivers as
some of the main parameters that can be adjusted. The settings vary according to the target of
the investigation; deeper investigations require lower frequencies and generally give lower
resolution. Collection of sub-bottom profiler (SBP) data (e.g. TOPAS system) is often done
simultaneously with heavier equipment (airguns, 'boomer', 'sparker'). SBP gives better
resolution than seismic data, but poorer penetration.
In more recent years, 3D high resolution seismic data has become available to the academia
and has had a tremendous impact on our understanding of offshore mass-wasting processes.
Such data allow for accurate visualization of modern and buried landslide events in
comparable (or superior) resolution to that achieved by multi-beam systems (Figure 135). One
of the greatest strengths of the 3D seismic method is the dense, regular sampling of data over

SafeLand - FP7
Page 217 of 401
Deliverable 4.1
the region of interest. The high-spatial resolution provided by 3D seismic data allows for a
detailed seafloor analysis and an exceptional visualization of the internal geometry mass
transport deposits. This increases the possibility to erect sophisticated kinematic and dynamic
models for mass-wasting processes. Application of 3D seismic interpretation has also proven
to be a powerful tool for geohazard identification and assessment of slope failure in offshore
deepwater areas (e.g., Steffens et al. 2004; Heini and Davies 2006; Frey-Martnez et al.
2009). A current and promising research field in the study of offshore geohazards is the
integration of 3D seismic attributes (e.g., instantaneous frequency, reflection strength,
coherency) with borehole log data and insitu geotechnical tests to make predictions of
subsurface physical properties over wide areas (e.g., Pinson et al. 2008).
Figure 135 : Perspective view of a 3-D seismic volume with cut-out from the Kumano Basin 3D seismic
reflection survey (from Bangs et al. 2009).
3.3. SUMMARY AND CONCLUSIONS

During the last two decades the development of high quality seafloor imaging techniques,
differential global positioning systems and high resolution seismics have given us the
opportunity to view the underwater morphology in a realistic way. These techniques are now
routinely used in order to build precise geological and geotechnical models, to conduct detail
morphological analyses of the seafloor and to map factors of importance for slope instability.
Because of this the use of multi-beam techniques and high resolution seismic data is today
essential to any offshore geoscientific activity and positively impact on any research planning.
4. STAKES OF THE AERIAL GEOPHYSIC

Airborne geophysics has been intensively applied for exploration of raw materials within the
last decades. Recently it developed into a promising approach for landslide investigations and
rapid mapping (Nakazato et al. 2006, Pfaffhuber et al. 2010, Sasaki & Nakazato 2004, Supper
et al. 2008). An applied example is shown within the case study 13. One reason for this is the
significant technology improvement during the last 10-15 years, providing accurate and well
calibrated data.
SafeLand - FP7
Page 218 of 401

Rev. No: 2
Date: 2012-02-15
The airborne geophysical technology uses sensors that are either mounted on an aircraft or
dragged on a cable several tens of meters below the aircraft (Figure 136). Commonly used
platforms are fixed wing airplanes or helicopters. Due to the fact that mass movements are
usually located in rugged terrains and that a maximum distance between sensor and
topography has to be kept due to the rapid signal decay, helicopters are in most cases the only
choice for the acquisition platform for such surveys.
Figure 136 : Helicopter borne electromagnetic system of the Geological Survey of Austria.
One of the big advantages of airborne geophysical measurements is that large areas can be
surveyed within relatively short survey times. Additionally it is the only remote sensing
method with which not only information about the surface but also of the internal structure
can be derived. Although the use of helicopters seems to be quite expensive the method is
very effective since several different sensor systems of different kind can be combined, which
acquire all data at the same time and at the same location. Further, if a comparable amount of
data had to be acquired on the ground, the costs would be magnitudes higher as with airborne
geophysics.
Due to the significant improvements in data acquisition as well as in data processing and
interpretation during the last years more delicate results could be expected by incorporating
latest improvements in airborne technology. Therefore several test studies are planned within
the SafeLand project to advance interpretation for landslide mapping and to explore the

SafeLand - FP7
Page 219 of 401
Deliverable 4.1
prospects of this methodology and evaluate the usability of the results for fast detection and
mapping of landslides. Results, interpretations and implications will be shown in the
deliverable 4.5 of the SafeLand project, Evaluation report on innovative monitoring and
remote sensing methods and future technology.

SafeLand - FP7
Page 220 of 401

Part E Geotechnical Ground-Based Monitoring Systems
Rev. No: 2
Date: 2012-02-15
PART E
GEOTECHNICAL
GROUND-BASED
MONITORING SYSTEMS
SECTION 1: Extensometers
p. 222
SECTION 2: Inclinometers
p. 224
SECTION 3: Piezometers
p. 229
SECTION 4: Contact Earth Pressure Cells
p. 235
SECTION 5: Multi-parametric In-Place Systems
p. 236
Sections 1, 2, 3, 4 and 5
- Author: E. Abbiate1, M. Lovisolo1
- Reviewer: J. Gili2, C. Michoud3
1
. Centro Servizi di Geoingegneria S.r.l., Acqui Terme, Italia. mario.lovisolo@csgsrl.eu

2
3
Abbiate E., Lovisolo M., 2010. Geotechnical Ground-Based Monitoring Systems. In the Deliverable 4.1 of the European project
SAFELAND: Review of Techniques for Landslide Detection, Fast Characterization, Rapid Mapping and Long-Term Monitoring.
Edited in 2010 by Michoud C., Abelln A., Derron M.-H. and Jaboyedoff M. Available at http://www.safeland-fp7.eu

SafeLand - FP7
Page 221 of 401
Deliverable 4.1
1. INTRODUCTION
Geotechnical instrumentation plays an important role in monitoring field performance. It
allows to identify ground geomorphologic evolution with the following goals:
Evaluate current and future stability condition for monitoring works at risk through the
study and prediction of movements and stress evolution;
Use warning systems;
Choose the best consolidation works
Evaluate mitigation interventions
Observational method (earth structures application)
In this part will be described some specific geotechnical ground-based monitoring systems
that provide measurements of deformation (extensometers and inclinometers), groundwater
pressure (piezometers) and total stress in soil (contact earth pressure cells).
2. EXTENSOMETERS
2.1. PROBE EXTENSOMETERS
Probe extensometers are used to measure the change in distance between two or more points
within a drilled hole in soft ground, by means of a portable probe containing a reed switch.
The probe is inserted into a casing in the drill hole in which the reference points, each of
which contains an array of bar magnets, have been fixed in a way to surround the casing on
the outside. The distance between the two points is determined by measurements of probe
position. For determination of absolute deformation data either one measuring point must be
at a location not subject to deformation or its position must be determined by surveying
methods. The pipe may be vertical (providing measurements of settlement or heave),
horizontal (providing lateral deformation measurements) or inclined. Typical applications of
probe extensometers are monitoring vertical compression within embankments, settlement
alongside excavations, heave at the base of open cut excavations and lateral deformation of
embankments. They are alternatives to fixed borehole extensometers, allowing for more
measuring points and minimizing the cost of permanently installed instrumentations, but
measurements are less precise.
To go further, please read Dunnicliff (1988-1993).
2.2. FIXED BOREHOLE EXTENSOMETERS

Fixed borehole extensometers installed from ground surface may be used in soft ground or
rock and may be Single Position (SPBX) for settlement measurements at one specific
elevation or Multiple Position (MPBX) for measurements at several elevations. They monitor
the changing distance between two or more points along the axis of the borehole, without use
of a movable probe. When the location of one measurement point is determined with respect
to a fixed reference datum, the devices also provide absolute deformation data. Typical

SafeLand - FP7
Page 222 of 401

Rev. No: 2
Date: 2012-02-15
applications are monitoring deformations around underground excavation in rock and behind
the faces of excavated slopes. Fixed borehole extensometers are also used for monitoring
consolidation settlements in soil, bottom heave in open cut excavations and strain in concrete
structures. The distance from the face of the collar anchor to the end of the rod is measured
using either a mechanical or an electrical transducer (Figure 137).
To go further, please read Burland et al. (1972) and Dunnicliff (1988-1993).
Figure 137 : Fixed borehole extensometer (Dunnicliff et al., 2005).
2.3. WIRE EXTENSOMETERS

The wire extensometer is a simple and low-cost device that allows the measurement of the
relative displacement between two points, one in the landslide mass that is in motion and the
other in stable ground (the borehole has been drilled vertically through the landslide body and
crosses the surface of failure). In Figure 138, we can observe instrumentation functioning. At
the left is visualized the device as installed; at the right its illustrated the situation after some
sliding displacement of the upper mass downhill. The device consists of a protected steel wire
anchored to a fixed point inside the borehole below the slip surface of the landslide. The
opposite end of the wire is attached to a frame outside the borehole that is anchored to the
landslide surface and holds a pulley. The wire is placed around the pulley and is kept in
tension by means of a counterweight. The movement of the landslide displaces the wire with
the consequent rotation of the pulley. The amount of wire displacement is measured by a
potentiometer which provides an electrical signal that is transmitted to a data logger. One
advantage of the system application is in combination with inclinometric probe measurements
in active landslides, extending the technical limitation of the probe inclinometer (typically a
few centimetres of displacement along the slide surface could stop the probe lowered inside
the pipe, requiring a new drilling activity). In the description of Vallcebre case study,
(Corominas et al., 2000) give exhaustive explanations about wire extensometer functioning.

SafeLand - FP7
Page 223 of 401
Deliverable 4.1
To go further, please read also Kovari (1985; 1988).
Figure 138 : Scheme of the wire extensometer (modified from Corominas et al., 2000).
3. INCLINOMETERS
3.1. PROBE INCLINOMETERS
Probe inclinometers are defined as devices for monitoring deformation normal to the axis of a
pipe by means of a probe passing along the pipe (Figure 139). The casing has tracking
grooves to guide the sensing probe for orientation both parallel to and at right angles to the
axis of the excavation. After the installation of the casing, the probe (containing a gravitysensing transducer designed to measure inclination with respect to the vertical) is lowered on
a graduated cable to the bottom of the hole and winched upward, with stops at (1m, 0.5 m, or
2-foot) intervals for collection of inclination data by means of a readout unit at the ground
surface. Typical applications of the inclinometers are:
The determination of the zone of a landslide movement; however, the probe

inclinometers do not allow continuous recording of the displacements and cannot
work after displacements of only a few centimetres;
Monitoring of horizontal movements of earth dams, embankments in soft soil and

along excavations or tunnels;
Monitoring of deviations from vertical of bulkheads or retaining walls.

SafeLand - FP7
Page 224 of 401

Rev. No: 2
Date: 2012-02-15
Figure 139 : Inclinometer (Dunnicliff et al., 2005).
The probe of the inclinometers has wheels that allow it to lowering parallel the pipe stretch.
Executing two measures in the orthogonal planes its possible to determine the components of
horizontal movement of any section. After the pipe installation, a first sequence of measures
is made to define pipes inclination. Comparing these first values with successive
inclinometric measures inclination variations can be measured.
Inclinometer data measures the tilt of the probe and determines the shape of the inclinometer
casing. Figure 140 presents schematics of the inclinations that are being measured. Changes
in shape could represent ground deformation or movement.

SafeLand - FP7
Page 225 of 401
Deliverable 4.1
Figure 140 : Measurement principles (Machan & Bennett, 2008).
The field accuracy of the instrument has been evaluated on a large scale project based on
measurements taken in USA and Canada (Mikkelsen) is 7.8 mm in 30 m depth. This value
is the sum of random and systematic error (Figure 141). Random error accumulates at a rate
of square roots of the number of readings, typically is 0.16 mm for each reading. The
systematic error cumulates arithmetically with an indicative 0.11 mm per reading.
The main factors that influence data error are:
Bias shift and sensitivity drift of the transducer;
Mechanical coupling of the combination wheels and casing;
Rotation error (orientation of pipes grooves if not compensated regularly can be

significative in long boreholes);
Positioning error (depth)
Temperature drift of transducer and analog reading units.
To go further, please read Mikkelsen (2003 b).

A good indication to increase accuracy in general is considering also in the design stage,
inclinometric pipe with larger diameters available: deformations of pipe along the time could
open or break soon the gooves and the casing will be unusable.

SafeLand - FP7
Page 226 of 401

Rev. No: 2
Date: 2012-02-15
Figure 141 : Total and random errors in inclinometer data. (Mikkelsen, 2003 b)
3.2. AUTOMATIC PROBE INCLINOMETER SYSTEM

The system developed by CNR-Italy represents a technological evolution of Automated
Inclinometer System, developed to measure a more accurate and reliable depth displacements
in a landslide areas and to determine the speed of movement (Figure 142). The System,
controlled by a PC, consists of a biaxial inclinometric probe with the novelty by a integration
of a electronics for data acquisition and storage, an electric motor with encoder for precision
driving and control in continuous position of the probe. By suitably programming the
electronic controller, the probe periodically goes down into the tube, then goes up, making
some stops for the measures at the desired levels. The data is acquired and stored by
electronics on board the probe and at the end of the cycle of measurement are transmitted
wireless mode (other news) at PC for data processing. This involves the elimination of
expensive cables connecting the probe at the PC (large section) and especially susceptible to
wear (for the high number of cycles of measurements). These major technological innovations
have led to a significant reduction in size, weight and consumption, which can lead to
electricity through solar panels and battery. The system is also designed for remote control
and you can set thresholds to alert / alarm system including the speed of displacements
recorded.
To go further, please read Lollino (1992) and updatings.

SafeLand - FP7
Page 227 of 401
Deliverable 4.1
Figure 142 : Automatic probe inclinometer (courtesy of Ing Lollino, CNR-IRPI Italy).
3.3. IN-PLACE INCLINOMETERS

In-place inclinometers are typically used for monitoring subsurface deformations around
excavations when rapid monitoring is required or when instrumented locations are difficult to
access for continued manual readings and also is used to monitor landslide areas. The device
consists of a series of gravity-sensor transducers (uniaxial or biaxial), joined by articulated
rods, positioned at intervals along the borehole axis or concentrated in zones of expected
movement (Figure 143). The main advantages include rapid automatic data logging and realtime monitoring, while a disadvantage is the great complexity and expense of the hardware.
To go further, please read Dunnicliff and La Fonta (2001) and La Fonta and Beth (2002),
Machan G. and Bennett V.G. (2008).
In-place inclinometers can be used in combination with a conventional inclinometer in two
possible modes:
a)
In-place version is installed before the conventional inclinometer to define the location
of any transverse deformation, with minimal labour costs for reading. If deformation occurs,
the in-place version is removed and the zone is monitored by a conventional inclinometer
probe.
b)
Conventional inclinometer is first installed to indicate any deformation and the inplace system is installed later to minimize subsequent effort and to provide an alarm trigger.

SafeLand - FP7
Page 228 of 401

Rev. No: 2
Date: 2012-02-15
Figure 143 : In-place inclinometer (Dunnicliff et al., 2005).
4. PIEZOMETERS
Piezometers are instruments used to monitor water level in the ground and to determine the
value of pore water pressure. To measure the pressure the piezometers are sealed within the
ground so that they respond only to groundwater pressure around themselves. Piezometers
should ideally have a very short screen and filter zone: if this is located at a specific isolated
depth, the piezometer is defined punctual, while, if the piezometer has a filter on all its length,
is defined windowed. Piezometers can be grouped into those that have a diaphragm between
the transducer and the pore or joint water and those that do not. Instruments in the first group
are piezometers with pneumatic, vibrating wire and electrical resistance strain gage
transducers; instruments in the second group are open standpipe and twin-tube hydraulic
piezometers. Generally the term piezometer indicates a device that measures the interstitial
water pressure in a specific ground layer; otherwise it will be used the term observation
well that suggests an instrument that has no subsurface seals and that creates a vertical
connection between strata. When we want to define the groundwater level we use observation
wells and open piezometers, while, when we want determine the pressure of interstitial fluid,
piezometers with pressure transducers are applied. The main problem with the piezometers is
the time-lag between the variation of piezometric level in the aquifer and the respective
variation in the piezometer.
To go further, please read Dunnicliff (1988-1993) and Mikkelsen (2003 a).
4.1. OBSERVATION WELLS

An observation well consists in a plastic or steel pipe, installed in a borehole, that is slotted
(windowed) in the lower section (Figure 144). The cavity between the pipe and the
borehole is filled with sand/gravel, creating a vertical connection with the strata. So, to
prevent water infiltration from the surface, the upper part is sealed with concrete and
bentonite or quarry- clay. For the same reason it is placed at the top of the pipe a cap that has
a hole of connection with the outside in other that air pressure overtopping water level is same

SafeLand - FP7
Page 229 of 401
Deliverable 4.1
as atmospheric pressure. Applications of observation wells are limited at the case of

continuously permeable ground in which groundwater pressure increases uniformly with
depth. If the well crosses rock or soil strata with different hydraulic conductivity, it will be
registered the layer with maximum value of conductivity. To obtain the piezometric level a
measure dipmeter probe suspended to a marked cable, is lowered in the pipe. When the probe
contacts the surface of the water, a led illuminates and the beeper sounds; the operator then
can read the depth-to-water measurement (a reference point on top of the pipe or on surface
should be selected for multiple readings). The resolution is 1cm.
Figure 144 : Observation well (Dunnicliff et al., 2005).
4.2. OPEN STANDPIPE PIEZOMETERS

They are different from observation wells because, to measure the interstitial pressure of a
specific layer, the cavity is sealed with concrete and bentonite over the monitoring layer
(Figure 145). When there is the possibility of undesired infiltrations from the bottom it is
placed under the layer an impermeable pad.
The components are identical in principle to components of an observation well, with the
addition of seals. Also an appropriate stopcock cover can be used, ensuring that venting of the
standpipe is not obstructed.
The main advantages of open standpipe piezometer are:
Reliability of collected data;
Long term monitoring;

SafeLand - FP7
Page 230 of 401

Can be used to evaluate the hydraulic conductivity;
Integrity of the seals can be evaluate after the installation;
It can be converted to diaphragm piezometer;
Rev. No: 2
Date: 2012-02-15
The main limitations of open standpipe piezometer are:
Long time lag;
The device is subject to damage during the installation and by vertical compression of
soil around standpipe.
To obtain a more rapid answer in soils with permeability lower then 10-6 m/s is employed a
Casagrande piezometer. It is always an open standpipe piezometer, but it is composed by a
filter element, consisting in cylinder of porous stone, linked with the surface by the means of
a tube with a little section.
Various methods are available for reading open standpipe piezometers, most of which involve
sounding the elevation of the water surface with a probe. The most commonly used probe is
an electrical dipmeter, described before.
Moreover a pneumatic, vibrating wire or electrical resistance strain gage pressure transducer
can be inserted into the open standpipe piezometer below the lowest possible piezometric
level, thereby allowing continuous readings to be made.
The transducer can be connected to a acquisition system (data collector) and downloaded in
field manually by a notebook (RS 232 serial port or wireless) or downloaded with a GSMGPRS modem connection by a remote location (monitoring centre).

SafeLand - FP7
Page 231 of 401
Deliverable 4.1
Figure 145 : Open standpipe piezometer (Dunnicliff et al., 2005).
4.3. TWIN-TUBE HYDRAULIC PIEZOMETERS

This device was developed for installation in the foundations and fill during construction of
embankment dams. It consists of a porous filter element connected to two plastic tubes that
allow, through water circulation, the elimination of gas bubbles eventually penetrated within
the piezometer. Gas bubbles tend to get together both in the filter element and in the tubes: the
creation of discontinuity in the liquid caused not realistic pressure readings. The gas must be
removed by flushing: this is the reason for requiring two tubes. Plastic tubes are connected
with pressure measure instruments like manometers and electric transducers. The piezometric
elevation is determined by adding at the filter elevation a quantity corresponding at the
register pressure.
The main advantages of twin-hydraulic piezometer are:
Inaccessible components have no moving parts;
It is reliable;
Long successful performance record;
When installed in fill, integrity can be checked after the installation;
Piezometer cavity can be flushed;
It can be used to measure permeability.
The main limitations are:

SafeLand - FP7
Page 232 of 401

Rev. No: 2
Date: 2012-02-15
Application generally limited to long-term monitoring of pore water pressure in

embankment dams;
Tubing must not be significantly above minimum piezometric elevation;
Periodic flushing may be required.
4.4. PNEUMATIC, VIBRATING WIRE AND ELECTRICAL RESISTANCE

PIEZOMETERS
Their installation is analogue to open standpipe installation. In these piezometers the
difference is that the water crosses the porous filter entering in a cavity where it acts on a
rubber or metal membrane.
4.4.1. Pneumatic piezometer

In this type of piezometer introducing gas behind the membrane (by the means of a tube),
when gas pressure is just higher than pore water pressure, there is an extroversion of the
membrane that allow to gas to enter in a tube connected with the surface. Obtaining the gas
presence in this tube it is possible to measure balance pressure. In some models oil is used
instead of gas (Figure 146).
The main limitation of these devices is that membrane extroversion reduces volume available
to interstitial water: consequently pore water pressure increases. If the gas pressure increase is
not slow and gradual we can carry out wrong pore water pressure readings. This problem is
bigger in clay soils.
Figure 146 : Pneumatic piezometer (Dunnicliff et al., 2005)

SafeLand - FP7
Page 233 of 401
Deliverable 4.1
4.4.2. Vibrating wire piezometer

The vibrating wire piezometer has a metallic diaphragm separating the pore water from the
measuring system. A tensioned wire is attached to the midpoint of the diaphragm: the
diaphragm deflection causes changes in wire tension (Figure 147). Since the oscillation
frequency wire exchanges with its tension, pinching with an electromagnet the wire in
midpoint and measuring with another coil oscillation frequency we can determine wire
tension and then pore water pressure.
The principal limitation consists in fact that wire can work loose causing a variation of
vibration frequency. Also, the wire can be subject to corrosion phenomena.
Instead, the advantage of vibrating wire piezometer are:
-
easy readings;
short time lag;
no freezing problems;
independence of reading elevation from the height where is placed the piezometer;
possibility to measuring negative pore water pressure.
Figure 147 : Vibrating wire piezometer (Dunnicliff et al., 2005).
4.4.3. Electrical resistance piezometer

These devices are based on the concept that the resistance exchanges with length which in
turn exchanges with the vary of stress affecting the conductor. There are used especially for in
situ tests when there is the necessity to have little instruments and it is possible to remove

SafeLand - FP7
Page 234 of 401

Rev. No: 2
Date: 2012-02-15
periodically the device. They are dividing in two categories on the base of wire installation:
bonded and unbonded electrical resistance strain gage piezometers.
The advantages of electrical piezometer are the same of vibrating wire piezometer advantages,
but unbonded version can provides temperature and dynamic measurements.
5. CONTACT EARTH PRESSURE CELLS

Earth pressure cells are used to determine distribution, intensity and direction of total stress.
On the base of their employ in different kind of measurements they are subdivided in:
Embedment earth pressure cells (measurements within a soil mass, like embankment
dam or fill overlying a tunnel);
Contact earth pressure cells (measurements at the interface of a structural element, like
retaining walls, tunnels, piles and below shallow foundations).
There are two models of cells that can be used for measurements within the ground:
diaphragm cells and hydraulic cells. The first type is characterized by a rigid circular
membrane that is deflected from external soil pressure: the deformation is captured by an
electrical resistance transducer directly linked to the interior surface of the cell or by a
vibrating wire transducer. The two faces in diaphragm cells are independent and can be active
or not. If either the faces are active and are in similar contact with the soil, the measurements
are important to understand the quality of the installation. The hydraulic type, instead, is
constructed from two circular or rectangular stainless steel plates welded together around their
periphery and separated by a narrow gap filled with hydraulic fluid. It is very important that
the cell is filled with de-aired liquid and that no gas bubbles are trapped during the filling.
External pressure squeeze the two plates together creating an equal pressure in the internal
fluid. A length of stainless steel tubing connects the fluid filled cavity to a pressure transducer
that converts the fluid pressure into an electrical signal transmitted by cable to the readout
location. The active faces of hydraulic cells can be thin (when either the faces are active) or
more thick (in the case of only an active face).
The types of contact earth pressure cells are similar to those utilize within a soil mass (Figure
148). In fact, some embedment earth pressure cells can be used directly as contact cells.
However there are some exceptions. First, a hydraulic cell with a thin active face couldnt use
because its not possible to install the cell completely sticking to the structure. Second, a cell
must be used with two active faces because the measurements collected from the external
surface are more subject to errors. Contact earth pressure cells provide a direct means of
measuring total pressures, for example the combination of effective soil stress and pore water
pressure. Measurements of total stress at the face of the structural elements are not plagued by
so many errors associated with measurements within a soil mass; however, cell stiffness and
the influence of temperature are often critical. The principal requirement for the installation of
a cell at the interface of a structure is that the sensitive face of the cell is absolutely flush with
the surface of structural element.
To go further, please read Dunnicliff (1988-1993).

SafeLand - FP7
Page 235 of 401
Deliverable 4.1
Figure 148 : Contact earth pressure cell (Dunnicliff et al., 2005).
6. MULTIPARAMETRIC IN PLACE SYSTEMS

Although in the environmental field several applications are known about multiparametric
devices (probe for water quality, chemical and physical parameters) in the geotechnical field
there are only a few applications. One of the main advantage is the correlation of the
geoindicators in the same time and site conditions with a significative cost reduction for the
drilling activities, control units/transmission systems specially when piezometric,
inclinometric and extensometric measurements are required in the same area.
6.1. DMS SYSTEM

The DMS is an in place system specifically designed for the real time stability control of
landslides, cuttings, engineering works. It allows the differential monitoring of the major
physical and mechanical parameters of the soil-bedrock and of the structures in 2D/3D as well
as of the piezometric level, accelerometric variations, temperature.
DMS column is like a spiral cord composed of a sequence of hard tubular modules (hosting
sensors and AD control), connected with each other by special joints, endowed with the
degree of freedom specific for the type of measurement performed (Figure 149). Depending
on the instrumentation version, vertical, horizontal and inclined installations are possible.
D.M.S. 3D system has been specifically studied for 3D measurement in drilling holes, so as to
record deformations directly on rocks or soils, and to execute multipiezometric measurements
in the same monitoring hole. The system is composed of modules (E.D. 80 millimeter, with
packers for direct hole anchorage), linked by special 3D flexible joints with high traction
resistance (up to 100 KN) and maximal extension > 100 mm/module.

SafeLand - FP7
Page 236 of 401

Rev. No: 2
Date: 2012-02-15
Figure 149 : Left - D.M.S. 3D column. Right - Installation and continuous monitoring of D.M.S. column.
Thanks to the modular structure DMS can operate in harsh environments with continuous
adaptability to bend and twist of the casing (or borehole), also with high deformations rate. It
is possible to use a wide range of transducers on a modular basis that may be interfaced with
the same digital control unit: mono/biaxial inclinometers, extensometers, accelerometers,
piezometers, thermometers, digital compasses.
Advantages compared to traditional instrumentation are:
Simultaneous multiparametric monitoring in the same casing/borehole
2D/3D deformation analysis
Continuous in place monitoring H24 suitable for Early Warning applications (self tests
on board and functionality checks guarantee a reliable monitoring)
High adaptability of the monitoring column to deformations
High space resolution of measurements

SafeLand - FP7
Page 237 of 401
Deliverable 4.1
Easy installation in place, and possibility to recover the instrumentation after the
monitoring period
Modularity and cost reduction
For column with considerable length (>30m) installation is made easier by means of a DMS
reeler, which allows to install and retrieve the column inside the drilling hole/casing. The
robust structure of the columns allows also to transport and install the system directly inside
the casing by means of a helicopter (mountain areas and harsh environments).
To go further, please read Lovisolo and Della Giusta (2005).

SafeLand - FP7
Page 238 of 401

Part F Other Techniques
Rev. No: 2
Date: 2012-02-15
PART F
OTHERS TECHNIQUES
SECTION 1: Global Navigation Satellite Systems (GNSS, GPS )
p. 240
SECTION 2: Introduction to Core Logging
p. 247
Sections 1
- Author: J. Gili1
- Reviewer: T. Eiken2,3
1
j.gili@upc.edu
. Department of Geosciences, Universitetet i Oslo, Norway.
3
2
Gili J., 2010. Global Navigation Satellite Systems. In the Deliverable 4.1 of the European project SAFELAND: Review of Techniques for
Landslide Detection, Fast Characterization, Rapid Mapping and Long-Term Monitoring. Edited in 2010 by Michoud C., Abelln A.,
Derron M.-H. and Jaboyedoff M. Available at http://www.safeland-fp7.eu
Section 2
- Authors: C. Michoud1, L.H. Blikra2, A. Abelln1, M.-H. Derron1, M. Jaboyedoff1
- Reviewer: M. Levisolo3
1
. Institute of Geomatics and Risk Analysis (IGAR). Universit de Lausanne, Switzerland. clement.michoud@unil.ch
. knes/Tafjord Early Warning Centre, Stranda, Norway.
. Centro Servizi di Geoingegneria S.r.l., Acqui Terme, Italia.
2
3
Michoud C., Blikra L.H., Abelln A., M.-H. Derron, Jaboyedoff M., 2010. Introduction to Core Logging. In the Deliverable 4.1 of the
European project SAFELAND: Review of Techniques for Landslide Detection, Fast Characterization, Rapid Mapping and LongTerm Monitoring. Edited in 2010 by Michoud C., Abelln A., Derron M.-H. and Jaboyedoff M. Available at http://www.safelandfp7.eu

SafeLand - FP7
Page 239 of 401
Deliverable 4.1
1. GLOBAL NAVIGATION SATELLITES SYSTEMS (GNSS, GPS)

1.1. INTRODUCTION
The GPS is a radionavigation, timing and positioning system with a wide set of applications:
from air, sea and terrestrial navigation, to environmental studies, natural resources
management, GIS data capture, surveying and geodetic global measurements. By tracking the
electromagnetic waves that the GPS satellites are sending continuously to the world, the
system can obtain the antenna position (Longitude, Latitude, and Height, or X, Y, Z coordinates).
In 1995 the Global Positioning System (GPS) became fully operational. Nowadays, at least
four satellites are available worldwide, day and night. The equipment is reliable, light, cheap
and easy to use. New procedures, methods and software have been developed to assist in the
field data capture and in the post-processing. As a consequence, the GPS equipment is
progressively more and more used for a wide range of engineering applications, such as the
establishment of control points for photogrammetry and remote sensing images, positioning
and levelling of boreholes and wells, off-shore surveys, environmental studies and
monitoring.
The GPS is based on the NAVSTAR constellation of satellites, deployed by the Department
of Defence (DoD) of the United States of America, for strategic purposes. There is a parallel
development, by Russia (from the former Soviet Union), based on the GLONASS
constellation; although it is not fully deployed, the GLONASS system has been reactivated in
the last years. Nevertheless, receivers able to track satellites of both constellations already
exist. Nowadays, the european navigation and positioning system, called GALILEO, is under
implementation and advanced deployment. It will be interoperable and complementary with
GPS and GLONASS. The Chinese Beidou system is also partially deployed. Other satellite
navigation systems are announced. The common acronym for all these satellite positioning
systems is GNSS (Global Navigation Satellite Systems). Fortunately, all the GNSS systems
are very similar in concept and in operation. It is a matter of fact that most of the users speak
about GPS even when they are using also other satellites. Step by step, we may evolve to the
use of the general acronym (GNSS).
Although the GPS was first conceived for military purposes, the civilian use of the systems is
allowed and very common. In the last decade the GNSS has been used extensively for
Geodesy and Surveying, modifying and enhancing the conventional survey procedures.
1.2. BASIC PRINCIPLE

The satellites compose the spatial sector. When fully deployed, each GNSS constellation has
a basic setup of 24 to 30 satellites, deployed in several orbital planes. The orbits have mean
altitudes between 20000 and 23000 Km over the Earth surface and are approximately circular.
Each satellite completes approximately two turns around the Earth per day. They send
continuously electromagnetic sinusoidal waves to the Earth, in several carrier frequencies
within the L band (i.e., for the GPS the carrier waves are L1 and L2, with wavelengths 1 19
cm, 2 24.4 cm). These carrier waves are modulated according to digital binary codes that
have an overall appearance of randomly distributed zeroes and ones. For this reason, they are
SafeLand - FP7
Page 240 of 401

Rev. No: 2
Date: 2012-02-15
called pseudo-random codes. For instance, the GPS satellites send two codes: the C/A or
Coarse Acquisition code, only on the L1 carrier, and the P or Precise code (usually
encrypted), on both L1 and L2. New GPS satellites will have a C code also on L2.
The GNSS system calculates the user position using a set of measurements called the
observables. They consist of data derived from the electromagnetic waves received from each
satellite. Dependent on the type of receiver and purpose, the computations can be made using
either the code or the phase of the carrier, or a combination of both. Positioning can be in real
time or with post-processing. The codes are widely used in navigation and in coarse
positioning (10 m typical precision, Figure 150).
Figure 150 : GPS/GNSS three-dimensional spatial multilateration of a point based on pseudo-ranges (PRi)
derived from code observation with different satellites (SVi). (, , h) are respectively Longitude, Latitude and
the height of the point above the reference ellipsoid. (X, Y, Z) are the global geocentric cartesian co-ordinates
(Gili et al., 2000)
Sophisticated geodetic receivers are used when higher accuracy is required, as in monitoring.
The computations are based in the phase of the carrier wave, no longer in the pseudorange
measured from the code. Additionally, double frequency or dual frequency receivers are
advised. Several (two or more) receiver + antenna have to track simultaneously the carrier
waves, and register the change in phase (phase angle) between the satellite and the antenna
with time. The calculations are more involved, with algorithms (single differences, double
differences, etc.) extending over observation time, usually from several minutes to several
hours. The clue point is the ambiguity resolution, that is, the determination of the integer
number of complete wavelengths, N, that fits between a given satellite and the antenna (see
Figure 151). This operation can be identified with an initialization period that has to be spent
in some kinematic methods. The most common processing procedures actually compute
relative positions of the points in the space. The vector that links two station points is called
baseline (Figure 151). A new point (R) is positioned by adding the baseline to the coordinates of a previously known one, called base point (B). Often, these calculations are
carried out after the fieldwork, during the post-processing, merging the data files from several
user receivers. Thus, the results may achieve very high precision, up to few mm.

SafeLand - FP7
Page 241 of 401
Deliverable 4.1
Figure 151 : GPS/GNSS carrier phase measurements. D is the distance between the antenna and a given
satellite, the wavelength, the phase angle and N an integer number (Gili et al., 2000).
1.3. FIELD PROCEDURES

There are several procedures or methods for data capture and treatment depending on the
purpose and precision required for the GNSS survey:
Considering the number of receivers involved, the positioning may be absolute or

autonomous positioning, when using one single unit (low precision), and relative
positioning when two or more units are used coordinately (observing at the same time
and sharing data). The term differential GPS (DGPS) refers to a medium precision
mode based on the coarse code: the users receiver derives the position as in the
Figure 150, but enhancing the pseudo-ranges with a set of corrections supplied by an
infrastructure of reference stations that covers the area.
But, in all the applications requiring high precision, two or more units (receiver plus
antenna) observing the carrier phase are needed. In some surveying cases, the
reference stations might belong to a network of base stations, public or commercial,
spread over the territory. For monitoring applications, however, it is preferred to have
the base station(s) in a stable outcrop, very close to the site, as the precision degrades
with the distance between stations.
According to the receiver movement, the surveys may be static, kinematic, or a

combination of both. Several procedures are available, the Static method, the FastStatic, the Pseudo-Kinematic, the Stop-and-Go, and others. Measurement of landslide
displacements can be undertaken by means of either static and/or kinematic methods.
The choice depends on practical local considerations, namely, the accessibility, the
number of points, the precision, the obstacles and the distance from point to point.
The first precision GPS method was the Static one, used for the observation of
geodetic networks covering hundreds of kilometres; the logging sessions last for
several hours in order to achieve the target precision. Nowadays, the most productive
methods available for determining single points with precision of millimetres or

SafeLand - FP7
Page 242 of 401

Rev. No: 2
Date: 2012-02-15
centimetres, are the Fast-Static (FS, also called Rapid-Static) and the Real Time
Kinematic (RTK), see Figure 152. The first one is a method with post-processing, an
evolution of the former Static positioning. In a typical case, with baselines around 1
km and 6, 5 or 4 common satellites in the sky, each session of logging needs around
10, 15 or 20 minutes respectively (although it is advisable to register some time in
excess). In the RTK method, the information of code and carrier phase observables
received at both ends of the baseline (Base station and Rover station, Figure 152) is
merged to compute the precise position on the spot. For this, our Base must transmit
some information for each observation epoch to the Rover. The communication Baseto-Rover can be by radio (i.e. a UHF/VHF data link, quite directional), or via GSM,
GPRS or internet.
The main characteristics of the quoted methods are summarized in the Table 21.
It is also available the RTK-Network method: the local Base is substituted by several
permanent stations that constitute a network surrounding the area. Through
GSM/GPRS or via the internet, the rover can connect to this general network to obtain
some net-interpolated solutions (from the so-called Virtual Reference Station, VRS).
However, with the present networks density and availability, this quite new method is
less suitable for a precise monitoring as the precision degrades with the distance to the
real Base.
Figure 152 : Fast-Static (F.S., or Rapid Static) and Real Time Kinematic (R.T.K.) GNSS methods. The station
over a known fixed point is referred as Base, and the receiver that moves from point to point as Rover (Gili et
al., 2000).

SafeLand - FP7
Page 243 of 401
Deliverable 4.1
Table 21 : Main characteristics of main GNSS/GPS methods (modified from Kahmen, 1997, and Gili et al.,
2000).
Method
Obs.time, per
Post-process
Strength
Typical max.
Typical
point, after
against lost of
baselines
baseline plani-
initialisation
signal
(km)
metric errorb, c
General Static
1 to several hours
Yes
robust
50 to 100 a
(51)mm+(10.1)ppm
Fast-Static or
8 to 20 min
Yes
robust
15 to 20
5mm+1ppm
1 to 10 s
No
sensitive
10
10mm+2ppm
Rapid-Static
R.T.K.
a)
Easily baselines can reach about 50 km. Depending on observation time, type of postprocessing and required precision, it may be extended up to thousands of km. b) The vertical
or elevation error is roughly 1.5 times the planimetric one. This factor can increase up to 3 or
4 with the latitude. c) 1 ppm means one part per million, or one additional millimetre per
kilometre of measured baseline.
Depending on when the solution is obtained, the procedure may be in real time (when
the position is obtained in the field, the RTK for instance), or with post-process (as the
FS). As discussed previously, the post-processing methods are used only in certain
precision procedures; special software (with appropriate algorithms, time averaging,
filters and network adjustment) is run in order to obtain the final position.
1.4. SOME CONSIDERATIONS ON THE USE OF GNSS IN LANDSLIDE

MONITORING AND ITS PRECISION
Field surveys are usually carried out with a given frequency (i.e. monthly, weekly) with
removable GPS-GNSS stations. The difference between the present values and the initial coordinates, gives the movement of the target. The main results of a GNSS positioning are the
coordinates of the occupied points at a certain time. Then, we can obtain displacements as
differences of positions along the time; velocities by derivatives; accelerations by double
derivatives, etc. The results are discontinuous over time, and related to the cumulative
movements of surface points.
Although the costs is still high, it is also possible to automate the procedure for the continuous
monitoring of the displacements with GPS stations working permanently on the slide area
(Malet et al., 2002). This solution, among others, is however justified to ensure the
monitoring of sites with high risk for the people downhill, and is the only GPS setup eligible
as an early warning system.
In the usual manual GPS surveys for monitoring, the precision of the positions is highly
dependent on both the equipment and, fundamental, the selected method. When using a
relative GPS high precision method we can reach cm to mm accuracy, depending on the total
log time, the number of satellites under tracking, their geometry (the spatial distribution on
the users horizon...), and on the network configuration (number of receivers working
together, redundancy of the net, fixed base points...). Usually, atmospheric delays, the orbital
parameters and the co-ordinate transformation are the most important factors in terms of
SafeLand - FP7
Page 244 of 401

Rev. No: 2
Date: 2012-02-15
positioning error. Fortunately, the effect of these factors can be avoided or reduced when
measuring displacements with GPS. Using two or more receivers -relative positioning-, at
points located, for instance, up to 20 km, these errors have a similar influence in both baseline
ends, and cancel.
When working with centimetre precision at several tents of kilometres, and without direct line
of sight, even in the night or raining, it is clear that the advantage of satellite positioning over
conventional surveying increases as the working area enlarges. On the other hand, the antenna
must have a good view of the sky: the buildings, the forest and the slopes of narrow step
valleys are obstructions to the direct reception of the radio signal that might difficult the
positioning.
Another important source of error is the setting-up of the antennas over the physical marks in
the field. Due to access, to environmental considerations and/or budget constraints, the
marking of the control points is quite light. The geodetic-type concrete cylinders are seldom
used. Instead, some special antenna mountings can be used to allow the forced centring
(Schfer 2008). In other situations, it is better to station the antenna over a surveying tripod
than sustaining it over the telescopic pole (Figure 153).
Figure 153 : Different antenna set-up: antenna directly over a rock outcrop; setting-up with tripod and optical
plummet; antenna over a telescopic pole.
Another issue when talking about precision of the results is the postprocessing software in
use. There are two fundamental approaches. The first one are the Standard or commercial
GNSS software, provided by the Receiver manufacturer, like TGO from Trimble, the LGO
from Leica, the TopconTools from Topcon, etc. They use to provide excellent results for
small and medium networks as the ones deployed usually around the landslide area. They are
quite user friendly, but the baselines are processed with the default parameters and the user
has few chances to choice or change them. On the other hand, we have the scientific GNSS
software like BERNESE (Univ. of Berna) or GIPSY (JPL) or the PPP (Precise Point
Positioning) techniques. There, the user might select hundreds of parameters and models
(troposphere, ionosphere, meteo, precise orbits, antenna offsets, etc). However, this kind of
software must be run by highly skilled operators, besides they need very long logging
sessions (day to week). Fortunately, this high level software is only essential for long
baselines (>30 km), and wide networks.
No big networks of control points are usually observed around a landslide monitored with
satellite positioning. A common situation, due to accessibility, personnel and equipment
SafeLand - FP7
Page 245 of 401
Deliverable 4.1
limitations is to use a set of radiations from one base point. It is suggested, if possible, to setup a second base receiver in another fixed point; this will considerably strengthen the results.
The reoccupation (at different hour of the day) and the incorporation of some fixed points (for
result checking) are also advisable strategies.
In the Table 22, the actual planimetric (X-Y) and elevation errors for a typical landslide
monitoring setup are given, when using the FS or the RTK method (Gili et al., 2000).
Table 22 : Example of planimetric and elevation errors for a typical GPS landslide monitoring (modified from
Gili et al., 2000)
RMS (1 , 68%)
GPS method
Planimetry
Elevation
FS or Rapid Static
12 mm
18 mm
RTK
16 mm
24 mm
a)
The GPS observation consists in a single radiation from one base point. The baselines range
between 1 and 1.5 km. The errors are distributed normally. The setting-up methods
considered are: tripod in the base, and telescopic pole with bubble in the rover side.
1.5. EXAMPLES
In the last decade landslide monitoring with GPS has experienced a rapid development
because it has demonstrated to be a reliable technique to follow the slide motion especially for
slow movements.
An early result was reported by Bonnard et al. (1996) in instrumented landslides located in the
Swiss Alps: using several receivers simultaneously, they obtained a precision of about one
centimetre. This precision was higher than that achieved there using ordinary surveying
techniques of triangulation.
Brunner et al. (2007) describes a long experience of monitoring a deep seated mass movement
in Gradenbach (Austria) with relative GPS. The authors reported motion discrepancies (1 )
of 4 mm in horizontal direction and 8 mm in vertical direction during the last seven years
(17 GPS surveying campaigns with at least 48 hours of observation; choke-ring antennas,
post-processing software Bernese 5.0, etc).
1.6. TRENDS
As previously quoted, it is possible nowadays to install permanent GPS stations in the
landslide acquiring continuously the GPS observables (Malet et al., 2002). The data can be
remotely transmitted to the Early Warning Centre, where it can be processed in near real time
(hourly sessions, for instance).
According to ClimChalp (2008), continuous monitoring with low-cost GPS receivers is
currently object of research. Another way to adjust the installation budget is to implement a
multi-antenna scheme, where one receiver is able to sequentially log or process the GPS
signal acquired by several antennas spread in the area (Chen et al., 2005).

SafeLand - FP7
Page 246 of 401

Rev. No: 2
Date: 2012-02-15
2. INTRODUCTION TO CORE LOGGING

All techniques described in the previous chapters are employed to investigate landslide
surfaces and sub-surfaces by geophysical measurements from plane or at the ground surface.
In most cases, there are arising a series of questions regarding the interpretation of such data,
and there will be a need for confirmation and comparison by direct measurements in the
subsurface. This part gives a short introduction of different methods and instrumentation for
subsurface investigations (cf. Table 23). Three main references on this topic are
recommended: McGuffey et al, in Turner and Schuster (1996), the US Army Corps of
Engineers (2001) and Bell (2007).
Borings should be performed in order to obtain the following main data:
Rock or soil description and characteristics
Sliding surfaces
Geomechanical characteristics
Hydrogeological parameters
Displacements
Following techniques can be used:
Drilling registrations and observations (Figure 156)
Core logging (Figure 154 and Figure 155)
Geophysical borehole logging
In situ testing, for example hydrological pumping
In place equipments.
Table 23 resumes some of the data that can be collected in connection with boreholes. The US
Army Corps of Engineers (2001) gives details on the listed techniques and tests, such as the
famous Standard or Cone Penetration Test (SPT, CPT) which assess the shear strength of
soils. McGuffey et al, in Turner and Schuster (1996), provides more details on geophysical
well logging, such as the Neutron-Neutron log which gives information on natural moisture
content above water table and porosity below water table. Other geophysical logging sensors
give data on P and S velocity, resistivity, conductivity, density, gamma and temperature.
Borehole investigations are one of the most important datasets for investigating large
landslides, and are crucial for detecting subsurface deformation and sliding surfaces. In
addition to geophysical logging and hydrological testing, two areas of investigations are
important:
Core description and logging

SafeLand - FP7
Page 247 of 401
Deliverable 4.1
In-place equipment for investigating deformations and pore pressure (inclinometers,

extensometers and piezometers)
Ideally one should perform core drilling giving rock cores, were the rock characteristics and
possible sliding planes can directly be documented and studied. Examples of two sliding
planes at the knes rockslide are shown in Figure 154 and Figure 155 (Ganerd et al., 2007).
Destructive drillings are often performed in landslide studies, and then rock characteristics
can be studied by camera or televiewer equipment.
Figure 154 : Core logging in knes. Lower borehole shows a crushed zone at 115 m depth, possible
representing a sliding zone.
Figure 155 : Core logging in knes. Middle borehole, 32 m depth: crushed and brecciated zone with silt and
clay which represents the upper sliding zone, also documented by displacement in DMS column (Differential
Monitoring of Stability, in place instrumentation).
The documentation of the subsurface displacements is the next important step in the
subsurface investigations. This can be done by different approaches (cf. chapters 2.F.):

SafeLand - FP7
Page 248 of 401

Rev. No: 2
Date: 2012-02-15
Manual inclinometric probes, Inclinometer casing are established in the borehole and
measurements are taken manually at different time periods.
In-place inclinometers or extensometers measuring continuous. This can be combined

also with piezometers measuring at different levels. The challenge is to find the best
localization of sensors in order to measure the displacements.
DMS columns. It allows the differential monitoring of displacement 2D/3D along the
overall landslide/rockslide domain as well of the major physical and mechanical
parameters piezometric level, pore pressure, accelerometric variations, temperature.
System is installed in place and measurements are in continuous.
Figure 156 : example of drilling report. The header indicates technical information on the boring, as the
sampling method and the equipment. Then drillers reported lithologies, moistures, voids and water tables. (US
Army Corps of Engineers, 2001)

SafeLand - FP7
Page 249 of 401
Deliverable 4.1
Table 23 : Methods and type of data that can be achieved by performing boreholes.

SafeLand - FP7
Page 250 of 401
3. Application to Landslides: Case Studies
Rev. No: 2
Date: 2012-02-15
APPLICATION TO LANDSLIDES:
CASE STUDIES

SafeLand - FP7
Page 251 of 401
3. Application to Landslides: Case Studies
Rev. No: 2
Date: 2012-02-15
3. APPLICATION TO LANDSLIDES: CASE STUDIES

Passive optical sensors
1. Landslide monitoring - Monitoring Tessina landslide, Italy, from optical remotely sensed
imagery.
2. Landslide mapping - Landslide mapping by textural analysis of ATM data: a case study
from the LOS Vlez district, south eastern Spain.
Active optical sensors
3. Landslide characterization and monitoring; displacement detection; collapse

characterization - Using TLS for rockslide monitoring. Case study of the Eiger collapse, Swiss
Alps.
4. Structural analysis; Discontinuity orientation - Structural analysis using color coded
information derived from 3D point clouds.
5. Rock face monitoring: rockfall detection; Spatial prediction - Using TLS for rockfall
detection and prediction.
6. Susceptibility assessment at regional scale - Susceptibility assessment at regional scale
based on areal LIDAR datasets. Application to the canton of Vaud, western Switzerland.
7. Landslide inventory and susceptibility mapping at regional scale - The application of
LIDAR-derived images for landslide inventory and susceptibility mapping in the Flemish
Ardennes, Belgium.
Active microwave sensors
8. Landslide detecting - Characterization and analyses of surface deformation, mass

movements and alpine region lifting with PSInSAR techniques/methods.
9. Landslide detection and mapping at regional scale - Landslide detection and mapping at the
basin scale by means of radar-interpretation.
10. Landslide characterization and monitoring at local scale - Integration of PSInSAR data
and ground-based instrument measurements for landslide characterization and monitoring at
the local scale.
11. Slow-moving landslide detection, mapping, characterization, monitoring - A new
approach to the use of DInSAR data in landslide studies at different scales: the case study of
National Basin Authority of Liri-Garigliano and Volturno rivers.

SafeLand - FP7
Page 253 of 401
Deliverable 4.1
Geophysical investigations
12. Rock slope monitoring - Microseismic monitoring at the unstable rock slope at knes,
Norway.
13. Landslide investigations - Landslide of Rindberg / Sibratsgfll investigated with airborne
geophysical survey.
Geotechnical investigations
14. Geotechnical investigations and instrumentation - Site-specific geotechnical investigations

for landslide characterization and detection. Application at the knes rockslide, western
Norway.
15. Landslide monitoring - From early warning to site-specific kinematic analysis: the case of
Bagnaschino landslide.
Multidisciplinary approaches
16. Ground surface displacement measurement - Detecting and mapping landslides in Salazie
area, La Runion Island, using remote sensing techniques.
17. Long term monitoring, characterization and modelling - Vallcebre landslide, Spain,
monitoring, from wire extensometers to GB-InSAR.

SafeLand - FP7
Page 254 of 401
3. Application to Landslides
Case Study 1
Rev. No: 2
Date: 2012-02-15
CASE STUDY 1:
Title: Monitoring Tessina landslide, Italy, from optical remotely sensed imagery.
Application: Landslide monitoring.
Technique: Multi-temporal image change detection.
Main references: This case study is based on following publication: Hervs et al. (2003).
Contributors: JRC (Hervs J., Barredo J.I., Rosin P.I., Van Den Eeckhaut M.).
ABSTRACT
In this study an image-processing method to map and monitor landslide activity using multitemporal optical
imagery is proposed. The method entails automatic change detection of suitably pre-processed (geometrically
registered and radiometrically normalised) sequential images, followed by thresholding into landslide-related
change pixels. Subsequent filtering based on the degree of rectangularity of regions followed to eliminate pixel
clusters corresponding to man-made land use changes.
The application of this method is illustrated in the complex Tessina landslide in the Eastern Italian Alps. The study
focused on discriminating the changes caused by the major reactivation of April 1992. The method has been
devised for optical remote sensing imagery in general, but due to the absence of high-resolution satellite imagery
covering that period, digital images derived by scanning existing aerial photograph diapositives at 1 m x 1 m
resolution have been used. The results show that the tested threshold method is able to classify image pixels
according to landslide activity conditions.
1. INTRODUCTION
Collecting information on landslide activity over wide
areas is a crucial task for landslide hazard
assessment. Field techniques, despite being very
precise, are usually not sufficient to achieve this goal,
since they mostly provide point-based measurements.
Mainly because of its synoptic view and its capability
for repetitive observations, optical (visible-infrared)
remotely sensed imagery acquired at different dates
and at high spatial resolution can be considered as an
effective complementary tool for field techniques to
derive such information (Soeters and van Westen,
1996).
The insufficient spatial resolution provided until very
recently by most spaceborne earth observation
systems inhibited the use of satellite remote sensing in
the optical region of the electromagnetic spectrum
(Soeters and van Westen, 1996; Mantovani et al.,
1996). On non-stereoscopic digital imagery, efforts
have mainly concentrated on extracting possible
indirect landslide indicators such as land cover
disruption patterns, specific sun-shading features of
hummocky surfaces and scarps and atypical
lithological occurrence patterns (e.g. McKean et al.,
1991; Hervs and Rosin, 1996; Hervs et al., 1996).
Therefore the launch of e.g. IKONOS, Quickbird, Eros
and ALOS, Geoeye satellite imagery at very high
spatial resolution had opened new perspectives for
monitoring landslides.
The objective of this study was to test and image
processing method to map ground-surface changes
caused by landsliding, using multitemporal high-

SafeLand - FP7
resolution optical remotely sensed imagery. The

method
includes change detection and thresholding of suitably
pre-processed images acquired before and after the
landslide event.
For this test, the Tessina landslide in Belluno province,
Northeast Italy, initiated in 1960 and reactivated
several times afterwards (Pasuto et al., 1993; Angeli et
al., 1994; Turrini et al., 1994; Mantovani et al., 2000)
has been selected. The Tessina landslide has a large
source area with rotational slides and mudflows
depositing the material onto a wide flat area making up
the upper accumulation zone. From there, the material
feeds downstream through a narrow and steep canal
filling up the valley and approaching the villages of
Funs and Lamosano. The landslide developed in the
Tessina valley between altitudes of 1220 and 625 m
asl, with a total longitudinal extension of nearly 3 km
and a maximum width of about 500 m. This study
focusses on the latest major reactivation in 1992 when
the easternmost part (area 70,000 m2 with a volume of
3.106 m3) of the source area collapsed as a result of
high precipitation and snow melting.
2. MATERIAL AND METHODS

Since high spatial resolution satellite imagery was not
available at the time of this event, simulated IKONOS
digital imagery at 1 m x 1 m resolution from existing
aerial photographs acquired in 28/9/88 and 7/10/94 at
1:75,000 scale has been used in this experiment. The
images were orthorectified and radiometrically
corrected (Hall et al., 1991; Hill and Sturm, 1991). We
Page 255 of 401
Deliverable 4.1
refer to Hervs et al. (2003) for more detailed

information.
In remote sensing, a number of methods for land cover
change detection are possible (Singh, 1989). In
general, two basic approaches can be considered:
either analysis of independently classified images (i.e.
post-classification techniques) or simultaneous
analysis of multitemporal images (i.e. pre-classification
techniques) followed by thresholding to discard
differences possibly related to non-surface changes
(Singh, 1989). For panchromatic imagery, the latter
approach appears more suitable. Although a number
of image processing techniques can be applied to
derive the simple pixel intensity change image, it is
usually accomplished either by image differencing or
image ratioing. Image ratioing involves a simple ratio
operation between the values of corresponding pixels
of two registered images of different dates (on a bandby-band basis if images are multispectral). With this
operation, more weight is progressively given to pixel
differences closer to zero. As a result, the numeric
scale is neither symmetrical nor linear, thus making
the subsequent thresholding process difficult. To
facilitate thresholding, a log transformation can be
applied to the ratio image to make the scale linear and
symmetrical about zero. Image differencing in turn
involves the subtraction of pixel values between both
images. Unlike the ratio technique, the difference
method highlights all cases of change to the same
extent. To map ground surface changes due to
landsliding
using
radiometrically
normalised,
multitemporal remotely sensed orthoimagery, we have
developed a method based on digital change detection
techniques. The method encompasses two main
steps: image differencing and subsequent thresholding
into classes of changes in connection with landslide
activity. The thresholding procedure here is essential
to discriminate true change image pixels by filtering
out from the change image most of the noise due to
scene changes between the image acquisitions and
residual differential illumination and geometric
misregistration effects.
Bilevel image thresholding into simple change and
no change classes most often requires the
occurrence of two distinct modes in the image intensity
histogram. In addition, these two peaks should not be
too similar in size and they should have approximately
normal distributions (Rosin, 2002). However, in
change detection analysis of remotely sensed images
over inland areas, the image histogram may often
contain only one obvious peak, as is the case for the
Tessina images. This causes difficulties for most
popular thresholding algorithms (Rosin, 2002). In order
to correctly threshold difference images with unimodal
histograms, Rosin et al. (2000) tested a number of
algorithms on images of the Tessina area. From these,
the best results were obtained with a relatively simple
thresholding algorithm proposed earlier by Rosin
(2001). The algorithm consists of thresholding at the
corner of the histogram. To find this corner, a straight
line is first drawn from the peak (largest bin) to the

SafeLand - FP7
high end (first empty bin following the last filled bin) of
the histogram (Fig. 1). The threshold point is selected
at the pixel intensity value that maximises the
perpendicular distance between the line and the
histogram distribution. The distance along the
perpendicular from a histogram point (xi, yi) to the
line (x1, y1)->(x2, y2) is derived as follows.
( y1 y2 ) xi ( x1 x2 ) yi x2 y1 x1 y2
( x1 x2 ) 2 ( y1 y2 ) 2
eq.1
Not only is this algorithm simple and robust, but it has

been shown to work effectively in a variety of
applications such as edge, motion and texture
thresholding. The thresholding is applied to the
absolute pixel difference values (i.e. negative values
are made positive). The resulting binary image can
then be classified into positive and negative categories
of change by checking for each above threshold value
the sign of the original signed difference image.
Depending on the spectral characteristics of land
cover and underlying soil/bedrock on unstable slopes,
such a classification may enable a better identification
and monitoring of slide masses.
Fig. 1. Difference image histogram showing threshold value

(T) corresponding to the detected corner.
Simple thresholding may often not be sufficient to

identify changes truly caused by landsliding. An
additional processing step is needed to eliminate
undesirable
noise.
Possible
polygonal-shaped
manmade features (e.g. random land use changes,
crop parcels, buildings and road segments) were
filtered out using rectangularity filtering preceded by
area filtering to avoid selection of very small pixel
groups as perfectly rectangular (Rosin et al., 2000).
Page 256 of 401
Case Study 1
3. RESULTS
The method entailed: (1) the creation of an image
expressing the differences in pixel brightness between
the two input images, (2) preliminary thresholding of
such an image using the histogram corner algorithm
(Rosin, 2001; Fig. 1), (3) splitting the thresholded
image into positive changes (i.e. increase in pixel
brightness from the older image to the newer one) and
negative changes (i.e. decrease in pixel brightness)
and (4) filtering out rectangular-shaped image blobs
(that is, clusters of adjacent pixels).
Rev. No: 2
Date: 2012-02-15
The final result of the image processing sequence is

shown in Fig. 2. Here, positive pixel intensity changes
are colored in red. Within the Tessina landslide, these
represent ground change patterns mostly associated
with the reactivation occurred in 1992, as documented
by Pasuto et al. (1993) and Angeli et al. (1994), and
illustrated by both on-site photographs and low-altitude
aerial photographs taken at the time of the
reactivation.
Fig. 2. Surface changes in the Tessina landslide between 1988 and 1994 illustrated on the 1994 orthophotograph. Legend: (1)
within the landslide body, positive pixel intensity changes represent both new soil outcrops and remobilised soil as a result of
landslide reactivation; zones highlighted are explained in the text; (2) negative pixel intensity changes due to vegetation growth
or soil moisture increase. Most changes outside the landslide correspond to land use change. Coordinates in UTM.

SafeLand - FP7
Page 257 of 401
Deliverable 4.1
Five major change zones can be observed. First, the

zone detached in 1992 from the northeastern landslide
crown, which produced a landslide surface increase of
70,000 m2 (zone A in the figure), thus showing fresh
soil outcrops with higher pixel intensity in the 1994
image. New soil outcrops together with fresh
remobilised material can also be observed along the
flanks of the narrow steep canal (zone B) because of
remobilisation and depletion of the previous mudflow
in this zone and subsequent minor sliding of the main
flanks. Zone C, adjacent to Funs, appears irregularly
reworked by small mudflows and remobilisation of the
former mudflow, as well as by man-made drainage
channels (depicted as linear blobs) dug following the
1992 event. Zone D shows a mudflow running off the
right flank of the landslide body, opposite to the village
of Funs. Here, the higher brightness fresh slide mass
partly filled a vegetated small depression originally
made up by a side stream, after breaching a levee.
Much of the former mudflow zone southwest of Funs
appears, however, not affected by the 1992
movements. In this zone, the remobilised material,
already highly fluidised, was funneled along the
previously dug central drainage channel into the
narrower valley section below, near the village of
Lamosano, where the new runout material is clearly
shown (zone E). New linear structures consisting of
protection walls and concrete partly paving the
Tessina stream floor, built in 1992, are also apparent
at the north end of Lamosano because of their
significant pixel brightness increase in the 1994 image.
Areas in yellow, which represent negative pixel
intensity changes, appear mostly associated with
either soil moisture increase or simple land cover
changes, including the growth of natural vegetation on
temporarily stable sectors of the landslide body. While
vegetation growth may be an indicator of ground
stability, soil moisture conditions (which can change
dramatically over short periods of time) may conceal
the brightness properties of the underlying ground,
especially on panchromatic images.
Apparent changes detected outside Tessina landslide
are for example changes related to vegetation
regrowth on stable sectors of gully walls northwest of
Lamosano.
4. CONCLUSIONS
A method based on automatic digital image change
detection and thresholding techniques using optical
remotely sensed imagery has been successfully
applied for monitoring landslide-related ground surface
changes (mainly land cover and exposed soil
conditions). Although tested with very high resolution
aerial photographs, the method is suited to monitor
landslide occurrence and activity over extensive areas
using the latest generation of very high resolution
satellite imagery. It is particularly useful for monitoring
surface changes caused by moderate-velocity
landslides, even if involving high internal deformation,

SafeLand - FP7
in humid and heavily vegetated areas, where the

application of more accurate radar interferometry
techniques is often not possible.
REFERENCES
Angeli, M.-G., Gasparetto, P., Menotti, R.M., Pasuto,
A., Silvano, S., 1994. A system of monitoring and
warning in a complex landslide in Northeastern
Italy. Landslide News 8, 1215.
Hall, F.G., Strebel, D.E., Nickenson, E., Goetz, S.J.,
1991. Radiometric rectification: toward a common
radiometric response among multidate multisensor
images. Remote Sensing of Environment 35, 11
27.
Hervs, J., Rosin, P.L., 1996. Landslide mapping by
textural analysis of Daedalus ATM data. Proc.
Eleventh Thematic Conf. Applied Geologic Remote
Sensing, Las Vegas, Nevada, vol. 2. ERIM, Ann
Arbor, Michigan, USA, pp. 394402.
Hervs, J., Rosin, P.L., Fernndez-Renau, A., Gmez,
J.A., Len, C., 1996. Use of airborne multispectral
imagery for mapping landslides in Los Velez
district, south eastern Spain. In: Chacon, J.,
Irigaray, C., Fernndez, T. (Eds.), Landslides.
Balkema, Rotterdam, pp. 353 361.
Hervs, J., Barredo, J.I., Rosin, P.L., Pasuto, A.,
Mantovani, F., Silvano, S., 2003. Monitoring
landslides from optical remotely sensed imagery:
the case history of Tessina landslide, Italy.
Geomorphology 54, 63-75.
Hill, J., Sturm, B., 1991. Radiometric correction of
multitemporal Thematic Mapper data for use in
agricultural land-cover classification and vegetation
monitoring. International Journal of Remote
Sensing 12, 1471 1491.
Mantovani, F., Pasuto, A., Silvano, S., Zannoni, A.,
2000. Data collection aiming at the definition of
future hazard scenarios of the Tessina landslide.
International Journal of Applied Earth Observation
and Geoinformation 2 (1), 3340.
McKean, J., Buechel, S., Gaydos, L., 1991. Remote
sensing and landslide hazard assessment.
Photogrammetric
Engineering
and
Remote
Sensing 57 (9), 1185 1193.
Pasuto, A., Silvano, S., Bozzo, G.P., 1993. The
Tessina landslide (Belluno, Italy). In: Panizza, M.,
Soldati, M., Barani, D. (Eds.), Proc. First European
Intensive Course on Applied Geomorphology,
Pubblicazioni Istituto di Geologia, Universita degli
Studi di Modena, Italy, pp. 63 69.
Pasuto, A., Silvano, S., Tecca, P.S., Zannoni, A.,
1999. Convenzione tra la Regione del Veneto e l
IRPI-CNR per lo studio della frana del Tessina in
Comune di Chies dAlpago (BL). Relazione finale.
IRPI-CNR, Padova, Italy.
Rosin, P.L., 2001. Unimodal thresholding. Pattern
Recognition 34 (11), 20832096.
Page 258 of 401
Case Study 1
Rosin, P.L., Hervs, J., Barredo, J.I., 2000. Remote

sensing image thresholding for landslide motion
detection. Proc. 1st Int. Workshop on Pattern
Recognition Techniques in Remote Sensing,
Andorra, 10 17.
Singh, A., 1989. Digital change detection techniques
using remotely sensed data. International Journal
of Remote Sensing 10, 989 1003.
Soeters, R., van Westen, C.J., 1996. Slope instability
recognition, analysis and zonation. In: Turner, K.,
Schuster, R.L. (Eds.), Landslides Investigation and
Mitigation, Transportation Research Board, Special
Report
247,
National
Research
Council,
Washington DC, pp. 129 177.

SafeLand - FP7
Rev. No: 2
Date: 2012-02-15
Turrini, M.C., Abu-Zeid, N., Semenza, E., Semenza,

P., El-Naqa, A., 1994. New studies on the Tessina
landslide (Belluno, Italy). Proc. 7th Int. IAEG
Congress. Balkema, Rotterdam, pp. 1667 1675.
Page 259 of 401
Case Study 2
Rev. No: 2
Date: 2012-02-15
CASE STUDY 2:
Title: Landslide mapping by textural analysis of ATM data: a case-study from the Los Vlez
district, south eastern Spain.
Application: Landslide mapping.
Technique: Semi-automated texture enhancement using band 11 (8.5 m - 13.5 m;
capturing thermally emitted energy) from Daedalus AADS 1268 Airborne Thematic Mapper
(ATM) data.
Main references: This case study is based on following publication: Hervs and Rosin
(1996), Hervs et al. (1996) and Hervs and Rosin (2001).
Contributors: JRC (Hervs J., Rosin P.I., Van Den Eeckhaut M.).
ABSTRACT
In this study two statistical approaches to semi-automated texture enhancement and discrimination for landslide
mapping in semi-arid, sedimentary terrain from Daedalus ATM data were evaluated. A supervised texture
discrimination technique was applied, based on calculating similarity between a reference texture spectrum
obtained from training samples and spectra from moving image windows. The results are compared with those
from interpreting a set of popular texture measures from the literature, derived from grey level co-occurrence
matrix statistics. In this comparison, interpretation is facilitated by statistical selection of the best combination of
three measures using a sequential forward search algorithm. It is concluded that the texture spectrum based
discrimination technique proves superior to using pre-defined sets of texture measures, since it is able to highlight
areas on imagery which are often associated with disrupted, displaced land masses.
1. INTRODUCTION
Landslides may produce complex slope surface

patterns not readily distinguished visually on remotely
sensed data. Therefore approaches to landslide
mapping with remotely sensed digital imagery, are
commonly based on a combination of textural and
spectral enhancements (Eyers et al, 1995). Automatic
texture analysis has been successful in discriminating
regular patterns in many machine vision applications.
However, a more serious challenge is posed when it is
applied to remotely sensed data because of the
irregular and less distinctive patterns usually occurring
in nature.
2.1 Study area
In this study we evaluate two statistical approaches to

texture enhancement and discrimination for semiautomated mapping hummocky slid terrain from
Daedalus ATM data. The first approach regards the
use of texture measures (Haralick et al, 1973; Haralick
1979;
Baraldi and Parmiggiani, 1995). These measures
contain information about the spatial distribution of
local grey-level variations within a band. In the second
approach, image texture is defined in terms of
frequency distribution of local intensity variation by
means of its texture spectrum (Wang and He, 1990).
From this, a supervised texture discrimination
technique is applied to identify areas with possible
slides. Statistical and empirical methods to select the
best texture measures and algorithm parameters, as
well as the introduction of subtle variations to the
original algorithms are also discussed.

SafeLand - FP7
The semi-arid moderate-relief area of Los Vlez in SE

Spain was selected for this study. This site includes
complex terrain patterns in terms of both landforms
and land use, which are very representative of the
various natural conditions found in extensive areas of
south and south-east Spain (Hervs and Rosin, 1996;
Hervs et al., 1996).
Landslides occur mainly on dark green and greenishbrown clays and marls of Lower Miocene age, which
often includes gypsum crystals and some sandstone
layers locally. They appear at the foot of the limestone
and dolomite ridges and hills, and at gully sides. Often
these formations are capped by cemented
conglomerate colluvium which slides together with the
underlying materials. Using the nomenclature from
UNESCO (1993) and IAEG (1990), the slope
movements correspond to rotational slides, both single
and multiple, and composite rotational-slides rock-falls.
Scarps are only apparent in some slides. The surface
dimensions of the displaced masses range from 100 to
800 metres either along or across the major movement
direction. Most slides show a characteristic hummocky
surface made up of an irregular succession of
depletion and accumulation zones.
2.2 Digital data
Page 261 of 401
Deliverable 4.1
For this study Daedalus AADS 1268 Airborne

Thematic Mapper (ATM) data were used. The ATM
sensor acquires imagery in 11 spectral bands from the
visible to the thermal infrared. A number of swathes
were flown by INTA of Torrejn, Spain, on 26th April
1994 just before noon. A 6051 883 pixel swathe with
a nominal pixel size of 3.5 m at nadir was then
selected. This covers an area approximately 16 km
long by 2.5 km wide. A ground survey was undertaken
simultaneously with the ATM flights.
The wide field of view of the ATM causes some
panoramic distortion across the swath, which was
corrected during pre-processing, as were the roll
effects (see Hervs et al., 1996). No further geometric
correction was attempted since it was not regarded
relevant for this kind of investigation. Possible acrosstrack differential sun illumination effects were reduced
to a minimum by flying swathes at a time when high
sun elevation angle is relatively high, while its azimuth
is far from the across-track scanning direction. This
effect was not immediately apparent in the data,
therefore no attempt was made to correct for it.
We further refer to Hervs and Rosin (2001) where the
same methods were adopted to SPOT panchromatic
images of 10 m x 10 m resolution and IRS-1C
panchromatic images of 5 m x 5 m.
2.3 Texture measure approach
Statistical texture measures are concerned with the
spatial distribution and spatial dependence among the
image pixel values in a local area. These measures
are extracted from grey-level co-occurrence matrices
(Haralick et al, 1973; Haralick, 1979). In this work we
evaluate the usefulness for landslide mapping of 14
textural measures selected from among those reported
successful in the literature for various remote sensing
applications. Some of these have been originally
proposed by Haralick et al (1973), such as Angular
Second Moment, Contrast, Correlation, Variance,
Inverse Difference Moment, Sum Average, Sum
Variance, Sum Entropy, Entropy, Difference Variance,
Difference Entropy and Second Measure of
Correlation. Others such as Recursivity and Inverse
Recursivity were proposed by Baraldi and Parmiggiani
(1995).
Because it was impossible to test the 14 texture
features for all bands, it was decided after both visual
inspection and preliminary processing of the ATM data
(see Hervs et al., 1996) to use only band 11 (8.5 m 13.5 m). This band captures thermally emitted energy
with dominant effect of surface temperature. Daytime
thermal imagery in the area is sensitive to topography
and therefore to geomorphology while is less sensitive
to land use and lithological variations. Basic spectral
enhancement techniques for characterising possible
variations of vegetation patterns, soil moisture or
lithologies between the slid and surrounding non-slid
areas using other bands generally proved inefficient.
This was mainly due to the steady drought conditions
in the area, the single time acquisition and the lack of

SafeLand - FP7
significant lithological changes around the displaced

mass.
2.4 Texture spectrum approach
The texture spectrum approach (Wang and He, 1990)
first considers texture units (Ei) by analysing each 33
window. The central value Vc is compared to its 8
nearest neighbours Vi or, alternately, to longer
distance neighbours, which are assigned one of three
values Ei determined by
0 if Vi Vc
Ei 1 if Vi Vc
2 if V V
i
c
From Ei the texture unit number is generated by

8
NTU 3i 1 Ei
with NTU {0, 1, 2, ,
i 1
656}
For each ww window in the image a texture spectrum
is generated by histogramming the texture unit
numbers that occur within the window (Figure 1).
Similarity between textures can now be compared by
taking the absolute differences between their texture
spectra. Thus, for two texture spectra Ai and Bi where i
{0,1,2,...,6560}, the similarity is
6560
S Ai Bi
i 0
For our purposes we generate the training texture

spectrum Ti from the area(s) of interest (of arbitrary
size and shape), corresponding to known landslide(s),
over ATM band 11. This spectrum is normalised to a
frequency by dividing each element of the histogram
by the area of the region. Supervised texture
discrimination is then carried out by assigning to each
pixel in the image the similarity S calculated from the
difference between Ti and the image texture spectrum
Wi(x,y), calculated in the ww window centred at (x,y)
S x, y
6560
T W ( x, y )
i 0
Since in most cases pixels do not have precisely

identical values to their neighbours a more efficient
version is to generate texture units using only two
values
0 if Vi Vc
Ei
1 if Vi Vc
while in the situation that Vi = Vc either the texture unit
number can be discarded, or Ei can be randomly
Page 262 of 401
Case Study 2
assigned either 0 or 1, and the resulting texture unit

number retained. We take the latter approach. Using 2
rather than 3 values means that much shorter texture
spectra are generated (containing 256 rather than
6561 values), providing a substantial improvement in
efficiency for only a slight degradation in quality. In
either version image pixels are subsequently
normalised to [0, 255].
Rev. No: 2
Date: 2012-02-15
to discriminate between rough and smooth surfaces.

However, in the study area the distinction between
known landslides, small gullies and cultivation terraces
is generally not possible in any of these texture
images. The Entropy image (Figure 2c), as first-ranked
by the SFS algorithm, does not appear in general
superior to Recursivity and Second Measure of
Correlation for mapping possible sliding. Nonetheless,
these measures are proved superior to other
randomly-selected measures.
3. RESULTS
The 3 statistically-selected texture measures are able
Fig. 1. Schematic flow chart of the texture spectrum approach with training texture spectrum Ti and image texture spectrum Wi

SafeLand - FP7
Page 263 of 401
Deliverable 4.1
In a false colour composite of these three texture

images discrimination of gullies and slides is improved
with respect to single measures, but this is not yet a
very useful product to differentiate slides by its own.
Furthermore, hummocky slid terrain cannot be
distinguished from closed pine trees stands. These
texture measures can be useful for landslide mapping
only as a preliminary image segmentation technique.
The texture spectrum method outputs an image where
very low pixel value areas are often associated with
landslides and disrupted terrain. From testing a few
variations of the possible texture spectrum algorithm
parameters, it was found that selecting a minimum of 2

training sites, a moving window (ww) of 81 81
pixels and a pixel increment of 4 for moving the
window both in the row and column directions, resulted
in a satisfactory image to map potential and existing
slide areas. In this density-sliced image (Figure 2b),
forest stands and regularly spaced trees crops are
clearly discriminated from known displaced land
masses. Incipient badlands and some area with
irregular limestone exposures appear undifferentiated
from landslides. The former may hinder detecting
existing instability on gully sides.
Fig. 2. (a) ATM Band 11. Known landslide sites are circled. (b) Texture Spectrum Based Segmentation. (c) Entropy Image

SafeLand - FP7
Page 264 of 401
Case Study 2
In our data, larger algorithm window sizes and pixel

increments tend to smooth the image, making the site
discrimination more difficult. Smaller windows and
pixel increments can result in relatively noisy images,
enhancing non-relevant spatial frequency content.
4. CONCLUSIONS
Two semi-automated, statistical texture approaches for
mapping landslide geomorphologic features in semiarid terrain have been analysed using single
acquisition, high spatial resolution optical data.
Some predefined texture measures can be useful in
this type of research for preliminary image
segmentation. In addition, the ability of some of the
local statistics properties as edge enhancers becomes
apparent. Selected measures using a discriminant
function prove to be a more objective approach, and
can thus be used irrespective of the study area. This
approach can be fairly automated, although it requires
substantial computer processing. Texture spectrum
based discrimination proves a more suitable and costeffective method, since it is able to highlight areas on
imagery often associated to sliding, and suspect areas
where detailed site investigations are needed.
However, in both approaches landslides are not easily
differentiated
from
dense
erosion
patterns.
Investigations are continuing to test the methodology
to lower spatial resolution images and further tune the
algorithm parameters.
Rev. No: 2
Date: 2012-02-15
Hervs, J., Rosin, P.L., Fernndez-Renau, A., Gmez,

J.A., Len, C., 1996. Use of airborne multispectral
imagery for mapping landslides in Los Velez
district, south eastern Spain. In: Chacon, J.,
Irigaray, C., Fernndez, T. (Eds.), Landslides.
Balkema, Rotterdam, pp. 353 361.
Hervs, J., Rosin, P.L., 2001. Tratamiento digital de
imgenes de teledeteccin en el espectro ptico
para
el
reconocimiento
y
control
de
deslizamientos. Proc. V Simposio Nacional sobre
Taludes y Laderas Inestables. Madrid, 27-30
November 2001. CEDEX, Ministerio de Fomento,
Madrid, pp. 63-74.
IAEG Commission on Landslides, 1990. Suggested
nomenclature for landslides. Bull. International
Association of Engineering Geology 41, 13-16.
UNESCO Working Party on World Landslide
Inventory, 1993. A suggested method for
describing the activity of a landslide. Bull.
International Association of Engineering Geology
43, 53-57.
Wang, L. He, D.C., 1990. A new statistical approach
for texture analysis. Photogram. Eng. Remote
Sensing 56, 61-66.
REFERENCES
Devijver, P.A., Kittler, J., 1982. Pattern Recognition. A
Statistical Approach, Prentice-Hall, Englewood
Cliffs, New Jersey, USA.
Eyers, R.,J. Moore, J. Hervs, Liu, J.G., 1995.
Landslide mapping using digital imagery: a case
history from south east Spain, Proc. 31st Annual
Conf.: Geohazards and Engineering Geology,
Coventry, UK, 10-14 September 1995. The
Geological Society, pp. 379-388.
Haralick, R.M., Shanmugan, K., Dinstein, I., 1973.
Textural features for image classification, IEEE.
Trans. Syst. Man, Cybern. Vol. SMC-3, 6, 610-621.
Haralick, R.M., 1979. Statistical and structural
approaches to texture, Proc. IEEE 67, 786-804.
Hervs, J., Prieto, C., 1981. Estudio de movimientos
del terreno en la provincia de Granada, Ministerio
de Obras Pblicas, Transportes y Medio Ambiente,
Direccin General de Carreteras, Madrid, Spain.
Internal Report, 3.
Hervs, J., Rosin, P.L., 1996. Landslide mapping by
textural analysis of Daedalus ATM data. Proc.
Eleventh Thematic Conf. Applied Geologic Remote
Sensing, Las Vegas, Nevada, vol. 2. ERIM, Ann
Arbor, Michigan, USA, pp. 394402.

SafeLand - FP7
Page 265 of 401
Case Study 3
Rev. No: 2
Date: 2012-02-15
CASE STUDY 3:
Title: Using TLS for rockslide monitoring. Case study of the Eiger collapse, Swiss Alps.
Application: Landslide characterization and monitoring; Displacement detection; Collapse
characterization.
Technique: Terrestrial Laser Scanner (TLS); Terrestrial LIDAR.
Main references (text extracted from): Oppikofer et al. (2008); Oppikofer (2009).
Contributor: NGU (Oppikofer, T.) and UNIL (Jaboyedoff, M.).
ABSTRACT
Landslides are a significant natural hazard in mountainous regions and are often triggered by external factors,
such as earthquakes, rainfall, permafrost thawing or retreat of glaciers. A large landslide occurred in the Swiss
Alps on 13 July 2006, when portions of an immense rock spur on the eastern flank of the Eiger peak collapsed.
Here we use field observations and terrestrial laser scanning data to record and quantify the relative motion along
the various blocks of rock that form this spur. The data show that during two years of observation the blocks
moved relative to one another by up to tens of metres along fractures that can be related to pre-existing
discontinuities. Rates of motion and deformation were high throughout July 2006, particularly in the northern part
of the spur that partially collapsed on 13 July. The rates decreased considerably during the subsequent months,
although a slight increase was noted in June and July 2007, but disintegration of the rock mass continued until
Summer 2009, when the last remnants of the rock spur collapsed. These observations are consistent with
instability of the spur initiated by subsidence of a single block at the rear, which acted as a wedge and
disintegrated over time owing to loss of lateral confinement.
1. INTRODUCTION
High-resolution digital elevation models obtained from
aerial or terrestrial laser scanning (ALS and TLS,
respectively) are used to detect landslides and
rockfalls (Slob and Hack, 2004; Rosser et al., 2005;
Abelln et al., 2006; Bitelli et al., 2004; Rosser et al.,
2007; Bauer et al., 2005; Derron et al., 2005; Schulz,
2007; Agliardi and Crosta, 2003; McKean and Roering,
2004). Most landslide monitoring techniques provide
information on only a few selected points and not the
whole landslide surface. Sequential terrestrial laser
scanning, on the other hand, allows for analysis
of movement direction and velocity, displacement and
volume change. These three-dimensional data open
new perspectives in predicting mass wasting and
deciphering failure mechanisms.
In this case study, we focus on a collapse at the
Eigers eastern flank (Fig. 1), one of Switzerlands
most famous peaks. The instability of this 2 million m3,
compact Jurassic limestone mass, was caused by the
Lower Grindelwald glaciers retreat, which has been
accelerated since 1935 by global warming. At the last
glacier maximum in 1860, the ice reached the back
scars level. The instability was subject to particularly
high glacial compression due to its position on a
topographic ridge. The removal of the glacier buttress
and subsequent decompression allowed the spur to
fail. On 10 June 2006 rockfalls of several hundred m3,
first slope movements and two steep 250-m-long
valley-parallel open cracks were observed. One of

SafeLand - FP7
these fractures forms the back-crack of the spur, and

the other separates the spur into a front and a rear
block (Fig. 1b,c). In addition, these blocks are
separated into a main part and a northern part by a
subvertical bedding plane that is nearly perpendicular
to the open cracks (Fig. 1c).
Fig 2a (20 June 2006) shows the scar one week
after the opening of the back-crack (dashed red line).
Only small displacements are visible in this figure,
besides some rockfall activity on the southeastern
part. Fig. 2b (13 July 2006) shows that the rear part
moved downwards by several meters and started to
break up, while a partial collapse occurred in the
northern part (before the first terrestrial laser scans).
Fig 2c (17 July 2006) shows the scar of the 13 July
2006 collapse in the northern part of the rock spur. Fig
2d (2 August 2006) shows the continuation of the
movement in the rear part. The last free standing
column in the northern part also collapsed, and a small
rockfall scar is visible on the front block (dashed red
line). Figs 2e to g (30 August 2006, 22 September
2006 and 1 December 2006, respectively) show
ongoing displacements, break-up of the rear block and
the opening of a new middle fracture (dashed red line)
within the front block. Figs 2h and i (11 June 2007 and
14 August 2008) show the dismantling of the front
block with an increase of rockfall activity and several
partial collapses. Finally, Fig 2j shows one of the final
stages of this rockslide, where only small parts of the
rear block and a 90 m high, blade-shaped needle
remained until it collapsed in summer 2009.
Page 267 of 401
Deliverable 4.1
Fig. 1. a) Situation map of the Eiger rockslide; b) Picture of the back-crack at the beginning of the monitoring campaign (Picture:
Geotest AG); c) Frontal view of the Eiger rockslide showing the division of the spur into a front and a rear block separated by a
steep middle fracture (green line) and the northern rear and front blocks are separated from the main blocks by a bedding plane
(blue line).
At the beginning of the monitoring campaign, slope

movements were very fast (several dm per day) and
caused frequent rockfalls and partial collapses.
Initially, the back-crack opening was measured by
perilous hand measurements, followed by total station
surveys. As the installed reflective targets were rapidly
destroyed due to collapses and rockfalls, a contactless
and reflectorless monitoring technique terrestrial
laser scanning (TLS) had to be used. On 11 July
2006, we deployed a long-range terrestrial laser
scanner to monitor the rockslide and acquired
sequential TLS point clouds on 12, 13, 17 and 25 July,
22 September and 1 December 2006, on 11 June and
17 July 2007 and on 14 August and 21 October 2008.
The comparison of sequential point clouds reveals the
instabilitys displacements and enables the calculation
of rockfall volumes. A more detailed explanation of this
case study can be found in Oppikofer et al., 2008 and
Oppikofer, 2009.

TLS is based on the reflectorless and contactless
acquisition of a point cloud of the topography using the
pulsed laser time-of-flight technique (Slob and Hack,
2004). For this study, we used an Optech ILRIS3D
terrestrial laser scanner system. TLS point clouds,
were acquired from two viewpoints located on the
opposite side of the valley (Sites A and B in Fig. 1a), at
a distance of 600m and 530m respectively. From both
viewpoints the resulting point cloud has an average
resolution of 18.0 cm on the rock face and 20.0 cm at
the back-crack. This spacing is approximately twice
the magnitude of the laser spot size (11.4 cm at a
distance of 600 m), which is significantly more than the
proposed ideal spacing of 0.859 times the beamwidth
(Lichti and Jamtsho, 2006). This however has the
advantage of drastically reducing acquisition time and
data file size by still providing high-resolution point
clouds.

SafeLand - FP7
We aligned the scans from both viewpoints with

Innovmetric PolyWorks software using first manual
point-pair matching and afterwards the surface-tosurface iterative closest point algorithm implemented
in PolyWorks. We progressively reduced the maximum
distance between the two scans to minimize the
alignment errors (1a < 2.5 cm). This error is mainly
due to the spot spacing and the fact that consequently
not always the same points of the quite irregular rock
surfaces were measured. In comparison, the
instruments precision is given by the manufacturer as
7 mm on the range and 8 mm on the position at a
distance of 100 m. Georeferencing is achieved by
alignment of the TLS point cloud onto the highresolution aerial digital elevation model with a cell size
of 2m (DEM-CS provided by swisstopo), with an
absolute positioning error of 14.5 cm (1gr). We used
the same alignment procedure for the alignment of
sequential TLS point clouds, but restraining the
iterative alignment to the stable part around the
rockslide only (1 a < 2.5 cm). Thus, the overall
alignment error of sequential scans is equal to
approximately 6 cm on single points. The
displacements measured in the 2006 Eiger rockslide
are much bigger than instrumental and data treatment
errors. Considering entire point clouds, for example by
interpolated surfaces, errors are significantly lower
than on single points.
Pre- and postfailure three-dimensional models on the
basis of the interpolated TLS point clouds permitted us
to obtain precisely the volume of rockfalls by
calculating the volume contained between the two
surfaces using PolyWorks. The comparison of the
aligned sequential TLS point clouds reveals
differences (expressed as shortest distance) between
the two topographic surfaces. These positive or
negative differences are interpreted as slope
movements: positive differences stand for an advance
of the rockslide and debris accumulation on the scree
slopes, whereas negative differences are related to
subsidence or rockfalls.
Page 268 of 401
Case Study 3
Rev. No: 2
Date: 2012-02-15
Fig. 2. From a to j, photographs of the evolution of the Eiger collapse. See text for a detailed explanation.

SafeLand - FP7
Page 269 of 401
Deliverable 4.1
More precise information on the slope movements is

obtained from displacement vectors of different
rockslide parts. They are calculated by manually
identifying corresponding points on view-shaded
representations of both point clouds. Ideal point pairs
are summits of blocks or small spurs, intersections or
end points of discontinuity traces and so on. The
vector between a point pair provides the displacement
direction and velocity of the point. Multiple vectors on
each rockslide compartment provide the mean
displacement vector and error. The errors in the
average displacement length and orientation are
obtained by dividing the standard deviations by the
square root of the number of vectors and are equal to
10.1 cm and 2.7 (aperture of cone centred on the
mean orientation and including 68% of the values),
respectively. The relatively high error is due to the high
point spacing (~18 cm), the measurement error on a
single point and the fact that the slope is breaking
apart.
3. RESULTS
Sequential TLS point clouds permitted to follow the
slope movements in 3D using shortest distance
comparisons with displacements of 80 cm/d for the
rear block, 20 cm/d for the front block and locally even
up to 125 cm in 24 hours (between 12 and 13 July
2006). This movement monitoring provided a first
assessment of the slope movements and highlighted
the different displacement directions and velocities of
the rear relative to the front block, with mainly negative
differences in the rear (downward displacements) and
positive differences at the front block (advances on
basal sliding surface). After the very fast initial
displacement rates in summer 2006, the velocities
decreased in the following months (4 cm/d and 1.5
cm/d for the rear and front blocks, respectively, during
winter 2006/2007). In one year, the total
displacements reached 15 m for the front block and 50
m for the rear parts. These high displacements went
along with the progressive lateral dismantling and
break-up of the rear block.
A detailed explanation of the rockslide evolution, using
the comparison between two consecutive TLS point
clouds, is shown in Fig 3:
TLS comparison between 11 and 12 July 2006 (Fig
3a) displays the downward movement of the rear part
(~80 cm/day), the advance by 125 cm/day of two
columns in the northern block, as well as the 2530 m3
large rockfall on the northern block and the scree
deposits on the southern side.
a)
TLS comparison between 12 and 13 July
2006 (Fig 3b) highlights the ongoing downward
movement of the rear part (~75 cm/day) and the fast
movement of the columns in the northern block (up to
150 cm/day) accompanied by two rockfalls of 118 and
329 m3. These higher movement rates on these

SafeLand - FP7
columns and the small rockfalls can be interpreted as

precursory signs for the partial collapse on 13 July
2006. Photographs of this collapse indicate that it
started with the failure of this fast-moving column,
which is valuable information for rockslide monitoring
and especially spatial and temporal failure prediction
through the possibility to detect areas with high
displacement rates by TLS, as discussed in Abelln et
al. (2009).
b)
2006 (Fig 3c) reveals the extent of the northern flanks
partial collapse that occurred at 19h35 on 13 July 2006
(grey area in Fig 3c) and the associated scree
deposits in the valley. Furthermore the continuous
downward movement of the rear block can be
quantified to 65 cm/year.
c)
2006 (Fig 3d) shows the ongoing advance of the
northern part, besides the advance of the front block
(~1.25 m; ~15 cm/day) and the downward movement
of the rear part (~4 m or 50 cm/day in average, up to
10 m on the left flank). A 4460 m3 large rockfall on the
front block and the associated deposits can also be
seen in this comparison of sequential TLS point
clouds.
d)
TLS comparison between 25 July and 22
September 2006 (Fig 3e) displays the dismantling of
the southern flank of the rear block and the breakingup of the northern part. A small collapse along the
vertical fracture separating the northern front block
from the front block led to the accumulation of up to 15
m-high scree deposit. The front block advanced by 6
to 7 m (~12 cm/day), while the rear block moved
downwards by more than 15 m (~30 cm/day).
e)
TLS comparison from 22 September to 1
December 2006 (Fig 3f) shows the downward
movement by about 8 m (~11 cm/day) of the rear
block and rear of the northern block, together with the
advance of the front block advanced (~2.5 m;
~4 cm/day). A new subvertical fracture, parallel to the
back-crack, separated the front block into two parts,
but no differential movement is detected between the
two. Note also the 3280 m3 rockfall of an overhanging
block on the back-crack.
f)
TLS comparison from 1 December 2006 to
17 July 2007 (Fig 3g) shows the continuation of the
movement of the rear block (~15 m; ~7 cm/day) and
its breaking-up, while the displacement of the front
block reached up to 6 m (~2.5 cm/day). The
comparison with a TLS dataset acquired on 11 June
2007 reveals that during winter 2006/2007 the
movement rates were 3-4 times smaller than in the
beginning of summer 2007, indicating possible
seasonal variations in the rockslide activity.
Page 270 of 401
Case Study 3
Rev. No: 2
Date: 2012-02-15
Fig. 3. Sequential comparison of TLS point clouds. Negative values equal downward displacements and/or
detachment of material. Positive values equal outward displacement and/or debris accumulation
.

SafeLand - FP7
Page 271 of 401
Deliverable 4.1
g)
Finally, TLS comparison from 17 July 2007 to
14 August 2008 (Fig 3h) displays the slope
displacements for both blocks with up to 30 m for the
rear block that continued to break-apart laterally and
approximately 10 m for the front block. This
comparison also testifies the increase in rockfall
activity with two large rockfalls on the front block, as
well as a partial collapse of approximately 60'000 m3
on its north-eastern corner (Fig 3h). Until 21 October
2008, the front block continued its break-up with
several smaller and larger events according to eyewitnesses and only a blade-shaped, 90 m high
needle remained.
Displacement vectors measured for the different
blocks provide a more precise analysis of the slope
movements. The sliding directions remained fairly
constant until autumn 2006, but the velocities
decreased. Since autumn 2006, the rear block's sliding
direction turned more to the East owing to the lateral
dismantling of the rock mass. Based on these 3D
displacement data and on a structural analysis of the
spur, a coherent geometrical model for the 2006 Eiger
rockslide was proposed. In the early stages of the
rockslide, its mechanism was a bilinear wedge failure
(Norrish and Willey, 1996). The downward-moving rear
block acted as an active wedge splitting the front block
from the stable rock mass and pushing it passively
along a complex sliding surface with an average dip
angle of ~33 (see Oppikofer et al., 2008 for more
details).
4. CONCLUSIONS
The study of the 2006 Eiger rockslide demonstrates
the necessity of a global analysis of 3D displacements
on the entire rockslide, in order to better understand
the failure mechanism and construct a consistent
structural and geometrical instability model. Such
knowledge is crucial for landslide hazard assessment,
because it opens new perspectives in characterizing
and understanding the slope movements and helps in
designing continuous monitoring techniques. The
ability to predict failure is crucial to hazard assessment
and monitoring of landslides (Crosta and Agliardi,
2003). Comparisons of sequential TLS data enable the
detection of areas with the highest displacement rates,
which are more susceptible to future failure. The
partial collapse of 13 July 2006 underlines the link
between maximum velocities and failure locations.
Knowing the most active zones on a landslide,
permanent survey techniques can be designed for
precise temporal failure prediction.
REFERENCES
Abelln, A. Jaboyedoff, M., Oppikofer, T., Vilaplana,
J.M. Detection of millimetric deformation using a
Terrestrial Laser Scanner: Experiment and
application to a rockfall event. Natural Hazards and
Earth System Science, 9, 365-372 (2009).

SafeLand - FP7
Abelln, A., Vilaplana, J. M. & Martnez, J. Application

of a long-range terrestrial laser scanner to a
detailed rockfall study at Vall de Nria (Eastern
Pyrenees, Spain). Eng. Geol. 88, 136148 (2006).
Agliardi, F. & Crosta, G. B. High resolution threedimensional numerical modelling of rockfalls. Int. J.
Rock Mech. Min. 40, 455471 (2003).
Bauer, A., Paar, G. & Kaltenbck, A. in Geoinformation for Disaster Management (eds van
Oosterom, P., Zlatanova, S. & Fendel, E. M.) 393
406 (Springer, Berlin, 2005).
Beniston, M. August 2005 intense rainfall event in
Switzerland: Not necessarily an analog for strong
convective events in a greenhouse climate.
Geophys. Res. Lett. 33, L05701 (2006).
Bitelli, G., Dubbini, M. & Zanutta, A. Proc. XXth ISPRS
Congress, Istanbul Vol. 35, 246251 (International
Society for Photogrammetry and Remote Sensing,
Istanbul, Turkey, 2004).
Coe, J. A. et al. Seasonal movement of the
Slumgullion landslide determined from Global
Positioning
System
surveys
and
field
instrumentation, July 1998March 2002. Eng.
Geol. 68, 67101 (2003).
Crosta, G. B. & Agliardi, F. Failure forecast for large
rock
slides
by
surface
displacement
measurements. Can. Geotech. J. 40, 176191
(2003).
Derron, M.-H., Jaboyedoff, M. & Blikra, L. H.
Preliminary assessment of rockslide and rockfall
hazards using a DEM (Oppstadhornet, Norway).
Nat. Haz. Earth Syst. Sci. 5, 285292 (2005).
Eberhardt, E., Stead, D. & Coggan, J. S. Numerical
analysis of initiation and progressive failure in
natural rock slopesthe 1991 Randa rockslide. Int.
J. Rock Mech. Min. 41, 6987 (2004).
Evans, S. G. & Clague, J. J. Recent climatic change
and catastrophic geomorphic processes in
mountain environments. Geomorphology 10, 107
128 (1994).
GK/SCNAT & VAW/ETHZ. The Swiss Glaciers,
Yearbooks of the Glaciological Commission of the
Swiss
Academy
of
Science
(SAS).
(http://glaciology.ethz.ch/swiss-glaciers) (Labratory
of Hydraulics, Hydrology and Glaciology (VAW) of
ETH Zrich, 2006).
Guarnieri, A., Pirotti, F., Pontin, M. & Vettore, A. 3rd
IAG/12th FIG Symp., Baden, Austria (International
Association of Geodesy & International Federation
of Surveyors, Baden, Austria, 2006).
Hopkin, M. Eiger loses face in massive rockfall.
News@Nature
<http://www.nature.com/news/2006/060717/full/060
717-3.html> (2006).
Jaboyedoff, M., Baillifard, F., Bardou, E. & Girod, F.
The effect of weathering on Alpine rock instability.
Q. J. Eng. Geol. 37, 95103 (2004).
Lichti, D. D. & Jamtsho, S. Angular resolution of
terrestrial laser scanners. The Photogrammetric
Record 21, 141160 (2006).
Page 272 of 401
Case Study 3
Lindenbergh, R. & Pfeifer, N. Proc. 7th Conf. Optical

3D Measurement Techniques, Vienna, Austria Vol.
2, 6170 (Vienna University of Technology,
Vienna, Austria, 2005).
Liniger,
M.
Die
Herausforderung
der
Gefahrenprognose
bei
Massenbewegungen:
Rutsch- und Sturzprozesse Markus Liniger. Bull.
Appl. Geol. 11, 7588 (2006).
McKean, J. & Roering, J. Objective landslide detection
and surface morphology mapping using highresolution airborne laser altimetry. Geomorphology
57, 331351 (2004).
Mencl, V. Mechanics of landslides with non-circular
slip surfaces with special reference to the Vaiont
slide. Geotechnique 16, 329337 (1966).
Messerli, B., Messerli, P., Pfister, C. & Zumbuhl, H. J.
Fluctuations of climate and glaciers in the Bernese
Oberland, Switzerland, and their geoecological
significance, 1600 to 1975. Arct., Alp. Res. 10,
246260 (1978).
Munich, Re. Topics Geo - Annual Review: Natural
Catastrophes
2005
(Munich
Reinsurance
Company, Munich, Germany, 2006).
Norrish, N. I. &Wyllie, D. C. in Landslides Investigation
and Mitigation (eds Turner, A. K. & Schuster, R. L.)
391425 (Transportation Research Board, National
Research
Council,
National
Academy
Press,Washington DC, 1996).
OcCC. Climate Change and Switzerland in 2050.
Impacts on Environment, Society and Economy
(OcCC/ProClim-, Bern, Switzerland, (2007).
Oppikofer T., Detection, analysis and monitoring of
slope movements using high-resolution digital
elevation models. PhD thesis, Institute of
Geomatics and Analysis of Risk, University of
Lausanne, Switzerland, 191 p. (2009)
Oppikofer T., Jaboyedoff M., Keusen H.-R., Collapse
at the eastern Eiger flank in the Swiss Alps. Nature
Geosciences 1(8), 531-535 (2008).
Petrascheck, A. & Hegg, C. (eds) Hochwasser 2000 Les crues 2000 (Federal Office for Water and
Geology, Berne, Switzerland, 2002).
Rosser, N. J., Petley, D. N., Dunning, S. A., Lim, M. &
Ball, S. Rock mechanics: Meeting societys
challenges and demands. in Proc. 1st Canada
U.S. Rock Mechanics Symposium, Vancouver,
Canada, May 2731, 2007 (eds Eberhardt, E.,
Stead, D. & Morrison, E.) 113120 (Taylor and
Francis, London, 2007).
Rosser, N. J., Petley, D. N., Lim, M., Dunning, S. A. &
Allison, R. J. Terrestrial laser scanning for
monitoring the process of hard rock coastal cliff
erosion. Q. J. Eng. Geol. 38, 363375 (2005).
Sartori, M., Baillifard, F., Jaboyedoff, M. & Rouiller, J.
D. Kinematics of the 1991 Randa rockslides
(Valais, Switzerland). Nat. Haz. Earth Syst. Sci. 3,
423433 (2003).
Schulz,W. H. Landslide susceptibility revealed by
LiDAR imagery and historical records, Seattle,
Washington. Eng. Geol. 89, 6787 (2007).

SafeLand - FP7
Rev. No: 2
Date: 2012-02-15
Slob, S. & Hack, R. in Engineering Geology for

Infrastructure Planning in Europe. A European
Perspective (eds Hack, R., Azzam, R. & Charlier,
R.) 179190 (Springer, Berlin, 2004).
Wieczorek, G. F. in Landslides Investigation and
Mitigation (eds Turner, A. K. & Schuster, R. L.) 76
90 Transportation Research Board, National
Research
Council,
National
Academy
Press,Washington DC, (1996).
Page 273 of 401
Case Study 4
Rev. No: 2
Date: 2012-02-15
CASE STUDY 4:
Title: Structural analysis using colour coded information derived from 3D point clouds.
Application: Structural analysis; Discontinuity orientation.
Technique: Aerial Laser Scanner (ALS); Terrestrial Laser Scanner (TLS).
Main references (text extracted from): Metzger et al., 2008 ; Jaboyedoff et al., 2007.
Contributor: UNIL (Metzger, R., Jaboyedoff, M.)
ABSTRACT
Current laser scanning systems, either mounted over aerial or terrestrial sensors, allow for massive acquisition of
million 3D points. One of the main current issues in laser scanner processing is the managing of this information
and its transformation into a certain parameter, e.g. dip direction and dip angle. In this case study, we present the
application of Coltop 3D software developed by our group (Metzger et al., 2008; Jaboyedoff et al., 2007) for
structural analysis purposes. Each point of the point cloud dataset can be represented through its colour defined
by the local dip and strike direction. The method for the basically in a three step process: (1) the spatial indexing
of massive unstructured point clouds; (2) the computation of the local dip and strike direction, by means of Eigen
values analysis; (3) the unique data representation according to their dip and strike direction, by means of a lower
Schmidt projection. Such analysis allows for the quick identification of structural features that may condition
landslide occurrence, such as the different joints sets, foliation, faults, etc.
1. INTRODUCTION
matrix C is defined over a local neighborhood

surrounding a point of interest as:
The use of LIDAR derived High Resolution Digital

Elevation Models (HRDEM) is currently widely used on
landslide detection, characterization, monitoring and
modelling.
Althoug GIS tools allow the analysis of topography
using slope, slope aspect, second derivative,
curvature, flow paths, etc., only a few softwares are
dedicated to the analysis of the relief structure. A
pioneering attempt of merging slope angle and slope
aspect in a single map was made by Brewer and
Marlow (1993), representing slope angle and slope
aspect with different colours. Using the dip and strike
direction of each cell, a DEM can be theoretically
represented with a map having a unique colour for
each spatial orientation. The application of this colour
code information may help in the 3D characterization
of geological structures.
2. METHOD
2.1 Principles of the software
Basically, the software allows for the visual
representation of dip and strike directions using a HSI
(Hue Saturation Intensity) wheel. The color displayed
in the screen is linked to the normal pole of the
sourronding points, as depicted in figure 1.
In order to deal with the huge number of datasets, a
structure based on octrees (an index based on spatial
portioning) is used, which allows for fast localization of
points within a given region, low consumption of RAM,
and hard drive access minimization. The covariance

SafeLand - FP7
where the entries for a neighborhood containing k

points are defined as:
with E(value) being the excepted value or the mean

value, (E(x)=x), and var(x) and cov(x,y) denoting the
variance of x and the covariance between x and y
respectively (Belton and Lichti, 2002). Since C is a
symmetric and positive semidefinite, its associated
eigenvalues i are greater than (or equal) to zero. The
local normal vector is given by the associated
eigenvector ei with the smallest eigenvalue. The
direction of the normal vector is the same as the one
found by least squares plane fitting, since the two
methods are equivalent (Shakarji, 1998).
2.2 Outputs
Following information is extracted from the LIDAR
point cloud using COLTOP 3D software:
Page 275 of 401
Deliverable 4.1
Dip angle and dip direction of a given point of

the point cloud.
Histogram of the orientations of a given area
of the point cloud.
Vegetation extraction, based on the high
surface variability of the point cloud on certain
areas.
All the extracted information can be visualized either in
a stereonett plot or in a table wit the numeral values.
3. DISCUSSION AND CONCLUSIONS

3D point clouds acquired from airborne or groundbased Lidar permit a rapid structural analysis. This is
useful since joints and instabilities are often in
inaccessible zones. The colors obtained from grid
DEMs using the Hue Saturation Index in COLTOP-3D
permit an easy detection of the main features of a
relief, such as the main joint sets shaping rock faces.
The colored surfaces and their interactivity, allow for a
detailed structural analysis. Unstructured clouds of 3D

data points can serve as a basis for surface
reconstruction by triangulation.
Thus, the color
representation, based on the dip direction and the dip
angle of the surface, can be assigned to each triangle.
This color representation forms a simple way to quickly
obtain information for slope analysis. COLTOP-3D has
shown its efficiency using square DEM grids. The
difficulty to implement a true 3D version comes from:
(1) storage and access of huge Lidar data sets; (2)
octree classification and triangulation of data points;
(3) extraction of 3D surface resulting from the
triangulation, and; (4) representation of the 3D surface
according to a Hue Saturation Index wheel using the
dip direction and dip angle. The promising preliminary
results presented here indicate that these new DEM
analysis tools will greatly help structural geologists and
rock mechanics engineers by supplementing part of
the classical field work and permitting to contribute to
more quantitative field work analysis.
Fig. 1. Principle of the COLTOP 3D software. The color of each sector is defined using the normal vector of the surface
terrain; the HIS wheel is also plotted over a stereonett projection.
REFERENCES
Belton, D. and Lichti, D. 2005. Classification and
Feature Extraction of 3D Point Clouds from
Terrestrial Laser Scanners, in SSC2005: Spatial
Intelligence, Innovation and Praxis: Proceedings of
National Biennal Conference of the Spatial
Sciences Institute 2005, Melbourne, Vic, pp. 39-48,
CD Publication.
Brewer, C. A. and Marlow, K. A. 1993, Color
representation of aspect and slope simultaneously.
Proceedings, Eleventh International Symposium on
Computer-Assisted Cartography (Auto-Carto-11),
Minneapolis, 328-337.

SafeLand - FP7
Metzger, R., and Jaboyedoff, M.: COLTOP 3-D: A

software dedicated to analyze relief using large
DEM and massive 3-D imaging cloud points,
Geophysical Research Abstracts, 10, 08785, 2008.
Jaboyedoff, M., Metzger, R., Oppikofer, T., Couture,
R., Derron, M.-H., Locat, J., and Turmel, D.: New
insight techniques to analyze rock-slope relief
using DEM and 3D-imaging cloud points:
COLTOP-3D software, in: Rock mechanics:
Meeting Societys challenges and demands, Proc.
of the 1st CanadaU.S. Rock Mechanics
Symposium, ed by: Eberhardt, E., Stead, D., and
Morrison, T., Taylor & Francis, 1, 6168, 2007.
Shakarji, C. 1998. Least-Squares Fitting Algorithms of
the NIST Algorithm Testing System. Journal of
Research of the National Institue of Standards and
Technology. Vol. 103, 6.
Page 276 of 401
Case Study 5
Rev. No: 2
Date: 2012-02-15
CASE STUDY 5:
Title: Using TLS for rockfall detection and prediction.
Application: Rock face monitoring; Rockfall detection; Spatial prediction.
Technique: Terrestrial Laser Scanner (TLS); Terrestrial LIDAR.
Main references: Abelln et al. (2010); Abelln et al., 2009; Abelln, 2009.
Contributor: UNIL (Abelln, A.); Univ. Barcelona (Vilaplana, J.M.; Calvet, J.)
ABSTRACT
This case shows an application of a Terrestrial Laser Scanner (TLS) for rock face monitoring. The pilot study area
corresponds to the main scarp of an old slide located at Puigcercs (Pallars Jussa, Spain). Five different point
clouds of the slope were acquired between 2007 and 2009 in an ongoing project. The first part of this research
shows the main rockfalls that occurred during the time span: geometry, frequency, volume, etc... The second
deals with the spatial prediction of rockfalls through the detection of a rockfall precursory indicator (small scale
pre-failure deformation). The spatial location of different rockfall events were predicted through the developed
methodology. This research offers interesting prospects in rockfall risk management: the TLS instrumental offer
interesting prospects not only for the spatial location of rockfalls but also for their spatial prediction.
1. INTRODUCTION
The possibility of acquiring datasets with high accuracy
and spatial resolution are currently opening up new
ways to visualize, model and interpret Earth surface
processes. One of these new remote sensing tools, a
Terrestrial Laser Scanner (TLS) was used in this study
for the detection and spatial prediction of rockfalls.
Minor scale rockfalls are the most frequent type of
landslides on cliffs and steep slopes in mountain
areas. Since this phenomena is the fastest type of
landslide, its impact energy and geological hazard can
reach very high values.
The monitored rock face (Fig.1) is the main scarp of a
landslide that took place in January 1881. This event
was described as a complex roto-translational slide,
evolving into an earth flow in its zone of accumulation
(Vidal 1881; Corominas and Alonso, 1984). The
displaced material occupied an area of 80,000 m2. The
scar consists on an alternation of grey marls,
sandstones, silt and clays. Minor rockfalls currently
constitute the main geomorphological process.
This scar was selected as a pilot area for rockfall

monitoring using a TLS.

2.1 Terrestrial Laser Scanner (TLS)
A TLS ILRS3D (Optech) was employed in this study.
This instrument consists of a transmitter/receiver of
infrared laser pulses and a scanning device. Distance
measurement (range,) is based on the time-of-flight
(t) of the laser pulse. A complete explanation of the
material and method is described in chapter 3 of this
deliverable.
2.2 Data acquisition
Data acquisition was performed from a single station.
First dataset was acquired on September 2007;
subsequent datasets were acquired 59, 135, 303 and
409 days after the first data acquisition.
Fig. 1. Study area: main scarp of an old landslide. The monitored rock face consists in an alternation of grey marls, silt
and sandstones.

SafeLand - FP7
Page 277 of 401
Deliverable 4.1
Three scans were collected in each fieldwork

campaign in order to totally cover the rock face. Main
range distances are comprised between 125 and 175
meters. The mean point spacing ranged from 4 to 6
cm. The standard deviation of the instrumental at the
specific condition of the study area was assessed as a
1.68cm.
2.3 Comparison of sequential scans
3D variations of the terrain on different time periods
were detected through a comparison of sequential TLS
datasets. The method for alignment (ICP) and
comparison (point to surface method) is described in
chapter 3 of this deliverable. Using this approach, two
different type of slope processes were recorded using
this approach: rockfalls and pre-failure deformation.
2.4 Rockfall detection
The detection of the different rockfalls occurred during
the monitoring period is based on the methodology
described in Lim et al., 2006 and Rosser et al., 2005. It
basically consists in a non-continuous acquisition of
TLS datasets, alignment in a common reference
system, difference calculation and volume extraction.
2.5 Detection of the pre-failure deformation
The pre-failure deformation on certain sectors of the
rock face may be used as a precursory indicator prior
to the failure (Abelln, 2009). This deformation is
quantified in our study through the displacement
measured at each point of the TLS point cloud.
Millimetric displacements cannot be detected with
certainty using the original, unprocessed TLS
datasets. Nevertheless, a nearest neighbour (NN)
filtering technique can be applied to the original
dataset, enabling the accurate detection of these
displacements (Abelln et al., 2009).
3. RESULTS
3.1 Rockfall detection
Rockfalls occurred during the time span can be
accurately detected using TLS. The rockfall events
occurred in the study area are shown in Fig. 2.
Volumes of events A, B, C, D and E range from 1 to
100m3. Event A is the event of the highest magnitude
(87m3). A more detailed description of the rockfall
frequency and magnitude during the different periods
analysed is shown in Blanchard et al., (2008) and
Abelln et al. (accepted).
3.2. Spatial prediction of rockfall
As stated above, the spatial prediction of rockfall is
based on the detection of the pre-failure deformation.
In order to spatially detect this precursory indicator
over the rock face, a detailed mesh was generated
(resolution: 5cm). The scattering of the differences
calculation was reduced by applying the NN filtering
technique described in Abelln et al. (2009). During
the time span, two areas showed a centimetric
deformation: areas B and F. These areas showed the
following characteristics: (a) a vertical discontinuity
delimiting the moving block from the stable part of the
slope, (b) an increase in the horizontal displacement
upwards, which is characteristic of a toppling
deformation and (c) a centimetric value of the prefailure deformation lower than 10 cm.
The pre-failure deformation in area B was detected in
period i, being followed by a failure in this area (Event
B) during the next period of the study, meaning that
this pre-failure deformation can be used as a
precursory indicator or rockfall events.
Fig. 2. Rockfall detection using TLS. The detached and accumulated materials of all the rockfalls that occurred during the
time span are encircled with a white line. The higher magnitude (Event A) rockfall is enlarged.

SafeLand - FP7
Page 278 of 401
Case Study 5
Similarly, the pre-failure deformation in area F was

detectable during periods ii and iii. The maximum
value of the pre-failure deformation in this sector of the
rock face has the same value as the deformation that
preceded the event B, with the result that a failure is
very likely in area F. New datasets acquired in
November 2009 confirmed the failure of this area.
Rev. No: 2
Date: 2012-02-15
In our case study, the relationship between pre-failure

deformation and rockfall failure was proved in two
different examples (Areas B and F). Furthermore, new
datasets acquired during 2009 confirmed new rockfall
events preceded by precursory deformation.
Nevertheless, no precursory indicator was detected
prior to some metric scale rockfalls (events A,C,D and
E). This could be due either to an insufficient temporal
resolution (no detection is possible if temporal
resolution of the study is lower than the temporal
length of the pre-failure deformation) or to an
insufficient accuracy of the method for detecting the
precursory deformation.
New cases studies are needed in order to validate the
hypothesis stated in this case study: the applicability of
the method over different study areas should be
tested. Different materials, failure mechanisms and
maximum range may influence the reliability of the
method for detecting rockfall pre-failure deformation.
5. CONCLUSION
Fig. 3. Pre-failure deformation in area F during period iii. As

predicted, a rockfall event occurred in the area few months
later.
The frequency and magnitude of the main rockfalls

that occurred during the time span was analyzed.
Furthermore, the spatial prediction of events B and F
was carried out through the detection of the pre-failure
deformation over the whole rock face. These results
proved that TLS instrumental can be used not only for
detection but also for prediction of minor scale rockfall.
4. DISCUSSION
The first part of the research deals with the detection
of the main rockfalls that occurred during the time
span: rockfall frequency, magnitude, geometry, etc.
The subvertical discontinuities generated after the
1881 landslide constitute the surface of detachment of
the current rockfalls. Toppling is the main failure
mechanism of these rockfalls. A direct relationship
between the intensity of the rainfall and rockfall
frequency is discussed in Abelln et al., (2010).
The second part of this case study deals with the
spatial prediction of rockfalls. This detection was
performed through the detection of the pre-failure
deformation over the whole rock face. Terzaghi (1950)
pointed out the importance of detecting slow surface
movements preceding catastrophic landslides in soils,
even if this deformation is not always detectable using
conventional techniques.
There is scant literature on coupling spatial and
temporal prediction of minor scale rockfalls, such as
those discussed in this study. Rosser et al., (2007)
demonstrated that rockfall frequency may be used as
a precursory indicator of greater failures. One question
that remains to be resolved is as follows: are all
rockfalls preceded by a precursory indicator? Drawing
an analogy with the observations of Terzaghi (1950)
on soil slides: are we currently failing to detect the
phenomena that precede minor rockfalls?

SafeLand - FP7
REFERENCES
Abelln, A. 2009. Improvements in our understanding
of rockfall phenomenon by Terrestrial Laser
Scanning. Emphasis on change detection and its
application to spatial prediction. PhD Thesis,
RISKNAT group, Faculty of Geodynamics and
Geophysics, University of Barcelona, Spain,
185pp.
Abelln, A. Jaboyedoff, M., Oppikoffer, T., Vilaplana,
J.M., 2009. Detection of millimetric deformation
using a Terrestrial Laser Scanner: Experiment and
application to a rockfall event. Natural Hazards and
Earth System Science, 9, 365-372.
Abelln, A., Calvet, J., Vilaplana, J.M., Blanchard, J.,
2010. Detection and spatial prediction of rockfalls
by means of terrestrial laser scanning monitoring.
Geomorphology.
doi:10.1016/j.geomorph.2010.03.016
Blanchard, J,. Calvet, J,. Abelln, A,. Garca, D.,
Khazaradze, G., Vilaplana, J.M., 2008: Estudio del
escarpe del deslizamiento de Puigcercs mediante
lser escner terrestre. Conca de Tremp,
Catalunya. Geotemas 10, 1389-1392.
Corominas, J., Alonso, E., 1984. Inestabilidad de
laderas en el Pirineo cataln. Tipologia y causas.
Inestabilidad de laderas en el Pirineo. Ponencias y
Comunicaciones. ETSICCP-UPC, C.1-C.53.
Page 279 of 401
Deliverable 4.1
Crosta, G.B., Agliardi, F., 2004. Failure forecast for

large rock slides by surface displacement
measurements. Canadian Geotechnical Journal,
40(1), 176-191
Lim, M., Petley, D.N., Rosser, N.J., Allison, R.J., Long,
A. J., Pybus, D., 2006. Combined digital
photogrammetry and time-of-flight laser scanning
for monitoring cliff evolution. Photogrammetric
Record, 20(1), 109-129.
Rosser, N.J., Petley, D.N., Lim, M., Dunning, S.A. &
Allison, R.J., 2005. Terrestrial laser scanning for
monitoring the process of hard rock coastal cliff
erosion. Quarterly Journal of Engineering Geology
and Hydrogeology, 38(4), 363-375.
Rosser, N.J., Lim, N, Petley, D.N., Dunning, S. &
Allison, R.J., 2007. Patterns of precursory rockfall
prior to slope failure. Journal of Geophysical
Research, Vol. 112, No. F4.
Vidal, L.M., 1881. Nota acerca de los hundimientos
ocurridos en la Cuenca de Tremp (Lrida) en
Enero de 1881. Boletin de la Comisin del Mapa
Geolgico de Espaa, Tomo VIII:113-129.

SafeLand - FP7
Page 280 of 401
Case Study 6
Rev. No: 2
Date: 2012-02-15
CASE STUDY 6:
Title: Susceptibility assessment at regional scale based on areal LIDAR datasets.
Application to the canton of Vaud, western Switzerland.
Application: Susceptibility assessment at regional scale.
Technique: Aerial Laser Scanner (ALS) ; Aerial LiDAR.
Main references: Jaboyedoff et al., 2008; Horton et al., 2008 (Section 3, debris flow); Loye
et al., 2009 (Section 4, rockfalls).
Contributor: UNIL (Jaboyedoff, M., Pedrazzini, A., Loye, A., Horton, P., Abelln, A.,
Michoud, C., Derron, M.H.).
ABSTRACT
This case study shows a susceptibility assessment at regional scale. The goal of the study was to develop and
apply different methods in order to provide a fast overview of potential events related to slope mass movement
(landslides, rockfalls and debris flows) at regional scale. The present method was developed for the Canton of
Vaud (2800 km2), western Switzerland. A summary of the methodology for the different phenomena is showed as
follows: (a) LANDSLIDES: Landslide inventory was performed based on the geomorphological observations of
High Resolution DEM, orthophotos and field work campaigns through expert criteria; (b) ROCKFALLS:
susceptibility analysis was performed through the identification of potential rock fall source areas based on a
geomorphometric analysis of DEM together with the calculation of the run-out assessment according to the 3D
shallow angle method; (c) DEBRIS FLOWS: susceptibility analysis was divided into two main steps: source area
identification and spreading area assessment. Both analyses were carried out based on HRDEM. The estimation
of the spreading area was carried out coupling flow direction and run-out distance calculations. The results of this
study allow a quick overview of areas potentially reached by the specific hazard. Results were validated using
datasets from the historical record, showing good agreement with past events and the knowledge of the local
geologists and guides. These susceptibility maps will help the decision-makers of the Vaud canton to prioritize
area of interest for the creation of more detailed, site specific hazard maps.
1. INTRODUCTION
During the last decade, innovative applications in
remote sensing techniques have been developed. The
possibility of acquiring datasets with high accuracy and
spatial resolution mounted on terrestrial, aerial and/or
satellite instrumentals is currently opening up new
ways of visualizing, modelling and interpreting Earth
surface processes. In this case study, High resolution
Digital Elevation Models (HRDEM) acquired from
Aerial Laser Scanner (ALS) was used for a multihazard susceptibility assessment.
Landslides, debris flows and rockfalls, which occur
mainly in mountainous regions, are a severe potential
danger for inhabited regions. Landslide risk in
mountainous areas is increasing as the population and
economic activity increase. The results of this study
allow a quick overview of areas potentially reached by
the specific hazard: landslides, rockfalls or debris
flows. The identification of the most susceptible areas
at regional scale plays a key role for the prioritization
of more detailed, site specific studies. Susceptibility
maps indicate the potentially unstable area, in order to
assess whether an area is potentially endangered or
not by slope. Nevertheless, such maps do not give any
information on the intensity and/or the frequency of
occurrence of the slope movements. This kind of map

SafeLand - FP7
belongs to the first step of the working process leading

toward detailed local danger maps (Lateltin, 1997).
This case study introduces an overview of the
methodologies applied for the following slope
processes: Landslides, debris flows and rockfalls. A
more detailed explanation of the proposed approach
can be found in Jaboyedoff et al., (2008), Horton et al.,
(2008); Loye et al. (2009).
2. LANDSLIDE INVENTORY AND

SUSCEPTIBTIBILITY
2.1 Method
The method focused on the use of GIS tools and GIS
Data (i.e. LiDAR-DEM of 1m x 1m cell size). The
potential instabilities (deep landslides) were identified
by means of a systematic approach over the 3200 km2
of the canton of Vaud, based on LiDAR-DEM analysis,
information provided from the 1:25,000 geological
maps as well as orthophotos. The validation of the
results was based on a review of geological data, field
observations and comparison with available historical
inventories (Noverraz, 1995).
Page 281 of 401
Deliverable 4.1
2.1.1 Landslide inventory

A landslides inventory was performed using following
information: (a) previous inventories (DUTI program;
Noveraz, 1995); (b) a detailed investigation of the
1:25,000 topographic map; (c) geological maps; (d)
LiDAR-DEM (MNT-MO, 2005 SIT) hillshade; (e)
morpho-structural analysis of LiDAR-DEM data; (f)
orthophotos.
area identification and debris flow propagation. Both

the sources identification and the spreading area
assessment are based on a regular grid with a
resolution of 10 meters, which is needed to capture the
variability of the topographic form for hillslopes (Wilson
1996). A more detailed explanation can be found in
(Horton et al., 2008).
2.1.2 Landslide susceptibility

The potential landslide areas were obtained by: (a) the
mapping of geomorphic traces visually analyzed using
the LiDAR-DEM hillshade; (b) the mapping of the
landslide activity carried out through the analysis of the
orthophotos and (c) the use of the historical record,
including previous landslide inventory maps. As
described by Ardizzone et al., (2007), such fine-scale
morphological topographic analysis is a relevant
approach for delineating potential landslides.
According to a systematic approach based on
morphological features (slope failures, scars, deposits,
sagging) all kinds of topographic irregularities (slope
statistics, etc.) were taken as evidence to delineate
current and ancient landslide prone areas. Moreover,
this approach was associated with the analysis of the
1:25000 geological maps (www.swisstopo.ch), using
the sensitivity of lithologies to landsliding, structural
elements such as fault systems and tectonic lines. The
slopes at rivers edges were also considered because
of the effect of enhanced erosion in such a context
(distortion of watercourses, occurrence of natural
dams, etc.) and were therefore used as criterion to
detect landslide activity. Then, the landslide
susceptibility was subdivided in three categories from
proven landslides (clear geomorphologic signatures) to
zones that possess one or more criteria susceptible to
contain landslides, but without any clear features.
The three main susceptibility classes are described as
follow: (i)
Class 1: Zone characterized by the
occurrence of landslides detected by a typical and
complete set of morphological landslide features or
based on previous inventories; (ii) Class 2: Zone
characterized by the occurrence of possible landslides
deduced by some of the morphological evidence but
which cannot be verified without a detailed field work;
(iii) Classes 3: This class was subdivided in three subclasses: 3a) areas characterized by an important
fluvial erosion; 3b) areas showing a rugged topography
(depression, scars, etc.); 3c) areas characterized by a
mean slope value between 18-37, which define a
range of slope suitable to develop landslides (Van
Westen et al, 1997).
3. DEBRIS FLOWS
3.1 Methodology
A new model was developed for a regional debris flow
susceptibility assessment using Matlab environment.
The methodology results in a 2 steps work: source

SafeLand - FP7
Fig. 1. Close up of the landslide prone areas map based on

DEM (MNT-MO, 2005 SIT), orthophotos and geological
map analysis. Each class corresponds to different
morphological evidence and to different types of processes,
as is described in the text. Figure from Horton et al., 2008
3.1.1 Source area identification

Three main criteria are relevant for a debris flow
initiation: sediment availability, water input and
geometry of the slope (Takahashi, 1981; Rickenmann
and Zimmermann 1993, Delmonaco et al., 2003): (i)
Sediment availability was linked to the lithology, as
some geological units produce more or less debris and
fine particles prone to be eroded, such as flysch and
marls; (ii) Water input was estimated using the upslope
contributing area; (iii) Geometry of the slope: the
inclination of each cell (slope angle) is a key
parameter in the triggering of the Debris Flow. Finally,
the plan curvature and the land use map were added
to increase the detection quality.
3.1.2 Spreading area assessment
The debris flow spreading was mathematically
estimated by two types of algorithms: flow direction
algorithms and run-out distance calculations:
(i) Flow direction algorithms: these algorithms
apportion the flow from one cell to its eight neighbours.
The final probabilities are function of the slope and the
persistence. Best fitting with occurred debris flows was
observed using Holmgren's (1994) algorithm.
(ii) Run-out distance calculations: these algorithms
control the distance reached by the debris flow and
reduce their divergence, influencing the flow direction.
The probable maximum run-out is characterized by an
average slope angle of 11 (Huggel et al. 2002).
Page 282 of 401
Case Study 6
3.2 Results
The results are based on the extreme events threshold
for the source areas previously identified (see Horton
et al., 2008). The spreading areas of all sources are
combined by keeping the maximum probability values
The result is the total area exposed to debris flow
spreading, with an associated qualitative probability
qualifying the susceptibility potential, as shown in
figure 2. A surface with a red colour has a higher
probability to be reached by a debris flow than a yellow
one. More accurate identification of the source and
spreading areas are obtained using a HRDM derived
from LIDAR datasets. A good coherence between the
simulation results and field observations was observed
on specific catchments where major debris flow events
occurred.
Rev. No: 2
Date: 2012-02-15
(SAD), characteristic of the main morphological units

(see Loye et al., 2009 for a further explanation).
As regards the interpretation, the peak of each fitted
curve can be correlated to major morphological units.
Two thresholds slope angle were encountered: (a) the
intersection between cliff and steep slopes and (b) the
cliff outcrop only. The maximum of the Gaussian
distribution can be assumed as an apparent stable
slope angle.
The slope angle frequency histogram was performed
on a DEM of 1m x 1m cell size (MNT-MO, 2005
SIT). For each of the five geologically-based units
areas of the Canton of Vaud, a slope angle frequency
distribution was computed and plotted (see Loye et al.,
2009). Slope gradients steeper than the angle defined
at the intersection between the estimated Cliffs and
Mountain sides population were also considered as
potential rockfall source areas, independently of the
lithology and surface cover. This corresponds to the
slope above which the Cliffs data distribution
becomes dominant over the other morphometric
classes (Cliff slope angle limit).
4.2 Run-out zones assessment.
Fig. 2. Spreading results for extreme events for the

Anzeindaz Solalex (Switzerland) region, with representation
of the probability values. (DEM by SITVD. Figure from Horton
et al., 2008
4. ROCKFALLS
One of the most challenging tasks for rockfall
prediction is to define the potential rockfall source
areas and the accurate determination of the reach
susceptibility. The methodology developed in this
section (Loye et al., 2009) results in a 2 steps work:
source area identification and block propagation.
Assessment of the runout zones from the potential

source areas described above was performed by
means of the CONEFALL software (Jaboyedoff and
Labiouse, 2003), which simply implements shallow
angle method (Toppe, 1987; Evans and Hungr, 1988)
computed in a GIS environment with software
CONEFALL (Jaboyedoff and Labiouse, 2003). Thus,
this software allows the estimation of the maximum
runout length in 3D by assuming a given aperture
angle (90- p) centred on the source point (See
Jaboyedoff et al., 2008).
Critical angles for each major unit are listed in Loye et
al. (2009). CONEFALL was applied to each potential
source zone with an aperture angle of 33 for all units.
This was established by comparison with rockfall
events observed on orthophotos and fieldwork
undertaken on test zones. Figure 3 shows the results
of the rockfall source and propagation areas.
4.1 Source area Identification

Unstable rockfall source areas are found in most
cases in ustable, rocky steep slopes. Rockfall
phenomenon is assumed to be controlled by rock type,
structural conditions (discontinuities, foliation, etc.) and
slope gradient.
It can be considered that the type of relief (e.g. glacial,
alluvial, plains, etc.) and the mechanical properties of
the rocks are characteristic for a given morphotectonic
setting (Strahler, 1950). Hence, main morphological
units of the topography (varying randomly around its
mean slope gradient) such as steep slopes, cliffs, etc.,
can be expressed in a unique Slope Angle Distribution

SafeLand - FP7
Fig. 3. Close-up of the Indicative rockfall hazard map.

Rockfall source zones are drawn in red and the runout
perimeter in orange SWISSIMAGE 2004 swisstopo
(DV012716). Figure from Horton et al., 2008
Page 283 of 401
Deliverable 4.1
5. RESULTS AND DISCUSSION

Currently, modern acquisition techniques (Airborne
Laser Scanning, Photogrammetry) enable to obtain a
regional scale DEM with a resolution as high as 1m
making the topographical analysis more and more
relevant since the modelled slope angles are very
close to reality. The quality of the results depends on
the DEM resolution and on the availability of other
different spatial information: the higher the accuracy
and resolution of a given DEM, the more accurate the
results. All the information related to the landslide
susceptibility classes were stored in a GIS database
and form an interactive landslides susceptibility map
all over the territory, which consist of a complete
description of information (distinctive morphological
sign, class, etc.). These characteristics enable an
overview of the state of knowledge of the landslide
prone areas.
ALS proved useful for the updating of previous
landslide inventories (section 2). The limiting factor in
interpreting LIDAR derived DEM hillshade 2D and its
3D visualization depends principally to the data
artifacts. They are mostly related to the occurrence of
locally very dense vegetation cover or occurrence of
clouds during the data acquisition as well as some
steep rugged topography that truncates the laser
signal during the data acquisition. Displaying hillshade
can have some kind of shading effects as well, which
creates inaccurate or even incorrect morpho-structural
features.
Process based modelling of debris flow is difficult
because of the complexity of the phenomenon and the
variability of controlling factors. The automatic
identification of sources areas and the estimation of
debris flow spreading, based on GIS tools, provide a
substantial basis for a preliminary susceptibility
assessment at a regional scale.
The method used in section 4 for source detection
areas is a rather conservative way to consider cliff
faces and slopes surfaces to be potentially instable
over a certain slope gradient. Likewise, CONEFALL
allows quick but accurate delineation of potential
rockfall prone perimeters, resembling with a 3D
trajectography model, although it doesnt require
physically-based parameters acquired on field.
6. GENERAL CONCLUSION
Nowadays, the availability of LiDAR-DEM represents
an important tool either to obtain a rapid overview of
potential unstable areas or to perform more detailed
studies as well. The methods analyzed in this case
study show the potential of the DEM-based analysis
for the estimation of the susceptibility of common slope
process phenomena in mountain areas. These
methods, relatively easy to use, allow a quick overview
of the potential hazard affecting a large territory by
using LIDAR-DEM. As a result, significant

SafeLand - FP7
information of the zones where a more detailed

analysis must be carried out is provided.
REFERENCES
Ardizzone, F. Cardinali, M., Galli, M., Guzzetti, F., and
P. Reichenbach 2007. Identification and mapping
of recent rainfall-induced landslides using elevation
data collected by airborne Lidar. Nat. Hazards
Earth Syst. Sci., 7, pp. 637-650.
Delmonaco, G., Leoni, G., Margottini, C., Puglisi, C.
and Spizzichino, D. 2003. Large scale debris-flow
hazard assessment: a geotechnical approach and
GIS modelling. Natural Hazards and Earth System
Sciences, Vol. 3, pp. 443-455.
Holmgren, P. 1994. Multiple flow direction algorithms
for runoff modelling in grid based elevation models:
An empirical evaluation. Hydrological Processes,
Vol. 8(4), pp. 327-334.
Horton, P., Jaboyedoff, M., Bardou , E. 2008. Debris
flow susceptibility mapping at a regional scale. In
Locat, J., Perret, D., Turmel, D., Demers, D., and
Leroueil, S., 2008. Proceedings of the 4th Canadian
Conference on Geohazards. From Causes to
Management. Presse de lUniversit Laval,
Qubec, 594p.
Huggel, C., Kaab, A., Haeberli, W., Teysseire, P. and
Paul, F. 2002. Remote sensing based assessment
of hazards from glacier lake outbursts: a case
study in the Swiss Alps. Canadian Geotechnical
Journal, Vol. 39(2), pp. 316-330.
Hungr O. and Evans S. G. 1988. Engineering
evaluation
of
fragmental
rockfall
hazard.
Proceedings of the 5th International Symposium on
Landslides, Lausanne, pp. 685-590.
Jaboyedoff, M. and Labiouse, V. 2003. Preliminary
assessment of rockfall hazard based on GIS data.
ISRM 2003 - Technology roadmap for rock
mechanics, South African Institute of Mining and
Metallurgy.
Jaboyedoff, M., Pedrazzini, A., Horton, P., Loye, A.,
Surace. 2008. Preliminary slope mass movements
suceptibility mapping using LIDAR DEM.
Proceedings,
61st
Canadian
Geotechnical
Conference and 9th Joint CGS/IAH-CNC
Groundwater Conference. Edmonton, AB.
Lateltin, O., 1997. Recommandations - Prise en
compte des dangers dus aux mouvements de
terrain dans le cadre des activits de
lamnagement du territoire. Office fdral de
lEnvironnement, des forts et du paysage.
Loye, A., Jaboyedoff, M., and Pedrazzini, A. 2009.
Identification of potential rockfall source areas at a
regional
scale
using
a
DEM-based
geomorphometric analysis, Nat. Hazards Earth
Syst. Sci., 9, 1643-1653.
Noverraz F. 1995. Carte des instabilits de terrain du
Canton de Vaud. Ecole Polytechnique Fdrale de
Lausanne 33p.
Rickenmann, D. and Zimmermann, M. 1993. The 1987
debris flows in Switzerland: documentation and
analysis. Geomorphology, Vol. 8(2-3), pp. 175-189.
Page 284 of 401
Case Study 6
Rev. No: 2
Date: 2012-02-15
Strahler, A. N. 1950. Equilibrium theory of erosional

slopes approached by frequency distribution
analysis, Am. J. Sci., 248, 673696, 800814.
Takahashi, T. 1981. Estimation of potential debris
flows and their hazardous zones: Soft
countermeasures for a disaster. Natural Disaster
Science, Vol. 3(1), pp. 57-89.
Toppe, R. 1987. Terrain models a tool for natural
hazard mapping. In Salm, B. and Gubler, H.,
editors, Avalanche formation, movement and
effects. IAHS Publication no. 162, 62938.
Van Westen, C.J., Rengers, N., Terlien, M. and
Soeters, R. 1997. Prediction of slope instability
phenomena through GIS-based hazard zonation.
Geol. Rundschau, 86, pp. 404-414.
Wilson, J.P. 1996. GIS-based land surface/subsurface
modeling: new potential for new models, Third
International Conference / Workshop on Integrating
GIS and Environmental Modeling, Santa Fe, New
Mexico, pp. 21-26.

SafeLand - FP7
Page 285 of 401
Case Study 7
Rev. No: 2
Date: 2012-02-15
CASE STUDY 7:
Title: The application of LiDAR-derived images for landslide inventory and susceptibility
mapping in the Flemish Ardennes, Belgium.
Application: Landslide inventory and susceptibility mapping at regional scale.
Technique: Aerial LiDAR.
Main references: This case study is based on following publications: Van Den Eeckhaut et
al. (2006; 2007a,b; in press).
Contributor: JRC (Van Den Eeckhaut M., Poesen J., Vandekerckhove L., Hervs J.).
ABSTRACT
Large, deep-seated soil slides are common features in the Flemish Ardennes (Belgium). As many of these old (>
100 years) landslides are located under forest in this hilly region, aerial photo interpretation is not an appropriate
landslide mapping method. Van Den Eeckhaut et al. (2007a) tested the potential of airborne LiDAR images for
mapping old active and dormant landslides in a 125 km test area in the Flemish Ardennes, and concluded that
that large-scale LiDAR-derived hillshade and contour line maps analyzed by experienced geomorphologists can
significantly improve field survey-based inventories of landslides with a subdued morphology in hilly regions. In the
same area LiDAR-derived slope and aspect maps were applied successfully, together with lithological maps, for
statistical landslide susceptibility mapping (Van Den Eeckhaut et al., 2006). This study presents the LiDARderived landslide inventory and susceptibility map of the complete Flemish Ardennes (720 km), and is a good
example of science meeting policy as the scientific results are used by qualified authorities to inform local
people. On the website of the Environment, Nature and Energy Department of the Flemish Government
(http://www.lne.be/themas/bodem/grondverschuivingen) the landslide inventory and landslide susceptibility maps
are available and prevention and mitigation measures are presented.
1. INTRODUCTION
The Flemish Ardennes is a hilly region marked by old,
large, deep-seated landslides with an affected area of
at least 1 ha and an assumed shear plane depth of
more than 3 m (Van Den Eeckhaut et al., 2005). Due
to the vicinity of cities such as Brussels and Ghent
(Figure 1), the forested hills of this region have
become attractive residential areas, and it appears
that the expansion of settlements and human
interference has increased vulnerability to landsliding.
Indeed, reactivations of old landslides, causing
damage to public and private property have been
documented, mostly during winter and spring, after
periods of persistent rainfall. This has resulted in an
increased interest in the process of landsliding and its
related problems by the local and regional authorities.
Optimal approaches to reduce landslide risk generally
comprise a mix of four strategies (Schuster and
Kockelman, 1996), among which reduction of
development in landslide-prone areas, is the most
effective and economical. To apply this strategy, the
locations of landslide-prone areas must be known. As
it is believed that past and present landslide locations
are the key to prediction of future landslide locations
(Carrara et al., 1995), a detailed landslide inventory
map is indispensable for the production of landslide
susceptibility maps (Ardizzone et al., 2002). Therefore
this study provides an overview of the LiDAR-derived

SafeLand - FP7
landslide inventory and susceptibility map of the

complete Flemish Ardennes (720 km).
2. MATERIALS AND METHOD

2.1 LiDAR and landslide inventory mapping
A detailed description of the LiDAR-data can be found
in Van Den Eeckhaut et al. (2007a). For the complete
Region of Flanders, LiDAR data was collected in 2001
and 2002. The original laser pulses had an average
pulse density of 1 per 4 m and the post-processing
was conducted by the vendor. The Terrascan software
was used to automatically remove undesired returns
and to create the bald earth DEM. Afterwards a
manual check followed. The database finally provided
(AGIV, 2004) has a uniform point density of at least 1
per 20 m. The accuracy of the altitude depends on the
soil cover and decreases with increasing vegetation
heights from 7 cm for freshly cut lawns to 20 cm for
forests and pastures (GIS-Vlaanderen, 2003).
A DEM with a 2 m x 2 m resolution was derived from
the LiDAR data using TIN interpolation. Production
artifacts were removed with low pass filtering. Then, a
slope map, contour line map (2 m interval) and two
hillshade maps were created. The two hillshade maps
had a sun elevation angle of 30 and a sun azimuth
angle of 315 and 45. No vertical exaggeration was
used.
Page 287 of 401
Deliverable 4.1
Fig. 1. Study area in the Flemish Ardennes (720 km ): (A) Landslide inventory overlain on LiDAR-derived hillshade map. White
rectangles show excerpts of (B) and (C); (B) Excerpt of the hillshade map showing shallow active landslides under pasture; (C)
Excerpt of the hillshade map showing three deep-seated rotational earth slides under forest.
The identification of landslides on hillshade and

contour line maps is based on the recognition of
landslide characteristics (e.g. main scarps, reverse
slopes, convex landslide foots) and alterations of the
drainage system (e.g. a main valley stream pushed
away by a landslide; Wills and McCrink, 2002; Van
Den Eeckhaut et al., 2007a).

SafeLand - FP7
Therefore between three and seven experts were

asked to digitize the presumed landslides on a vector
map which could be overlain with the LiDAR-derived
maps.
Afterwards, the LiDAR-derived landslide inventory map
was compared with the landslide inventory produced
from intensive field surveys before LiDAR data
Page 288 of 401
Case Study 7
became available for the study area (Van Den

Eeckhaut et al., 2005).
2.2 LiDAR-derived maps for landslide susceptibility
mapping
Understanding the role of individual factors controlling
landslide location is important for predicting where
landslides can occur in the future, i.e. to ascertain
landslide susceptibility. In the past three decades, a
large number of qualitative and quantitative methods
were applied and tested to assess landslide
susceptibility (e.g. Carrara et al., 1995; Guzzetti et al.,
1999; Glade and Crozier, 2005). For modeling
landslide susceptibility on a regional scale, logistic
regression is nowadays a common statistical modeling
technique (e.g. Carrara et al. 1995; Greco et al.,
2007). We adopted logistic regression to find the bestfitting model describing the relationship between the
dependent variable (i.e. the presence or absence of
depletion areas of large, deep-seated landslides) and
a set of independent variables (i.e. terrain height,
slope gradient, aspect, plan and profile curvature,
Tertiary geology, soil drainage, distance to rivers and
distance to faults). A 10 m x 10 m resolution was used
and all topographical data was derived from LiDAR.
3. RESULTS AND DISCUSSION

3.1 Internal morphology of landslides
On the LiDAR-derived hillshade map, the internal
morphology of the landslides is clearly visible. Figure
1C shows three deep-seated rotational earth slides, all
located under forest with the main scarps and several
reverse slopes clearly distinguishable. The main
scarps are indicated in white, because they are
oriented to the light source located in the northwest.
The reverse slopes, which are oriented towards the
east and thus lying in the shadow. Also several ditches
draining ponding water from some of the reverse
slopes are visible, and the curved form of the brook at
the foot of the most eastern landslide indicates that it
has been pushed to the west by the displaced slope
material.
3.2 Landslide inventory
The regional inventory (Figure 1A) shows 210
landslides, equivalent to a density of one landslide
every 3.4 km. However, landslides are not equally
distributed throughout the area. 77.6% (n=163) of the
total number of observed landslides are deep-seated,
i.e. the shear surface was estimated to be deeper than
3 m, and larger than 1 ha (average 4 ha; e.g. Figure
1C). The estimation of shear surface depth is based
upon the deformation of the terrain and the height and
shape of the main scarp. These deep-seated
landslides are mainly classified as rotational earth
slides (Cruden and Varnes 1996) Shallow landslides
represent 22.3% (n=47) of all landslides in the
inventory, and individually occupy an area of less than

SafeLand - FP7
Rev. No: 2
Date: 2012-02-15
1 ha (average 0.5 ha; Figure 1B). They have an

estimated depth of less than 3 m, and are classified
chiefly as rotational slides with a flow component at the
toe. Many of the shallow failures occurred inside preexisting deep-seated landslides.
Important here is that the analysis of the LiDARderived maps resulted in an improvement of the field
survey-based landslide inventory map. First, the
boundaries of landslides under forest could be
mapped more accurately and secondly LiDAR
derivates allowed experts to have a complete overview
of the large landslides.
3.3 Landslide susceptibility map
Several models based on different combinations of
independent variables were evaluated with the same
evaluation parameters as explained in Van Den
Eeckhaut et al. (2006; i.e. Area under Receiver
Operating Characteristic Curve; Kappa value and
other parameters calculated from confusion matrices).
From this evaluation procedure, we found that the best
landslide susceptibility model can be written as:
p
log(
)
1 p
= - 13.418 + (0.386 x slope
gradient) + (2.520 x NW) + (2.948 x W) + (2.043 x SW)

+ (2.399 x S) + (1.653 x SE)+ (2.337 x GeVl) + (2.407
x GeMe) + (1.488 x Tt) + (1.381 x KoAa)
(1)
where p is a value between 0 and 1 reflecting the
probability of occurrence of a landslide depletion
(initiation) area. Slope gradient, the presence of
northwest (NW), west (W), southwest (SW), south (S)
and southeast (SE) oriented hillslope sections, and the
presence of lithological formations with an important
clay content (GeVl, GeMe, Tt and KoAa) are the
independent
variables
influencing
landslide
occurrence. Slope gradient is the most important
environmental factor controlling landslide occurrence,
and landslides are expected to occur on hillslopes with
a slope gradient above 0.10 m.m-1 and a southeast to
northwest orientation where a lithology rich in swelling
clays is located at relatively shallow depth. Landslide
accumulation zones are difficult to delineate by this
model as the landslide debris is sometimes deposited
on slope sections with a 0.05 m.m-1 slope gradient.
Figure 2 shows the classified landslide susceptibility
map produced from model (1). 7% of the study area
(i.e. ca. 7 times the area mapped as affected by
landslides) is classified with very high to moderate
susceptibility to landslides and 84% of the mapped
landslide grid cells are correctly classified susceptible.
4. CONCLUSIONS
The results of this study revealed that for hilly regions
affected by active and dormant landslides high-quality
landslide inventory maps can be obtained through the
combination of the analysis of LiDAR-derived maps in
Page 289 of 401
Deliverable 4.1
Fig. 2. Landslide susceptibility map of the Flemish Ardennes (720 km )
a GIS environment, where possible by different

experts, and detailed field checks. Indeed, for
landslide inventory mapping the analysis of LiDAR
derivates and field surveys is complementary. Field
checks remain necessary to exclude false positives
(e.g. quarries) and to map small and shallow
landslides, and landslides that occurred after LiDARdata was collected.
The results further show that high resolution slope and
aspect maps derived from LiDAR represent important
tools for landslide susceptibility mapping.
REFERENCES
AGIV, 2004. DHM Vlaanderen, LIDAR hoogtepunten
brondata, GIS-Vlaanderen, AGIV.

SafeLand - FP7
Ardizzone F, Cardinali M, Carrara A, Guzzetti F,

Reichenbach P. 2002. Impact of mapping errors on
the reliability of landslide hazard maps. Natural
Hazards and Earth System Sciences 2: 3-14.
Carrara A, Cardinali M, Guzzetti F, Reichenback P.
1995. GIS technology in mapping landslide hazard.
In Geographical Information Systems in Assessing
Natural Hazards, Carrara A, Guzzetti F (eds).
Kluwer Acad. Publ.: Dordrecht; 135-176.
Cruden DM, Varnes DJ. 1996. Landslide Types and
Processes. In Landslides, Investigation and
Mitigation, Transportation Research Board.
National Research Council, Special Report 247,
Turner AK, Schuster RL (eds). National Academy
Press: Washington, DC; 36-71.
GIS-Vlaanderen, 2003;
http://www.gisvlaanderen.be/gis/diensten/geovlaanderen/?artid=240
Page 290 of 401
Case Study 7
Rev. No: 2
Date: 2012-02-15
Glade, T. and Crozier, M. J., 2005. The nature of

landslide hazard impact. In: Glade, T, Anderson,
M.G. and Crozier, M. J., (Eds.) Landslide Hazard
and Risk. Wiley, Chichester, pp.43-74
Greco, R., Sorriso-Valvo, M., Catalano, E., 2007.
Logistic Regression analysis in the evaluation of
mass movements susceptibility: The Aspromonte
case study, Calabria, Italy. Engineering Geology
89, 47-66.
Guzzetti, F., Carrara, A., Cardinali, M., Reichenbach,
P., 1999. Landslide hazard evaluation: a review of
current techniques and their application in a multiscale study, Central Italy, Geomorphology, 31,
181-216.
Schuster, R.L., Kockelman, W.J., 1996. Principles of
landslide hazard reduction. In Landslides
Investigation and Mitigation, National Research
Council, Special Report 247, Turner AK, Schuster
RL (Eds.). National Academy Press: Washington,
DC; pp. 91-103.
Van Den Eeckhaut, M., Poesen, J., Verstraeten, G.,
Vanacker, V., Moeyersons, J., Nyssen, J., Van
Beek, L.P.H., 2005. The effectiveness of hillshade
maps and expert knowledge in mapping old deepseated landslides. Geomorphology 67: 351-363.
Van Den Eeckhaut, M., Vanwalleghem, T., Poesen, J.,
Govers, G., Verstraeten, G., Vandekerckhove, L.,
2006. Prediction of landslide susceptibility using
rare events logistic regression: a case-study in the
Flemish Ardennes (Belgium). Geomorphology 76,
392-410.
Van Den Eeckhaut, M., Poesen, J., Verstraeten, G.,
Vanacker, V., Nyssen, J., Moeyersons, J., Van
Beek, L.P.H., Vandekerckhove, L., 2007a. The use
of LIDAR-derived images for mapping old
landslides under forest. Earth surface processes
and landforms 32, 754-769.
11. Van Den Eeckhaut, M., Poesen, J., Verstraeten,
G., 2007b. Vervolgproject - Opstellen van een
gevoeligheidskaart
met
betrekking
tot
grondverschuivingen voor de Vlaamse Ardennen.
Rapport In opdracht van Vlaamse Overheid, Dpt
LNE,
ALBON,
Brussel,
pp.
97.
http://www.lne.be/themas/bodem/grondverschuivin
gen/grondverschuivingen
Van Den Eeckhaut, M., Poesen, J., Vandekerckhove,
L., Van Gils, M., Van Rompaey, A., in press.
Human-environment interactions in residential
areas susceptible to landsliding: the Flemish
Ardennes case-study. Area.
Wills CJ, McCrink TP. 2002. Comparing landslide
inventories, the map depends on the method.
Environmental and Engineering Geoscience 8:
279-293.

SafeLand - FP7
Page 291 of 401
Case Study 8
Rev. No: 2
Date: 2012-02-15
CASE STUDY 8:
Title: Characterization and analyses of surface deformations, mass movements and alpine
region lifting with PSInSAR method in NW Slovenia.
Application: Landslide detecting.
Technique: PSInSAR.
Main references: This case study is based on following publications: Komac and Bavec,
2007a, Komac and Bavec, 2007b.
Contributor: GeoZS (Carman, M., Kumelj, S., Jemec, M.).
ABSTRACT
PSInSAR method was used for mass movement detection on an area which spreads over 700km2 and lies in the
NW part of Slovenia. The research area forms the eastern flank of the Alpine arch and is tectonically still active.
For the analytical purposes 57 images of descending orbit from satellites ERS 1 and ERS 2 were used. The time
span of the acquired images was from April 1992 to December 2000. The average signal reflector (PS) density for
the area was 23 per km2. Altogether 16304 permanent scatters were detected. For the best 10 % (1646 PS), time
series of displacements were acquired. The results also show a constant uplift of Alps and they indicate that the
uplift is of higher magnitude than it was considered until now. The relative uplift in relation to the reference point in
the Alpine foreland ranges up to 3.35 mm per year.
1. INTRODUCTION
Slovenia lays in the eastern flank of Alpine arch and
an active tectonics present are main driving forces for
mass movements in the region, especially the northwestern part of Slovenia. The research area was
chosen as the study area due to its neotectonic activity
(Poljak et al., 2000; Zupani et al., 2001; Grenerczy
et al., 2005) and due to a number of landslide, rockfall
and debris flow occurrences (Komac et al., 2005). The
proved active tectonics of the region is a consequence
of compression field with approximate tension field 1
in N-S direction (Placer, 1998; Grenerczy et. al., 2005;
Weber et al., 2006; Rinar et al., 2007). There are
several major active faults (Idrija, Sava and Ravne
fault) and an active South Alps thrust (Poljak, 2000).
These faults are most probably right displacement
faults, at least this can be stated for Ravne fault
(Zupani et al., 2001; Bajc et al., 2001). The area is
subdivided into several nappes (Julian nappe, Tolmin
nappe, Trnovski nappe; after Placer, 1998). Wider
area is classified as of middle seismic activity area
(Poljak et al., 2000). During the InSAR data
acquisition, between April 1992 and December 2000, a
major earthquake occurred in the research area with
magnitude of MW = 5.6 (Bajc et al., 2001; Gosar et al.,
2001; Zupani et al., 2001).
The research area consists mainly of Mesozoic
carbonate rocks, some flysch clastites, in the northern
part of Paleozoic clastic and carbonate rocks. The net
of fluvial and glacial valleys is filled with Quaternary
sediments. (Buser, 1987; Jurkovek, 1987).
Modern satellite radar permanent (also persistent)
scatterer interferometric technique (PSInSAR) enables
very accurate monitoring of relative vertical

SafeLand - FP7
displacement velocities of observed surfaces and

grounds (Ferretti et al., 2001; Ferretti et al., 2005;
Brgmann et al., 2006; Dixon et al., 2006; Ferretti &
Crespa, 2006). Technique is also very useful in
geology for monitoring coseismic and aseismic
tectonic displacements (Massonnet et al., 1993;
Massonnet et al., 1994; Dixon, 1995; Peltzer et al.,
1996; Massonnet et al., 1996; Peltzer et al., 1999),
slow moving landslides (Ferretti et al; 2001; Colesanti
et al., 2003a; Hilley et al., 2004), and swelling of
ground or subsidence (Carnec and Delacourt, 2000;
Ferretti et al., 2000; Colesanti et al., 2003b; Vasco &
Ferretti, 2005).
2. RESEARCH AREA
PSInSAR technique is also very useful in geology for
monitoring
coseismic
and
aseismic
tectonic
displacements, slow moving landslides and swelling of
ground or subsidence. The research area is located in
the NW part of Slovenia, bordering the Italy and
covering the south-eastern flank of Alps, the area of
Julian Alps (Figure 1).
To assess the tectonic and landslide mass movements
in the research area, geological (Buser, 1987;
Jurkovek, 1987), structural data, digital elevation
model (SMA, 2001), landslide occurrences (Komac et
al., 2005), seismic (Placer, 1998; Poljak et al., 2000;
Bajc et al., 2001; Gosar et al., 2001; Zupani et al.,
2001; Grenerczy et al., 2005; Weber et al., 2006;
ivi, 2006; Rinar et al., 2007), rainfall (ARSO,
2006) and geophysical (Permanent Scatterrer
InSAR; T.R.E., 2006) data were used. Analyses were
focused into assessment of applicability of PSInSAR
technique for monitoring the uplifts or subsidence of
Page 293 of 401
Deliverable 4.1
masses, either as a consequence of endogenic

(tectonics) or exogenic (gravitation, climate) forces.
The PSInSAR method shown as important tool for

characterisation and detecting tha landslides. The
research was primarily focused on tectonic uplifting.
Fig. 2. Vertical displacements (mm) related to mass

movements on four sites, Jablenca (LS 1), Koritnica (LS 2),
Modrej (LS 3), and Ilovica (LS 4). Monthly rainfall cumulative
values are represented with bars.
Fig. 1. The research area. The reference was set in the town
of Tolmin represented in the Figure 1 with the star. Landslide
sites are marked with LS 1 to LS 4.
3 SLOPE MASS MOVEMENTS

Slope mass movements in the research area, detected
with the PSInSAR, are more related to tectonic activity
than to other triggering factors (i.e. rainfall, human
activity). A general overlook of the research area
indicates that majority of the subsiding PS indicate
their connection to scree deposits and gravitational
processes related to them. Analyses of average
displacements have been conducted on several sites,
Jablenca (6 PS), Koritnica (2 PS), Modrej (6 PS), and
Ilovica (1 PS), marked as sites LS1, LS2, LS3, and
LS4 respectively in Figure 1. From the movements of
observed sites, trends of environs have been deducted
to eliminate aseismic displacements in the surrounding
areas. Displacement results have been compared to
rainfall and seismic data to assess the causes of
displacements. In case of Jablenca landslide it can be
concluded that displacements are more governed by
seismic activity than by rainfall, while for Koritnica and
Ilovica sites the rainfall seems to be the main
triggering factor, in case of later with a delay of one to
two months, most probably due to its geological setting
(shale with sandstone and limestone). In case of
Modrej movements are most probably a combination
of several factors, since neither seismic activity or
rainfall are evidently related to displacements. Figure 2
represents the comparison of landslide displacements
with monthly rainfall cumulative values.
4 CONCLUSIONS

SafeLand - FP7
In case of uplifting the reasons are more or less clear,

it is the consequence of active tectonics and those
extreme ones are probably the result of locally limited
conditions. On the contrary, the subsidence or
moving downwards is more of a complex character. It
is probably the combination of simultaneous influence
of tectonics and gravitation. Most extreme movements
are most probably indications of gravitational mass
movements landslides. To a certain degree mass
displacements in form of landslides can be monitored,
although same problems as with tectonic activity exist.
The uplifting velocity values have to be interpreted
very carefully, since the horizontal displacements can
blur real vertical displacements noticeably.
Further investigations would have to be upgraded with
ground measurements and with InSAR data from
ascending orbits to cover the south-facing slopes
where majority of landslides in the area occur. Also
more precise geological assessment of individual PS
would have to be conducted.
REFERENCES:
Bajc, A., Aoudia, A. Sarao, P. Suhadolc, 2001. The
1998 Bovec-Krn mountain (Slovenia) earthquake
sequence, Geophys. Res. Lett., vol 28, p.p. 1839
1842.
Buser, S., 1987. Basic geological map of SFRY,
sheet Tolmin and Udine, 1:100.000, Federal
Geological Survey, Belgrade.
Brgmann, G. Hilley, A. Ferretti and F. Novali, 2006.
Resolving vertical tectonics in the San Francisco
Bay Area from permanent scatterer InSAR and
GPS analysis, Geology, Vol 34/3, p.p. 221-224.
Colesanti, A. Ferretti, C. Prati and F. Rocca, 2003a .
Monitoring Landslides and Tectonic Motion with
the Permanent Scatterers Technique, Engineering
Geology, vol 68, p.p. 3-14.
Carnec and C. Delacourt, 2000. Three years of
mining
subsidence
monitored
by
SAR
interferometry, near Gradane, France, Journal of
Applied Geophysics, vol 43, p.p. 43-54.
Page 294 of 401
Case Study 8
Colesanti, A. Ferretti, F. Novali, C. Prati and F. Rocca,

2003b. SAR Monitoring of Progressive and
Seasonal Ground Deformation Using the
Permanent
Scatterers
Technique,
IEEE
Transactions on Geoscientific Remote Sensing, vol
41, p.p. 1685-1700.
Dixon, F. Amelung, A. Ferretti, F. Novali, F. Rocca, R.
Dokka, G. Sella, S.W. Kim, S. Wdowinski and D.
Whitman, 2006. Subsidence and flooding in New
Orleans, Nature, vol 441, p.p. 587588.
Ferretti, M. Bianchi, C. Prati and F. Rocca, HigherOrder Permanent Scatterers Analysis, EURASIP
Journal on Applied Signal Processing, vol 20, p.p.
32313242, 2005
Ferretti, C. Prati and F. Rocca, 2001. Permanent
Scatterers
in
SAR
Interferometry,
IEEE
Transactions On Geoscience And Remote
Sensing, vol 39/1, p.p. 820.
Ferretti and S. Crespa, 2006. Advances in Differential
SAR
Interferometry:
from
DInSAR
to
PSInSARTM, CSTARS Height Workshop
October, 24th 2006, Miami, Florida.
Gosar, R. Stopar, M. Car and M. Mucciarelli, 2001.
The earthquake on 12 April, 1998 in Krn
mountains (Slovenia): Ground motion amplification
study using microtremorsand modelling based on
geophysical data, Journal Appl. Geophys., vol
47/2, p.p. 153167.
Grenerczy, G. Sella, S.Stein and A. Kenyeres, 2005.
Tectonic implications of the GPS velocity field in
the northern Adriatic region, Geophys.Res.Lett.,
vol
32,
article
L16311,
doi:
10.1029/2005GL022947.
Jurkovek, B., 1987. Basic geological map of SFRY,
sheet Beljak and Pontebba, 1:100.000, Federal
Geological Survey, Belgrade.
Komac, M., Bavec, M., 2007a. Detection of mass
movements in Alpine Slovenia using PSInSAR
data. 32nd International Symposium on Remote
Sensing of Environment, San Jos, 25-29 June
2007. Sustainable development through global
earth observations.
Komac, M., Bavec, M., 2007b. PSInSAR data analysis
- an insight into active tectonics and mass
movements in west Slovenia. V: Mednarodna
konferenca Gospodarjenje z odpadki, okoljska
geotehnologija
in
trajnostni
razvoj:
ICWMEGGSD'07 - GzO'07, Avgust 28.-30.,2007,
Ljubljana,
Slovenija.
Ljubljana:
Naravoslovnotehnika fakulteta, Oddelek za
geotehnologijo in rudarstvo, 2007, 10 p.

SafeLand - FP7
Rev. No: 2
Date: 2012-02-15
Komac, J. inigoj, M. Krivic, . Kumelj, K. Hribernik

and A. Vehovec, 2005. Pregled in analiza
podatkov v obstojeih bazah plazov za novelacijo
baze gis_ujME, Geoloki zavod Slovenije, 51 str.,
Ljubljana.
Massonnet, D., M. Rossi, C. Carmona, F. Adragna, G.
Peltzer, K. Feigl and T. Rabaute, 1993. The
displacement field of the Landers earthquake
mapped by radar interferometry, Nature, vol 364.
Massonnet, D., K.L. Feigl, M. Rossi and F. Adragna,
1994.
Radar
interferometric
mapping
of
deformation in the year after the Landers
earthquake, Nature, vol 369.
Massonnet, D., W. Thatcher and H. Vadon, 1996.
Detection of post-seismic fault zone collapse
following the Landers earthquake, Nature, vol 382.
Placer, L., 1998. Contribution to the macrotectonic
subdivision of the border region between southern
Alps and External Dinarides, Geologija, vol 41,
p.p. 223255, 1998.
Poljak, M., 2000. Structural-Tectonical Map of
Slovenia 1 : 250.000, Geological Survey of
Slovenia.
Poljak, M. ivi and P. Zupani, 2000. The
seismotectonic charcteristics of Slovenia, Pure
Applied Geophysics, vol 157, p.p. 2755.
Rinar, B. Koler and M. Bavec, 2007. Recentna
aktivnost regionalnih geolokih struktur v zahodni
Sloveniji, Geologija, vol 50/1 (in press), Ljubljana.
Vasco &. Ferretti, 2005. On the use of quasi-static
deformation to understand reservoir fluid flow,
Geophysics, vol 70/4, p.p. O13-O27.
Zupani, I. Ceci, A. Gosar, L. Placer, M. Poljak and
M. ivi, 2001. The earthquake of 12 April 1998
in the Krn Mountains (Upper Soa valley, Slovenia)
and its seismotectonic characteristics, Geologija,
vol 44/1, p.p. 169192.
Weber, M. Vrabec, B. Stopar, P. Pavlovi Preeren
and T. Dixon, 2006.The PIVO-2003 experiment: a
GPS study of Istria peninsula and Adria microplate
motion, and active tectonics in Slovenia, In: Pinter,
N., Grenerczy, G., Weber, J., Stein, S., Medak, D.
(eds): The Adria microplate: GPS geodesy,
tectonics and hazards. NATO Science Series, IV,
Earth and Environmental Sciences, vol 61, p.p.
305-320, Springer.
Page 295 of 401
Case Study 9
Rev. No: 2
Date: 2012-02-15
CASE STUDY 9:
Title: Landslide detection and mapping at the basin scale by means of PSI technique. The
case study of the Arno River Basin, Italy.
Application: Landslide detection and mapping at regional scale.
Technique: PSI (Persistent Scatterers Interferometry).
Main references: This case study is based on the following publications: Lu et al., 2009;
Farina et al., 2006; Canuti et al., 2008.
Contributor: UNIFI (Tofani V., Catani F., Casagli N.)
ABSTRACT
SAR interferometry (InSAR) has already shown its significance for landslide mapping. However, temporal
decorrelation and atmospheric disturbances limit the usefulness of traditional differential InSAR techniques. The
recently developed Permanent Scatterers (PS) technique removes the temporal decorrelation and atmospheric
artifacts by generating radar benchmarks derived from a multi-interferogram analysis of SAR images. PS are
suitable for investigating slow moving landslides because they are able to detect ground displacements with the
accuracy of millimeters.
The PS analysis was applied at a regional scale for the integration of the PS measurements within a landslide
inventory for the presence of a high number of mass movements. About 350 ERS 1/ERS2 SAR images have
been interferometrically processed by means of the PS technique, with the detection of about 600,000 PS.
However, because of the large number of PS that can be identified, the effective extraction of information useful
for landslide studies from this PS technique sometimes remains difficult. With the aim of mapping landslides
rapidly and (semi-) automatically, the hotspot analysis on the PS present within the Arno river basin (Italy) using
spatial statistics approach has been performed. Four years (2003-2006) of RADARSAT SAR images within the
basin have been processed so as to identify slow moving landslides. Gi* statistics for the local test on PS datasets
and the kernel function for PS density estimation based on the Gi* values have been applied. The output is the
hotspot map which emphasizes the existing mass movement. This methodology offers an innovative tool for
extracting useful information from PS, thus providing an effective way of landslide rapid mapping.
1. INTRODUCTION
Remote sensing is useful for landslide studies. SAR
interferometry (InSAR) is an important branch of
remote sensing (Bamler & Hartl 1998) and is a
valuable tool for landslide mapping and monitoring
(Corsini et al. 2006). Combined with both amplitude
and phase parameters, an interferogram can be
generated with the radar images of the same area.
After unwrapping interferogram fringes, it is capable to
detect ground movement with millimetric accuracy
(Massonnet & Feigl 1998). However, the usefulness of
traditional differential InSAR (DInSAR) techniques is
limited by factors such as temporal decorrelation and
atmospheric disturbances (Fruneau et al. 1996,
Massonnet & Feigl 1998, Kimura & Yamaguchi 2000,
Ferretti et al. 2001).
The
multi-image
Persistent
Scatterers
SAR
Interferometry (PSI) technique (Ferretti et al. 2001,
Colesanti et al. 2003, Mora et al. 2003, Werner et al.
2003, Duro et al. 2005), has showed its capability to
provide information about ground deformations over
wide area with millimetric preci-sion, making this
technique suitable for both regional and slope scale
mass movements investigations. In particular the
Permanent Scatterers (PS) technique patented by the
Politecnico di Milano and is commercially available

SafeLand - FP7
through the POLIMI spin-off company TeleRilevamento Europa (TRE). PS technique produces
radar benchmarks derived from a multi-interferogram
analysis of SAR images. The temporal decorrelation
and atmospheric artifacts can be meanwhile estimated
and removed (Ferretti et al. 2001). Some successful
cases have shown the suitability of PS for investigating
slow moving landslide (Colesanti et al. 2003, Casagli
et al. 2005, Farina et al. 2006).
In this contribution examples of the application of the
PS technique to landslide mapping at regional scale
are described. The study area is the Arno river basin
located in the Northern Apennines, Italy.
In particular the updating of the landslide inventory
map carried by means of the ERS and ENVISAT PS
processed from 11 years (1992-2001) is described.
The methodology adopted relies on the possibility of
assigning a spatial meaning to the point-wise ground
displacement measurements provided by the PS
technique, through the interpretation aerial-photos and
optical satellite imagery, topographic maps and
ancillary data.
Besides with the intention of developing an effective
and semi-authomatic procedure for landslide mapping
from PS, a spatial statistical approach on the PS
analysis in the same area is described. The aim is to
Page 297 of 401
Deliverable 4.1
employ PS processed from 4 years (2003-2006) of

RADARSAT SAR images to identify slow moving
landslides within the basin. We consider this spatial
statistics approach as an ef-fective way for landslide
mapping, thus providing an innovative approach for
the rapid extraction of useful information from PS.
2. STUDY AREA
The Arno river basin is located in the central Italy
(Figure 1). The total area of the whole basin is about
9130 km2. The basin is across of Apennines chain. As
a result, 7190 km2 of the basin is situated in the
mountainous and hilly area. The basin is very
susceptible to landslides. More than 27,000 landslides
were mapped. The affected landslide area is more
than 800 km2. These landslides are dominated by
earth slides and flows (about 74%) as well as shal-low
landslides and creeps (Catani et al. 2005a, Fa-rina et
al. 2006). Most of these slide movements are slow and
intermittent, accompanied with accelerations due to
the prolonged and intensive rainfall. The concentration
of precipitation periods also accounts for the landslides
activity transition from dormant to active (Catani et al.
2005b). Considering the high density population within
the basin, the prevalence of landslides in the Arno
river basin poses an high risk. More than 16,000 civil
buildings, 460 industrial areas and 350 km roads are
affected by landslides. In addition, ca. 6 billions Euro
losses are expected in the upcoming 30 years (Catani
et al. 2005b).
Fig. 1. The location of the Arno river basin.
3. UPDATING OF LANDSLIDE INVENTORY

MAP
3.1 Methodology
Landslide inventories report the spatial disribution of
existing slope movements, frequently including details
about the landslides typologies and their state of
activity (Wieczorek, 1984). Today the most common
methodology for their compilation is based on aerialphoto interpretation, on field surveys and on the
collection of local databases (Crozier, 1984, Soeters
and Van westen 1996). The Arno River Basin Authority
produced an inventory map of active, dormant and

SafeLand - FP7
inactive landslides at a reference mapping scale of

1:10,000, dating back to 2003. The map follows the
landslide classification and the terminology proposed
by Cruden and Varnes (1996), with a few slight
changes due to the scope and to the extent of the
analyzed basin. Within the Arno river basin more than
27,000 landslides have been mapped. Based on this
inventory map, PS have been used for discovering
previously undetected landslides and to redifine the
pereimeter and state of activity of the existing ones.
By interferometrically processing more than 350 ERS1 and ERS-2 SAR images spanning 10 years (19922002), approximately 600,000 PS have been detected
within the whole basin. 44.8% of the derived PS are
located in the hilly and mountain areas with a density
of 37 PS/km2.
By combining PS information with optical images such
as SPOT5, color aerial photos, and a 10-meter
resolution digital elevation model (DEM) we performed
a geomorphologic analysis for landslide detection.
Attention was focused on the spatial distribution and
the state of ac-tivity (active, dormant or inactive,
according to Cruden and Varnes (1996)) of both
already mapped landslides and new unstable areas
(Fig. 1). In particular the main benefits obtained from
the application of the PS technique to landslides
regard:
- Better defined boundaries of already detected mass
movements and/or;
- Better defined states of activity and/or;
- The detection of previously unmapped unstable
areas.
The first effort was devoted to the detection of the
presence of PS within or close to mapped mass
movements. Once that the PS information have
confirmed the boundaries and state of activities of the
mapped landslides a new field with the avarage
velocities, computed on the two different time intervals
(1992 2002 and 19992002), has been added in the
attributes table. In case of differences in the spatial
distribution, e.g. PS in movement close to the landslide
boundaries, or in the state of activity, e.g. PS stable
within an active landslide, the support of multitemporal aerial-photos or SPOT5 image was
employed. The interpretation of these data was aimed
at detecting morphological or vegetation features
indicative of a variation of the landslide boundaries or
a change of its state of activity (Figure 2).
Furthermore, a strong effort was spent on the interpretation of moving PS located far from any mapped
landslide. To limit possible misinterpretations, attention
was focused only on groups of PS characterized by
displacement rates (over 34 mm/year) that are significant with respect to the technique precision. Sparse
radar benchmarks showing lower rates were
neglected.
The support of photo-interpretation and contour lines
analysis for detecting diagnostic morphologies induced
by slope instabilities was fundamental at this phase, as
were field surveys for particularly problematic areas.
Page 298 of 401
Case Study 9
Rev. No: 2
Date: 2012-02-15
4. HOT SPOT ANALYSIS

4.1 Methology
Fig. 2. Example of landslide inventory map modification with

the PS-based methodology: (a) the pink polygon represents
the originally mapped landslide while the red outline indicates
the updated boundary from the PS method; (b) the aerialphoto (1996) of the mapped landslide; (c) cracks on a
building located within the unstable area, where the PS
record velocities up to 5 mm/year.
Table 1. Statistical data about the results of PS-methodology

applied on the Arno basin landslide inventory map (AdBA is
the acronym of Autorit di Bacino del F. Arno).
Number of landslides
mapped by AdBA
Number of landslides
mapped by AdBA with PS
information
%
for
number
of
landslides mapped by
AdBA with PS information
Area of AdBA landslides
(km2)
Area
of
landslides
information (km2)
% of area of landslides
information
Landslides
density
(km2/km2 as a percentage
of whole river basin)
27,270
1664
6.1%
The effective extraction of useful information from PS

technique for landslide studies is sometimes difficult
due to the large number of PS that can be present,
thereby entailing long interpretation times. In the
places where there are a lot of stable reflectors such
as buildings and bared rocks, the PS density is higher
than 500 PS/km2. Large number of PS calls for an
efficient approach of data processing. Still, with the
quick development of new SAR sensors and efforts of
increasing PS density, the PS data is expected to be
updated more frequently with higher density. Thus, an
effective approach of data interpretation is needed.
With the intention of developing an effective procedure
for landslide mapping from PS, a spatial statistical
approach on the PS analysis is introduced. The aim is
to employ PS processed from 4 years (2003-2006) of
RADARSAT SAR images to identify slow moving
landslides within the basin. We consider this spatial
statistics approach as an ef-fective way for landslide
mapping, thus providing an innovative approach for
the rapid extraction of useful information from PS.
The purpose of the hotspot analysis is to identify
concentrations of high velocity PS. The analysis is
based on the two statistics approaches: Getis-Ord Gi*
statistics and kernel density estimation.
The Getis-Ord Gi* statistics is a kind of local spatial
statistics which represents the association up to a
specified distance. In the study we apply Gi* statistics
to evaluate the spatial clustering of neighboring PS.
The Gi* index is defined as follows (Getis & Ord 1996):
151.3
18.9%
8.8%
Gi* (d )
In order to highlight the percentage of information

coming from the proposed methodology and to also
evaluate its effectiveness, a simple statistical analysis
of the final landslide inventory map was performed
(Table 1).
223 new landslides have been mapped while 1660
existing landslides have been modified regarding the
state of activity or the boundary throughout the whole
basin for the new landslide inventory. The newly
mapped landslides through PS information represent
6.8% of the total landslides in the final inventory.
* *
j wij ( d ) x j Wi x
s* (nS1 j * ) Wi*2 / (n 1)
0.5
While in the Equation 1,
Wi* j wij (d ) wij , S1 j * j wij 2 and
3.2 Results

SafeLand - FP7
802.9
s*2
x j2
x*
wij(d) is the spatial weights vector. It is defined within

the searching distance d from the PS whose velocity is
x. Each single PS is at a site i with its neighbours j
within the distance d. In the study, all the PS are
treated as the same weight, namely 1.
In order to define the searching distance of d, for each
pixel (10m) of the DTM of the Arno river basin, both
the shortest path to a channel (d1) and the ridge (d2)
are calculated based on steepest descent direction.
Page 299 of 401
Deliverable 4.1
The searching distance of each pixel (di) is calculated

as the mean value of d1 and d2, simulated as a
measure of landslide length along the slope. The
estimation of the searching distance d is based on the
mean value of all DTM pixels. In the case of the Arno
River basin, the searching distance d is 114m
compared with a DTM of 10m resolution.
We apply the Gi* statistics in our study and perform
the local test on PS datasets. We choose the velocity
of PS as the weighting factor and calculate Gi* index
for each single PS. The Gi* index measures
concentrations of high velocity PS for the entire study
area. The larger (positive) the Gi* index is, the more
intense the clustering of high velocity values, with the
PS moving direction towards Line-of-Sight (LOS) of
the satellite. The smaller (negative) the value is, the
more intense the clustering of low velocity values
(negative), with the PS moving direction away from
LOS.
The following procedure is to estimate PS density
using kernel calculation. The kernel density estima-tion
(Silverman 1986) uses a kernel estimator which is
defined as:
f ( x)
x Xi
1 n
)
K(
nh i 1
h
where h is the window width, x-Xi is the distance to

each PS i. K is the quadratic kernel function defined
as:
3
(1 x 2 ),| x | 1
4
K ( x) 0, x 1
K ( x)
We employ this kernel function for PS density

estimation, choosing the previous derived Gi* index as
the weight. The output is a smooth density map adding the values of kernel function. It indicates the existing hotspot of high velocity mass movement.
4.2 Results
Figure 3 shows one part of the hotspot map derived
from the same SAR image frame. The area is within
the Arno river basin covering the Pistoia-Prato-Firenze
and the Mugello basin. Both the ascending (Fig 3a)
and descending (Fig 3b) hotspot map are displayed
based the result of kernel density estimation.
Clustering of positive velocity PS (moving to-wards
LOS) is rendered with blue hotspot while clustering of
negative velocity PS (moving away from LOS) is
rendered with red color. The deeper color it displays,
the more intense clustering of higher velocity PS. The
radius of the hotspot implies the dimension of the
potential landslide-affected area. Figure 4 is the
hotspot map with the overlay of both ascending and
descending hotspot map. The magenta areas are the
combination of red and blue hotspots. They indicate
the different moving directions of PS from different
ascending and descending orbits. Still, the deep blue
and deep red hotspot indicate the same moving
directions detected from the ascending and
descending PS.

SafeLand - FP7
Fig. 1. Part of the hotspot map of the Arno river basin,

including the Pistoia-Prato-Firenze and Mugello basin area:
a) hotspot map derived from kernel density estimation using
ascending RADARSAT PS; b) hotspot map derived from
kernel density estimation using descending RADARSAT PS.
Red Hotspot (low negative kernel density) indicates the
clustering of high velocity PS moving away from satellite
along LOS whereas blue hotspot (high positive kernel
density) implies the clustering of high velocity PS moving
towards satellite along LOS.
The clustering of high velocity PS is able to detect

landslide considering LOS of satellite and down-slope
landslide movement can especially be detected from
PS moving away from sensor. According to the
previous studies and existing landslide inventory map,
some of the detected hotspots are confirmed to be the
mass movements resulting from slow moving
landslides. Hence, the hotspot map is expected to be
an important source for the following study of landslide inventory updating. However, detection errors
exist when such mass movement is related to other
geophysical processes that PS can identify. The major
errors are attributed to uplift and ground subsidence.
Moreover, the hotspot sometimes shows the
limitations in separating landslide movement from
other geo-processes. For example, if an area is subject to both landslide and subsidence, the hotspot
analysis possibly only yields the result of one big
hotspot. This hotspot however fails to separate these
two kinds of mass movement. Such problems are
mainly resulted from the current limitations of PS
technology for detailed interpretation of mass
movement.
Page 300 of 401
Case Study 9
Fig. 4. The hotspot map overlaid with ascending and

descending data. The magenta hotspot indicates the
clustering of high velocity PS detected by both ascending and
descending PS, with one moving direction away from LOS
and another moving towards LOS. The deep red and blue
hotspots imply the clustering of high velocity PS detected by
both ascending and descending data, with the same moving
direction along LOS.
5. DISCUSSION AND CONCLUSION

In this contribution the usefulnees of the PS technique
for landslide detection and mapping at regional scale
is discussed.
The application of this method at the regional scale, as
a support for the production of a landslide inventory
map, was performed on the Arno river basin, in Central
Italy.
The main benefit of the methodology regards the
capability
to
perform
ground
displacement
measurements over urbanized areas where risk
conditions are usually higher. The positive feedback
implies that within the next years, due to the
imminenet launch of the space-borne SAR missions
with a acquisition parameters better tuned for
landslides investigations, the methodology has the
potential to become a fully operational tool for
landslide mapping.
On the other hand the effective extraction of useful
information from PS technique for landslide studies is
sometimes difficult due to the large number of PS that
can be present. Large number of PS calls for an
efficient approach of data processing.
At this regard the hotspot analysis employs the spatial
statistics analysis of PS in order to emphasize the
landslide-affected area. The output map is the hotspot
visualization emphasizing the clustering of high
velocity PS. The approach is considered to be an
example of effectively extracting useful information
from large number of PS. The derived hotspot is
considered to be useful for the further landslide
inventory mapping and risk management.
REFERENCES
Bamler, R. & Hartl, P. 1998. Synthetic aperture radar
interferometry. Inverse Problems 14: 1-54.

SafeLand - FP7
Rev. No: 2
Date: 2012-02-15
Canuti P., Casagli N., Catani F., Falorni G., Farina P.

2007. Integration of Remote Sensing Techniques
in different stages of landslide response. In
Progress in Landslide Sciences Eds. Kyoji Sassa,
Hiroshi Fukuoka, Fawu Wang, Gonghui Wang
Casagli, N., Guerri, L., Righini, G., Ferretti, A.,
Colombo, D. & Prati, C. 2005. Integrated use of PS
and very high resolu-tion optical images for
supporting landslide risk manage-ment. URSI,
Symposium on Microwave Remote Sensing of the
Earth, Oceans, Ice and Atmosphere, 20-21 April
2005, Ispra, Italy.
Catani, F., Farina, P., Moretti, S., Giovanni, N. &
Strozzi, T. 2005a. On the application of SAR
interferometry to geo-morphological studies:
estimation of landform attributes and mass
movements. Geomorphology 66: 119-131.
Catani, F., Casagli, N., Ermini, L., Righini, G. &
Menduci, G. 2005b. Landslide hazard and risk
mapping at catchment scale in the Arno River
basin. Landslides 2: 329-342.
Colesanti, C., Ferretti, A., Prati, C. & Rocca, F. 2003.
Monitor-ing landslides and tectonic motions with
the Permanent Scatterers Technique. Engineering
Geology 68: 3-14.
Corsini, A., Farina, P., Antonello, G., Barbieri, M.,
Casagli, N., Coren, F., Guerri, L., Ronchetti, F.,
Sterzai, P. & Tarchi, D. 2006. Space-borne and
ground-based SAR interferometry as tools for
landslide hazard management in civil protec-tion.
International Journal of Remote Sensing 27: 23512369.
Crozier, M.J., 1984. Field assessment of slope
instability. In: Brundsen, D., Prior, D. (Eds.), Slope
Instability. John Wiley and Sons, Chichester, pp.
103140.
Cruden, D. M. & Varnes, D. J. 1996. Landslides types
and processes. In: Landslides Investigation and
Mitigation: Transportation Research Board, A.K.
Turner and R.L. Schuster (eds.), US National
Research
Council,
Special
Re-port
247,
Washington DC, p. 36-75.
Farina, P., Colombo, D., Fumagalli, A., Marks, F. &
Moretti, S. 2006. Permanent Scatterers for
landslide investigations: outcomes from the ESASLAM project. Engineering Geol-ogy 88: 200-217.
Ferretti, A., Prati, C. & Rocca, F. 2000. Non-linear
subsidence rate estimation using Permanent
Scatterers in Differential SAR interferometry. IEEE
Transactions on Geoscience and Remote Sensing
38: 2202-2212.
Ferretti, A., Prati, C. & Rocca, F. 2001. Permanent
Scatterers
in
SAR
interferometry.
IEEE
39: 8-20.
Fruneau, B., Achache, J. & Delacourt, C. 1996.
Observation and modeling of the Saint-Etienne-deTinee landslide using SAR interferometry.
Tectonophysics 265(3): 181-190.
Getis, A. & Ord J.K. 1996. Local spatial statistics: an
over-view. In John Wiley and Sons (eds), Spatial
Analysis: Mod-eling in a GIS Environment : 261.
Page 301 of 401
Deliverable 4.1
Kimura, H. & Yamaguchi, Y. 2000. Detection of

landslide
ar-eas
using
satellite
radar
interferometry: Photogrammetric Engineering and
Remote Sensing 66(3): 337-344.
Lu P., Casagli N., Catani F., Tofani V. (2009) Hotspot
analysis of Permanent Scatterers (PS) for slow
moving landslides detection. In: Landslide
Processes: From Geomorphological Mapping To
dynamic Modelling. Proceedings of the Landslide
Processes Conference, A tribute to Dr. Theo van
Asch, Strasbourg, 6-7 February 2009.
Mora, O,, Mallorqu, JJ, Broquetas, A. (2003). Linear
and nonlinear terrain deformation maps from a
reduced set of inter-ferometric SAR images. IEEE
Transactions on Geoscience and Remote
Sensing,41(10), 2243-2253.
Massonnet, D. & Feigl, K. L. 1998. Radar
interferometry and its application to changes in the
earths surface. Reviews of Geophysics 36(4): 441500.

SafeLand - FP7
Soeters, R. & Van Western C. J. 1996. Slope

instability recog-nition, analysis and zonation.
Turner AK, Schuster RL (eds) Landslides:
investigation and mitigation. Sp. Rep.247,
Transportation Research Board, National research
Council, National Academy Press, Washington,
DC, pp129-177.
Werner, C.L., Wegmuller, U., Wiesmann, A., Strozzi,
T., (2003). Interferometric Point Target Analysys
for Deformation Mapping. IGARSS Proceedings
2003, Toulouse.
Wieczorek, G (1983). Preparing a detailed landslideinventory map for hazard evaluation and reduction.
Bulletin of the Association of Engineering
Geologists, 21(3), 337 342.
Page 302 of 401
Case Study 10
Rev. No: 2
Date: 2012-02-15
CASE STUDY 10:

Title: Integration of PS-InSAR data and ground-based instrument measurements for
landslide monitoring at the local scale.
Application: Landslide characterization and monitoring at local scale.
Technique: PSI (Persistent Scatterers Interferometry).
Main references: This case study is based on the following publications: Casagli et al.,
2009; Farina et al., 2006.
Contributor: UNIFI (Tofani V., Catani F., Casagli N.)
ABSTRACT
Interferometric SAR, whether satellite- or ground-based (InSAR and DInSAR) are the techniques most researched
during the last decade for slope motion monitoring and characterization. In particular the multimage Persistent
Scatterers SAR Interferometry (PSI) technique (Ferretti et al. 2001, Colesanti et al. 2003, Mora et al. 2003,
Werner et al. 2003, Duro et al. 2005), has showed its capability to provide information about ground deformations
over wide area with millimetric precision, making this technique suitable for both regional and slope scale mass
movements investigations. The PSI analysis has been applied for the monitoring of single well-known slope
movement, the Carbonile landslide, located in the Arno river basin, Northern Apennines. In this case study the use
of InSAR for the monitoring of single slow landslides threatening built-up areas has provided satisfactory results,
allowing the measurement of superficial deformations with high accuracy on the landslide sectors characterized by
a good radar reflectivity and coherence. Besides the PS-InSAR technique has been integrated with classic
ground-based monitoring techniques in order to facilitate the interpretation of landslide behaviour and kinematics.
1. INTRODUCTION
Monitoring means the comparison of landslide
conditions like areal extent, speed of movement,
surface topographyand soil humidity from different
periods in order to assess landslide activity (Mantovani
et al., 1996).
The measurement of superficial displacements
induced by a slope movement often represents the
most effective method for defining its behavior,
allowing the observation of response to triggering
factors and the assessment of effectiveness of
corrective measures (Farina et al., 2006).
Different techniques are available for measurements of
the ground displacements, starting from the traditional
inclinometers, extensometers, topographic surveys,
until more recent applications such as GPS, aerial
photogrammetry, LIDAR measurements (Angeli et al.,
2000; Gili et al., 2000; Kaab, 2000; Hervas et al.,
2003; McKean and Roering, 2004).
Interferometric SAR, whether satellite or ground-based
(InSAR and DInSAR) are the techniques most
researched during the last decade for slope motion
monitoring (Metternicht et al., 2005).
In particular the multi-image Persistent Scatterers SAR
Interferometry (PSI) technique (Ferretti et al. 2001,
Colesanti et al. 2003, Mora et al. 2003, Werner et al.
2003), has showed its capability to provide information
about ground deformations over wide area with
millimetric precision, making this technique suitable for
both regional and slope scale mass movements
investigations.
In
particular
the

SafeLand - FP7
Permanent Scatterers (PS-InSAR) technique

patented by the Politecnico di Milano and is
commercially available through the POLIMI spin-off
company Tele-Rilevamento Europa (TRE). PS
technique produces radar benchmarks derived from a
multi-interferogram analysis of SAR images. The
temporal decorrelation and atmospheric artifacts can
be meanwhile estimated and removed (Ferretti et al.
2001).
The InSAR technique, by providing an accurate
measurement of ground displacements without the
necessity of positioning any targets on the ground and
without any physical contact with the slope, is best
suited for assessing the temporal evolution of slow
landslides (up to a few centimeters per year).
Unfortunately, incidence angle, spatial resolution,
wavelength, and revisiting time interval of the
operational sensors are not optimal compared to the
particular spatial and temporal pattern of all the types
of movement we are dealing with.
In this contribution an example of monitoring of the
Carbonile slow moving landslide by means of PS
technique is reported. This example provides the
combined use of underground and superficial data for
a better understanding of the landslide geometry and
its deformation pattern.
2. CARBONILE LANDSLIDE
Carbonile, a small village located in Northern
Apennines, Tuscany and it has been affected by
Page 303 of 401
Deliverable 4.1
different slope instability problems since 1984. This

site has been selected as case study in the progect
ESA-SLAM (Farina et al., 2006). These movements
have seriously endangered the village and its 200
residents. Damage to cultivated areas, buildings and
infrastructures have been recorded since 1984. For
these reasons the whole zone was mapped as
exposed to the highest landslide hazard level within
the Hydrogeological Setting Plan of the River Basin
Authority.
The landslides correspond to complex movements
dominated by earth-slides with translational and rotational components occurring on multiple shear
surfaces
and with non-uniform distribution of
velocities.
The
inclinometric
readings
show
displacement rates ranging from slow to very slow
(IUGS/WGL, 1995).
The advanced PS processing performed on the
ERS1-ERS2 descending dataset has allowed the
detection of 310 radar benchmarks within the study
site (2.7 km2) resulting in a PS spatial density of 114
PS/ km2 (Figure 1).
The PS velocities indicate that the central part of the
ancient landslide is stable, while three zones, two on
the sides of Carbonile and one in the upper part of the
slope, are characterized by significant movements,
with deformation rates up to 12 mm/year.
Fig.1. Aerial-photo of the Carbonile landslide with the location

of the Permanent Scatterers. The colors of the PS are
related to the average annual velocity along the satellite line
of sight, estimated between 1992 and 2002.
In order to assist the geological interpretation of the

main landslides present in the Carbonile area, an
integration of PS data with inclinometric readings and
borehole samplings was attempted. The analysis was
performed along different profiles. The main rupture
surfaces, placed at depths ranging from 3 to 13 m,
were interpreted from the inclinometric readings, while
boreholes provided the stratigraphic information.
Profile BB (Figure 2) represents an example of the
combined use of underground and superficial data for
a better understanding of the slope instability
geometry. Down-slope movements with velocities of
up to 15 mm/ year were recorded by the inclinometers
between 1987 and 1992 (tubes 4 and 22) along a main

SafeLand - FP7
failure surface. The PS located in the lower part of the

slope also indicate ground displacements compatible
with the movement measured by the inclinometer, in
terms of both line of sight velocity (about 710
mm/year) and direction of movement. The PS have
negative velocities, meaning that they are moving
away from the satellite sensor, which is compatible
with the SW down-slope direction of movement
recorded by the inclinometric tubes.
Fig.2. Cross-sections of the Carbonile area showing both

subsurface information from inclinometric readings and
borehole samplings and superficial movement from the PS
analysis. Dashed lines indicate the main sliding surfaces, as
inferred from the inclinometric measurements. Continuous
arrows represent the PS displacement vectors as measured
along the satellite line of sight (L.O.S.). Arrows length is
proportional to the average PS velocity (computed over the
period 19922002) and the vector versus indicates the
velocity sign. The upper image shows profile BB and while
the bottom profile 55.
A quantitative comparison of PS and inclinometric

measurements was performed using inclinometric tube
27 (by considering the average deformations along the
tube), located along profile 5 (Figure 2), and the three
closest PS. For the comparison all the data were
projected along the EW movement direction recorded
by the inclinometer, with a 10 gradient equal to local
slope of the terrain. Even though the overlap in the
time interval covered by the two datasets was not
complete (the PS analysis ranges from 1992 up to
2002, while inclinometric readings were acquired from
1990 to 1996), the radar measurements are of the
same order of magnitude as the inclinometric
readings. The projected average velocity (obtained
Page 304 of 401
Case Study 10
from the linear fitting of the displacement

measurements) recorded by the inclinometric tube
between 1992 and 1996 is 3.4 mm/year, while
velocities measured by the PS range from 7.0 to 13.7
mm/year, with an average value of 9.5 mm/year. The
difference between the two sets of measurements was
ascribed to the different types of movement measured
by the two techniques (superficial vs. deep
deformations) and to the low sensitivity of PS
measurements to horizontal movements.
3. CONCLUSION
The Carbonile case study illustrates how it is possible
to use Permanent Scatterers for the monitoring of
single landslides inducing high risk scenarios.
The advanced processing of ERS1-ERS2 descending
dataset spanning a time interval from 1992 to 2002
has allowed an accurate analysis of temporal and
spatial displacement fields as well as the identifaction
of the most hazardaous areas.
In order to assist the geological interpretation of the
landslide both concerning the landslide geometry and
deformation pattern, a comparison of the PS
measurements and inclinometers measurements was
carried out. The quantitative comparison has
demostrated that the measurements acquired by the
two instrumentions are of the same order of magnitude
thus proving that SAR intereferometry is a powerful
tool for landslide monitoring.
Rev. No: 2
Date: 2012-02-15
Kaab A. (2002). Monitoring high-mountain terrain

deformation from repeated air and spaceborne
optical data: examples using digital aerial imagery
and
ASTER
data
ISPRS,
Journal
of
Photogrammetry and Remote Sensing, 57(1-2),
39-52
Mantovani F., Soeters R., Van Westen C. J. (1996)
Remote sensing techniques for landslide studies
and hazard zonation in Europe, Geomorphology,
15( 3-4): 213-225.
McKean J., Roering J. (2004). Objective landslide
detection and surface morphology mapping using
high-resolution
airborne
laser
altimetry.
Geomorphology, 57, 331 351.
Metternicht, G., Hurni L., Gogu R. (2005) Remote
sensing of landslides: An analysis of the potential
contribution to geo-spatial systems for hazard
assessment
in
mountainous
environments.
Remote Sensing of Environment, Vol. 98, no23, pp. 284-303.
REFERENCES
Angeli M., Pasuto A., Silvano S. (2000), A critical
review of landslide monitoring experiences.
Engineering Geology 55, 133-147.
Casagli N., Tofani V., Adler R. (2009), A look from
space. In Landslides, Disaster Risk Reduction,
Sassa K. and Canuti P. eds.
Colesanti, C., Ferretti, A., Prati, C. & Rocca, F. (2003).
Monitoring landslides and tectonic motions with the
Permanent Scatterers Technique. Engineering
Geology 68: 3-14.
Farina, P., Colombo, D., Fumagalli, A., Marks, F. &
Moretti, S. (2006). Permanent Scatterers for
landslide investigations: outcomes from the ESASLAM project. Engineering Geology 88: 200-217.
Ferretti, A., Prati, C. & Rocca, F. (2001). Permanent
Scatterers
in
SAR
interferometry.
IEEE
39: 8-20.
Gili J.A., Corominas J., Rius J. (2000) Using Global
Positioning System techniques in landslide
monitoring. Engineering Geology 55, 95-113.
Hervas J., Barredo J., Rosin P., Pasuto A., Mantovani
F., Silvano S. (2003). Monitoring landslides from
optical remotely sensed imagery: The case story of
Tessina landslide, Italy. Geomorphology, 54, 63
75.

SafeLand - FP7
Page 305 of 401
Case Study 11
Rev. No: 2
Date: 2012-02-15
CASE STUDY 11:

Title: A new approach to the use of DInSAR data in landslide studies at different scales: the
case study of National Basin Authority of Liri-Garigliano and Volturno rivers.
Application: Slow-moving landslide detection; Mapping; Characterization; Monitoring.
Technique: DInSAR.
Main references: This case study is based on the following publications: Cascini et al.,
2010; Cascini et al., 2009; Cascini et al., 2008.
Contributors: UNISA (Cascini L., Peduto D., Fornaro G.)
ABSTRACT
Within the general framework of the landslide risk analysis, in the last decades significant efforts have been made
by the scientific community towards the development of topics dealing with the hazard analysis (including the
landslide characterization and the analysis of frequency) and the consequence analysis.
With reference to slow-moving landslide characterization, the contribution deriving from remote sensing
techniques can be helpful in detecting large areas and in analyzing both the state of activity and the kinematical
characteristics of the instability phenomena. In particular, the use of remote sensing techniques such as
Differential SAR Interferometry (DInSAR) has been already dealt with in the scientific literature via a number of
case studies. However, standardized procedures for the proper interpretation and the confident use of DInSAR
data have not been fully investigated and validated, although algorithms for image processing have become more
and more sophisticated. Moreover, the diffusion of these data when they are not validated can dangerously bring
to misleading interpretations and unjustified warnings.
Starting from current limits of the applicability to landslide, this contribution shows the potential of innovative
procedures based on the joint use of DInSAR data, at both full- and low- resolution, with the help of considerations
on the acquisition geometry of sensors, simplified landslide kinematical models and landslide-induced damage to
structures/infrastructures.
1. INTRODUCTION
This study has been carried out jointly by the group of
Geotechnical Engineering of Salerno University (Italy)
and IREA-CNR (Italy) as a development of the
expertise achieved in the exploitation of DInSAR data
for the analysis of subsidence phenomena at different
scales (Cascini et al., 2006; 2007). The passage from
subsidence to landslides is not straightforward due to
the complexity of instability phenomena and current
limitations in both the sensor acquisition geometry and
the availability/interpretation of DInSAR data on slopes
(Cascini et al., 2010). To this end, the present analysis
introduces a new methodology for DInSAR data
interpretation in areas for which a proper
geomorphological and topographic knowledge is
available.
The first step is the generation of the a priori DInSAR
landslide visibility map (described in details in Cascini
et al., 2009). Then, DInSAR data interpretation is
based on the joint use of remote sensed data and
simplified geomorphological models. The reported
procedure is tested at both medium (i.e. 1:25,000
scale according to Fell et al., 2008) and large scales
(i.e. 1:5,000 scale, Fell et al., 2008) within a sample
area extending for around 500 km2 in the territory of
National Basin Authority of Liri-Garigliano and Volturno
(NBA-LGV) rivers (Central-Southern Italy) (Fig.1).

SafeLand - FP7
2. THE TEST AREA AND THE USED

DATASET
Thirty-three images (track 308 - frame 2765), acquired
over descending orbits of the ERS1-ERS2 satellite
systems (period March 1995 - February 2000), have
been processed via the Enhanced Spatial Differences
(ESD) approach (Fornaro et al., 2009a,b), which
represents an upgrading of the original SBAS
algorithm (Berardino et al., 2002). These algorithms
allow the generation of both low-resolution DInSAR
maps (with pixel on the ground of approximately 80 x
80 m) and full-resolution maps (with pixel on the
ground of approximately 10 x 10 m).
For the test area both base and thematic maps are
available at 1:25,000 scale. The base maps were
produced in 2001 as results of the activities of the
PSAI
project
(Piano
Stralcio
per
lAssetto
Idrogeologico), carried out by a group of experts and
technicians working for NBA-LGV in accordance with
the Act of Italian Parliament (L. 365/2000), aimed to
zone both landslide hazard and risk all over the Italian
territory (Cascini, 2008). The geological map highlights
that the bedrock mainly consists of Upper Miocene
arenaceous units mantled by Quaternary Age
superficial deposits, characterized by talus and alluvial
fans. Landslide phenomena cover around 5% of the
whole territory as it can be noticed in the available
Page 307 of 401
Deliverable 4.1
landslide inventory map, derived from aerial

photographs and surface surveys. This map furnishes
detailed information for each mapped phenomena with
reference to location, typology, state of activity and
areal extension (Cascini et al., 2005).
used to distinguish in advance whether an area is

expected to be visible from space-borne SAR sensors
thus driving data-users through the image dataset
selection. Once SAR images have been processed, if
an adequate knowledge of landslide phenomena is
available, a procedure for 1D-LOS DInSAR data
projection can be implemented to generate the
advanced DInSAR landslide velocity map.
As for the scale of the study, low-resolution DInSAR
data can be used for landslide analyses at 1:25,000
scale, according to the dimension of both the landslide
phenomena and the coherent DInSAR pixels on the
ground; whereas full-resolution DInSAR data allow
studies at more detailed scale (i.e.1:5,000) according
to the almost point-wise information and the dimension
of single portions of landslides and structures/
infrastructures.
Fig. 1. The test area.
Owing to the phase ambiguity limitation of DInSAR

data processing (for more details refer to Cascini et al.,
2010), the analysis of landslides focuses on the
typology of phenomena ranging from extremely to very
slow velocity classes (i.e. lower than 1,6m/year
according to Cruden and Varnes, 1996). In the study
area a total number of 897 slow-moving landslides are
mapped (Peduto, 2008; Cascini et al.,2009); according
to Varnes (1978) they are classified as: 204 rotational
slides, 238 earth flows, 78 rotational slides-earth flows,
336 creeps, 33 earth flows - creeps, 8 deep-seated
gravitational
movements.
On
the
basis
of
geomorphological criteria, three different states of
activity are distinguished for these landslides, defined
as follows: active (including active, reactivated and
suspended), dormant and inactive (relict)
phenomena (Cruden and Varnes, 1996). The selected
landslide typologies exhibit a significant predominance
of dormant phenomena (428) on active ones (92). This
is confirmed by the available dataset on damage to
buildings and roads interacting with the displaced
masses that includes: 557 phenomena (around 62% of
the total) interacting with properties; 183 damage
surveys and only 40 damaged facilities inventoried in
the study area (Cascini et al., 2008), corresponding to
30% of surveyed buildings and 24% of investigated
roads.
3. THE PROCEDURE PROPOSED FOR

DINSAR APPLICABILITY TO LANDSLIDES
The framework sketched in Fig. 2 describes the
procedure developed for DInSAR data analysis at
different scales. The first step consists of the
generation of the a priori DInSAR landslide visibility
map (Peduto, 2008; Cascini et al., 2009), which can be

SafeLand - FP7
Fig. 2. The framework of DInSAR data analyses for landslide

studies (Cascini et al., 2010).
3.1 A priori visibility map

The choice of the most suitable SAR image dataset
represents a key step for DInSAR data exploitation in
landslide studies, since the visibility of a certain portion
of a slope depends on several factors such as slope
aspect and inclination, vegetation cover, presence of
buildings/infrastructures.
The role played by the aspect angle and the slope
inclination, which have direct impact on the feasibility
of DInSAR deformation monitoring, has been already
discussed in Colesanti & Wasowski (2006); Peduto
(2008); Cascini et al. (2009).
Following the procedure described in details in Peduto
(2008) and Cascini et al. (2009) here an example of
the a-priori DInSAR landslide visibility map over a
portion of the study area, obtained via low-resolution
ESD DInSAR data, is presented at 1:25,000 scale
(Fig.3). The input data consist of the following
available maps: landslide inventory map; aspect map;
slope angle map; land-use map; urbanised area map.
Particularly, by intersecting the aspect map and the
slope angle map visible/ visible with difficulty/ not
visible areas are zoned.
Page 308 of 401
Case Study 11
By merging the land-use map and the urbanised area

map, vegetated areas are removed from those
portions classified as visible according to geometric
considerations, thus obtaining the so called a priori
DInSAR landslide visibility map. The validation of the
map for the whole test area (partially shown in Fig. 3)
proves that 67% out of a total of 215 low-resolution
DInSAR coherent pixels intersecting landslide affected
areas concentrate on visible areas; 19% lay on areas
visible with difficulty; only 14% can be found in areas
assumed as not visible. The described procedure can
be easily applied at 1:5,000 scale if base and derived
maps are available at this scale.
Rev. No: 2
Date: 2012-02-15
geomorphological scheme of the landslide (see

Cascini et al., 2010) on which each DInSAR coherent
pixel is located. Finally, the projected velocity value is
assumed as a reliable value if the condition number
(see the Appendix in Cascini et al., 2010) does not
exceed the fixed threshold.
Fig. 4. Flow-chart for the generation of the advanced DInSAR

landslide velocity map (Cascini et al., 2010).
Fig. 3. The ``a priori DInSAR landslide visibility map'': (a) on

descending orbits with low-resolution DInSAR coherent pixel
distribution; (b) on ascending orbits (Cascini et al., 2009).
3.2 1D LOS DInSAR data projection and the

generation of the advanced DInSAR landslide velocity
map
The procedure for the projection of 1D LOS DInSAR
velocity vectors, based on both geomorphological and
DEM models, is described in Peduto 2008 and Cascini
et al., 2010. The flow-chart for the generation of the
advanced DInSAR landslide velocity map is shown in
Fig. 4.
On this map DInSAR coherent pixels are represented
depending on several factors to be taken into account.
Particularly, conditions of movement/no-movement
depend on whether the mean velocity of the DInSAR
coherent pixel exceeds the fixed velocity threshold
(Cascini et al., 2010). Moreover, the direction of
movement is assumed consistent with the

SafeLand - FP7
4. LOW-RESOLUTION LANDSLIDE
ANALYSIS
The advanced low-resolution DInSAR landslide
velocity map was developed at 1:25,000 scale for all
the test area in ADB-LGV according to the schemes in
Fig. 4 so as to analyse rotational slides, earth flows
and rotational slides-earth flows, whose total amount is
553; 185 (around 33%) of those resulted covered by
DInSAR data. Moreover, the map highlighted that
almost 84% of the DInSAR covered dormant
landslides (144) exhibit evidence of no-movement. On
the other hand, the percentage of active landslides
(25) with moving coherent DInSAR pixels is about
24%, on the average (Cascini et al., 2008). An
example is reported with reference to the
municipalities of Frosinone and Torrice (Lazio Region)
in Fig. 5 where very few moving low-resolution
DInSAR pixel are detected over an area of about 3 km
x 6 km where dormant phenomena prevail.
Within this area, the directions of movement (see the
four red arrows), derived by modelling the available 1D
LOS DInSAR data, seem congruent with the
Page 309 of 401
Deliverable 4.1
assumption of the geomorphological schemes

(downward direction along slope).
Moreover, over the test area the highest projected
mean velocities are attained in the main body,
independently of the landslide typology (Fig. 6).
5. FULL-RESOLUTION LANDSLIDE
ANALYSIS
Analyses of landslide phenomena at more detailed
scale (i.e. 1:5,000) can exploit full-resolution DInSAR
data following the flow-chart in Fig. 4. However, since
these analyses call for significant computational efforts
they should be concentrated on limited areas. Taking
this in mind two main goals were pursued: the
preliminary analysis of landslide features (i.e. check of
mapped boundaries; detection of ground displacement
out of mapped areas); an insight into different
kinematic behaviour characterizing different portions of
the same phenomenon.
5.1 Analysis of landslide features
Fig. 5. An example of advanced low-resolution DInSAR

landslide velocity map for the municipality of Frosinone and
Torrice (Lazio Region, Italy). 1) Not moving DInSAR coherent
pixel or on flat areas; 2) DInSAR coherent pixel moving on
vertical direction; 3) dormant rotational slide; 4) active
rotational slide; 5) dormant earth flow; 6) active earth flow; 7)
dormant rotational slide earth flow; 8) active rotational slide
earth flow; 9) creep phenomenon (Cascini et al., 2010).
Fig. 6. DInSAR mean velocities recorded within covered

landslides: 53 rotational slides; 93 earth flows; 39 rotational
slides earth flows (Cascini et al., 2010).
The advanced low-resolution map were then used to

detect new landslide phenomena within the test area in
ADB-LGV by extending the analysis of moving/not
moving coherent pixels on those portions of the
territory mapped as hollows in the geomorphological
map at 1:25,000 scale (Cascini et al., 2009). These
zones (1263 in the investigated area) are
characterized by geomorphological settings quite
similar to landslide-affected areas, also exhibiting the
same landslide predisposing factors. Indeed, as for 63
hollows a clear evidence of movement was recorded;
this can provide elements for a check/update of the
landslide inventory map that represents the starting
point for the landslide risk analysis as described in Fell
et al.(2008).

SafeLand - FP7
In order to check possible changes in landslide

boundaries, the entire full-resolution coherent pixel
dataset was projected assuming translational
movements along the steepest slope direction for the
pixels out of the mapped landslides. Several examples
are reported in Cascini et al. (2010); here, a portion of
the municipality of Frosinone (Fig. 7) highlighting
evidences of movements both inside and outside the
landslide boundaries is shown. Particularly, special
attention is worth being paid to full-resolution DInSAR
coherent pixels exhibiting mean velocity values
exceeding 0.3 cm/year within two dormant earth flows
(labelled with letters A and D) and creep zones
(labelled with letters B and C) as well as the areas
framed with the circle and the square, respectively. As
for the zone framed with the circle, a cross check via
the geomorphological map shows that the buildings in
the area were built on an hollow on which the DInSAR
data show evidences of movement for the period
1995-2000. The area framed by a rectangle highlights
the presence of moving full-resolution DInSAR
coherent pixels in proximity of a landslide classified as
active earth flow in the landslide inventory map at
1:25,000 scale.
Fig. 7. Overview of Advanced full-resolution DInSAR

landslide velocity map for a portion of the municipality of
Frosinone (Lazio Region, Italy). 1) Not moving DInSAR
coherent pixel or on flat areas; 2) not projected translational
displacement owing to high condition number; 3) dormant
rotational slide; 4) active rotational slide; 5) dormant earth
flow; 6) active earth flow; 7) creep phenomenon (Cascini et
al., 2010).
Page 310 of 401
Case Study 11
In this case further analyses carried out via a map at

1:5,000 scale of the landslide affected areas (see
Cascini et al., 2010) allowed to point out that the
moving coherent pixel falls within a small active
rotational slide earth flow; on the contrary, the stable
coherent pixels are located within the head of the
dormant rotational slide earth flow. This highlights
that full-resolution DInSAR data reliability must be
necessarily checked via thematic and topographic
maps at 1:5,000 scale.
5.2 The case study of La Consolazione landslide
La Consolazione landslide is a rotational slide earth
flow, located within Torrice municipality, which twice
reactivated in the last decades (1986; 1990). The
landslide (around 12 ha) exhibits the typical shape of
rotational slide earth flow with a width ranging from
100-150m (in the head where the rotational
mechanism develops) up to 250m (in the accumulation
zone where the enlarging earth flow prevails).
Furthermore, in the upper accumulation zone
(between the elevations of 200 and 220 m. a.s.l.)
Rev. No: 2
Date: 2012-02-15
evidence of creep phenomena (e.g. ripples on the

ground surface) can be distinguished.
The available documents report that in 1986 only the
narrow upper portion (Fig. 8a) reactivated; then, in the
90s subsequent reactivations involved the portion of
the landslide that stretches from the creeping zone
down to the toe.
The above information allows the identification of three
main portions: the old main scarp and the old terraced
accumulation zone with evidence of cracks; the recent
minor scarp (215 m a.s.l.) bordering the reactivated
earth flow; the old accumulation zone.
Referring to the state of activity, the upper portion of
the landslide, as well as the old accumulation zone
appears as dormant; whereas the remaining portion
corresponds to the reactivated earth flow.
Fig. 8a shows the advanced full-resolution DInSAR
velocity map superimposed to both the building map
and the landslide map at 1:5,000 scale. Full-resolution
DInSAR data concentrate on 4 buildings within the
landslide affected area.
Fig. 8. La Consolazione landslide: a) Map of the landslide with buildings and Advanced full-resolution DInSAR velocity data; b)
Longitudinal cross section of the landslide with buildings and DInSAR velocity vectors; c) and d) photo of two damaged buildings
framed in the circle and corresponding to the orange one in the section: 1)terraced old accumulation zone; 2)reactivated old
accumulation zone; 3) old accumulation zone; 4) old scarp; 5) recent scarp; 6)cracks; 7)reactivated earth flow; 8) dormant earth
flow; 9) dormant rotational slide; 10) cross-section; 11) not moving DInSAR coherent pixel or moving coherent pixel on flat
areas; 12) DInSAr coherent pixel moving on vertical direction; 13) not projected translational displacement owing to high
condition number; 14) damaged building; 15) building without damage survey; 16)building without damage survey (Cascini et al.,
2010).

SafeLand - FP7
Page 311 of 401
Deliverable 4.1
The analysis of mean velocity values for 1995-2000

period highlights that the full-resolution coherent pixels
located near of the old scarp exhibit no evidence of
movement, whereas those located within the
reactivated portion are moving with mean velocity
values higher than 0.5 cm/year (Fig. 8a and b).
As for the landslide mechanism, Fig. 11b shows the
longitudinal cross section of the instability
phenomenon and the DInSAR velocity vectors.
Finally, referring to the buildings in the area, it can be
noticed that: i) the evidence of movement recorded to
two buildings, framed with the circle in Fig. 9a and
located in the central portion of the reactivated
landslide, matches the damage occurrence as
observed by the damage survey (Figs. 9 c and d); ii)
the unchecked buildings located in the lower portion of
the landslide exhibiting evidence of movement need a
damage survey in order to assess their structural
integrity.
6. CONCLUSIONS
In the last decade the use of remote sensing data
derived from DInSAR techniques has rapidly grown
thanks to the development of enhanced image
processing algorithms and their increased availability
that, as in the case of the Italian territory, can reach
the total coverage. Consequently, there is the need of
a confident use of these data so as to overcome some
problems that still arise in the analysis of slow-moving
landslides. In this regard, this research introduces an
innovative procedure that allows a significant
improvement of the common 1D LOS velocity maps at
both low- and full-resolution.
The results obtained seem particularly appealing
considering the enhanced capabilities of the newest
sensor (e.g. TerraSAR_X, COSMO/SKYMED, etc)
which will offer high resolution DEMs also allowing
improved spatial resolution, three times higher data
acquisition frequency and an increase in the sensitivity
to temporal decorrelation via the reduction of the
wavelength.
REFERENCES
Berardino P., Fornaro G., Lanari R. and Sansosti E.
(2002). A New Algorithm for Surface Deformation
Monitoring based on Small Baseline Differential
SAR Interferograms. IEEE Trans. Geosci. and
Remote Sens., 40 (11), 2375-2383.
Cascini L., Bonnard Ch., Corominas J., Jibson R.,
Montero-Olarte J. (2005). Landslide hazard and
risk zoning for urban planning and development.
State of the Art Report (SOA7). Proceedings of the
International Conference on Landslide Risk
Management, Vancouver (Canada). O. Hungr, R.
Fell, R. Couture and E. Eberthardt (eds.). Taylor
and Francis, London, 199-235.

SafeLand - FP7
Cascini L., Fornaro G., Peduto D. (2010). Advanced

low- and full-resolution DInSAR map generation for
slow-moving landslide analysis at different scales.
In
press
on
Engineering
Geology,
doi:10.1016/j.enggeo.2010.01.003. now available
online
at:
http://dx.doi.org/10.1016/j.enggeo.2010.01.003
Cascini L., Fornaro G., Peduto D. (2009). Analysis at
medium scale of low-resolution DInSAR data in
slow-moving landslide-affected areas. ISPRS
Journal of Photogrammetry and Remote Sensing,
64(6),598-611. doi:10.1016/j.isprsjprs.2009.05.003
Cascini L., Ferlisi S., Peduto D., Pisciotta G., Di
Nocera S., Fornaro G. (2008) Multitemporal
DInSAR data and damages to facilities as
indicators for the activity of slow-moving landslides.
In: Landslides and Engineered Slopes. From the
Past to the Future. Chen Z., Zhang J., Li Z., Wu F.,
Ho K. (eds.). Proceeding of the 10th International
Symposium on Landslides and Engineered Slopes,
30 June-4July 2008, Xian (China), Taylor and
Francis Group, London. Vol. II, pp. 1103-1109.
ISBN 978-0-415-41196-7.
Cascini L., Ferlisi S., Peduto D., Fornaro G., Manunta
M. (2007) Analysis of a subsidence phenomenon
via DInSAR data and geotechnical criteria. Italian
Geotechnical Journal, 4, 50-67.
Cascini L., Ferlisi S., Fornaro G., Lanari R., Peduto D.,
Zeni G. (2006).Subsidence monitoring in Sarno
urban area via multitemporal DInSAR technique.
International Journal of Remote Sensing Vol.27,
No. 8, 20 April 2006, 1709-1716. Taylor and
Francis.
ISSN
0143-1161,
DOI:
10.1080/01431160500296024.
Fell R., Corominas J., Bonnard C., Cascini L., Leroi E.,
Savage W.Z. (2008). Guidelines for landslide
susceptibility, hazard and risk zoning for land-use
planning. Engineering Geology, 102 (3-4), pp. 99111.
Fornaro G., Pauciullo A., Serafino F. (2009a)
Deformation monitoring over large areas with
multipass differential SAR interferometry: A new
approach based on the use of spatial differences,
International Journal of Remote Sensing, 30 (6),
1455 - 1478.
Fornaro G., Reale D., Serafino F. (2009b). FourDimensional SAR Imaging for Height Estimation
and Monitoring of Single and Double Scatterers",
IEEE Trans. Geosci. Remote Sens., 47 (1), 224237.
Peduto D. (2008). Analysis of ground deformations
related to subsidence and landslide phenomena
via DInSAR techniques. Ph.D.Thesis (In English).
University of Salerno, Italy, May 2008. Tutor:
Cascini L.; co-tutor: Fornaro G..
Page 312 of 401
Case Study 12
Rev. No: 2
Date: 2012-02-15
CASE STUDY 12:

Title: Microseismic monitoring at the unstable rock slope at knes, Norway.
Application: Rock slope monitoring.
Technique: Microseismic monitoring.
Main references: mainly based on Roth et al., 2006 and Roth and Blikra, 2009.
Contributors: NORSAR (Roth, M.).
ABSTRACT
The unstable rock slope at knes has an estimated volume of about 70 million m3, and parts of the slope are
moving at a rate between 2-15 cm/year. Amongst many other direct monitoring systems we have installed a smallscale seismic network in order to monitor microseismic events related to the movement of the slope. The network
has been operational since November 2005 with only a few short-term outages. Seismic data are transferred in
real-time from the site to NORSAR for automatic detection processing. The resulting detection lists and charts and
the associated waveform are forwarded immediately to the early warning centre of the Municipality of Stranda.
Seismic monitoring provides independent and complementary data to the more direct monitoring systems at
knes. We observe increased seismic activity in periods of heavy rain fall or snow melt, when laser ranging data
and extensometer readings indicate temporary acceleration phases of the slope. The seismic network is too small
and the velocity structure is too heterogeneous in order to obtain reliable locations of the microseismic events.
End of 2009 we installed a high-sensitive broadband seismometer (60 s - 100 Hz) in the middle of the unstable
slope. This will allow us to better constrain the locations of the microseismic events and to monitor local/regional
and global seismicity.
1. INTRODUCTION
Microseismic monitoring is a well-established
technique for risk evaluation and production
optimization in mines. During production the stress
field in the subsurface changes and sudden stress
releases in form of microearthquakes occur. Similar
we expect to observe microseismic events at a moving
rock slope. There might be microseismic events
associated directly with the movement in form of stickslip movements on a sliding plane or shear-failures of
intact rock during the formation of a sliding plane.
Furthermore, we expect secondary microseismic
events in form of individual rock-falls and small-scale
slides induced by the movement of the slope.
Microseismic monitoring is an indirect tool in that one
measures effects of the movement and not directly the
movement itself as done for instance by
extensometers etc. However, in contrary to point
measurements, a seismic monitoring system can
provide an overview over the seismic activity in the
entire volume of interest. Depending on the instrument
sensitivity, the installation, the site and the ambient
noise conditions even the tiniest signals can be
detected (on a rocky subsurface for instance a
hammer blow can be easily detected in 100 m
distance).
Changes of the seismic activity can be indicative for
changes of the slope dynamic or the slope movement
rate. The seismic systems at knes are considered as
an important part of the overall monitoring system
providing independent and complementary input to the
National centre for rock slide monitoring at Stranda.

SafeLand - FP7
2. AKNES ROCKSLOPE
The Aknes rock slope is located in the county of Mre
og Romsdal, Norway approximately halfway between
Bergen and Trondheim (Figure 1). Already in 1985
geological mapping revealed the instability of the site
and continuous extensometer measurements showed
an opening of fractures in the upper part of the slope
with a rate of several centimeters per year. The
volume of the instability was estimated to be about 6
million m3 at that time.
Later investigations based on additional field studies,
high-resolution aerial photography and detailed
mapping concluded that the volume is much larger, i.e.
40 70 million m3, and that a sudden failure of the
slope could generate a local tsunami in the Storfjord
system.
The upper part of the instability is clearly defined by
distinct back-scarps and fractures some of them
several meters wide and 20 - 30 m deep. The lower
limit of the unstable slope is more difficult identify,
because it is covered with boulders and dense
vegetation. However, several springs can be found at
about 150 m above sea level, which may represent the
outlets of water running along detachment faults. The
area of the unstable part is about one km2 (Figure 2).
Core drillings in summer 2005 revealed fault gauges in
a depth of 30 to 40 m in the centre of the slope and in
the upper part of the site fault gauges could be found
even down to 100 m depth.
Page 313 of 401
Deliverable 4.1
consumption and a GSM telephone relay. It is

powered by a set of batteries which in turn is charged
by a diesel generator. Data are transmitted via radio
link and internet to NORSAR for automatic processing.
During the first 9 months of operation data were
recorded in continuous mode and in August 2006 the
system was switched to triggered-mode recording with
higher sampling rate (1000 Hz).
Fig. 1. Location of the knes rock slope.
Fig.3. Digital elevation map with the positions of the

geophone network and the seismic station AKN (black),
helicopter pads (yellow), three extensometers (green), two
laser ranges (magenta) and the bunkers with power supply
and radio communication (red).
Fig.2. Outline of the unstable area (picture take from the

opposite shore of the fjord).
3. INSTRUMENTATION
There are several real-time monitoring systems (e.g.
optical total station, GPS, laser ranging system,
extensometers, crackmeters, tiltmeters, borehole
DMS), installed at knes. In addition to those
instruments we installed a small-scale geophone
network in 2005 and a broadband seismic station in
2009.
The seismic network consists of 8 three-component
geophones (4.5 Hz) covering an area of about 250 x
150 m in the upper part of the slope (Figure 3). The
geophones are connected by armored cables with the
central acquisition system (Figure 4) in Bunker I. The
acquisition system customized by NORSAR engineers
includes a 24-channel digitizer (Geode, Geometrics), a
GPS clock, an industry computer with low power

SafeLand - FP7
Fig. 4. Up: Data acquisition system of geophone network

installed in the upper bunker. Down: Installation of a 3component geophone.
Page 314 of 401
Case Study 12
The broadband station AKN (international registry of

seismograph stations) has been installed on an
outcrop in the middle of the unstable area (Figure 5).
The station is connected to Bunker II and provides
continuous real-time data to the early warning centre
in Stranda (http://www.aknes.no/) and to NORSAR.
The sensor is a Guralp ESPC with a frequency band
from 1/60 to 100 Hz in order to monitor the local
microseismic activity as well as the global/regional
seismicity. The broadband seismometer has a
sensitivity which is about 65 times higher than that of
the installed geophones and the sensor will be able to
record the same microseismic events as seen by the
geophone network. Both systems are synchronized by
GPS clocks and therefore we expect to obtain better
constraints for the locations of the observed
microseismic events.
Rev. No: 2
Date: 2012-02-15
Fig. 6. Up (black) and Down (red) times of the 24 channels of

the geophone network for the time period Oct 2005 - Apr
2010.
The state-of-health of the geophone network is being

checked continuously and seismic data are transferred
automatically from knes to NORSAR using ftp. In
triggered-mode the average daily data volume is about
250 MByte. The data are submitted to an automatic
event detection (Signal-to-noise ratio > 2 on at least 10
channels within a common time window of 1 s). The
results are displayed in near real-time (~10-15 min
delay)
on
http://www.norsar.no/pc-47-48-LatestData.aspx. In addition we automatically generate plots
of the event waveforms and make them available to
the knes/Tafjord data base.
The broadband station AKN has been online without
any data outage so far. The waveform data are
publicly available from NORSAR or from ORFEUS
(http://www.orfeus-eu.org/) and waveform plots can be
accessed at http://www.norsardata.no/NDC/heliplots/.
A real-time data stream is forwarded to the early
warning centre in Stranda, Mre og Romsdal.
5. OBSERVATIONS
Fig.5. Up: Installation of the broadband seismic station AKN.

Down: Seismometer pit with sensor, power/communication
module and data acquisition module.
4. PERFORMANCE AND DATA PROCESSING

The robustness of the system is a very important
issue, because the knes site is remote and can only
be accessed by helicopter. The seismic network has
been operational without major outages since 2005
(Figure 6) and it is performing very stable. So far only
two complete outages happened: one due to a failure
of the central power supply, the other due to a failed
software update. Partial outages have been caused by
cable damages from rock falls and water leakage into
the geophones.

SafeLand - FP7
Figure 7 displays an example for a microseismic event

(left) and a rock fall (right) recorded with the geophone
network. The plots show the 24 channels of the 8
three-component geophones for a time window of 17
seconds. The microseismic event has a duration of
about 1-3 seconds and it is relatively strong with a high
signal-to-noise ratio on all channels. The rock fall has
a total duration of more than 10 seconds with multiple
signal arrivals.
Figure 8 shows the signal amplitudes of reviewed
events for the time period from January 2006 until
March 2009. The data base has been inspected
visually and all obvious non-local events, man-made
events and other noise events have been removed.
The signal amplitudes vary over a range of about 3
magnitudes with most of the events having amplitudes
less than 1 mm/s.
Page 315 of 401
Deliverable 4.1
Fig.7. Left: Microseismic event, Right: rockfall.

Fig. 10. Air temperature measurements.
Fig. 8. Signal amplitudes of reviewed events.
Another way to illustrate the seismic activity is to plot

the cumulative signal amplitudes (Figure 9). A
constant slope of this curve is indicative for a constant
rate of seismic activity. Kinks and steps in this curve
are Changes of the slope are the result of temporal
clusters of small events and/or the occurrence of
larger individual events. The change of seismic activity
might be indicative for acceleration phase of the knes
slope. We observe changes in seismicity mainly
during autumn (2007, 2008), when heavy rainfalls
occurred and during spring (2007, 2008, 2009)
seasons, when air temperature stays above 0 degrees
(Figure 10) and snow melt starts.
During this periods we also have indications for slope
acceleration from the extensometer (Figure 11) and
laser measurements (Figure 12).
Fig. 11. Extensometer measurements. Acceleration phases in

autumn 2007 at extensometer 2 and 3. Instrument outages
spring and summer 2008.
Fig. 12. Laser measurements. Acceleration phase in spring

2008 on lower laser.
6. CONCLUSIONS
Fig. 9. Cumulative signal amplitudes.

SafeLand - FP7
The seismic monitoring systems at knes are

operating reliable and provide independent real-time
input to the early warning centre in Stranda. We
observe a higher level of microseismicity during snow
Page 316 of 401
Case Study 12
Rev. No: 2
Date: 2012-02-15
melt, which coincides with temporary acceleration

phases of the slope. Microseismic monitoring is
complementary to the direct measurements at the
unstable rock slope, but it is important. All
measurements have uncertainties and systems can
fail. Therefore, an early warning system that is based
on different technologies and that correlates
observations from independent real-time data will
perform more stable and reliable.
REFERENCES
Roth, M., Dietrich, M., Blikra, L. H. & I. Lecomte 2006.
Seismic monitoring of the unstable rock slope site
at knes, Norway. 19th annual Symposium on the
Application of Geophysics to Engineering and
Environmental Problems, Seattle, USA
Roth, M. & L. H. Blikra 2008. Microseismic monitoring
at the knes rock slope, Norway. EGU General
Assembly 2008, Vienna, Austria.
Roth, M. & L. H. Blikra 2009. Seismic monitoring of the
unstable rock slope at knes, Norway. EGU
General Assembly 2009, Vienna, Austria.
Roth, M. and L. H. Blikra (2009): Seismic Monitoring of
the Unstable Rock Slope at knes, Norway.
Geophysical Research Abstracts, 11, EGU20093680, Vienna, Austria
WEBPAGES
Seismic network latest automatic processing results:
http://www.norsar.no/pc-47-48-Latest-Data.aspx.
Broadband station AKN current data:
http://www.norsardata.no/NDC/heliplots/
Broadband station AKN historic data:
http://www.norsardata.no/NDC/data/
http://www.norsardata.no/NDC/lpdata/
Early warning centre Stranda (in Norwegian):
http://www.aknes.no/

SafeLand - FP7
Page 317 of 401
Case Study 13
Rev. No: 2
Date: 2012-02-15
CASE STUDY 13:

Title: Landslide of Rindberg / Sibratsgfll investigated with airborne geophysical survey.
Application: Landslide investigations.
Technique: Airborne geophysical survey.
Main references: Based on the paper: Supper et al. 2008.
Contributors: GSA (Supper R., Baron I.).
ABSTRACT
After a large landslide event in Sibratsgfll/Austria several exploration methods were evaluated on their
applicability to investigate and monitor landslide areas. The resulting optimised strategy consists of the combined
application of airborne electromagnetics, ground geoelectrical measurements and geoelectrical monitoring
combined with hydrological and geological mapping and geotechnical modeling. Interdisciplinary communication
and discussion was the primary key to assess this complicated hazard situation.
1. INTRODUCTION
In spring 1999, after a short period of heavy
precipitation and the rapid melting of snow, a
catastrophic landslide was triggered on the South-flank
of the Rubach Valley near Sibratsgfll in the province
of Vorarlberg (Austria). Shortly after the first slide
activity was observed, the State Department of
Avalanche and Torrent Control authorized preliminary
geological investigations. As a follow up of this first
phase of investigation, a complex research program
was initiated. The final goal of this study was to
develop an operative strategy to optimise measures in
case of future events. The applied methods
incorporated geo-morphological, hydro-geological and
geophysical surveys of the area. It was determined
that airborne geophysical measurements are a
valuable tool to obtain a quick overview of the
geological situation. Furthermore they can help to
detect areas susceptible to a high sliding risk, to assist
the follow up geological and hydrological mapping
program and to optimise further ground-geophysical
surveys. In a second step, ground geoelectrical
surveys
were
used
to
achieve
advanced
understanding of the internal structure of the landslide.
The location of survey lines was planned according to
the resistivity pattern derived from the airborne
electromagnetic survey. Based on these findings and
on the results of the geo-hydrological mapping
program, boreholes were drilled to calibrate the
geoelectrical results. Labaratory tests were performed
on soil and rock samples to determine the
geotechnical parameters the main subsurface units.
Additionally, geophysical and hydrophysical logging
were carried out. Based on these results a
geotechnical subsurface model was created and
parameters and conditions of safety and failure were
calculated. Finally a multi parameter monitoring
network was installed and has been operated since

SafeLand - FP7
2002. In this paper we focus on the geophysical

aspects of the strategy.
2. THE GEOLOGICAL FRAMEWORK

The research area is entirely located within the
Feuersttter Unit, which is characterized by
extensive rock disruption. The landslide area itself is
mainly composed of rocks of the Junghansen and
Schelpen series. These sub-units consist of marl and
schist with highly variable stability as a result of
tectonic fracturing. Due to their low resistance against
weathering, rocks degrade under the influence of
water into deeply weathered granular soils. Within
these areas, all primary scarps are located. The
Junghansen and Schelpen series are embedded into
the more stabile Feuersttter sandstones and
limestones of the Aptychen series. Fig.1 shows the
results of the geological mapping of the landslide area
(Jaritz et al, 2004) and surroundings.
3. THE AIRBORNE GEOPHYSICAL SURVEY

Soon after the landslide event, a high resolution, multiparameter airborne survey was performed. The main
part of the airborne system (Motschka, 2001)
consisted of a GEOTECH-Bird of 5.6 m length and
140 kg weight.. For the Sibratsgfll survey,
electromagnetic alternating fields with frequencies of
3200Hz (coaxial coil) and 7190Hz (coplanar coil) were
used, thus allowing investigation depths up to 70 m
below ground surface. The electromagnetic bird was
supplemented by a laser altimeter, two differential
GPS sensors, one of them located in the bird and the
other one in the helicopter, a Cs-magnetometer, a
gamma spectrometer and a passive microwave soil
moisture sensor.
Page 319 of 401
Deliverable 4.1
Profile MONITORING
Fig. 1. Geological overview map of the landslide Rindberg.
Out of the set of different airborne techniques, the

electromagnetic method was expected to deliver the
most useful results for the investigation of landslide
areas. This is due to the fact that with this method
depth-specific sounding can be performed. Moreover
the derived parameter, namely the electrical bulk
resistivity of the subsurface, directly depends on fluid
content, soil porosity and clay content. These
parameters describe indirectly the internal lithological
composition of the landslide. Using the results of a
two-layer inversion cross-correlated with remote
sensing data (satellite images, ortho-photographs and
digital terrain data from airborne laser scanning) and
results from geological mapping, a conceptual model
of the subsurface structure was derived.
Fig. 2 shows the results of the inversion of the
electromagnetic data using a homogenous halfspace
model (Ahl, 2007). This approach is very useful in
providing a first overview over geological structures in
a research area.
The territory affected by movement of clay and marl
material is clearly detectable in the central zone of the
landslide, exhibiting medium to low resistivities (35-100
Ohmm). Sandstones (resistivity>175 Ohmm) of the
Feuersttter series, building up the unmoved frame of
the landslide on the western side, can be delineated
laterally with high resolution due to their contrast in
resistivity to the surrounding materials.
However, ground moraines, defining the south-western
border of the landslide, and limestone material from

SafeLand - FP7
the Aptychen series, dominating thenorth-eastern part

of the landslide, show the same resistivity signature.
Therefore they cannot be distinguished from the
sandstone using only electromagnetic results.
Incorporating additionally results from the gamma ray
mapping, particularly the potassium distribution (fig.3),
a more accurate differentiation between Aptychen
series (low potassium content) and sandstones
(medium potassium values) was possible.
It is remarkable that the distribution of maximum
potassium content follows the main flow track of the
landslide (fig.3). The reason for that is not obvious.
This might be due to the fact that the top layers in the
landslide material are mainly mixtures of marl and
schist originating from Junghansen and Schelpen
series, which show also high potassium contents at insitu position. Destruction of the vegetation cover and
dehiscing of the sward by high movement rates could
be other causes. Within a second processing step, the
geoelectric results as well as information from core
drillings were used to constrain the inversion of the
airborne electromagnetic data. This improved
resistivity model was consequently used to refine the
subsurface structural model, based on the complex
interpretation of all available data.
Furthermore it has to be pointed out that all airborne
geophysical results had been very valuable for
mapping geologists as they helped to optimize the
follow-up mapping procedures. Additionally it helped to
minimize actual field work, which is often very difficult
and time consuming in the rugged terrain of unstable
slopes.
Page 320 of 401
Case Study 13
Rev. No: 2
Date: 2012-02-15
Fig. 2. Results of the inversion of airborne electromagnetic mapping: homogenous halfspace inversion - Parameter: Resistivity
[Ohmm].
Fig. 3. Results of the inversion of airborne mapping: Parameter: Gamma ray spectroscopy Potassium [ppm].
4. CONCLUSIONS
Several methods were evaluated to design an
improved interdisciplinary strategy for immediate
measures to be applied in case of future landslide
events. The strategy allows to quickly assess the

SafeLand - FP7
prevailing hazard situation and to develop and

recommend effective mitigation measures.
The resulting optimised approach consists of the
application of airborne electromagnetics, ground
geoelectrical
measurements
and
geoelectrical
monitoring combined with hydrological and geological
mapping and geotechnical modeling. Interdisciplinary
Page 321 of 401
Deliverable 4.1
communication and discussion was the primary key to

access the complicated hazard situation in the case of
the large-scale landslide event evaluated at Rindberg.
Acknowledgements.
Winkler, E., Seiberl W., Ahl A.: Interpretation Of

Airborne Electromagnetic Data with Neural
Networks: In Geophysical Applications of Artificial
Neural Networks and Fuzzy Logic, Sandham,W.
and Leggett M. (Eds.), Kluwer Academic
Publishers, 2003.
This study was financed by grants from the State

Department of Torrent and Avalanche Control (WLV),
the Federal Ministry of Transport, Innovation and
Technology (BMVIT), the Federal Ministry of Science
and Research (BMWF), the State of Vorarlberg and
internal research funds of the Geological Survey of
Austria. We further like to thank Grete Whrer WLV
(logistical support), Martina Mittelberger Land
Surveying Office Feldkirch (GPS) and Robert Hofmann
(geotechnical consulting) for fruitful cooperation.
REFERENCES
Ahl, A., Winkler E., Bieber G., Rmer A.: Probabilistic
inversion of aeroelectromagnetic data with a
homogeneous halfspace model: Journal of Applied
Geophysics, submitted, 2007.
Avdeev A.: Three dimensional electromagnetic
modelling and inversion from theory to practice,
Surveys in Geophysics 26:767-799, 2005.
Jaritz W, Reiterer A, Supper R.: Landslide Rindberg
(Vorarlberg):
Multidiscipline
Research,
Proceedings of the 10th Interpraevent Congress,
Riva del Garda, 2004.
Motschka, K.: Aerogeophysics in Austria: Bulletin of
the Geological Survey of Japan Vol. 52 No. 2/3,
pp. 8388, Tsukuba, Japan, 2001.
Pedler, W. H., Head, C. L., and Williams, L. L.:
Hydrophysical
Logging:
A
New Wellbore
Technology for Hydrogeologic and Contaminant
Characterization of Aquifers, Proceedings of
National Groundwater Association 6th National
Outdoor Action Conference, pp 1701-1715, 1992.
Seiberl, W., Ahl A., Winkler E.: Interpretation of
airborne electromagnetic data with neural
networks: Exploration Geophysics, 29, 152-156,
1998.
Sengpiel, K., Siemon B.: Examples of 1-D inversion of
multifrequency HEM data from 3-D resistivity
distributions: Expl. Geophys., 29, 133-141, 1998.
Supper, R.; Ahl, A. ; Rmer, A.; Jochum, B.; Bieber,
G.: A complex geo-scientific strategy for landslide
hazard mitigation from airborne mapping to
ground monitoring, Advances in Geosciences, 14,
1-6, 2008.

SafeLand - FP7
Page 322 of 401
Case Study 14
Rev. No: 2
Date: 2012-02-15
CASE STUDY 14
Title: Site-specific geotechnical investigations for landslide characterization and detection.
Application at the knes rockslide, western Norway.
Application: Geotechnical investigations and instrumentation.
Technique: Core drilling; Bedrock core logging and testing; Geophysical borehole logging;
Hydrological testing; Borehole inclinometers (DMS).
Main references: See references list at the end of the case study.
Contributor: knes/Tafjord Early Warning Centre (Blikra L.H. & Kristensen L.); ICG Geological Survey of Norway (Oppikofer T.)
ABSTRACT
Site-specific geotechnical data are essential for the design and implementation of monitoring systems and earlywarning of large landslides. This data gives us important information about the characterization and strength of
geological structures and the kinematics of the unstable areas. The case study shows some of the investigation
methods used for the knes rockslide in western Norway. The methods used include geological field mapping,
geophysical surface measurements and measurements of deformations. One of the most important investigations
is coming borehole drilling. Bedrock-core logging and testing, geophysical logging and instrumentation in order to
investigate subsurface deformations are crucial data in order to understand geometry, volumes and kinematics of
the rockslide. Large efforts have also been on the investigations of the hydrological system. The full
understanding of the geological model and the kinematics will only be achieved by doing detailed subsurface
investigations and instrumentation in boreholes.
1. INTRODUCTION
Site-specific geotechnical data are essential for the
design and implementation of monitoring systems and
early-warning of large landslides. This data gives us
important information about the characterization and
strength of geological structures and the kinematics of
the unstable areas. They are key data in order to
perform reliable numerical models and to construct
geological models.
focusing on the knes rockslide and other large

unstable slopes (Blikra, 2008). The knes rockslide
with an estimated volume of 30 to 54 Mm3 is moving
with a velocity of up to 8 cm/year, and its failure may
trigger a devastating tsunami in the fjord.
The case study gives an overview of the geotechnical
investigations and instrumentation that have been
performed in order to be able to construct a reliable
geological model and as an important basis for the
design of monitoring systems.
2. GEOLOGICAL SETTING
Fig. 1. The location of the knes rockslide in the inner part of

Storfjorden, western Norway
The case presented here is the knes rockslide,

situated on the north-western flank of the
Sunnylvsfjord in Western Norway (Fig. 1). The
knes/Tafjord project initiated in 2004 was a large
investigation, monitoring and early-warning project

SafeLand - FP7
knes rockslide is located in the Western Gneiss

Region and the bedrock is constituted of medium
grained granitic gneisses and granodiorite gneisses of
Proterozoic age (Ganerd et al., 2008) with bands and
lenses of mafic material. Thee gneisses have
generally well developed foliation surfaces and mineral
banding (Braathen et al., 2004) and numerous
centimetric to decametric, close to tight folds are
visible in the field. At knes there occur biotite-rich
layers up to 20 cm thick and coincide with zones of
high fracture frequencies and sliding surfaces is likely
located within these mica-rich layers (Ganerd et al.,
2008).
The morphological investigations
characteristic features (Fig. 2):
show
several
Page 323 of 401
Deliverable 4.1
An about 500m more or less continuous back

crack (Upper tension fracture).
A large depression in the upper western corner of
the rockslide, developed into a graben structure.
The total vertical displacement is from 20-30 m.
A series of tension fractures from the upper to the
middle part of the slope.
Prominent slide scars along the southwestern
canyon. Historical data indicates a slide in the
upper part in the late 1800, and slides also in
1940 and 1960.
Small slide scars in the lower part of the rockslide.
Large blocks or parts of the rocks is coming out of
the slope at two particular areas, one in the
middle part and one area in the lowermost part.
Distinct water springs at the lowermost part of the
slope at about 100masl. However, there are also
smaller springs in the middle part of the slide
area.
measured by periodic GPS and total station

demonstrates that there is a movement of between 2
and 4 cm/year. The south-western flank has a much
larger movement with a general trend of 6-8 cm a
year, but locally up to 15 cm/year. The same trend can
also be seen on the measurements done by groundbased radar campaigns placed in Oaldsbygda on the
other side of the fjord.
Fig. 3. Shaded relief map of the knes area showing

movement data measured by periodic GPS and total station.
The extent of the possible unstable area is shown in colour.
The fast moving upper western part is indicated by red
colures, and the interpreted compressional zones are
indicated.
4. DRILLING AND CORING
Fig. 2. Morphological features of the knes rockslide (Blikra,

2008).
3. SURFACE DISPLACEMENTS
The movement of the knes area has been measured
by a series of different methods, including GPS, total
station, ground-based radar, extensometers and single
lasers (Fig. 3). The large tension cracks in the upper
part are more than 500 m long and limit the total
volume which is moving. The movement data

SafeLand - FP7
An important part of the investigations at knes has

been the core drilling, including 7 deep boreholes at
150-200 m depth at three different locations (Fig. 3).
Logging of different types of data was performed
during the drilling operation. Important are especially
the water pressure, water flow, the system pressure
and the penetration velocity (Fig. 4). These data are
important information at an early stage and gives
information about the hydrological conditions and the
occurrence of large fractures. Note for example the
reduced water pressure and system pressure at 35-50
m depth due to large open fractures. Similar reduced
water pressure can be seen at 120 m depth.
Page 324 of 401
Case Study 14
Rev. No: 2
Date: 2012-02-15
Fig. 4. Data from the drilling operation at the middle site showing water pressure, water flow, system pressure, and penetration
velocity.
The bedrock cores were of excellence quality and

gave important information about the rock quality and
the characteristics of fractures and weak zones
(Ganerd et al., 2007). Some zones were totally
crushed, while others have well-defined breccias with
silt and clay (Fig. 5). We experienced some core loss
at zones characterized by crushed rocks. All cores
were carefully logged, and the fracture frequency gives
a good picture of the down-hole variation (Fig. 6, right).
higher velocity with less change is documented until

125 m depth.
5. GEOPHYSICAL LOGGING
All boreholes were logged by different methods before
the in-place instrumentation (Fig. 6). The resistivity
data gives good indications of the water conditions in
fractures and the seismic P-velocity gives overview of
the rock quality and fractures. The seismic velocity
from just below the water level (52 m depth) and down
to about 105 m is characterized by large variations in
velocity indicating many fractures. Below this zone, a

SafeLand - FP7
Fig. 5. Drilling core at 34 m depth at the middle borehole at

knes, showing a crushed and brecciated zone with silt and
clay, representing the upper sliding zone. This sliding plane is
also documented by displacement in DMS column
(Differential Monitoring of Stability, in place instrumentation).
Page 325 of 401
Deliverable 4.1
6. HYDROLOGICAL TESTING
A vital part of stability considerations of large
landslides is the hydrological conditions. An integrated
approach combining heat pulse flowmeter logging with
a new protocol of dynamic fluid electric conductivity
(DFEC) were performed by ETH (Switzeralnd) for the
identification and characterization of groundwater flow
(Thny, 2008). A pulse flow-meter logging was also
performed, in addition to large-scale tracer
experiments.
The dynamic fluid electric conductivity logging was
performed in the following way:
1. The borehole were first flushed with water
with a higher salinity that the water in the
system
2. The conductivity were logged several times
after the flushing in order to observe the time
changes
The test gives very interesting data showing clearly
were we have stagnant conditions and were we
distinct water circulation (Fig. 7). The profile shows
generally stagnant flow conditions above 87 m and
below 119 m, but with downward oriented flow in
between, and upward directed flow in the uppermost
10 meters of the logged section. The loss of drilling
fluid pressure from 87 to 115 m depth and the highly
conductive outflow at 119 m depth clearly indicate the
occurrence of highly permeable rock masses between

87 and 120 m depth (Thny, 2008).
7. SUBSURFACE DISPLACEMENTS
Tree of the boreholes has been instrumented by the
DMS (Differential Monitoring of Stability) system. The
DMS is a multiparametric column for investigations
and permanent monitoring of subsurface movements;
temperature and water pressure, and have been
developed and patented by the Italian company CSG
srl. The DMS column is like a sensorized spiral cord or
snake composed of a sequence of hard tabular
modules connected to each other by special
2D/3Djoints. The present system at knes composes
3 continuous columns 50, 100 and 120 m long
measuring the movement in 2D. The 120 m long
column consists of totally 245 sensors. The sensors
composes inclinometers, temperature sensors and in
selected modules pietzometers and digital compass.
Figure 8 show example of data from the first campaign
at the middle location. It clearly documents the sliding
zone at 33-34 m depth, but also movement at deeper
parts.
The instrumentation in boreholes is essential for both
the investigations of subsurface characteristics (sliding
planes, volumes etc) and for operative early-warning.
Fig. 6. Geophysical logging upper borehole (Rnning et al., 2006).

SafeLand - FP7
Page 326 of 401
Case Study 14
Rev. No: 2
Date: 2012-02-15
Fig. 7. Measured and simulated electric conductivity profiles

in upper borehole. From Thny (2008).
Fig. 8. Displacement in upper part of middle borehole

from a 10 month campaign. Data is from a DMS
system of 50 in-place inclinometric sensors.
8. GEOLOGICAL MODEL
Numerous geological, geophysical and geotechnical
studies were made on the knes rockslide to better
understand its mechanism and to locate sliding
surfaces (e.g. Rnning et al., 2006; Blikra, 2008;
Ganerd et al., 2008). The slope movements are
monitored by GPS, total station, extensometers, laser
distance meters, ground-based radar, InSAR and
terrestrial laser scanning (Kveldsvik et al., 2006; Blikra,
2008). A detailed summary of these studies can be
found in Oppikofer (2009). The geological model is not
yet totally understood, but the instability is clearly
controlled by the structural pattern of the gneissic
rocks, with the step back fracture following a sharp
fold, and the sliding zones being parallel to the foliation
planes further down slope. The upper part of the
rockslide is characterized by tension, while several
parts in the lower area show compression features
(Fig. 2 and 9). However, this only gives the general
picture, as the geometry and subsurface deformation

SafeLand - FP7
is much more complex (e.g. Jaboyedoff, 2010). The

rockslide is composed of several individual blocks with
different surface movement directions, and also
differential movement at different depths.
Fig. 9. A proposed general geological model for a 2D length

profile at knes. The colours show the 2D resistivity data,
with blue codes reflecting low resistivity.
Page 327 of 401
Deliverable 4.1
9. CONCLUSIONS
Monitoring and early-warning related to large
landslides requires detailed investigations with the use
of different methods. These include geological field
mapping, geophysical surface measurements and
measurements of deformations. One of the most
important data is coming from borehole drilling.
Bedrock-core logging and testing, geophysical logging
and instrumentation in order to investigate subsurface
deformations are crucial data in order to understand
geometry, volumes and kinematics of the rockslide.
Large efforts have also been on the investigations of
the hydrological system.
Thny, R.2008: Dynamic fluied electric conductivity

logging for identification and characterization of
preferential groundwater flow in the knes
rockslide (Norway). MSc Thesis, Engineering
Geology, ETH Zurich.
Ganerd, G.V., Grneng, G., Aardal, I.B., Kveldsvik,
V., 2007. Core logging of seven boreholes from
knes, Stranda municipality, Mre and Romsdal
County. NGU report no. 2007.020, pp. 222.
REFERENCES
Blikra. L.H. (2008): The knes rockslide; monitoring,
threshold values and early-warning. In: ZUYU
Chen; Jian-Min Zhang; Ken Ho; Fa-Quan Wu;
Zhong-Kui Li (Eds). Landslides and Engineered
Slopes. From the Past to the Future. Proceedings
of the 10th International Symposium on Landslides
and Engineered Slopes, 30 June - 4 July 2008,
Xi'an, China. Taylor and Francis. ISBN: 978-0-41541196-7.
Ganerd, G., Grneng, G., Rnning, J.S., Dalsegg,
E., Elvebakk, H., Tnnesen, J.F., Kveldsvik, V.,
Eiken, T., Blikra, L.H. and Braathen, A. (2008).
Geological model of the knes rockslide, western
Norway. EngineeringGeology102, 118.
Braathen, A., Blikra, L.H., Berg, S.S. and Karlsen, F.
2004: Rock-slope failures of Norway; type,
geometry, deformation mechanisms and stability.
Norwegian Journal of Geology (NGT) 84, 67-8.
Kveldsvik, V. (2008). Static and dynamic stability
analyses of the 800 m high Aknes rock slope,
western Norway. Ph. D. thesis, Norwegian
University of Science and Technology (NTNU),
Dept. of Geology and Mineral Resources
Engineering. 128.
Oppikofer, T. 2009: Detection, analysis and monitoring
of slope movements by high-resolution digital
elevation models. Ph.D. thesis, University of
Lausanne.
Jaboyedoff, M., Oppikofer, T., Derron, M-H., Bhme,
M., Blikra, L.H. & Saintot, A. 2010. Complex
landslide behavior and structural control: a 3D
conceptual model of knes rockslide, Norway.
Special Publications of the Geological Society of
London (in press).
Rnning et al., 2006 Rnning, J.S., Dalsegg, E.,
Elvebakk, H., Ganerd, G. & Tnnesen, J.F. 2006:
Geofysiske mlinger knes og Tafjord, Stranda og
Nordal kommuner, Mre og Romsdal. Norges
geologiske underskelse Rapport 2006.002 (in
Norwegian).

SafeLand - FP7
Page 328 of 401
Case Study 15
Rev. No: 2
Date: 2012-02-15
CASE STUDY 15:

Title: From early warning to site-specific kinematic analysis: the case of Bagnaschino
landslide.
Application: Landslide monitoring.
Technique: Differential Monitoring of Stability (D.M.S.) system.
Main references: Unpublished report, specific for the Safeland project.
Contributors: CSG S.r.l. (Levisolo, M).
ABSTRACT
This case study shows the capabilities of a technique for landslide monitoring: Differential Monitoring of Stability
(D.M.S.) system.
The instrumentation is installed on Bagnaschino landslide that is located in the Municipality of Torre Mondov
(Cuneo, Piedmont) and is characterized by a composite movement. During the observation period (28th October
2008 13th July 2009) it was possible to monitor different kinematics and different weather conditions in
continuous. D.M.S. column allowed to investigate 5 triggering events and their relative period of stasis. In spite of
a displacement of approximately 60 cm that occurred at 7 m blg, the integrity of D.M.S. column was preserved.
Continuous monitoring of the landslide allowed to notice weak deep creep in the interval 30-44 m blg in addition to
considerable shallow movements. The system was efficient also with the presence of elevated displacements
unfitting with the functionality of the traditional borehole instrumentation.
1. HISTORY OF THE SITE

1.1 Characteristics of the landslide
During the flood event occurred in Piedmont in 1994
the highway n. 194 was seriously damaged by a
landslide activated in Bagnaschino site (Torre
Mondov, Cuneo) (fig. 1-2).
D.M.S.
Fig. 2. DEM of Bagnaschino landslide and D.M.S. position.
Fig. 1. Localization of Bagnaschino site (Google Earth).

SafeLand - FP7
In a first time, Casotto Torrent strongly damaged the

road with wide erosions and, in a second time, in
proximity of km 1+400, a large landslide invaded the
carriage way. The phenomenon is classified as a
complex landslide movement. It was considered as a
reactivation of a deep-seated gravitational slope
deformation. The estimated value of area and volume
involved in the phenomenon are respectively 150.000
m2 and 1.2 million m3. Geotechnical investigations
allowed to understand the stratigraphy of the landslide.
At the top there is a thick layer of eluvial-colluvial
overburden (up to 20-30 m in the upper part)
composed by heterogeneous debris inside an
abundant sandy clay loam matrix. Below this layer,
there is a strongly alterate bedrock mainly composed
by laminated micaschistes overlapping on metabasites
and prasinites. Because of heavy rainfalls of flood
Page 329 of 401
Deliverable 4.1
event of 1994, there was a saturation of overburdens

material that slid over the bedrock, overflowing in the
slope lower part.
1.2 Projects and works
In 1998 it was made a General Project relative to the
reconstruction of the road sections damaged during
the flood event of 1994.
After a first analysis it was hypothesized to divert the
highway on opposite slope by means of two bridges.
However this solution was much onerous and also the
opposite slope was characterized by unacceptable
geotechnical and global stability characteristics.
So the attention has been focused on the slope on
landslide with further investigations through boreholes
with piezometers and inclinometers. Following there
are the main works realized:
modelling of detachment part of the landslide;
regularization and profiling of the upper slope
of landslide zone for the formation of a single
sloping plane;
excavation of unstable material with the
creation of n4 benches;
building of draining trenches and subhorizontal drains on the slope;
natural engineering works;
creation of little channels for water draining;
profiling of the most steep areas;
realization of monitoring and control works
(piezometers and inclinometers).
Finally, it was realized an artificial gallery with the aim
of road protection against the material that could come
from the slope.
2. MONITORING SYSTEM WITH D.M.S.

COLUMN
In order to monitor in continuous the stability
conditions, the province of Cuneo (Civil Protection
Office) established in 2008 a slope monitoring plan
with a D.M.S. system. The monitoring instrumentation
is composed of n1 multiparametric D.M.S. column 60
m long installed in inclinometric pipe DN 80 mm. The
column is connected to a local control unit equipped
with solar cell power supply and to a GSM data
transmission. The column is composed of a sequence
of rigid tubular modules containing various sensors,
connected each other by means of special 2D/3D
flexible joints having 2 or 3 freedom degrees: this
allows to single units to adapt with borehole
characteristics and soil movements and to conserve
azimuth direction compared with an external reference
system.
In order to monitor the conditions of the groundwater,
there were installed n2 piezometric sensors.
The instrumentation was installed on 28th October
2008 (fig. 3) and immediately linked in situ in wireless
remote mode through a GSM network.

SafeLand - FP7
Fig. 3. Installation of D.M.S. column.
The acquisition unit performs in continuous:

transducers energization;
sensorial data acquisition;
data storage;
alarm thresholds management on 4 levels.
Following there are described the technical
characteristics of sensors, control unit and data
transmission applied for the monitoring of
Bagnaschino landslide:
Biaxial inclinometric sensors
Range (deg)
Resolution (deg)
Accuracy (deg)
Linearity (25% range)
Access time (ms)
Power (V)
Temp. op. range (C)
Temp. storage (C)
20
0.005
0.02
0.2
160
7-12 DC
+70/-20
+90/-40
Piezometric sensors
Range (psi)
Resolution (% FS)
Linearity (% FS)
Power (V)
Temp. op. range (C)
Temp. storage (C)
0-30-100
0.01
0.02
7-12 DC
+70/-20
+90/-40
Control unit and data transmission unit

Communication
Applicable sensors
Data storage ability
Power (V)
Case
Battery
Remote control with

GSM-GPRS/wireless in
situ with serial RS 232
or RS 485
Max n128 DIGITALANALOGICAL
Circular buffer
7-12 DC
IP68
12 V, 5 years duration
The data in binary format of D.M.S. column are

transmitted by communication protocol RS 485 to the
control and recording unit, where they can be acquired
in continuous or at regular intervals defined by the
user by means of the GEOMASTER management
software.
The program allows remote control of the D.M.S.
monitoring, real-time data visualization, data download
(memory buffer) and the setting of alarms and phone
Page 330 of 401
Case Study 15
Rev. No: 2
Date: 2012-02-15
numbers of the control staff on duty. Remote control,

download, storage and processing of data can be
executed through a server unit in the monitoring centre
or, in emergency conditions, through a mobile remote
control unit equipped with integrated GSM-GPRS
wireless transmission (fig. 4).
Fig. 4. Remote control of D.M.S. system.
3. MONITORING RESULTS AND DISCUSSION

3.1 Contribute for the triggering thresholds study
Correlation between precipitations data and D.M.S.
displacement allowed to identify critical events that
have reactivated the landslide on the sliding surface at
7 meters blg, with direction 30 NE. During the
observation period it was possible to monitor in
continuous different kinematics and different weather
conditions. D.M.S. column allowed to investigate 5
triggering events and their relative period of stasis,
with a clear delay time after rain events or snow
melting. Weather data are been taken from database
of two climatic station in proximity of Bagnaschino site
Rainfalls start
Displacement start
Rainfalls type
Snow at ground
Temperature rise
Concomitant factors
Rainfall [mm]
Rainfalls duration [h]
Critical intensity [mm/h]
Total cumulative rainfall [mm]
Cumulative rainfall event [mm]
Total cumulative displacement
[mm]
Cumulative displacement event
[mm]
1st EVENT
12/12/2008 0.00
16/12/2008 6.00
Snow
Yes
Yes
(Viola and Pamparato). Following there are described

the characteristics of each event:
First event: on 28th November 2008 there was
the first snowfall (one of the most intense of
the last century in the area) that was followed
by some rainy days and finally by another
snowfall on 13-19th December. In the same
days there was a temperature rise that
caused the partial snow melting and
subsequently the first movement read by
D.M.S. column.
Second event: on 1st March 2009 there was a
light rainfall following by a strong temperature
rise (thermal zero at 1500 m asl) that caused
the second landslide activation on 2nd March
2009 at 20:03, 37 hours after the rainfall
beginning.
Third event: on 31st March 2009 at 06:00 a
strong rainfall began and lasted for some
days. After 30 hours, the landslide moved.
Fourth event: this event is linked to more
rainfalls occurred in the days 16th-22nd April
2009 and is different from the previous events
because of a lower movement velocity
(displacement about 10 mm).
Fifth event: on 26th April 2009 the strongest
spring rainfall started and after about 29
hours (27th April, 08:00) the landslide moved.
This heavy rainfall lasted for some days: the
cumulative displacement was 299.7 mm in
only two days. The roll axis on the involved
D.M.S. module reached its saturation angle
(tilt>20): the further displacement is
calculated with the interpolation of its pitch
axis, still active.
In the following table (table 1) and diagram (fig. 5)
each triggering event has been described in detail
considering also rain, cumulative rain, snow events
and temperature.
3rd EVENT
31/03/2009 6.00
01/04/2009 12.00
Rain
Yes
No
-
70
84
1.786
190
150
2nd EVENT
01/03/2009 6.00
02/03/2009 20.00
Rain/snow
Yes
Yes
Snow melting 120
mm
44
96
1.708
354
164
63
30
2.100
480
180
160
138
1.159
590
110
77.6
29
2.676
800
220
11.5
160.6
209.0
225.0
524.7
11.5
149.1
48.4
10.0
299.7
Snow melting 90 mm
4th EVENT
16/04/2009 6.00
22/04/2009 0.00
Rain
No
No
5th EVENT
26/04/2009 3.00
27/04/2009 8.00
Rain
No
No
Table 1. Rainfall thresholds events.

SafeLand - FP7
Page 331 of 401
Deliverable 4.1
DMS TORRE MONDOVI'
LEGEND
Cumulative displacement [mm]
Viola temperature [C]
Viola cumulative rain [mm]
Viola daily rain [mm]
Pamparato cumulative rain [mm]
Pamparato daily rain [mm]
Fourth triggering event

Fifth triggering event
22/04/2009 00:29
27/04/2009 00:29
Delay time 29h
25
100
800
20
15
300
10
600
400
5
200
80
60
Daily rain [mm]
Thermal zero
1500m asl
400
120
500
1000
30
Temperature [C]
600
Third triggering event

01/04/2009 22:36
Delay time 30h
First triggering event

16/12/2009 06:00
Second triggering event

02/03/2009 20:03
Displacement analysis - Ref. 5-8m bgl

10/11/2008 - 30/04/2009
40
0
200
100
20
-5
29-apr-09
19-apr-09
9-apr-09
30-mar-09
20-mar-09
10-mar-09
28-feb-09
18-feb-09
8-feb-09
29-gen-09
19-gen-09
9-gen-09
30-dic-08
20-dic-08
10-dic-08
30-nov-08
20-nov-08
10-nov-08
-10
Snow
Fig. 5. Displacement analysis with the individuation of 5 triggering events.
For each event it was calculated a particular value,

the critical intensity, that correspond to the ratio
between precipitation quantities (calculated in mm)
that caused triggering movement and its duration
(calculated in hours). The interpolated line in the
following bi-logarithmic plot (fig. 6) can be
considered a site-specific deterministic approach to
the limit equilibrium threshold that separates the
stability and instability field.
The relation that links precipitation intensity and

duration can be represented with the following
equation:
I = c + a*D b
where, in this case:
c=0
a = 9.1376
b = 0.3901
The next graph (fig. 7) shows the trend of threshold
elaborated, by means of real field monitored
displacement
data, on Bagnaschino landslide
compared to a review of empirical/statistical
thresholds resulted in slope failures, run out (Caine,
1980; Cancelli & Nova, 1985; Ceriani et al., 1994;
Crosta & Frattini, 2001, etc).
Fig. 6. Plot of intensity-duration thresholds.

SafeLand - FP7
Continuous monitoring of the landslide allowed to

notice weak deep creep in the interval 30-44 m blg
in addition to considerable shallow movements. The
activation of deep movements is delayed in respect
to shallow movements, with not unambiguous
behaviour.
Page 332 of 401
Case Study 15
Rev. No: 2
Date: 2012-02-15
3.3 Removal of D.M.S. column
BAGNASCHINO
Fig. 7. Comparison of Bagnaschino threshold with

literature empirical/statistic thresholds approach.
3.2
Relationship
displacement
between
rainfalls
On 13th July 2009 the D.M.S. column was removed.

The D.M.S. column allowed to obtained with
continuity the kinematics of the landslide in action,
not only limited to the initial stages of trigger, but
also during the evolution up to achievement of stasis
conditions.
The integrity of D.M.S. column is preserved in spite
of the displacement of 60 cm occurred at 7 m blg
(fig. 9-10). The excavation realized subsequently to
release the column confirmed depth, direction and
extent of the displacement, allowing the complete
recovery of the instrumentation and the repair of the
inclinometric pipe that is replaced and protected by
another pipe with a larger diameter.
amd
The following plot showing the relationship between

cumulative rainfall for each event and the relative
displacement.
In the graph we can notice two outlier points: its
rainfall value, in default of certain data, has been
calculated assuming the total contribute of snow
melting at ground previous the event..
Fig. 9. Removal of D.M.S. column.
Fig. 8. Plot showing relationship between rainfalls and

displacement (* these points are been obtained considering the
total snow melting at ground).
In rainfall interval included between 100 and 300

mm, data interpolation can be a line whose equation
is:
y = 2.6338x 280.76
To ensure the linear behaviour in the first part of the
curve rainfalls-displacements, it will be necessary to
individualize at least a significant triggering event
characterized only by rainfalls not snowy.

SafeLand - FP7
Fig. 10. The deformed module of D.M.S. column showing

the displacement of about 60 cm.
4. CONCLUSIONS
This case study shows the potential of the D.M.S.
system for landslides continuos monitoring. The
Bagnaschino landslide monitoring shows that
D.M.S. column allowed to investigate 5 triggering
events multilevel and their relative period of stasis.
Page 333 of 401
Deliverable 4.1
Moreover D.M.S. system allowed to obtain with

continuity the kinematics of the landslide in action,
(initial stages of triggering, evolution and
achievement of stasis conditions). The registration
of several triggering events in a short monitoring
period is a particularity of D.M.S. system. In fact, in
spite of a displacement of about 60 cm, the integrity
of D.M.S. column is preserved and the involved
modules are still active. Elevated displacements
influence the functionality of the traditional borehole
instrumentation after only a few centimetres.
REFERENCES
Baldesi, L. 2006. Studio cinematico dei dissesti ed
analisi dei fattori di innesco nella media ed alta
Val Nure (Piacenza); I casi di monitoraggio in
continuo di Farini e Casale-Colla. Universit
degli Studi di Milano-Bicocca, unpublished
thesis.
Cannon, S.H., Ellen, S.D., 1988. Rainfall that
resulted in abundant debris-flow activity during
the storm of January 35, 1982, in the San
Francisco
Bay
region,
California:
U.S.
Geological Survey Professional Paper 1434, 27
33.
Caine, N. 1980. The rainfall intensity-duration
control of shallow landslides and debris flows.
Geogr. Ann. A, vol. 62, pp. 23-27.

SafeLand - FP7
Page 334 of 401
Case Study 16
Rev. No: 2
Date: 2012-02-15
CASE STUDY 16:

Title: Detecting and mapping landslides in Salazie area, La Runion Island, using Remote
Sensing techniques.
Application: Ground surface displacement measurement.
Technique: Remote Sensing: radar interferometry, image correlation.
Main references: Delacourt et al., 2009; Cruchet, 1994.
Contributors: BRGM (Raucoules D., de Michele M.).
ABSTRACT
This case study shows the capabilities of a technique for landslide monitoring: Differential Monitoring of Stability
(D.M.S.) system.
The instrumentation is installed on Bagnaschino landslide that is located in the Municipality of Torre Mondov
(Cuneo, Piedmont) and is characterized by a composite movement. During the observation period (28th October
2008 13th July 2009) it was possible to monitor different kinematics and different weather conditions in
continuous. D.M.S. column allowed to investigate 5 triggering events and their relative period of stasis. In spite of
a displacement of approximately 60 cm that occurred at 7 m blg, the integrity of D.M.S. column was preserved.
Continuous monitoring of the landslide allowed to notice weak deep creep in the interval 30-44 m blg in addition to
considerable shallow movements. The system was efficient also with the presence of elevated displacements
unfitting with the functionality of the traditional borehole instrumentation.
1. INTRODUCTION
Due to its specific geological and climatic conditions,
landslides are a major hazard affecting La Runion
Island in the Indian Ocean (Cruchet 1994). Several
large landslides (up to several square kilometres)
severely affect infrastructures (especially roads) and
habitations mostly in the region of les Cirques. The

presented study (carried out in the framework of
MVterre-1 project, 2001-2002, funded by EU, Reunion
region and BRGM) focuses on Cirque de Salazie and
more precisely on the Hellbourg Landslide (figures 1
and 2).
Fig. 1. La Runion Island. The red frame locates the Salazie circle. The main GPS network is located.

SafeLand - FP7
Page 335 of 401
Deliverable 4.1
Presently, most of the techniques for monitoring

landslide
displacements
are
derived
from
measurements of reference stations. Conventional
geodetic techniques and extensometry techniques are
the most commonly used. GPS measurements are an
alternative (figure 1 shows the current GPS network
installed by the Regional Geological Survey of la
Runion on the study area). Moreover, due to spatial
and temporal heterogeneities of the displacements,
such ground based measurements are not sufficient to
fully describe the velocity field of a landslide.
Remote sensing imagery is a powerful tool for
landslide monitoring because it offers synoptic views
that can be obtained at different time intervals.
Differential SAR interferometry (DINSAR) of ERS (C
band) has shown its capability for deriving high
accuracy maps (at centimeter level Massonnet and
Feigl, 1998) of landslide displacement. However, this
technique is affected by severe geometrical and
environmental limitations. Due to the dense vegetation
cover of the study site, coherence of C-band (i.e.
Radarsat data) interferograms is poor: we therefore
assessed L-band JERS-1 data because the band is
more penetrative.
In the recent years, new techniques based on
correlation of optical images for the production of
deformation maps have been developed. Those
techniques have been successfully applied to the
measurement of coseismic deformation (Michel et al.,
1999) and more recently on landslides by combining
both aerial photographs and Quick-Bird Images
(Delacourt et al, 2004). For our test site, we used
aerial photographs and SPOT 5 optical images.
2. RESULTS
We have therefore tested two remote sensing
techniques to observe landslides in the La Runion
Island (Hellbourg area, Cirque de Salazie).
The INSAR analysis, carried out with a reduced set of
6 JERS-1 scenes, shows the interest of the L-band
SAR interferometry (figure 3). Both the shape and rate
of the Hellbourg landslide have been obtained.
Deformations larger than one fringe (i.e. about 11.5
cm) were observed. These results show the interest of
using L-band data in a context where C-band is not
relevant due to the vegetation cover.
Fig. 3. Sum of two 3 months JERS-1 interferograms (1 fringe

= 5.6 cm in line of sight) overlain on the topographic map of
the landslide. In red, GPS displacements for the period 19951999. Black lines represent positions of observed crack
boundaries of the landslide. Orange line shows the position of
the Hellbourg road severely affected by the landslide.
Fig. 2. Airborne photograph of the Hellbourg Landslide.

SafeLand - FP7
Correlation of optical data can be used for deriving

deformation maps of landslides that have a nearly
constant displacement rate over several years. This
technique can be applied either with two scenes
acquired at two periods by the same type of sensor, or
with different sensors (for example aerial photography
combined to very high resolution satellite imagery).
This technique allows the mapping of a large
displacement field. Our preliminary results in
correlating of SPOT 5 and aerial optical images have
permitted the detection of displacements of about 5m
over a period of five years (figure 4).
Page 336 of 401
Case Study 16
Rev. No: 2
Date: 2012-02-15
Fig. 4. a) Correlation Spot 5 and air-borne image (1997-2002) b) N-S displacement along profile A c) N-S displacement along
profile B.
Both the geometry and cinematic evolution of the two

landslides studied in the Hellbourg area by optical
correlation are compatible with the InSAR results.
3. PERSPECTIVES
In further works (to be carried out in the framework of
Mvterre-2 project, 2009-2013), we plan to use the
presented techniques to the whole Cirque de Salazie
and in particular Grand Ilet and Hellbourg landslides.
For this second step of the MvTerre Project we have a
particular interest for the data provided by new
sensors: Formosat (optical), TerraSAR-X and CosmoSkymed (X-band radar).
Although (due to the vegetated cover of the area) we
do not expect exploitable results with X-band
interferometry, the high resolution and repeatitivity of
those sensors should be suitable for multi-temporal
image correlation in a displacement monitoring
perspective.
The combination of results from the different sensors
should provide 3D displacements and their evolution
with time. Validation using the GPS network data and
interpretation will be carried out, in particular in order
to correlate the results with displacement acceleration
in relation with seasonal conditions.

SafeLand - FP7
REFERENCES
Cruchet, M., 1994, p. 57. Approche des problmes de
stabilit des pentes la Runion. Rapport
BRGM/RR-37962-FR.
Delacourt, C., Allemand, P., Casson, B., Vadon, H.,
2004, Velocity field of the "La Clapire" landslide
measured by the correlation of aerial and
QuickBird images, Geophys. Res. Lett, 31, p. 15.
Delacourt, C. , Raucoules, D. Moulic, S.L. Carnec,
C., Feurer, D., Allemand, P., Cruchet, M., 2009,
Observation of a large landslide on la reunion
island using differential sar interferometry (JERS
and Radarsat) and correlation of optical (Spot5 and
aerial) images, Sensors , Volume 9, Issue 1,
January 2009, Pages 616-630
Massonnet, D., Feigl, K., 1998, Radar interferometry
and its application to the changes of the Earths
surface, Review of Geophysics, 36 (4), pp 441500.
Michel, R., Avouac, J.-P., 2002, Deformation due to
the 17 August 1999 Izmit, Turkey, earthquake
measured from SPOT images, Journal of
Geophysical Research B: Solid Earth, 107 (4), pp.
2-1.
Page 337 of 401
Case Study 17
Rev. No: 2
Date: 2012-02-15
CASE STUDY 17
Title: Monitoring the Vallcebre landslide, Spain: from wire extensometers to GB-InSAR.
Application: Long term monitoring, characterization and modelling.
Technique: Surveying; Photogrammetry; GPS; Inclinometers; Wire-extensometers;
Piezometers; Rain-gages; DInSAR; GBSAR.
Main references: This case study is based mainly in Corominas et al., 2005.
Contributors: UPC (Gili, J.)
ABSTRACT
This case study presents the work carried out for monitoring the displacements of Vallcebre landslide (Eastern
Pyrenees, Spain). The landslide, which extends over an area of 0,8 Km2 and involves more than 20 million cubic
meters, has experienced displacements as large as 1 m per year in some points. It has been periodically
monitored since 1987, using a wide range of surface and in-hole techniques, successively along time:
triangulation with theodolite, Terrestrial Photogrammetry, Electronic Distance Measurement, GPS, inclinometers
and wire extensometers, piezometers, rain gages, satellite based DInSAR, and GBSAR. Somehow, the Vallcebre
monitoring is a showroom of the evolution along the last decades of the measuring technologies, which have been
applied in the landslide as a natural laboratory or as a demonstration site. After the cross-checking of the different
systems and the assessment of their advantages and drawbacks, we conclude that, even though wire
extensometers and inclinometers may have the highest precision, in practice, all the systems have their own role
in giving meaningful data for the monitoring. When possible, the complementary use of some of them is strongly
recommended..
1. INTRODUCTION
The Vallcebre landslide is located in the Eastern
Pyrenees, 140 Km north of Barcelona, Spain. The
situation, geological context of the landslide, and a
complete geomorphological description can be found
in Corominas et al. (2005). The mobilized material
consists of a set of shale, gypsum and claystone
layers gliding over a thick limestone bed. The average
slope of the whole landslide is about 10. The
movement affects an area of 0.8 km2, which shows
superficial cracking and distinct ground displacements.
classical surveying and photogrammetry, and using

GPS from 1995. In 1996, this site was included in the
frame of the NEWTECH Project funded by the
European Commission.
G1
G6
G18
G17
G8
G14
G3
G16
G4
B7
G5
Cal Bassenc
The landslide is a translational slide with a stair-shape

profile. Fig.1 shows a geomorphologic sketch of the
landslide and the location of the monitored points and
boreholes. The most active area is the lower unit, the
toe of which is being eroded continuously by the
Vallcebre torrent. Fig. 2 presents a geological crosssection of the landslide
Cal Monjo
B5
B3
T
EN
RR
TO
B6
G2
Cal Genet B13

G11
B12
B1
LL
G15
B8
AR
250m
LEGEND
SCARP
B10
G9
G121
UPSLOPE DIPPING
SURFACE
G12
TORRENT OF
VA
LL
CE
BR
E
G13
G131
G2
MONITORING POINT
AND DISPLACEMENT
B2
BOREHOLE
TORRENT
CRACKED HOUSE
GRABEN
As in most of real landslides, the structure and

behaviour is not simple. Very often, the measurement
of displacements is the simplest way to observe the
evolution of a landslide and to analyze either the
kinematics of the movement, the response to the
triggering conditions (i.e. rainfall) or the efficiency of
corrective measures. In certain situations, remedial
works are not feasible for budget, technical or social
constraints, and monitoring is one of the few mitigation
actions that can be adopted.
Several measuring techniques of these methods have
been used in Vallcebre since 1987, beginning with

SafeLand - FP7
DIRECTION OF
DISPLACEMENT
Fig. 1. Geomorphologic scheme and monitoring network of

Vallcebre landslide. Base points (E3, E4 and E5) are outside
this map, around 1 Km to the NW. Sample mean yearly rates
of displacement are shown (enlarged by a factor of 100).
Fourteen boreholes were drilled in the slope and

equipped with inclinometers, wire extensometers and
piezometers. Recently, Radar Interferometry is being
tested in the Vallcebre natural Lab.
Page 339 of 401
Deliverable 4.1
Fig. 2. Geological cross-section showing B1 to B9 instrumented boreholes. The materials are: (1) Limestone (bedrock), (2)
Laminated clay siltstone, (3) Clay siltstone with gypsum micro-modules, (4) Gravel with silty matrix (superficial formation).
2. MONITORING SYSTEMS USED IN

VALLCEBRE
In this section, a general overview of the monitoring
techniques that have been used consecutively along
the years in Vallcebre is described.
2.1 Terrestrial photogrammetry
The first monitoring network established on the
landslide was based on terrestrial (or close-range)
photogrammetry. A total of seven campaigns were
performed at the landslide foot between 1987 and
1992.
Stereopairs were taken with a Wild P32 metric camera
(fig.3). The ground control surveys were carried out by
triangulation with Wild T2 theodolite, with an expected
error of 2 mm. The stereocompilation was done with
an analytical stereo-plotter Wild AC1 at the Institut
Cartogrfic de Catalunya. The residuals after the least
squares adjustment were around 9 mm (RMS), giving
an error-to-distance ratio of 9 mm / 45 m (1/5000) for
well defined points (as clear as pointing targets).
additional points in the middle unit were monitored with

single distance variation measurements (fig. 4 b).
During 1994 and 1995, a triangulateration (fig. 4 c)
from new base points E1 and E2 was extended to 16
points spread out through the whole landslide. The
angle measurements were carried out with a Wild T2
theodolite, and the EDM with a Wild DIOR 3002S.
During the period 1987-1995, displacements as large
as 8 m were observed at the lower unit, decreasing
uphill. The rate of movement was strongly dependent
on the rainfall. Rates of about 4 m per year were
observed at certain points near the toe in the rainy
years, while almost no displacement occurred during
periods of drought. In the upper landslide units, the
rate of displacement was significantly smaller, in the
range of 10 to 30 cm/year. This stage allowed the
identification of the most active sectors within the
landslide and a rough assessment of the relation
between rainfall and surface displacements was
envisaged (Fig. 5).
The results for each survey were the change in the

coordinates of main points (displacements) and maps
with contour lines. In general, the application of this
method is not straight forward because of the difficulty
to found the proper set-up with an adequate view over
the hill slope. Also, dedicated and costly equipment
has to be available. Regardless of that, the terrestrial
photogrammetry provides information continuous over
the space, but very scarce along time.
2.2
Triangulation and EDM
Since 1987, geodetic measurements with theodolite

and EDM (Electronic Distance Measurement) were
carried out. Since 1987 until 1992, triangulation was
undertaken in the same area than the photogrammetry
(lower unit, fig. 4 a). In the period 1988-1994, three

SafeLand - FP7
Fig. 3. Wild P32 camera for terrestrial photogrammetry over a

theodolite.
Page 340 of 401
Case Study 17
2.3
Rev. No: 2
Date: 2012-02-15
The GPS surveys
The GPS technology has been used since the early

1990s to monitor structural and terrain deformations
(see i.e. Lovse et al., 1995). An overview of the
applications of GPS to structural and terrain monitoring
can be found in Gili et al. (2000).
Theoretically, GPS a priori errors have to be more
balanced in the three axes, when compared to
theodolite and EDM ellipses of error. This point was
confirmed by the results of several monitoring
campaigns, and led us to check the feasibility of using
GPS techniques to perform systematic monitoring of
the Vallcebre landslide. GPS preliminary tests were
made in July 1994. Five baselines, from 800 to 1500
m, were observed with two Topcon Geodetic
Receivers GP-R1, following the Static method. The
results were promising due to easy-to-use, precision
and matching with surveying measurements. Also, a
movement simulation test was made in July 1994.
A displacement of 10 mm North was imposed to the
antenna, and the static observation repeated. The new
computations gave a movement vector of {E= 0
mm, N= 16 mm}.
Fig. 5. First Relationship between accumulated rainfall (mm,

solid line) and displacements (mm, right scale, line with dots).
The observed discrepancy (0 mm, 6 mm) falls within

the expected error.
In December 1995, a complete EDM and GPS survey
was carried out, to link the measurements taken with
classical methods with the first campaign with GPS.
Two Trimble 4000 SSi dual frequency receivers were
used (Fig. 6).
Fig. 6. Examples of Vallcebre monitoring points with different

antenna set-up methods: telescopic pole (left); tripod with
optical plummet over an inclinometer borehole (right).
Fig. 4. Triangulation, single

triangulateration schemes.

SafeLand - FP7
distance
variation
and
New base points were established (E3, E4 and E5,

engraved directly on stable surface of the limestone
layer) for stability and visibility requirements, and most
of the old targets were recovered with minor
modifications. As new points have been added since
1996, the monitoring network has now 40 points (Fig.
1 and 7): most of them are points engraved in rock
blocks outcropping in the hillside or the top of the
casing of the inclinometric boreholes; several steel
rods, stakes and poles have been used as well. Four
additional points are on the limestone around the
sliding zone. These points were the fixed points used
to check the GPS accuracy. This network allowed both
the measurement of displacements and the
comparison with movements obtained with the
borehole
equipment
(inclinometers
and
wire
extensometers).
Page 341 of 401
Deliverable 4.1
G1
2.4 In-hole measurements
G18
G17
G14
Between July 1996 and April 1998, 16 boreholes were

drilled in the slope, and equipped with inclinometers,
wire extensometers and open standpipe piezometers.
In 2004-2005 multipoint piezometers were installed in
three additional boreholes.
G8
G16
G5
B6
G2
B5
G3
G4 B7
B16
B3
B12
B1
B8
B13
G11
G15
B10
G9
G121
G12
G13
G131
E3
500m
Fig. 7. Part of Vallcebre landslide GPS monitoring network:

solid and dashed lines are respectively GPS-FS and GPSRTK baselines during a given campaign.
Inclinometers and piezometers were standard devices.

As an example, Fig. 9 shows a typical record from
inclinometer B7, where the slip surface is clearly
marked. Measurements were made every 2-3 weeks
until they went out of order. On the other hand, a wire
extensometer equipment was especially built following
a previous design of Angeli et al. (1988). It consists of
a protected steel wire anchored to the bedrock, below
the slip surface, inside a piezometric pipe (fig. 10). The
wire is kept in tension by means of a pulley and a
counterweight of which rotation is continuously
recorded using a potentiometer. This system has
proven to work properly with landslide displacements
much larger than those required to broken
inclinometric pipes. Full details of the wire
extensometer and the correction of the reading of the
wire to derive the landslide displacement can be found
in Corominas et al. (2000).
Fig. 8. Partial view of the GPS monitoring in Vallcebre.

Displacement in plan from the initial position. A maximum
error fringe (horizontal dotted line) has been added to
highlight the tolerance threshold. Target numbers correspond
to those indicated in figure 1.
The GPS methods used are either static (Fast-Static

or Quick-Static) or kinematic (Real Time Kinematic),
depending on the sky visibility of the points (the FS
method is more robust against temporary signal
interruption).
24 GPS campaigns have been carried out since 1995,
with Trimble (4000 SSi) and Topcon (Hyper Pro)
receivers. Part of the results can be appreciated in fig.
8.

SafeLand - FP7
Fig. 9. Horizontal displacement profile obtained from

inclinometer B7.
A main advantage is that the piezometers and wire

extensometers were installed with an automatic
recording system.
Page 342 of 401
Case Study 17
Rev. No: 2
Date: 2012-02-15
monitoring, although some previous experiences

exists (i.e. Squarzoni et al., 2003; Tarchi et al., 2003).
DInSAR: In November 2006 we installed seven corner
reflectors to enhance the performance of DInSAR
techniques (figure 12) with SAR data of the ASAR
sensor within the ESA Envisat satellite mission. The
cross-checking of the DInSAR results versus the
displacements derived from the wire extensometers
has been carried out along 2007 and 2008, and some
preliminary results have been obtained (Gili et al.,
2009).
Fig. 12. The DInSAR technique (left, Crosetto et al, 2009) has
been used to compute the displacements of several corner
reflectors (right) since November 2006.
Fig. 10. Sketch of the borehole wire extensometer
(Corominas et al., 2000). In the picture, the black cable for
the piezometer can also be appreciated.
GBSAR first surveys: The Institute of Geomatics owns

a GBSAR (Ground Based Synthetic Aperture Radar)
system (IBIS-L model from Ingegneria Dei Sistemi
IDS; working in the Ku band, 17.1 GHz). Although the
2009/2010 wintertime has been quite crude, two
preliminary GBSAR campaigns (February 9th and
April) have been carried out in Vallcebre (figure 13), in
order to start the GBSAR monitoring. This will help the
SAFELAND project to assess the advantages and
drawbacks of this technique as landslide monitoring
and early warning system.
Fig. 11. Wire extensometer measurements at borehole B2.

Each minor tick in the time axis corresponds to 1 week.
The readings were taken every 20 min and stored until

they were downloaded: each month, manually, in the
beginning; by GSM data transmission nowadays.
Without a continuous measurement (fig. 11), crucial
information would be lost near the concentrated rainfall
periods characteristics of the Mediterranean climate. In
Vallcebre it has been possible to establish that there is
a perfect synchronism between changes in
displacements and the groundwater level inside the
slope.
2.5 Radar Interferometry
We are in the firsts steps for the application of both the
DInSAR and the GBSAR techniques to landslide

SafeLand - FP7
Fig. 13. The IBIS-LIDS GBSAR system installed in front of

the Vallcebre landslide by the Institute of Geomatics for the
first tests. A number of mini-corners (right) were deployed in
the scene.
3. SUMMARY AND CONCLUSION

It is out of the scope of this extended abstract to
present all the cross-checks and precision evaluation
for the different methods installed in Vallcebre. Some
of these analyses can be found in the references. Only
a few conclusions will be highlighted here.
Page 343 of 401
Deliverable 4.1
The GPS measurements show better trends and

stability than the EDM, at least with the Vallcebre
conditions and equipment. The expected error (1)
established for a single GPS Fast Static positioning is
12 mm in plan and 18 mm in elevation.
The precision achieved with the GPS set-up in
Vallcebre, although enough for the monitoring goals,
may be increased using improved setting up of the
antennas over the benchmarks and/or extended
observation times.
On the other hand, the inclinometric measurements
are precise within the 1-2mm, but the readings were
only possible until 5 to 20 cm of total displacement for
the top of the boreholes, due to the casing deformation
in the failure zone. For the extensometers, the results
show an excellent match with the inclinometers and
the GPS (fig. 11). Compared with the GPS, the wire
extensometer technique has the advantage of being
an automatic and continuous measuring system which
can be correlated with rainfall and piezometric heads.
But the installation of in-hole devices had been
possible only when additional funding was obtained in
order to drill the boreholes and to instrument them. All
the piezometers and the wire extensometers installed
in Vallcebre are still operational. The wire cumulative
displacements are of several meters since the
installation in 1996. With readings each 20 minutes,
especially during critical rainy events when other
systems are not in operation, it is a very valuable data
for the correlation with the rain and the water pressure
inside the slope.
Regarding the SAR interferometry first results: the
DInSAR computations in the 7 corner reflectors have
exhibited a nice relationship with the wire
extensometer measurements. The terrestrial GBSAR
is only in the first phase and it needs some additional
campaigns to establish their strengths and
weaknesses.
Corominas, J., Moya, J., Lloret, A., Gili, J.A., Angeli,

M.G., Pasuto, A. and Silvano, S. (2000).
Measurement of landslide displacements using a
wire extensometer. Engineering Geology 55, pp.
149-166.
Corominas, J., Moya, J., Ledesma, J., Lloret, J., Gili,
J.A. (2005). "Prediction of ground displacements
and velocities from groundwater level changes at
the Vallcebre landslide (Eastern Pyrenees, Spain)".
Landslides 2, pp. 8396.
Crosetto M., Monserrat O., Pozzoli A. & Gili J.A.
(2009). Deteccin y medida de deformaciones del
terreno utilizando interferometra diferencial SAR
(in Spanish). Proceed. VII Simposio Nacional
sobre Taludes y Laderas Inestables, Barcelona,
oct. 2009.
Gili, J.A.; Corominas, J., Rius, J. (2000). Using Global
Positioning System Techniques in landslide
monitoring. Engineering Geology 55, pp. 167-192.
Gili, J.A., Crosetto, M. & Serral, D. (2009).
Reflectores Radar en el deslizamiento de
Vallcebre. Validacin y primeros resultados (in
Spanish). Proceed. VII Simposio Nacional sobre
Taludes y Laderas Inestables, Barcelona, oct.
2009.
Lovse, J.W., Teskey, W.F., Lachapelle, G., Cannon,
M.E. (1995). "Dynamic Deformation Monitoring of
Tall Structure Using GPS Technology". Journal of
Surveying Engineering - ASCE. Vol. 121, No. 1,
pp. 35-40.
Squarzoni, C., C. Delacourt, P. Allemand (2003). Nine
years of spatial and temporal evolution of the La
Valette landslide observed by SAR interferometry.
Engineering Geology, Vol. 68, 1-2, pp. 53-66.
Tarchi, D, N. Casagli, R. Fanti, D.D. Leva, G. Luci, A.
Pasuto (2003). Landslide monitoring by using
ground-based SAR interferometry: an example of
application to the Tessina landslide in Italy.
Engineering Geology, Vol. 68, 1-2, pp. 15-30.
Rates of displacement of the Vallcebre landslide were

up to 130 mm/month at the lower unit during rainy
periods, with an average of 30 mm/month. These
figures fall to zero during the drought periods.
All the results are in fairly good accordance when
adequate corrections and error considerations are
done. The redundancy between methods is very
advisable to detect malfunctions of the devices or to
cover gaps with no data. The different methods should
not be considered as excluding alternatives but as
complementary systems to be applied progressively or
in different areas.
REFERENCES
Angeli, M.G., Gasparetto, P., Silvano, S., Tonetti, G.
(1988). An automatic recording system to detect
the critical stability of slopes. Proc. 5th Int.
Symposium on Landslides, Lausanne, Vol. 1. A.A.
Balkema, Rotterdam, pp.375-378.

SafeLand - FP7
Page 344 of 401
4. Discussion
Rev. No: 2
Date: 2012-02-15
DISCUSSION

SafeLand - FP7
Page 345 of 401
4. Discussion
Rev. No: 2
Date: 2012-02-15
4. DISCUSSION
4.1. ADVANTAGES AND LIMITATIONS OF EACH METHOD
The main advantages and limitations of each of the techniques described in previous chapters
are discussed in this section. These characteristics are basically mentioned in terms of spatial
and temporal resolution, accuracy, spatial coverage, cost, etc. This chapter is intended to be a
link or an introductory section to the deliverable 4.4 (Guidelines for the selection of
appropriate remote sensing technologies for monitoring different types of landslides) of the
SafeLand project.
Table 24 : Advantages and limitations of passive optical sensors.
Airphotos
Ground based
imaging
Advantages
Limitations
- High spatial and temporal resolution (on demand,

e.g. 50 pixels per m2 at 1 picture per day)
- Limited coverage in gentle slopes
- Large coverage on steep slopes
- Inter-visibility is required (optical line of sight)
- Images obtained at a low cost
- Others: Solar illumination affects datasets, e.g.

presence of shadows in the pictures; data
acquisition is not possible during night.
- High flexibility (i.e. easy data acquisition and high

portability)
- Very large coverage on gentle slopes
- Limited temporal resolution
- Historical record available: Images since the

middle of the XX century
- Intervisibility is required (optical line of sight)
- Very high spatial resolution
- Others: reduced coverage in steep slopes and cliffs

(steep slopes are misrepresented)
Satellite imaging
- Low spatial resolution of historical images (pixels

at meter or decameter level)
- Very large spatial coverage (up to several thousand

km2)
- Cost: low resolution images (e.g. landsat, ASTER)
can be obtained at a low price
- Short revisit times with very high resolution
pointable sensors
- Others: Historical record available: Images since
70s; Panchromatic / multispectral images are
frequently used
- In some sensors, low temporal resolution

(monthly acquisition)
- Low accuracy (depending on the sensor)
- Intervisibility is required (optical line of sight)
- Others: low coverage in steep slopes and cliffs
(i.e., steep slopes are misrepresented); Post
processing is needed (filtering)
- Image quality depends on atmospheric conditions
- Very high resolution images from commercial
satellites (e.g. QuickBird, IKONOS, GeoEye) are
very expensive
- Geometric distortions due to different view angles
of pointable sensors.

SafeLand - FP7
Page 347 of 401
Deliverable 4.1
Table 25 : Advantages and limitations of active optical sensors.
Advantages
Limitations
Electronic
Distance Meter
- High resolution (e.g. 200 point per m2).
Aerial Laser
Scanners
(ALS)
- Temporal and spatial resolution on demand
Terrestrial Laser
Scanners
- High accuracy (millimeter level).
- Considerable maximum range (up to several km)

- Others: 3D information; High flexibility (i.e., easy
set-up and portability); Feasibility of automation of
the process (e.g. using a robotic total station);
suitable for early warning systems
- Good accuracy (centimeter level)

- Large coverage on steep slopes.
- Others: 3D information; High flexibility (i.e., easy
set-up and portability)
- High resolution in gentle surfaces
- Very large coverage (thousand km2)
- Others: 3D information; fast data acquisition;
considerable maximum range (few thousand km)
- Low resolution (singular measurements do not

provide a complete image of the object)
- Others: Intervisibility is required (optical line of
sight)
Relatively
(< 700m)
low
maximum
range
- Others: Post processing is needed (aligning,

filtering, etc.); intervisibility is required (optical
line of sight)
- Centimeter to decimeter accuracy level
- Post processing is needed (filtering)
- Others: Post processing is needed (aligning,
filtering, etc.); reduced coverage in steep slopes
(cliffs are misrepresented)
Table 26 : Advantages and limitations of active microwave sensors.
Ground Based InSAR
Interferometric
Radar Distance
Meters
Advantages
Limitations
- Very high accuracy: sub millimeter level.

- High range (up to 5km)
- High temporal resolution (on demand, up to 1
measurements each 10 sec.)
- Others: not affected by atmospheric artifacts;
- Very high temporal resolution (on demand) when
using a fixed GB-SAR (e.g. 1 measurement each
10min), allowing for real time monitoring
- Very high accuracy (millimeter level)
- Range up to 3 km
- High spatial resolution (pixels at a meter level);
Ability to monitor movements over a long period;

SafeLand - FP7
- Low spatial resolution (singular measurements do

not provide a complete image of the object)
- Others: a heavy setup is needed in case of real
time monitoring
- Others: Loss of coherence in vegetated areas;

Atmospheric artifacts are possible; Post processing
is needed. A heavy setup is needed in case of real
time monitoring
Page 348 of 401
4. Discussion
Rev. No: 2
Date: 2012-02-15
- Low spatial resolution, specially in areas with

steep slopes or presence of vegetation
DInSAR
- Large coverage
- Others: Historical record: images available since
1991; Ability to monitor movements over a long
period; Only 2 timages are needed; No cost
installation and maintenance of equipment on the
ground
- Low temporal resolution (monthly values). Not

suitable for fast movements
- Others: Loss of coherence in vegetated areas and
when dealing with fast movements; Atmospheric
artifacts are possible; Post processing is needed; A
Digital elevation model is needed; Time
decorrelation problems
Advanced DInSAR
- Low spatial resolution, specially in areas with

steep slopes or presence of vegetation
- Application to different scales of study
- High accuracy in the measurement of the
displacements (mm level).
- Others: Historical record: images available since
1991; Ability to monitor movements over a long
period; No cost installation and maintenance of
equipment on the ground
- Low temporal resolution (monthly values). Not

suitable for fast movements
- Others: Requires the presence of fixed points on
several sequences of radar images; Landslide
displacements are measured parallel to the satellite
Line Of Sight; Loss of coherence in vegetated
areas; Atmospheric artifacts are possible; Post
processing is needed; The cost can be substantial
when dealing with high resolution datasets;
Requires a large set of images
Table 27 : Advantages and limitations of geophysical investigations.
Advantages
- Higher resolution than aerial approaches.
- Cheaper than boreholes and most methods are nonintrusive.
Ground-Based
- Derives from the surface information about

subsurface structures and provides either curves
(1D), vertical sections (2D), or cubes (3D) of the
mapped parameters.
- Can gather several types of parameters (elastic,
electrical, electromagnetic, etc).
- Wide range of portable and light instruments
nowadays and several standardized software.
- Should be used to plan location of boreholes for
ground proofing (cost reduction).
- Should be used to interpolate structures between
boreholes for 2D-3D mapping and in order to catch
the lateral variability of the ground.
- Hydrogeophysics is in full expansion and
geophysics can help understand the hydrology of a
site.
- Can be used for monitoring.

SafeLand - FP7
Limitations
- May be difficult to relate measured parameters
(elastic, electrical, etc) to soil/rock properties.
- Measurements should always be calibrated with
borehole testing and geophysical logging.
- Often non-uniqueness of the results which should
always be provided with errors and modeling
controls. Not sufficient to produce nice color
maps and state that one color is one type of soil: not
that simple.
- Some approaches are still heavy to carry out and
may be challenging on difficult ground (e.g., some
landslide sites).
- Still heavy equipment and requires electrical
power (batteries, etc).
- Application for site monitoring is still a subject of
research for several methods, though some are quite
standard already (e.g., micro-seismics).
- Geophysical results are difficult to interpret if not
done in tight cooperation with other geoscientists
(geology, geotechnique, hydrogeology, glaciology,
etc).
Page 349 of 401
Aerial
Deliverable 4.1
- Wide coverage, several parameters recorded at the

same time
- Only remote sensing technique that can look into
the subsurface (not just surface)
- Application for site monitoring is still a subject of

research
Table 28 : Advantages and limitations of geotechnical methods.
Extensometer
Advantages
Limitations
Probe:
Probe:
- Low cost survey
- Measurements are less precise than using a fixed

borehole;
Fixed borehole:
- A borehole with casing is required
- High accuracy
Fixed borehole:
Wire:
- High accuracy
- Others: simple device; real time monitoring is
possible; low-cost device
Wire:
Manual probe:
- Post processing is needed, correction of
systematic errors
- No continuous recording of the displacement
Inclinometer
Manual probe:
- Low cost survey
- Access limited due to a localized deformation of

the casing of only a few centimeters
In-place:
- A borehole with 4 grooves casing is required
- Rapid automatic data

monitoring is possible
logging;
real
time
- Remote control
Automatic probe:
- High accuracy
- Remote control; real time monitoring is possible
In-place:
- Great complexity; Cost of the instrumentation
- Long term stability
Automatic Probe:
- Costs of the instrumentation
- Access limited due to a localized deformation of
the casing of only a few centimeters

SafeLand - FP7
Page 350 of 401
4. Discussion
Rev. No: 2
Date: 2012-02-15
Observation well:
- Can be more economic
Open standpipe:
- High accuracy
Observation well:
- Reliability of collected data
- Single layer aquifer
- Long-term monitoring
- Time-lag between the variation of piezometric

level in the aquifer and the respective variation in
the piezometer
- Can be used to evaluate the hydraulic conductivity

- Can be converted to diaphragm piezometer
Twin-tube hydraulic:
Piezometer
- High accuracy

Open standpipe:
- Inaccessible components have no moving parts
- Device subject to damage during the installation

and vertical compression
- Reliable
- A borehole is required (+standpipe)
- Long successful performance record
Twin-tube hydraulic:
- Piezometer cavity can be flushed
- Application generally limited to long-term

monitoring of pore water pressure in embankments
dams
- Can be used to measure permeability

Pneumatic:
- High accuracy
- Tubing must not be significantly above minimum

piezometric elevation
Vibrating wire:
- Periodic flushing may be required
- Easy readings
- A borehole is required (+twin tubes)
- Short time lag
Pneumatic:
- No freezing problems
- Membrane extroversion reduces volume available

to interstitial water (pore water pressure increases)
- Independence of reading elevation from the height

where is placed the piezometer
- A borehole is required (+twin tubes)
- Possibility to measure negative pore water

pressure
Vibrating wire:
Electrical resistance:
- Zero drift
- Easy readings
- A borehole is required
- Short time lag
Electrical resistance:
- No freezing problems
- Wire subject to corrosion phenomena
- Independence of reading elevation from the height

where is placed the piezometer
Earth
pressure
cells
- Possibility to measure negative pore water

pressure
- High accuracy
- Direct measurement

- Not post processing is needed.

SafeLand - FP7
Page 351 of 401
Multiparametric in place
systems (DMS)
Deliverable 4.1
- Multiparametric continuous monitoring in the

same casing/borehole
- 2D/3D deformation analysis
- Early Warning applications
- High adaptability of the monitoring column to
larger deformations
- Easy installation (set-up);
- Cost of the instrumentation (in particular 3D

versions of the DMS)
- The instrumentation can be recover after the

monitoring period
- Modularity
Core
Logging
GNSS
Table 29 : Advantages and limitations of other techniques.
Advantages
Limitations
- High positioning accuracy is possible (mm level)

with precision GNSS

- Spatial and temporal resolution on demand
- Others: low signal when data acquisition

performed in narrow valleys or close to buildings,
trees, etc.; More elaborated signal post processing
is required for millimeter level accuracy
- Others: Intervisibility is not required.; low cost;

easy data acquisition; suitable for early warning
systems
- Direct observation of subsurface materials
- Direct observation of shear surface; allows for
geological and geotechnical investigations (see
section 5.1.5)

- Very high cost
4.2. APPROPRIATE TECHNIQUES ACCORDING TO LANDSLIDE SITUATION

Landslides occur within different soil and rock contexts, destabilized by different internal
settings, triggered by different external factors including human activities (Terzaghi, 1950;
Jaboyedoff & Derron, 2005; and more generally the Deliverable 1.1 of the SafeLand project).
With the case studies presented in this document, the reader will get an idea of the effective
implementation, the results that can be obtained and the reliability of these techniques in
specific situations. In addition to this document, Deliverable 4.4 of the SafeLand project
(Guidelines for the selection of appropriate remote sensing technologies for monitoring
different types of landslides) is specifically dedicated to this topic.

SafeLand - FP7
Page 352 of 401
4. Discussion
Rev. No: 2
Date: 2012-02-15
4.3. INNOVATIVE TECHNIQUES AND FURTHER DEVELOPMENTS

Researchers always aim to increase the level of the knowledge to overcome the actual
technical limitations, either by developing new (1) algorithms, (2) softwares or (3)
instrumentations. One or two examples of innovative techniques for each of these categories
are presented below.
The innovative applications (e.g. methods, algorithms, hardware or software) will be
discussed in detail in Deliverable 4.5 of the SafeLand project (Evaluation report on
innovative monitoring and remote sensing methods and future technology).
(a) Algorithm developments

An important challenge in DInSAR for landslide studies is to increase the number of
permanent scatterers or targets in non-urban areas, and much of the research is focused on
development of algorithms that can process and interpret the increased volume of data more
efficiently. For example, the development of the SqueeSAR algorithm by T.R.E. (Novali et
al., 2009) allows to drastically increase the density of points in mountainous areas (Figure
157), keeping a very high spatial resolution.
Figure 157: Localization of scatterers, using the SqueeSAR algorithm (from Navali et al., 2009). The high
density of points allows for a better delimitation and interpretation of SAR data for landslide investigations in
non-urban areas (T.R.E. website, 2010. Available at http://www.treuropa.com/treuropa/tabid/178/Default.aspx).
(b) Software developments

Softwares for radar images and interferometry processing have also experienced rapid
development during the last decade. Current efforts are focused on creating more userfriendly softwares, allowing a democratization of methods by non-specialists. For instance,
ESA in collaboration with the Delft University of Technology is currently developing a
complete toolbox for radar images and interferometry processing, which will soon be
available for free, NEST-DORIS (Figure 158, Marinkovic et al., 2009).

SafeLand - FP7
Page 353 of 401
Deliverable 4.1
Figure 158 : Early version of the software NEST-DORIS, demonstrated during the Fringe 2009 (Marinkovic et
al., 2009)
Another example is the new software Coltop 3D, developed at the IGAR (University of
Lausanne) for post-processing of TLS datasets and now available in its commercial version
(contact: http://www.terranum.ch). It allows one to work on huge dataset (up to 200 million of
points) and to perform structural analysis, providing discontinuities identification, dips
measurement and color coding of rock face (Jaboyedoff et al., 2007).
Figure 159 : Illustation of a TLS dataset imported in Coltop 3D (left) and the field corresponding (right)
(Carrea et al., 2010)

SafeLand - FP7
Page 354 of 401
4. Discussion
Rev. No: 2
Date: 2012-02-15
(c) Instrumental developments

Hardware and tools for remote sensing and monitoring are continously being improved. For
example, a new TLS developed by Optech, the LYNX Mobile Mapper, allows scanning
transportation corridors from a car by simply scanning along the road (Figure 160 and Figure
161). It can be very useful to assess the rockfall susceptibility along transportation corridors,
saving a lot of time for the point cloud acquisition and post-processing (Carrea et al., 2010).
Figure 160 : Up. Point cloud acquired thanks to LYNX Mobile Mapper, highlighting main discontinuity sets.
Low. Corresponding outcrop on field. (Carrea et al., 2010)
Figure 161 : LYNX Mobile Mapper apparel on the roof of the car. (Optech website, 2010. Available at
http://www.optech.ca/press_LYNX_Release.htm)

SafeLand - FP7
Page 355 of 401
5. Conclusions
Rev. No: 2
Date: 2012-02-15
CONCLUSIONS

SafeLand - FP7
Page 357 of 401
5. Conclusions
Rev. No: 2
Date: 2012-02-15
5. CONCLUSIONS
Since the 90s, innovative applications in monitoring and remote sensing techniques have
been developed. The possibility of acquiring terrain information (height, displacement, land
use, etc.) with high accuracy and high spatial resolution is currently opening up new ways of
visualizing, modelling and interpreting Earth surface processes such as landslides, debris
flows, rockfalls, etc. These new sensors can be mounted on terrestrial, aerial and/or satellite
instrumentals, covering a full spectra of accuracies, resolutions, points of view, etc.
Geophysical and geotechnical investigations can also bring additional information on subsurface processes and movements, which are essential for monitoring and early-warning
systems.
The main part of this deliverable provides a comprehensive review of the state of the art for
different techniques (remote sensing and ground-based) used for landslide detection, fast
characterization, rapid mapping and long-term monitoring. Together with the technical
information, an effort was made to provide examples of applications of each technique to
specific landslide case studies, as described in chapter 3.
This Deliverable attemps to provide useful information not only to researchers, but also to
stakeholders who look for a better understanding of important questions related to landslide
hazard and risk, namely where, how and when a landslide will occur?

SafeLand - FP7
Page 359 of 401
6. References
Rev. No: 2
Date: 2012-02-15
REFERENCES

SafeLand - FP7
Page 361 of 401
6. References
Rev. No: 2
Date: 2012-02-15
6. REFERENCES
6.1. GENERAL REFERENCES WITHIN THE DELIVERABLE
Carrea, D., Longchamp, C., Jaboyedoff, M., Choffet, M., Derron, M.-H., Michoud, C., Pedrazzini, A., Conforti,
D., Leslar, M., Tompkinson, W., 2010. Fast rockfall hazard assessment along a road section using the new
LYNX Mobile Mapper Lidar. EGU General Assembly, EGU2010-5178.
Jaboyedoff, M. & Derron, M.-H. 2005. Hazard assessment within an Integrated Risk Assessment Process for
Landslides (IRAPL). Proc., International Conference on Landslide Risk Management, Vancouver, Balkema.
in press.
Jaboyedoff M., Metzger R., Oppikofer T., Couture R., Derron M.-H., Locat J. Turmel D., 2007. New insight
techniques to analyze rock-slope relief using DEM and 3D-imaging cloud points: COLTOP-3D software. In
Eberhardt, E., Stead, D and Morrison T. (Eds.): Rock mechanics: Meeting Societys Challenges and
demands (Vol. 1), Taylor & Francis. pp. 61-68.
Marinkovic, P., Hanssen, R., Minchella, A., Engdahl, M., 2009. NEST-DORIS: InSAR extension for NEST. In:
Fringe 2009 Workshop. Advances in the Science and Applications of SAR Interferometry. Organized by the
European Space Agency. ESA ESRIN Frascati (Italia), 30th November 4th December 2009.
Novali, F., Ferretti, A., Fumagalli, A., Rocca, F., Prati, C., Rucci, A., 2009. The Second Generation PSInSAR
approach: SqueeSAR. In: Fringe 2009 Workshop. Advances in the Science and Applications of SAR
Interferometry. Organized by the European Space Agency. ESA ESRIN Frascati (Italia), 30th November 4th
December 2009.
Optech website, 2010. Available at http://www.optech.ca/press_LYNX_Release.htm
SafeLand deliverable 1.1, 2010. Landslide triggering mechanisms in Europe Overview and State of the Art.
Deliverable 1.1 of the European project SafeLand. Edited in 2010 by Crosta G., Agliardi F., Frattini P.,
Sosio R.
Terzaghi, K., 1950. Mechanism of Landslides. The Geological Society of America, Engineering Geology
(Berkley) Volume. Pp 83-123
T.R.E. website, 2010. Available at http://www.treuropa.com /treuropa/tabid/178/Default.aspx
6.2. SPECIFIC REFERENCES FOR THE CHAPTER 2A

Aizu, Y., and T. Asakura (1987), Principles and development of spatial filtering velocimetry, Applied Physics B:
Lasers and Optics, 43(4), 209-224.
Ando, S. (1986), A velocity vector field measurement system based on a spatio-temporal image derivative,
Transactions of the Society of Instrument and Control Engineers, 22(12), 1330-1336.
Arattano, M., and P. Grattoni (2000), Using a fixed video camera to measure debris flow surface velocity, in
Debris-Flow Hazards Mitigation: Mechanics, Prediction, and Assessment, edited by G. F. Wieczorek and N.
D. Naeser, pp. 273-281, Balkema., Rotterdam.
Arattano, M., and L. Marchi (2000), Video-derived velocity disribution along a debris flow surge, Physics and
Chemistry of the Earth, Part B: Hydrology, Oceans and Atmosphere, 25(9), 781-784.
ASTER-GDEM-VALIDATION-TEAM (2009), ASTER Global DEM Validation Summary ReportRep., 28 pp,
METI/ERSDAC NASA/LPDAAC USGS/EROS.
Avrithis, Y. S., A. D. Doulamis, N. D. Doulamis, and S. D. Kollias (1999), A Stochastic Framework for Optimal
Key Frame Extraction from MPEG Video Databases, Comput Vis Image Und, 75(1-2), 3-24.
Baatz, M., C. Hoffmann, and G. Willhauck (2008), Progressing from object-based to object-oriented image
analysis, edited, pp. 29-42.
Baldi, P., M. Fabris, M. Marsella, and R. Monticelli (2005), Monitoring the morphological evolution of the
Sciara del Fuoco during the 2002-2003 Stromboli eruption using multi-temporal photogrammetry, ISPRS
Journal of Photogrammetry and Remote Sensing, 59(4), 199-211.
Baltsavias, E. P. (1999), On the performance of photogrammetric scanners, in Photgrammetric Week 99', edited
by D. Fritsch and E. Spiller, pp. 155-173, Wichmann, Heidelberg, Stuttgart, Germany.
SafeLand - FP7
Page 363 of 401
Deliverable 4.1
Barlow, J., S. Franklin, and Y. Martin (2006), High spatial resolution satellite imagery, DEM derivatives, and
image segmentation for the detection of mass wasting processes, Photogrammetric Engineering and Remote
Sensing, 72(6), 687-692.
Barron, J. L., D. J. Fleet, and S. S. Beauchemin (1994), Performance of optical flow techniques, International
Journal of Computer Vision, 12(1), 43-77.
Bendjebbour, A., Y. Delignon, L. Fouque, V. Samson, and W. Pieczynski (2001), Multisensor image
segmentation using Dempster-Shafer fusion in Markov fields context, Geoscience and Remote Sensing,
IEEE Transactions on, 39(8), 1789-1798.
Berthier, E., H. Vadon, D. Baratoux, Y. Arnaud, C. Vincent, K. L. Feigl, F. Rmy, and B. Legrsy (2005),
Surface motion of mountain glaciers derived from satellite optical imagery, Remote Sensing of Environment,
95(1), 14-28.
Bitelli, G., M. Dubbini, and A. Zanutta (2004), Terrestrial Laser Scanning and Digital Photogrammetry
techniques to monitor landslide bodies, in ISPRS, edited, Istanbul.
Blaschke, T. (2010), Object based image analysis for remote sensing, ISPRS Journal of Photogrammetry and
Remote Sensing, 65(1), 2-16.
Borghuis, A. M., K. Chang, and H. Y. Lee (2007), Comparison between automated and manual mapping of
typhoon-triggered landslides from SPOT-5 imagery, International Journal of Remote Sensing, 28, 1843
1856.
Bovolo, F., L. Bruzzone, L. Capobianco, A. Garzelli, S. Marchesi, and F. Nencini (2010), Analysis of the Effects
of Pansharpening in Change Detection on VHR Images, Geoscience and Remote Sensing Letters, IEEE,
7(1), 53-57.
Brady, B. T., and G. A. Rowell (1986), Laboratory investigation of the electrodynamics of rock fracture, Nature,
321, 488-492.
Brandes, J. (2003), Further Developments of Film Emulsions, in Photogrammetric Week 2003, edited by
D.Fritsch, pp. 89-93, Stuttgart, Germany.
Brckl, E., F. K. Brunner, and K. Kraus (2006), Kinematics of a deep-seated landslide derived from
photogrammetric, GPS and geophysical data, Engineering Geology, 88(3-4), 149-159.
Bubenzer, O., and A. Bolten (2008), The use of new elevation data (SRTM/ASTER) for the detection and
morphometric quantification of Pleistocene megadunes (draa) in the eastern Sahara and the southern
Namib, Geomorphology, 102(2), 221-231.
Byksalih, G., and K. Jacobsen (2007), Comparison of DEM Generation by Very High Resolution Optical
Satellites, in New Developments and Challenges in Remote Sensing, edited by EARSeL, pp. 627-637,
Millpress, Rotterdam , Warschau.
Canty, M. J., A. A. Nielsen, and M. Schmidt (2004), Automatic radiometric normalization of multitemporal
satellite imagery, Remote Sensing of Environment, 91(3-4), 441-451.
Cardenal, J., E. Mataa, J. L. Perez-Garcia, J. Delgado, M. A. Hernandez, A. Gonzalez, and J. Diaz-de-Teran
(2008), CLOSE RANGE DIGITAL PHOTOGRAMMETRY TECHNIQUES APPLIED TO LANDSLIDES
MONITORING paper presented at ISPRS Congress 2008, Commission VIII, Beijing, 3-11 July 2008.
Carr, L., and E. Rathje (2008), The Use of Remote Sensing to Identify Landslides Caused by the 2004 Niigataken Chuetsu Earthquake in Japan, in 6th International Workshop on Remote Sensing for Disaster
Management Applications, edited, Pavia, Italy.
Casson, B., C. Delacourt, and P. Allemand (2005), Contribution of multi-temporal remote sensing images to
characterize landslide slip surface. Application to the La Clapire landslide (France), Nat. Hazards Earth
Syst. Sci., 5(3), 425-437.
Chandler, J., P. Ashmore, C. Paola, M. Gooch, and F. Varkaris (2002), Monitoring River-Channel Change
Using Terrestrial Oblique Digital Imagery and Automated Digital Photogrammetry, Annals of the
Association of American Geographers, 92(4), 631-644.
Chandler, J. H., J. G. Fryer, and A. Jack (2005), Metric capabilities of low-cost digital cameras for close range
surface measurement, The Photogrammetric Record, 20(109), 12-26.
Chang, L.-W. (2006), Combining Feature Extraction and Contextual Classification for Landslide Identification
based on Multispectral Imagery, 65 pp, National Cheng Kung University, Tainan.

SafeLand - FP7
Page 364 of 401
6. References
Rev. No: 2
Date: 2012-02-15
Chang, Y.-L., L.-S. Liang, C.-C. Han, J.-P. Fang, W.-Y. Liang, and K.-S. Chen (2007), Multisource Data Fusion
for Landslide Classification Using Generalized Positive Boolean Functions, IEEE Transactions on
Geoscience and Remote Sensing, 45(6), 1697-1708.
Chavez, P. S. (1988), An improved dark-object subtraction technique for atmospheric scattering correction of
multi-spectral data, Remote Sensing of Environment, 24(459-479).
Chavez, P. S. (1996), Image-Based Atmospheric Corrections - Revisited and Improved, Photogrammetric
Engineering & Remote Sensing, 62(9), 1025-1036.
Chen, X., L. Vierling, and D. Deering (2005), A simple and effective radiometric correction method to improve
landscape change detection across sensors and across time, Remote Sensing of Environment, 98(1), 63-79.
Cheng, K. S., C. Wei, and S. C. Chang (2004), Locating landslides using multi-temporal satellite images,
Advances in Space Research, 33(3), 296-301.
Cheng, P., and C. Chaapel (2008), Automatic DEM Generation-Using WorldView-1 Stereo Data with or without
Ground Control, Geoinformatics, Oktober/November 2008.
Choudhary, A., M. Pal, S. Banerjee, and S. Chaudhury (2008), Unusual Activity Analysis Using Video Epitomes
and pLSA, in Indian Conference on Computer Vision, Graphics & Image Processing, edited, pp. 390-397.
Clark, R. N., G. A. Swayze, K. E. Livo, R. F. Kokaly, T. V. V. King, J. B. Dalton, J. S. Vance, B. W. Rockwell, T.
Hoefen, and R. R. McDougal (2002), Surface reflectance calibration of terrestrial imaging spectroscopy
data: A tutorial using AVIRIS, in 10th Airborne Earth Science Workshop, edited, p. 21, JPL Publication.
Coe, J. A., J. P. McKenna, J. W. Godt, and R. L. Baum (2009), Basal-topographic control of stationary ponds on
a continuously moving landslide, Earth Surface Processes and Landforms, 34(2), 264-279.
Coimbra, M. T., and M. Davies (2005), Approximating optical flow within the MPEG-2 compressed domain,
Circuits and Systems for Video Technology, IEEE Transactions on, 15(1), 103-107.
Costa, J. E., and G. P. Williams (1984), Debris-flow dynamics (videotape), in Open File Report, edited, United
States Geological Survey.
Cramer, M. (2001), On the use of direct georeferencing in airborne photogrammetry, in 3rd International
Symposium on Mobile Mapping Technology, edited, p. 13, Cairo, digitale Publikation auf CD, 13 Seiten.
Cramer, M. (2009), Digital Photogrammetric Camera Evaluation The Initiative of the German Society of
Photogrammetry, Remote Sensing and Geoinformation (DGPF), ISPRS Highlights, 7(March 2009 Issue),
http://www.isprshighlights.org/nieuws/item.php?nieuws_id=89.
Cramer, M., H. Krau, K. Jacobsen, M. von Schnermark, and N. S. Haala, V. (2009), Das DGPF-Projekt zur
Evaluierung digitaler photogrammetrischer Kamerasysteme, in DGPF Yearly Conference, edited by E.
Seyfert, DGPF, Jena, Germany.
Crippen, R. E. (1992), Measurement of Subresolution Terrain Displacement using SPOT Panchromatic Imagery,
Episodes, 15(1), 56-61.
Crosta, G. (1997), Evaluating rock mass geometry from photographic images, Rock Mechanics and Rock
Engineering, 30(1), 35-58.
Danneels, G., E. Pirard, and H.-B. Havenith (2007), Automatic landslide detection from remote sensing images
using supervised classification methods, in Geoscience and Remote Sensing Symposium, edited, IGARSS,
Barcelona, Spain.
Datcu, M., H. Daschiel, A. Pelizzari, M. Quartulli, A. Galoppo, A. Colapicchioni, M. Pastori, K. Seidel, P. G.
Marchetti, and S. D'Elia (2003), Information mining in remote sensing image archives: system concepts,
Geoscience and Remote Sensing, IEEE Transactions on, 41(12), 2923-2936.
Delacourt, C., P. Allemand, B. Casson, and H. Vadon (2004), Velocity field of the La Clapire landslide
measured by the correlation of aerial and QuickBird satellite images, Geophys. Res. Lett., 31.
Delacourt, C., D. Raucoules, S. L. Moulic, C. Carnec, D. Feurer, P. Allemand, and M. Cruchet (2009),
Observation of a Large Landslide on La Reunion Island Using Differential Sar Interferometry (JERS and
Radarsat) and Correlation of Optical (Spot5 and Aerial) Images, Sensors, 9(1), 616-630.
Delacourt, C., P. Allemand, E. Berthier, D. Raucoules, B. Casson, P. Grandjean, C. Pambrun, and E. Varel
(2007), Remote-sensing techniques for analysing landslide kinematics: a review, Bulletin de la Societe
Geologique de France, 178(2), 89-100.

SafeLand - FP7
Page 365 of 401
Deliverable 4.1
Dewitte, O., J. C. Jasselette, Y. Cornet, M. Van Den Eeckhaut, A. Collignon, J. Poesen, and A. Demoulin (2008),
Tracking landslide displacements by multi-temporal DTMs: A combined aerial stereophotogrammetric and
LIDAR approach in western Belgium, Engineering Geology, 99(1-2), 11-22.
Dong, J., D. Zhuang, Y. Huang, and J. Fu (2009), Advances in Multi-Sensor Data Fusion: Algorithms and
Applications, Sensors, 9(10), 7771-7784.
Du, Y., P. M. Teillet, and J. Cihlar (2002), Radiometric normalization of multitemporal high-resolution satellite
images with quality control for land cover change detection, Remote Sensing of Environment, 82(1), 123134.
Du, Y., J. Cihlar, J. Beaubien, and R. Latifovic (2001), Radiometric normalization, compositing, and quality
control for satellite high resolution image mosaics over large areas, IEEE Transactions on Geoscience and
Remote Sensing,, 39, 623634.
Dymond, J. R., A.-G. Ausseil, J. D. Shepherd, and L. Buettner (2006), Validation of a region-wide model of
landslide susceptibility in the Manawatu-Wanganui region of New Zealand, Geomorphology, 74(1-4), 7079.
Ebert, A., N. Kerle, and A. Stein (2009), Urban social vulnerability assessment with physical proxies and spatial
metrics derived from air- and spaceborne imagery and GIS data, Natural Hazards, 48(2), 275-294.
Eckardt, A., A. Brner, and F. Lehmann (2009), The Bright Future of High Resolution Satellite Remote Sensing
Will Aerial Photogrammetry Become Obsolete?, in Photgrammetric Week 09', edited by D. Fritsch, pp.
127-134, Wichmann, Stuttgart, Germany.
Eckhardt, D. W., J. P. Verdin, and G. R. Lyford (1990), Automated update of an irrigated lands GIS using SOPT
HRV imagery, Photogrammetric Engineering and Remote Sensing, 56(11), 1515-1522.
Elvidge, C. D., D. Yuan, R. D. Werackoon, and R. S. Lunetta (1995), Relative radiometric normalization of
Landsat Multispectral Scanner (MSS) data using an automated scattergram controlled regression,
Photogrammetric Engineering and Remote Sensing, 61(10), 1255-1260.
Enet, F., and S. T. Grilli (2007), Experimental Study of Tsunami Generation by Three-Dimensional Rigid
Underwater Landslides, Journal of Waterway, Port, Coastal, and Ocean Engineering, 133(6), 442-454.
Ergun, B., T. Kavzoglu, I. Colkesen, and C. Sahin (2010), Data filtering with support vector machines in
geometric camera calibration, Opt. Express, 18(3), 1927-1936.
Euroconsult (2008), Earth Observation Remote Sensing Trends, in Advisory Committee on Commercial Remote
Sensing, edited by A. Keith.
Euroconsult (2009), Satellite-Based Earth Observation, Market Prospects to 2018, edited.
Eyers, R., J. M. Moore, J. Hervas, and J. G. Liu (1998), Integrated use of Landsat TM and SPOT panchromatic
imagery for landslide mapping: case histories from southeast Spain, Geological Society, London,
Engineering Geology Special Publications, 15(1), 133-140.
Fernndez, T., J. Jimnez, P. Fernndez, R. E. Hamdouni, F. J. Cardenal, J. Delgado, C. Irigaray, and J.
Chacn (2008), Automatic Detection of Landslide Features with Remote Sensing Techniques in the Betic
Cordilleras (Granda, Southern Spain), paper presented at ISPRS Congress 2008, Commission VIII, Beijing,
3-11 July 2008.
Ferrier, G. (1995), Evaluation of apparent surface reflectance estimation methodologies, International Journal
of Remote Sensing, 17(12), 22912297.
Fourniadis, I. G., J. G. Liu, and P. J. Mason (2007), Landslide hazard assessment in the Three Gorges area,
China, using ASTER imagery: Wushan-Badong, Geomorphology, 84(1-2), 126-144.
Fujisawa, K., and J. Ohara (2008), Simultaneous monitoring of a collapsing landslide with video cameras,
Advances in Geosciences, 14, 183187.
Galli, M., F. Ardizzone, M. Cardinali, F. Guzzetti, and P. Reichenbach (2008), Comparing landslide inventory
maps, Geomorphology, 94(3-4), 268-289.
Gao, B.-C., M. J. Montes, C. O. Davis, and A. F. H. Goetz (2009), Atmospheric correction algorithms for
hyperspectral remote sensing data of land and ocean, Remote Sensing of Environment, 113(Supplement 1),
S17-S24.
Garzelli, A., F. Nencini, and L. Capobianco (2008), Optimal MMSE Pan Sharpening of Very High Resolution
Multispectral Images, Geoscience and Remote Sensing, IEEE Transactions on, 46(1), 228-236.

SafeLand - FP7
Page 366 of 401
6. References
Rev. No: 2
Date: 2012-02-15
Genevois, R., A. Galgaro, and P. R. Tecca (2001), Image analysis for debris flow properties estimation, Physics
and Chemistry of the Earth, Part C: Solar, Terrestrial & Planetary Science, 26(9), 623-631.
Gianinetto, M. (2008), Multi-Scale Digital Terrain Model Generation Using Cartosat-1 Stereo Images for The
Mausanne Les Alpilles Test Site, in The International Archives of the Photogrammetry, Remote Sensing and
Spatial Information Sciences, edited, pp. 1331-1336.
Gianinetto, M., and M. Scaioni (2008), Automated Geometric Correction of High Resolution Pushbroom
Satellite Data, Photogrammetric Engineering and Remote Sensing, 74, 107-115.
Gislason, P. O., J. A. Benediktsson, and J. R. Sveinsson (2006), Random Forests for land cover classification,
Pattern Recognition Letters, 27(4), 294-300.
Haeberlin, Y., P. Turberg, A. Retire, O. Senegas, and A. Parriaux (2005), Validation of Spot-5 Satellite
Imagery for Geological Hazard Identification and Risk Assessment for Landslides, Mud and Debris Flows
in Matagalpa, Nicaragua, in XXth ISPRS Congress, edited, Istanbul, Turkey.
Hall, F. G., D. E. Strebel, J. E. Nickeson, and S. J. Goetz (1991), Radiometric rectification: Toward a common
radiometric response among multidate, multisensor images, Remote Sensing of Environment, 35(1), 11-27.
Haralick, R. M., and L. Shapiro (1985), Survey: Image segmentation techniques., Computer Vision, Graphics,
and Image Processing 29(100-132).
Haralick, R. M., K. Shanmugam and I. H. Dinstein (1973), Textural Features for Image Classification., IEEE
Transactions on Systems, Man and Cybernetics, 3(6), 610-621.
Hervs, J., and P. L. Rosin (1996), Landslide mapping by textural analysis of ATM data, in 11th Thematic
Conference on Applied Geologic Remote Sensing, edited, pp. 394-402, Las Vegas, USA.
Hervs, J., and P. Bobrowsky (2009), Mapping: Inventories, Susceptibility, Hazard and Risk, in Landslides
Disaster Risk Reduction, edited by K. Sassa and P. Canuti, pp. 321-349, Springer, Heidelberg.
Hervs, J., P. L. Rosin, A. Fernandez-Renau, J. A. Gomez, and C. Leon (1996), Use of airborne multispectral
imagery for mapping landslides in Los Velez, south eastern Spain, in Landslides: 8th International
Conference and Field Trip on Landslides, edited by J. Chacon, C. Irigaray and T. Fernandez, Balkema.
Rotterdam. , Granada, Spain.
Hervs, J., J. I. Barredo, P. L. Rosin, A. Pasuto, F. Mantovani, and S. Silvano (2003), Monitoring landslides
from optical remotely sensed imagery: the case history of Tessina landslide, Italy, Geomorphology, 54(1-2),
63-75.
Hild, F., B. Raka, M. Baudequin, S. Roux, and F. Cantelaube (2002), Multiscale Displacement Field
Measurements of Compressed Mineral-Wool Samples by Digital Image Correlation,, Apllied Optics, 41,
6815-6828.
Hirschmugl, M., H. Gallaun, R. Perko, and M. Schardt (2005), Pansharpening - Methoden fr digitale, sehr
hoch auflsende Fernerkundungsdaten, in Angewandte Geoinformatik, edited by J. Strobl, T. Blaschke and
G. Griesebner, pp. 270-276, Wichmann, Salzburg, Austria.
Holzer-Popp, T., M. Bittner, E. Borg, S. Dech, T. Erbertseder, B. Fichtelmann, and M.Schroedter (2001), Das
automatische Atmosphrenkorrekturverfahren DurchBlick, in Fernerkundung und GIS. Neue Sensoren innovative Methoden, edited by T. Blaschke, Herbert Wichmann Verlag, Heidelberg.
Honda, K., and M. Nagai (2002), Real-time volcano activity mapping using ground-based digital imagery,
ISPRS Journal of Photogrammetry and Remote Sensing, 57(1-2), 159-168.
Hong, G. (2007), Image Fusion, Image Registration, and Radiometric Normalization for High Resolution Image
ProcessingRep., Department of Geodesy and Geomatics Engineering
University of New Brunswick, New Brunswick, Canada.
Hong, G., and Y. Zhang (2008), A comparative study on radiometric normalization using high resolution
satellite images, Int. J. Remote Sens., 29(2), 425-438.
Honkavaara, E., et al. (2009), Digital Airborne PhotogrammetryA New Tool for Quantitative Remote
Sensing?A State-of-the-Art Review On Radiometric Aspects of Digital Photogrammetric Images, Remote
Sensing, 1(3), 577-605.
Hovius, N., C. P. Stark, and P. A. Allen (1997), Sediment flux from a mountain belt derived by landslide
mapping, Geology, 25(3), 231-234.
Hsieh, C.-C. (2006), Subpixel Change Detection and Identification Based on Spectral Unmixing: An Application
to Change Detection of Landslide, 65 pp, National Cheng Kung University, Tainan.

SafeLand - FP7
Page 367 of 401
Deliverable 4.1
Hu, Y., and C. V. Tao (2002), Updating Solutions of the Rational Function Model Using Additional Control
Information, Photogrammetric Engineering & Remote Sensing, 68(7), 715-723.
Ilstad, T., F. V. De Blasio, A. Elverhi, C. B. Harbitz, L. Engvik, O. Longva, and J. G. Marr (2004), On the
frontal dynamics and morphology of submarine debris flows, Marine Geology, 213(1-4), 481-497.
Imaizumi, F., S. Tsuchiya, and O. Ohsaka (2005), Behaviour of debris flows located in a mountainous torrent on
the Ohya landslide, Japan, Canadain Geotechnical Journal, 42(3), 919-931.
Inaba, H., Y. Itakura, and M. Kasahara (2000), Surface velocity computation of debris flows by vector field
measurements, Physics and Chemistry of the Earth, Part B: Hydrology, Oceans and Atmosphere, 25(9),
741-744.
Itakura, Y., H. Inaba, and T. Sawada (2005), A debris-flow monitoring devices and methods bibliography, Nat.
Hazards Earth Syst. Sci., 5(6), 971-977.
Jacobsen, K. (2006), Pros and cons of the orientation of very high resolution optical space images, International
Archives of Photogrammetry, Remote Sensing and Spatial Information ScienceISPRS Symposium, 36(1), 7.
Janzen, D. T., A. L. Fredeen, and R. D. Wheate (2006), Radiometric correction techniques and accuracy
assessment for Landsat TM data in remote forested regions, Canadian Journal of Remote Sensing, 32(5),
330-340.
Jarvis, A., H. I. Reuter, A. Nelson, and E. Guevara (2008), Hole-filled SRTM for the globe Version 4, available
from the CGIAR-CSI SRTM 90m Database, edited.
Jensen, J. R. (1983), Urban/Suburban Land Use Analysis, in Manual of Remote Sensing, edited by R. N. Colwell,
pp. 15711666, American Society of Photogrammetry, Falls Church, VA.
Jensen, J. R. (2005), Introductory digital image processing : a remote sensing perspective, Third edition ed., 526
pp., Prentice-Hall, Upper Saddle River.
Jibson, R. (2006), The 2005 La Conchita, California, landslide, Landslides, 3(1), 73-78.
Joyce, K. E., G. D. Dellow, and P. J. Glassey (2008), Assessing image processing techniques for mapping
landslides, in IEEE International Geoscience and Remote Sensing Symposium, edited, pp. 12311234,
Boston, USA.
Joyce, K. E., S. E. Belliss, S. V. Samsonov, S. J. McNeill, and P. J. Glassey (2009), A review of the status of
satellite remote sensing and image processing techniques for mapping natural hazards and disasters,
Progress in Physical Geography, 33(2), 183-207.
Kb, A. (2002), Monitoring high-mountain terrain deformation from repeated air- and spaceborne optical
data: examples using digital aerial imagery and ASTER data, ISPRS Journal of Photogrammetry and
Remote Sensing, 57(1-2), 39-52.
Kb, A. (2004), Mountain glaciers and permafrost creep. Methodical research perspectives from earth
observation and geoinformatics technologies. Habilitation thesis, , 205 pp, University of Zurich.
Kb, A. (2005), Combination of SRTM3 and repeat ASTER data for deriving alpine glacier flow velocities in
the Bhutan Himalaya, Remote Sensing of Environment, 94(4), 463-474.
Kalantari, M., A. Hashemi, F. Jung, and J. P. Guedon (2009), A New Solution to the Relative Orientation
Problem using only 3 Points and the Vertical Direction, doi: arXiv:0905.3964v1 [cs.CV].
Kandawasvika, A. M. (2009), On Interoperable Management of Multi-Sensors in Landslide Monitoring
Applications, 172 pp, Universitt der Bundeswehr, Erlangen.
Kaufman, Y. J. (1989), The atmospheric effect on remote sensing and its corrections, in Theory and Applications
of Optical Remote Sensing, edited by G. Asrar, Wiley New York.
Kaufman, Y. J., A. E. Wald, L. A. Remer, G. Bo-Cai, L. Rong-Rong, and L. Flynn (1997), The MODIS 2.1μm channel-correlation with visible reflectance for use in remote sensing of aerosol, Geoscience and
Remote Sensing, IEEE Transactions on, 35(5), 1286-1298.
Kavzoglu, T., and I. Colkesen (2009), A kernel functions analysis for support vector machines for land cover
classification, International Journal of Applied Earth Observation and Geoinformation, 11(5), 352-359.
Kennedy, R. E., P. A. Townsend, J. E. Gross, W. B. Cohen, P. Bolstad, Y. Q. Wang, and P. Adams (2009),
Remote sensing change detection tools for natural resource managers: Understanding concepts and
tradeoffs in the design of landscape monitoring projects, Remote Sensing of Environment, 113(7), 13821396.

SafeLand - FP7
Page 368 of 401
6. References
Rev. No: 2
Date: 2012-02-15
Korup, O., M. J. McSaveney, and T. R. H. Davies (2004), Sediment generation and delivery from large historic
landslides in the Southern Alps, New Zealand, Geomorphology, 61, 189207.
Koyama, T., H. Inaba, Y. Itakura, and M. Kasahara (2000), Detection of Debris Flow Based on MPEG Video,
Transactions of the Society of Instrument and Control Engineers, 36(11), 1047-1049.
Kraus, K. (2007), Photogrammetry : geometry from images and laser scans, Second edition ed., 459 pp., De
Gruyter, Berlin.
Ladstdter, R., and V. Kaufmann (2005), Terrestrisch-photogrammetrische Dokumentation des Blockgletschers
im ueren Hochebenkar., in Internationale Geodtische Woche, edited by G. Chesi and T. Weinhold, pp.
92-101, Wichmann, Heidelberg, Obergurgl, Austria.
Laliberte, A., and A. Rango (2009), Texture and Scale in Object-Based Analysis of Subdecimeter Resolution
Unmanned Aerial Vehicle (UAV) Imagery, IEEE Transactions on Geoscience and Remote Sensing, 47(3),
761-770.
Laliberte, A., C. Winters, and A. Rango (2008), A procedure for orthorectification of sub-decimeter resolution
imagery obtained with an unmanned aerial vehicle (UAV) in American Society for Photogrammetry and
Remote Sensing Annual Conference, edited, Portland, OR.
Leprince, S. (2008), Monitoring Earth Surface Dynamics With Optical Imagery, 252 pp, California Institute of
Technology, Pasadena.
Leprince, S., S. Barbot, F. Ayoub, and J. P. Avouac (2007a), Automatic and Precise Orthorectification,
Coregistration, and Subpixel Correlation of Satellite Images, Application to Ground Deformation
Measurements, Geoscience and Remote Sensing, IEEE Transactions on, 45(6), 1529-1558.
Leprince, S., F. Ayoub, Y. Klinger, and J.-P. Avouac (2007b), Co-Registration of Optically Sensed Images and
Correlation (COSI-Corr): an Operational Methodology for Ground Deformation Measurements, in
Geoscience and Remote Sensing Symposium, edited, pp. 27002702, Institute of Electrical and Electronics
Engineers, Barcelona, Spain.
Lim, M., N. J. Rosser, R. J. Allison, and D. N. Petley (2010), Erosional processes in the hard rock coastal cliffs
at Staithes, North Yorkshire, Geomorphology, 114(1-2), 12-21.
Lu, D., P. Mausel, E. Brondzio, and E. Moran (2004), Change detection techniques, International Journal of
Remote Sensing, 25(12), 2365-2401.
Luhmann, T., S. Robson, S. Kyle, and I. Harley (2006), Close range photogrammetry : principles, techniques and
applications, 500 pp., Whittles, Caithness.
Mantovani, F., R. Soeters, and C. J. Van Westen (1996), Remote sensing techniques for landslide studies and
hazard zonation in Europe, Geomorphology, 15(3-4), 213-225.
Mantovani, F., A. Pasuto, S. Silvano, and A. Zannoni (2000), Collecting data to define future hazard scenarios
of the Tessina landslide, International Journal of Applied Earth Observation and Geoinformation, 2(1), 3340.
Marcelino, E. V., A. R. Formaggio, and E. E. Maed (2009), Landslide inventory using image fusion techniques
in Brazil, International Journal of Applied Earth Observation and Geoinformation, 11(3), 181-191.
Marcelino, E. V., L. M. G. Fonseca, F. Ventura, and A. N. C. S. Rosa (2003), Evaluation of HIS, PCA and
Wavelet Transform fusion techniques for the identification of landslide scars using satellite data, in IX
Simpsio Brasileiro de Sensoriamento Remoto, edited, pp. 487494, Belo Horizonte, Brasil.
Marchi, L., M. Arattano, and A. M. Deganutti (2002), Ten years of debris-flow monitoring in the Moscardo
Torrent (Italian Alps), Geomorphology, 46(1-2), 1-17.
Marpu, P. R., I. Niemeyer, S. Nussbaum, and R. Gloaguen (2008), A procedure for automatic object-based
classification, edited, pp. 169-184.
Martha, T., N. Kerle, C. J. van Westen, and K. Kumar (2009a), Characterising spectral, spatial and
morphometric properties of landslides for semi-automatic detection using object-oriented methods,
Geomorphology, In Press, Corrected Proof.
Martha, T., N. Kerle, J. V. Jetten, C. v. Westen, and K. V. Kumar (2009b), Landslide Volumetric Analysis Using
Cartosat-1 Derived DEMs, in Geoscience and Remote Sensing Letters, edited.
Martha, T., N. Kerle, C. van Westen, V. Jetten, and K. Vinod Kumar (2010a), Effect of Sun Elevation Angle on
DSMs Derived from Cartosat-1 Data. , Photogrammetry and Remote Sensing, Article In Press.

SafeLand - FP7
Page 369 of 401
Deliverable 4.1
Martha, T. R., N. Kerle, V. Jetten, C. J. van Westen, and K. Vinod Kumar (2010b), Landslide Volumetric
Analysis Using Cartosat-1-Derived DEMs, Geoscience and Remote Sensing Letters, IEEE, PP(99), 1-5.
Maruya, M., and H. Ohyama (2008), Accurate DEM and Ortho-Rectified Image Production from ALOS/PRISM,
paper presented at Geoscience and Remote Sensing Symposium, 2008. IGARSS 2008. IEEE International.
Mason, P. J., M. S. Rosenbaum, and J. M. Moore (1998), Digital image texture analysis for landslide hazard
mapping, Geological Society, London, Engineering Geology Special Publications, 15(1), 297-305.
Mavrantza, O., and D. Argialas (2008), An object-oriented image analysis approach for the identification of
geologic lineaments in a sedimentary geotectonic environment, edited, pp. 383-398.
Moine, M., A. Puissant, and J.-P. Malet (2009), Detection of landslides from aerial and satellite images with a
semi-automatic method.
Application to the Barcelonnette basin (Alpes-de-Haute-Provence, France), in International Conference
Landslide Processes, edited by J.-P. Malet, A. Rematre and T. Bogaard, pp. 63-68, Strasbourg, France.
Murphy, M. A., and W. P. Burgess (2006), Geometry, kinematics, and landscape characteristics of an active
transtension zone, Karakoram fault system, Southwest Tibet, Journal of Structural Geology, 28(2), 268-283.
Nagasubramanian, V., P. Radhadevi, R. Ramachandran, and R. Krishnan (2008), 3D Reconstrution with
Rational Function Model, Journal of the Indian Society of Remote Sensing, 36(1), 27-35.
Nelson, T., H. G. Wilson, B. Boots, and M. A. Wulder (2005), Use of ordinal conversion for radiometric
normalization and change detection, International Journal of Remote Sensing, 26(3), 535 - 541.
Nichol, J., and M. S. Wong (2005), Satellite remote sensing for detailed landslide inventories using change
detection and image fusion, International Journal of Remote Sensing, 26(9), 1913 - 1926.
Nichol, J. E., A. Shaker, and M.-S. Wong (2006), Application of high-resolution stereo satellite images to
detailed landslide hazard assessment, Geomorphology, 76(1-2), 68-75.
Niethammer, U., S. Rothmund, and M. Joswig (2009), UAV-based remote sensing of the slow-moving landslide
Super-Sauze, paper presented at International Conference Landslide Processes: from geomorpholgic
mapping to dynamic modelling, CERG Editions, Strasbourg, February 6 to 7, 2009.
Niethammer, U., M. R. James, S. Rothmund, J. Travelletti, and M. Joswig (forthcoming), Very high spatial
resolution monitoring of the Super-Sauze landslide with an UAV-based remote sensing technique,
Engineering Geology.
Ochiai, H., Y. Okada, G. Furuya, Y. Okura, T. Matsui, T. Sammori, T. Terajima, and K. Sassa (2004), A fluidized
landslide on a natural slope by artificial rainfall, Landslides, 1(3), 211-219.
Over, M., B. Schottker, M. Braun, and G. Menz (2003), Relative radiometric normalisation of multitemporal
Landsat data-a comparison of different approaches, paper presented at Geoscience and Remote Sensing
Symposium, 2003. IGARSS '03. Proceedings. 2003 IEEE International.
Owen, L. A., U. Kamp, G. A. Khattak, E. L. Harp, D. K. Keefer, and M. A. Bauer (2008), Landslides triggered by
the 8 October 2005 Kashmir earthquake, Geomorphology, 94(1-2), 1-9.
Pal, R., Pal, K. (1993), A review on image segmentation techniques, Pattern Recogn, 26(9), 1277-1294.
Paolini, L., F. Grings, J. A. Sobrino, J. C. J. Muoz, and H. Karszenbaum (2006), Radiometric correction effects
in Landsat multi-date/multi-sensor change detection studies, International Journal of Remote Sensing,
27(4), 685 - 704.
Park, N.-W., and K.-H. Chi (2008), Quantitative assessment of landslide susceptibility using high-resolution
remote sensing data and a generalized additive model, Int. J. Remote Sens., 29(1), 247-264.
Petley, D. N., W. D. O. Crick, and A. B. Hart (2002), The use of satellite imagery in landslide studies in high
mountain area, in 23rd Asian Conference on Remote Sensing edited, Kathmandu, Nepal.
Petrie, G. (2008), paceborne Digital Imaging Sensors and Systems, in Advances in Photogrammetry, Remote
Sensing and Spatial Information Sciences, edited by Z. Li, J. Chen and E. Baltsavias, pp. 29-43, CRC Press,
London.
Pierrot-Deseilligny, M., and N. Paparoditis (2006), A multiresolution and optimization-based image matching
approach: An application to surface reconstruction from SPOT5-HRS stereo imagery, in ISPRS Workshop
On Topographic Mapping From Space (With Special Emphasis on Small Satellites), edited, Ankara, Turky.
Poon, J. (2007), Spatial Information Generation from High-Resolution Satellite Imagery, 160 pp, University of
Melbourne, Melbourne.

SafeLand - FP7
Page 370 of 401
6. References
Rev. No: 2
Date: 2012-02-15
Popp, T. (1995), Correcting atmospheric masking to retrieve the spectral albedo of land surfaces from satellite
measurements, International Journal of Remote Sensing, 16(18), 3483 - 3508.
Proske, H., K. Granica, M. Hirschmugl, and M. Wurm (2008), Landslide Detection and Susceptibility Mapping
using innovative Remote Sensing Data Sources, in Intraprevent, edited, pp. 219 - 229, Dornbirn, Austria.
Puymbroeck, N. V., R. Michel, R. Binet, J.-P. Avouac, and J. Taboury (2000), Measuring Earthquakes from
Optical Satellite Images, Appl Optics, 39(20), 3486-3494.
Qunot, G. M., J. Pakleza, and T. A. Kowalewski (1998), Particle image velocimetry with optical flow,
Experiments in Fluids, 25(3), 177-189.
Rahman, H., and G. Dedieu (1994), SMAC: A simplified method for the atmospheric correction of satellite
measurements in the solar spectrum, International Journal of Remote Sensing, 15(1), 123143.
Rau, J.-Y., L.-C. Chen, J.-K. Liu, and T.-H. Wu (2007), Dynamics Monitoring and Disaster Assessment for
Watershed Management Using Time-Series Satellite Images, Geoscience and Remote Sensing, IEEE
Transactions on, 45(6), 1641-1649.
Reulke, R. (2003), Film-based and Digital Sensors Augmentation or Change in Paradigm?, in
Photogrammetric Week 2003, edited by D. Fritsch, pp. 41-52, Stuttgart, Germany.
Richards, J. A., and X. Jia (2006), Remote Sensing Digital Image Analysis, Springer, Berlin, Germany.
Richter, R. (1990), A fast atmospheric correction algorithm applied to Landsat TM images, International
Journal of Remote Sensing, 11(1), 159 - 166.
Richter, R. (1998), Correction of Satellite Imagery Over Mountainous Terrain, Appl Optics, 37(18), 4004-4015.
Richter, R., T. Kellenberger, and H. Kaufmann (2009), Comparison of Topographic Correction Methods,
Remote Sensing, 1(3), 184-196.
Riedel, T., C. Thiel, and C. Schmullius (2008), Fusion of multispectral optical and SAR images towards
operational land cover mapping in Central Europe, edited, pp. 493-511.
Robson, S., and M. James (2007), Photogrammetric image sequence processing to determine change in active
lava flows, in Annual Conference of the Remote Sensing & Photogrammetry Society, edited, Newcastle upon
Tyne.
Roessner, S., H.-U. Wetzel, H. Kaufmann, W. Kornus, M. Lehner, P. Reinartz, and R. Mueller (2000), Landslide
investigations in Southern Kyrgyzstan based on a Digital Elevation Model derived from MOMS-2P data, in
IAPRS, , edited, pp. 1259-1266, Amsterdam.
Roncella, R., G. Forlani, and F. Remondino (2005), Photogrammetry for geological applications: automatic
retrieval of discontinuity in rock slopes, in SPIE-IS&T Electronc Imaging, edited, pp. 17-27, San Jose,
California.
Sadasiva Rao, B., A. S. R. K. V. M. Mohan, K. Kalyanaraman, and K. Radhakrishnan (2006), Evaluation of
Cartosat-I Stereo Data of Rome, in International Symposium on Geospatial Databases for Sustainable
Development,, edited, Goa, India.
Sarkar, A., A. Banerjee, N. Banerjee, S. Brahma, B. Kartikeyan, M. Chakraborty, and K. L. Majumder (2005),
Landcover classification in MRF context using Dempster-Shafer fusion for multisensor imagery, Image
Processing, IEEE Transactions on, 14(5), 634-645.
Scherler, D., S. Leprince, and M. R. Strecker (2008), Glacier-surface velocities in alpine terrain from optical
satellite imagery--Accuracy improvement and quality assessment, Remote Sensing of Environment, 112(10),
3806-3819.
Schmidt, S., and M. Nsser (2009), Fluctuations of Raikot Glacier during the past 70 years: a case study from
the Nanga Parbat massif, northern Pakistan, Journal of Glaciology, 55(194), 949-959.
Schneevoigt, N. J., S. van der Linden, H.-P. Thamm, and L. Schrott (2008), Detecting Alpine landforms from
remotely sensed imagery. A pilot study in the Bavarian Alps, Geomorphology, 93(1-2), 104-119.
Schpfer, E., S. Lang, and F. Albrecht (2008), Object-fate analysis - spatial relationships for the assessment of
object transition and correspondence, edited, pp. 785-801.
Schott, J. R., C. Salvaggio, and W. J. Volchok (1988), Radiometric scene normalization using pseudoinvariant
features., Remote Sensing of Environment, 26(1), 1-16.
Schowengerdt, R. A. (2007), Remote Sensing - Models and Methods for Image Processing, 3rd ed., 509 pp.,
Academic Press, Burlington.

SafeLand - FP7
Page 371 of 401
Deliverable 4.1
Schroeder, T. A., W. B. Cohen, C. Song, M. J. Canty, and Z. Yang (2006), Radiometric correction of multitemporal Landsat data for characterization of early successional forest patterns in western Oregon, Remote
Sensing of Environment, 103(1), 16-26.
Serpico, S. R., and G. Moser (2007), MRF-based remote sensing image classification with automatic model
parameter estimation. , in Signal and Image Processing for Remote Sensing, edited by C. H. Chen, CRC
Press, Boca Raton, FL.
Sertel, E., S. H. Kutoglu, and S. Kaya (2007), Geometric correction accuracy of different satellite sensor images:
application of figure condition, International Journal of Remote Sensing, 28(20), 4685 - 4692.
Singhroy, V., K. E. Mattar, and A. L. Gray (1998), Landslide characterisation in Canada using interferometric
SAR and combined SAR and TM images, Advances in Space Research, 21(3), 465-476.
Song, C., C. E. Woodcock, K. C. Seto, M. P. Lenney, and S. A. Macomber (2001), Classification and Change
Detection Using Landsat TM Data: When and How to Correct Atmospheric Effects?, Remote Sensing of
Environment, 75(2), 230-244.
Stewnius, H., C. Engels, and D. Nistr (2006), Recent developments on direct relative orientation, ISPRS
Strang, G. a. N., and T. Nguyen (1996), Wavelets and Filterbanks, Wellesley Press, Cambridge.
Sturzenegger, M., and D. Stead (2009), Quantifying discontinuity orientation and persistence on high mountain
rock slopes and large landslides using terrestrial remote sensing techniques, Natural Hazards and Earth
System Sciences, 9, 267287.
Take, W. A., M. D. Bolton, P. C. P. Wong, and F. J. Yeung (2004), Evaluation of landslide triggering
mechanisms in model fill slopes, Landslides, 1(3), 173-184.
Tatem, A. J., S. J. Goetz, and S. I. Hay (2008), Fifty Years of Earth Observation Satellites: Views from above
have lead to countless advances on the ground in both scientific knowledge and daily life, American
Scientist., 96(5), 390-398.
Tenenbaum, D. E., L. E. Band, S. T. Kenworthy, and C. L. Tague (2006), Analysis of soil moisture patterns in
forested and suburban catchments in Baltimore, Maryland, using high-resolution photogrammetric and
LIDAR digital elevation datasets, Hydrological Processes, 20(2), 219-240.
Tokumaru, P. T., and P. E. Dimotakis (1995), Image correlation velocimetry, Experiments in Fluids, 19(1), 1-15.
Toutin, T. (2004), Review article: Geometric processing of remote sensing images: models, algorithms and
methods, International Journal of Remote Sensing, 25(10), 1893 - 1924.
Toutin, T. (2006), Comparison of 3D Physical and Empirical Models for Generating DSMs from Stereo HR
Images, Photogrammetric Engineering & Remote Sensing, 72(5), 597604.
Travelletti, J., C. Delacourt, G. Koval, J.-P. Malet, J. Schmittbuhl, and D. B. v. Dam (2009), A multi-temporal
image correlation method to characterize landslide displacement fields, Geophysical Research Abstracts,
11.
Tsutsui, K., S. Rokugawa, H. Nakagawa, S. Miyazaki, T. Chin-Tung Cheng Shiraishi, and S.-D. Yang (2007),
Detection and Volume Estimation of Large-Scale Landslides Based on Elevation-Change Analysis Using
DEMs Extracted From High-Resolution Satellite Stereo Imagery IEEE Transactions on Geoscience and
Remote Sensing Letters, 45(6), 1681 - 1696
van Asselen, S., and A. C. Seijmonsbergen (2006), Expert-driven semi-automated geomorphological mapping for
a mountainous area using a laser DTM, Geomorphology, 78(3-4), 309-320.
van Westen, C. J., and F. L. Getahun (2003), Analyzing the evolution of the Tessina landslide using aerial
photographs and digital elevation models, Geomorphology 54(12), 7789.
Velloso, M. L. F., F. J. de Souza, and M. Simoes (2002), Improved radiometric normalization for land cover
change detection: an automated relative correction with artificial neural network, paper presented at
Geoscience and Remote Sensing Symposium, 2002. IGARSS '02. 2002 IEEE International.
Vermote, E. F., D. Tanr, J. L. Deuz, M. Herman, and J. J. Morcrette (1997), Second simulation of the satellite
signal in the solar spectrum, 6S: An overview, IEEE Trans. Geoscience and Remote Sensing, 35(3), 675
686.
Walstra, J., J. H. Chandler, N. Dixon, and T. A. Dijkstra (2007), Aerial photography and digital
photogrammetry for landslide monitoring, Geological Society, London, Special Publications, 283(1), 53-63.

SafeLand - FP7
Page 372 of 401
6. References
Rev. No: 2
Date: 2012-02-15
Wang, Z. H. (2004), RS+GCPs select basic information of landslide, The Chinese Journal of Geological Hazard
and Control, 15(1), 94-101.
Weber, D., and A. Herrmann (2000), Contribution de la photogrammtrie numrique l'tude spatio-temporelle
de versants instables: l'exemple du glissement de Super-Sauze (Alpes-de-Haute-Provence). , Bulletin de la
Socit gologique de France, 171(6), 637648.
Wen, G., S.-C. Tsay, R. F. Cahalan, and L. Oreopoulos (1999), Path radiance technique for retrieving aerosol
optical thickness over land, Journal of Geophysical Research, 104(D24), 3132131332.
White, D. J., W. A. Take, and M. D. Bolton (2003), Soil deformation measurement using particle image
velocimetry (PIV) and photogrammetry, Gotechnique, 53(7), 619631.
Whitworth, M., D. Giles, and W. Murphy (2002), Identification of landslides in clay terrains using Airborne
Thematic Mapper (ATM) multispectral imagery, SPIE.
Whitworth, M., D. Giles, and W. Murphy (2005), Airborne remote sensing for landslide hazard assessment: a
case study on the Jurassic escarpment slopes of Worcestershire, UK, Quarterly Journal of Engineering
Geology and Hydrogeology, 38(3), 285-300.
Whitworth, M., D. Giles, and I. Anderson (2006), Landslide imaging techniques for urban geoscience
reconnaissance in 10th IAEG Congress - Engineering geology for tomorrow's cities, edited, Nottingham,
United Kingdom.
Wills, C. J., and T. P. McCrink (2002), Comparing Landslide Inventories: The Map Depends on the Method,
Environmental and Engineering Geoscience, 8(4), 279-293.
Wolff, K., and A. Gruen (2008), Up to Date DSM Generation Using High Resolution Satellite Image Data, in
21th ISPRS Congress, edited, pp. 1103-1108, International Archives of Photogrammetry, Remote Sensing
and Spatial Information Sciences, Beijing.
Xu, J.-F., and J.-F. Huang (2008), Empirical Line Method Using Spectrally Stable Targets to Calibrate IKONOS
Imagery, Pedosphere, 18(1), 124-130.
Yaallah, S. M., and M. R. Saradjian (2005), Automatic normalization of satellite images using unchanged pixels
within urban areas, Information Fusion, 6, 235-241.
Yamaguchi, Y., S. Tanaka, T. Odajima, T. Kamai, and S. Tsuchida (2003), Detection of a landslide movement as
geometric misregistration in image matching of SPOT HRV data of two different dates, International
Journal of Remote Sensing, 24(18), 3523-3534.
Yang, X., and C. P. Lo (2000), Relative Radiometric Normalization
Performance for Change Detection from
Multi-Date Satellite Images, Photogrammetric Engineering & Remote Sensing, 66(8), 967-980.
Yin, H.-Y., Y. Lin, J.-C. Lien, B.-J.Lee, T.-Y. Chou, Y.-M. Fang, H.-P. Lien, and Y.-H. Chang (2007), The Study
of On-site and Mobile Debris Flow Monitoring Station, in 2nd International Conference on Urban Disaster
Reduction, edited, Taipei, Taiwan.
Youssef, A., N. Maerz, and A. Hassan (2009), Remote sensing applications to geological problems in Egypt: case
study, slope instability investigation, Sharm El-Sheikh/Ras-Nasrani Area, Southern Sinai, Landslides, 6(4),
353-360.
Yuan, D., and C. D. Elvidge (1996), Comparison of relative radiometric normalization techniques, ISPRS
Zakeri, A., K. Heg, and F. Nadim (2008), Submarine debris flow impact on pipelines -- Part I: Experimental
investigation, Coastal Engineering, 55(12), 1209-1218.
Zhang, H., J. E. Fritts, and S. A.Goldman (2008a), Image segmentation evaluation: A survey of unsupervised
methods, Comput Vis Image Und, 110(2), 260-280.
Zhang, L. (2005), Automatic Digital Surface Model (DSM) Generation from Linear Array Images, 219 pp, Swiss
Federal Institute of Technology Zurich.
Zhang, L., L. Yang, H. Lin, and M. Liao (2008b), Automatic relative radiometric normalization using iteratively
weighted least square regression, International Journal of Remote Sensing, 29(2), 459 - 470.
Zhang, Z., G. H., Z. W., and Y. Zhang (2005), Application of remote sensing to study of landslide, paper
presented at Geoscience and Remote Sensing Symposium, 2005. IGARSS '05. Proceedings. 2005 IEEE
International.

SafeLand - FP7
Page 373 of 401
Deliverable 4.1
Zitov, B., and J. Flusser (2003), Image registration methods: a survey, Image Vision Comput, 21(11), 9771000.
6.3. SPECIFIC REFERENCES FOR THE CHAPTER 2B

Abelln, A. 2009. Improvements in our understanding of rockfall phenomenon by Terrestrial Laser Scanning.
Emphasis on Change detection and its application to spatial prediction. PhD Thesis RISKNAT group,
University of Barcelona, 185 pp.
Abelln, A., Calvet, J., Vilaplana, J.M., Blanchard, J., 2010. Detection and spatial prediction of rockfalls by
means of terrestrial laser scanner monitoring. Geomorphology vol. 119(3-4), 162-171. doi:10.1016/
j.geomorph.2010.03.016.
Abelln, A., Vilaplana, J.M., Martnez, J., 2006. Application of a long-range Terrestrial Laser Scanner to a
detailed rockfall study at Vall de Nria (Eastern Pyrenees, Spain). Engineering Geology 88, 136148.
Abelln, A., Jaboyedoff, M., Oppikofer, T., Vilaplana, J. M., 2009. Detection of millimetric deformation using a
terrestrial laser scanner: experiment and application to a rockfall event, Nat. Hazards Earth Syst. Sci., 9,
365-372.
Agliardi, F., Crosta, G.B., 2003. High resolution three-dimensional numerical modelling of rockfalls.
International Journal of Rock Mechanics and Mining Sciences 40, 455471.
ASPRS LiDAR Committee, 2004. ASPRS Guidelines Vertical Accuracy Reporting for LiDAR Data. Available at
http://www.asprs.org/society/committees/lidar/Downloads/Vertical_Accuracy_Reporting_
for_Lidar
_Data.pdf.
ASPRS LIDAR committee, 2004. ASPRS Guidelines: Vertical Accuracy Reporting for Lidar Data. Ed: M. Flood
Baillifard, F., Jaboyedoff, M., Rouiller, J.-D., Robichaud, G., Locat, P., Locat, J., Couture, R., Hamel, G., 2004.
Towards a GIS-based rockfall hazard assessment along the Quebec City Promontory, Quebec, Canada.
Pages 207213 of: Lacerda, W.A., Ehrlich, M., Fontoura, A.B. & Sayo, A. (eds), Landslides: Evaluation
and Stabilization. London: Taylor & Francis.
Baillifard, F., Jaboyedoff, M., Sartori, M., 2003. Rockfall hazard mapping along a mountainous road in
Switzerland using a GIS-based parameter rating approach. Natural Hazards and Earth System Sciences,
3(5), 435442.
Baltsavias, E. P., 1999. Airbone laser scanning: basic relations and formulas. ISPRS Journal of
Photogrammetry & Remote Sensing 54:199-214.
Bauer, A, Paar, G., Kaltenbck, A., 2005. Mass movement monitoring using terrestrial laser scanner for rock
fall management, Proceedings of the First International Symposium on Geo-Information for Disaster
Management, pp. 393406 Delft, The Netherlands.
Besl, P., McKay, N., 1992.A method for registration of 3-D shapes. IEEE Transactions on Pattern Analysis and
Machine Intelligence 14:239-256.
Bornaz, L., Rinaudo, F., 2004. Terrestrial Laser Scanner Data processing. Proceedings of XX ISPRS Congress,
12-23 July 2004, Istanbul, Turkey, IAPRS vol. XXXXV, part B5: pp. 514-519.
Chandra, A.M., 2005. Higher surveying. Chapter 1: Triangulation and trilateration. New age international
publishers, 424 pp.
Chen, Y., Medioni, G., 1992. Object Modelling by Registration of Multiple Range Images. Image and Vision
Computing 10, 145155.
Dallaire E. E., 1974. Electronic distance measuring revolution well under way. Civil Engineering 44(10): 66-71.
Delacourt C., Allemand P, Casson B., Vadon H., 2004. Velocity field of the "La Clapire" landslide measured by
the correlation of aerial and QuickBird satellite images. Geophysical Research Letters 31, No. 15, L15619,
10.1029/2004GL020193.
Dorren, L.K.A., Seijmonsbergen, A.C., 2003. Comparison of three GIS-based models for prediction rockfall
runout zones at a regional scale. Geomorphology 56, 4964.
Frattini, P., Crosta, G., Carrara, A. & Agliardi, F. 2008. Assessment of rockfall susceptibility by integrating
statistical and physically-based approaches. Geomorphology, 94(3-4), 419437.

SafeLand - FP7
Page 374 of 401
6. References
Rev. No: 2
Date: 2012-02-15
Gokceoglu, C., Sonmez, H., Ercanoglu, M., 2000. Discontinuity controlled probabilistic slope failure risk maps
of the Altindag (settlement) region in Turkey. Engineering Geology, 55(4), 277296.
Gnter, A., 2003. Slopemap: programs for automated mapping of geometrical and kinematical properties of
hard rock hill slopes, Computer & Geosciences 29/7, 865-875.
Guzzetti, F., Reichenbach, P. & Wieczorek, G. F. 2003. Rockfall hazard and risk assessment in the Yosemite
Valley, California, USA. Natural Hazards and Earth System Sciences, 3(6), 491503.
Habib, A., Van Rens, J. 2008. Quality Assurance and Quality Control of LiDAR Systems and Derived Data.
American Society for Photogrammetry and Remote Sensing (ASPRS) committee, 15pp.
<http://www.asprs.org/society/committees/lidar/AKAM_LiDAR_Calibration.pdf>
Harding, D., 2009. Pulsed Laser Altimeter Ranging Techniques and Implications for Terrain Mapping, in
Topographic Laser Ranging and Scanning: Principles and Processing, Jie Shan and Charles Toth, eds.,
CRC Press, Taylor & Francis Group, pp. 173 194.
Heritage GL, Large ARG. 2009. Laser scanning for the environmental sciences. Wiley-Blackwell, London.
Hiremagalur J, Yen KS, Akin K, Bui T, Lasky, TA, Ravani B (2007) Creating Standards and specifications for
the use of Laser Scanning in CalTrans projects. Technical report n F/CA/RI/2006/46, California
Department of Transportation, US.
Hoek, E., 2000. Analysis of rockfall hazards. Rock Engineering: course notes, p. 115-136. Rocscience. Available
at <www.rocscience.com>.
Hoek, E., Bray, J., 1981. Rock Slope Engineering. 3rd edn. London: E & FN Spon.
Horton, P., Jaboyedoff, M., Bardou , E. 2008. Debris flow susceptibility mapping at a regional scale. In Locat,
J., Perret, D., Turmel, D., Demers, D., and Leroueil, S., 2008. Proceedings of the 4th Canadian Conference
on Geohazards. From Causes to Management. Presse de lUniversit Laval, Qubec, 594p.
Ingensand, H., 2006. Metrological aspects in terrestrial laser-scanning technology. In Proceedings of the 3rd
IAG/12th FIG Symposium, Baden, Austria.
Jaboyedoff, M. & Labiouse, V. 2003. Preliminary assessment of rockfall hazard based on GIS data. Pages 575
578 of: 10th International Congress on Rock Mechanics ISRM 2003 - Technology roadmap for rock
mechanics. South African Institute of Mining and Metallurgy, Johannesburgh, South Africa.
Jaboyedoff, M., Baillifard, F., Couture, R., Locat, J. & Locat, P. 2004c. New insight of geomorphology and
landslide prone area detection using Digital Elevation Model(s). Pages 191198 of: Lacerda, W.A., Ehrlich,
M., Fontoura, A.B. & Sayo, A. (eds), Landslides: Evaluation and Stabilization. London: Taylor & Francis.
Jaboyedoff, M., Baillifard, F., Couture, R., Locat, J. , Locat, P. 2004. New insight of geomorphology and
landslide prone area detection using Digital Elevation Model(s). Pages 191198 of: Lacerda, W.A., Ehrlich,
M., Fontoura, A.B. & Sayo, A. (eds), Landslides: Evaluation and Stabilization. London: Taylor & Francis.
Jaboyedoff, M., Metzger, R., Oppikofer T., Couture, R., Derron, M.-H., Locat, J., Turmel, D., 2007. New insight
techniques to analyze rock-slope relief using DEM and 3D-imaging cloud points: COLTOP-3D software.
In: Eberhardt, E., Stead, D., Morrison, T. (Eds.) Rock Mechanics: Meeting Society's Challenges and
Demands (Vol. 2). Taylor & Francis, pp. 61-68.
Jaboyedoff, M., Pedrazzini, A., Loye, A., Horton, P. & Surace, I. 2008. Preliminary slope mass movements
susceptibility mapping using Lidar DEM. Pages 419426 of: Proceedings of the 61st Canadian
Geotechnical Conference, vol. 1.
Jaboyedoff, M., Choffet, M., Derron, M.-H., Horton, P., Loye, A., Longchamp, C., Mazotti, B., Michoud, C.,
Pedrazzini, A., 2010. Preliminary slope mass movements susceptibility mapping using DEM and LiDAR
DEM. Submitted in march 2010. 58p.
Kamerman, G. W., 1993. Laser radar (chapter 1). In: Fox CS (ed) The Infrared and Electro-Optical Systems
Handbook. Infrared Information Analysis Center, Ann Arbor, Michigan, 176.
Kemeny, J., Turner, K. Norton, B., 2006. LIDAR for Rock Mass Characterization: Hardware, Software,
Accuracy and Best-Practices, Proceedings of the Workshop on Laser and Photogrammetric Methods For
Rock Mass Characterization: Exploring New Opportunities, Golden, CO.
Kemeny, J., Turner, K., 2008. Ground-Based LiDAR: Rock Slope Mapping and Assessment. Federal Highway
Administration report, FHWA-CFL/TD-08-006. Available at http://www.iaeg.info/portals/0/Content/
Commissions/Comm19/GROUND-BASED LiDAR Rock Slope Mapping and Assessment.pdf

SafeLand - FP7
Page 375 of 401
Deliverable 4.1
Lan, H., Martin, C. D; Zhou, C; Lim, C.H., 2010. Rockfall hazard analysis using LiDAR and spatial modeling.
Geomorphology, 118(1-2), 213-223. doi:10.1016/j.geomorph.2010.01.002.
Lato, M., Diederichs, M., Hutchinson, D.J., Harrap., R., 2009. Optimization of LiDAR scanning and processing
for automated structural evaluation of discontinuities in rockmasses. International Journal of Rock
Mechanics & Mining Sciences 46, 194199.
Lichti D.D., Jamtsho M., 2006. Angular resolution of terrestrial laser scanners. Photogrammetric Record
21(114), 141160.
Lichti, D.D., 2005. Spectral filtering and classification of terrestrial laser scanner point clouds,
Photogrammetric Record 20, pp. 218240
Lichti, D.D., 2007. Error modelling, calibration and analysis of an AM-CW terrestrial laser scanner system,
ISPRS Journal of Photogrammetry and Remote Sensing 61 (5), 307324.
Lindenbergh, R. and Pfeifer, N., 2005. A statistical deformation analysis of two epochs of terrestrial laser data
of a lock. Proceedings of the 7th Conference on Optical 3D. Measurement Techniques, Vienna, Austria, 6170.
Loye, A., Jaboyedoff, M., and Pedrazzini, A.: Identification of potential rockfall source areas at a regional scale
using a DEM-based geomorphometric analysis, Nat. Hazards Earth Syst. Sci., 9, 1643-1653,
doi:10.5194/nhess-9-1643-2009, 2009.
Malet, J.P., Maquaire, O., Calais, E., 2002. The use of Global Positioning System techniques for the continuous
monitoring of landslides: application to the Super-Sauze earthflow (Alpes-de-Haute-Provence, France).
Geomorphology 43, 33-54.
Manetti L., Steinmann G., 2007. 3DeMoN ROBOVEC Integration of a new measuring instrument in an
existing generic remote monitoring platform. 7th International Symposium on Field Measurements in
Geomechanics, 24-27 September 2007, Boston, MA, USA.
McKean, J., Roering, J., 2004. Objective landslide detection and surface morphology mapping using highresolution airborne laser altimetry. Geomorphology, 57, pp.331351.
Meentemeyer, R. K., Moody, A., 2000. Automated mapping of conformity between topographic and geological
surfaces. Computers & Geosciences, 26(7), 815829.
Miko, M., Vidmar, A., and Brilly, M.: Using a laser measurement system for monitoring morphological changes
on the Strug rock fall, Slovenia, Nat. Hazards Earth Syst. Sci., 5, 143-153, 2005
Monserrat, O., Crosetto, M., 2008. Deformation measurement using terrestrial laser scanning data and least
squares 3D surface matching. ISPRS Journal of Photogrammetry and Remote Sensing, 63(1), 142-154.
Oppikofer T., 2009. Detection, analysis and monitoring of slope movements using High-Resolution Digital
Elevation Models. PhD thesis, Institute of Geomatics and Analysis of Risk, University of Lausanne,
Switzerland, 191 pp.
Oppikofer, T., Jaboyedoff, M., Blikra, L., Derron, M.-H., Metzger, R., 2009. Characterization and monitoring of
the knes rockslide using terrestrial laser scanning. Nat. Hazards Earth Syst. Sci., 9, 1003-1019, 2009
Oppikofer, T., Jaboyedoff, M., Keusen, H.-R., 2008. Collapse of the eastern Eiger flank in the Swiss Alps. Nature
Geosciences 1(8), 531-535.
Optech, 2009. ILRIS-3D Intelligent Laser Ranging and Imaging System, from Optech,
http://www.optech.ca/i3dprodline-ilris3d.htm, access 12 June 2009.
Pedrazzini, A. Abellan, A., Oppikofer, T., Ambrosi, C., Spataro, A., Jaboyedoff, M., 2010. Retrogressive
landslide monitoring by TLS: precursory displacements and final collapse. Case study at Val Canaria
(Ticino, Switzerland). Geophysical Research Abstracts Vol. 12, EGU2010-12160, 2010
Petrie and Toth, 2008. Introduction to laser ranging, profiling and scanning. In Topographic Laser Ranging and
Scanning: principles and processing. Edited by: Shan, J., Toth, C. K., CRC Press, Taylor & Francis, 590p.
Rosser, N.J., Lim, N, Petley, D.N., Dunning, S., Allison, R.J., 2007. Patterns of precursory rockfall prior to slope
failure. Journal of Geophysical Research, Vol. 112, No. F4, doi:10.1029/2006JF000642.
Rosser, N.J., Petley, D.N., Lim, M., Dunning, S. A., Allison, R.J., 2005. Terrestrial laser scanning for monitoring
the process of hard rock coastal cliff erosion. Quarterly Journal of Engineering Geology and Hydrogeology,
38(4), 363-375.
Rouiller, J., Jaboyedoff, M., Marro, Ch., 1998. Pentes instables dans le penique valaisan. Ed V/d/f. Zurich. 239p.

SafeLand - FP7
Page 376 of 401
6. References
Rev. No: 2
Date: 2012-02-15
Settles, E., Gttle, A., Von Poschinger, A., 2008. Slope Monitoring Methods - A State of the Art Report.
ClimChalp, Interreg III B, Alpine Space, Work package 6, Munich, 165pp.
Shan J, Toth K., 2008. Topographic laser ranging and scanning principles and processing. CRC Press, Taylor
& Francis Group, LLC.
Slob, S., Hack, R., 2004. 3D Terrestrial Laser Scanning as a new Field Measurement and Monitoring Technique.
In: Hack, R., Azzam, R., Charlier, R. (eds), Engineering Geology for Infrastructure Planning in Europe: A
European perspective. Springer, Berlin. Lecture notes in Earth Sciences, n.104, pp179-189.
Sturzenegger, M., Stead, D., 2009. Quantifying discontinuity orientation and persistence on high mountain rock
slopes and large landslides using terrestrial remote sensing techniques. Nat. Hazards Earth Syst. Sci., 9,
267287.
Teza, G., Galgaro, A., Zaltron, N., Genevois, R., 2007. Terrestrial laser scanner to detect landslide displacement
fields: a new approach. International Journal of Remote Sensing, Volume 28, Issue 16, pages 3425 3446.
Travelletti, J., Oppikofer, T., Delacourt, C., Malet, J.P., Jaboyedoff, M. 2008. Monitoring landslide
displacements during a controlled rain experiment using a long-range terrestrial laser scanning (TLS).
International Archives of Photogrammetry and Remote Sensing, 37 (B5), 485490.
USACE, 2002. Engineering and Design - Structural Deformation Surveying. U.S. Army Corps of Engineers
Manual EM 1110-1-1804, 292pp
USACE, 2007. Control and topographic surveying. U.S. Army Corps of Engineers Manual EM 1110-1-1005,
498pp.
Van Den Eeckhaut, M., Poesen, J., Verstraeten, G., Vanacker, V., Moeyersons, J., Nyssen, J., and van Beek, L. P.
H. (2005). The effectiveness of hillshade maps and expert knowledge in mapping old deep-seated landslides,
Geomorphology, 67(34), 351363.
Vilajosana, I., Suriach, E., Abelln, A., Khazaradze, G., Garcia, D., Llosa, J., 2008. Rockfall induced seismic
signals: case study in Montserrat, Catalonia, Nat. Hazards Earth Syst. Sci., 8, 805-812.
Voegtle, T., Schwab, I., Landes, T., 2008. Influences of different materials on the measurement of a Terrestrial
Laser Scanner (TLS). Proc. of the XXI Congress, The International Society for Photogrammetry and Remote
Sensing, ISPRS2008, Vol. XXXVII, Commission V, 3-11 July 2008, Beijing, China, 1061-1066.
Vosselman, G., Maas, H.-G., 2009. Airborne and terrestrial laser scanning. Whittles. Publishing, Dunbeath,
Caithness, Scotland. 336 p
Wehr, A., Lohr, U., 1999. Airborne laser scanning - an introduction and overview. ISPRS Journal of
Photogrammetry and Remote Sensing 54:68-82.
6.4. SPECIFIC REFERENCES FOR THE CHAPTER 2C

Adam, N., B. Kampes, M. Eineder, J. Worawattanamateekul, and M. Kircher., The development of a scientific
permanent scatterer system, paper presented at ISPRS Hannover Workshop, Inst. for Photogramm. and
Geoinf., Hannover, Germany. 2003.
Allievi, J., Prati, C., Rocca, F., Savio, G., Arrigoni, M., Zanoletti, L. Combined use of artificial and permanent
scatterers. Proceedings of the ESA Envisat Symposium, 6-10 September 2004, Salzburg (Austria). 2004.
Antonello, G., Casagli, N., Farina, P., Guerri, L., Leva, D., Nico, G., Tarchi, D. SAR interferometry monitoring
of landslides on the Stromboli Volcano. Proceedings of FRINGE 2003 Workshop, 1-5 December 2003,
ESA/ESRIN, Frascati, Italy. 2003.
Arnaud A., Adam N., Hanssen R., Inglada J., Duro J., Closa J., and Eineder M. ASAR ERS interferometric phase
continuity. International Geoscience and Remote Sensing Symposium, 21-25 July 2003, Toulouse (France).
2003.
Atzori S., Manunta M., Fornaro G., Ganas A., Salvi S., "Postseismic displacement of the 1999 Athens
earthquake retrieved by the Differential Interferometry by Synthetic Aperture Radar time series", J.
Geophys. Res., 113, B09309, doi:10.1029/2007JB005504. 2008.
Bamler, R., Hartl, P. Synthetic aperture radar interferometry. Inverse Problems 14. 54p. 1998.
Bamler, R. A Comparison of Range-Doppler and Wavenumber Domain SAR Focussing Algorithms. IEEE Trans.
Geosci. Remote Sens.,, 30, pp. 706-713. 1992.

SafeLand - FP7
Page 377 of 401
Deliverable 4.1
Beckmann P. and A.Spizzichino, The Scattering of Electromagnetic Waves from Rough Surfaces, Norwood, MA:
Artech House. 1987.
Berardino, P., Fornaro, G., Lanari, R., Sansosti, E. A new algorithm for surface deformation monitoring based
on Small Baseline Differential SAR Interferograms. IEEE Trans. Geosci. Remote Sens.. 40 (11), 2375-2383.
2002.
Berardino, P., Costantini, M., Franceschetti, G., Iodice, A., Pietranera, L., Rizzo, V. Use of differential SAR
interferometry in monitoring and modelling large slope instability at Maratea (Basilicata, Italy).
Engineering Geology 68, pp 31-51. 2003.
Bernardini G., Ricci, P., Coppi, F. A Ground Based Microwave Interferometer with imaging capabilities for
remote sensing measurements of displacements, 7th Geomatic Week/3rd Int. Geotelematics Fair, Barcelona,
Spain, 2007 February 20-23. 2007.
Blanco-Sanchez P., Mallorqui J., Duque S., Monnells D. The Coherent Pixels Technique (CPT): An Advanced
DInSAR Technique for Nonlinear Deformation Monitoring. Pure and Applied Geophysics, vol. 165, n. 6, pp.
1167-1193. 2008.
Bovenga, F., Nutricato, R., Refice, A., Wasowski, J. Application of multi-temporal differential interferometry to
slope instability detection in urban/peri-urban areas. Eng. Geol.. 88, 219-240. 2006.
Brgmann, R., Hilley, G.E., Ferretti, A. Resolving vertical tectonics in the San Francisco Bay Area from
Permanent Scatterer InSAR and GPS Analysis. Geology. 34, 221-224. 2006
Canadian Center for Remote-Sensing. http://cct.rncan.gc.ca. 2008.
Canuti, P., Casagli, N., Farina, P., Ferretti, A., Marks, F., Menduni, G. Land subsidence in the Arno river basin
studied through SAR interferometry. Proc. of SISOLS 2005, 7th Int. Symp. on Land Subsidence. Shanghai,
China, 23-28 Oct. 2005. 1, 407-416. 2005.
Carnec, C., Massonnet, D., King, C. Two examples of the use of SAR interferometry on displacement fields of
small spatial extent. Geophysical Research Letters, vol. 23 no. 4. pp 3579-3582. 1996.
Casagli N., Tibaldi A., Merri A., Del Ventisette C., Apuani C., Guerri L., Fortuny-Guasch J., Tarchi D.
Deformation of Stromboli Volcano (Italy) during the 2007 eruption revealed by radar interferometry,
numerical modelling and structural geological field data. Journal of Volcanology and Geothermal Research
182 (2009) 182200. 2003.
Casu F., Manzo M., Lanari. R. A quantitative assessment of the SBAS algorithm performance for surface
deformation retrieval from DInSAR data. Remote Sens. Of the Environment, vol. 102, pp. 195-210. 2006.
Casu, F., Manzo, M., Pepe, A., Lanari R. SBAS-DInSAR Analysis of Very Extended Areas: First Results on a
60,000 km2 Test Site. IEEE Geosci. and Remote Sens. Letters, vol. 5, no. 3, pp 438-442. 2008
Catani, F., Farina, P., Moretti, S., Nico, G., Strozzi, T. On the application of SAR interferometry to
geomorphological studies: Estimation of landform attributes and mass movements. Geomorphology 66. pp
119-131. 2005
Catani, F., Casagli, N., Ermini, L., Righini, G., Menduni, G. Landslide hazard and risk mapping at catchment
scale in the Arno River basin. Landslides. 2, 329-342. 2005b
Colesanti, C., and Wasowski, J. Satellite SAR interferometry for wide-area slope hazard detection and sitespecific monitoring of slow landslides. Proc. 9th Int. Symp. on Landslides, 28 Jun. - 2 Jul., 2004, Rio de
Janeiro, 795-802. 2004.
Colesanti, C., and Wasowski, J. Investigating landslides with space-borne Synthetic Aperture Radar (SAR)
interferometry. Eng. Geol.. 88, 173-199. 2006.
Colesanti, C., Le Moulic, S., Bennani, M., Raucoules, D., Carnec, C., Ferretti, A. Detection of mining related
ground instabilities using the Permanent Scatterers technique - A case study in the east of France. Int. J.
Remote Sens.. 26 (1), 201-207. 2005.
Colesanti., C., Ferretti, A., Prati, C., Rocca, F. Monitoring landslides and tectonic motions with the Permanent
Scatterers Technique. Eng. Geol.. 68, 3-14. 2003.
Colombo, D., Farina, P., Moretti, S., Nico, G., Prati, C. Land subsidence in the Firenze-Prato-Pistoia basin
measured by means of spaceborne SAR interferometry. Proc. IGARSS 2003, Toulouse, France. 2003.
Costantini, M., Iodice, A., Magnapane, L., Pietranera L. Monitoring terrain movements by means of sparse SAR
differential interferometric measurements. Proc. IGARSS 2000, Honolulu, Hawaii, USA. 7, 3225-3227.
2000.

SafeLand - FP7
Page 378 of 401
6. References
Rev. No: 2
Date: 2012-02-15
Costantini, M., S. Falco, F. Malvarosa, F. Minati, A new method for identification and analysis of persistent
scatterers in series of SAR images. in Proc. Int. Geosci. Remote Sensing Symp. (IGARSS), Boston MA, 7-11
July 2008, pp. 449-452. 2008.
Costantini, M., S. Falco, F. Malvarosa, F. Minati, F. Trillo. Method of Persistent Scatterers Pairs (PSP) and
High Resolution SAR Interferometry. Proc. Int. Geosci. Remote Sensing Symp. (IGARSS), Cape Town, South
Africa, 12-17 July 2009. 2009.
Crosetto M. and Pasquali P. DSM generation and deformation measurement from SAR data. Chapter 12 in
Advances in Photogrammetry, Remote Sensing and Spatial Information Science, ISPRS Congress Book,
edited by P. Aplin, 2008. 2008.
Crosetto M., Crippa B., Biescas E. Early detection and in-depth analysis of deformation phenomena by radar
interferometry. Eng Geol, 79 (1-2), pp. 81-91. 2005a.
Crosetto M., Crippa B., Biescas E., Monserrat O., Agudo M. State-of-the-Art of Land Deformation Monitoring
Using Differential SAR Interferometry. ISPRS Hannover Workshop 2005. High Resolution Earth Imaging
for Geospatial Information 17-20 May 2005, XXXVI, Part 1/w3. 2005b.
Crosetto, M., A. Arnaud, J. Duro, E. Biescas, and M. Agudo. Deformation monitoring using remotely sensed
radar interferometric data, paper presented at 11th FIG Symposium on Deformation Measurements, Patras
Univ., Santorini, Greece. 2003.
Crosetto, M., Biescas, E., Duro, J., Closa, J., and Arnaud, A. Generation of Advanced ERS and Envisat
Interferometric SAR Products Using the Stable Point Network Technique, Photogramm. Eng., 74(4), 443
451, 2008.
Czuchlewski, K., Weissel, J.K., Kim, Y. Polarimetric synthetic aperture radar study of the Tsaoling landslide
generated by the 1999 Chi-Chi earthquake, Taiwan. J. Geophys. Res., 108. 2003.
De Maio, A., Fornaro G., Pauciullo A. Detection of Single Scatterers in Multdimensional SAR Imaging. IEEE
Trans. Geosci. Remote Sens, July 2009, 47 (7), pp. 2284-2997. 2009.
Dixon, T.H., Amelung, F., Ferretti, A., Novali, F., Rocca, F., Dokka, R., Sella, G., Kim, S.W., Wdowinski, S.,
Whitman, D. Subsidence and flooding in New Orleans - A subsidence map of the city offers insight into the
failure of the levees during Hurricane Katrina. Nature. 441, 587-588. 2006.
Dubois, P.C., van Zyl, J.J., and Engman, E.T.. An Empirical Soil Moisture Estimation Algorithm Using Imaging
Radar. IEEE Trans. Geosci. Remote Sens., Vol. 33, No. 4, pp. 915-926, 1995.
Dubois, P.C., van Zyl, J.J., and Engman, E.T., Measuring Soil Moisture with Imaging Radars. IEEE Trans.
Geosci. Remote Sens., Vol. 33, No. 4, pp. 915-926. 1995.
Duro J., Closa J., Biescas E., Crosetto M., Arnaud A. High resolution differential interferometry using time
series of ERS and Envisat SAR data. Actes del congrs 6a. Setmana Geomtica (CD-ROM). Barcelona
(ES), 2005.2.8-11.. 2005.
Duro, J., Inglada, J., Closa, J., Adam, N., Arnaud, A. High resolution differential interferometry using time
series of ERS and ENVISAT SAR data. Proc. of FRINGE 2003, 1-5 Dec. 2003, Frascati, Italy. 2003.
ESA Radar courses. http://earth.esa.int/applications/data_util/SARDOCS/spaceborne/Radar_Courses/. 2010.
Farina, P., Casagli, N., Ferretti, A. Radar-interpretation of InSAR measurements for landslide investigations in
civil protection practices. Proc. 1st North American Landslide Conference, Vail, Colorado, 3-8 Jun. 2007.
272-283. 2008.
Farina, P., Colombo, D., Fumagalli, A., Gontier, E., Moretti, S. Integration of Permanent Scatterers Analysis
and High Resolution Optical Images within Landslide Risk Analysis. Proc. of FRINGE 2003, 1-5 Dec. 2003,
Frascati, Italy. 2003.
Farina, P., Colombo, D., Fumagalli, A., Marks, F., Moretti, S. Permanent Scatterers for landslide
investigations: outcomes from the ESA-SLAM project. Eng. Geol.. 88, 200-217. 2006.
Ferretti, A., Prati, C., Rocca, F. Nonlinear subsidence rate estimation using Permanent Scatterers in Differential
SAR Interferometry. IEEE Trans. Geosci. Remote Sens.. 38, 2202-2212. 2000.
Ferretti, A., Prati, C., Rocca, F. Permanent Scatterers in SAR interferometry. IEEE Trans. Geosci. Remote Sens.
39 (1), 8-20. 2001.
Ferretti, A., Monti-Guarnieri, A., Prati, C., Rocca, F., Massonnet, D. Lichtenegger, J. InSAR Principles :
Guidelines for SAR Interferometry Processing and Interpretation. ESA Publications, 234p. 2007.

SafeLand - FP7
Page 379 of 401
Deliverable 4.1
Fornaro, G., Monti Guarnirei, A. Minimum Mean Square Error Space-Varying Filtering of Interferometric SAR
Data. IEEE Trans. Geosci. Remote Sens.,, 40, pp. 11-21. 2002.
Fornaro G., Pauciullo A., Serafino F. Deformation Monitoring over large areas with Multipass Differential SAR
Interferometry: a new approach based on the use of Spatial Differences. Int. Journal of Remote Sens.,. vol.
30, no. 6, April 2009, pp 1455-1478. 2009.
Fornaro G., Reale D., Serafino F. Four-Dimensional SAR Imaging for Height Estimation and Monitoring of
Single and Double Scatterers. IEEE Trans. Geosci. Remote Sens., vol. 47 (1), pp. 224-237. 2009.
Fornaro G., Serafino F., Reale D. 4D SAR Imaging: the Case Study of Rome. doi 10.1109/LGRS.2032133. 2009.
Fortuny-Guash, J., Sieber, A.J. Fast algorithm for near-field synthetic aperture radar processor. IEEE Trans.
Antennas Propagat., vol. 42, pp. 14581460. Oct. 1994.
Fortuny-Guasch J. A Fast and Accurate Far-Field Pseudopolar Format Radar Imaging Algorithm, IEEE Trans.
Geosci. Remote Sens. 47 (4), 1187 1196. April 2009.
Fruneau, B., Apache, J., Delacourt, C. Observation and modeling of the Saint-Etienne-de-Tine landslide using
SAR interferometry. Tectonophysics 265. pp 181-190. 1996.
Gamma Remote Sensing. GammaSar user manual, Bern. 2008.
Hajnsek, I., Pottier, E., and Claude, S. Inversion of Surface Parameters From Polarimetric SAR. IEEE Trans.
Geosci. Remote Sens., Vol. 41, No. 4, pp. 727-744, 2003.
Hajnsek, I., Jagdhuber, T., Shon, H., and Papathanassiou, K.P. Potential of Estimating Soil Moisture Under
Vegetation Cover by Means of PolSAR. IEEE Trans. Geosci. Remote Sens., Vol. 42, No. 2, pp. 442-454,
2009.
Hanssen, R.F. Radar Interferometry: Data Interpretation and Error Analysis. Kluwer Academic Publishers.
308p. 2001.
Hanssen, R.F. Satellite radar interferometry for deformation monitoring: a priori assessment of feasibility and
accuracy. International Journal of Applied Earth Observation and Geoinformation 6, pp 253-260. 2005.
Herrera G., Davalillo J. C., Mulas J., Cooksley G., Monserrat O. and Pancioli V. (2009b) - Mapping and
monitoring geomorphological processes in mountainous areas using PSI data: Central Pyrenees case study.
Nat. Hazards Earth Syst. Sci., 9, 15871598, 2009.
Herrera, G., Fernandez, J.A., Tomas, R., Cooksley, G., Mulas, J., (2009a). Advanced interpretation of
subsidence in Murcia (SE Spain) using A-DInSAR data - modelling and validation. Nat. Hazards Earth Syst.
Sci.. 9, 647-661.
Hilley, G.E., Burgmann, R., Ferretti, A., Novali, F., Rocca, F., (2004). Dynamics of slow-moving landslides from
Permanent Scatterer analysis. Science. 304 (5679), 1952-1955.
Hooper A. and Pedersen R. (2007) - Deformation due to Magma Movement and Ice Unloading at Katla
Volcano, Iceland, Detected by Persistent Scatterer InSAR. Proceedings ENVISAT Symposium, Montreux
2007.
Hooper A., Segall P., Zebker H. (2007) - Persistent Scatterer InSAR for Crustal Deformation Analysis, with
Application to Volcn Alcedo, Galpagos. J. Geophys. Res., 112, B07407, doi:10.1029/2006JB004763.
Hooper, A., Zebker, H., Segall, P., Kampes, B. (2004). A new method for measuring deformation on volcanoes
and other natural terrains using InSAR persistent scatterers. Geophys. Res. Lett.. 31 (23), 5, 2004.
Jagdhuber, T., Hajnsek, I., Papathanassiou, K.P., and Bronstert, A., Soil Moisture Estimation Using A MultiAngular Modified Three-Component Polarimetric Decomposition, Proc. of the IEEE Geosci. and Remote
Sens. Symposium, IGARSS09, Cape Town, 2009.
Lanari R., Casu F., Manzo M., Lundgren P. Application of the SBAS-DInSAR technique to fault creep: a case
study of the Hayward fault, California. REMOTE SENSING OF ENVIRONMENT, vol. 109, pp.20-28. 2007.
Lanari, R., Berardino, P., Borgstrm, S., Del Gaudio, C., De Martino, P., Fornaro, G., Guarino, S., Ricciardi,
G.P., Sansosti, E., Lundgren, P. The use of IFSAR and classical geodetic techniques for caldera unrest
episodes: application to the Campi Flegrei uplift event of 2000. J. Volcanol. Geotherm. Res.. 133, 247-260.
2004.
Lauknes, T.R. Long-Term Surface Deformation Mapping using Small-Baseline Differential SAR Interferograms.
Faculty of Science, University of Troms. 88p. 2004.

SafeLand - FP7
Page 380 of 401
6. References
Rev. No: 2
Date: 2012-02-15
Le Moulic, S., Raucoules, D., Carnec, C., King, C. A least squares adjustment of multi-temporal InSAR data:
application to the ground deformation of Paris. Photogrammetric Engineering and Remote Sensing, 71 (2),
pp. 197-204. 2005.
Le Toan, T. Beaudoin, A. and Guyon, D. Relating Forest Biomass to SAR Data. IEEE Trans. Geosci. Remote
Sens., Vol. 30, No. 2, pp. 403-411. 1992.
Lemoine, G.F., Hoekman, D.H., and Sieber, A.J. Polarimetric Contrast Classification of Agricultural Fields
using MESTRO-1 AIR-SAR Data. Int. J. Remote Sensing, Vol. 15, No. 14, pp. 2851-2869. 1994.
Leva, D., Nico, G., Tarchi, D., Fortuny-Guasch, J., Sieber, A.J. Temporal analysis of a landslide by means of a
ground-based SAR Interferometer. IEEE Trans. Geosci. Remote Sens., vol. 41, no 4, Part 1, pp.745 752.
April 2003.
Lin, D.S., Wood, E.F., Beven, K. and Saatchi, S. Soil Moisture Estimation over Grass-covered Areas Using
AIRSAR. Int. J. Remote. Sensing, Vol. 15, No. 11, pp. 2323-2343. 1994.
Luzi, G., Pieraccini, M., Mecatti, D., Noferini, L., Guidi, G., Moia, F., Atzeni C. Ground-Based Radar
Interferometry for Landslides Monitoring: Atmospheric and Instrumental Decorrelation Sources on
Experimental Data. IEEE Trans. Geosci. Remote Sens., vol. 42, no 11, pp 2454 2466. November 2004.
Luzi, G., Pieraccini, M., Mecatti, D., Noferini, L., Macaluso, G., Galgaro, A., Atzeni, C. Advances in ground
based microwave interferometry for landslide survey: a case study. International Journal of Remote
Sensing, Vol. 27, No. 12 / 20, pp. 2331 2350. June 2006.
Luzi, G., Pieraccini, M., Mecatti, D., Noferini, L., Macaluso, G., Tamburini, A., Atzeni, C. Monitoring of an
Alpine Glacier by Means of Ground-Based SAR Interferometry. Geoscience and Remote Sensing Letters ,
Vol. 4, No 3, pp. 495-499. July 2007.
Luzi, G., Noferini, L., Mecatti, D., Macaluso, G., Pieraccini, M., Atzeni, C., Schaffhauser, A., Fromm, R.,
Nagler, T. Using a Ground-Based SAR Interferometer and a Terrestrial Laser Scanner to Monitor a SnowCovered Slope: Results From an Experimental Data Collection in Tyrol (Austria). IEEE Transaction on
Geoscience and Remote Sensing, vol.47, no.2 , February 2009, Page(s): 382-393. 2009.
Luzi, G. Ground based SAR interferometry: a novel tool for Geoscience. In: Geoscience and Remote Sensing,
New Achievements, edited by: P.Imperatore & D. Riccio, ISBN: 978-953-7619-97-8. In print.
Lyons, S., and D. Sandwell. Fault creep along the southern San Andreas from interferometric synthetic aperture
radar, permanent scatterers, and stacking, J. Geophys. Res., 108(B1), 2047, doi:10.1029/ 2002JB001831.
2003.
Massonnet, D., Feigl, K.L. Radar interferometry and its application to changes in the Earths surface. Reviews
of Geophysics 36. pp 441-500. 1998.
Mensa, D. High Resolution Radar Cross-Section Imaging. Artech House, Boston. 1991.
Metternicht, G., Hurni, L., Gogu, R. Remote sensing of landslides: An analysis of the potential contribution to
geo-spatial systems for hazard assessment in mountainous environments. Remote Sens. Environ.. 98 (2),
284-303. 2005.
Michoud, C., Jaboyedoff, M., Pedrazzini, A. Application de linterfromtrie diffrentielle Radar dans la
surveillance du glissement de La Frasse, Suisse. Mmoire de Master s Sciences de lEcole Lmanique des
Sciences de la Terre et de lEnvironnement. 95p. 2009.
Mora, O., Mallorqui, J., Broquetas, A. Linear and non-linear terrain deformation maps from a reduced set of
interferometric SAR images. IEEE Trans. Geosci. Remote Sens.. 41 (10), 2243-2253. 2003.
Musson, R.M.W., Haynes, M., Ferretti, A. Space-based tectonic modelling in subduction areas using PSInSAR.
Seismol. Res. Lett.. 26 (9), 1913-1926. 2004.
Noferini, L., Pieraccini, M., Mecatti, D., Luzi, G., Tamburini, A., Broccolato, M., Atzeni, C. Permanent
scatterers analysis for atmospheric correction in Ground Based SAR Interferometry. IEEE Trans. Geosci.
Rem. Sens., vol. 43, no 7, pp. 1459-1471. 2005.
Norland, R. Differential Interferometric Radar for Mountain Rock Slide Hazard Monitoring. Technical report,
4p. Available on the knes/Tafjord Beredskap IKS website: http://www.aknes-tafjord.no. 2006.
Oh, Y., Sarabandi, K. and Ulaby, F.T. An Empirical Model and an Inversion Technique for Radar Scattering
from Bare Soil Surfaces. IEEE Trans. Geosci. Remote Sens., Vol. 30, No. 2, pp. 370-381. March 1992.
Peckham, G. Instrumentation and measurements in atmospheric remote sensing. Reports on Progress in Physics
54, pp. 531-577. 1991.

SafeLand - FP7
Page 381 of 401
Deliverable 4.1
Pepe A., Lanari L. On the Extension of the Minimum Cost Flow Algorithm for Phase Unwrapping of
Multitemporal Differential SAR Interferograms. IEEE Trans. Geosci. Remote Sens., vol. 44, no. 9, pp. 23742383. 2006.
Pieraccini, M., Casagli, N., Luzi, G., Tarchi, D., Mecatti, D., Noferini, L., Atzeni, C. Landslide monitoring by
ground-based radar interferometry: a field test in Valdarno (Italy). International Journal of Remote
Sensing, 24 6, pp. 1385-1391. 2002.
Pipia, L., Fabregas, X., Aguasca, A., Lopez-Martinez, C., Mallorqui, J., Mora, O. A Subsidence Monitoring
Project using a Polarimetric GB-SAR Sensor. The 3rd Int. Workshop POLinSAR 2007 Frascati, Italy on 2226 January 2007.
Prati, C. Introduction to SAR and InSAR. 4e Confrence canadienne sur les gorisques : des causes la gestion.
Workshop, Universit Laval. 2008.
Preissner, J. The influence of the atmosphere on passive radiometric measurements. Proceedings of the
Symposium on Millimeter and Submillimeter Wave Propagation Circuits: AGARD Conference, VOl. 245,
pp. 1-13. 1978.
Reale, D., Pascazio, V., Schirinzi, G., Serafino, F. 3D Imaging of Ground based SAR Data. Geoscience and
Remote Sensing Symposium, 2008 IGARSS2008. IEEE International Volume 4, 7-11 July 2008.
Righini, G., Casagli N., Del Ventisette, C., Costantini, M., Malvarosa, F., Minati, F. Space-borne SAR Analysis
for Landslides Mapping in the Framework of the PREVIEW Project. Proc. 1st World Landslide Forum,
Tokyo, Japan. 26-27. 2008.
Rignot, E., Zimmerman R., and van Zyl, J.J. Spaceborne Applications of P. Band Imagin Radars for Measuring
Forest Biomass. IEEE Trans. Geosci. Remote Sens., Vol. 33, pp. 1162-1169. 1995.
Rudolf, H., Leva, D., Tarchi, D., Sieber, A.J., A mobile and versatile SAR system. Proc. IGARSS99, Hamburg,
pp. 592594. 1999.
Salvi, S., Atzori, S., Tolomei, C., Allievi, J., Ferretti, A., Rocca, F., Prati, C., Stramondo, S., Feuillet, N. Inflation
rate of the Colli Albani volcanic complex retrieved by the permanent scatterers SAR interferometry
technique. Geophys. Res. Lett.. 31, 1-4. 2004.
Sansosti, E., Berardino, P., Manunta, M., Serafino, F., Fornaro, G. Geometrical SAR Image Registration. IEEE
Trans. Geosci. Remote Sens, vol. 44, no.10, pp. 2861-2870. 2006.
SARMap courses. Synthetic Aperture Radar, SARScape The SAR Guidebook. 268p. 2008
Shi, J., Wang, J., Hsu, A. Y., ONeill, P.E. and Engman, E.T. Estimation of Bare Surface Soil Moisture and
Surface Roughness Parameter Using L-band SAR Image Data. IEEE Trans. Geosci. Remote Sens., Vol. 35,
No. 5, pp. 1254-1266. 1997.
Shi, J., and Dozier, J. Estimation of Snow Water Equivalence Using SIR-C/X-SAR, Part I: Inferring Snow
Density and Subsurface Properties. IEEE Trans. Geosci. Remote Sens., Vol. 38, No. 6, pp. 2465-2474,
November 2000.
Squarzoni, C. Mesure des champs de dplacement de surface et modlisation numrique des glissements de
terrain. Diplme de doctorat de lUniversit Claude Bernard Lyon 1 et lUniversita degli Studi di Padova.
221p. 2003.
SRTM factsheet. http://www2.jpl.nasa.gov/srtm/factsheets.html. 2005.
Stramondo, S., Bozzano, F., Marra, F., Wegmuller, U., Cinti, F.R., Moro, M., Saroli, M. Subsidence induced by
urbanisation in the city of Rome detected by advanced InSAR technique and geotechnical investigations.
Remote Sens. Environ.. Vol. 112, 3160-3172. 2008.
Strozzi, T., Matzler, C. Backscattering Measurements of Alpine Snowcovers at 5.3 GHz and 35 GHz. IEEE
Trans. on Geoscience and Remote Sensing, Vol. 36, No. 3, pp. 838-848. May 1998.
Strozzi, T., Wegmller, U., Mtzler, C. Mapping Wet Snowcovers with SAR Interferometry. Int. J. Remote Sens.,
Vol. 20, No. 12, pp. 2395-2403. 1999.
Strozzi T., Wegmuller U., Keusen H.R., Graf K., Wiesmann A. Analysis of the terrain displacement along a
funicular by SAR interferometry. IEEE Geoscience and Remote Sensing Letters, January 2006, vol. 3, no. 1,
pp.15-18. 2006.
Tarchi, D., Ohlmer, E., Sieber, A.J. Monitoring of structural changes by radar interferometry. Res. Nondestruct.
Eval. 9, 213 225. 1997.

SafeLand - FP7
Page 382 of 401
6. References
Rev. No: 2
Date: 2012-02-15
Tarchi, D., Rudolf, H., Luzi, G., Chiarantini, L., Coppo, P., Sieber, A.J. SAR interferometry for structural
changes detection: a demonstration test on a dam. Proc. IGARSS99, Hamburg, pp. 15221524. 1999
Tarchi, D., Casagli, N., Fanti, R., Leva, D., Luzi, G., Pasuto, A., Pieraccini, M., Silvano, S. Landside Monitoring
by Using Ground-Based SAR Interferometry: an example of application to the Tessina landslide in Italy,
Engineering Geology 68, pp.15-30. 2003a.
Tarchi,D., Casagli, N., Moretti, S., Leva, D., Sieber, A.J. Monitoring landslide displacements by using groundbased radar interferometry: Application to the Ruinon landslide in the Italian Alps, J. Geophys. Res., 108,
10.1-10.14. 2003b.
Tizzani, P., Berardino, P., Casu, F., Euillades, P., Manzo, M., Ricciardi, G.P., Zeni, G., Lanari, R. Surface
deformation of Long Valley caldera and Mono Basin, California investigated with the SBAS-InSAR
approach. Remote Sens. Environ.. 108, 277-289. 2007.
Tsang, L., J.A.Kong, and R.T.Shin, Theory of Microwave Remote Sensing, New York: John Wiley, 1985.
Ulaby, F.T., R.K.Moore, and A.K.Fung, Microwave Remote Sensing, vol. II, Reading, MA: Addison-Wesley,
1982.
Vilardo, G., Ventura, G., Terranova, C., Matano, F., Nard, S. Ground deformation due to tectonic,
hydrothermal, gravity, hydrogeological, and anthropic processes in the Campania Region (Southern Italy)
from Permanent Scatterers Synthetic Aperture Radar Interferometry. Remote Sens. Environ.. 113 (1), 197212. 2009.
Wasowski J., D. Casarano, C. Lamanna, F. Bovenga, D. Conte, R. Nutricato, P. Berardino, M. Manzo, A. Pepe,
G. Zeni, R. Lanari, A. Refice. A comparative analysis of DInSAR results achieved by the SBAS and SPINUA
techniques: the Maratea valley case study, Italy. Proc. Envisat Symposium 2007, Montreux, Switzerland
2327 April 2007 (ESA SP-636, July 2007). 2007.
Wegmuller U., Werner C., Strozzi T., Wiesmann A. Monitoring mining induced surface deformations. Proc. of
IGARSS 2004, Anchorage (USA). Sept. 2004.
Wegmuller U., Werner C., Strozzi T., Wiesmann A. Application of SAR interferometric techniques for surface
deformation monitoring. 3rd IAG / 12th FIG Symposium, Baden, May 22-24, 2006.
Werner C., Wegmuller U., Strozzi T., Wiesmann A. Interferometric Point Target Analysis for deformation
mapping. Proc. of IGARSS 2003, Toulouse (Francia), Luglio 2003. 2003a.
Werner C., Wegmuller U., Strozzi T., Wiesmann A. Interferometric Point Target Analysis with JERS L-band
data. Proc. of IGARSS 2003, Toulouse (Francia), Luglio 2003. 2003b.
Werner, C., Strozzi, T., Wesmann, A., Wegmuller, U. Gammas portable radar interferometer. 13th FIG
Symposium on Deformation Measurements and analysis, LNEC, Lisbon May 12-16 2008. 2008.
Wiley, C.A. Pulsed Doppler Radar Methods and Apparatus. United States Patent No. 3196436. 1954.
Woodhouse, I.H. Introduction to Microwave Remote Sensing. CRC Taylor & Francis, 370p. 2006.
Zebker, H.A., Rosen, P.A., Hensley, S. Atmospheric effects in interferometric synthetic aperture radar surface
deformation and topographic maps. Journal of Geophysical research 102. pp 7547-7563. 1997.
6.5. SPECIFIC REFERENCES FOR THE CHAPTER 2D

Agnesi, V., Camardab, M., Conoscentia, C., Di Maggio, A., Dilibertoc, I., Madoniac, P., Rotiglianoa, E., 2005.
A multidisciplinary approach to the evaluation of the mechanism that triggered the Cerda landslide (Sicily,
Italy). Geomor., 65, 101-116.
Almendros, J. Ibez, G. Alguacil, E. Del Pezzo, 1999. Array analysis using circular-wave-front geometry: an
application to locate the nearby seismo-volcanic source. Geophys. J. Int., 136,159-170.
Amitrano D., Gaffet S., Malet J.P., Maquaire O., 2007. Understanding mudslides through micro-seismic
monitoring: the Super-Sauze (South-East French Alps) case study. Bull. Soc. Gol. Fr., 178 (2), 149-157.
Amitrano D., Grasso J.-R., Senfaute G., 2005. Seismic precursory patterns before a cliff collapse and criticalpoint phenomena. Geophysical Research Letters, 32, 8, L08314, doi:10.1029/2004GL022270.
Amitrano D., Senfaute G., Grasso J.R., Got J.L., Gaffet S., Clement C., 2004. Potential of the seismic monitoring
for the understanding of gravitational instability. AGU Fall Meeting, paper nH44A-06, San Francisco.

SafeLand - FP7
Page 383 of 401
Deliverable 4.1
Arattano M., 1999. On the use of seismic detectors as monitoring and warning systems for debris flows. Natural
Hazards, 20, 197-213.
Arattano M., Marchi L., 2005. Measurements of debris-flow velocity through cross-correlation of
instrumentation data. Natural Hazards and Earth System Sciences, 5, 137-142.
Bangs, N.L.B., Moore, G.F., Gulik, S.P.S., Pangborn, E.M., Tobin, H.J., Kuramoto, S., Taira, A., 2009: Broad,
weak regions of the Nankai Megathrust and implications for shallow coseismic slip. Earth and Planetary
Science Letters 284: 44-49.
Barnhardt, W.A. and Kayen, R.E., 2000. Radar structure of earthquake-induced coastal landslides in
Anchorage, Alaska. Environ. Geosciences, 7, 38-45.
Barton, N., 2006. Rock Quality, Seismic Velocity, Attenuation and Anisotropy. Taylor and Francis, UK and
Netherlands, 729p.
Batayneh, A.T. and Al-Diabat, A.A., 2002. Application of a two-dimensional electrical tomography technique for
investigating landslides along the Amman-Dead Sera Highway, Jordan. Env. Geol., 42, 399-403.
Bessason B., Eiriksson G., Thorarinsson O., Thorarinsson A., Einarsson S., 2007. Automatic detection of
avalanches and debris flows by seismic methods. Journal of Glaciology, 53, No 182, 461-472.
Bichler A., Bobrowsky P., Best M., Douma M., Hunter J., Calvert T., Burns R., 2004. Three-dimensional
mapping of a landslide using a multi-geophysical approach: the Quesnel Forks landslide. Landslides, 1 (1),
29-40.
Blha P., Mrlina J., Nevara J., 1998. Gravimetric investigation of slope deformations. Expl. Geophys, RemoteSens. and Env. J., 1, 21-24.
Bolt, B. A., 1982. Inside the Earth. San Francisco: W.H. Freeman.
"Bondevik, S.,Lvholt, F., Harbitz, C., Mangerud, J., Dawson A., Svendsen, J.I., 2005: The Storegga Slide
tsunami; comparing field observations with numerical simulations. Marine and Petroleum Geology, 22:
195-208. "
Borle, Y., and Jaboyedoff, M., 2007. Phnomnes rosifs dans le bassin versant de Lourtier (Valais, Suisse):
tude dune instabilit rocheuse et tablissement dun bilan sdimentaire laide dune cartographie de
terrain dtaille : mthodologie et application. Master thesis in Engineering Geology. IGAR-UNIL.
Bouillon, A.-L., 2005. Geophysics for geohazards on land: state-of-the-art, case studies and education: Dipl.
Eng. Geophysics, University of Strasbourg, ICG Report 2005-T1-1, NGI Report 20051108-1,
http://www.geohazards.no/projects/geophys/ICG_ALB_Report.pdf
Brodsky E. E., 2003. Landslide basal friction as measured by seismic waves. Geophysical Research Letters, 30,
No. 24, 1-5.
Brckl E., Mertl S., 2006. Seismic monitoring of deep-seated mass movements. In: Disaster Mitigation of Debris
Flows. Slope Failures and Landslides, 571-580.
Bruno, F. and Marillier, F., 2000. Test of high-resolution seismic reflection and other geophysical techniques on
the Boup landslide in the Swiss Alps. Surveys in Geophys., 21, 333348.
Burtin A., Bollinger L., Cattin R., Vergne J., Nbelek J. L., 2009. Spatio-temporal sequence of Himalayan debris
flow from analysis of high-frequency seismic noise. J. Geophys. Res., 114, 1-15.
"Butler, D., 2005. Near Surface Geophysics, SEG investigations in geophysics series; no. 13, Ed. Butler, ISBN 156080-130-1."
Cadman J.D., Goodman R.E., 1967. Landslide noise. Science, 158, 1182-1184.
Caris, J.P.T. and Van Asch, Th.W.J, 1991. Geophysical, geotechnical and hydrological investigations of a small
landslide in the French Alps. Eng. Geol., 31, 249-276.
Cha, Y., Kang, J., Jo, C.H., 2006. Application of linear-array microtremor surveys for rock mass classification
in urban tunnel design. Expl. Geophys., 37, 108-113.
Chambers, J.E., Meldrum, P.I., Gunn, D.A., Wilkinson, P.B., Kuras, O., Weller, A.L., Ogilvy, R.D., 2009.
Hydrogeophysical Monitoring of Landslide Processes Using Automated Time-Lapse Electrical Resistivity
Tomography (ALERT). Extended abstract, Near Surface Conference, Dublin.
Cole S. E., Cronin S. J., Sherburn S. and Manville V., 2009. Seismic signals of snow-slurry lahars in motion: 25
September 2007, Mt Ruapehu, New Zealand. Geophysical Research Letters, 36, 1-5.

SafeLand - FP7
Page 384 of 401
6. References
Rev. No: 2
Date: 2012-02-15
Cole S. E., Cronin S. J., Sherburn S., Manville V., 2008. Seismic signals associated with the 2007 lahars of Mt.
Ruapehu, New Zealand. AGU, San Francisco, USA, Fall Meeting 2008, Abstract-based Poster.
Cruden, D.M. and Varnes, D.J., 1996. Landslide types and processes. In: Landslides investigation and
mitigation, Transportation Research Board, Special Report 247, National Academy of Sciences. Washington
DC., USA, 36-75.
Dan, G., Sultan, N. & Savoye, B. 2007: The 1979 Nice harbour catastrophe revisited: Trigger mechanism
inferred from geotechnical measurements and numerical modeling. Marine Geology 245, 40-64.
Dasios, A., McCann, C., Astin, T.R., McCann, D.M., Fenning, P., 1999. Seismic imaging of the shallow
subsurface: shear-wave case histories. Geophysical Prospecting, 47, 565-591.
De Angelis S., Bass V., Hards V., Ryan G., 2007. Seismic characterization of pyroclastic flow activity at
Soufrire Hills Volcano, Montserrat, 8 January 2007. Nat. Hazards Earth Syst., 7, 467-472.
Del Gaudio, V.,Wasowski, J., Pierri, P., Mascia, U., Calcagnile, G., 2000. Gravimetric study of a retrogressive
landslide in Southern Italy. Surveys in Geophys., 21, 391-399.
Demoulin, A., Pissart, A., Schroeder, C., 2003. On the origin of late Quaternary palaeolandslides in the Lige (E
Belgium) area. Int. J. Earth Sci. (Geol Rundsch), 92, 795-805.
Deparis J., Jongmans D., Cotton F., Baillet L., Thouvenot F., Hantz D., 2008. Analysis of rock-fall and rock-fall
avalanche seismograms in the French Alps. Bull. of the Seism. Soc. of America, 98-4, 1781-1796.
Dixon N., Hill R., Kavanagh J., 2003. Acoustic emission monitoring of slope instability: development of an
active waveguide system. Geotech. Eng., 156, 83-95.
Dixon N., Kavanagh J., Hill R., 1996. Monitoring landslide activity and hazard by acoustic emission. J. Geol.
Soc. China, 39, 437-464.
Eberhardt E., Spillmann T., Maurer H., Willenberg H., Loew S., Stead D., 2004. The Randa Rockslide
Laboratory: Establishing brittle and ductile instability mechanisms using numerical modelling of
microseismicity. Proc. 5th Int. Symp. on Landslides, Rio de Janeiro, Brazil Balkema, Rotterdam, 481-487.
Esposito A. M., Giudicepietro F., Scarpetta S., DAuria L., Marinaro M. And Martini M., 2006. Automatic
Discrimination among Landslide, Explosion-Quake, and Microtremor Seismic Signals at Stromboli Volcano
Using Neural Networks. Bulletin of the Seismological Society of America, 96, No. 4A, 1230-1240.
Ferrucci, F., Amelio, M., Sorriso-Valvo, M., Tansi, C., 2000. Seismic prospecting of a slope affected by deepseated gravitational slope deformation: the Lago Sackung, Calabria, Italy. Eng. Geol., 57, 53-64.
Foti, S., 2000, Multistation methods for geotechnical characterization using surface waves: PhD thesis,
Politecnico di Torino, Italy.
French Association for quality of applied geophysics, Applied Geophysics, Code of Practice, 1992.
http://www.agapqualite.org/
Frey-Martinez, J. 2009: 3D Seismic Interpretation of Mass Transport Deposits: Implications for Basin Analysis
and Geohazard Evaluation. In D.C. Mosher et al. (eds.), Submarine Mass Movements and Their
Consequences, Advances in Natural and Technological Hazards Research, Vol 28: 667-684.
Gallipoli, M., Lapenna, V., Lorenzo, P.,Mucciarelli, M., Perrone, A., Piscitelli, S., Sdao, F. 2000. Comparison
of geological and geophysical prospecting techniques in the study of a landslide in southern Italy. European
J. Env. Eng. Geophys., 4, 117-128.
Ganerd, G.V., Grneng, G., Aardal, I.B., Kveldsvik, V., 2007. Logging of drill cores from seven boreholes at
knes, Stranda municipality, Mre and Romsdal County. Geological Survey of Norway, report 2007.020.
Ge M., 2005. Efficient mine microseismic monitoring. International Journal of Coal Geology, 44-56.
Ghose, R., 2004. Model-based integration of seismic and CPT data to derive soil parameters. In: (Ed.) Proc.
10th European Meeting of Environmental and Engineering Geophysics, Utrecht, The Netherlands, EAGE
Publications , Houten, The Netherlands, paper B019, 4 p.
Gibert, D. and Pessel, M., 2001. Identification of sources of potential fields with the continuous wavelet
transform: Application to self-potential profiles. Geophys. Res. Lett., 28, 1863-1866.
Glade, T., Stark, P., Dikau, R., 2005. Determination of potential landslide shear plane depth using seismic
refraction. A case study in Rheinhessen, Germany. Bull. Eng. Geol. Environ., 64,151-158.
Goodman R. E., Blake W., 1965. An investigation of rock noise in landslides and cut slopes. In: Rock Mechanics
and Engineering geology Supplementum II, Safety in Rock Engineering, 15th Symposium of the Austrian
Regional group (i. f.) of the International Society for Rock Mechanics, Salzburg, 24-25 September 1964.
SafeLand - FP7
Page 385 of 401
Deliverable 4.1
Grandjean, G. 2006. A seismic multi-approach method for characterizing contaminated sites. Journal of Applied
Geophysics., 58, 2, 87-98.
Grandjean, G., and Bitri, A., 2006. 2M-SASW: Multifold multichannel seismic inversion of local dispersion of
Rayleigh waves in laterally heterogeneous subsurfaces: application to the Super-Sauze earthflow, France.
Near Surface Geophys., 367-375.
Green, M., and Cunningham, D. 1998: Seabed imaging techniques. Hydro International, 2(1): 4 pp.
Hack, R. , 2000. Geophysics for slope stability. Surveys in Geophys., 21, 423-448.
Haflidason H., Sejrup H.P., Nygrd A., Mienert J., Bryn .P, Lien R., Forsberg C.F., Berg K., Masson D. 2004:
The Storegga Slide: architecture, geometry and slide development. Mar Geol 213:201234.
Hardy H. R. Jr., 1992. Laboratory studies relative to the development of mechanical waveguides for acoustic
emission monitoring of geological structures: Int. J. of Non-destructive Testing and Diagnostics, XIII, 2, 3238.
Hardy, H.R. and E.J. Kimble. 1991. Application of high-frequency AE/MS techniques to rock slope monitoring.
Vth Conf. AE/MS Geol. Str. and Mat., edited by Hardy, Trans Tech Publication, Germany, The
Pennsylvania State University, 457-477.
Harp E. L. and Jibson R. W., 1995. Seismic instrumentation of landslides: building a better model of dynamic
landslide behavior. BSSA, 85, No. 1, 93-99.
Havenith, H.B., Jongmans, D., Abdrakmatov, K., Trefois, P., Delvaux, D., Torgoev, A. 2000. Geophysical
investigations on seismically induced surface effects, case study of a landslide in the Suusamyr valley,
Kyrgyzstan. Surveys in Geophys., 21, 349-369.
Helmstetter A., Garambois S., 2009. Seismic monitoring of Schilienne Rockslide (French Alps): analysis of
seismic signals and their correlation with rainfalls. J. Geophys. Res., submitted paper.
Helmstetter, A., and Garambois, S., 2010. Seismic monitoring of Schilienne Rockslide (French Alps): analysis
of seismic signals and their correlation with rainfalls, J. Geophys. Res., 115, F03016,
doi:10.1029/2009JF001532.
Hermanns, R.L., Blikra, L.H., Naumann, M., Nilsen, B., Panthi, K.K., Stromeyer, D., Longva O., 2006: Examples
of multiple rock-slope collapses from Kfels (tz valley, Austria) and western Norway. Engineering
Geology, 83: 94-108.
Herrmann, R.B., 2002. Computer programs in seismology. Department of Earth and Atmospheric Sciences,
Saint Louis University.
Huang C.-J., Shieh C.-L., Yin H.-Y., 2004. Laboratory study of the underground sound generated by debris
flows. J. Geophys. Res., 109, 1-11.
Huang, C.-J., Yin H.-Y., Chen C.-Y., Yeh C.-H., Wang C.-L. 2007. Ground vibrations produced by rock motions
and debris flows, J. Geophys. Res., 112, F02014, doi:10.1029/2005JF000437.
Huggel C., Caplan-Auerbach J., Waythomas C. F., Wessels R. L., 2007. Monitoring and modeling ice-rock
avalanches from ice-capped volcanoes: A case study of frequent large avalanches on Iliama Volcano,
Alaska. J. of Volcanology and Geothermal res., 168, 114-136.
Hughes-Clarke, J.E., Mayer, L.A., and Wells, D.E. 1996: Shallow-water imaging multibeam sonars: A new tool
for investigating sea floor processes in the coastal zone and on the continental shelf. Marine Geophysical
Researches, 18: 607-629.
Hunter, J. A., Burns, R. A., Good, R. L., Oullan, S. E., Pugin, A., Crow, H., 2010. Near-surface geophysical
techniques for geohazards investigations: some Canadian examples, The Leadind Egde, August, 964-975.
Internationational Centre for Geohazards (ICG), Oslo, Geophysics for Geohazards Theme.
http://www.geohazards.no/projects/geophys.htm
Itakura M., Kaburaki H. And Arakawa C., 2005. Branching mechanism of intergranular crack propagation in
three dimensions. Physical Review E, 71, 055102(R) 1-4.
Jeannin, M., Garambois, S., Gregoire, S., Jongmans, D., 2006. Multi-configuration GPR measurements for
geometrical fracture characterization in limestone cliffs (Alps). Geophysics, 71, 885-892.
Jongmans, D., and Garambois, S., 2007. Geophysical investigation of landslides: A review. Bulletin Socit
Gologique de France 178, 2. Pp 101-112.
Jongmans, D., Hemroulle, P., Demanet, D., Renardy, F., Vanbrabant, Y., 2000. Application of 2D electrical and
seismic tomography techniques for investigating landslides. European J. Env. Eng. Geophys., 5, 75-89.
SafeLand - FP7
Page 386 of 401
6. References
Rev. No: 2
Date: 2012-02-15
Joswig M., 2005. Nanoseismic Monitoring: Method and First Applications, Elaborate Tutorial on Nanoseismic
Monitoring.
Kearey, Ph., Brooks, M., Hill I., 2002. An introduction to geophysical exploration. Wiley-Blackwell Publishing,
262p.
Kennedy, B.A., and K.E. Niermeyer, 1971. Slope monitoring systems used in the prediction of a major slope
failure at the Chuquicamata Mine, Chile. In: Proceedings of the Conference on Planning Open Pit Mines,
Balkema, Johannesburg, 215-225.
Kishimura K. and Izumi K., 1997. Seismic signals induced by snow avalanche flow. Natural Hazards, 15, 89100.
Kndel, K., G. Lange, H.-J. Voigt, 2007. Environmental geology, Handbook for field methods and case studies,
Ed: Bundesanstalt fr Geowissenschaften und Rohstoffe, ISBN 978-3-540-74669-0 Springer Berlin
Heidelberg New York, 1357p.
Koerner R. M., Mc Cabe W. M., Lord A. E., 1981. Acoustic emission behaviour and monitoring of soils: Acoustic
Emission in Geotechnical Practice, ASTM STP 750, American Society for Testing and Materials, 93-141.
Kolesnikov Y.I., Nemirovich-Danchenko M.M., Goldin S.V., Seleznev V.S., 2003. Slope stability monitoring from
microseismic field using polarization methodology. Nat. Haz. Earth Sys. Sc., 3, 515-521.
Kousteni A., Hill R., Dixon N., Kavanagh J., 1999. Acoustic emission technique for monitoring soil and rock
slope instability. Proc. Int. Symp. On Slope Stability Engineering: Geotechnical and Geoenvironmental
Aspects, Matsuyama, Shikoku, Japan Balkema, Rotterdam, 151-156.
LHeureux, J.-S., Hansen, L., Longva, O., Emdal, A., Grande, L. 2010: A multidisciplinary study of submarine
slides at the Nidelva fjord delta, Mid-Norway - Implications for hazard assessments. Norwegian Journal of
Geology.
La Rocca M., Galluzzo D., Saccorotti G. Tinti S., Cimini G. B. and Del Pezzo E., 2004. Seismic signals
associated with landslides and with tsunami at Stromboli Volcano, Italy. Bulletin of Seismological Society of
America, 94, No. 5, 1850-1867.
Lapenna, V., Lorenzo, P., Perrone, A., Piscitelli, S., Rizzo, E., Sdao F., 2005. 2D electrical resistivity imaging of
some complex landslides in Lucanian Apennine chain, southern Italy. Geophysics, 70, B11-B18.
Lapenna, V.,Lorenzo, P., Perrone, A., Piscitelli, S., Rizzo, E., Sdao F., 2003. High-resolution geoelectrical
tomographies in the study of the Giarrossa landslide (Potenza, Basilicata). Bull. Eng. Geol. Env., 62, 25968.
Lastras, G., Canals, M., Urgeles, R., Hughes-Clarke, J. E. & Acosta, J. 2004: Shallow slides and pockmark
swarms in the Eivissa Channel, western Mediterranean sea. Sedimentology 51, 114.
Lavergne, M., 1986. Mthodes Sismiques, Ed : TECHNIP, Paris, 207p.
Lavigne F., 2000. Lahars hazard micro-zonation and risk assessment in Yogyakarta city, Indonesia. GeoJournal
, 49, No 2, 131-138.
Lebourg, T., Binet, S., Tric, E., Jomard H., El Bedoui, S., 2005. Geophysical survey to estimate the 3D sliding
surface and the 4D evolution of the water pressure on part of a deep seated landslide. Terra Nova, 17,
399-406.
Lecomte, I., Thollet, I., Juliussen, H., Hamran, S.-E., 2008. Using geophysics on a terminal moraine damming a
glacial lake: the Flatbre debris flow case, Western Norway, Advances in Geosciences, 14, 301-207, ICG
contribution 191.
Lin, C.-P., and Chang T.-S., 2004. Multi-station analysis of surface wave dispersion. Soil Dynamics and
Earthquake Engineering 24, 877886.
Locat J. & Lee, H.J. 2002: Submarine landslides: advances and challenges. Canadian Geotechnical Journal 39,
193-212.
Lockner D.A., 1993. The role of acoustic emission in the study of rock fracture. Int. J. Rock Mech. Min. Sci. and
Geomech. Abstr., 30, No 7, 883-899.
Loke, M. H., 1999, Electrical imaging surveys for environmental and engineering studies: A practical guide to
2-D and 3-D surveys, www.geometrics.com, copyright by Dr. M.H. Loke.
Longva, O., Janbu, N., Blikra, L.H. & Boe, R. 2003: The 1996 Finneidfjord slide: seafloor failure and slide
dynamics. In: Locat, J. & Mienert, J. (eds.): Submarine Mass Movements and Their Consequences, 531538. Dordrecht, Netherlands: Kluwer academic Publishers.

SafeLand - FP7
Page 387 of 401
Deliverable 4.1
Lord A. E., Fisk C. L., Koerner R. M., 1982. Utilisation of steel rods as AE waveguides: J. Geotech. Eng. Div.,
Proc. ASCE, 108, 300-305.
Marchi L., Arattano M., Deganutti A. M., 2002. Ten years of debris-flow monitoring in the Moscardo Torrent
(Italian Alps), Geomorphology, 46, 1-17.
Marescot, L., Monnet, R., Chapellier, D., 2008. Resistivity and induced polarization surveys for slope instability
studies in the Swiss Alps. Engineering Geology 98, pp 1828.
Maurer, H., Spillmann, T., Heincke, B., Hauck, C., Loew, S., Spingman, M., Green, A., 2010. Geophysical
characterization of slope instabilities, First Break, 28, August, 53-61.
Mauritsch, H.J., Seiberl, W., Arndt, R., Romer, A., Schneiderbauer, K., Sendlhofer, G.P. 2000. Geophysical
investigations of large landslides in the Carnic region of southern Austria. Eng. Geol., 56, 373388.
McCann, D.M. and Forster, A., 1990. Reconnaissance Geophysical Methods in Landslide Investigations.
Engineering Geology 29 (1), 59-78.
Menke, W., 1989. Geophysical Data Analysis: Discrete Inverse Theory. International Geophysics Series, Ed. R.
Dmowska and J. R. Holton, Academic Press, Inc., Harcourt Brace Jovanovich, Publishers. Volume 45.
Mric O., Garambois S., Malet J.P., Cadet H., Guguen P., Jongsmans D., 2007. Seismic noise-based methods
for soft-rock landslide characterisation. Soc. gol. Fr., 178 (2), 137-148.
Mric, O., Garambois, S., Malet, J.-P., Cadet, H., Guguen, P., Jongmans, D., 2007. Seismic noise-based
methods for soft-rock landslide characterization. Bull. Soc. Geol. France, 178, 2, 137-148.
Mric, O., Garambois, S., Orengo Y., 2006. Large gravitational movement monitoring using a spontaneous
potential network. In: Proc. 19th Annual meeting of SAGEEP, Seattle, USA, EEGS Ed., Denver, USA, 6 p.
Mric, O.,Garambois, S., Jongmans, D., Wathelet, M., Chatelain, J.-L., Vengeon, J.-M., 2005. Application of
geophysical methods for the investigation of the large gravitational mass movement of Schilienne, France.
Can. Geotech. J., 42, 1105-1115.
Milsom, J., 2003. Field Geophysics, Third Edition. The Geological Field Guide Series, Wiley, ISBN 0-47084347-0.
Mourot P., 2008. Mthodes et outils pour lauscultation et la surveillance des instabilits gravitaires. Thse de
Doctorat, Laboratoire de Gophysique Interne et de Tectonophysique, Universit de Savoie, Chambery.
Mucciarelli M., Gallipoli M. R., 2001. 10 years of Nakamuras technique: a review of theory, experiments and
applications. Bollettino di Geofisica Teorica ed Applicata, 42, 3-4, 255-266.
Nakamura Y., 1989. A method for dynamic characteristics estimation of subsurface using microtremor on
ground surface. Quar. Report. Railway. Tech. Res. Institute, 30, 25-33.
Nakamura Y., 1989. A method for Dynamic Characteristics estimation of subsurface using microtremor on the
ground surface. Quarterly Report of railway Technical Research Institute (RTRI), 30, No. 1.
Nakazato H., Kuroda S., Okuyama T., Sasaki Y., 2006. The aim at a rich rural village utilizing water and soil.
Improvement of production basis for improving productivity and exhibiting many-sided functions, and
development of management techniques. Improvement of airborne electromagnetic method and threedimensional resistivity distribution exploration in landslide areas. Mizu to Tsuchi o Ikashi Yutakana Noson
o Mezashite Saishin Nogyo Kogaku Kenkyu Seikashu Heisei 18nen, pp 216-217.
Nanoseismic monitoring - http://nanoseismic.net
Nguyen, F., Garambois, S., Jongmans, D., Pirard, E., Locke, M., 2005. Image processing of 2D resistivity data
to locate precisely faults. J. App. Geophys., 57, 260-277.
Nicholson, C., 1992. Recent developments in rockburst and mine seismicity research. In: T.A. Wawersik (Ed.),
Rock Mechanics, Balkema, Rotterdam, 1079-1086.
Novosad S., Blaha P., Kneijzlik J., 1977. Geoacoustic methods in the slope stability investigation. Bull. of the
Int. Ass. of Engineering Geology, 16, 228-231.
Obert , L., 1977. The microseismic method: discovery and early history. First conf. on acoustic emission /
microseismic activity in geologic structures and materials, Trans Tech Publications, 11-12.
Obert L., 1941. Use of subaudible rock noises for prediction of rock bursts. United States Bureau of Mines RI
3555.
Obert L., Duvall W., 1942. Use of subaudible rock noises for prediction of rock bursts. Part II, United States
Bureau of Mines RI 3654.

SafeLand - FP7
Page 388 of 401
6. References
Rev. No: 2
Date: 2012-02-15
Obert L., Duwall W., 1945a. Microseismic method of predicting rock failure in underground mining. Part I,
General Methods, United States Bureau of Mines RI 37m97.
Obert L., Duwall W., 1945b. Microseismic method of predicting rock failure in underground mining. Part II,
Laboratory Experiments, United States Bureau of Mines RI 3903.
Obert L., Duwall W., 1957. Microseismic method of determining the stability in underground workings. United
States Bureau of Mines Bulletin 573.
Observatoire des Instabilits de Versants - http://eost.u-strasbg.fr/omiv/Super_Sauze_intro.html
Palacky, G.J., 1987. Resistivity characteristics of geologic targets. Geosciences Journal, 3, 138-144.
Park, C.B., Miller, R.D., Xia, J., 1999. Multichannel analysis of surface waves. Geophysics 64, 800808.
Park, C.B., Miller, R.D., Xia, J., 2001. Offset and resolution of dispersion curve in multichannel analysis of
surface waves (MASW): Proceedings of the SAGEEP 2001, Denver, Colorado, SSM4.
Patella, D, 1997. Introduction to ground surface self-potential tomography. Geophys. Prospect., 45, 653681.
Pettinelli, E., Beaubien, S., Tommasi, P., 1996. GPR investigations to evaluate the geometry of rock slides and
bulking in a limestone formation in northern Italy. European J. Env. Eng. Geophys., 1, 271-286.
Pfaffhuber A.A., Grimstad E., Domaas U., Auken E., Halkjr M., 2010. The hunt for sliding planes in a phyllitic
rock slide inWestern Norway using airborne electromagnetic mapping, EGU General Assembly 2010,
Geophysical Research Abstracts, 12, EGU2010-4492.
Pinson, L., Henstock, T.J., Dix, J., Bull, J.M. 2008: Estimating quality factor and mean grain size of sediments
from high-resolution marine seismic data. Geophysics,VOL. 73, NO. 4
Polom, U., Hansen, L., Sauvin, G., LHeureux, J-S., Lecomte, I., Krawczyk, C. M., Vanneste, M., Longva, O.,
2010. High-resolution SH-wave Seismic Reflection for Characterization of Onshore Ground Conditions in
the Trondheim Harbor, Central Norway, Geophysical References, Society of Exploration Geophysicists, in
press.
Pugin, A.J.M., Larson, T.H., Sargent, S.L., McBride, J.H, Bexfield, C.E., 2004. Near surface mapping using
SH-wave and P-wave seismic land-streamer data acquisition in Illinois, U.S.: The Leading Edge, 23, 677682.
Rankka, K., Andersson-Skld, Y., Hulten, C., Larsson, R., Leroux, V., and Dahlin, T., 2004, Quick clay in
Sweden, Swedish Geotechnical Institute, report 65.
Renalier F., Bivre, G., Jongmans, D., Campillo, M., Bard, P.-Y., 2010. Characterization and monitoring of
unstable clay slopes using active and passive shear wave velocity measurements, SEG reference
publications, in press.
Renalier F., Jongmans, D., Campillo, M., Bard, P.-Y., 2010. Shear wave velocity imaging of the Avignonet
landslide (France) using ambient noise cross correlation, Journal of Geophysical Research, 115, F03032,
14p.
"Reynolds, J.M. 1992. ""The identification and mitigation of glacier-related hazards: examples from the
Cordillera Blanca, Peru."" In: McCall, G.J.H., Laming, D.C.J. and Scott, S. (Eds), Geohazards, London,
Chapman & Hall, 143-157."
Reynolds, J.M., 1997. An introduction to applied and environmental geophysics, John Wiley & Sons, Chichester,
806 pp., ISBN 0-471-95555-8.
Reynolds, J.M., 2006. Role of geophysics in glacial hazard assessment, First Break, 24, no. 8, August, 61-66.
Roch, K.H., Schwatal, B., Bruckl, E., 2006. Potentials of monitoring rock fall hazards by GPR: considering as
example the results of Salzburg. Landslides, 3, 87Rmoen, M., Pfaffhuber, A.A., Karlsrud, K., Helle, T. E., 2010. The use of a CPTU-probe with resistivity module
on marine sediments. CPT10: 2nd international symposium on Cone Penetration Testing, Huntington
Beach, California.
Rnning, J.S., Dalsegg, E., Elvebakk, H., Ganerd, G., Tnnesen, J.F., 2006. Geofysiske mlinger knes og
Tafjord, Stranda og Nordal kommuner, Mre og Romsdal. Geological Survey of Norway, report 2006.02, in
Norwegian.
Rnning, J.S., Dalsegg, E., Heincke, B., Tnnesen, J.F., 2007. Geofysiske mlinger p bakken ved knes og ved
Hegguraksla, Stranda og Nordal kommuner, Mre og Romsdal. Geological Survey of Norway, report
2007.26, in Norwegian.

SafeLand - FP7
Page 389 of 401
Deliverable 4.1
Roth M., Dietrich M., Blikra L. H., Lecomte I., 2005. Seismic monitoring of the unstable rock slope site at knes,
Norway. NORSAR Report.
Rouse C., Styles P. and Wilson S. A., 1991a. Acoustic Emissions from two Landslip Areas in South Wales. Z.
Geomorph. N. F., Suppl.-Bd. 83, 135-154.
Rouse C., Styles P., Wilson S.A., 1991b. Microseismic emissions from flowslide-type movements in South Wales.
Engineering Geology, 31, 91-110.
Rubin, Y., and Hubbard, S.,(2005. Eds: Springer, The Netherlands, 530 pp.
Sailhac, P., and Marquis, G., 2001. Analytic potentials for the forward and inverse modeling of SP anomalies
caused by subsurface fluid flow. Geophys. Res. Lett., 28, 1851-1854.
Sasaki Y., Nakazato H., 2004. Inversion of airborne EM data accounting for terrain and inaccurate flight height,
SEG Expanded Abstracts 23, 648.
Schmutz, M., Albouy, Y., Gurin, R., Maquaire, O., Vassal, J., Schott, J.-J., Desclotres, M., 2000. Joint
electrical and time domain electromagnetism (TDEM) data inversion applied to the Super Sauze earthflow
(France). Surveys in Geophys., 21, 371-390.Seigel, H.O., 1995 A guide to high precision land gravimeter
surveys. Scintrex Limited, Concord, Ontario, Canada, p122
Seabeam 1999: Multibeam Sonar - Theory of Operation. 2101-8289
Senfaute G., Duperret A., Lawrence J. A., 2009. Micro-seismic precursory cracks prior to rock-fall on coastal
chalk cliffs: a case study at Mesnil-Val, Normandie, NW France. Nat. Hazards Earth Syst. Sci., 9, 16251641.
SESAME European research project, 2004. Guidelines for the implementation of the H/V spectral ratio
technique on ambient vibrations - Measurements, processing and interpretation. WP12 Deliverable
D23.12, European Commission Research General Directorate, Project No. EVG1-CT-2000-00026, p62.
"Sherif, R.E., 2002. Encyclopedic Dictionary of Applied Geophysics, fourth edition, Society of Exploration
Geophysicists, ISBN 9781560801184 (13); 1560801182 (10), 429 p."
Shiotani T., 2006. Evaluation of long-term stability for rock slope by means of acoustic emission technique.
NDT&E International, 39, 217-228.
Socco, V. and Jongmans, D., 2004. Special issue on Seismic Surface Waves. Near Surf. Geophys., 2, 163-258.
Solberg, I.L., Rnning, J.S., Dalsegg, E., Hansen, L., Rokoengen, K., Sandven, R., 2008. Resistivity
measurements as a tool for outlining quick-clay extent and valley-fill stratigraphy: a feasability study from
Buvika, central Norway. Canadian Geotechnical Journal, 45: 210-225, doi:10.1139/T07-089.
Spillmann T., Maurer H., Green G. A., Heincke B., Willenberg H., Husen S., 2007. Microseismic investigation of
an unstable mountain slope in the Swiss Alps. J. Geophys. Res., 112.
Steffens G.S., Shipp R.C., Prather B.E., Nott J.A., Gibson J.L., and Winker C.D. 2004: The use of near seafloor
3D seismic data in deepwater exploration and production. In Davies R.J., Cartwright J.A., Stewart S.A.,
Lappin M., and Underhill J.R. (eds.), 3D Seismic Technology: Application to the Exploration of
Sedimentary Basins. Geol Soc Spec Publ 29:3543.
Stuart G., Murray T., Brisbourne A., Styles P. And Toon S., 2005. seismic emissions from a surging glacier:
Bakaninbreen, Svalbard. Annals of Glaciology, 42, 151-157.
Supper, R., Ahl, A., Rmer, A., Jochum, B., Bieber, G., 2008. A complex geo-scientific strategy for landslide
hazard mitigation - from airborne mapping to ground monitoring, Advances in Geosciences, 14, 1-6.
Supper, R., and Rmer, A., 2003. New achievements in developing a high speed geoelectrical monitoring system
for landslide monitoring. In: Proc. 9th Meeting Env. Eng. Geophys., Prague, Czech Republic, EAGE
Publications, Houten, The Netherlands EEGS Ed., O-004.
Suriach E., Sabot F., Furdada G., Vilaplana J. M., 2000. Study of seismic signals of artificially released snow
avalanches for monitoring purposes. Phys. Chem. Earth (B), 25, 721-727.
Suriach E., Vilajosana I., Khazaradze G., Biescas B., Furdada G., Vilaplana J. M., 2005. Seismic detection and
characterization of landslides and other mass movements, Natural Hazards and Earth System Sciences, 5,
791-798.
Suwa H., Mizuno T., Suzuki S., Yamamoto Y., Ito K., 2008. Sequential processes in a landslide hazard at a slate
quarry in Okayama, Japan. Nat. Hazards, 45, 321-331.

SafeLand - FP7
Page 390 of 401
6. References
Rev. No: 2
Date: 2012-02-15
Suwa H., Yamakoshi T., Sato K., 2000. Relationship between debris-flow discharge and ground vibration,
Debris-Flow Hazards Mitigation: Mechanics, Prediction, and Assessment, Proceedings of the Second
International Conference on Debris-Flow Hazards Mitigation, Taipei/Taiwan 16-18 August 2000, 311-318.
Swiss Geophysical Commission, Quality guidelines for geophysical methods. http://www.sgpk.ethz.ch/
Tonnellier A., Malet J.-P., Joswig M., Walter M., Corsini A. 2010. Landslides under microscope: understanding
slow-moving landslides through passive micro-seismic monitoring. International Conference TRAMM
Triggering of Rapid Mass Movements in Steep Terrains - Mechanisms and Risk, 21-25th April 2010,
Mont Verit, Switzerland. Abstract-based poster.
US Army Corps of Engineers, 1995. Geophysical Exploration for Engineering and Environmental
Investigations. Manual No. 1110-1-1802. 207 p.
US Army Corps of Engineers, 2001. Geotechnical Investigations. Engineering and Design. Manual No. 1110-11804. 449 p.
Van Westen, C.J., 2004. Geo-Information tools for landslide risk assessment: an overview of recent
developments. In: Proc. 9th International. Symp. Landslides, Rio de Janeiro, Brazil, Balkema, Rotterdam,
39-56.
Walter M., Joswig M., 2008. Seismic monitoring of fracture processes generated by a creeping landslide in the
Vorarlberg Alps. First Break, 26, 131-135.
Walter M., Joswig M., 2009b. Seismic characterisation of slope dynamics caused by softrock-landslides: The
Super-Sauze case study, In: Malet, J.-P., Rematre, A., Boogard, T.A. (Eds.): Proceedings of the
International Conference on Landslide Processes: from geomorpholgic mapping to dynamic modelling,
Strasbourg, CERG Editions, 215-220.
Walter M., Niethammer U., Rothmund S., Joswig M., 2009a. Joint analysis of the Super-Sauze (French Alps)
mudslide by nanoseismic monitoring and UAV-based remote sensing. First Break, 27, 53-60.
Wang H., Ge M., 2007. Acoustic emission/microseismic source location analysis for a limestone mine exhibiting
high horizontal stresses. Int. J. of Rock Mechanics & Mining Sciences, 45, 720-728.
Wiechert D., Horner R. B., Evans S. G., 1994. Seismic Signatures of Landslides: the 1990 Brenda Mine Collapse
and the 1965 Hope Rockslides. Bull. of the Seism. Soc. of America, 84-5, 1523-1532.
Wilkinson, P.B., Chambers, J.E., Meldrum, P.I., Gunn, D.A., Kuras, O., Ogilvy, R.D., 2010. Tracking the
movements of electrodes on an active landslide over time using only timelapse resistivity data. Extended
abstract, Near Surface Conference, Zurich.
Wilkinson, P.B., Chambers, J.E., Meldrum, P.I., Gunn, D.A., Kuras, O., Ogilvy, R.D., 2010. Tracking the
movements of electrodes on an active landslide over time using only timelapse resistivity data. Extended
abstract, Near Surface Conference, Zurich.
Willenberg H., Evans K.F., Eberhardt E., Loew S., Spillmann T., Maurer H., 2004. Geological, geophysical and
geotechnical investigations into the internal structure and kinematics of un unstable sliding mass in
cristalline rock. 9th Internat. Symp. on Landslides, Rio de Janeiro, 489-494.
Willenberg, H., Evans, K.F., Eberhardt, E., Loew, S., Spillmann, T., Maurer, H.R., 2004. Geological,
geophysical and geotechnical investigations into the internal structure and kinematics of an unstable,
complex sliding mass in crystalline rock. In: Proc. 9th International. Symp. Landslides, Rio de Janeiro,
Brazil, Balkema, Rotterdam, 489-494.
Wisn, R., Auken, E., Dahlin, T., 2005. Combination of 1D laterally constrained inversion and 2D smooth
inversion of resistivity data with a priori data from boreholes. Near Surf. Geophys., 3, 71-79.
Wisen, R., Christiansen, A.V., Auken, E., Dahlin, T., 2003. Application of 2D laterally constrained inversion and
2D smooth inversion of CVES resistivity data in a slope stability investigation. In: Proc. 9th Meeting Env.
Eng. Geophys., Prague, Czech Republic, EAGE Publications, Houten, The Netherlands, O-002.
"Wust-Bloch G. H., Joswig M., 2006. Pre-collapse identification of sinkholes in unconsolidated media at Dead
Sea area by ""nanoseismic monitoring"" (graphical jackknife location of weak sources by few, low-SNR
records). Geophys. J. Int."
Xia, J., Miller, R.D., Park, C.B., 1999. Configuration of near surface shear wave velocity by inverting surface
wave. Proceedings of Symposium on the Application of Geophysics to Engineering and Environmental
Problems 99, pp. 95104.

SafeLand - FP7
Page 391 of 401
Deliverable 4.1
Yang C., Luo Z., Hu G., Liu X., 2007. Application of a microseismic monitoring system in deep mining. J. of Uni.
of Science and Technology of Beijing, 14-1, 6-8.
Zhang, S., Hong Y., Yu B. 2004. Detecting infrasound emission of debris flows for warning purposes. In: Proc.
10th Congress INTERPRAEVENT 2004, pp. Z/359Z/364, Riva del Garda, Trento, Italy, 24-27 May 2004.
Zhdanov, M.S., 2002. Geophysical inverse theory and regularization problems. Elsevier, Amsterdam NewYork,628 pp.
Zimmer V. L., Stock G. M., Sitar N., 2008. Seismic Monitoring of Rock Falls in Yosemite National Park. AGU,
San Francisco, USA, Fall Meeting 2008, Abstract-based Poster.
Zobin V. M., Placencia I., Reyes G. and Navarro C., 2009. The characteristics of seismic signal produced by
lahars and pyroclastic flows: Volcn de Colima, Mexico. Journal of Volcanology and Geothermal Research,
179, 157-167.
Zyatev G. G., Merkulov V. P., Nikolsky A. A., 1998. The reflection of landslip processes in geophysical fields,
60th EAGE Conf. and techn. Exhib., Leipzig, 8-12 June 1998, Oral and Poster Present., Geophys. Div.,
Extend. Abstr. Book, Leipzig, 121.
6.6. SPECIFIC REFERENCES FOR THE CHAPTER 2E

Burland, J.B, More,J.F.A., Smith, P.D.K. A simple and precise borehole extensometer. Geotechnique, Vol. 22,
No1, pag 174-177. 1972.
Corominas, J., Moya, J., Lloret, A., Gili, J.A., Angeli, M.G., Pasuto, A. Silvano, S. Measurement of landslide
displacements using a wire extensometer. Engineering Geology, vol. 55, issue 3, pp. 149-166. 2000.
Dunnicliff, J., assisted by Green, G.E. Geotechnical instrumentation for monitoring field performance. WileyInterscience Publication, 577 p. 1988-1993.
Dunnicliff, J., La Fonta, J.G. In place inclinometers. A significant test program. Geotechnical News, vol. 19,
n1, March, pp. 33-34. 2001.
Dunnicliff J. et al. Geotechnical instrumentation for field measurements. Training course, March 13-15, 2005.
The University of Florida. 2005.
Kovari, K. Detection and monitoring of structural deficiencies in the rock foundation of large dams. 5th Int.
Conf. Large Dams. Lausane, pp. 695-719. 1985.
Kovari, K. Methods of monitoring Landslides. Proc. 5th Int. Symp. On Landslides. Lausane, vol. 3, 14 p. 1988.
La Fonta, J.G., Beth, M.B. In place inclinometers, a significant test program. SolData-France. 2002.
Lollino, G. Automated Inclinometric System. Proceedings of the Sixth International Symposium on Landslides.
10-14 February 1992, Christchurch, New Zealand. 1992.
Lovisolo, M., Della Giusta, A. Precision of D.M.S. columns from real time in-place measurements and
improvement in micro-movements analysis with early warning function. In: Computational Methods and
Experimental Measurements XII (Brebbia, C.A. and Carlomagno, G.M.). WITpress, Southampton, Boston,
pp.177-186. 2005.
Machan, G., Bennett, V.G. Transportation research circular n E-C129. Use of Inclinometers for Geotechnical
Instrumentation on Transportation Projects - State of the Practice. Transportation Research Board,
Washington. 2008.
Mikkelsen E. Piezometers in fully grouted boreholes. In: Field Measurements in Geomechanics. OSLO (Myrvoll
F., ed.). Balkema Publishers, pp. 545-554. 2003a.
Mikkelsen E. Advances in inclinometer data analysis. In: Field Measurements in Geomechanics. OSLO (Myrvoll
F., ed.). Balkema Publishers, pp. 555-567. 2003b.
6.7. SPECIFIC REFERENCES FOR THE CHAPTER 2F

Bell, F.G. Engineering Geology. Second edition. Butterworth-Heinemann. 584 p. 2007.
Bonnard, Ch., Noverraz, F., Dupraz, H. Long-term movements of substabilized versants and climatic changes
in the Swiss Alps. Proceed. 7th Int. Symp. On Landslides, Trodheim, Vol. 3, 1525-1530. 1996.

SafeLand - FP7
Page 392 of 401
6. References
Rev. No: 2
Date: 2012-02-15
Brunner, F.K., Woschitz, H., Macheiner, K. Monitoring of deep-seated mass movements. In: Proc. of. The 3rd
International Conference on Structural Health Monitoring of Intelligent Infrastructure, Vancouver, Canada.
CD-ROM, 10 p. 2007.
Chen, Y., He, X., , Ding, X., Sang, W. Steep-Slope Monitoring: GPS Multiple-Antenna System at Xiaowan Dam.
GPS World Magazine, November 2005, 4p. 2005.
ClimChalp. Slope Monitoring Methods, A State of the Art Report. ClimChalp is an European Regional
Development Project, within the framework of the INTERREG IIIB Alpine Space Programme. The Report of
the Work Package 6, was released in Munich, 28.2.2008. 179 p. 2008.
Gili, J.A., Corominas, J., Rius, J. Using Global Positioning System techniques in landslide monitoring.
Engineering Geology, 55, pp. 167-192. 2000.
Malet, J.P., O. Maquaire and E. Calais. The use of Global Positioning System techniques for the continuous
monitoring of landslides: application to the Super-Sauze earthflow (Alpes-de-Haute-Provence, France).
Geomorphology, Volume 43, Issues 1-2, 1 February 2002, pp 33-54. 2002.
McGuffey, V.C., Modeer, V.A. Jr, Turner, A.K. Subsurface exploration. In: Landslides, investigation and
mitigation. Tuner, A.K., Schuster, L., editors. Special report 247, Transportation Research Board, National
Research Council. 678 p. 1996.
Schfer, T. "GPS observations at Mt. Hochstaufen. Practice Example A11 in ClimChalp (2008), pp. 137-142.
2008.
US Army Corps of Engineers. Geotechnical Investigations. Engineering and Design. Manual No. 1110-1-1804.
449 p. 2001.

SafeLand - FP7
Page 393 of 401
7. Glossary
Rev. No: 2
Date: 2012-02-15
GLOSSARY

SafeLand - FP7
Page 395 of 401
7. Glossary
Rev. No: 2
Date: 2012-02-15
7. GLOSSARY
Accuracy (applied for TLS)
This parameter measures the degree of closeness of a measurement

to the real value. The accuracy of a laser measurement is a main
function of the range, the reflectivity of the material, the complexity of
the scanned surface and the angle of incidence. For monitoring
purposes, the accuracy of a given TLS dataset can be estimated
through the estimation of the precision of the measurement, i.e.,
through the comparison of two different point clouds acquired
consecutively
Affine Polynomial Model
Set of equations describing the relation of an image to the local

reference system in terms of translation, rotation, scaling and
distortion
Airborne Electromagnetics
Airborne geophysical method to determine the distribution of the

specific electrical resistivity within the subsurface
Airborne geophysics
Geophysical approach which uses sensors that are either mounted

on an aircraft or dragged on a cable several tens of meters below the
aircraft (an airplane or a helicopter)
Ascending / Descending
orbit
South-North / North-South trajectory of the platform.
Azimuth (SAR context)
Distance along the spaceborne flight direction. The azimuthal

resolution can be performed by Synthetic Aperture Radar techniques.
Change Detection (Lidar)
The study of the changes or differences that occur during a given

portion of time. The sign of the differences depend on the
computation technique. But it is common to attribute negative signs to
areas that are behind or below the reference surface (i.e. detachment
of material) and positive values to areas that are in front or above the
reference surface (i.e. scree deposits, advance of a landslide etc.).
Change Detection (Passive

Optical Data)
Detection and characterization of episodic or sudden changes by

comparison of at least two different images of the same earth surface
area. Main criteria for changes are variations in the surface
reflectance in several multispectral bands.
Comparison of TLS datasets The process of point cloud alignment plus change detection. As an
output of the procces, changes in the slope during the time span (i.e.
displacements, rockfall occurrence, deposition, etc), are recordered
Complex signal
Each pixel is registered by a complex number, keeping the

information on the amplitude and the phase of the return signal. This
is the raw data for an InSAR study.
Data Fusion
Methods for the combination of the data to optimize the utilization of

large volumes of data from multiple sources.
Digital Image Correlation
Comparison of grey value co-occurrence within a search window in

two images. Points of maximum correlation are used for the coregistration of the two images or two evaluate the displacement of
pixels or group of pixels.

SafeLand - FP7
Page 397 of 401
Deliverable 4.1
DInSAR
Differential InSAR. Technique aiming at separating topographic and

deformation components of the InSAR phase information. It is based
on subtraction of the topographic component (derived from short time
span interferograms or simulation using DEM) in order to retrieve the
deformation component and map the ground surface deformation.
SBAS and PSI are advanced evolutions of the DINSAR technique.
Ephemerides
Satellite position and velocity vectors at any time t
Fringe
Line of an interferogram corresponding of one color cycle.
G.C.P
For Ground Control Point: an easily identifiable object with known

coordinates used for the geometrical correction of maps and remote
sensing imagery.
Gamma-Ray spectroscopy
Measurement of natural gamma radiation at differen energy values
Geometric correction
Modeling of systematic and non-systematic geometric errors to derive

a relationship between the image and a global coordinate system
GNSS
Global Navigation Satellite System, a radionavigation, timing and

positioning system that can provide the user 3D position. This
acronym is the common one to enclose the GPS (USA), the
GLONASS (former soviet union), the GALILEO (EU), the Beidou
(China), and other future satellite navigation systems.
GPS
See GNSS
Ground range
Distance along the perpendicular of the flight direction on the

grounds surface. The range resolution can be performed emitting
chirped pulses.
Ground Sampling Distance
Measure of the distance between two adjacent samples in an image
Ground-based geophysics
Geophysical approach using sensors located either on the ground

(non-intrusive methods) or in boreholes. Also called near-surface
geophysics.
Homogenous half space

model
Numerical model which describes the ground subsurface with a

uniform electrical resistivity 1 derrived from a multilayer AEM
dataset
Image Mining
Automation of the extraction of implicit knowledge, image data

relationship, or other patterns not explicitly stored in the imagery in
large archives.
InSAR
SAR interferometry. Technique based on the difference of the phase

values between two SAR acquisition. The phase difference
(interferometric phase) contains geometric information that could be
related to ground surface altitude (in order to produce DEM) or
deformation.
Intensity
The intensity is defined as the amount of reflected signal with respect

to the emitted one. Its value is usually normalized on a 0-255 scale.
the intensity parameter mainly depends on the range, angle of
incidence, soil moisture, and object material

SafeLand - FP7
Page 398 of 401
7. Glossary
Rev. No: 2
Date: 2012-02-15
Joint interpretation
Used in ground-based geophysics to describe an interpretation

process (manual or automated) where results obtained by different
methods (e.g., after inversion for each method) are somehow merged
to reach a more complete conclusion about the ground under
investigation.
Joint inversion
Used in ground-based geophysics to describe an inversion approach

where parameters obtained from different methods (e.g., elastic and
electrical) are obtained together during the same inversion process,
these parameters constraining each others.
LIDAR
Acronym for Light Detection and Ranging. Optical remote sensing

technology measures the properties of scattered light to find range
and/or other information about a distant target. This can be an Aerial
(ALS) or Terrestrial (TLS) Laser Scanning instrument
Look angle
Angle between the vertical and the LOS.
LOS
For Line of Sight: an imaginary line from the sensor (optical, radar,
laser, etc) to the surveyed point. Direct visibility along this line is
mandatory when distance calculation.
Master / Slave images
Reference / comparative images in interferometry.
Model inversion
Find the most suitable parameter model fitting to a set of

measurements
Nadir
Projection on the Earths surface of the platforms trajectory
Near / Far range
Part of the image closest to / furthest from the Nadir.
Object-Oriented Image
Analysis
Grouping of image pixels into homogenous and meaningful objects

and consequent image analysis incorporating additional object
features such as texture, shape and neighborhood relationships.
Penetration depth
Ground-based geophysics: characterize the depth which can be

reached and investigated depending of the type of method (seismic,
ERT, GPR, etc), the geometry of the survey (e.g., distance between
emitters and receivers), the used frequencies, etc. Note that there is
often a natural trade-off between penetration depth and resolution:
the higher the depth, the lower the resolution because high
frequencies - good for high resolution - cannot propagate far due to
attenuation, etc.
Point Cloud
A large quantity of 3D points acquired by the ALS/TLS through a

rangefinder method, also referred to as ALS/TLS scans or datasets.
A point cloud can be considered a data point cloud or reference
point cloud
Pseudo depth sections
Representation of geoelectric measurement results; apparent

resistivities are displayed according to the separation of current
electrodes, not real depths
Radiometric correction
Pre-processing steps that involve correction of sensor, atmospheric

or topographic effects on the recorded radiance in one image
(absolute correction), or target the equalization of two or more
images that have been recorded under distinct conditions.

SafeLand - FP7
Page 399 of 401
Deliverable 4.1
Range
The distance between the sensor (i.e. TLS, GB-SAR) and the portion
of the terrain for which the coordinates are acquired
Rational Function Model
Ratio of two polynomial functions relating 3-D surface coordinates to

2-D image coordinates.
Resolution (airborn imaging) This parameter determines the level of detail that can be observed
from a scanned point cloud. It can be divided into range and angular
(or spatial) resolution. The estimation of the TLS resolution is usually
misunderstood and commonly interpreted as equal to the point
spacing or sampling interval, obviating the influence of the laser
beam width.
Resolution (ground-based
geophysics)
Resolution - in the strict mathematical sense of an inversion problem

- characterizes whether searched parameters can be independently
determined or resolved from a certain data set and assuming a model
between data and the searched parameters. In a simpler manner,
and considering imaging (e.g., seismic, ERT, etc), the level of
resolution tells how detailed an image can be. In seismic imaging, a
rule of thumb is to state that resolution is a quarter of the minimum
wavelength (velocity divided by frequency).
SAR
Synthetic Aperture Radar. System designed for providing higher

resolution than conventional radar systems (synthetic aperture
compared to real aperture). Processing of the complex value
(amplitude related to the backscaterred energy and phase related to
the distance between the sensor and the targets ) of the
backscatered signal acquired with a sampling rate compatible with
the expected resolution, allows to relocate the targets contributing to
the signal in order to obtain a finer image. Typically, spaceborne SAR
resolutions range from the metre to tens of metres.
Scan of Reference
The first TLS dataset acquired for a certain area. Subsequent

datasets are aligned towards this dataset.
Shannon sampling theorem
Fundamental work from information theory stating that a signal can

be lossless reconstructed from samples taken at intervals two times
shorter than the highest frequency of the signal.
Slant range
distance along the perpendicular of the azimuth on the SAR image.

The slant range can be different of the ground range due to
geometrical distortion.
Spot Dimension
The projection of the laser beam on the ground or target area.

Although the laser signal is highly collimated, the spot dimension
increases its value with the distance (e.g. 2.1 mm at 50 meters and
9.6 mm at 500 m). This parameter is also referred to as laser
beamwidth at a given distance.
Stereoprocessing
Techniques involving the identification of corresponding objects in

two or more images and the derivation of their 3D coordinates by
point triangulation.
Supervised Classification
Areas with a know class memberships in an image are used as

samples to train a classification algorithm. The trained algorithm is
then used to classify the rest of the image or other scenes.

SafeLand - FP7
Page 400 of 401
7. Glossary
Rev. No: 2
Date: 2012-02-15
TLS
Acronym for Terrestrial Laser Scanner. This instrument consists of a

transmitter/receiver of infrared laser pulses and a scanning device
mounted over a ground-based platform. Range measurements with
centimetric accuracy can be undertaken at a high speed rate. As a
result, 3D coordinates of million points are obtained.
Tomography
The word "tomography" is derived from the Greek tomos (part) and
graphein (to write). Tomography is imaging by sections or sectioning,
through the use of any kind of penetrating wave. In ground-based
geophysics tomography can be performed in various ways, e.g.,
using traveltimes of elastic waves (seismic) or electromagnetic waves
(Ground Penetrating Radar - GPR), or apparent resistivity
measurements (Electrical Resistivity Tomography - ERT), etc.
Track / Frame
Est-West / North-South limit of a SAR image.
Two strata model
Numerical model which describes the ground subsurface with 2

layers of different thickness (h) and resistivity () derrived from a
multilayer AEM dataset
Unsupervised Classification
Automatic categorization or clustering of a digital image data into a

number of predefined or statistical optimal classes.

SafeLand - FP7
Page 401 of 401

Deliverable 4.1 Safeland

Uploaded by

Copyright:

Available Formats

Deliverable 4.1 Safeland

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Deliverable 4.1 Safeland

Uploaded by

Copyright:

Available Formats

Grant Agreement No.

7th Framework Programme

Deliverable Leader: Unil

SafeLand deliverable 4.1

To quote this document:

Grant Agreement No.: 226479

SafeLand deliverable 4.1

Note about contributors

Lead partner responsible for the deliverable:

Partner responsible for quality control:

were written and reviewed by at least two specialized experts of independent

Grant Agreement No.: 226479

SafeLand deliverable 4.1

Grant Agreement No.: 226479

SafeLand deliverable 4.1

Two strata model

Grant Agreement No.: 226479

SafeLand deliverable 4.1

Geographic Object-Based Image Analysis

Grant Agreement No.: 226479

SafeLand deliverable 4.1

Grant Agreement No.: 226479

SafeLand deliverable 4.1

DATA AND TECHNIQUES .......................................................................................................... 23

PART A - PASSIVE OPTICAL SENSORS ..................................................................................................... 25

SafeLand deliverable 4.1

Grant Agreement No.: 226479

SafeLand deliverable 4.1

PIEZOMETERS ................................................................................................................................... 229

APPLICATION TO LANDSLIDES: CASE STUDIES ................................................................ 253

CONCLUSIONS ......................................................................................................................... 359

REFERENCES ........................................................................................................................... 363

Grant Agreement No.: 226479

SafeLand deliverable 4.1

Specific references for the chapter 2B......................................................................................... 374

GLOSSARY ................................................................................................................................ 397

Grant Agreement No.: 226479

Grant Agreement No.: 226479

Grant Agreement No.: 226479

Grant Agreement No.: 226479

2. Data and Techniques

DATA AND TECHNIQUES

Grant Agreement No.: 226479

2. Data and Techniques

2. DATA AND TECHNIQUES

PART A: Passive Optical Sensors

PART B: Active Optical Sensors

PART C: Active Microwave Sensors

PART D: Geophysical Investigations

PART E: Geotechnical Ground-Based Monitoring Systems

PART F: Other Techniques

Grant Agreement No.: 226479

Grant Agreement No.: 226479

2. Data and Techniques

SECTION 1: General Pre-Processing Step

SECTION 2: Ground-Based Imaging

SECTION 3: Aerial Imaging

SECTION 4: Satellites Imaging

Grant Agreement No.: 226479

1. GENERAL PRE-PROCESSING STEPS