ΠΑΝΕΠΙΣΤΗΜΙΟ ΠΑΤΡΩΝ ΤΜΗΜΑ ΓΕΩΛΟΓΙΑΣ ΤΟΜΕΑΣ ΕΦΑΡΜΟΣΜΕΝΗ ΓΕΩΛΟΓΙΑΣ & ΓΕΩΦΥΣΙΚΗΣ ΕΡΓΑΣΤΗΡΙΟ ΣΕΙΣΜΟΛΟΓΙΑΣ
Παθητική Σεισμική Συμβολομετρία & Ανισοτροπία Εγκαρσίων Κυμάτων στη διερεύνηση του Γήινου φλοιού
ΔΗΜΗΤΡΙΟΣ Ν. ΓΙΑΝΝΟΠΟΥΛΟΣ
Γεωλόγος, M.Sc. Εφαρμοσμένης Γεωλογίας & Γεωφυσικής
Διδακτορική Διατριβή
ΠΑΤΡΑ 2016
Passive Seismic Interferometry
& Shear Wave Splitting in the investigation of the Earth's crust: application to the Corinth Rift, Greece
a dissertation presented
by
Dimitrios N. Giannopoulos
to
The University of Patras, Department of Geology,
Seismological Laboratory
in partial fulfillment of the requirements
for the degree of
Doctor of Philosophy
in Seismology
Patras
2016
Frontpiece*
© Alexandros Maragos
ΗΡΑΚΛΕΙΤΟΣ, 535 - 475 Π.Κ.Χ.
HERACLITUS, 535 - 475 BCE
v
University of Patras Faculty of Sciences Department of Geology Sector of Applied Geology & Geophysics
Laboratory of Seismology
Supervisory Committee Members:
Ass. Prof. Efthimios Sokos (Supervisor) (Department of Geology, University of Patras)
Prof. G-Akis Tselentis (National Observatory of Athens - Institute of Geodynamics)
Prof. Georgios Oikonomou (Department of Physics, University of Patras)
This to certify that we have read this dissertation and that in our opinion it is fully adequate, in scope of quality, as a dissertation for the degree of Doctor of Philosophy.
Examining Committee Members:
Assoc. Prof. Efthimios Sokos (Supervisor) (Department of Geology, University of Patras) ______
Prof. G-Akis Tselentis (Department of Geology, University of Patras | National Observatory of Athens - Institute of Geodynamics) ______
Prof. Georgios Oikonomou (Department of Physics, University of Patras) ______
Prof. Ioannis Koukouvelas (Department of Geology, University of Patras) ______
Prof. Anastasia Kiratzi (Department of Geophysics, Aristotle University of Thessaloniki) ______
Prof. Panayiotis Papadimitriou (Faculty of Geology-Geoenvironment, University of Athens) ______
Ass. Researcher Dr. Christos Evangelidis (National Observatory of Athens - Institute of Geodynamics) ______
vii
Η έγκριση της διδακτορικής διατριβής από το Τμήμα Γεωλογίας του Πανεπιστημίου
Πατρών, δεν υποδηλώνει ότι το Τμήμα αποδέχεται τις απόψεις του συγγραφέα
(Ν.5343/1932, Άρθρο 202, Παρ. 2).
The approval of this doctoral thesis by the Department of Geology, University of
Patras, does not imply that the Department accepts the author's opinions
(L.5343/1932, Article 202, Paragraph 2).
© Copyright by Dimitrios N. Giannopoulos 2016
All Rights Reserved
ix
Acknowledgements
This thesis is not only the result of personal effort and work, but it is also the result of the knowledge, help and work of many others. Following the completion of my doctoral thesis I would like to express my sincere thanks to all those who substantially contributed to this study.
First I would like to express my thankfulness to my supervisor Assoc. Professor
Efthimios Sokos for his outstanding guidance throughout the whole period of my studies. I thank him for always having an open door for me, for his valuable help, patience, discussions, for everything. Without all these, this thesis would not have been possible.
I am grateful to my supervisory committee members, the Director of Geodynamic
Institute - National Observatory of Athens Prof. G-Akis Tselentis and Prof.
Georgios Oikonomou from the Department of Physics, University of Patras, for the scientific advices and for their overall assistance.
I would like to mention one by one expressing my sincere gratitude to Anne
Deschamps, Diane Rivet both from Géoazur, Université de Nice Sophia Antipolis,
France, Hélène Lyon-Caen from Ecole Normale Supérieure, Paris, France and
Pascal Bernard from Institut de Physique du Globe de Paris, France, for their valuable contribuiton to the initiation of the Ambient Noise Tomography study and for the great hospitality during two very productive visits I had in Nice and Paris.
Aurelien Mordret from Department of Earth, Atmospheric and Planetary Sciences,
MIT, MA, USA for providing important software tools used in the tomographic
xi
inversions. Konstantinos I. Konstantinou from the Department of Earth Sciences,
National Central University of Taiwan for his valuable contribution to the shear-wave anisotropy analysis. My sincere thanks to Paris Paraskevopoulos and Athanasios
Lois, both from Department of Geology, University of Patras, who have been involved in different parts of this study. I am grateful to all for their active contribution, feedback, ideas, comments that have improved the quality of this thesis.
Finally, but most prominently in rank, my ultimate acknowledgement to my beloved parents, Mando and Nikos, and to my brother, Spyros, for their love, the continuing support and encouragement.
Dimitrios N. Giannopoulos
* In the image of the frontpiece Alexandros Maragos captured Panachaiko mountain range and the gulf of Patras, Greece, at sunset, the land where I was born, grew up, studied...
The photo was awarded by National Geographic (http://on.natgeo.com/1NE66jd) as the
"Photo of the Day" in December 7, 2015.
xii
Dedicated to my parents
and to my brother for their love and support
xiii
Table of contents
Frontpiece iv
Committee members vii
Copyright ix
Acknowledgements xi
Dedication xiii
Table of contents xv
List of figures xix
List of tables xxix
List of abbreviations xxxi
Chapters
1. Introduction - Thesis outline...... 33
1.1 Introduction...... 33
1.2 Passive Seismic Interferometry - Part I...... 33
1.3 Shear-Wave Splitting - Part II...... 35
1.4 Thesis objectives - Outline...... 37
2. Corinth Rift...... 41
2.1 Geotectonic Seismotectonic setting...... 41
Part I
3. Passive Seismic Interferometry...... 53
3.1 Ambient Seismic Noise...... 53
3.2 Theory...... 57
xv
3.3 Ambient Noise Tomography using Seismic Interferometry...... 60
4. Ambient Noise Tomography of the western Corinth Rift...... 65
4.1 Introduction...... 65
4.2 Data and processing workflow...... 65
4.3 Pre-processing & cross-correlations computation...... 67
4.4 Dispersion curve measurements...... 76
4.5 2-D Rayleigh-wave group velocity maps...... 79
4.6 Resolution Assessment...... 88
4.7 Inversion of local dispersion curves - Depth inversion...... 88
5. Interpretation & Discussion...... 99
5.1 Introduction...... 99
5.2 Rayleigh-wave group velocity maps...... 100
5.3 Shear-wave velocity structure at depth...... 103
5.4 Overall interpretation...... 108
5.5 Conclusions...... 112
Part II
6. Shear Wave Splitting...... 117
6.1 Introduction...... 117
6.2 What causes shear wave splitting...... 117
6.3 Variations in shear wave splitting parameters...... 120
6.4 Shear wave window...... 122
6.5 Applications...... 123
7. Shear Wave Splitting analysis in the western Corinth Rift...... 125
xvi
7.1 Introduction...... 125
7.2 The January 2010 Efpalio earthquakes...... 126
7.3 Data...... 128
7.4 Methodology...... 129
7.4.1 Shear wave splitting measurements...... 129
7.4.2 Estimation of VP/VS ratios...... 132
8. Shear Wave Splitting Results...... 133
8.1 Spatio-temporal grouping of the data...... 133
8.2 A first overview of the measurements...... 135
8.3 Within and close to the rupture areas...... 137
8.4 Outside of the rupture areas...... 138
8.5 Validation of observations...... 140
9. Interpretation & Discussion...... 155
9.1 Stress field and fast polarization directions...... 155
9.2 Possible causes of δt, VP/VS variations...... 157
9.3 Conclusions...... 165
10. Epilogue...... 167
10.1 Summary of thesis work...... 167
10.2 Conclusion – Future work...... 170
Appendix A...... 173
Appendix B...... 203
Reference list...... 209
xvii
List of Figures
2.1 Plate boundaries and the general geodynamic pattern of the broader Hellenic region (continental Greece, Ionian and Aegean Seas). Figure is from Vött (2007)……………………………………………………………………………………….. 41
2.2 Tectonic map of the western Corinth Rift. Major onshore and offshore fault traces of the area as in Ford et al. (2013), Palyvos et al. (2008), Flotté et al. (2005), McNeill et al. (2005), Bell et al. (2009), and Taylor et al. (2011). GPS displacement vectors from Avallone et al. (2004). Figure is from Beckers et al. (2015)……………………………………………………………………………………. 43
2.3 (a) Geological map of the Corinth Rift (EHF: East Helike Fault; KrF: Krathis Fault; DF: Derveni Fault; PMF: Pirgaki-Mamoussia Fault) showing the Hellenide nappes (pre-rift basement) on which the rift is superimposed. (b) Regional context of the Corinth Rift above the active Hellenic subduction zone and lying between the Cefalonia Fault (KF) and the North Anatolian Fault (NAF). Subduction of the African plate generated the Aegean Volcanic Arc (AVA). The dotted black box in (a) delimits the inset map (c) with selected seismic profiles from seismic reflection data. The numbering of off-shore seismic lines is that of Taylor et al. (2011). The red box in (a) delimits the zone of on-shore – off-shore correlations that are presented in Figure 2.4. Figure from Hemelsdael & Ford (2014)………………………………………………………... 47
2.4 Stratigraphic correlation between off-shore and on-shore domains in the central Corinth Rift basin. Right panel: off-shore syn-rift succession interpreted from seismic profile L07 (see Figure 2.3) after time-depth conversion. Sediments are divided into two main sedimentary units, Unit A and Unit B. The ages of the reflection boundaries are taken from Taylor et al. (2011). Left panel: synthetic sedimentary and lithostratigraphic log of the on- shore syn-rift stratigraphy around the Akrata area, following Ford et al. (2013). Rose diagrams represent foreset dip directions of the Middle group and Upper group Gilbert-type deltas. Question marks are used for uncertain correlations. Both logs are represented at the same vertical scale. Figure from Hemelsdael & Ford (2014)………………………………………………………………. 48
xix
2.5 Un-interpreted and interpreted complete seismic lines L14 (a) and L29 (b) (see Figure 2.3). Colors are those used in the seismic log in Figure 2.4. Figure from Hemelsdael & Ford (2014)………………………………………………………………. 49
3.1 The mode noise model across the Hellenic Unified Seismological Network (HUSN) (red line) as it was estimated by Evangelidis et al. (2012) from the minimum of all station probability density function (PDF) mode noise levels. The corresponding PDF mode noise model for the continental United States (blue dashed line) as estimated by McNamara & Buland (2004) is shown as bold dashed line. The shaded zone marks the area between the minimum tenth and ninetieth percentiles of all of the HUSN station power spectral density distributions, representing the 80% confidence interval of the minimum noise levels in Greece. An approximate classification of the seismic noise according to the frequency content is also presented. Figure is from Evangelidis et al. (2012)……………………………………………………………………………………….. 56
3.2 Schematic illustration of an ideal situation when the spatial distribution of the noise sources (circles) is homogeneous and the cross-correlation function of the two ambient noise signals is symmetric. The positive and negative lags
correspond to the empirical Green's function between and . Figure from Garnier & Papanicolaou (2009)………………………………………………………… 58
3.3 Schematic illustration of a case in which the noise sources are spatially localized and the cross-correlation function is not symmetric. Figure from Garnier & Papanicolaou (2009)………………………………………………………… 59
3.4 Schematic illustration of a case in which the noise sources are spatially localized and the orientation of the station-pair is perpendicular to the main direction of the energy flux from the ambient noise sources. In this case the coherent part of the cross-correlation function can by difficult to distinguish. Figure from Garnier & Papanicolaou (2009)…………………………………………. 60
3.5 Comparison between group and phase velocities. (a) Definition of group (U) and phase (V) velocities and (b) arrivals of dispersive surface-waves at different receivers. Figure is from Sheriff and Geldart (1995)……………….. 62
4.1 Map of the western Corinth Rift and seismic stations used in this study. Broadband seismometers are displayed with triangles and short-period
seismometers are displayed with rectangles, where green, red and blue colors
xx
signify stations operated by the University of Patras Seismological Laboratory (UPSL), the Corinth Rift Laboratory (CRL) and the Seismological Laboratory of Athens University (NKUA), respectively. The inter-station path coverage of all available stations is also presented (bottom map)………………………………. 66
4.2 A general schematic representation of the data processing scheme………...... 67
4.3 Example of a 1-day long trace of ambient seismic noise (raw data) recorded at SERG station. Black arrows mark seismic events...... 68
4.4 Example of a raw (upper panel) and a spectrally whitened (bottom panel) between 0.07 and 2.5 Hz amplitude spectra for 1-day long vertical component data of SERG station...... 70
4.5 Simple schematic illustration of the empirical Green's function extraction process from ambient seismic noise. After the necessary pre-processing of the vertical components of two daily recordings of noise data from AGRP and KALE stations, the cross-correlation computation resulted in the emergence of the fundamental mode of the Rayleigh-waves (grey circle)...... 71
4.6 Example of Rayleigh-wave energy emergence for increasingly long time-series. Stacking of daily cross-correlations between the vertical-vertical components of AGRP and KALE stations are presented (band-pass filtered between 0.4 and 1 Hz)...... 73
4.7 Example of Rayleigh-wave vertical-vertical reference (stacked over 3 yrs) correlation functions sorted by increasing inter-station distance (band-pass filtered between 0.07 and 2.5 Hz)...... 73
4.8 Daily cross-correlations between the station pair AGRP- KALE. A characteristic example of a time-shift which was detected after the preprocessing and cross-correlations computation. The red arrows indicate the ~8 s time-shift observed between the ~680th and the ~800th day (the duration of the studied time-period is 1095 days, hence day 0 corresponds to 01/01/2012 and day 1095 to 31/12/2014)…………………………………………………………….. 74
4.9 Daily cross-correlations between the station pair AGEO - LAKA. A characteristic example of rejecting cross-correlations of bad quality (black braces) (the duration of the studied time-period is 1095 days, hence day 0
corresponds to 01/01/2012 and day 1095 to 31/12/2014)……………………………. 75
xxi
Cross-correlations between (a) AGRP and PANR stations and between (b) 4.10 AGEO and AIOA stations from ~3 yrs of ambient noise filtered for different period bands and computed for the vertical components of noise records. (c) and (d) amplitude spectra of AGRP and PANR, and AGEO and AIOA cross- 77 correlations…………………………………………………………………………………
4.11 Three characteristic examples of dispersion curve picking on the frequency- time (period-velocity) diagrams for vertical-vertical component correlation between the station-pairs (a) EFP - PSAR, (b) AIOA - TRAZ and (c) SERG - TEME. The colored background represents the time-frequency diagram with the warm colors showing the large amplitudes. Black dots represent the relative maxima of the diagram, while the white circles highlight the automatic picking of these points. The black line is a five-order polynomial fitting to the automatic picks…………………………………………………………… 78
4.12 Every Rayleigh-wave dispersion curve measured in this study is plotted on a frequency-time diagram (a). The average dispersion curve with its standard deviation is also shown as thick black line. A diagram showing the number of measurements as a function of the period (b) and (c) a probability density function plot of the dispersion curves…………………………………………………. 79
4.13 Rayleigh-wave group velocity measurements and ray-path coverage at 1 s, 2 s, 3 s, 4 s, 5 s and 6 s, respectively. The thick red line shows the limit of the study area connecting cells of the geographical grid with no measurements. Seismic stations are shown as black triangles………………………………………………..... 84
4.14 Velocity error introduced by excluding the topography and the bathymetry of the study area. The black curve shows the cumulative number of ray-paths with errors smaller than a certain threshold. 100% of the inter-station paths have error smaller than 1%, while the 90% of them has errors smaller than 0.5%...... 85
4.15 The variance reduction of the travel-time residuals between 1 and 6 s with 0.02 s step………………………………………………………………………………….. 86
4.16 Rayleigh-wave group velocity maps at 1 s, 2 s, 3 s, 4 s, 5 s and 6 s. Seismic stations are shown as black triangles. For each period, the variance reduction (VarRed) between data computed from an initial homogeneous model with an
average velocity and the final model, as well as the final mean velocity
xxii 86
(Vmean) are shown at the lower left corner of each frame. Major fault traces shown are: 1=East Eliki, 2=West Eliki, 3=Psathopyrgos, 4=Marathias, 5=Trizonia, 6=Aigion, 7=Kamarai/Selianitika. The respective fault numbers are shown in 4.16a………………………………………………………………………... 87
4.17 The ray-path density at 1 s, 2 s, 3 s, 4 s, 5 s and 6 s. Seismic stations are shown as blue triangles………………………………………………………………….. 88
4.18 Spatial resolution maps at 1 s, 2 s, 3 s, 4 s, 5 s and 6 s. Seismic stations are shown as red triangles…………………………………………………………………… 90
4.19 Schematic illustration of the misfit calculation between a theoretical dispersion curve (thick black line) and a measured dispersion curve (thin gray line) with its uncertainties. The misfit value is the normalized area (dS1 + dS2) / S. Figure is from Mordret et al. (2014)...... 92
4.20 Inverted shear-wave velocity model for the whole study area. (a) Synthetic dispersion curves overlaid by the average local dispersion curve with error bars and (b) the associated inverted models. Both synthetic dispersion curves and models are colored according to their misfit. The thick black lines are the dispersion curve and model with the minimum misfit value……………………… 93
4.21 (a) An example of the 2D marginals of all pairs of parameters derived from the 1D inversion at an individual point of the study area. The model was parameterized with four layers over a half space with seven unknown parameters, the velocities in the four layers (V1, V2, V3 and V4 (m/s)) as well as the depth of the three associated interfaces (D1, D2 and D3 (m)). The images depict the sampling procedure of the Neighborhood Algorithm across each 2D parameter-space of the whole model-space and the gradual sampling towards the most promising regions with the lowest misfit. The colors correspond to the misfit value (m/s). Note the relative difficulty of the Neighborhood Algorithm to converge the sampling within the 2D parameter- spaces between the depth and the velocities of the superficial layers. The lower panel (b) shows the evolution of the misfit during the inversion. After eleven iterations the misfit achieves the lowest values and does not improve any more. The colors correspond to the number of models for each iteration…………………………………………………………………………………….. 94
xxiii
4.22 Vertical cross-sections along the grey lines sketched in Figure 4.16a according to the results of the depth inversion. The thick dashed black lines show the intersection between the profiles. Major faults (black lines) are presented and numbered as in Figure 4.16a. The faults geometry approximation is taken from Bernard et al. (2006)………………………………………………………………. 95
4.23 Slices through the 3D shear-velocity models at (a) 800m, (b) 2000m, (c) 3900m and (d) 5100m below sea level. Major fault traces are presented and numbered (top left panel) as in Figure 4.16a……………………………………………………… 96
5.1 On-shore and off-shore cross-section presented from west (a) to east (b) which cut through the easternmost edge of the study area in a ~N-S direction. Colors are those used in Figures 2.4 and 2.5 in Chapter 2. EHF: East Heliki Fault; PMF: Pirgaki-Mamoussia Fault; DF: Derveni Fault and AF: Aigion Fault. See Figure 2.3 in Chapter 2, for the exact location of the presented faults and profiles. Figure from Hemelsdael & Ford (2014)…………………………………….. 102
5.2 Superposition of the average 1D Vs model derived from the present study and the three crustal models proposed for the western Corinth Rift by Latorre et al. (2004), Rigo et al. (1996) and Novotny et al. (2000)…………………………………. 106
5.3 Tomographic models of shear-wave velocities in the western Corinth Rift derived by (a) Latorre et al. (2004) and (b) Gautier et al. (2006). Map views show velocity layers between 0 and 5 km depth. Stations are plotted as triangles and major faults are drawn in dark lines (Psa: Psathopyrgos fault; Ai: Aigion fault; He: Heliki fault and Py-Ma: Pyrgaki-Mamoussia fault). Earthquakes are displayed as dark dots. Parts of the models not crossed by rays have been masked. Well-resolved areas are outlined with a white contour. Figure modified from Latorre et al. (2004) and Gautier et al. (2006)…………….. 108
5.4 Schematic structural cross-section through Aigion fault, in the Aigion harbor. On the left, depths have been estimated from seismic refraction data (Naville et al. 2004) while on the right depths correspond to cuttings and core analysis (Lemeille et al. 2004; Rettenmaier et al. 2004). Velocities indicated on the right part of the figure correspond to P-wave arrivals from sonic logs, while on the left part velocities are those estimated from seismic refracted data. The pore pressure values correspond to measurements obtained during the drilling
phase for the platy limestone and after drilling operation for the limestone
xxiv
encountered below the fault…………………………………………………………….. 110
6.1 Schematic illustration of shear-wave splitting caused by the presence of stress-
aligned microcracks oriented parallel to the maximum horizontal stress, σH.
The fast shear-waves are typically parallel to the direction of σH. σh and σv denote the minimum horizontal stress and the vertical stress, respectively. Figure modified after Crampin & Peacock (2008)…………………………………… 119
6.2 Schematic illustration of the Anisotropic Poro-Elasticity (APE) model of the evolution of the geometry of stress-aligned fluid-saturated micro-cracks under
changing conditions of increasing stress, σH. The mechanism of deformation is fluid-movement along pressure gradients between neighboring micro-cracks. Figure modified after Crampin & Peacock (2008)…………………………………… 119
6.3 Schematic illustration of the shear wave window. The incidence angle is the angle from vertical at which the shear wave energy is arriving at the recorder. The shear wave window is the vertical cone (Volti & Crampin 2003) extending downward from the seismic station……………………………………………………. 122
7.1 Map of the study area in western Corinth Rift. Seismic stations used for the SWS analysis are shown as triangles, where green and red colors signify stations operated by the University of Patras Seismological Laboratory (UPSL) and Corinth Rift Laboratory (CRL) in collaboration with the Seismological Laboratory of Athens University (NKUA), respectively. The seismic events (colored circles) from which the valid splitting results were obtained, the Efpalio earthquakes epicenters (Efp1 and Efp2 as stars), major cities (squares) and major fault traces of the area are also shown. As in Doutsos & Poulimenos (1992), Flotté et al. (2005), Papanikolaou et al. (1997), Valkaniotis (2009) the major faults shown are: 1=Psathopyrgos, 2=Trizonia, 3=Trikorfo, 4=Filothei, 5=Marathia, 6=Antirio, 7=Drosato, 8=Efpalio, 9=Selianitika, 10=Aigion and 11=off shore fault related to Efpalio sequence (according to Sokos et al. 2012) and other on- and off-shore faults. The orientation of the principal stress axes after Kokkalas et al. (2006) is shown at the top left-hand corner. The depths of the events are color coded according to the color scale (bottom-right). The diameters of the circles are proportional to the magnitudes……………………… 127
7.2 Two examples (a) and (b) of valid splitting measurements of the shear waves recorded at the EFP station for two events that occurred in January 19th and
28th, respectively. Upper panels: Contour diagrams of the cross-correlation
xxv
coefficient in the (φ, δt) space. The preferred solutions of (φ, δt) corresponding to the maximum value (dot) are shown within the 95% confidence regions (dotted lines). As measurement uncertainties/errors we consider the range of the possible extreme values of φ and δt, respectively, within the confidence region (red lines in 7.2a). Lower panels: Superposition of the horizontal components (upper traces), and the corrected fast and slow components (lower traces) once the splitting effects have been removed. Particle motions diagrams are shown to the right of each sub-panel……………………………………………… 131
8.1 The distribution of the studied seismic events that occurred before (a) and after (b) the Efpalio earthquakes. Vertical cross-sections along the study area depicting the previous seismicity are also presented (a1, a2, b1, b2)… The shaded rectangles represent the projections on the surface of the rupture areas of the Efpalio earthquakes (Efp1 and Efp2). The approximate dimensions of the rupture areas were calculated according to Wells & Coppersmith (1994). The dashed ellipses were created for the need of the spatial grouping of the data that are considered to be within and close to the rupture areas. The ellipses were designed approximately according to the slight expansion of the early aftershocks beyond the calculated boundaries of the rupture areas during the first ~30 days after the main shocks following Sokos et al. (2012). Seismic stations, major cities and major fault traces are also presented as in Figure 7.1…………………………………………………………………………………………… 134
8.2 Maps of the western end of the Corinth Rift showing rose diagrams of the measured fast shear wave polarization directions. Seismic stations, major cities and major fault traces are also presented as in Figure 7.1. (PERIOD I: time period before the 1st Efpalio event (January 2009 - Efp1), PERIOD II: time period after the 1st Efpalio event until the end of the aftershock sequence (Efp1 - end of May 2010) and PERIOD III: time period after the end of the aftershock sequence (June 2010 - December 2010))………………………………… 136
8.3 Diagrams showing the variation of the measured shear wave time delays δt
(a), fast polarization directions φ (b) and VP/VS ratios (c) with time from events located close/within the rupture areas. The approximate dimensions of the rupture area are shown in Figure 8.1. The time delays were normalized according to the hypocentral distances. Black vertical bars represent measurement errors……………………………………………………………………… 139
xxvi
8.4 Diagrams showing the variation of the measured shear wave time delays δt
(a), fast polarization directions φ (b) and VP/VS ratios (c) with time from events located outside the rupture areas. The approximate dimensions of the rupture area are shown in Fig. 8.1. The time delays were normalized according to the hypocentral distances. Black vertical bars represent measurement errors……... 141
8.5 Diagrams showing the variation of (a) the measured shear wave time delays
δt, (b) fast polarization directions φ and (c) VP/VS ratios from January 2009 to December 2010 for event multiplets. The time delays were normalized according to the hypocentral distances. Black vertical bars represent measurements errors…………………………………………………...... 143
8.6 Diagrams showing the variation of the measured shear wave time delays δt
(a), fast polarization directions φ (b) and VP/VS ratios (c) from all the available data recorded between January 2009 and December 2010 per station. The time delays were normalized according to the hypocentral distances. Black vertical bars represent measurement errors. The names of the stations are shown at the top left corner of each panel………………………………………………………... 145
8.7 Diagrams showing the variation of the measured shear wave non-normalized time delays δt with time. Black vertical bars represent measurement errors……………………………………………………………………………………….. 147
8.8 Diagram showing the variation of the measured shear wave non-normalized (raw) time delays from the whole dataset. Black vertical bars represent measurement errors……………………………………………………………………… 147
8.9 Diagrams showing the variation of the measured shear wave non-normalized (raw) time delays δt per station. Black vertical bars represent measurement errors. The names of the stations are shown at the top left corner of each panel………………………………………………………………………………………… 148
8.10 Equal area projections of the fast polarization directions. (a) whole time period (Jan 2009 - Dec 2010), (a1) time period before the 1st Efpalio event (January 2009 - Efp1 (Jan. 18, 2010)), (a2) time period after the 1st Efpalio event until the end of the aftershock sequence (Efp1 - end of May 2010) and (a3) time period after the end of the aftershock sequence (Jun 2010 - Dec 2010). The radius of the plots is scaled to an incidence angle of 45°…………………………… 149
xxvii
8.11 Waveforms of "earthquake multiplets" or "similar earthquakes" recorded by the EFP, SERG and ROD stations. We considered the similar earthquakes/earthquake multiplets as a pair of earthquakes consisting of an earthquake that occurred before the Efpalio events and an earthquake that occurred after that, with a cross-correlation coefficient greater than 0.7, similar magnitude and spaced in distances less than the mean horizontal and vertical location error. A 0.05 - 5 Hz band-pass filter was used in pre- processing of the seismic waveforms as the waveforms cross-correlation coefficient is more stable at lower frequencies (e.g., Shearer 1997; Shearer et al. 2005)…………………………………………………………………………………….. 152
9.1 The percentage of the relative change between the observed average VP/VS
ratios, VP and VS velocities and the background values derived from the velocity model proposed by Latorre et al. (2004). Validation of the variation of
the average values of the splitting parameters and VP/VS ratios that was observed through the studied time period, after the application of non- parametric hypothesis testing is also presented. A two-sample Kolmogorov-
Smirnov (KS) test (Gibbons 1971) was applied for the time delays and VP/VS ratios, while a statistical test relative to directional data (Trauth 2010) was applied for the polarization directions (for details about the statistical testing, see Appendix B)…………………………………………………………………………… 159
9.2 The same as presented in Figure 9.1, but for each station separately...... 161
9.3 A plot showing the variation of the distances r of the January 18th to February 25th, 2010 aftershocks from the focus of the first Efpalio event (January 18th, 2010) versus their occurrence times t…………………………………………………. 163
xxviii
List of Tables
7.1 Data availability (shear-wave splitting measurements) for the studied time period. Months with at least some data are plotted in grey…………………………………………………………………………………. 129
8.1 Summary of the average values of the shear-wave splitting parameters measured per seismic station for the whole dataset………………………… 137
8.2 Comparison of the average values of the studied parameters (φ, δt,
VP/Vs, VP and Vs) before and after the Efpalio earthquakes, derived from the earthquake multiplets………………………………………………... 144
8.3 Pairs of earthquake multiplets with the corresponding splitting
parameters (φ and δt) as well as the corresponding values of VP/Vs ratio,
VP and Vs…………………………………………………………………………... 151
xxix
List of Abbreviations (in alphabetical order)
APE Anisotropic poro-elasticity: a model for the stress-induced evolution of fluid-saturated micro-cracked rock
ANT Ambient Noise Tomography
CC(s) Cross-correlation(s)
DC(s) Dispersion curve(s)
EDA Extensive dilatancy anisotropy: the distributions of stress-aligned micro- cracks in almost all in situ rocks in the Earth's crust
EGF Empirical Green's Function
FTAN Frequency-Time Analysis
GF Green's function
PSI Passive Seismic Interferometry
SNR Signal-to-noise ratio
SWS Shear-wave splitting
VP P-wave velocity
VS S-wave velocity
δt Time delay
σH Principal axes of maximum horizontal stress
φ Fast polarization direction
xxxi
Chapter 1
Introduction Thesis Outline
1.1 Introduction
The Corinth Rift is an active intra-continental structure on the western edge of the
Aegean Arc, Greece. It is a region which has been in the limelight of the geo-scientific interest for more than three decades, as it is one of the most seismically active rifts in the Euro-Mediterranean region, presenting one of the highest geodetically measured rates of extension. Despite the large amount of multidisciplinary observations and the proposed models, there is still a considerable discussion concerning the relationship between seismic activity, fault mechanics and the rifting process. In the framework of the ongoing research aimed to understand the tectonic evolution of the area, this thesis is trying to investigate the velocity structure and the physical characteristics of the upper crust beneath the western Corinth Rift derived by two different techniques:
Passive Seismic Interferometry using ambient seismic noise and Shear-Wave
Splitting using locally recorded shear-waves. The dissertation is organized into two major parts wherein each of the two aforementioned techniques is analyzed.
1.2 Passive Seismic Interferometry - Part I
In the frame of the first part, we use passive seismic interferometric techniques to constrain the velocity structure in the upper crust of the western Corinth Rift. The method of Seismic Interferometry we refer to, aims to retrieve the Green's function of the medium between two sensors by just utilizing ambient seismic noise recordings at
33 Chapter 1 Introduction Thesis Outline the Earth's surface. Seismic Interferometry is considered as a revolutionary method characterized by a rapid development especially during the last decade. The idea was first proposed by Claerbout (1968) according to which "by cross-correlating random wave fields of ambient seismic noise recorded on two locations on the Earth's surface, we can retrieve the wave field that would be recorded at one of the locations if there was a source at the other". This conjecture was first proved by Wapenaar (2003), and since then, a plethora of studies have been performed using ambient noise Seismic
Interferometry showing how useful the Green's function extraction can be in practical applications, either by monitoring temporal changes in material properties associated with volcanic processes and fault zones or by imaging the subsurface of the Earth at different scales (from reservoir to continental scale). Following this rule, we applied
Passive Seismic Interferometry technique on long time-series of ambient seismic noise recorded at all available seismic stations which are deployed in the western Corinth
Rift. In general, we cross-correlated all vertical-vertical component time-series from all stations, turning each station into a virtual source emitting Rayleigh-waves
(Green's functions). Then Rayleigh-wave group velocity dispersion curves were measured for each station-pair by applying frequency-time analysis and finally we performed straight-ray surface-wave tomographic inversions to produce an ambient noise tomography of the western Corinth Rift at different periods between 1 and 6 s.
Since no other ambient noise tomography has been performed until today in the area, the first part of this thesis is considered as the first attempt to study the crustal velocity structure of the Corinth Rift by using ambient seismic noise recordings. In general this thesis is the second ambient noise tomography performed on a local scale in Greece, after a high-frequency ambient noise tomography of Migdonia basin, northern Greece, by Hannemann et al. (2014). In general, our results highlight the
34 Chapter 1 Introduction Thesis Outline complexity of the geological and tectonic regime of the western Corinth Rift.
Tomographic images at periods up to ~3 s (up to ~2.5 km depth) revealed that the overall distribution of the Rayleigh-wave velocity is mostly coincident with the geology context of the area. Low velocity zones are mostly observed where Plio-Quaternary syn-rift sediments are present, while higher velocities are due to the pre-rift basement structures. At periods above ~3 s, where Rayleigh-wave velocity measurements are sensitive to deeper structures, the results highlight a low velocity zone located in the southern part of the Corinth Rift with a preferential elongation in the WNW-SSE direction, sub-parallel to the strike of the rift. The low velocity anomaly is the most interesting and profound feature in the velocity maps between 4 and 6 s. Interestingly, this zone is also highlighted at depths even greater than ~5 km (especially around
Aigion area), namely much deeper than the estimated sedimentary layer in the area.
We interpret the presence of the low velocity anomaly below the southern part of the rift, in relation with the present-day active tectonic regime giving also special attention on possible involvement of fluid circulation processes at depth within a highly fractured crust.
1.3 Shear-Wave Splitting Part II
Motivated by previous studies in the literature noting that possible fluid circulation processes at depth play a key role in the overall evolution of the western Corinth Rift
(crustal stress, faulting, seismicity etc.) and prompted by the observations derived from our ambient noise tomography study, in the second part of this thesis, we concentrated on local shear-waves to investigate possible changes in the seismic waves propagation properties of the crust, related to the occurrence of seismic events in the area. Shear-wave splitting is a phenomenon in which shear-waves are separated into
35 Chapter 1 Introduction Thesis Outline two components with almost vertical polarizations and different propagation velocities. This can occur during shear-wave propagation through an anisotropic medium (Crampin & Chastin 2003; Crampin & Peacock 2005). The two splitting parameters that can be measured through a shear-wave splitting analysis are the polarization direction of the fast component of the splitted shear-waves, and the time- delay between the two components. Significant variations of the crustal anisotropy parameters have been observed in relation to earthquakes worldwide, reflecting spatial and/or temporal changes in the characteristics of the medium and the stress field. Towards this purpose, we performed a local shear-wave splitting analysis in a time-period that includes the occurrence of two moderate earthquakes. On January
2010 two earthquakes of MW ~5.2 occurred near the village of Efpalio located at the northern coast of the western Corinth Rift. For the needs of our study, we used a 2- year long dataset, covering one year before and one year after the January 2010
Efpalio earthquakes, to study the temporal variability of the splitting parameters related to the earthquakes occurrence. In general, our analysis revealed the presence of an anisotropic upper crust in the western Corinth Rift, as well as a significant temporal variation of the splitting parameters in relation with the earthquakes. In order to have additional information about the average properties of the medium along the studied ray- paths, we accompanied each shear-wave splitting measurement with apparent VP, VS and average VP/VS ratio calculations. A distinct increase in time-delay values and VP/VS ratios was observed soon after the Efpalio earthquakes, followed by a decrease after the end of the aftershock sequence. The measurements of the apparent
VP and VS showed that the observed changes in VP/VS ratios after the Efpalio earthquakes were due to an increase in VP and a decrease in VS. Considering the above, we attempted to interpret the causative factors of the observed temporal
36 Chapter 1 Introduction Thesis Outline variations associated with the Efpalio earthquakes, in terms of the regional stress field and fluids involvement. We suggest that a migration of over-pressured fluids through the earthquake produced fractured damage zone is most probably the main cause of the observed increase in time delays and VP/VS ratios. The observed variations in the time-delays and VP/VS ratios after the Efpalio earthquakes seemed to be slightly stronger close to the hypocentral areas, which are possibly reflecting minor post-earthquake related changes in the properties of the crust. Fast shear-wave polarizations present a general E-W orientation, which is in agreement with the regional stress field. On the contrary, the Efpalio earthquakes seemed to have little or no influence on this parameter since fast polarization directions did not present any significant change after the Efpalio earthquakes.
1.4 Thesis objectives Outline
In the view of the above background, the present thesis revolves around the following objectives:
To present the first noise-based surface-wave tomography of the western Corinth
Rift.
To attempt constraining the 3D shear-velocity structure of the upper crust under
the western Corinth Rift by means of ambient noise tomography.
Probe possible relationship between regions of low velocity zones at depth with the
present-day tectonic regime and possible fluid circulation processes.
To study the anisotropy pattern in the upper crust of the western Corinth Rift.
To investigate possible temporal changes in seismic propagation velocities (seismic
anisotropy) associated with earthquake occurrences.
37 Chapter 1 Introduction Thesis Outline
To examine possible factors and processes that lead to changes in the anisotropy
parameters.
To give a perspective on the use of noise-based imaging techniques in the
investigation of the structure of the Corinth Rift, considering the method as an
additional and important tool that can provide complementary information in
addition to the existing knowledge.
To make a contribution to our understanding of the crustal structure and
deformation in the western Corinth Rift.
Set the basis for further investigations and give a motivation for future research
studies to exploit the observations of the current thesis.
As it is already mentioned, the thesis is divided in two major parts and it is structured, including the current chapter, into 10 chapters.
In Chapter 2 we introduce the structure of western Corinth Rift which is the target area of this thesis. We provide information about the general tectonic and geological setting of the area.
Chapter 3 is the start of the first major part of this dissertation. This chapter gives an introduction to the field of Passive Seismic Interferometry. We discuss about the nature of ambient seismic noise, the principles, theory and potential applications of
Passive Seismic Interferometry.
In Chapter 4 it is described the first application of ambient noise tomography in the western Corinth Rift. We describe the pre-processing/preparation of the ambient noise data and the methods used to extract Rayleigh-wave group velocity dispersion curves from cross-correlations between all available pairs of stations. After describing the
38 Chapter 1 Introduction Thesis Outline tomography formalism adopted for our inversions, we regionalize the group velocity measurements into 2D Rayleigh-wave group velocity maps. Finally we locally invert these maps to obtain a simple four-layer 3D shear-wave velocity model of the upper ~8 km of the crust.
In Chapter 5 we interpret and discuss the tomography results in relation with the geotectonic regime of the area. Since no other noise-based tomography study was performed in the area, we compare and discuss our results in the light of previous earthquake tomography studies.
Chapter 6 constitutes the beginning of the second part. This chapter provides an introduction to Shear-Wave Splitting. The underlying theory of Shear-Wave Splitting phenomenon, the parameters which can be determined through a local shear-wave anisotropy analysis as well as potential applications of the phenomenon.
In Chapter 7 we perform a shear-wave splitting analysis, including apparent VP, VS and average VP/VS ratio calculations in the epicentral area of the January 2010
Efpalio earthquakes. After providing information about the Efpalio earthquakes, we describe the methodologies we adopted for both splitting parameters and VP/VS ratios measurements.
Chapter 8 provides a detailed representation of the results derived from the anisotropy analysis giving major emphasis on the temporal variability of the derived anisotropy parameters and VP/VS ratios before and after the Efpalio earthquakes occurrence.
39 Chapter 1 Introduction Thesis Outline
In Chapter 9 we make a detailed interpretation and discussion of the shear-wave splitting analysis. We interpret and discuss the implications of our understanding of the anisotropy pattern in the study area.
Chapter 10 is the conclusion chapter (Epilogue) of this thesis. We briefly summarize our understanding of the structure and deformation mechanism in the upper crust of the western Corinth Rift and we discuss possible future research.
40
Chapter 2
Corinth Rift
2.1 Geotectonic Seismotectonic setting
Fig. 2.1: Plate boundaries and the general geodynamic pattern of the broader Hellenic region (continental Greece, Ionian and Aegean Seas). Figure modified from Vött (2007).
The tectonic and geological setting of central and western Greece is complex. This complexity is due to the co-existence of different types of tectonic regimes (extensional, contractional, strike-slip, transpressional, transtensional etc.) all within a relatively small region. The tectonic framework of western Greece is dominated by the large dextral strike-slip fault off the coast of Cefalonia Island, where the change between continent-continent collision in the north and ocean-continent subduction in the south occurs (Underhill 1989; Sachpazi et al. 2000 among many others) (Figure 2.1). During
41
Chapter 2 Corinth Rift
the Eocene, central and western Greece was affected by the Alpine collision, which led to the formation of the Hellenic mountain range (Doutsos et al. 2006). Mesozoic and
Early Cenozoic carbonate rock propagated westwards along north-south-striking and east-dipping thrust faults (Xypolias & Doutsos 2000). From the early Pliocene to the present, an extensional stress field has progressively prevailed in the previous compressional tectonic regime in central and western Greece (Doutsos & Kokkalas
2001). As a result, numerous extensional basins started to develop during the Late
Plio-Pleistocene. Today a series of ~E-W-trending graben systems cut the Hellenic mountain chain. The most prominent of which are the Amvrakikos Gulf, the Aitolo-
Akarnanian basin, the Gulf of Patras and the Corinth Rift (Kokkalas et al. 2006)
(Figure 2.1).
The Corinth Rift (CR) is a continental rift which separates main-land central Greece to the north from Peloponnese to the south (Figures 2.1 and 2.2). It is approximately
120 km long and 10-20 km wide, with an E-W orientation, extending from the Gulf of
Patras in the west, to the Gulf of Alkionides in the east. It is one of the most seismically active extensional rifts in the Euro-Mediterranean region (Armijo et al.
1996), with geodetically measured rates of extension varying from ~5 mm/yr at the eastern part to ~15 mm/yr at the western part and localized on a 10-20km wide zone (e.g.
Briole et al. 2000; Avallone et al. 2004; Bernard et al. 2006). Due to the observed high extension rates, the Corinth Rift has been identified as a site of major importance for earthquake studies. Several studies have been carried out investigating the cause of the high strain rates of the Corinth Rift. Some researchers have linked the rift formation to the westward motion of the Anatolian micro-plate and the propagation of the North Anatolian fault (e.g. Jackson 1994; Le-Pichon et al. 1995). Other studies related the rift formation to the roll-back of the subducting African plate
42
Chapter 2 Corinth Rift
Fig. 2.2: Tectonic map of the western Corinth Rift. Major onshore and offshore fault traces of the area as in Ford et al. (2013), Palyvos et al. (2008), Flotté et al. (2005), McNeill et al. (2005), Bell et al. (2009), and Taylor et al. (2011). GPS displacement vectors from Avallone et al. (2004). Figure is from Beckers et al. (2015).
43
Chapter 2 Corinth Rift
(e.g. Le-Pichon & Angelier 1979; Hatzfeld et al. 1997) and others to a combination of the previous processes (e.g. McClusky et al. 2000; Doutsos & Kokkalas 2001). GPS measurements (e.g. Briole et al. 2000; Avallone et al. 2004; Bernard et al. 2006), earthquake focal mechanism solutions (e.g. Rigo et al. 1996; Bernard et al. 1997;
Godano et al. 2014), as well as neo-tectonic fault analyses (e.g. Jackson et al. 1982;
Doutsos & Poulimenos 1992; Kokkalas et al. 2006; Zygouri et al. 2008) indicate an approximately N-S direction of extension.
The high extension rates in the Corinth Rift, especially at the western part, are accompanied by a high level of seismic activity which is characterized by frequent earthquake swarms and also by the occurrence of moderate to large earthquakes.
Although the Gulf of Corinth is geographically limited, numerous on- and offshore earthquakes with magnitudes up to 6.7 were instrumentally recorded or historically reported (Papadopoulos 2000). Some of the large and destructive earthquakes that occurred during the last decades are the Alkionides seismic sequence in 1981 involving three strong events of MS 6.7, 6.4 and 6.4 (Jackson et al. 1982), the MS 5.9 Galaxidi earthquake in 1992 (Hatzfeld et al. 1996), the MS 5.4 Patras earthquake in 1993
(Tselentis et al. 1994; Karakostas et al. 1994) and the MS 6.2 Aigion earthquake in
1995 (Tselentis et al. 1996; Bernard et al. 1997). The CR has major north-dipping normal faults located on the southern margins and antithetic south-dipping faults on the northern margins. The major north-dipping faults are organized in a right- stepping en echelon pattern with a strike that ranges between W-E and WNW-ESE and a steep dip of 55o-70o, resulting in the subsidence of the northern coast and on the upward displacement of the main footwalls (Armijo et al. 1996; Kokkalas et al. 2006;
Zygouri et al. 2008). Such case is the Eliki-Aigion-Kamarai-Psathopyrgos fault system
(Figure 2.2), which is one of the youngest fault systems in the Corinth Rift, related to
44
Chapter 2 Corinth Rift
the opening phase of the gulf that initiated during Pliocene (Micarelli et al. 2003;
Hemelsdael & Ford 2014).
Even though the Corinth Rift has been studied extensively and many tectonic models have been proposed, the understanding of the crustal structure is such that the relationship between major faults and seismicity at depth is still an open debate.
Regarding the fault geometry four models have been proposed, among others. For instance, Rigo et al. (1996), following King et al. (1985) and Doutsos & Poulimenos
(1992), proposed that the micro-seismicity is related to a low-angle seismically active detachment zone lying at ~9-11 km depth. According to Rigo et al. (1996) micro- seismicity in the western part of the Corinth Rift occurs at the intersection between the major north-dipping normal faults that crop out along the southern part of the rift and a low-angle (15o ±10o) north-dipping detachment zone (see fig. 12 in Rigo et al.
1996). On the contrary, Sorel (2000) suggested a model according to which an active low-angle detachment fault, more than ~70 km long, accommodates most of the extension of the Corinth Rift (Khelmos detachment, see fig. 3 of Sorel 2000). Following this speculative assumption, all other north-dipping normal faults cropping out on the
Peloponnese are considered secondary listric structures which intersect the Khelmos detachment at a relative shallow depth only. Hatzfeld et al. (2000) suggested that micro-seismicity is probably related to a seismic-aseismic (brittle-ductile) transition zone at ~8-12 km depth resulted from the rapid extension rate of the western Corinth
Rift, and not to the low-angle north-dipping active faults as previously proposed. More recently, Lambotte et al. (2014) proposed a new mechanical model of the rifting process, which involves a non-elastic uniform deformation at depth under the Corinth
Rift's axis. According to this model, an aseismic N-S opening of the rift is coupled with the downward and northward growth of a yet immature detachment, not yet
45
Chapter 2 Corinth Rift
connected to the ductile lower crust, and contributes most of the opening rate measured from GPS at the surface. The reported seismicity fluctuation would possibly result from small strain instabilities, undetected by continuous GPS and possibly related to non stationary migration of fluid pulses.
The stratigraphy of the western Corinth Rift reflects the present-day and Plio-
Quaternary tectonic processes. Limestone nappes are outcropping almost everywhere to the north, whereas to the south these nappes are covered by thick layers of syn-rift sediments, which in some extent are subject to the rapid uplift. The limestone nappes only outcrop on the footwall of the southern active faults towards mainland
Peloponnese (Doutsos & Poulimenos 1992; Ghisetti & Vezzani, 2005) (Figure 2.2).
Several both on-shore (e.g., Leeder et al. 2012; Roberts et al. 2009; Ford et al. 2013) and off-shore studies (e.g., Sachpazi et al. 2003; Lykousis et al. 2007; Bell et al. 2008,
2009; Taylor et al. 2011) have been performed aiming to extract and quantify the syn- rift sedimentation structure of the Corinth Rift. For instance, multichannel 2D seismic profiles and high resolution 2D seismic data collected during the Maurice Ewing
(EW0108) 2001 geophysical survey (Taylor et al. 2011) and the 1997 SEISGRECE survey (Sachpazi et al. 2003) constrained successfully the geometry of the main faults and the off-shore syn-rift succession of the rift. Hemelsdael & Ford (2014) used on- shore sedimentary, structural and geomorphological data, as well as off-shore seismic reflection data from the 2001 geophysical survey (Taylor et al. 2011) to perform a stratigraphic correlation between off-shore and on-shore domains in the central
Corinth Rift basin. Selected seismic lines and the zone of the on-shore off-shore stratigraphic correlation are indicated in Figure 2.3.
The on-shore syn-rift sediments of the western and central Corinth Rift have been
46
Chapter 2 Corinth Rift
PMF
Fig. 2.3: (a) Geological map of the Corinth Rift (EHF: East Helike Fault; KrF: Krathis Fault; DF: Derveni Fault; PMF: Pirgaki-Mamoussia Fault) showing the Hellenide nappes (pre-rift basement) on which the rift is superimposed. (b) Regional context of the Corinth Rift above the active Hellenic subduction zone and lying between the Cefalonia Fault (KF) and the North Anatolian Fault (NAF). Subduction of the African plate generated the Aegean Volcanic Arc (AVA). The dotted black box in (a) delimits the inset map (c) with selected seismic profiles from seismic reflection data. The numbering of off-shore seismic lines is that of Taylor et al. (2011). The red box in (a) delimits the zone of on-shore – off-shore correlations that are presented in Figure 2.4. Figure from Hemelsdael & Ford (2014).
separated into three lithostratigraphic groups (Nixon et al. 2016 and references therein). A Lower Group characterized by alluvial to lacustrine sediments deposited in the late Pliocene, a Middle Group dominated by lacustrine fan deltas that built during a period of rift deepening and northward migration, and an Upper Group characterized by alternating marine and lacustrine sediments (e.g., Leeder et al. 2012;
Ford et al. 2013). Up to ~2.5 km of syn-rift sediments have accumulated in the off- shore Corinth Rift. These include two main sedimentary units, a lower thick "early- rift" unit and a shallower "late-rift" unit (Unit B and Unit A, respectively, as they were introduced by Taylor et al. 2011). These main sedimentary units are separated by a
47
Chapter 2 Corinth Rift
basin-wide unconformity, which is locally angular and marks an abrupt change in the seismic reflection character (Nixon et al. 2016). These units vary considerably in thickness across the rift. Figure 2.4 and 2.5 show the on-shore off-shore correlation of the aforementioned lithostratigraphic groups and units and examples of interpreted seismic profiles, respectively, both from Hemelsdael & Ford (2014).
Fig. 2.4: Stratigraphic correlation between off-shore and on-shore domains in the central Corinth Rift basin. Right panel: off-shore syn-rift succession interpreted from seismic profile L07 (see Figure 2.3) after time-depth conversion. Sediments are divided into two main sedimentary units, Unit A and Unit B. The ages of the reflection boundaries are taken from Taylor et al. (2011). Left panel: synthetic sedimentary and lithostratigraphic log of the on-shore syn-rift stratigraphy around the Akrata area, following Ford et al. (2013). Rose diagrams represent foreset dip directions of the Middle group and Upper group Gilbert-type deltas. Question marks are used for uncertain correlations. Both logs are represented at the same vertical scale. Figure from Hemelsdael & Ford (2014).
48
Chapter 2 Corinth Rift
Fig. 2.5: Un-interpreted and interpreted complete seismic lines L14 (a) and L29 (b) (see Figure 2.3). Colors are those used in the seismic log in Figure 2.4. Figure from Hemelsdael & Ford (2014).
Variations of the sediments thickness and the relative thickness of the two off-shore sedimentary units through the whole basin reveals the time-space evolution of Corinth
Rift infill (Sachpazi et al. 2003). The maximum sediment thickness was measured in the main basin north of the Xylocastro fault (Taylor at al. 2011) (see Figure 2.3).
49
PART I
Chapter 3
Passive Seismic Interferometry
3.1 Ambient Seismic Noise
The field of Seismology has considerably changed over the past few decades based on the expansion of the computational power with advancements in software and hardware. These advancements allowed seismologists to exchange, store and process in a more efficient way extremely large volume of data. This led as well to a significant improvement of the processing techniques and the mathematical representation of the seismic wave field. In the long run, all these improvements are being made for one main purpose among others, to improve the resolution and the accuracy of seismic imaging and thus to understand better the deformation of the crust.
One of the most common limitations in seismological studies and specifically in seismic imaging of the Earth's interior is the spatio-temporal distribution of seismic sources. With receivers being inexpensive and more manageable compared to active sources, and considering also the continuously increasing interest in study on areas of great geological and tectonic interest with no or low seismic activity, it has been the focus of one branch of Seismology to develop and apply "earthquake-independent" methodologies based on extracting the impulsive response solely between receivers.
Such methodology, which utilizes recordings of random wave fields, commonly called
Passive Seismic Interferometry (PSI), is becoming an increasingly well established tool to determine the seismic velocity structure of the Earth's subsurface.
In Seismology two types of signals are considered to be random wave fields. The first is seismic coda waves, which are mainly the multiple scattered parts of the seismic
53 Chapter 3 Passive Seismic Interferometry waveforms (Paul et al. 2005) and the second is ambient seismic noise, namely whatever is recorded when no identifiable active source is emitting, and which is superimposed on all recorded data (Curtis et al. 2006). The majority of the current literature deals much more with the utilization of ambient seismic noise and less with the investigation of the origins and the nature of seismic noise wave fields. Despite that, some studies have primarily concentrated on understanding the composition of seismic noise, achieving a significant progress on this topic (from Longuet-Higgins
1950 to Kedar & Webb 2005). A comprehensive literature review of the current knowledge about the ambient seismic noise wave field given by Bonnefoy-Claudet et al. (2006) revealed an overall agreement between older and recent studies about the origin of seismic noise and its frequency dependence. According to Bonnefoy-Claudet et al. (2006) and references therein, ambient noise has basically two different origins: natural or man-made/cultural and they can be classified according to their frequency content. Based on a general approximation, at periods above 1s (below 1 Hz) the sources are natural (ocean, large-scale meteorological conditions etc.), at intermediate periods 0.2 - 1 s (~1 to 5 Hz) the sources are either natural (local meteorological conditions e.g. local storms, air pressure changes, turbulent wind or wind-induced vibrations of trees/buildings) or man-made (urban) and at very low periods below 0.2 s
(higher frequencies > 5 Hz) the sources are essentially cultural. Cultural sources are numerous such as car and train traffic, industrial machines, explosions or the exploitation of underground (hydrocarbon, hot waters) reservoirs, just to name a few.
Another important classification of ambient seismic noise concerns the period band between ~4 and 20 s. Ambient seismic noise in the period band < 20 s, commonly referred to as microseisms, are the continuous oscillation of the solid ground produced by oceanic and atmospheric forcing which are recorded everywhere in the world
54 Chapter 3 Passive Seismic Interferometry independently from earthquake activity (Stutzmann et al. 2000). More specifically, the seismic noise spectra display two strong peaks at the microseisms band below 20 s, one pick around 14 s and another one around 7 s, which denote what are called respectively primary (10 - 20 s) and secondary (4 - 10 s) microseisms. The primary microseism is generated mostly near the coastlines when ocean waves reach shallow waters and interact with the shallow seafloor and the shore (Hasselmann 1963). The secondary microseism is the strongest noise peak with almost double-frequency signals relative to the primary microseism. It is generated by the interaction of coastline-reflected ocean gravity waves travelling in opposite directions (Longuet-
Higgins 1950). According to Nishida et al. (2008), Fukao et al. (2010) and others, surface-waves consisted of both Rayleigh and Love-waves, are present in significant quantities within the microseism band. Seismic noise of longer periods above 20 s, referred to as the "Earth's hum", is observed in the continuous background oscillations in long-period seismic spectra. Many researchers attributed this long-period ambient noise to atmospheric motions (e.g. Tanimoto & Um 1999; Ekstrom 2001), others (e.g.
Tanimoto 2005; Rhie & Romanowicz 2006) suggested that the origin of the long-period noise could be related to so-called ocean infragravity waves, long-period (>50 s) sea surface gravity waves, while Rhie & Romanowicz (2004) proposed that the origin may involves a coupling process between atmosphere, ocean and seafloor. The frequency content of ambient noise and the aforementioned frequency limits concerning the classification of the ambient seismic noise sources may vary seasonally and from site to site. A general picture of the ambient noise level and characteristics can be given by a power spectral density analysis and its corresponding probability density function
(Figure 3.1).
55 Chapter 3 Passive Seismic Interferometry
Fig. 3.1: The mode noise model across the Hellenic Unified Seismological Network (HUSN) (red line) as it was estimated by Evangelidis et al. (2012) from the minimum of all station probability density function (PDF) mode noise levels. The corresponding PDF mode noise model for the continental United States (blue dashed line) as estimated by McNamara & Buland (2004) is shown as bold dashed line. The shaded zone marks the area between the minimum tenth and ninetieth percentiles of all of the HUSN station power spectral density distributions, representing the 80% confidence interval of the minimum noise levels in Greece. An approximate classification of the seismic noise according to the frequency content is also presented. Figure is from Evangelidis et al. (2012).
The development of Passive Seismic Interferometry triggered a shift in the way seismologists think about the parts of the signal that until recently were discarded of most analyses. The origin and the nature of the ambient seismic noise wave field are essential as the use of seismic noise is more and more popular for seismic imaging purposes. Especially considering the growing number of processing techniques based on the noteworthy assumption that the noise wave field is predominantly consisting of fundamental mode Rayleigh-waves (Bonnefoy-Claudet et al. 2006).
56 Chapter 3 Passive Seismic Interferometry
3.2 Theory
Passive Seismic Interferometry is based on a conjecture stated initially by Claerbout
(1968), according to which the cross-correlation (CC) of ambient seismic noise recorded at two stations yields a deterministic wavelet, or an approximation (of the surface- wave part) of the Green's function (GF) of the medium between the two stations, as if one station was a source and the other a receiver. Since then, this conjecture has been both theoretically and experimentally proven and applied by many authors with different modifications (e.g. Weaver & Lobkis 2001, 2002; Wapenaar et al. 2002, 2004;
Shapiro & Campillo 2004; Shapiro et al. 2005; Roux et al., 2005; Wapenaar &
Fokkema 2006; Campillo & Paul 2003; Derode et al. 2003; van Tiggelen 2003; Malcolm et al. 2004; Snieder 2004 and many others).
Let and denote the time-dependent wavefields recorded by two stations at and positions. The cross-correlation function between the two wave fields over the time interval with a time lag is given (Garnier & Papanicolaou 2009) by:
(3.1)
In theory, the recovery of Green's functions from Seismic Interferometry using ambient seismic noise requires isotropic distribution of noise sources or energy of diffuse wave fields (e.g., Snieder 2004; Roux et al. 2005). However, in the physical world this requirement is difficult to satisfy, and for this reason the GF inferred from
Seismic Interferometry, hereinafter referred to as the empirical Green's function
(EGF), deviates from the true Green's function (GF). In a homogeneous medium with randomly distributed sources, it has been shown (Snieder 2004; Roux et al. 2005) that:
57 Chapter 3 Passive Seismic Interferometry
(3.2)
where G is the Green's function. The above approximate equality points out that a time-symmetrized Green's function can be obtained from the cross-correlation if there is enough noise source diversity (Garnier & Papanicolaou 2009, see Figure 3.2). In this
(ideal) case the wave field at any seismic stations is equipartitioned due to the superposition of uncorrelated plane waves coming from different directions. Then in particular the travel-time of the emerged wave field can easily be recovered from the singular support of the cross correlation.
Fig. 3.2: Schematic illustration of an ideal situation when the spatial distribution of the noise sources (circles) is homogeneous and the cross-correlation function of the two ambient noise signal is symmetric.
The positive and negative lags correspond to the empirical Green's function between and . Figure is from Garnier & Papanicolaou (2009).
As it was mentioned before, the situation in which the noise sources are homogeneously distributed over the space (Figure 3.2) is rarely encountered in the real world. A deviation from this ideal situation is commonly observed when the spatial distribution of the noise sources has a limited diversity. The directivity of the
58 Chapter 3 Passive Seismic Interferometry recorded wave fields affects the quality of the EGF. According to Garnier &
Papanicolaou (2009) if the distribution is spatially localized, then the flux of the wave field energy is not isotropic, and the cross-correlation function is not symmetric (see
Figure 3.3). Another important factor that affects the quality and the symmetry of the cross-correlation function is the temporal/seasonal variations of the distribution of the noise sources. Finally, in some situations it may be difficult also to distinguish the coherent part of the cross correlation function. For instance in cases when the energy of the ambient wave field is recorded simultaneously at the seismic stations, the cross- correlation may be weak or lack energy (see Figure 3.4).
Fig. 3.3: Schematic illustration of a case in which the noise sources are spatially localized and the cross- correlation function is not symmetric. Figure is from Garnier & Papanicolaou (2009).
59 Chapter 3 Passive Seismic Interferometry
Fig. 3.4: Schematic illustration of a case in which the noise sources are spatially localized and the orientation of the station-pair is perpendicular to the main direction of the energy flux from the ambient noise sources. In this case the coherent part of the cross-correlation function can by difficult to distinguish. Figure is from Garnier & Papanicolaou (2009).
3.3 Ambient Noise Tomography using Seismic Interferometry
Surface-waves are the main extracted waves from the ambient seismic noise, because they dominate the Green's function between receivers located at the surface and also because ambient seismic noise is excited preferentially, as we already mentioned in
Paragraph 3.1, by superficial sources such as oceanic microseisms and atmospheric disturbances (Shapiro & Campillo 2005). More particularly, the emergence of either
Rayleigh or Love-waves from the ambient seismic noise depends on which components of ground motion are analyzed. If we assume two stations with three-component sensors, the cross-correlation between all components produces nine inter-component combinations, ZZ, ZR, ZT, RZ, RR, RT, TZ, TR and TT, where Z, R and T are the vertical, radial and transverse components of the noise signals, respectively. Rayleigh- waves can be retrieved from either cross-correlation of vertical components (Z-Z) or cross-correlation of radial components (R-R) and Love-waves can be retrieved from
60 Chapter 3 Passive Seismic Interferometry cross-correlation of transverse components (T-T) (Roux 2009). However, processing of
R and T components is subject to possible errors in instrument orientation (azimuth) and more attention is usually required in such cases.
The emerging surface-waves from Passive Seismic Interferometry are dispersive as expected for Rayleigh-waves inside the Earth, namely, the dependence of Rayleigh- wave velocity on period (the long periods arrive before the short periods) as well as the dependence of Rayleigh-wave depth sensitivity on period (the long periods influence deeper portion of the Earth) (Shapiro & Campillo 2004). According to Sheriff &
Geldart (1995) an important consequence of surface-wave dispersion characteristics is the existence of a group velocity. Usually, when talking about the velocity of surface- waves propagation, we commonly use the term "phase velocity" because it is the distance covered per unit time by a point of constant phase, such as a peak or a trough. But for a dispersive medium, this is not the velocity with which a pulse of energy (the emerged surface-wave in our case) travels, as it is composed of several single period signals, each one travelling with its own velocity. Figure 3.5 visualizes this concept. For the wave packet shown in Figure 3.5, the group velocity can be determined by drawing the envelope of the wave packet (the double curve ABC and
AB'C) and measuring the distance that the envelope covers in unit time. The phase velocity is determined approximately by the rate of advance of a single phase break.
For an accurate estimation of the phase velocity, the wave packet should be decomposed into its period components (e.g. Fourier analysis) before the measurement of the velocity of each component.
Since the retrieved empirical Green's function between two stations represents the seismic response as if one of the two stations acts as an impulsive source, carrying the
61 Chapter 3 Passive Seismic Interferometry
Fig. 3.5: Comparison between group and phase velocities. (a) Definition of group (U) and phase (V) velocities and (b) arrivals of dispersive surface-waves at different receivers. Figure is from Sheriff & Geldart (1995).
signature of the velocity structure between the stations, the inter-station travel-times for surface-waves on multiple paths within a seismic network can then be employed in a tomographic inversion to image the seismic velocity perturbations across the network coverage. As a result, the successes of surface-wave emergence by Passive
Seismic Interferometry were immediately followed by a revolutionary technique of noise-based surface-waves imaging, commonly called ambient noise tomography
(ANT). ANT techniques have been developed rapidly during the last decade and they have been recently applied for imaging the velocity structure at local (e.g. Brenguier et al. 2007; Mordret et al. 2015; Obermann et al. 2016, especially for imaging volcanic
62 Chapter 3 Passive Seismic Interferometry edifices), regional (e.g. Shapiro et al. 2005; Sabra et al. 2005) and continental (e.g.
Yang et al. 2008; Bensen et al. 2008) scale. Other researchers used ambient noise correlations for monitoring velocity changes related to either volcanic activity (e.g.,
Sens-Schönfelder & Wegler 2006; Brenguier et al. 2008b) or earthquake occurrences
(e.g., Brenguier et al. 2008a; Clarke et al. 2011). In this case, relative changes in the arrival time/travel time of the empirical Green's functions are computed and these variations can be interpreted as changes in the velocity structure over time
(Ratdomopurbo & Poupinet 1995). More recently, ambient seismic noise recordings have been successfully used for the first time in the Hellenic region by Evangelidis et al. (2016) to monitor relative velocity variations related to the 2014 Northern Aegean earthquake in Greece. The authors detected a clear and intense seismic velocity drop co-seismic to the major seismic event. Most of the recent studies investigating temporal changes in velocity structure using ambient noise correlations have shown that ANT could be used effectively as a monitoring tool by analyzing and processing the noise data on daily basis in a time-lapse mode, in areas with significant scientific interest such as active fault zones.
Ambient noise tomography could be particularly useful and advantageous within the context of modern seismic arrays where computing the cross-correlations between all possible pairs of stations results in very dense and azimuthally well distributed ray- path coverage. Some of the principal advantages of ANT rely on the following facts:
- Ambient seismic noise is available everywhere and at any location.
- ANT is totally independent of the spatio-temporal distribution of seismicity.
- Provides in some cases a much needed alternative to traditional earthquake tomography in aseismic areas (e.g. continental interiors).
63 Chapter 3 Passive Seismic Interferometry
- Provides complementary geophysical information by significantly improving the
lateral resolution in areas with poor ray coverage.
- ANT is a non-destructive and environment-friendly method of seismic imaging.
- Major importance also for the exploration industry is the fact that noise data are
collected without any active sources as well as at a very low cost.
64
Chapter 4
Ambient Noise Tomography of the western Corinth Rift
4.1 Introduction
The primary purpose of this study is to attempt for the first time to constrain the velocity structure of the upper crust in western Corinth Rift by means of Passive
Seismic Interferometry and its application of Ambient Noise Tomography. We analyzed Rayleigh-wave empirical Green's functions emerging from long-term observations of propagating ambient seismic noise at pairs of all available stations in the area. We used the emerged waveforms to measure Rayleigh-wave group velocity dispersion curves and to obtain 2D group velocity maps at different periods. Finally we inverted the group velocity measurements at depth attempting to assess a 3D shear- wave velocity structure of the western Corinth Rift.
4.2 Data and processing workflow
The raw data are referred to a period of three years (from January 2011 to December
2014) and include continuous seismic recordings from 22 seismic stations located at the western Corinth Rift. We used the vertical component of the recordings from both broadband and short-period seismic sensors maintained by the Seismological
Laboratory of Athens University (NKUA, http://dggsl.geol.uoa.gr/), the University of
Patras Seismological Laboratory (UPSL, http://seismo.geology.upatras.gr/) and the
Corinth Rift Laboratory (CRL, http://crlab.eu/) (Figure 4.1).
65 Chapter 4 ANT of the western Corinth Rift
Fig. 4.1: Map of the western Corinth Rift and seismic stations used in this study. Broadband seismometers are displayed with triangles and short-period seismometers are displayed with rectangles, where green, red and blue colors signify stations operated by the University of Patras Seismological Laboratory (UPSL), the Corinth Rift Laboratory (CRL) and the Seismological Laboratory of Athens University (NKUA), respectively. The inter-station path coverage of all available stations is also presented (bottom map).
66 Chapter 4 ANT of the western Corinth Rift
In Figure 4.2 we present the general steps of the data processing procedure that we followed during the ambient noise tomography of the western Corinth Rift.
Fig. 4.2: A general schematic representation of the data processing scheme.
4.3 Pre-processing & cross-correlations computation
Before the computation of the cross-correlations, a signal pre-processing was applied station by station on 1-day long traces of seismic noise (raw data) (Figure 4.3). Pre- processing of the raw data is considered as one of the most important phases in surface-wave studies based on ambient seismic noise (Bensen et al. 2007). The main purpose of the pre-processing is to accentuate the frequency band of the ambient noise
67 Chapter 4 ANT of the western Corinth Rift by attempting also to remove different kinds of irregularities that tend to obscure our ambient noise data.
Fig. 4.3: Example of a 1-day long trace of ambient seismic noise (raw data) recorded at SERG station. Black arrows mark seismic events and local perturbations.
Bensen et al. (2007) was one of the first studies that provided a basic guidance regarding the principal steps that someone needs to follow for the ambient noise data pre-processing. Below we list all the steps that we followed for the pre-processing of our data, after adopting the proposed procedure of Bensen et al. (2007) to our case:
(1) Removal of the mean and the trend of the signal
(2) Decimation to 5 Hz
Decimation to 5 Hz allowed us to form a homogeneous data set for the analysis
and also, concerning the large volume of the ambient noise recordings, we achieved
a significant reduction of the processing duration.
(3) Band-pass filtering between 0.07 - 2.5 Hz (~ 0.4 - 14 s)
A frequency band that the amplitude of the ambient seismic noise in the western
Corinth Rift is expected to be high (Evangelidis et al. 2012).
68 Chapter 4 ANT of the western Corinth Rift
(4) Removal of the instrumental response
The available seismic network of the study area (see Figure 4.1) consists of
stations with different types of instrumentation with both broad-band and short-
period sensors. The removal of the instrument response allowed us to include and
process data between all available station-pairs.
(5) Elimination of signal parts with amplitude greater than 10 times the standard deviation
In the first instance, this helps to remove the effect (peaks) of local seismic events or other glitches.
(6) Spectral whitening between 0.07 - 2.5 Hz (~ 0.4 - 14 s)
Spectral whitening, also called frequency-domain or spectral normalization, is
applied in order to broaden the band of the ambient noise signal in the upcoming
cross-correlations during the following processing step and also to eliminate
degradation caused by possible monochromatic noise sources (Bensen et al. 2007).
An example of spectral normalization (whitening) is presented in Figure 4.4.
(7) Discarding signals with amplitude greater than 3 times the standard deviation
(8) One-bit normalization
One-bit normalization is one of about five commonly used methods of time-
domain/temporal normalization. Temporal normalization is applied in order to
reduce the effects of parasitic signals and glitches (e.g. earthquakes, random local
perturbations etc.). We decided to use the most effective method, the one-bit
normalization, which keeps only the sign of the raw signal by setting all positive
amplitudes with a "1" and all negative amplitudes with a "-1". It is an effective,
69 Chapter 4 ANT of the western Corinth Rift
and quite aggressive at the same time method which is used continuously in
several ambient noise studies.
Fig. 4.4: Example of a raw (upper panel) and a spectrally whitened (bottom panel) between 0.07 and 2.5 Hz amplitude spectra for 1-day long vertical component data of SERG station.
After the completion of the previous procedure we computed daily cross-correlations of the pre-processed noise records between the vertical components of all possible station pairs (Figure 4.1b). In general, this yielded a total of about 250000 daily CCs (~1095 days × 231 station-pairs). Figure 4.5 presents a simple schematic illustration of the empirical Green's function extraction process from ambient seismic noise via cross-
70 Chapter 4 ANT of the western Corinth Rift correlation computation. In this example, after the necessary pre-processing of the vertical components of two daily recordings of noise data from AGRP and KALE stations, the cross-correlation computation resulted in the emergence of the fundamental mode of the Rayleigh-waves (grey circle).
Fig. 4.5: Simple schematic illustration of the empirical Green's function extraction process from ambient seismic noise. After the necessary pre-processing of the vertical components of two daily recordings of noise data from AGRP and KALE stations, the cross-correlation computation resulted in the emergence of the fundamental mode of the Rayleigh-waves (grey circle).
The positive lag part (positive times) of the cross-correlation is called the "causal" signal and the negative lag part (negative times) is called the "acausal" signal. The emerged waveforms in the positive and negative lags represent waves travelling in opposite directions between the stations (Bensen et al. 2007). We have already noted in Chapter 3, that if the sources of ambient noise are distributed homogeneously in azimuth around the station pair, the causal and acausal signals would be almost identical. However in the real world, as well as in our case, considerable asymmetry in
71 Chapter 4 ANT of the western Corinth Rift amplitude is observed between the two lags, which possibly indicates differences in the source process and in distance from the source in the directions radially away from the station-pair. In order to reduce these effects, we treat the two-sided daily cross- correlations in two ways. We stacked all daily CCs to desired number of days, and more specifically we stacked them over the whole recording period (3 yrs) to build a reference stack, and then we merged the two-sided reference signals into one-sided by averaging their positive and negative lag times, which is usually called the
"symmetric" signal. It has been shown that the distribution of the ambient sources becomes more random when averaged over long times (Shapiro et al. 2005). Figure 4.6 displays an example of a better and optimized emergence of the Rayleigh-waves after stacking for increasingly long time-series up to 30 days for the same station-pair
AGRP-KALE as presented in Figure 4.5. Note the asymmetry of the emerged
Rayleigh-wave energy between the time lags. Figure 4.7 presents examples of reference Rayleigh-wave vertical-vertical correlation functions stacked over the whole studied period and sorted by increasing inter-station distance.
It is noteworthy to say the following. Before stacking the daily cross-correlations, a primary quality control of the daily cross-correlations accompanied by a visual inspection of the data is strongly recommended. Initially, by stacking daily cross- correlations over a time period as much longer as possible, the construction of a reliable reference empirical Green's function can be achieved, eliminating (among others) effects of seasonal variations of the ambient seismic noise and its sources.
However this does not always lead to a desirable result. Because daily cross- correlations of bad quality included in the stacking procedure could dramatically affect the quality of the final/reference stack. For this reason before stacking we rejected daily cross-correlation functions presenting a signal-to-noise ratio (SNR) lower than a
72 Chapter 4 ANT of the western Corinth Rift
Fig. 4.6: Example of Rayleigh-wave energy emergence for increasingly long time-series. Stacking of daily cross-correlations between the vertical-vertical components of AGRP and KALE stations is presented (band-pass filtered between 0.4 and 1 Hz).
Fig. 4.7: Example of Rayleigh-wave vertical-vertical reference correlation functions (3 yrs stacks) sorted by increasing inter-station distance (band-pass filtered between 0.07 and 2.5 Hz).
73 Chapter 4 ANT of the western Corinth Rift specific value (SNR lower than 1.5 at both lags), as well as cross-correlations which were displaying clear and significant clock errors. Figures 4.8 and 4.9 present an example of a significant time-shift (error) observed between the station-pair AGRP -
KALE and an example of removing daily cross-correlations of "bad" quality from the station-pair AGEO - LAKA, respectively. Regarding the first case, we suppose that the time-shift was mainly caused by AGRP station since it appears when AGRP data are cross-correlated with data from other stations. It has been shown through the previous procedure that time-shift errors can be easily detected through a simple visual inspection of the daily cross-correlations between two stations. Detection and correction of time-shift errors, such as the observed ~8 s time-shift between AGRP -
KALE, is something that has to be taken under consideration for future work.
Fig. 4.8: Daily cross-correlations between the station pair AGRP- KALE. A characteristic example of a time-shift which was detected after the preprocessing and cross-correlations computation. The red arrows indicate the ~8 s time-shift observed between the ~680th and the ~800th day (the duration of the studied time-period is 1095 days, hence day 0 corresponds to 01/01/2012 and day 1095 to 31/12/2014).
74 Chapter 4 ANT of the western Corinth Rift
Fig. 4.9: Daily cross-correlations between the station pair AGEO - LAKA. A characteristic example of rejecting cross-correlations of bad quality (black braces) (the duration of the studied time-period is 1095 days, hence day 0 corresponds to 01/01/2012 and day 1095 to 31/12/2014).
75 Chapter 4 ANT of the western Corinth Rift
4.4 Dispersion curve measurements
The previous processing steps ended up with a total of 231 reference vertical-vertical component CCs for each inter-station ray-path. Under the ambient noise tomography formalism assumed in this study, the reference CCs of stacked signals for the whole recording period are approximately coincident with the Rayleigh-wave Green's function between the station-pairs (one station of each station-pair can be thought of as the source and the other as the receiver) and can be treaded analogously. In Figure
4.10a and 4.10b, for the AGRP - PANR inter-stations path of ∼45 km and for the
AGEO - AIOA path of ∼8 km (Figure 4.1), the ~3 yrs cross-correlation functions were filtered in different period bands. We observe the dispersive character of the reconstructed Rayleigh-waves, with the longer period waves preceding the short period waves, for both long and short inter-station distances. The dominant frequency content of the cross-correlation functions between AGRP - PANR and AGEO - AIOA range between 0.15 Hz (~6.6 s) and 0.8 Hz (~ 1.25 s), and 0.15 Hz (~6.6s) and 1.05 Hz
(~0.9 s), respectively (Figure 4.10c and 4.10d). The frequency contents and the
Rayleigh-wave dispersion presented in Figure 4.10 suggest that we can utilize the cross-correlation functions, having a sufficient vertical resolution needed to resolve the shallow velocity structure of the western Corinth Rift based on the analysis of the dispersion characteristics. It is therefore appropriate to apply a common Frequency-
Time Analysis (FTAN) method (Levshin et al. 1989) to the reference CCs and measure the group velocities of the emerged Rayleigh-waves (surface-wave dispersion between all station-pairs) similarly to the standard approach in group velocity tomography from earthquake recordings. The FTAN method is generally based on a band-pass filtering of the signals around different period bands, the calculation of the envelope function of each filtered trace and the representation of the amplitude of the envelopes
76 Chapter 4 ANT of the western Corinth Rift of each filtered trace as a 2D complex function of time and period. We can then identify (pick) at each period the maximum amplitudes of the resulting envelopes and the corresponding group velocities as the ratio of time to the already known inter- station distances of each station-pair. So, the group velocity as a function of period, namely the dispersion curves (DCs) can be measured.
Fig. 4.10: Cross-correlations between (a) AGRP and PANR stations and between (b) AGEO and AIOA stations from ~3 yrs of ambient noise filtered for different period bands and computed for the vertical components of noise records. (c) and (d) amplitude spectra of AGRP and PANR, and AGEO and AIOA cross-correlations.
Commonly, due to the large amount of data, the estimation of the group velocity dispersion curves is performed automatically. However, in order to avoid in some cases an incorrect picking of the dispersion curve in the time-frequency diagram, e.g. possibly due to the presence of high-amplitude surface-wave overtones, we used a
77 Chapter 4 ANT of the western Corinth Rift
Graphical Users Interface implemented by Mordret et al. (2014) & (2015) that allows analyst contribution to the picking of the DCs preventing possible false detections.
Figure 4.11 shows an example of dispersion curve picking on the time-frequency diagrams for vertical-vertical component correlations between stations EFP-PSAR,
AIOA-TRAZ and SERG-TEME. The relative maxima of the diagrams corresponding to the fundamental mode are interpolated and smoothed by a fifth-order polynomial.
Fig. 4.11: Three characteristic examples of dispersion curve picking on the frequency-time (period- velocity) diagrams for vertical-vertical component correlation between the station-pairs (a) EFP - PSAR, (b) AIOA - TRAZ and (c) SERG - TEME. The colored background represents the time-frequency diagram with the warm colors showing the large amplitudes. Black dots represent the relative maxima of the diagram, while the white circles highlight the automatic picking of these points. The black line is a five- order polynomial fitting to the automatic picks.
It is worth noting that the dispersion curve measurements were done at periods corresponding to inter-station distances larger than 1.5 wavelengths. So, depending on the period we wanted to study, station-pairs with inter-station distances smaller than
1.5 wavelengths were discarded. Figure 4.12 presents every Rayleigh-wave dispersion curve that was measured and used for this study plotted on a frequency-time diagram.
The average dispersion curve with its standard deviation is also shown (top panel).
The middle panel shows a diagram of the number of measurements as a function of the period while the lower one presents a probability density function plot of the
78 Chapter 4 ANT of the western Corinth Rift dispersion curves. Note the two regions of high probability density (red color) around 2 s and 5 s period and the region of intermediate to high probability density (green color) reflecting a quite sufficient data availability across ~1 to ~6 s period.
Fig. 4.12: Every Rayleigh-wave dispersion curve measured in this study is plotted on a frequency-time diagram (a). The average dispersion curve with its standard deviation is also shown as thick black line. A diagram showing the number of measurements as a function of the period (b) and (c) a probability density function plot of the dispersion curves.
4.5 2-D Rayleigh-wave group velocity maps
For the tomographic inversions of the group velocities (dispersion curves) we followed the method for surface-wave tomography of Barmin et al. (2001) and more specifically we followed a Cartesian version of this method which was implemented by Mordret et
79 Chapter 4 ANT of the western Corinth Rift al. (2013). The ray-theoretic method of Barmin et al. (2001) is mainly based on minimizing a regularization function composed of a Gaussian-shaped lateral smoothing function and a constraint on the amplitude of the perturbation weighted by local ray-path density.
We adopted the approaches of Barmin et al. (2001) and Mordret et al. (2013) on our data and below we describe in more detail the method we followed for the inversions.
Using ray theory, the forward problem for Rayleigh-wave tomography consists of estimating a frequency-dependent travel-time t of Rayleigh-waves from a set of 2D group velocity maps. So, a group travel time t along a ray-path p can be computed as
(4.1)
where s is the distance along the ray-path and U is the group velocity. Since Rayleigh- wave travel-times are inversely related to velocities, we treat eq. 4.1 as follows. The travel-time perturbation δt relative to a reference group velocity distribution is expressed by
(4.2)
Defining a model , δt becomes a linear function of m. For the ith ray-path we
have:
(4.3)
We assume that the observed travel-times tobs are a sum of real travel-time t and an
obs observational error ε (t = t + ε) and define a datum where is the travel-time for the reference model U0. This results in the following:
(4.4)
80 Chapter 4 ANT of the western Corinth Rift
To estimate the model m we minimize the following penalty function:
(4.5) where G is the forward operator, d is the data vector and C is the covariance matrix of the data. Specifically, the first term of the penalty function (eq. 4.5) represents the deviation of the model from the data, the second term is the spatial smoothing condition and the third term is the damping constraint that penalizes the weighted norm of the model. For an arbitrary function f(r) the norm can be defined as:
(4.6)
The spatial smoothing condition F involves a correlation length σ:
(4.7)
where K is a smoothing kernel defined as follows:
(4.8)
normalized such as .
The third and final term of the penalty function constrains the amplitude of the perturbations depending on local ray-path density:
(4.9) where ρ is the ray-path density around x and λ is a user-defined constant. α, β, σ and λ are all user-defined parameters that were determined through systematic testing of the misfit evolution. The velocity distribution is discretized with a Cartesian geographical grid where each cell has a constant velocity. Let N be the number of data
(ray-paths), i = 1,...,N, and M be the number of cells in the model, j, k = 1,...,M, then
81 Chapter 4 ANT of the western Corinth Rift m(x) = m is a M-long vector containing the slowness for every cell in the grid and eq.
4.5 can be (re)written in matrix form as follows:
(4.10) where Q is the regularization matrix defined as
(4.11)
The matrix G is a matrix containing the length of every ray-path in every cell of the model:
(4.12)
th th where is the length of the i ray-path in the j cell of the model and U0 is the initial group velocity in the jth cell. The matrices F and H are matrices and their components are defined by:
(4.13)
where is the Kronecker symbol and with being the distance in the jth and kth cells.
(4.14)
where ρj is the number of ray-paths crossing the jth cell.
With these definitions, the minimum mmin of the function S is:
(4.15) and the inversion operator G*, is defined as follows:
82 Chapter 4 ANT of the western Corinth Rift
(4.16)
Finally, we define the resolution matrix as:
(4.17)
Taking into account the data availability (available measurements per period, Figure
4.12b) we have tested different sizes of grid spacing to achieve the best compromise between models parameterization, spatial resolution and a reliable representation of the velocity features. After these tests we concluded to the following inversion grid step. We used grids with two different grid spacing for the 2D group velocity models.
For periods T < 2 s and 4 > 5 s we used a 17 × 10 grid with a cell size of 0.04o × 0.04o
(~4.1 km EW and ~4.4 km NS) while for periods 2 s ≤ T ≤ 4 we used a denser 22 × 13 grid with a cell size of 0.03o × 0.03o (~3.1 km EW and ~3.3 km NS). We inverted the
Rayleigh-wave group velocities at 26 periods between 1 and 6 s with a step of 0.2 s.
Following Moschetti et al. (2007) and Mordret et al. (2015), we performed the inversion in two main steps. The initial model for the inversion is characterized by a constant velocity which is taken as the mean group velocity for each period. In the first step, we inverted a very smooth map that was used to detect and exclude measurement outliers. We discarded measurements having travel-time residuals greater than two times the standard deviation. In the second step, we used the remaining measurements to build the final group velocity maps for each period. Figure
4.13 presents the resulted ray-path coverage at six periods between 1 and 6 s, in which ray-paths are colored according to the measured group velocities. The geographical grid is also presented for each period. Detailed maps showing the ray-path coverage at all studied periods with the step of 0.2 s are presented in Appendix A.
83 Chapter 4 ANT of the western Corinth Rift
Fig. 4.13: Rayleigh-wave group velocity measurements and ray-path coverage at 1 s, 2 s, 3 s, 4 s, 5 s and 6 s, respectively. The thick red line shows the limit of the study area connecting cells of the geographical grid with no measurements. Seismic stations are shown as black triangles.
84 Chapter 4 ANT of the western Corinth Rift
During the inversion procedure the topography and the bathymetry of the rift were not taken into account. One could argue that the topography and the bathymetry of the study area could affect the accuracy of the measurements, especially the velocity measurements of short-period Rayleigh-waves. For this reason we estimated the surface-wave velocity errors as the relative difference in distance between the stations with or without taking into account the relief of the area. The errors are thus the relative difference between the length of the horizontal line that connects each station- pair and the length of the line that follows the shape of the relief (including elevation and bathymetry) between the stations. We found errors < 1% (see Figure 4.14) on group velocity measurements, with more than 90% of the inter-station ray-paths having errors < 0.4%. Considering the observed variation of the group velocity measurements per period (Figure 4.12a), these low error values caused by the flat topography approximation should be definitely considered negligible.
. Fig. 4.14: Velocity error introduced by excluding the topography and the bathymetry of the study area. The black curve shows the cumulative number of ray-paths with errors smaller than a certain threshold. 100% of the inter-station paths have error smaller than 1%, while the 90% of them has errors smaller than 0.5%.
85 Chapter 4 ANT of the western Corinth Rift
Applying the ray-theoretic formulation described above we managed to derive group velocity maps of the study area. Maps of final 2D Rayleigh-wave group velocity models of the western Corinth Rift at six different periods from 1 to 6 s are presented in
Figure 4.16. The final mean velocity and the variance reduction between data computed from a homogeneous model with an average velocity and the final model are shown at the lower left corner of each map. The mean velocity increases with increasing period from ~1.4 km/s at 1 s to ~2.2 km/s at 6s, as the penetration depth of
Rayleigh-waves is considered to increase with increasing periods. The mean velocity seems also to be in good agreement with the average of the dispersion curves shown in
Figure 4.12a. The variance reduction of the travel-time residual generally display values higher than 50%, with an exception of the variance reduction around 1 s period, indicating that the derived velocity maps fit the data relatively well (see Figure 4.15).
2D Rayleigh-wave group velocity maps of the study area at all studied periods with a step of 0.2 s are presented in Appendix A. Figure 4.17 shows the path density (number of ray-paths per cell of the geographical grid) in a map view for the corresponding periods. Path density decreases in general with increasing period, with the exception of the path density at 1 s period.
Fig.4.15: The variance reduction of the travel-time residuals between 1 and 6 s with 0.02 s step.
86 Chapter 4 ANT of the western Corinth Rift
Fig. 4.16: Rayleigh-wave group velocity maps at 1 s, 2 s, 3 s, 4 s, 5 s and 6 s. Seismic stations are shown as black triangles. For each period, the variance reduction (VarRed) between data computed from an initial homogeneous model with an average velocity and the final model, as well as the final mean velocity (Vmean) are shown at the lower left corner of each frame. Major fault traces shown are: 1=East Eliki, 2=West Eliki, 3=Psathopyrgos, 4=Marathias,
5=Trizonia, 6=Aigion, 7=Kamarai/Selianitika. The respective fault numbers are shown in 4.16a. The thick grey lines (4.16a) denote examples of extracted 2D profiles (S1, S2 and S3) at depth shown in Figure 4.22.
87 Chapter 4 ANT of the western Corinth Rift
Fig. 4.17: The ray-path density at 1 s, 2 s, 3 s, 4 s, 5 s and 6 s. Seismic stations are shown as blue triangles.
88 Chapter 4 ANT of the western Corinth Rift
4.6 Resolution Assessment
Concerning the resolution assessment of the group velocity maps, note that from eq.
4.17, following Barmin et al. (2001) and Mordret et al. (2013), (2015), we have already evaluated the resolution matrix (R) associated with each inversion. Each row of the resolution matrix is a resolution map defining the resolution at each cell of the model, thus in general the resolution matrix is quite large and the information it includes is somehow challenging to utilize. More specifically, each row of the resolution matrix
(resolution map) represents approximately the response of the tomographic procedure to a Delta-function type perturbation/anomaly, in a specified cell of the grid. In order to estimate the spatial resolution we fitted an ellipse to each resolution map that encloses 50% of the recovered amplitude of the perturbation and took its major axis as estimation for the spatial resolution. Figure 4.18 shows plots of the spatial resolution maps for the respective periods. The spatial resolution is limited to 8 km. Similarly with the previous images, spatial resolution maps of the study area at all studied periods with the step of 0.2 s are presented in Appendix A.
4.7 Inversion of local dispersion curves Depth inversion
All retrieved Rayleigh-wave group velocity maps can be considered as a set of local group velocity dispersion curves. Actually, for each cell of the geographical grid one can gather group velocity measurements at different periods forming with this way local dispersion curves. The inversion of a local DC can lead to a local 1D shear- velocity model and consequently the combination of all individual 1D models from all cells result in the 3D shear-velocity structure of the sub-surface. For the needs of our study, we gathered the group velocities with the already used step of 0.02 s period in
89 Chapter 4 ANT of the western Corinth Rift
Fig. 4.18: Spatial resolution maps at 1 s, 2 s, 3 s, 4 s, 5 s and 6 s. Seismic stations are shown as red triangles.
90 Chapter 4 ANT of the western Corinth Rift each model cell, using a grid of 30 × 17 cells with 0.02o × 0.02o cell size and we constructed local dispersion curves.
For the inversion at depth we adopted the Neighborhood Algorithm (NA), an optimized
Monte-Carlo global search technique developed by Sambridge (1999a, b) which has been efficiently used in various geophysics inversions. This search algorithm makes use of geometrical constructs known as Voronoi cells to derive the search path in a model-space. A model is a set of different parameters and the corresponding model- space is a multi-dimensional space having the same dimensions as the number of the parameters used to characterize the model. The model-space is specified by a priori range of values for each parameter. In the case of our local DC inversion problem, the model is a 1D layered shear-wave velocity profile with two parameters for each layer, the thickness and the shear-wave velocity.
The inversion at depth was performed by adopting a Neighborhood Algorithm code implemented and provided by A. Mordret, MIT, 2016. In a first step, the Neighborhood
Algorithm samples na models (1D layered shear-velocity profiles in our case) uniformly distributed in the model-space generating Voronoi cells (nearest neighbour regions associated with each model) with a locally uniform density. Then for each model, a theoretical DC is calculated (Herrmann & Ammon 2004) and the misfit between the theoretical and the observed local DC is computed and assigned to the corresponding
Voronoi cell (Mordret et al. 2014). The misfit between the observed local DC and the theoretical one is defined as the difference of the theoretical dispersion curve with the observed local DC, taking into account with its uncertainties and normalized by the area of the measured dispersion curve (see Figure 4.19 for a graphical explanation). In a second step, a number of nc cells with the lowest misfit are chosen and nb new
91 Chapter 4 ANT of the western Corinth Rift models are generated within each of the previous cells. Then, a new mesh of Voronoi cells are created based on every model including both previous and the new ones. The new misfits are calculated and the best nc models are selected to be re-sampled. The procedure is repeated nd times, where nd is the number of iterations. The main advantage of the Neighborhood Algorithm is that it is able to reduce the misfit efficiently by exhibiting complex self-adaptive behaviour in searching the model-space and concentrating sampling simultaneously on the most promising local regions of interest (Voronoi cells).
Fig. 4.19: Schematic illustration of the misfit calculation between a theoretical dispersion curve (thick black line) and a measured dispersion curve (thin gray line) with its uncertainties. The misfit value is the normalized area (dS1 + dS2) / S. Figure from Mordret et al. (2014).
Before inverting each localized DC for the 3D structure, we calculated and inverted the average DC of the local dispersion curves in order to assess an average 1D shear- velocity model of the whole study area. As we previously mentioned, in the case of a local DC inversion problem, the model is a 1D layered velocity profile with two parameters, the thickness and the shear-wave velocity. So one has to invert for 2n parameters, where n is the number of layers. The disadvantage of this parameterization is based on the fact that if n is chosen too large, the model-space becomes huge and it is extremely time-consuming to sample the volume of the model-
92 Chapter 4 ANT of the western Corinth Rift space densely. Furthermore, testing with different number of parameters showed that as the number of parameters grows, the non-singularity of the final solution becomes more and more critical and choosing the right model which has also a physical meaning during the sampling of the model-space is difficult. In order to avoid over- parameterization, for the average 1D shear-velocity profile of the whole study area, we parameterized the model with four layers over a half space with seven unknowns, the velocities in the four layers as well as the depth of the three associated interfaces.
Results of the inversion of the average local DC are presented in Figure 4.20.
Fig. 4.20: Inverted shear-wave velocity model for the whole study area. (a) Synthetic dispersion curves overlaid by the average local dispersion curve with error bars. The thick black line is the dispersion curve with the minimum misfit value and (b) the associated inverted models. The thick black line is the model with the minimum misfit value while the well and poorly resolved parts are shown with solid and dashed line, respectively. Both synthetic dispersion curves and models are colored according to their misfit.
The Neighborhood Algorithm run with na = 5000, nb = 500 nc = 2 and nd = 10. So, during this inversion 15 000 (na + nb × nc × nd) models have been sampled. In the last step, we calculated the best-fitting (lowest misfit) 1D depth profiles for all grid cells, resulting in the 3D shear-velocity structure (3D cube) of the study area. Then, from
93 Chapter 4 ANT of the western Corinth Rift the resulted 3D structure, we were able to extract 1D profiles as well as 2D vertical and horizontal sections (profiles and slices) at any desired location or depth, respectively.
Fig. 4.21: (a) An example of the 2D marginals of all pairs of parameters derived from the 1D inversion at a random location (cell of the grid) of the study area. The model was parameterized with four layers over a half space with seven unknown parameters, the velocities in the four layers (V1, V2, V3 and V4 (m/s)) as well as the depth of the three associated interfaces (D1, D2 and D3 (m)). The images depict the sampling procedure of the Neighborhood Algorithm across each 2D parameter-space of the whole model- space and the gradual sampling towards the most promising regions with the lowest misfit. The colors correspond to the misfit value (m/s). Note the relative difficulty of the Neighborhood Algorithm to converge the sampling within the 2D parameter-spaces between the depth and the velocities of the superficial layers. The lower panel (b) shows the evolution of the misfit during the inversion. After eleven iterations the misfit achieves the lowest values and does not improve any more. The colors correspond to the number of models for each iteration.
94 Chapter 4 ANT of the western Corinth Rift
As mentioned previously, the local inversion at depth for all grid cells resulted in the construction of a 3D shear-velocity model. The construction of the 3D matrix (cube) enables us to extract and plot any desired row(s) or column(s) of it. In the following two figures we present vertical 2D sections, an East-West (S1) and two North-South
(S2 and S3), cutting through the study area (Figure 4.22) and horizontal slices at different depths (800, 2000, 3900 and 5100 m) (Figure 4.23). Taking into account the retrieved Rayleigh-wave group velocity maps (see Figure 4.16) we decided to extract cross-sections cutting through the most interesting low velocity features primarily within the southern part of the rift. The locations of the cross-sections are sketched with thick grey lines in Figure 4.16a.
Fig. 4.22: Vertical cross-sections along the grey lines sketched in Figure 4.16a according to the results of the depth inversion. The thick dashed black lines show the intersection between the profiles. Major faults (black lines) are presented and numbered as in Figure 4.16a. Dashed rectangles indicate regions of lower velocities. The faults geometry approximation is taken from Bernard et al. (2006).
95 Chapter 4 ANT of the western Corinth Rift
Fig. 4.23: Slices through the 3D shear-velocity models at (a) 800m, (b) 2000m, (c) 3900m and (d) 5100m below the sea level. Major fault traces shown are: 1=East Eliki, 2=West Eliki, 3=Psathopyrgos, 4=Marathias, 5=Trizonia, 6=Aigion, 7=Kamarai/Selianitika.
We did not account for the western Corinth Rift relief (topography/bathymetry of
±350m) in our analysis. Besides the previously mentioned negligible influence on the group velocities (Figure 4.14), even for the lowest studied periods of around 1 s, the wavelength is estimated to be longer than the elevation or the bathymetry of the western Corinth Rift (λ ). Hence we consider that the Rayleigh waves at periods ≥ 1 s are affected negligibly by the flat topography approximation and the retrieved velocity perturbations are reliable.
The essential target of the current chapter was to describe the processing steps (pre- processing, noise cross-correlation, dispersion curve measurements, tomographic
96 Chapter 4 ANT of the western Corinth Rift inversion, inversion at depth) and the methodologies we implemented in order to perform a noise-based tomography of the western Corinth Rift. Results and images presented in this chapter will be discussed and interpreted in the following chapter.
97
Chapter 5
Interpretation & Discussion
5.1 Introduction
In the previous chapter we made use of ambient seismic noise recorded by the available seismological stations at the western Corinth Rift to reconstruct Rayleigh- wave empirical Green's functions between each pair of seismic sensors. We utilized these waveforms to perform an ambient noise tomography and then we inverted local dispersion curves extracted from Rayleigh-wave group velocity maps in order to constrain the 3D shear-wave velocity structure of the area up to ~6 km depth.
Taking into account the distribution of the available seismic network and the resulted ray-path coverage of the study area, the spatial resolution of the group velocity maps appeared to be poor towards the eastern and western edges where only a small number of ray-paths or some isolated rays are available. The most characteristic case is the region between the Rion straits and the Mornos delta at the western edge of the study area (see Figures 4.13, 4.16 - 4.18). In some cases as well, the irregular distribution of available ray-paths caused minor smearing effects/artifacts in the mapped group velocity perturbation such as the observed velocity features at the two corners of the easternmost edge. Note, for instance, the smearing effect with the elongated velocity feature in the 2 s group velocity map caused by the easternmost station in the northern coast (Figure 4.16b). It is important to keep in mind the above in the context of the following discussion of the results. Discussion and interpretation concerns only well recovered parts of the velocity models as given by the spatial resolution estimation.
99 Chapter 5 Interpretation & Discussion
5.2 Rayleigh-wave group velocity maps
Considering the resolution estimation at 1 s period (Figure 4.18a), a reliable interpretation of the Rayleigh-wave group velocity map (Figure 4.16a) can be made only for the eastern part of the study area, namely around Aigion, Trizonia and the region between them. The 2 s Rayleigh-wave group velocity map in Figure 4.16b highlights two distinct low velocity zones. A low velocity zone around Aigion area in the southeast (which is also highlighted in the well recovered part of the 1 s period map, Figure 4.16a), and a second one around Psathopyrgos area in the southwest which extends up to Mornos delta to the north. Higher velocities are imaged across the rest of the area, but mostly across the northern part of the rift. The lateral expansion of the lower and higher velocity zones at periods around 1 and 2 s is coincident with the geology context of the area (see the generalized tectonic map in Figure 2.2,
Chapter 2). The observed low velocities are due to the presence of Plio-Quaternary syn-rift sediments across the south coast, while higher velocities are due to the pre-rift basement structures which are outcropping almost everywhere to the north. The observed lateral expansion of the low velocity zones towards the off-shore region between Mornos delta and Psathopyrgos, as well as towards the off-shore area northern of Aigion are possibly due to the sediments of deltaic sequences which are accumulated in the basin. Sedimentation in the Corinth Rift (Late Quaternary) is mostly controlled by the sediment input from the surrounding rivers (Perissoratis et al. 2000). The largest rivers that are draining into the rift are the aforementioned
Mornos in the northwest and a series of rivers around the Aigion area (Vouraikos,
Kerynitis, Meganitis, Selinountas, Erineos) that enter the southern rift draining the northern Peloponnese. Deltaic sequences have been also documented at the western margin of the rift, between Mornos delta and Psathopyrgos (Perissoratis et al. 2000).
100 Chapter 5 Interpretation & Discussion
The 3 s Rayleigh-wave group velocity map (Figure 4.16c) shows quite similar features as the 2 s map. In this case, the previously observed two low velocity anomalies show an incipient tend to merge into a single WNW-ESE trending low velocity anomaly, extending across the south part of the study area from the Aigion area to the Mornos delta. The overall distribution of the low velocities at 3 s still demonstrate good agreement with the major geological features of the area, reflecting also relatively well the asymmetric geometry of the rift. As we expected, higher velocities at 3 s period are observed across the northern coast of the rift, where pre-rift basement structures are present.
At periods above 3 s, where group velocity measurements are sensitive to deeper structures of the crust, the dispersion maps in Figures 4.16d to 4.16f highlight an elongated low velocity zone trending across the southern part of the rift. This WNW- trending low velocity anomaly is also the most prominent feature in the velocity maps between 4 and 6 s Figures 4.16e and 4.16f). At a rough estimate, the characteristic depth of penetration of the Rayleigh-waves is wavelength-dependent. Assuming that the depth of penetration of Rayleigh-waves ranges approximately between one-third and one-half of their wavelengths (Lowrie 2007), and considering also the estimated mean velocities of the emerged Rayleigh-waves (shown at the lower left corner of each frame of Figure 4.16), Rayleigh-waves of 4 s period for instance, with a mean velocity of ~2 km/s, would be sensitive down to ~2.7 - 4 km. In general, the longer the period the deeper the Rayleigh-wave energy penetrates. In Chapter 2, it is referred that stratigraphic interpretation of the on-shore and off-shore data (sedimentary and lithostratigraphic logs, 2D seismic reflection profiles etc.) (Sachpazi et al. 2003;
Lykousis et al. 2007; Bell et al. 2008, 2009; Taylor et al. 2011 and others) which have revealed that the maximum thickness of the syn-rift sediments accumulated above the
101 Chapter 5 Interpretation & Discussion basement in the Corinth Rift is estimated to be about 2.5 km in the central rift's basin.
The sediment thickness varies considerably along the ~E-W rift axis, and specifically, it has been indicated that sediments thicken from the west towards the east part of the basin (Sachpazi et al. 2003). Therefore, the thickness of the accumulated syn-rift sediments within our study area is expected to be quite smaller than the maximum reported one from the literature (~2.5 km). In Figure 5.1 we present ~N-S cross- sections cutting through the easternmost part of the study area (from Hemelsdael &
Ford 2014, seismic lines are shown in Figure 2.3, Chapter 2). We observe that the thickness of the on-shore sediments varies from some tens of meters to a maximum of
~1 km, while the off-shore sediments present a maximum thickness up to ~1.5 km.
Fig. 5.1: On-shore and off-shore cross-section presented from west (a) to east (b) which cut through the easternmost edge of the study area in a ~N-S direction. Colors are those used in Figures 2.4 and 2.5 in Chapter 2. EHF: East Heliki Fault; PMF: Pirgaki-Mamoussia Fault; DF: Derveni Fault and AF: Aigion Fault. See Figure 2.3 in Chapter 2, for the exact location of the presented faults and profiles. Figure modified from Hemelsdael & Ford (2014).
102 Chapter 5 Interpretation & Discussion
Consequently, Rayleigh-waves at periods above 3 s which are characterized by a penetration depth of about 2.7 - 4 km begin to sense the faster shear-wave velocities in the pre-rift carbonates within the underlying basement, and the slow shear-wave velocities of the sediment layers are comprised by the higher velocities below.
Following the previous assumption, the seismic velocity features in the group velocity maps at periods above 3 s (Figures 4.16d, e & f) most likely reflect the characteristics of the underlying basement.
Comparing the orientation of the obtained low velocity anomaly with the orientation of the active normal faults at the southern coast, as well as with the orientation of the overall rift axis, a striking similarity is observed between them. Taking account the group velocity maps across all the studied periods from 1 to 6 s, we observe that the lower velocities around Aigion in the eastern part of the study area are bounded by the
Trizonia fault in the north and by the west Heliki fault on the southern side. In the western part, the low velocity patterns are spatially correlated mostly with the
Psathopyrgos fault.
5.3 Shear-wave velocity structure at depth
In order to infer the shear-wave velocity structure of the study area, we inverted the extracted Rayleigh-wave dispersion curves at depth by adopting a Monte-Carlo approach of the Neighborhood Algorithm (Sambridge 1999a, b). As it is already mentioned in the previous chapter, in the case of our inversion problem, the model is a
1D layered shear-wave velocity profile with two parameters for each layer, the thickness and the shear-wave velocity. In general, according to the adopted technique the model with the best data fit is specified after a uniform pseudo-random sampling
103 Chapter 5 Interpretation & Discussion of the parameter-space. The disadvantage of this approach relies on the fact that when the dimensionality (number of parameters) of the parameter-space is increased, the basic random generation of models becomes quite inefficient. This restriction allowed us to sample the model-space with a small number of layers (four), ensuing reliable results on the one hand, and images with slightly lower resolution on the other hand.
As a future work, further effort needs to be done to reduce the number of the free parameters to be inverted. It is beyond the scope of this study, but the parameterization of the shear-velocity profile at depth as a power-law for instance
(e.g., Wathelet et al. 2004), is strongly suggested to be introduced in such cases. With this way, a larger number of layers with their corresponding velocities could be modeled as a power-law, achieving a significant reduction of the tuning parameters to be inverted, and the construction of a more complex and detailed shear-velocity model at the same time. In summary, an empirical tuning of the parameters that control the inversion process will ensure among others the computational efficiency and it will provide images of the sub-surface with better resolution.
Despite the fact that our resolution is not enough to detect sharp variations at depth, results and images from the depth inversion reveal the presence of some distinctive velocity patterns which, as would normally be expected, are in agreement with the results of the ambient noise tomography. According to the available local dispersion curves extracted from the Rayleigh-wave group velocities, we were able to constrain the shear-velocity structure of the area up to ~6 km depth, considering the upper 4 km of the crust as the well resolved parts.
Regarding the 2D shear-wave velocity profiles and slices at depth (see Figures 4.22 &
4.23, Chapter 4), they show as expected, the presence of low velocity regions which
104 Chapter 5 Interpretation & Discussion correlate relatively well with the low velocity regions obtained from the noise tomography maps (Figure 4.16, Chapter 4), providing better constraints on the depth distribution of the detected low velocity zones. These low velocity zones are located mostly beneath the southern part of the rift, while their lateral expansion confirms their already mentioned spatial correlation with the major fault traces of the area.
From the shear-wave velocity profiles and slices, we clearly constrain low velocity regions, not only in the shallow part of the crust, where sediments are present, but also at deeper parts. More specifically, low velocities across the E-W profile S1 (Figure
4.22a) are clearly related with the Psathopyrgos fault and the western edge of the
Kamarai/Selianitika fault zones. In the N-S profile S2 (Figure 4.22b) we distinctly observe a region of lower velocities in the vicinity of the East Eliki, Aigion and
Kamarai/Selianitika faults. A minor low velocity zone is also observed at depth close to the Trizonia fault. Profile S3 (Figure 4.22c) presents similar velocity patterns in the N-
S direction around the Psathopyrgos fault as in S1. Slices through the well-recovered parts of the 3D shear-velocity model at different depths provide a better picture about the distribution of the observed low velocity zones as depth increases. The horizontal sections (Figure 4.23) clearly illustrate the WNW-ESE elongated low velocity zone across the southern margins of the rift. Notably, this zone is clearly observed not only beneath the Corinth Rift's basin, but also beneath the northern Peloponnese, even at depths greater than ~5 km within the deeper part of the crust, much deeper than the estimated total thickness of the sediments. Note for instance, the strong lateral variation of shear-velocities at 5100 m depth (Figure 4.23d), where the lowest shear- velocity values are detected beneath the Aigion area.
It is worth noting at this point that the derived shear-velocities from the current study by means of ambient-noise interferometry appear to be in general under-estimated in
105 Chapter 5 Interpretation & Discussion
relation with previous VS models proposed for the same region, and particularly those derived from local travel-time measurements. Such crustal models are the corresponding models proposed by Rigo et al. (1996) and Latorre et al. (2004) for the western Corinth Rift. On the contrary, the observed VS model from the current study displays higher values in the uppermost layers of the crust compared with that proposed by Novotny et al. (2000) inferred from surface-waves, but at depths greater than ~3 km, our model present lower VS values. According to the available local dispersion curves extracted from the Rayleigh-wave group velocities, we were able to constrain the shear-velocity structure of the area up to ~6 km depth. In Figure 5.2 we present the average 1D Vs model derived from the present study, along with the aforementioned crustal models of Rigo et al. (1996), Latorre et al. (2004) and Novotny et al. (2000) up to the aforementioned depth.
Fig. 5.2: Superposition of the average 1D Vs model derived from the present study and the three crustal models proposed for the western Corinth Rift by Latorre et al. (2004), Rigo et al. (1996) and Novotny et al. (2000).
106 Chapter 5 Interpretation & Discussion
Among other possible factors (e.g., different methodologies), the lower shear-velocities of our model compared with those of Rigo et al. (1996) and Latorre et al. (2004) are most probably due the high sensitivity of the Rayleigh-wave propagation velocity, especially of the low period Rayleigh-waves, to the very shallow structures of sediment depositions in the study area. This is also supported by the lower VS compared to that of Novotny et al. (2000) whose model derived by joint inversion of both Love and
Rayleigh-waves. Considering the fact that the 1D models of the Corinth Rift derived from measurements of first-arrival times are usually suffer from a limited resolution in the shallower crust between depths of 0 and 4 km (Latorre et al. 2004), Ambient
Noise Tomography possibly contributes to a more reliable understanding of the shallow structure of the crust by providing a crustal velocity model at a resolution higher than those attainable by travel-time tomography alone.
5.4 Overall interpretation
Since no other ambient noise tomography studies of our target area are available in the literature, we compare are results with those obtained by the tomographic studies of Gautier et al. (2006) and Latorre et al. (2004). In both studies a three-dimensional travel-time tomography of the western Corinth Rift was performed by re-analyzing data sets of passive seismological experiments. The outcome of the previous works was among others, the construction of detailed shear-wave velocity images down to ~11 km of the western Corinth Rift. In Figure 5.3 we show map views up to 5 km depth of the tomographic models modified from Gautier et al. (2006) and Latorre et al. (2004).
Large-scale features of our models are similar to those obtained by Latorre et al.
(2004) and Gautier et al. (2006). Gautier et al. (2006) and Latorre et al. (2004) detected a low velocity region which was found to be parallel to the trend of the rift.
107 Chapter 5 Interpretation & Discussion
Fig. 5.3: Tomographic models of shear-wave velocities in the western Corinth Rift derived by (a) Latorre et al. (2004) and (b) Gautier et al. (2006). Map views show velocity layers between 0 and 5 km depth. Stations are plotted as triangles and major faults are drawn in dark lines (Psa: Psathopyrgos fault; Ai: Aigion fault; He: Heliki fault and Py-Ma: Pyrgaki-Mamoussia fault). Earthquakes are displayed as dark dots. Parts of the models not crossed by rays have been masked. Well-resolved areas are outlined with a white contour. Figure modified from Latorre et al. (2004) and Gautier et al. (2006).
The detected low velocity anomaly from the previous studies shows a similar orientation and lateral evolution with increasing depth, being in an agreement with the observations from our ambient noise tomography models (Figure 4.16). On the southern coast, the lower velocities are spatially correlated with the major fault traces, such as the Heliki, Aigion, Kamarai/Selianitika and Psathopyrgos faults, an observation which is also supported by the previous studies. In the first instance, following Gautier et al. (2006) and Latorre et al. (2004), we suggest that the distribution of the lower velocities is possibly controlled by both the presence of the pre-rift deposits and the present-day active tectonic regime of the area. However, since the pre-rift sediments around the study area have thickness varying from a few hundreds of meters to about 2.5 km northwards into the basin (Hamelsdaël & Ford
108 Chapter 5 Interpretation & Discussion
2014), this is not sufficient enough to interpret the low velocity images at periods above 3 s affecting deeper parts of the crust, bellow the sediments.
A possible complementary explanation for the presence of the distinct low velocity zone at the south could possibly reflect a highly fractured upper crust as well as the involvement of fluid circulation processes at these parts of the crust. So the obvious question which arises from this hypothesis "are there any evidence of fluid circulation processes within the crust of the western Corinth Rift?".
The observatory of CRL (http://crlab.eu/) has been developed in the western Corinth
Rift since 1999 and one of its main objectives is to investigate the mechanics of the active faults in the western Corinth Rift with a special emphasis on the role of fluids
(e.g. Cornet et al. 2004; Bernard et al. 2006). Within the framework of the observatory of CRL several multidisciplinary studies have been performed towards this purpose.
For instance, the study concerning the 1995 Aigion earthquake and its aftershocks
(Bernard et al. 1997) has pointed out that either over-pressured fluids or stress rotations coming from crustal heterogeneities might have played a role in triggering the Aigion earthquake as well as other moderate earthquakes in the Corinth Rift.
Bourouis & Cornet (2009) indicated possible overpressure fluid diffusion processes in the western Corinth Rift, related to the seismically activated parts of the crust. The previous authors, utilizing also the hydraulic data from a deep well that intersected the Aigion Fault (Cornet et al. 2004), suggested that the overpressure conditions within the normal fault system of the southern coast are possibly due to the fact that the faults act as hydraulic barriers in the direction perpendicular to their strike, preventing fluid flows. In Figure 5.4 we present a schematic structural cross-section through the Aigion fault based, among others, on cuttings and core analysis from the
109 Chapter 5 Interpretation & Discussion drilling operation of the 1000 m-deep AIG10 borehole (Cornet et al. 2004, the AIG10 well observatory). In the case of the Aigion fault, it has been shown that the fault core is made of a thick impervious clay zone surrounded by permeable cataclastic zones.
Fig. 5.4: Schematic structural cross-section through Aigion fault, in the Aigion harbor. On the left, depths have been estimated from seismic refraction data (Naville et al. 2004) while on the right depths correspond to cuttings and core analysis (Lemeille et al. 2004; Rettenmaier et al. 2004). Velocities indicated on the right part of the figure correspond to P-wave arrivals from sonic logs, while on the left part velocities are those estimated from seismic refracted data. The pore pressure values correspond to measurements obtained during the drilling phase for the platy limestone and after drilling operation for the limestone encountered below the fault.
Existence of fluids in the cataclastic zones of faults in the western Corinth Rift has been also documented from geochemical analysis of fault gouges (Koukouvelas &
Papoulis 2009; Pick & Marty 2008; Baus et al. 2004). Other studies that have indicated the presence of fluids in the area close to the observed low velocity zone and their key role in the development of seismicity are among others the already
110 Chapter 5 Interpretation & Discussion
mentioned study of Latorre et al. (2004), according to which high VP/VS ratios were recognized between 7 and 9 km depth under the rift indicating fluid saturated rocks and Pacchiani & Lyon-Caen (2010) suggesting a fluid-driven spatio-temporal evolution of seismicity, studying the case of the 2001 Agios Ioannis earthquake swarm (located few km south of Aigion). The results of the ambient noise tomography maps at periods above 3 s, and those of the 3D shear-velocity model up to ~6 km depth depicting the presence of a distinct low velocity anomaly even at depths greater than
~5 km (Figure 4.23, Chapter 4) below the southern part of the Corinth Rift, could confirm to some extent the observed evidence of over-pressured fluids and fluid circulation processes at depth from the aforementioned literature. Taking into account the sedimentary structure of the broader study area as we have already reported it based on stratigraphic interpretation of seismic data, and more specifically considering the fact that the 1000 m-deep well of the AG10 drilling project (Figure 5.4) intersects the Aigion fault within the limestones of the Hellenide nappes at ~760 m depth (Cornet et al. 2004), it is strongly validated that the proposed evidence of fluid circulation processes within a highly fracture medium are located deep into the pre- rift basement and that are not affected by the local geological context. Such characteristic case is the observed low velocity zone around Aigion area (Figures 4.16 and 4.23). Moreover, since the distribution of the low velocity zones follows a WNW-
ESE direction being parallel/sub-parallel to the strike of the neighboring major fault traces, the relatively more permeable and highly fractured parts of the crust are strongly related with the major fault zones along the southern coast of the rift. In general, our findings strongly support the hypothesis of fluid interactions and their significant role to the overall evolution of the rift proposed by previous research studies in the literature.
111 Chapter 5 Interpretation & Discussion
5.5 Conclusions
Results presented in this study can be summarized as follows:
1. The Rayleigh-wave group velocity variations at periods up to 3 s are mostly due to
the lithological variation of the western Corinth Rift.
2. Syn-rift sediments within the rift basin and across the southern coast are clearly
presented as lower Rayleigh-wave group velocity features than the limestone
nappes which are outcropping mostly to the north.
3. In general, the ambient noise tomography maps reveal lateral variations which
clearly reflect an asymmetric velocity structure of the western Corinth Rift with the
southern margins having slower velocities than the northern ones.
4. A predominant low velocity region below the southern part of the rift was observed
at periods above 3 s. The proposed interpretation of this anomaly is a combination
of the present-day active tectonic regime and the possible involvement of fluid
circulation processes at depth within a highly fractured upper crust.
5. The 3D shear-velocity model inferred from the Rayleigh-wave group velocity
measurements demonstrated that the observed low velocity zones are extended at
depths much deeper than the sediment layers (> 2.5 km) of the study area,
reflecting a highly fractured pre-rift basement structure beneath the southern
margins of the western Corinth Rift, and they are strongly related with the
neighboring major fault traces.
6. The results of the current study confirm the assumption of the important role of
fluids in the seismotectonic evolution of the western Corinth Rift being in
112 Chapter 5 Interpretation & Discussion
agreement with previous research studies.
7. Our study efficiently showed the potential to use ambient seismic noise recordings
as a complementary tool towards the investigation of the velocity structure of a
complex geotectonic environment such as the Corinth Rift.
113
PART II
Chapter 6
Shear Wave Splitting
6.1 Introduction
Shear-wave splitting (SWS), also known as seismic birefringence or shear-wave double-refraction, is a phenomenon in which S-waves split into two components with almost vertical polarizations and different propagation velocities. It occurs during shear-wave propagation through an anisotropic medium (Crampin & Chastin 2003;
Crampin & Peacock 2005) (Figure 6.1). The two parameters that can be measured through a SWS analysis are, the polarization direction φ of the fast component of the split shear-waves and the time delay δt between the two components. Measurements of φ and δt values can provide information about the anisotropy of the volume sampled along the seismic ray. SWS techniques are useful tools for estimating the seismic anisotropy of either the upper crust by performing splitting measurements on local/direct shear-waves or the lower crust and upper mantle by performing measurements on core-refracted shear-wave phases such as SK(K)S or PK(K)S.
6.2 What causes shear wave splitting
Shear-wave splitting phenomenon is detected all over the Earth and it is widely observed in almost all types of rocks regardless the geological and tectonic regime. In general, shear-wave splitting is considered as key diagnostic phenomenon for seismic anisotropy. Comprehensive literature reviews (e.g., Crampin & Lovell 1991; Crampin
117 Chapter 6 Shear Wave Splitting
& Chastin 2003; Crampin & Peacock 2005) have suggested and concluded to four possible sources of such type of seismic anisotropy:
(i) aligned crystals
(ii) aligned grains (lithologic anisotropy)
(iii) aligned fine layering (structural anisotropy)
(iv) stress-aligned microcracks (stress-induced anisotropy)
Each of these sources under certain conditions could cause SWS locally, but more than
30 years of observations (from Crampin 1981 to Crampin et al. 2015) particularly suggest that stress-aligned (possibly fluid-saturated) vertically-oriented microcracks are likely the most possible cause of relatively uniform S-wave splitting throughout at least the uppermost 20 km of the crust, oriented parallel or sub-parallel (±10o to ±20o) to the direction of the maximum horizontal stress (σH).
Various models have been proposed to interpret the shear-wave splitting in the crust.
The most widely accepted physical model which delineates and explains the principal cause of SWS with the aforementioned mechanism (stress-aligned microcracks) is the
Extensive Dilatancy Anisotropy (EDA) model that was firstly introduced by Crampin
(1984) (Figure 6.1). EDA model has since been modified and led to the Anisotropic
Poro-Elasticity (APE) model (Crampin & Zatsepin 1997; Zatsepin & Crampin 1997).
Fluid-saturated microcracks are highly compliant, and the evolution of fluid-saturated grain-boundary cracks and pores in response to changing conditions in the rock mass has been successfully modeled with APE (Figure 6.2; Crampin & Chastin 2003).
118 Chapter 6 Shear Wave Splitting
Fig. 6.1: Schematic illustration of shear-wave splitting caused by the presence of stress-aligned microcracks oriented parallel to the maximum horizontal stress, σH. The fast shear-waves are typically parallel to the direction of σH. σh and σv denote the minimum horizontal stress and the vertical stress, respectively. Figure modified after Crampin & Peacock (2008).
Fig. 6.2: Schematic illustration of the Anisotropic Poro-Elasticity (APE) model of the evolution of the geometry of stress-aligned fluid-saturated microcracks under changing conditions of increasing stress,
σH. The mechanism of deformation is fluid-movement along pressure gradients between neighboring microcracks. Figure modified after Crampin & Peacock (2008).
119 Chapter 6 Shear Wave Splitting
According to APE-model, the principal mechanism of these changing conditions is fluid movement along pressure gradients between neighboring microcracks at different orientations to the stress field. According to APE also, for low-levels of deformation, below those at which rock mass fails by fracturing, any change of stress modifies the internal geometry of microcrack distributions.
6.3 Variations in shear wave splitting parameters
Shear-wave splitting is usually controlled by quite small differences in the velocities of the two split shear components (typically less than 5% according to Crampin &
Chastin 2003). One of the main advantages of shear-wave splitting is that by rotating the horizontal component seismograms into preferred orientations of polarization, or by plotting hodograms or polarigrams, the time-delay between the split shear-waves can be measured with great accuracy. Therefore, compared with other geophysical methods, shear-wave splitting is considered as a quite efficient tool for detecting and monitoring small variations of the rock mass properties.
Changes in SWS parameters have been observed worldwide, in relation to earthquakes, reflecting changes in the anisotropic characteristics of the medium and/or the stress field (Gao & Crampin 2004; Crampin & Gao 2012). These variations and specifically variations in SWS time-delays before and after strong earthquakes are quite complicated to be interpreted and have often been the subject of debate. Such debated cases are Peacock et al. (1988), Aster et al. (1990), Crampin et al. (1990),
(1991), Seher & Main (2004), Peng & Ben-Zion (2005), Munson et al. (1995), Liu et al.
(2004, 2005) and Crampin & Gao (2005). As it was mentioned, the principal driving mechanism for these changes is fluid migration between closely-spaced microcracks at
120 Chapter 6 Shear Wave Splitting different orientations to the stress field (Crampin 1999). The parameters that control changes in the microcrack geometry also control the splitting of shear-waves, so that changes in deformation can be directly monitored by analyzing the SWS. Since SWS has proven to be an effective indicator of anisotropy (due to stress deformation processes in fluid-saturated microcrack systems), there has been an increasing interest in monitoring changes in SWS beneath volcanic edifices (e.g., Gerst et al.
2004, Unglert et al. 2011) and geothermal fields (e.g., Rial et al. 2005 and references therein).
The Corinth Rift, the Aegean volcanic arc and surrounding areas in Greece have also been the target of several shear-wave anisotropy studies due to their active complex tectonics and high rate of seismicity. For instance, shear-wave splitting analyses have been performed in local earthquakes to investigate anisotropy in the upper crust in the Corinth Rift (e.g., Bouin et al. 1996; Bernard et al. 1997; Kaviris et al. 2008, 2010;
Papadimitriou et al. 1999, 2010), in the epicentral area of the 2008 Movri Mountain earthquake in NW Peloponnese (Giannopoulos et al. 2013), across the Aegean volcanic arc such as in Milos Island (Sachpazi et al. 1991) and in Santorini Volcanic Complex
(e.g., Konstantinou et al. 2013; Kaviris et al. 2015; Papadimitriou et al. 2015). In the
Aegean region, shear-wave splitting measurements aiming to investigate seismic anisotropy in the lower crust and upper mantle have been performed on teleseismic
SKS waves by Hatzfeld et al. (2001); Kreemer et al. (2004); Schmid et al. (2004);
Evangelidis et al. (2011) and Endrun et al. (2011). All these studies played an important role in providing constraints on the anisotropy pattern in both local and regional scale as well as on the possible origins of the detected seismic anisotropy.
121 Chapter 6 Shear Wave Splitting
6.4 Shear wave window
When shear-waves reach a recording site on the surface, at low incidence angles, they have severe interactions with the free Earth's surface that almost all the similarities with the incoming waveform are irretrievably lost. So, during a SWS analysis, the analyzed seismic events are required to have shear-wave arrivals at incidence angles lower than a critical value. The critical value of the incidence angle ic is given by:
VS ic (6.1) V
where VS and VP are the S-wave and P-wave velocities near the surface (Booth &
Crampin 1985). The 3-D vertical cone of incidence which is bounded by the eq. 6.1 is commonly called shear-wave window (SWW) (Figure 6.3).
Fig.6.3: Schematic illustration of the shear wave window. The incidence angle is the angle from vertical at which the shear wave energy is arriving at the recorder. The shear wave window is the vertical cone (Volti & Crampin 2003) extending downward from the seismic station.
122 Chapter 6 Shear Wave Splitting
Assuming a normal VS V ratio of about for a Poisson's ratio of 0.25, the critical
incidence angle is estimated close to 36o. However, because of the ray-curvature due to
low velocities at the layers near the surface, the effective SWW (straight line cone) can
usually be extended to ~45o (Volti & Crampin 2003). The above restriction will give
consistent polarization measurements, which would be mainly controlled by the
structure near and below the recording site, avoiding serious distortions and
contaminations by possible S-P conversions. Consequently, data selection related to
the SWW is an absolute requirement for valid observations of shear-wave splitting.
6.5 Applications
Considering the fact that SWS parameters are controlled by changing conditions in
the geometry and the evolution of the microcrack systems, SWS analyses can be used
to monitor, model and under appropriate circumstances (it is highly debated) predict
the response of the rock mass to these changing conditions. This calculability and
predictability has numerous potential applications ranging from industrial
hydrocarbon exploration and production to a huge variety of natural hazard
monitoring, management and mitigation (Crampin & Chastin 2003). Below we list
some promising and efficient applications of the SWS techniques: i. Hydrocarbons exploration and production industry
Reservoir crack characterization (predominant microcracks orientation,
density etc.)
Correlation of hydrocarbon production with the degree of SWS
Estimating orientations of hydraulic fractures
123 Chapter 6 Shear Wave Splitting
Production optimization by correlating the orientation of the fractures with the
most productive flow ii. Stress-forecasting earthquakes and other geological hazards
Monitoring stress accumulation or relaxation processes through temporal
variations of SWS time delays (up to now, Crampin et al. 1999 is the only
successful forecast of time and magnitude of an M 5 earthquake in SW
Iceland) iii. Structural engineering - Engineering geology
Monitoring any situation where the rockmass is likely to suffer deformation or
disturbance (slope stability in mines, ground stability of buildings, tunnels,
dams; etc.)
124
Chapter 7
Shear-Wave Splitting analysis in the western Corinth Rift
7.1 Introduction
The results from the Ambient Noise Tomography (ANT) at depths greater than ~3 km revealed among others, a region of decreased velocities across the southern part of the
Corinth Rift. We consider this low velocity zone to be related with the present-day active tectonic regime and as possible evidence of fluid circulation processes within a highly fractured upper crust. The observations from the ANT reflect an asymmetric velocity structure of the rift, with the southern parts having slower velocities than the northern ones. These characteristics are possibly representing an overall and more permanent in time picture of the rift's velocity structure. Following the above observation and motivated also by previous research findings suggesting a fluid-driven spatio-temporal evolution of the Corinth Rift's seismicity, the main objective of the second part of this thesis, is to investigate evidence of possible changes in the seismic propagation characteristics of the crust within a smaller time scale, related with the occurrence of critical events in the area, concentrating on the northern part of the western Corinth Rift.
We attempt to meet the objective of the second part by performing a shear-wave splitting analysis in the western Corinth Rift, and more specifically, by investigating the potential temporal variability of the shear-wave splitting parameters relative to earthquake occurrences in the area. For this purpose, we analyzed waveforms of local
125 Chapter 7 SWS analysis in the western Corinth Rift shear-waves recorded during the years 2009-2010 in the epicentral area of the
January 2010 Efpalio earthquakes. After describing the available data and the methods of the shear-wave splitting analysis that we followed, in the next chapters we will present and discuss the measured parameters and the detected variations. As a result, we provide evidence that the observed changes of the aforementioned parameters were possibly caused by temporally evolving conditions in the upper crust.
These changes are validated by comparing measurements from event multiplets that have occurred at different time periods (before and after the Efpalio earthquakes) but at the same focal area. Finally, we attempt to interpret the causative factors of the observed temporal variations, associated with the Efpalio earthquakes occurrence, in terms of the regional stress field and the possible involvement of fluid circulation processes.
7.2 The January 2010 Efpalio earthquakes
On January 18th, 2010 an earthquake of MW 5.3 (Efp1) occurred near the town of
Efpalio along the northern coast of the western Corinth Rift. Four days later, on
January 22nd, about 5 km to the north-east from the first earthquake another MW 5.2 event occurred (GMT 00:46) (Efp2, Figure 7.1). The two main shocks and the spatio- temporal evolution of the Efpalio sequence were thoroughly studied by Ganas et al.
(2013), Sokos et al. (2012), and Karakostas et al. (2012) among others. According to
Sokos et al. (2012), the January 18th and 22nd earthquakes were located at hypocentral depths of about 6.6 and 8.0 km, respectively. Both events according to the aforementioned studies exhibit normal faulting along E-W trending planes. However, the slip direction along the fault planes was rather controversial and there was no general agreement between the previous studies. Ganas et al. (2013) suggested that
126 Chapter 7 SWS analysis in the western Corinth Rift
Fig. 7.1: Map of the study area in western Corinth Rift. Seismic stations used for the SWS analysis are shown as triangles, where green and red colors signify stations operated by the University of Patras Seismological Laboratory (UPSL) and Corinth Rift Laboratory (CRL) in collaboration with the Seismological Laboratory of Athens University (NKUA), respectively. The seismic events (colored circles) from which the valid splitting results were obtained, the Efpalio earthquakes epicenters (Efp1 and Efp2 as stars), major cities (squares) and major fault traces of the area are also shown. As in Doutsos & Poulimenos (1992), Flotté et al. (2005), Papanikolaou et al. (1997), Valkaniotis (2009) the major faults shown are: 1=Psathopyrgos, 2=Trizonia, 3=Trikorfo, 4=Filothei, 5=Marathias, 6=Antirio, 7=Drosato, 8=Efpalio, 9=Kamarai/Selianitika, 10=Aigion and 11=off shore fault related to Efpalio sequence (according to Sokos et al. 2012) and other on- and off-shore faults. The orientation of the principal stress axes after Kokkalas et al. (2006) is shown at the top left-hand corner. The depths of the events are color coded according to the color scale (bottom-right). The diameters of the circles are proportional to the magnitudes.
the Efp1 event ruptured a blind, north-dipping fault accommodating north-south extension of the western Corinth Rift, while the Efp2 event possibly ruptured a fault with a weakly imaged south-dipping plane. Karakostas et al. (2012) proposed that both events ruptured faults dipping to the north. According to Sokos et al. (2012) the first event was related with a south-dipping fault plane, while the second event seemed to be related with a north-dipping. According to the previous study, these
127 Chapter 7 SWS analysis in the western Corinth Rift structures may coincide with already mapped surface traces of faults. The surface trace for the south-dipping fault of the first event is very well correlated with the
Trikorfo-Filothei south-dipping fault while the extrapolated surface termination of the assumed causative fault for the second event is located offshore, close to the north- dipping fault mapped by Papanikolaou et al. (1997).
7.3 Data
In the framework of the second part of the thesis, we used recordings from six permanent broadband stations, located on the western part of the Corinth Rift, operated under the framework of the Hellenic Unified Seismological Network (HUSN) and the Corinth Rift Laboratory (CRL, http://crlab.eu/) (Figure 7.1). Efpalio (EFP) and
Sergoula (SERG) stations were deployed by the University of Patras Seismological
Laboratory (UPSL, http://seismo.geology.upatras.gr/) in cooperation with Charles
University in Prague (http://geo.mff.cuni.cz/) while the other four seismic stations of
Trizonia (TRIZ), Kalithea (KALE), Rodini (ROD) and Lakka (LAKA) were deployed by the CRL in cooperation with the Seismological Laboratory of Athens University
(NKUA, http://dggsl.geol.uoa.gr/). Despite the fact that the available seismic network in the western Corinth Rift is one of the densest in Greece, many stations could not be used for a shear-wave anisotropy analysis. Since the critical values of the shear-wave splitting parameters (especially the shear-wave polarization directions) is often subject to instrument orientation errors and in order to avoid miscalculations and possible misleading observations, we did not include in our analysis recordings from the majority of the CRL's seismometers which are located in 60 - 190 m boreholes and characterized by unknown orientation. Lack of data, lack of seismicity, ambiguous S- phase arrivals, lack of valid anisotropy measurements within the shear-wave window
128 Chapter 7 SWS analysis in the western Corinth Rift of certain stations was some of other minor reasons which did not allowed us to use some other stations.
The waveform data studied here consists of the background seismicity and aftershocks of the MW 5.3 and 5.2 Efpalio earthquakes, recorded from January 2009 until
December 2010. The location parameters of the studied events were provided by the
Institute of Geodynamics - National Observatory of Athens (IGNOA, http://bbnet.gein.noa.gr/). The mean location errors in the horizontal and vertical directions did not exceed ±0.6 and ±1.9 km, respectively. A general map of western
Corinth Rift is already presented in Figure 7.1, showing the seismic stations and the events from which the shear-wave splitting parameters were measured. Table 7.1 presents the valid shear-wave splitting measurements (data availability) through the studied time period.
Table 7.1. Data availability (shear-wave splitting measurements) for the studied time period. Months with at least some data are plotted in grey.
2009 2010 J F M A M J J A S O N D J F M A M J J A S O N D EFP SERG TRIZ ROD LAKA KALE
7.4 Methodology
7.4.1 Shear wave splitting measurements
The technique we used to measure the splitting parameters was adopted from the cross-correlation method which was firstly introduced by Ando et al. (1983). As it was already mentioned, waveforms of local seismic events can be distorted by phase conversions in the case they reach the free surface at incidence angles larger than a
129 Chapter 7 SWS analysis in the western Corinth Rift critical value (Booth & Crampin 1985, see details in Chapter 5, Paragraph 5.3). For this reason, we selected seismic events that correspond to incidence angles < 45o, all located within the effective shear-wave window (Booth & Crampin 1985) of every station. Before applying the methodology of the splitting measurements, we visually inspected all the candidate seismic events and manually picked P- and S-wave arrivals. This procedure was mostly performed in order to select events with clear and impulsive S-wave arrival phases, as well as to exclude seismograms with controversial
S-phase records possibly due to their vicinity at the shear-wave window's limits or due to free-surface topographic irregularities.
According to method adopted for the SWS analysis, the seismograms were interpolated to 200 samples s-1, integrated to displacement and then band-pass filtered between 1 Hz and 10 Hz using a two-pole Butterworth filter. The measurement window for each waveform was defined in the following way: the start of the window was fixed 0.05s before the S-wave arrival while the endpoint was adjusted each time until the value of cross-correlation coefficient C between the fast and slow components was maximized. In this method, the seismograms of the horizontal components are both rotated in the horizontal plane at 1 degree increment of azimuth (α) from -90 to
90 degrees. Then, for each azimuth, the cross-correlation coefficient C is calculated between the two orthogonal seismograms for a range of time-delays (τ) in a selected time window, with 0.02s time-shift increment. When the absolute value of C becomes maximum, the corresponding values of azimuth (α) and time (τ) are chosen as the fast polarization direction and the time-delay between the splitted shear-waves, respectively. Following Kuo et al. (1994), the measurement's uncertainty is estimated using a t-test at a 95% confidence level on the values of C. Examples of two valid splitting measurements are presented in Figure 7.2.
130 Chapter 7 SWS analysis in the western Corinth Rift
We accept as valid the splitting measurements which conform to the following criteria:
(a) the C value is larger than 0.80
(b) the signal-to-noise ratio is larger than 2.5
(c) the change of the measured dt is less than 0.02s when the window size is varied by
±0.02s, and
(d) the change of the measured φ is less than 10° when the window size is varied by
±0.02s.
Finally, in order to verify to some extent the accuracy and the validity of the measurements, we made a final visual inspection to all the splitting measurements we considered as valid, verifying a clear linearity of the particle motion once the splitting effects (φ and δt) have been removed (see Figure 7.2).
Fig. 7.2: Two examples (a) and (b) of valid splitting measurements of the shear-waves recorded at the EFP station for two events that occurred in January 19th and 28th, respectively. Upper panels: Contour diagrams of the cross-correlation coefficient in the (φ, δt) space. The preferred solutions of (φ, δt) corresponding to the maximum value (dot) are shown within the 95% confidence regions (dotted lines). As measurement uncertainties we consider the range of the possible extreme values of φ and δt, respectively, in the confidence region (red lines in 7.2a). Lower panels: Superposition of the horizontal components (upper traces), and the corrected fast and slow components (lower traces) once the splitting effects have been removed. Particle motions diagrams are shown to the right of each sub-panel.
131 Chapter 7 SWS analysis in the western Corinth Rift
7.4.2 Estimation of VP/VS ratios
Following the approach of Nur (1972), under the assumption of linear ray paths, we calculated an average VP/VS ratio using the estimated travel times at each station for all the events that satisfied the splitting criteria:
V tS (7.1) VS t
with tS and t , where and are the arrival times of the S- and P- waves, respectively. The calculated VP/VS ratios are then compared to the corresponding time delays, giving additional information about the average properties of the medium along the ray paths.
132
Chapter 8
Shear Wave Splitting Results
8.1 Spatio-temporal grouping of the data
In order to present our findings as efficiently as possible, we separated the 2-year long dataset into three sub-periods. The time period before the occurrence of the Efpalio earthquakes, from January 2009 until January 18th, 2010 (hereafter called "Period I"), the period that began soon after the occurrence of the earthquakes, including the aftershock sequence, from January 22nd, 2010 until the end of June, 2010 (hereafter called "Period II"), and the remaining time period until the end of 2010 (hereafter called "Period III").
Another important factor for an efficient representation and interpretation of the measurements, in addition to the previous temporal separation of the data, is the spatial distribution of the studied seismic events. For this reason, we also grouped the data spatially to that located inside and very close to the rupture areas of the Efpalio earthquakes and those located outside. We calculated the approximate dimensions of the rupture areas based on the empirical relationship between the moment magnitude and the rupture area proposed by Wells & Coppersmith (1994) (Figure 8.1). Since the aftershock activity of the Efpalio earthquakes seemed to expand slightly into the surrounding area beyond the calculated boundaries of the rupture areas (see fig. 1b in
Sokos et al. 2012), for selecting the data which were close to the rupture area, we broadened the selection boundaries in accordance with the distribution of the early aftershocks during the first ~30 days after the Efpalio earthquakes as it is proposed by
Sokos et al. (2012).
133 Chapter 8 SWS Results
Fig. 8.1: The distribution of the studied seismic events that occurred before (a) and after (b) the Efpalio earthquakes. Vertical cross-sections along the study area depicting the previous seismicity are also presented (a1, a2, b1, b2). The shaded rectangles represent the projections on the surface of the rupture areas of the Efpalio earthquakes (Efp1 and Efp2). The approximate dimensions of the rupture areas were calculated according to Wells & Coppersmith (1994). The dashed ellipses were created for the need of the spatial grouping of the data that are considered to be within and close to the rupture areas. The ellipses were designed approximately according to the slight expansion of the early aftershocks beyond the calculated boundaries of the rupture areas during the first ~30 days after the main shocks following Sokos et al. (2012). Seismic stations, major cities and major fault traces are also presented as in Figure 7.1.
By separating the studied time period and the data in these ways, considering the occurrence of the Efpalio earthquakes as a significant time point of our study, we are actually focusing on time periods in which the properties of the upper crust possibly exhibit different characteristics.
134 Chapter 8 SWS Results
8.2 A first overview of the measurements
After the shear-wave analysis, considering the whole dataset, we obtained 439 valid splitting measurements derived from 416 seismic events with local magnitudes ≤ 3.3 and focal depths ranging from ~4 to ~15 km. Specifically, we obtained 108 valid measurements for the Period I, 257 for the Period II and 74 for the Period III.
The calculated time delays were normalized according to their hypocentral distances in order to exclude possible dependence of time delay values on event locations.
Therefore, when we talk about time delays in the following paragraphs, we are actually referring to normalized time delays (ms per km of hypocentral distance). A first overview of the results using the complete dataset shows that the time delays estimated for the eriod I had a mean value of 2.9 ±0.4 ms/km while after the Efpalio earthquakes, in Period II, there was an increase to a mean value of 5.5 ±0.5 ms/km. In
Period III the mean value of δt decreased in 3.6 ±0.4 ms/km. The mean fast polarization direction varied between 68o at SERG station, and 125o at KALE station, with a mean of 84° ±9°. Figure 8.2 shows rose diagrams of the fast shear-wave polarization directions as they have been measured during the three sub-periods for every single station. The VP/VS ratios exhibit an average value of 1.76 ±0.04 in eriod
I, while in Periods II and III, the average value significantly increased at 1.88 ±0.04 and 1.87 ±0.04, respectively. Table 8.1 gives a summary showing a list of all the available stations along with the average values of the shear-wave splitting parameters. In the following paragraphs, we present in detail the results of this study in accordance with the aforementioned spatio-temporal grouping of the data that we performed.
135 Chapter 8 SWS Results
Fig. 8.2: Maps of the western end of the Corinth Rift showing rose diagrams of the measured fast shear-wave polarization directions. Seismic stations, major cities and major fault traces are also presented as in Fig. 7.1. (PERIOD I: time period before the 1st Efpalio event (January 2009 - Efp1), PERIOD II: time period after the 1st Efpalio event until the end of the aftershock sequence (Efp1 - end of May 2010) and PERIOD III: time period after the end of the aftershock sequence (June 2010 - December 2010)).
136 Chapter 8 SWS Results
Table 8.1: Summary of the average values of the shear-wave splitting parameters measured per seismic station for the whole dataset a Nobs φ δt σ Station Period I Period II Period III Total (o) (ms/km) (ms/km) EFP 30 136 39 205 75 3.6 4.9 SERG 19 64 46 129 68 6.3 4.2 ROD 16 16 12 44 86 4.2 3.5 KALE 8 8 10 26 125 6.3 4.2 TRIZ 23 - - 23 122 3.1 2.7 LAKA 12 - - 12 108 1.2 0.9 a Nobs denote the number of observations per station, φ is the mean of the fast polarization directions based on directional statistics (see Appendix B), δt is the average time delays normalized according to the hypocentral distance and σ is the standard deviation of these values. Period I: time period before the 1st Efpalio event (January 2009 - Efp1), Period II: time period after the 1st Efpalio event until the end of the aftershock sequence (Efp1 - end of May 2010) and Period III: time period after the end of the aftershock sequence (June 2010 - December 2010).
8.3 Within and close to the rupture areas
A number of 261 valid splitting measurements were derived from seismic events that occurred within and close to the rupture areas of the Efpalio earthquakes. We obtained 42 valid measurements for the Period I, 167 for the Period II and 52 for the
Period III. As shown in Figure 8.1, EFP station is the only station which is located just above the rupture areas, making in this way the measurements from this station the most representative of the area very close to the rupture zones. Diagrams showing the variation of (a) the measured shear-wave time delays, (b) fast polarization directions and (c) VP/VS ratios from this part of the crust versus time are presented in Figure 8.3.
The time delays estimated for the eriod I had a mean value of 2.4 ±0.5 ms/km while after the Efpalio earthquakes, in Period II, there was an increase to a mean value of
5.8 ±0.5 ms/km. In eriod III the mean value of δt slightly decreased to 5.5 ±0.4 ms/km. The increase in time delays is clearly observed soon after the occurrence of the
Efpalio earthquakes, exhibiting a decreasing trend a few months later, after the end of
137 Chapter 8 SWS Results the aftershock sequence (Figure 8.3a). For the Period I, fast polarization directions show a mean value of 59o ±10o, while in Periods II and III, the mean values were estimated at 82o ±9o and 69o ±9o, respectively. The VP/VS ratios exhibit an average value of 1.78 ±0.03 in eriod I, while in eriods II and III, the average value significantly increased at 1.89 ±0.04 and 1.88 ±0.04, respectively. Corresponding errors for each VP/VS ratio have been estimated by using the uncertainties of P- and S-phase picks. It is noteworthy that average VP/VS ratio in Period I is about equal or somewhat lower than the background values estimated by several authors for the same region.
For instance, Latorre et al. (2004) observed VP/VS ratios about 1.78 for the first 15 km of the crust. Rigo et al. (1996) and Novotny et al. (2012) observed velocity ratios of about 1.83 and 1.80, respectively. We think that the velocity model proposed by
Latorre et al. (2004) is quite representative for the study area as it was derived during a passive tomography study using seismic events very close to the Efpalio area. The calculated 1.78 ±0.03 VP/VS ratio for Period I is about the same with that estimated from Latorre et al. (2004). While the increased observed average VP/VS ratios after the
Efpalio earthquakes are higher 6.2% for Period II, and 5.6% for Period III than the previously reported value.
8.4 Outside the rupture areas
Focusing on the area outside of the rupture zones of the Efpalio earthquakes, we obtained a total of 178 valid splitting measurements. Specifically, we obtained 43 valid measurements for the Period I, 72 for the Period II and 63 for the Period III. In this case, all seismic stations, except the EFP station, are located outside of the rupture areas (see Figure 8.1). Diagrams showing the variation of the measured shear-wave time delays (a), fast polarization directions (b) and VP/VS ratios (c) from this area are presented in Figure 8.4.
138 Chapter 8 SWS Results
Fig. 8.3: Diagrams showing the variation of the measured shear-wave time delays δt (a), fast polarization directions φ (b) and VP/VS ratios (c) with time from events located close/within the rupture areas. The approximate dimensions of the rupture area are shown in Figure 8.1. The time delays were normalized according to the hypocentral distances. Black vertical bars represent measurement errors.
139 Chapter 8 SWS Results
Measurements taken from outside the rupture areas show that the time delays estimated for the eriod I had a mean value of 3.4 ±0.3 ms/km while after the Efpalio earthquakes, in Period II, there was an increase to a mean value of 5.2 ±0.4 ms/km. In
Period III the mean value of δt slightly decreased in 4.9 ±0.4 ms/km. In this case also, an increase in time delays is clearly observed soon after the occurrence of the Efpalio earthquakes, exhibiting a decreasing trend a few months later, after the end of the aftershock sequence (see Figure 8.4a). For the Period I, fast polarization directions show a mean value of 96o ±9o, while in Periods II and III, the mean values were estimated at 74o ±9o and 69o ±9o, respectively. The VP/VS ratios exhibit an average value of 1.75 ±0.03 in eriod I, while in both eriods II and III, the average value significantly increased at 1.85 ±0.04. Similarly to the previous results from the events inside and close to the rupture areas, the average VP/VS ratio in Period I is lower than the background values. The calculated VP/VS ratio for Period I is lower 1.7% than that estimated from Latorre et al. (2004). While the average VP/VS ratios after the Efpalio earthquakes, for both Periods I and II, are higher 3.9%.
8.5 Validation of observations
In order to validate the observed changes of the parameters in time and exclude the possibility that different ray-paths bias the measurements values, we followed two procedures. Firstly, we followed a non-parametric hypothesis testing framework. A two-sample Kolmogorov- Smirnov (KS) test was applied to the time delays and VP/VS ratios (Gibbons 1971) and a statistical test appropriate for directional data to the fast polarization directions (Trauth 2010). The statistical tests were applied once between the datasets of Periods I and II and then between the datasets of Periods II and III, with a level of significance of 5% (for details about the statistical testing see Appendix
B).
140 Chapter 8 SWS Results
Fig. 8.4: Diagrams showing the variation of the measured shear-wave time delays δt (a), fast polarization directions φ (b) and VP/VS ratios (c) with time from events located outside the rupture areas. The approximate dimensions of the rupture area are shown in Fig. 8.1. The time delays were normalized according to the hypocentral distances. Black vertical bars represent measurement errors.
141 Chapter 8 SWS Results
The statistical tests revealed the following interesting observations:
(a) the change in time delays values derived from events inside and close to the
rupture zones was significant between the periods before and after the Efpalio
earthquakes
(b) the change in time delays values derived from events outside the rupture zones
was significant but less stronger than the previous one
(c) the difference in VP/VS values was significant between period I and II, and not for
the periods II and III
(d) φ values did not change significantly through the time periods for both the spatial
groups of data
Secondly, we searched our catalogue of the valid measurements for similar earthquakes (hereafter called "multiplets"). We considered an earthquake doublet as a pair of earthquakes consisting of an earthquake that occurred before the Efpalio events and an earthquake that occurred after that, with a cross-correlation coefficient greater than 0.7, similar magnitude and spaced in distances less than the mean horizontal and vertical location error. A 0.05 - 5 Hz band-pass filter was used in pre- processing of the seismic waveforms as the waveforms cross-correlation coefficient is more stable at lower frequencies (e.g. Shearer 1997; Shearer et al. 2005). Waveforms of seventeen such multiplets were found recorded by the EFP, SERG and ROD stations. Figure 8.5 shows diagrams with the variation of the measured (a) shear-wave time delays, (b) fast polarization directions and (c) VP/VS ratios from January 2009 to
December 2010 for event multiplets.
142 Chapter 8 SWS Results
Fig. 8.5: Diagrams showing the variation of (a) the measured shear-wave time delays δt, (b) fast polarization directions φ and (c) VP/VS ratios from January 2009 to December 2010 for event multiplets. The time delays were normalized according to the hypocentral distances. Black vertical bars represent measurements errors.
Considering the time delay measurements that were obtained from the multiplets only, δt had an average value of 1.73 ±0.4 ms/km before the Efpalio earthquakes and
2.81 ±0.4 ms/km after their occurrence, displaying a relative percentage increment greater than 50%. Fast polarization directions did not exhibit any significant variation before and after the occurrence of the Efpalio earthquakes, as an average direction of
101o ±11o and 97o ±9o was obtained respectively. Finally, concerning the VP/VS ratios, an average 1.80 ±0.037 was measured before and 1.91±0.042 after. Both the application of the statistical testing and the use of doublet earthquakes confirmed the same variation of the initial derived results before and after the Efpalio earthquakes.
Table 8.2 presents a comparison of the average values of all the studied parameters
143 Chapter 8 SWS Results before and after the Efpalio earthquakes as they were derived from the earthquake multiplets. According to the previous findings we suggest that our observation shows a robust temporal variation of the time delays and VP/VS during the years 2009-2010.
Table 8.2: Comparison of the average values of the studied parameters (φ, δt, VP/Vs, VP and Vs) before and after the Efpalio earthquakes, derived from the earthquake multiplets a
Before Efpalio After Efpalio Percentage of Parameters sequence sequence relative change φ (o) 101 97 -
δt (ms/km) 1.73 2.81 +53%
VP/Vs 1.80 1.91 +5%
VP (km/s) 6.01 6.11 +0.5%
Vs (km/s) 3.34 3.21 -5% a φ is the mean of the fast polarization directions based on directional statistics, δt is the average time delays, VP and Vs are the P- and S-velocity, respectively.
In the following pages we present detailed plots presenting diagrams with the variation of the measured parameters (both normalized and non-normalized δt, φ and
VP/VS ratios) for each individual station from January 2009 to December 2010. Equal area projections, as well as the waveforms of the multiplets that were used for the validation are also presented.
144 Chapter 8 SWS Results
Fig. 8.6: Diagrams showing the variation of the measured shear-wave time delays δt (a), fast polarization directions φ (b) and VP/VS ratios (c) from all the available data recorded between January 2009 and December 2010 per station. The time delays were normalized according to the hypocentral distances. Black vertical bars represent measurement errors. The names of the stations are shown at the top left corner of each panel (figure continues in the next pages).
145 Chapter 8 SWS Results
Fig. 8.6: For description see previous page.
146 Chapter 8 SWS Results
Fig. 8.6: For description see page 145.
Fig. 8.7: Diagrams showing the variation of the measured shear-wave non-normalized time delays δt with time. Black vertical bars represent measurement errors.
Fig. 8.8: Diagram showing the variation of the measured shear-wave non-normalized (raw) time delays from the whole dataset. Black vertical bars represent measurement errors.
147 Chapter 8 SWS Results
Fig. 8.9: Diagrams showing the variation of the measured shear-wave non-normalized (raw) time delays δt per station. Black vertical bars represent measurement errors. The names of the stations are shown at the top left corner of each panel.
148 Chapter 8 SWS Results
Fig. 8.10: Equal area projections of the fast polarization directions. (a) whole time period (January 2009 - December 2010), (a1) time period before the 1st Efpalio event (January 2009 - Efp1 (Jan. 18, 2010)), (a2) time period after the 1st Efpalio event until the end of the aftershock sequence (Efp1 - end of May 2010) and (a3) time period after the end of the aftershock sequence (June 2010 - December 2010). The radius of the plots is scaled to an incidence angle of 45° (figure continues in the next pages).
149 Chapter 8 SWS Results
Fig. 8.10: For description see previous page.
150 Chapter 8 SWS Results
Fig. 8.10: For description see page 149.
Table 8.3: Pairs of earthquake multiplets with the corresponding splitting parameters (φ and δt) as well as the corresponding values of VP/Vs ratio, VP and Vs. a
Origin time φ δt VP Vs VP/Vs (yymmddhhmmss) (degrees) (ms/km) (km/s) (km/s) 090816142141 150 1.53 1.80 6.40 3.56 1 100412061626 145 1.05 2.09 6.88 3.29 090816142141 150 1.53 1.80 6.40 3.56 2 100124210445 152 2.22 1.93 5.77 2.99 091211193311 133 2.14 1.90 6.22 3.27 3 100124210445 152 2.22 1.93 5.77 2.99 091211193311 133 2.14 1.90 6.22 3.27 4 100124002502 103 4.33 2.02 6.65 3.29 090626092305 63 0.59 1.82 5.75 3.16 5 100124041713 87 4.74 2.00 5.94 2.97 090815062952 162 0.59 1.95 6.02 3.09 6 100124002502 103 4.33 2.02 6.65 3.29 090626092305 63 0.59 1.82 5.75 3.16 7 100124002502 103 4.33 2.02 6.65 3.29 090816142141 150 1.53 1.80 6.40 3.56 8 100428215530 147 1.60 1.71 5.77 3.37 090815062952 162 0.59 1.95 6.02 3.09 9 100124210445 152 2.22 1.93 5.77 2.99 090626092305 63 0.59 1.82 5.75 3.16 10 100123204447 93 0.73 2.05 6.73 3.28 090517195310 60 2.81 1.72 5.49 3.19 11 100325124340 16 3.00 1.92 5.88 3.06 090517195310 60 2.81 1.72 5.49 3.19 12 100326224348 32 2.24 1.81 5.76 3.19 090517195310 60 2.81 1.72 5.49 3.19 13 100510191439 9 2.69 1.87 6.24 3.33 090507201003 53 2.97 1.71 6.55 3.84 14 101204053751 77 3.20 1.67 5.63 3.38 090508090336 51 2.80 1.63 6.14 3.76 15 101204053751 77 3.20 1.67 5.63 3.38 091007012012 58 2.82 1.70 5.89 3.46 16 101204053751 77 3.20 1.67 5.63 3.38 091109025849 108 4.68 1.59 5.99 3.77 17 100221203321 115 6.04 1.64 5.12 3.12 a φ and δt denote the fast polarization direction and the time delays, respectively.
151 Chapter 8 SWS Results
Fig. 8.11: Waveforms of "earthquake multiplets" or "similar earthquakes". Figure continues in the next pages, for description see next page.
152 Chapter 8 SWS Results
Fig. 8.11: Waveforms of "earthquake multiplets" or "similar earthquakes" recorded by the EFP, SERG and ROD stations. We considered the similar earthquakes/earthquake multiplets as a pair of earthquakes consisting of an earthquake that occurred before the Efpalio events and an earthquake that occurred after that, with a cross-correlation coefficient greater than 0.7, similar magnitude and spaced in distances less than the mean horizontal and vertical location error. A 0.05 - 5 Hz band-pass filter was used in pre-processing of the seismic waveforms as the waveforms cross-correlation coefficient is more stable at lower frequencies (e.g., Shearer 1997; Shearer et al. 2005).
153
Chapter 9
Interpretation & Discussion
9.1 Stress field and fast polarization directions
It has conclusively been shown that the Corinth Rift is one of the most representative and extensively studied areas of active extensional deformation worldwide (see
Chapter 2). Considering the study of Kokkalas et al. (2006), the extension of the rift is mainly controlled by WNW and ENE-striking normal faults. A detailed stress tensor analysis in the Corinth Rift, presented in the previous study, shows a σ3-axis in a nearly N-S direction (see Figure 7.1 and fig. 4 in Kokkalas et al. 2006) and small deviation from this general direction is due to the prevalence of one of these two main fault sets. The zones at the junction between the WNW and ENE-striking faults seem to be noticeable in terms of the Corinth Rift's seismicity. Areas near the bend of the two fault orientations, acted as initiators of moderate to large earthquakes at the past
(e.g. the 1993 Ms 5.6 Patras earthquake and the 1981 Alkionides earthquakes). The computed focal mechanisms of the Epf1 and Efp2 events showed T-axis azimuths of
187o and 1o, respectively (Sokos et al. 2012), observations that are in agreement in general with the trend of the seismo-tectonic, stress and strain regime of the rift.
The observed ENE-WSW (84o ±9o) direction of φ is in a good agreement with the regional stress and strain field. According to the mean values of fast polarization directions in Periods I, II and III derived from events both from inside and outside the rupture areas, and also taking into account their measurement errors, the fast polarization directions did not reveal any significant change. The Efpalio earthquakes
155 Chapter 9 Interpretation & Discussion seemed to have little or no influence on this parameter. We suggest that the observed
~E-W fast shear wave polarization direction is caused by pre-existing stress-aligned microcracks, oriented parallel or sub-parallel to the horizontal maximum stress axis, which is parallel to the trend of faulting and perpendicular to the N-S extension of the
Corinth Rift. The orientation of these microcracks most probably did not change after the Efpalio earthquakes.
Previous local shear-wave anisotropy studies performed in the same and nearby areas, such as the works of Bouin et al. (1996), Bernard et al. (1997), Kaviris et al. (2008),
(2010) and Papadimitriou et al. (1999), (2010), (2014), pointed out similar mean fast polarization directions (~E-W), to the ones inferred from the current work, being in agreement with the overall tectonic regime of the Corinth Rift. Minor differences observed between the previous studies could be possibly explained among others, by the different measurement methodologies that were adopted by each researcher.
Interestingly, the findings of the previous studies, including the finding of the current analysis, provide evidence that the polarization direction of the fast shear-waves did not exhibit significant variations during the last two decades, reflecting microcrack systems with relatively stable orientation through time.
Other research studies, such as Hatzfeld et al. (2001), Evangelidis et al. (2011) and
Endrun et al. (2011), concentrated on deeper parts of the lithosphere, investigating the azimuthal anisotropy in the lower crust and mantle across the Hellenic subduction zone and the Aegean region by performing SKS and surface wave anisotropy analyses.
It would be interesting to compare our measurements with fast polarization directions deduced for deeper parts of the lithosphere, however, the network coverage that was used from the previous studies was not dense enough in the broader region of our
156 Chapter 9 Interpretation & Discussion study area. Possible future measurements from a denser network around the study area combined with the measurements of the current work could allow us to study, for instance, the degree of the vertical coherence of deformation in the Corinth Rift.
9.2 Possible causes of δt, VP/VS variations
According to Crampin (1999) and several other studies afterwards, the principal cause of time delays variations is fluid migration along pressure gradients between closely- spaced microcracks and pores. We have already mentioned in Chapter 5 (in the frame of the interpretation and discussion of the ambient noise tomography) the findings of several studies (Cornet et al. 2004; Bernard et al. 2006; Koukouvelas & Papoulis 2009;
Pick & Marty 2008; Baud et al. 2004, Lattore et al. 2004, Gautier et al. 2006,
Pacchiani & Lyon-Caen 2010 among others), that indicated the presence of fluids in the western Corinth Rift and their key role in the mechanics of active faulting and the seismicity. Concerning the findings of these studies and the observed time delays variation that we detected from our analysis, we suggest that the Efpalio earthquakes caused a change in the properties of the crust, as well as in the pre-existing microcrack system geometry. The observed distinctive increase in the time delays on one hand and the maintenance of the same mean fast polarization direction before and after the earthquakes on the other hand, suggest that the cause of the observed variations in the splitting parameters was a possible migration of over-pressured fluids through the pre-fractured damage zone of the study area. Based on the fact that the increase in the time delays, after the Efpalio earthquakes, derived from the data located close to the rupture zones is more intense than the observed increase from the data located outside the rupture areas, we assume that the degree of the changes in the properties of the crust is stronger close to the rupture zones than away from them.
157 Chapter 9 Interpretation & Discussion
A similar example relating to migration of fluids in an over-pressured condition within a pre-fracture zone was examined for the case of the 2009 MW 6.3 L'Aquila earthquake in central Italy by Di Luccio et al. (2010). In summary, the previous study has revealed, among others, diffusion processes of over-pressured fluids within the pre- fractured crust following the occurrence of the MW 6.3 L'Aquila earthquake.
An additional interpretation tool concerning the possible involvement of over-pressure fluids in the study area is the VP/VS ratios measurements. Due to the sensitivity of
VP/VS ratios in pore fluids, this parameter is an appropriate tool not only for detecting fluid activities, but also for determining the possible fluid phase (gas or liquid phase).
VP/VS ratios reflect also the properties of the upper crust, varying with mineral, rock compositions and generally with lithology (Fernandez-Viejo et al. 2005). These kinds of changes in lithology occur during much larger time scales than the 2-yrs of our dataset. For this reason, variation in lithology does not seem to be a decisive factor that can influence the observed variations. Seismic tomography studies in different geological systems, like volcanic and geothermal, were successful to delineate zones of high or low VP/VS ratios (e.g., Gunasekera et al. (2003); Chiarabba & Moretti (2006) and others). These studies have highlighted the dominant role of fluids in influencing the values of VP/VS where liquids result in high VP/VS ratios, by lowering the VS, while gases shift the ratio to lower values as they affect more the VP in the rock. In order to investigate in more detail the causes of the increase of VP/VS after the Efpalio earthquakes, under the assumption of linear ray paths, we have already calculated apparent VP and VS velocities for the same set of events that had an estimated VP/VS ratio. We averaged these values for each of the three sub-periods and we also compared them with the background values for the top 15 km of the crust derived from the seismic velocity model of Latorre et al. (2004). The averaged values of the
158 Chapter 9 Interpretation & Discussion
measured VP/VS, VP and VS for each period along with the percentage of their relative change from the background values are shown in Figure 9.1.
Fig 9.1: The percentage of the relative change between the observed average VP/VS ratios, VP and VS velocities and the background values derived from the velocity model proposed by Latorre et al. (2004) from inside (a) and outside (b) of the rupture areas. Validation of the variation of the average values of the splitting parameters and VP/VS ratios that was observed through the studied time period, after the application of non-parametric hypothesis testing is also presented. A two-sample Kolmogorov-Smirnov
(KS) test (Gibbons 1971) was applied for the time delays and VP/VS ratios, while a statistical test relative to directional data (Trauth 2010) was applied for the polarization directions (for details about the statistical testing, see Appendix B).
The measurements of the apparent VP and VS shows that the observed changes in
VP/VS ratios after the Efpalio earthquakes were due to an increase in VP and a decrease in VS. This observation is reflected from data recorded from both inside and outside of the rupture areas. It is important to mention that despite the aforementioned variations in the VP and VS, their values were still higher than the background ones proposed by Latorre et al. (2004). The previous observed influence of
159 Chapter 9 Interpretation & Discussion
the fluids in the VP/VS ratios was also validated by the averaged apparent VP and VS values derived from the multiplets. In that case, the values of the calculated apparent velocities (see Table 8.2) show that the occurrence of the Efpalio earthquakes affected more the VS values than the corresponding VP since VS exhibited a 5.5% decrease after the Efpalio earthquakes, while VP increased by 0.5%. In Figure 9.2 a detailed representation of the percentage of the relative change between the observed average
VP/VS ratios, VP and VS velocities and the background values (Latorre et al. 2004) at each station is given. Based on the observed high VP/VS ratios after the Efpalio earthquakes and the fact that VS is decreasing while VP is increasing, we infer that pore and cracks space in the over-pressured cracked rock which is related to the seismically activated parts of the crust was most probably filled with liquid.
The three-dimensional travel-time tomography of the western Corinth Rift by Latorre et al. (2004) pointed out a quite complex structure, indentifying two distinct zones at depth exhibiting different characteristics. A shallow structure between 0-5 km and a deeper one between 7-11 km. The limit between the two zones (5 km -7 km) was suggested by the recovery of a large-scale vertical velocity anomaly and an increase in the seismicity rate (see figs. 13 to 16 in Latorre et al. 2004). At depths larger than 5-7 km, a significant increase in VP/VS ratios was found. Vertical VP/VS profiles indicated a possible correlation between the observed VP/VS anomalies and earthquake clusters located in the study area. According to authors of the aforementioned study, fluid saturation in fractured rocks could explain the high VP/VS ratio caused by the increase of VP and decrease of VS. More specifically, metamorphic processes involving phyllosilicate rocks may be responsible for the release of structural water by dehydration reactions. Because the previous authors and other studies (e.g. Xypolias
& Koukouvelas 2001; Dornsiepen et al. 2001) among others) have suggested the
160 Chapter 9 Interpretation & Discussion
Fig. 9.2: The same as in Figure 9.1.
161 Chapter 9 Interpretation & Discussion presence of phyllosilicate-rich rocks within the studied part of the crust, the high
VP/VS might be caused by the aforementioned metamorphic processes. One could argue that the fluids might have an alternative origin, suggesting for instance that fluids are possibly upper mantle sourced. A database of helium isotope measurements around
Greece and surrounding areas compiled by Pik and Marty (2009) does not seem to support such an explanation for our study area. The results of the previous work show a remarkable absence of mantle-helium signal in the Corinth rift fluids. Pik and
Marty (2009) interpreted the high proportion of crustal helium in the Corinth Rift suggesting that the fault system is rooted in the upper crust and is not connected at depth with zones where mantle-He has possibly been trapped.
Pore pressure diffusion in media with hydraulic diffusivity is one of the main mechanisms which control the triggering and the spatio-temporal evolution of the aftershock seismicity (Shapiro et al. 2003). When pore pressure diffusion processes operate, an envelope of cloud of events can be recognized in a plot of the distance of the pressure front from the fluid source versus time. Assuming a homogeneous isotropic medium, the distance of the pressure front from the triggering front (fluid source) can be approximated by the theoretical curve:
(9.1) where the distance r (radius of the triggering front) is a function of the diffusivity D and time t (time from the injection start) (Shapiro et al. 1997; Di Luccio et al. 2010). In the case of the Efpalio earthquake sequence, we analyzed the aftershocks distribution from the Efp1 up to February, 25th. The data used for this analysis consisted of 288 seismic events. The source parameters of the data were provided by Sokos et al.
(2012). In a distance versus time diagram (r-t) we identify a triggering front with a
162 Chapter 9 Interpretation & Discussion diffusivity of 4.5 m2s-1 (Figure 9.3). The procedure for fitting the theoretical curve of eq. 9.1 to the data was to select the farthest earthquakes that occurred in consecutive, non-overlapping time windows of five days, and search using a least squares search the diffusivity (D) value that provided the best fit. The estimated model of the spatio- temporal evolution of the aftershock seismicity (Figure 9.3) indicates that the diffusion process started soon after the Efp1 event since it was not observed any time lag between the Efp1 and the triggered earthquakes. The estimated diffusivity value of about 4.5 m2s-1 is within those reported in the literature (e.g., Talwani et al. 2007) and, in addition with the observed variation of the VP/VS ratios, it strongly supports the involvement of liquid fluids within the fractured medium. For the case of the study of the spatio-temporal evolution of an earthquake
Fig. 9.3: A plot showing the variation of the distances r of the January 18th to February 25th, 2010 aftershocks from the focus of the first Efpalio event (January 18th, 2010) versus their occurrence times t.
163 Chapter 9 Interpretation & Discussion swarm occurred in the southern coast of the western Corinth Rift (2001 Agios Ioannis earthquake swarm), Pacchiani & Lyon-Caen (2010) estimated a hydraulic diffusivity equal to 0.1 m2s-1.
It appears that the fluids in the epicentral area of the Efpalio earthquakes were more diffusible within a possibly more fractured crust. Another comparison can be made with the case of the L'Aquila earthquake, in which Di Luccio et al. (2010) estimated a
D value of about 4.5 m2s-1 for the foreshocks and a value of 80 m2s-1 for the aftershocks.
According to the authors, the value of 80 m2s-1, a threshold value for the diffusivity of crustal rocks, reflected the involvement of mantle sourced CO2-rich fluids. The difference between the D values derived from the previous studies and from the current work, reflecting different degrees of diffusion processes, is possibly due to the different properties of the fractured crust at the time periods when the studies were conducted and also due to the different magnitudes of the earthquakes of each case.
The largest event of the Agios Ioannis earthquake swarm (Pacchiani & Lyon-Caen
2010) had a moment magnitude of 4.3, while the L'Aquila earthquake, as previously mentioned, it was an MW 6.3 event.
Finally, it is noteworthy to mention the variation of the time delays in Period III and how it reflects the properties of the crust. After the significant increase that was observed in time delays after the Efpalio earthquakes, this parameter appears to have a decreasing trend, moving to background values (see Figures 8.3, 8.4, 8.6 - 8.9).
According to Crampin & Gao (2006), changes of stress affect the geometry of the microcrack system by changing the crack density and aspect-ratios, a process which is directly monitored by variations in average time delays. It seems possible that the cracks, started to close after the earthquakes due to stress relaxation and the crack density became smaller resulting in smaller values of time delays.
164 Chapter 9 Interpretation & Discussion
9.3 Conclusions
The results of the current shear-wave splitting analysis and the supplementary calculations can be summarized as follows: i) Shear wave splitting processes were observed in the western end of the Corinth Rift
during the years of 2009-2010. ii) Fast polarization directions presented a general E-W orientation, which is in
agreement with the regional stress and strain field. iii) Fast polarization directions did not change after the Efpalio earthquakes.
iv) A distinct increase in time delays and VP/VS ratios was observed soon after the
Efpalio earthquakes, followed by a decrease after the end of the aftershock
sequence. v) After the Efpalio earthquakes, a migration of over-pressured liquid fluids through
the fractured damage zone is possibly the main cause of the observed increase in
time delays and VP/VS ratios. The previous observed variations in the δt and VP/VS
ratio after the Efpalio earthquakes seemed to be slightly stronger close to the
rupture areas than outside of them, which are possibly reflecting different degree of
changes in the properties of the crust.
165
Chapter 10
Epilogue
10.1 Summary of thesis work
The principal objective of this doctoral dissertation was to investigate the structure and the properties of the upper crust beneath the western Corinth Rift through two different studies. Firstly, we applied for the first time in the area the technique of
Passive Seismic Interferometry and its application on Ambient Noise Tomography, a relatively new methodology for imaging the subsurface, to investigate the velocity structure of the area. We successfully applied this technique, we developed improvements in different parts of the processing methodology (e.g., increased the speed and the efficiency of the pre-processing tools by their further automating and streamlining), we gained valuable experience and improved our ability in processing and utilizing ambient seismic noise recordings for imaging purposes. Secondly, we concentrated on local shear-waves and performed seismic anisotropy measurements to investigate changes in the seismic propagation properties related to the occurrence of a critical event such as the January 2010 Efpalio earthquakes. The most important and interesting outcome of the pervious analysis was the detection of clear temporally evolving conditions in the upper crust prior and after the occurrence of the Efpalio earthquakes.
More specifically, in the frame of the first part, we analyzed Rayleigh-wave empirical
Green's functions emerging from cross-correlations of long-time series (3 yrs) of ambient seismic noise between pairs of all available stations in the western Corinth
167 Chapter 10 Epilogue
Rift. We used the emerged waveforms to measure Rayleigh-wave group velocity dispersion curves and obtain 2D group velocity maps. Finally we inverted these maps attempting to assess the 3D shear-wave velocity structure of the western Corinth Rift.
The results of Ambient Noise Tomography revealed the presence of interesting and distinctive velocity features. The estimated Rayleigh-wave group velocities are characterized by a strong lateral variation and they clearly reflect an asymmetric velocity structure of the western Corinth Rift, with the southern margins "slower" than the northern ones. The velocity perturbations observed at short periods (< 3 s) are interpreted mostly in accordance with the lithological variation. Of particular interest are the observed regions of decreased velocities at the south. The presence of a predominant low velocity zone below the southern part of the rift at periods above 3 s, at depths greater than 3 km, namely much deeper than the expected maximum thickness of the sediments, is considered the most interesting and profound feature in the velocity maps. It is characterized by a preferential WNW-ESE elongation, being sub-parallel to both the rift axes and the major active fault traces. This low velocity zone was also visible in horizontal and vertical depth sections of the inferred 3D shear- velocity model. We consider this low velocity anomaly to be related with the present- day active tectonic regime reflecting possible involvement of fluid circulation processes within and in the vicinity of these major faults at depth. These low velocity zones correspond to a highly fractured and fluid saturated crust within the more active southern margins of the western Corinth Rift, in contrast with the northern ones. Our findings strongly support the hypothesis of fluid interactions and their significant role to the overall evolution of the rift which is proposed by several research studies in the literature.
168 Chapter 10 Epilogue
Within the frame of the second part of the current thesis, the relevance of fluid circulation processes was also supported by the findings of the local shear-wave anisotropy analysis that we performed. During this analysis, we utilized shear-wave splitting on local seismic events in the epicentral area of the January 2010 Efpalio earthquakes and measured time delays and fast polarization directions one year before and one year after the earthquakes. Attempting to determine temporal variations on crustal rock properties related to the occurrence of the Efpalio earthquakes, we also used P and S travel-times in order to calculate apparent VP, VS and average VP/VS ratios along the ray-paths corresponding to each valid anisotropy measurement. The above analysis revealed a significant increase in both time delays between the fast and slow shear-wave components and VP/VS ratios soon after the
Efpalio earthquakes. Measurements of apparent VP and VS showed that the observed changes in VP/VS were affected mostly by a decrease in VS values. We interpreted the observed temporal variations as possible over-pressured fluids migration. We suggested that a migration of fluids (of crustal origin in the form of overpressured liquids) through the pre-fracture damage zone in the upper crust was triggered by the occurrence of the Efpalio earthquakes and caused the observed variations.
A notable conclusion drawn from both the shear-wave anisotropy analysis and the noise-based tomography includes evidence of fluids interactions not only within the more seismically active southern margins of the Corinth Rift, but also along the less active northern region. Furthermore, fluids interactions appear to be more intense and permanent in time in the southern part of the rift than the detected ones along the northern part which they mostly triggered by the Efpalio earthquakes occurrence. The observed variations in time delays and VP/VS ratio after the earthquakes were slightly
169 Chapter 10 Epilogue stronger close to the rupture areas than outside of them and they appeared to have a decreasing trend, moving to background values a few months later.
10.2 Conclusion Future work
In this thesis we intended to emphasize the usefulness of Passive Seismic
Interferometry technique as a complementary tool towards the investigation of the
Corinth Rift structure and we efficiently showed the potential to use ambient seismic noise recordings for Ambient Noise Tomography at a regional scale within a complex geotectonic environment. We believe that the current findings add to an ongoing and growing body of literature on understanding the seismotectonic evolution of the western Corinth Rift.
As with most research studies, new observations are commonly accompanied by new questions, hence there are many steps that can be taken to continue improving the methods and expand the research targets of this thesis. We hope that our observations will trigger further investigations on this topic. For instance, further investigation is needed to study in more detail the conditions needed for the presence of over- pressured fluids in an area which is characterized by extension, such the western
Corinth Rift, and how important could be the role of these fluids in the possibility of generation of slip along low dipping faults. Towards this direction, a time-lapse (4D)
Ambient Noise Tomography is recommended to be performed as a future work in order to delineate the areas of fluid saturation and their changes over time, since a considerable amount of data recorded during the last ~20 years in the Corinth Rift region is possibly available and appropriate for such a study. Recent studies of volcanoes and fault zones (e.g., Brenguier et al. 2008a, b; Clark et al. 2011) has successfully indicated that ambient seismic noise can be used as a monitoring tool by
170 Chapter 10 Epilogue utilizing the noise data even on a daily basis. Therefore, it could also be possible to monitor relative velocity variations and study potential changes in the elastic properties before and after the occurrence of moderate and major seismic events.
We would like also to mention the usefulness of a denser and extended seismic network made of broadband sensors around the western Corinth Rift in order to achieve more accurate images of smaller scale structures of both shallow and deep parts of the crust, and also to widen the frequency range of the reconstructed surface- waves and the Ambient Noise Tomography. Following the previous suggestion, adding some ocean bottom seismometers (OBS) in the center of the Corinth Rift's basin would greatly increase the area coverage (ray-path coverage) and the resolving potential of
Ambient Noise Tomography eliminating some uncertainty in our observations.
Finally, the reconstructed surface-waves from the ambient seismic noise could also be used to measure anisotropy with a relative good accuracy. In this thesis, we were able to use a small part of information which is contained in the whole set of noise cross- correlations between the sensors of the current network. We managed to reconstruct
Rayleigh-wave empirical Green’s functions from correlations between vertical-vertical component recordings. A more complete analysis of the whole dataset, including all three-component (Z, R, T) combination pairs of records, is suggested to be an additional subject of future work. For instance, the computation of the velocity variations of both quasi Rayleigh-waves and quasi Love-waves as a function of the azimuth will allow us to retrieve the fast direction and the amplitude of the azimuthal anisotropy across the network coverage. Shear-wave splitting measurements derived from seismic events can then be accompanied and compared with noise-based anisotropy analyses.
171
APPENDIX A
173
174
Appendix A ANT of the western Corinth Rift
Fig. A1: Maps presenting Rayleigh-wave group velocity measurements and ray-path coverage at 26 periods between 1 and 6 s with a step of 0.2 s. The red line shows the limit of the study area connecting grid cells with zero ray coverage. Seismic stations are shown as black triangles. Figure continued on the next pages.
* Note: Colormapping of velocities is presented with inverse colomaps within the Appendix, compared with those presented in the main thesis text.
175 Appendix A ANT of the western Corinth Rift
Fig. A1: See description at p. 175.
176 Appendix A ANT of the western Corinth Rift
Fig. A1: See description at p. 175.
177 Appendix A ANT of the western Corinth Rift
Fig. A1: See description at p. 175.
178 Appendix A ANT of the western Corinth Rift
Fig. A1: See description at p. 175.
179 Appendix A ANT of the western Corinth Rift
Fig. A1: See description at p. 175.
180 Appendix A ANT of the western Corinth Rift
Fig. A1: See description at p. 175.
181 Appendix A ANT of the western Corinth Rift
Fig. A2: Rayleigh-wave group velocity maps at 26 periods between 1 and 6 s with a step of 0.2 s. Seismic stations are shown as black triangles. For each period, the variance reduction (VarRed) between data computed from a homogeneous model with an average velocity and the final model, as well as the final mean velocity (Vmean) are shown at the lower left corner of each map. Major fault traces are presented as in Figure 4.16. Figure continued on the next pages.
182 Appendix A ANT of the western Corinth Rift
Fig. A2: See description at p. 182.
183 Appendix A ANT of the western Corinth Rift
Fig. A2: See description at p. 182.
184 Appendix A ANT of the western Corinth Rift
Fig. A2: See description at p. 182.
185 Appendix A ANT of the western Corinth Rift
Fig. A2: See description at p. 182.
186 Appendix A ANT of the western Corinth Rift
Fig. A2: See description at p. 182.
187 Appendix A ANT of the western Corinth Rift
Fig. A2: See description at p. 182.
188 Appendix A ANT of the western Corinth Rift
Fig. A3: The ray-path density at 26 periods between 1 and 6 s with a step of 0.2 s. Seismic stations are shown as blue triangles. Figure continued on the next pages.
189 Appendix A ANT of the western Corinth Rift
Fig. A3: See description at p. 189.
190 Appendix A ANT of the western Corinth Rift
Fig. A3: See description at p. 189.
191 Appendix A ANT of the western Corinth Rift
Fig. A3: See description at p. 189.
192 Appendix A ANT of the western Corinth Rift
Fig. A3: See description at p. 189.
193 Appendix A ANT of the western Corinth Rift
Fig. A3: See description at p. 189.
194 Appendix A ANT of the western Corinth Rift
Fig. A3: See description at p. 189.
195 Appendix A ANT of the western Corinth Rift
Fig. A4: Spatial resolution maps at 26 periods between 1 and 6 s with a step of 0.2 s. Seismic stations are shown as red triangles. Figure continued on the next pages.
196 Appendix A ANT of the western Corinth Rift
Fig. A4: See description at p. 196.
197 Appendix A ANT of the western Corinth Rift
Fig. A4: See description at p. 196.
198 Appendix A ANT of the western Corinth Rift
Fig. A4: See description at p. 196.
199 Appendix A ANT of the western Corinth Rift
Fig. A4: See description at p. 196.
200 Appendix A ANT of the western Corinth Rift
Fig. A4: See description at p. 196
201 Appendix A ANT of the western Corinth Rift
Fig. A4: See description at p. 196.
202
APPENDIX B
Appendix B
Statistical testing of the splitting parameters
B.1 Two-sample Kolmogorov-Smirnov test
In the geosciences, there are a lot of occasions where we want to test the hypothesis that two datasets derive from the same statistical distribution. However, the kind and sizes of the samples do not allow any assumption about their distribution (e.g.,
Gaussian distributed). In these cases, we need a non-parametric two-sample statistical test for checking this hypothesis. Such a test is the two-sample Kolmogorov-Smirnov
(KS) test (Gibbons 1971). The two-sample KS test is one of the most used and general non-parametric method for comparing two populations, as it is sensitive to differences in both the location and shape of the empirical cumulative distribution functions of the two samples. Specifically, let us assume X1, X2, two datasets, of length n and m, respectively, and let F1,n(x), F2,m(x) be their empirical distributions. We need to test under what circumstances the hypothesis F1,n(x) = F2,m(x) (the null hypothesis) is valid. In this case, the KS statistic is:
sup (B1) x
and the null hypothesis is rejected, with a level of significance α if:
(B2)
205 Appendix B Statistical Testing
where Ka is a constant threshold value that is derived from the Kolmogorov distribution. We applied this hypothesis-testing procedure in VP/VS ratios and time delays data, with a significance level a = 5%.
B.2 F statistic test for directional data
Let us consider: θ1, θ2,..., θn and φ1, φ2,..., φm as the two sets of azimuth measurements.
We wish to statistically test the hypothesis that they belong to the same distribution.
Before using the appropriate statistic, a specific analysis relative to the directional data has to be followed (Trauth 2010). The characteristics of the directional data are described by measures of central tendency and dispersion, which are similar to the statistical characterization of univariate datasets. Initially, we need to calculate the resultant or mean direction for the sets of angular data according to the relationships:
(B3)
(B4) and
(B5)
(B6)
The resultant directions of the data are given by:
(B7)
and
(B8)
206 Appendix B Statistical Testing and the mean resultant lengths are:
(B9)
and
(B10)
respectively.
The test statistic that we used for testing the similarity between the two mean directions is the F-statistic, and it is given by the relationship:
(B11)
where k is the concentration parameter, which can be obtained from tables using Rall,
R1 and R2 are the mean directions (resultants) of the two datasets, respectively, and
Rall is the resultant of the combined datasets. The calculated F-statistic is compared with the critical values from the standard F tables, and the two mean directions are not significantly different if the measured F is lower than the critical Fcr. As previously, we applied the aforementioned statistical test to our data with a level of significance 5%.
207
Reference list
Ando, M., Ishikawa, Y. & Yamazaki, F., 1983. Shear wave polarization anisotropy in
the upper mantle beneath Honshu, Japan, J. Geophys. Res., 88, 5850-5864.
Armijo, R., Meyer, B., King, G., Rigo, A. & Papanastassiou, D., 1996. Quaternary
evolution of the Corinth Rift and its implications for the Late Cenozoic evolution of
the Aegean. Geophys. J. Int., 126 (1), 11-53.
Aster, R., Shearer, P. & Berger, J., 1990. Quantitative measurements of shear-wave
polarizations at the Anza seismic network, southern California: implications for
shear-wave splitting and earthquake prediction, J. Geophys. Res., 95, 12449-12473.
Avallone, A., Briole, P., Agatza-Balodimou, A.M., Billiris, H., Charade, O., Mitsakaki,
C., Nercessian, A., Papazissi, K., Paradissis, D., & Veis, G., 2004. Analysis of eleven
years of deformation measured by GPS in the Corinth Rift Laboratory area.
Comptes Rendus Geosciences 336 (4-5), 301-311.
Barmin, M., Ritzwoller, M. & Levshin, A., 2001. A fast and reliable method for surface
wave tomography. Monitoring the Comprehensive Nuclear-Test-Ban Treaty:
Surface Waves. Springer, pp. 1351-1375.
Baud, ., Diraison, M., Geraud, Y., Reuschlé, T., Schmittbhul, J., Souque, C. & lace,
J., 2004. Circulation des fluides dans les systèmes de failles du Rift de Corinthe;
Données structurales, pétrophysiques de porosité et de géochimie; rapport GDR
"Corinth" (unpublished report, http://crllab.eu/).
Beckers, A., Hubert-Ferrari, A., Beck, C., Bodeux, S., Tripsanas, E., Sakellariou,
D. & De Batist, M., 2015. Active faulting at the western tip of the Gulf of Corinth,
Greece, from high-resolution seismic data. Mar. Geology, 360, 55-69.
209
Reference list
Bell, R. E., McNeill, L. C., Bull, J. M. & Henstock, T. J. 2008. Evolution of the offshore
western Gulf of Corinth, Geol. Soc. Am. Bull., 120 (1-2), 156-178.
Bell, R.E., McNeill, L.C., Bull, J.M., Henstock, T.J., Collier, R.E.L. & Leeder, M.R.,
2009. Fault architecture, basin structure and evolution of the Gulf of Corinth Rift,
central Greece. Basin Research 21 (6), 824-855.
Bensen, G., Ritzwoller, M., Barmin, M., Levshin, A., Lin, F., Moschetti, M., Shapiro,
N. M. & Yang, Y., 2007. Processing seismic ambient noise data to obtain reliable
broad-band surface wave dispersion measurements, Geophysical Journal
International, 169 (3), 1239-1260.
Bensen, G., Ritzwoller, M. & Shapiro, N., 2008. Broadband ambient noise surface
wave tomography across the united states, Journal of Geophysical Research: Solid
Earth (1978-2012), 113 (B5), 1.
Bernard, P., Chouliaras, G., Tzanis, A., Briole, P., Bouin, M., Tellez, J., Stavrakakis,
G. & Makropoulos, K., 1997. Seismic and electrical anisotropy in the Mornos delta,
Gulf of Corinth, Greece and its relationship with GPS strain measurements,
Geophys. Res. Lett., 24 (17), 2227-2230.
Bernard, P., Lyon-Caen, H., Briole, P., Deschamps, A., Pitilakis, K., Manakou, M.,
Boudin, F., Berge, C., Makropoulos, K., Diagourtas, D., Papadimitriou, P., Lemeille,
F., Patau, G., Billiris, H., Castarede, H., Charade, O., Nercessian, A., Avallone, A.,
Zahradnik, J., Sacks, S., & Linde, A., 2006. Seismicity, deformation and seismic
hazard in the western rift of Corinth: New insights from the Corinth Rift
Laboratory (CRL). Tectonophysics, 426 (1-2), p. 7-30.
Bonnefoy-Claudet, S., Cotton, F. & Bard, P.-Y., 2006. The nature of noise wave field
and its applications for site effects studies: A literature review. Earth-Sci. Rev., 79,
2050-227.
210
Reference list
Booth, D.C. and S. Crampin, S., 1985. Shear-wave polarizations on a curved wavefront
at an isotropic free-surface, Geophys. J. R. Astron. Soc., 83, 31-45.
Bouin, M., Téllez, J. & Bernard, ., 1996. Seismic anisotropy around the Gulf of
Corinth, Greece, deduced from three-component seismograms of local earthquakes
and its relationship with crustal strain, J. Geophys. Res., 101 (B3), 5797-5811.
Bourouis, S. & Cornet, F.H., 2009. Microseismic activity and fluid fault interactions:
some results from the Corinth Rift Laboratory (CRL), Greece. Geophysical Journal
International, 178 (1), 561-580.
Brenguier, F., Shapiro, N. M., Campillo, M., Nercessian, A. & Ferrazzini, V., 2007. 3-D
surface wave tomography of the Piton de la Fournaise volcano using seismic noise
correlations, Geophysical Research Letters, 34 (2), L2305.
Brenguier, F., Campillo, M., Hadziioannou, C., Shapiro, N. M., Nadeau, R. & Larose,
E., 2008a. Postseismic relaxation along the San Andreas Fault at Parkfield from
continuous seismological observations, Science, 321, 5895 1478-1481.
Brenguier, F., Shapiro, N. M., Campillo, M., Ferrazzini, V., Duputel, Z., Coutant, O. &
Nercessian, A., 2008b. Towards forecasting volcanic eruptions using seismic noise,
Nature Geoscience, 1, 126-130.
Briole, P., Rigo, A., Lyon-Caen, H., Ruegg, J.C., Papazissi, K., Mitsakaki, C.,
Balodimou, A., Veis, G., Hatzfeld, D. & Deschamps, A., 2000. Active deformation of
the Corinth rift, Greece: results from repeated Global Positioning System surveys
between 1990 and 1995. J. Geophys. Res., 105 (B11), 25605-25625.
Bouin, M., Téllez, J. & Bernard, P., 1996. Seismic anisotropy around the Gulf of
Corinth, Greece, deduced from three-component seismograms of local earthquakes
and its relationship with crustal strain, J. Geophys. Res., 101 (B3), 5797-5811.
211
Reference list
Campillo, M. & Paul, A., 2003. Long-range correlations in the diffuse seismic coda:
Science, 299, 547-549.
Chiarabba, C. & Moretti, M., 2006. An insight into the unrest phenomena at the
Campi Flegrei caldera from VP and VP/VS tomography. Terra Nova 18, 373-379.
Clarke, D., Zaccarelli, L., Shapiro, N.M. & Brenguier F., 2011. Assessment of
resolution and accuracy of the Moving Window Cross Spectral technique for
monitoring crustal temporal variations using ambient seismic noise, Geophys. J.
Int. 186 (2), 867-882.
Claerbout, J. F. 1968. Synthesis of a layered medium from its acoustic transmission
response: Geophysics, 33, 264-269.
Corinth Rift Laboratory team and RESIF Datacenter, 2013. CL - Corinth Rift
Laboratory Seismological Network (CRLNET); RESIF - Réseau Sismologique et
géodésique Français. https://doi.org/10.15778/RESIF.CL
Cornet, F. L. Doan, M.L., Moretti, I. & Borm, G., 2004. Drilling through the active
Aigion fault: The AIG10 well observatory, C.R. Geos., 336 (4-5), 395-406.
Crampin, S., 1981. A review of wave motion in anisotropic and cracked elastic-media,
Wave Motion, 3, 343-391.
Crampin, S., 1984. An introduction to wave propagation in anisotropic media:
Geophys. J. Roy. Astr. Sot., 76, 17-28.
Crampin, S., Booth, D.C., Evans, R., Peacock, S. & Fletcher, J.B., 1990. Changes in
shear wave splitting at Anza near the time of the North Palm Springs Earthquake,
J. Geophys. Res., 95, 11197-11212.
Crampin, S., Booth, D.C., Evans, R., Peacock, S., & Fletcher, J.B., 1991. Comment on
"Quantitative measurements of shear wave polarizations at the Anza seismic
network, Southern California: implications for shear wave splitting and earthquake
212
Reference list
prediction" by R.C. Aster, P.M. Shearer and J. Berger, J. Geophys. Res., 96, 6403-
6414.
Crampin, S. & Lovell, J.H., 1991. A decade of shear-wave splitting in the Earth’s
crust: what does it mean? what use can we make of it? and what should we do
next?. Geophys. J. Int., 107, 387-407.
Crampin, S., & Zatsepin, S.V., 1997. Modeling the compliance of crustal rock-II.
Response to temporal changes before earthquakes, Geophys. J. Int., 129, 495-506.
Crampin, S., 1999. Calculable fluid-rock interactions. J. Geol. Soc. 156, 501-514.
Crampin, S., Volti, T. & Stefansson, R., 1999. A successfully stress-forecast
earthquake, Geophys. J. Int., 138, F1-F5.
Crampin, S. & Chastin, S., 2003. A review of shear wave splitting in the crack-critical
crust, Geophys. J. Int., 155, 221-240.
Crampin, S. & Gao. Y., 2005. Comment on "Systematic analysis of shear-wave
splitting in the aftershock zone of the 1999 Chi–Chi, Taiwan, earthquake: shallow
crustal anisotropy and lack of precursory changes", by Liu, Teng, Ben-Zion, Bull.
Seismol. Soc. Am., 95, 354-360.
Crampin, S. & Peacock, S., 2005. A review of shear-wave splitting in the compliant
crack-critical anisotropic Earth. Wave Motion, 41, 59-77.
Crampin, S. & Gao, Y., 2006. A review of techniques for measuring shear wave
splitting above small earthquakes. Phys. Earth. Planet. Inter. 159, 1-14.
Crampin, S. & Peacock, S., 2008. A review of the current understanding of shear-wave
splitting and common fallacies in interpretation, Wave Motion, 45, 675-722.
Crampin, S. & Gao, Y., 2012. Plate-wide deformation before the Sumatra- Andaman
Earthquake, J. Asian Earth Sci., 46, 61-69.
Crampin, S., Gao, Y. & Bukits, J., 2015. A review of retrospective stress-forecasts of
213
Reference list
earthquakes and eruptions, Physics of the Earth and Planetary Interiors, 245, 76-
87, ISSN 0031-9201.
Curtis, A., Gerstoft, P., Sato, H., Snieder, R. & Wapenaar, K., 2006. Seismic
interferometry - turning noise into signal. The Leading Edge, 25, 1082-1092.
Derode, A., Larose, E., Campillo, M. & Fink, M., 2003. How to estimate the Green's
function of a heterogeneous medium between two passive sensors? Application to
acoustic waves: Applied Physics Letters, 83, 3054-3056.
Di Luccio, F., Ventura, G., Di Giovambattista, R., Piscini, A. & Cinti, F.R., 2010.
Normal faults and thrusts reactivated by deep fluids: The 6 April 2009 Mw 6.3
L'Aquila earthquake, central Italy, J. Geophys. Res., 115, B06315.
Dornsiepen, U.F., Manutsoglu, E. & Mertmann, D., 2001. Permian-Triassic
paleogeography of the external Hellenides, Palaeogeog. Palaeoclimat. Palaeoecol.,
172, 327-338.
Doutsos, T. & Kokkalas, S., 2001. Stress and deformation patterns in the Aegean
region, J. Struct. Geol., 23, 455-472.
Doutsos, T. & Poulimenos, G., 1992. Geometry and kinematics of active faults and
their seismotectonic significance in the western Corinth-Patras rift (Greece).
Journal of Structural Geology, 14 (6), 689–699.
Doutsos, T., Koukouvelas, I. & Xypolias, P., 2006. A new orogenic model for the
External Hellenides. In: Robertson, A.H.F., Mountrakis, D. (eds) Tectonic evolution
of the Eastern Mediterranean regions. Geological Society, London, Special
Publications, 260, 507-520.
Edrun, B., Lebedev, S., Meier, T., Tirel, C. & Friederich, W., 2011. Complex layered
deformation within the Aegean crust and mantle revealed by seismic anisotropy.
Nature Geoscience, 4, 203-207.
214
Reference list
Ekström, G. 2001. Time domain analysis of Earth's long period background seismic
radiation, J. Geophys. Res., 106, 26, 483-26, 493.
Evangelidis, C. P., Liang, W.-T., Melis, N. S. & Konstantinou, K. I., 2011. Shear wave
anisotropy beneath the Aegean inferred from SKS splitting observations. J.
Geophys. Res., 116, B04314.
Evangelidis, C. P. & Melis, N.S., 2012. Ambient Noise Levels in Greece as Recorded at
the Hellenic Unified Seismic Network. Bull. Seism. Soc. Am., 102, 6, 2507-2517,
Evangelidis, C., Daskalakis, E. & Tsogka, C., 2016. Monitoring seismic velocity
changes caused by the 2014 Northern Aegean earthquake using continuous ambient
noise records. Geophysical Research Abstracts, Vol. 18, EGU2016-12454.
Fernandez-Viejo, Clowes, G.R.M. & Welford, J.K., 2005. Constraints on the
composition of the crust and uppermost mantle in northwestern Canada: VP/VS
variations along Lithoprobe's SNorCLE. Canadian Journal of Earth Sciences, 42,
1205-1222.
Flotté, N., Sorel, D., Müller, C. & Tensi, J., 2005. Along strike changes in the
structural evolution over a brittle detachment fault: example of the Pleistocene
Corinth-Patras rift (Greece). Tectonophysics, 403, 77-94.
Ford, M., Rohais, S., Williams, E.A., Bourlange, S., Jousselin, D., Backert, N. &
Malartre, F., 2013. Tectono-sedimentary evolution of the western Corinth Rift
(Central Greece). Basin Research, 25 (1), 3-25.
Fukao, Y., Nishida, K. & Kobayashi, N., 2010. Seafloor topography, ocean infragravity
waves, and background Love and Rayleigh waves, J. Geophys. Res., 115, B04302.
Ganas, A., Chousianitis, K., Batsi, E., Kolligri, M., Agalos, A., Chouliaras, G. &
Makropoulos, K., 2013. The January 2010 Efpalion earthquakes (Gulf of Corinth,
215
Reference list
Central Greece): earthquake interactions and blind normal faulting, J. Seismology,
17, 2, 465-484.
Gao, Y. & Crampin, S., 2004. Observations of stress relaxation before earthquakes,
Geophys. J. Int., 157, 578-582.
Garnier, J. & Papanicolaou, J., 2009. Passive sensor imaging using cross correlations
of noisy signals in a scattering medium, SIAM J. Imaging Sci., 2, 396-437.
Gautier, S., Latorre, D., Virieux, J., Deschamps, A., Skarpelos, C., Sotiriou, A.,
Serpetsidaki, A. & Tselentis, A., 2006. A new passive tomography of the Aigion area
(Gulf of Corinth, Greece) from the 2002 dataset, Pure App. Geoph., 163 (2), 431-453.
Gerst, A. & Savage, M., 2004. Seismic anisotropy beneath Ruapehu volcano: a possible
eruption forecasting tool, Science 306, 1543–1547.
Ghisetti, F. & Vezzani, L., 2005. Inherited structural controls on normal fault
architecture in the Gulf of Corinth (Greece). Tectonics, 24.
Giannopoulos, D., Sokos, E., Konstantinou, K.I., Lois, A. & Tselentis, G., 2013.
Temporal variation of shear-wave splitting parameters before and after the 2008
Movri Mountain earthquake in northwest Peloponnese (Greece). Annals Of
Geophysics, 55 (5).
Gibbons, J.D., 1971. Nonparametric Statistical Inference, McGraw-Hill, New York.
Godano, M., Deschamps, A., Lambotte, S., Lyon-Caen, H., Bernard P. & Pacchiani, F.,
2014. Focal mechanisms of earthquake multiplets in the western part of the Corinth
Rift (Greece): influence of the velocity model and constraints on the geometry of the
active faults, Geophys. J. Int., doi: 10.1093/gji/ggu059.
Gunasekera, R.C., Foulger G.R. & Julian, B.R., 2003. Reservoir depletion at the
Geysers geothermal area, California, shown by four-dimensional tomography. J.
Geophys. Res. 108, 2134.
216
Reference list
Hannemann, K., Papazachos, C., Ohrnberger, M., Savvaidis, A., Anthymidis, M. &
Lontsi, A.M., 2014. Three-dimensional shallow structure from high-frequency
ambient noise tomography: new results for the Mygdonia basin-Euroseistest area,
northern Greece. J. Geophys. Res. Solid Earth, 119.
Hasselmann, K., 1963. A statistical analysis of the generation of microseisms, Rev.
Geophys., 1, 177-209.
Hatzfeld, D, Kementzetzidou, D., Karakostas, V., Ziazia, M., Nothard, S., Diagourtas,
D., Deschamps, A. Karakaisis, G., Papadimitriou, P., Scordilis, M., Smith, R.,
Voulgaris, N., Kiratzi, S., Makropoulos, K., Bouin, M.-P. & Bernard, P., 1996. The
Galaxidi earthquake of 18 November, 1992: a possible asperity within the normal
fault system of the Gulf of Corinth (Greece). Bull Seismol. Soc. Am., 86, 1987-1991.
Hatzfeld, D., Martinod, J., Bastet, G., & Gautier, P., 1997. An analog experiment for
the Aegean to describe the contribution of the gravitational potential energy. J.
Geophys. Res., 102, 649-659.
Hatzfeld, D., Karakostas, V., Ziazia, M., Kassaras, I., Papadimitriou, E., Makropoulos,
K., Voulgaris, N. & Papaioannou, C., 2000. Microseismicity and faulting geometry
in the Gulf of Corinth (Greece), Geophys. J. Int. 141, 438-456.
Hatzfeld, D., Karagianni, E., Kassaras, I., Kiratzi, A., Louvari, E., Lyon-Caen, H.,
Makropoulos, K., Papadimitriou, P., Bock, G. & and Priestley K., 2001. Shear wave
anisotropy in the upper mantle beneath the Aegean related to internal deformation.
J. Geophys. Res., 106 (B12), 30737-30753.
Herrmann, R. & Ammon, C., 2004. Surface waves, receiver functions and crustal
structure, Computer Programs in Seismology.
Hemelsdael, R. & Ford, M., 2014. Relay zone evolution: A history of repeated fault
propagation and linkage, central Corinth rift, Greece, Basin Research, 27.
217
Reference list
Jackson, J.A., 1994. Active tectonics of the Aegean region. Annu. Rev. Earth Planet.
Sci. Lett., 22, 239-271.
Jackson, J.A., Gagnepain, J., Houseman, G., King, G., Papadimitriou, P., Soufleris, P.
& Virieux, J., 1982. Seismicity, normal faulting and the geomorphological
development of the Gulf of Corinth (Greece): the Corinth earthquakes of February
and March 1981, Earth Planet. Sci. Lett. 57, 377-397.
Kaviris, G., Papadimitriou, P. & Makropoulos, K., 2008. An Overview of Anisotropy
Studies in Central Greece using recordings around the Gulf of Corinth (Greece) and
aftershocks of the 1999 Athens Earthquake, Short Papers, 31st ESC General
Assembly, Hersonissos, Crete, Greece, 215-223.
Kaviris, G., Papadimitriou, P. & Makropoulos, K., 2010. Anisotropy Study of the
February 4th 2008 Swarm in NW Peloponnesus (Greece), Bull. Geol. Soc. Greece,
XLIII/4, 2084-2092.
Kaviris, G., Papadimitriou, P., Kravvariti, Ph., Kapetanidis, V., Karakonstantis, A.,
Voulgaris, N. &Makropoulos K., 2015. A detailed seismic anisotropy study during
the 2011–2012 unrest period in the Santorini Volcanic Complex. Physics of the
Earth and Planetary Interiors, 238, 5-88.
Karakostas, B., Papadimitriou, E., Hatzfeld, D., Makaris, D., Makropoulos, K.,
Diagourtas, D., Papaioannou, Ch., Stavrakakis, G., Drakopoulos, J. & Papazachos,
B., 1994. The aftershock sequence of July 14, 1993 (MS=5.4) Patras earthquake. In:
Seventh congress of geological society, Greece, XXX/5:167-174.
Karakostas, V., Karagianni, E. & Paradisopoulou, P., 2012. Space–time analysis,
faulting and triggering of the 2010 earthquake doublet in western Corinth Gulf. Nat
Hazard 63, 1181-1202.
Kedar, S. & Webb, F.H., 2005. The Ocean's seismic hum, Science, 307, 682-683.
218
Reference list
King, G.C.P., Ouyang, Z.X., Papadimitriou P., Deschamps, A., Gagnepain, J.,
Houseman, G., Jackson, J.A., Soufleris, C. & Virieux, J., 1985. The evolution of the
Gulf of Corinth (Greece): an aftershock study of the 1981 earthquakes, Geophys. J.
R. Astr. Soc., 80,677-683.
Kokkalas, S., Xypolias, P., Koukouvelas, I.K. & Doutsos, T., 2006. Post-collisional
contractional and extensional deformation in the Aegean region. In: Dilek, Y.,
Pavlides, S. (Eds.), Post-Collisional Tectonics and Magmatism in the Eastern
Mediterranean Region. Special Papers, 409. Geological Society of America, 97-123.
Konstantinou, K.I., Evangelidis, C.P., Liang, W.-T., Melis, N.S. & Kalogeras, I., 2013.
Seismicity, VP/VS and shear wave anisotropy variations during the 2011 unrest at
Santorini caldera, southern Aegean, J. Volcanology & Geothermal Res., 267, 57-67.
Koukouvelas, I.K. & Papoulis, D., 2009. Fluid involvement in the Helike normal fault,
Gulf of Corinth Greece, J. Struct. Geol. 31, 237-250.
Kreemer, C., Chamo-Rooke, N. & Le Pichon X., 2004. Constraints on the evolution and
vertical coherency of deformation in the northern Aegean from a comparison of
geodetic, geologic and seismologic data, Earth Planet. Sci. Lett., 225, 329-346.
Kuo, B.Y., Chen, C.C. & Shin, T.C., 1994. Split S waveforms observed in northern
Taiwan: implications for crustal anisotropy, Geophys. Res. Lett. 21, 1491-1494.
Lambotte S., Lyon-Caen, H., Benard, P., Deschamps, A., Patau, G., Nercessian, A.,
Pacchiani, F., Bourouis, S., Drilleau M. & Adamova, P., 2014. Reassessment of the
rifting process in the Western Corinth Rift from relocated seismicity, Geophys. J.
Int., doi: 10.1093/gji/ggu096.
Latorre, D., Virieux, J., Monfret, T., Monteiller, V., Vanorio, T., Got, J.L. & Lyon-
Caen, H., 2004. A new seismic tomography of Aigion area (Gulf of Corinth, Greece)
from the 1991 data set. Geophysical Journal International 159 (3), 1013-1031.
219
Reference list
Leeder, M. R., Mark, D.F., Gawthorpe, R.L., Kranis, H., Loveless, S., Pedentchouk, N.,
Skourtsos, E., Turner, J., Andrews, J.E. & Stamatakis, M., 2012. A “Great
Deepening”: Chronology of rift climax, Corinth rift, Greece, Geology, 40 (11).
Le Pichon, X. & Angelier, J., 1979. The Hellenic arc and trench system: a key to the
neotectonic evolution of the eastern Mediterranean area. Tectonophysics, 60, 1-42.
Le Pichon, X., Chamot-Rooke, N., Lallemant, S., Noomen, R. & Veis, G., 1995.
Geodetic determination of the kinematics of the central Greece with respect to
Europe: Implication for eastern Mediterranean tectonics, J. Geophys. Res., 100, 12
675-12690.
Lemeille, F., Chatoupis, F., Foumelis, M., Rettenmaier, D., Unkel, I., Micarelli, L.,
Moretti, I., Bourdillon, C., Guermet, C. & Müller, C., 2004. Recent syn-rift deposits
in the hangingwall of the Aigion Fault (Gulf of Corinth, Greece), C. R. Geoscience,
336, 425-434.
Liu, Y., Teng, T.-L. & Ben-Zion, Y., 2004. Systematic analysis of shear-wave splitting
in the aftershock zone of the 1999 Chi–Chi, Taiwan, earthquake: shallow crustal
anisotropy and lack of precursory changes, Bull. Seismol. Soc. Am., 94, 2330-2347.
Liu, Y., Ben-Zion, Y. & Teng, T.-L., 2005. Reply to "Comments on 'Systematic analysis
of shear-wave splitting in the aftershock zone of the 1999 Chi-Chi, Taiwan,
Earthquake: shallow crustal anisotropy and lack of precursory changes' by Y. Liu,
T.-L. Teng, Y. Ben-Zion" by S. Crampin, Y. Gao, Bull. Seism. Soc. Am., 95, 361-366.
Longuet-Higgins, M.S., 1950. A theory of the origin of microseisms, Phil. Trans. Roy.
Soc., 243 (857), 1-35.
Levshin, A., Yanovskaya, T., Lander, A., Bukchin, B., Barmin, M., Ratnikova, L. & Its
E., 1989. Seismic Surface Waves in a Laterally Inhomogeneous Earth, Dordrecht.
220
Reference list
Lykousis, V., Sakellariou, D., Moretti, I. & Kaberi H., 2007. Late Quaternary basin
evolution of the Gulf of Corinth: Sequence stratigraphy, sedimentation, fault–slip
and subsidence rates, Tectonophysics, 440 (1-4), 29-51.
Malcolm, A. E., Scales, J. A. & van Tiggelen, B. A., 2004. Extracting the Green
function from diffuse, equipartitioned waves: Physical Review E, 70, 015601-1-
015601-4.
McClusky, S. et al., 2000. Global Positioning System constraints on the plate
kinematics and dynamics in the eastern Mediterranean and Caucasus. J. Geophys.
Res., 105, 5695-5719.
McNeill, L.C., Cotterill, C.J., Henstock, T.J., Bull, J.M., Stefatos, A., Collier, R.E.L.,
Papatheoderou, G., Ferentinos, G., Hicks, S.E., 2005. Active faulting within the
offshore western Gulf of Corinth, Greece: implications for models of continental rift
deformation. Geology, 33 (4), 241.
Micarelli, L., Moretti, I. & Daniel, J.M., 2003. Structural properties of rift-related
normal faults: the case study of the Gulf of Corinth, Greece. J. Geodyn. 36, 275-303.
Mordret, A., Landès, M., Shapiro, N.M., Singh, S., Roux, P. & Barkved, O., 2013. Near-
surface study at the Valhall oil field from ambient noise surface wave tomography.
Geophys. J. Int. 193 (3), 1627-1643.
Mordret, A., Landès, M., Shapiro, N.M., Singh, S. & Roux, ., 2014. Ambient noise
surface wave tomography to determine the shallow shear velocity structure at
Valhall: depth inversion with a Neighbourhood Algorithm. Geophys. J. Int. 198 (3),
1514-1525.
Mordret, A., Rivet, D., Landès, M. & Shapiro, N.M., 2015. 3-D shear-velocity
anisotropic model of iton de la Fournaise volcano (la Réunion island) from ambient
seismic noise. J. Geophys. Res. Solid Earth 120 (1), 406-427.
221
Reference list
Moschetti, M.P., Ritzwoller, M.H. $ Shapiro, N.M., 2007. Surface wave tomography of
the western United States from ambient seismic noise: Rayleigh wave group
velocity maps. Geochem. Geophys. Geosyst. 8 (Q08010).
Munson, C.G., Thurber, C.H., Li, Y. & Okubo, P.G., 1995. Crustal shear wave
anisotropy in southern Hawaii: spatial and temporal analysis, J. Geophys. Res.,
100, 20367-20377.
Naville, C., Serbutoviez, S., Moretti, I., Daniel, J.M., Throo, A., Girard, F., Sotiriou, A.,
Tselentis, A., Skarpzelos, C., Brunet, C. & Cornet, F.H., 2004. Pre-drill surface
seismic in vicinity of AIG10 well and post drill VSP, C. R. Geoscience, 336, 407-414.
Nishida, K., Kawakatsu, H., Fukao, Y. & Obara, K., 2008. Background Love and
Rayleigh waves simultaneously generated at the Pacific Ocean floors, Geophys. Res.
Lett., 35, L16307.
Nixon, C. W., Mcneill, L. C., Bull, J. M., Bell, R. E., Gawthorpe, R. L., Henstock, T. J.,
Christodoulou, D., Ford, M., Taylor, B., Sakellariou, D., Ferentinos, G.,
Papatheodorou, G., Leeder, M., R., Collier, R. E. Li., Goodliffe, A., M., Sachpazi, M.
& Kranis, H., 2016. Rapid spatiotemporal variations in rift structure during
development of the Corinth Rift, central Greece. Tectonics, 35 (5), 1225-1248.
Novotny O., Zahardnik, J. & Tselentis G-A., 2000. NW Turkey Earthquakes and the
crustal structure inferred from surface waves observed in the Corinth Gulf,
Greece. Physics Of The Earth, 117, 23, 456-473. Also published at Bulletin of the
Seismological Society of America, 91, 4, 875-879.
Nur, A., 1972. Dilitancy, pore fluids and premonitory variations of ts/tp travel times,
Bull. Seism. Soc. Am., 62, 1217-1222.
222
Reference list
Obermann, A., Lupi, M., Mordret, A., Jakobsdóttir, S. & Miller, S., 2016. 3D-ambient
noise Rayleigh wave tomography of Snæfellsjökull volcano, Iceland, Journal of
Volcanology and Geothermal Research, 317, 42-52.
Pacchiani, F. & Lyon-Caen, H., 2010. Geometry and spatio-temporal evolution of the
2001 Agios Ioanis earthquake swarm (Corinth Rift, Greece), Geophys. J. Int., v. 180
(1), pp. 59-72.
Palyvos, N., Lemeille, F., Sorel, D., Pantosti, D. & Pavlopoulos, K., 2008. Geomorphic
and biological indicators of paleoseismicity and Holocene uplift rate at a coastal
normal fault footwall (western Corinth Gulf, Greece). Geomorphology 96 (1-2), 16-
38.
Papadimitriou, P., Kaviris, G. & Makropoulos, K., 1999. Evidence of shear-wave
splitting in the eastern Corinthian Gulf (Greece), Phys, Earth Plan. Inter., 114 (1-2),
3-13.
Papadimitriou, P., Kaviris, G., Karakonstantis, A. & Makropoulos, K., 2010. The
Cornet seismological network: 10 years of operation, recorded seismicity and
significant applications, Annal Geol Des Pays Hell., 45, 193-208.
Papadimitriou, P., Kapetanidis, V., Karakonstantis, A., Kaviris, G., Voulgaris, N. &
Makropoulos, K., 2015. The Santorini Volcanic Complex: A detailed multi-
parameter seismological approach with emphasis on the 2011-2012 unrest period, J.
Geodyn., 85, 32-57.
Papadopoulos, G. A., 2000. A new tsunami catalogue of the Corinth Rift: 373 B.C.–
A.D. 2000. In: G. A. Papadopoulos (ed.), Historical Earthquakes and Tsunamis in
the Corinth Rift, Central Greece, Inst. Geodynamics, Natl. Observatory Athens, pp.
121-126.
223
Reference list
Papanikolaou, D., Chronis, G., Lykousis, V., Sakellariou, D., & Papoulia, I., 1997.
Submarine neotectonic structure of W. Korinthiakos Gulf and geodynamic
phenomena of the Egion earthquake. Proceedings of 5th Hellenic Congress of
Oceanography & Fisheries, Volume I. Kavala, Greece, pp. 415-418.
Paul, A. Campillo, M., Margerin, L., Larose, E. & Derode A., 2005. Empirical synthesis
of time-asymmetrical Green functions from the correlation of coda waves, J.
Geophys. Res., 110, B08302.
Peacock, S., Crampin, S., Booth, D.C. & Fletcher, J.B., 1988. Shear-wave splitting in
the Anza seismic gap, southern California: temporal variations as possible
precursors, J. Geophys. Res., 93, 3339-3356.
Peng, Z. & Ben-Zion, Y., 2005. Spatiotemporal variations of crustal anisotropy from
similar events in aftershocks of the 1999 M 7.4 Izmit and M 7.1 Düzce, Turkey,
earthquake sequences, Geophys. J. Int., 160, 1027-1043.
Pik, R. & Marty, B., 2009. Helium isotopic signature of modern and fossil fluids
associated with the Corinth rift fault zone (Greece): Implication for fault
connectivity in the lower crust, Chemical Geol., 266, 67-75.
Ratdomopurbo, A. & Poupinet G., 1995. Monitoring a temporal change of seismic
velocity in a volcano: Application to the 1992 eruption of Mt. Merapi (Indonesia),
Geophys. Res. Lett. 22, no. 7, 775-778.
Rettenmaier, D., Giurgea, V., Hötzl, H. & Förster, A., 2004. The AIG10 drilling project
(Aigion, Greece): interpretation of the litholog in the context of regional geology and
tectonics, C. R. Geoscience, 336, 415-423.
Rhie, J. & Romanowicz B., 2004. Excitation of Earth's continuous free oscillations by
atmosphere-ocean-seafloor, Nature, 431, 552-554.
224
Reference list
Rhie, J. & Romanowicz, B., 2006. A study of the relation between ocean storms and
the Earth's hum, Geochem. Geophys. Geosyst., 7, Q10004.
Rial, J.A., Elkibbi, M. & Yang, M., 2005. Shear-wave splitting as a tool for the
characterization of geothermal fractured reservoirs: lessons learned, Geothermics,
Volume 34, 3, 365-385, ISSN 0375-6505.
Rigo, A., Lyon-Caen, H., Armijo, R., Deschamps, A., Hatzfeld, D., Makropoulos, K.,
Papadimitriou, P. & Kassaras, I., 1996. A microseismic study in the western part of
the Gulf of Corinth (Greece): implications for large-scale normal faulting
mechanisms. Geophys. J. Int. 126 (3), 663-688.
Roberts, G. P., Houghton, S. L., Underwood, C., Papanikolaou, I., Cowie, P.A., van
Calsteren, P., Wigley, T., Cooper, F.J. & McArthur J.M. 2009. Localization of
Quaternary slip rates in an active rift in 105 years: An example from central Greece
constrained by 234 U- 230 Th coral dates from uplifted paleoshorelines, J. Geophys.
Res., 114 (B10), B10406.
Roux, P., 2009. Passive seismic imaging with directive ambient noise: application to
surface waves and the San Andreas Fault in Parkfield, CA, Geophysical Journal
International, 179 (1), 367-373. 54.
Roux, P., Sabra, K. G., Kuperman, W. & Roux, A., 2005. Ambient noise cross
correlation in free space: Theoretical approach, The Journal of the Acoustical
Society of America, 117, 79, 18.
Sabra, K., Gerstoft, P., Roux, P., Kuperman, W. & Fehler, M., 2005. Surface wave
tomography from microseisms in Southern California, Geophysical Research Letter,
32, L14311.
225
Reference list
Sachpazi M., Christophe Clement, C., Laigle, M., Hirn H. & Roussos, N., 2003. Rift
structure, evolution, and earthquakes in the Gulf of Corinth, from reflection seismic
images. Earth and Planetary Science Letters 216, 243-257.
Sachpazi, M., Hirn, A., Clément, C., Haslinger, F., Laigle, M., Kissling, E., Charvis, P.,
Hello Y., Lépine, J-C., Sapin M. & J. Ansorge J., 2000. Western Hellenic subduction
and Cephalonia Transform: Local earthquakes and plate transport and strain,
Tectonophysics, 319, 301-319.
Sachpazi, M. & Hirn, A., 1991. Shear-wave anisotropy across the geothermal field of
Milos, Aegean volcanic arc Geophys. J. Int., 107, 673-685.
Sambridge, M., 1999a. Geophysical inversion with a neighborhood algorithm. I.
Searching a parameter space. Geophys. J. Int. 138 (2), 479-494.
Sambridge, M., 1999b. Geophysical inversion with a Neighbourhood Algorithm -II.
Appraising the ensemble. Geophys. J. Int.138, 727-746.
Schmid, C., Van der Lee, S. & Giardini D., 2004. Delay times and shear wave splitting
in the Mediterranean region, Geophys. J. Int., 159, 275-290.
Sens-Schönfelder, C. & Wegler, U., 2006. Passive image interferometry and seasonal
variations of seismic velocities at Merapi volcano, Indonesia, Geophysical Research
Letters, 33 (21), L21302. 22, 79, 154.
Seher, T. & Main I.G., 2004. A statistical evaluation of a "stress-forecast" earthquake,
Geophys. J. Int., 157, 187-193.
Shapiro, N. M., & Campillo, M., 2004. Emergence of broadband Rayleigh waves from
correlations of the ambient seismic noise: Geophysical Research Letters, 31, L07614-
1-L07614-4.
Shapiro, N. M., Campillo, M., Stehly, L. & Ritzwoller, M.H., 2005. High resolution
surface-wave tomography from ambient seismic noise: Science, 307, 1615-1618.
226
Reference list
Shapiro, S.A., Huenges, E. & Borm, G., 1997. Estimating the crust permeability from
fluid‐injection‐induced seismic emission at the KTB site, Geophys. J. Int., 131,
F15-F18.
Shapiro, S. A., Patzig, R., Rothert, E.& Rindschwentner, J., 2003. Triggering of
seismicity by pore-pressure perturbations: Permeability‐related signatures of the
phenomenon, Pure Appl. Geophys., 160, 1051–1066.
Shearer, P. M., 1997. Improving local earthquake locations using the L1 norm
and waveform cross correlation: application to the Whittier Narrows, California,
aftershock sequence, J. Geophys. Res. 102, 8269-8283.
Shearer, P. M., Hauksson, E. & Lin, G. Q., 2005. Southern California hypocenter
relocation with waveform cross-correlation, part 2: Results using source-specific
station terms and cluster analysis, Bull. Seismol. Soc. Am. 95, 904-915.
Sheriff R.E. & Geldart, L.P., 1995. Exploration seismology, University Press,
Cambridge.
Snieder, R., 2004. Extracting the Green’s function from the correlation of coda waves:
A derivation based on stationary phase, Physical Review E, 69(4), 46610. 18, 48.
Stutzmann, E., Roult, G. & Astiz, L., 2000. Geoscope station noise level, Bull. Seism.
Soc. Am., 90, 690-701.
Sokos E, Zahradník, J., Kiratzi, A., Janský, J., Gallovič, F., Novotny, O., Kostelecký,
J., Serpetsidaki, A. & Tselentis, G-A., 2012. The January 2010 Efpalio earthquake
sequence in the western Corinth Gulf (Greece). Tectonophysics 530-531:299-309.
Sorel, D., 2000. A Pliocene and still-active detachment fault and the origin of the
Corinth-Patras rift, Greece, Geology, 28, 83-86.
Talwani, P., Chen, L. & Gahalaut, K., 2007. Seismogenic permeability, ks, J. Geophys.
Res., 112, B07309.
227
Reference list
Tanimoto, T., 2005. The oceanic excitation hypothesis for the continuous oscillations of
the Earth, Geophys. J. Int., 160, 276-288.
Tanimoto, T. & Um J., 1999. Cause of continuous oscillations of the Earth, J. Geophys.
Res., 104, 28, 723-28, 739.
Taylor, B., Weiss, J. R., Goodliffe, A. M., Sachpazi, M., Laigle, M. & Hirn, A., 2011.
The structures, stratigraphy and evolution of the Gulf of Corinth rift, Greece,
Geophys. J. Int., 185 (3), 1189-1219.
Tselentis, G., Melis, N. & Sokos, E., 1994. The Patras (July 14, 1993; Ms = 5.4)
earthquake sequence. In: Seventh congress geological society, Greece XXX/5:159-
166.
Tselentis, G., Melis, N., Sokos, E. & Papatsimpa, K., 1996. The Egion June 15, 1995
(6.2 Ml) earthquake, Western Greece, Pure App. Geophys. 147, 83-98.
Trauth, S.N., 2010. Matlab® Recipes for Earth Sciences, 3rd Edition, Springer-Verlag,
Berlin/Heidelberg.
Underhill, J., 1989. Late Cenozoic deformation of the Hellenide foreland, western
Greece. Geological Society of America Bulletin, 101. 613-634. van Tiggelen, B. A. 2003. Green function retrieval and time reversal in a disordered
world: Physical Review Letters, 91, 243904-1-243904-4.
Unglert K., Savage, M. K., Fournier, N., Ohkura, T. & Abe, Y., 2011. Shear wave
splitting, VP/VS and GPS during a time of enhanced activity at Aso caldera, Kyushu,
J. Geophys. Res., 116, B11203.
Valkaniotis, S., 2009. Correlation between neotectonic structures and seismicity in the
broader area of Gulf of Corinth (Central Greece). PhD Thesis, Aristotle University
of Thessaloniki, Thessaloniki.
228
Reference list
Volti, T. & Crampin, S., 2003. A four-year study of shear-wave splitting in Iceland: 1.
Background and preliminary analysis. In New insights into structural
interpretation and modeling. (ed. NIEUWLAND, D.A.), Geol. Soc. Lond., Spec.
Publ., 212, 117-133.
Vött, A., 2007. Relative sea level changes and regional tectonic evolution of seven
coastal areas in NW Greece since the mid-Holocene, Quaternary Science Reviews,
26, 7-8, 894-919, ISSN 0277-3791.
Wapenaar, K., 2003. Synthesis of an inhomogeneous medium from its acoustic
transmission response. Geophysics, 68, 1756-1759.
Wapenaar, K., Draganov, D., Thorbecke, J. & Fokkema, J., 2002. Theory of acoustic
daylight imaging revisited: 72nd Annual International Meeting, SEG, Expanded
Abstracts, 2269-2272.
Wapenaar, K., Thorbecke, J. & D. Draganov, D., 2004. Relations between reflection
and transmission responses of three-dimensional inhomogeneous media:
Geophysical Journal International, 156, 179-194.
Wapenaar K. & Fokkema, J., 2006. Green's function representations for seismic
interferometry, Geophysics, 71, SI33-SI46.
Weaver, R. L., & Lobkis, O.I., 2001. Ultrasonics without a source: Thermal fluctuation
correlations at MHz frequencies: Physical Review Letters, 87, 134301-1-134301-4.
Weaver, R. L., & Lobkis, O.I., 2002. On the emergence of the Green's function in the
correlations of a diffuse field: pulse-echo using thermal phonons: Ultrasonics, 40,
435-439.
Wells, D.L. & Coppersmith, K.J., 1994. New Empirical Relationships among
Magnitude, Rupture Length, Rupture Width, Rupture Area and Surface
Displacement, Bulletin of the Seismological Society of America, 84, 4, 974-1002.
229
Reference list
Xypolias, P. & Doutsos, T., 2000. Kinematics of rock flow in a crustal scale shear zone:
implication for the orogenic evolution of the SW Hellenides, Geological Magazine,
137, 81-96.
Xypolias, P. & Koukouvelas, I., 2001. Kinematic vorticity and strain rate patterns
associated with ductile extrusion in the Chelmos Shear Zone (External Hellenides,
Greece), Tectonophysics, 338, 59-77.
Yang, Y., Ritzwoller, M. H., Lin, F.-C., Moschetti, M. & Shapiro, N. M., 2008.
Structure of the crust and uppermost mantle beneath the western United States
revealed by ambient noise and earthquake tomography, Journal of Geophysical
Research : Solid Earth, 113 (B12). 48, 110, 129, 154.
Zatsepin, S.V. & Crampin, S, 1997. Modeling the compliance of crustal rock-I.
Response of shear-wave splitting to differential stress, Geophys. J. Int., 129, 477-
494.
Zygouri, V., Verrios, S. Kokkalas, S., Xypolias, P. & Koukouvelas, I., 2008. Scaling
properties within the Gulf of Corinth, Greece; comparison between offshore and
onshore active faults. Tectonophysics, 453, 193-210.
230