University Pierre and Marie Curie – Paris VI, France, Doctoral School “Geosciences and Natural Resources” (ED 398) and Bulgarian Academy of Sciences, Sofia, Bulgaria, National Institute of Geophysics, Geodesy and Geography Crustal deformations of shallow earthquakes in the Eastern Mediterranean studied by radar interferometry and seismology Maya ILIEVA, Meng ABSTRACT of dissertation for the degree of Doctor Paris 2011 The dissertation is prepared in the Laboratory of Geology, UMR CNRS 8538, École Normale Supérieure (ENS), Paris, France and the Department of Geodesy, National Institute of Geophysics, Geodesy and Geography – Bulgarian Academy of Sciences (NIGGG-BAS), Sofia, Bulgaria, in the framework of a bilateral doctorate between NIGGG-BAS and University Pierre and Marie Curie (UPMC) – Paris VI. The PhD student is started her work in a regular form of the doctorate in the year of 2005 in NIGGG-BAS and in 2007 in UPMC. The dissertation is considered on a meeting of the Department of Geodesy at NIGGGBAS which took place on 29.06.2010, and in December 2010 is considered and passed for defense from the “Science for the Earth and science for the Universe” scientific commission at the University Pierre and Marie Curie, Doctoral School “Geosciences and Natural Resources” (ED398). The Dissertation consists of 179 pages, developed in six parts with 152 figures, 24 tables and four appendixes, one in Chapter II and three in Chapter III, respectively. It is applied also a list of abbreviations. The list of the references contains 145 titles and 27 internet sites. The dissertation is presented for defence on 10 Feb 2011 from 14:00 in hall E314 in Laboratory of Geology, ENS, Paris, France to a jury composed of: Prof. Bertrand MEYER (UPMC, Paris) Dr. CR CNRS Alexis RIGO (OMP, Toulouse) Assoc. Prof. Mourad BEZZEGHOUD (CGE, Évora) Assoc. Prof. Bénédicte FRUNEAU (IFG, Marne-la-Vallée) Assist. Prof. Efthimios SOKOS (PSL, Patras) Assoc. Prof. Slaveyko GOSPODINOV (UACEG, Sofia) Dr. DR CNRS Pierre BRIOLE (ENS, Paris) Assoc. Prof. Dimitar DIMITROV (NIGGG, Sofia) Dr. DR CNRS Hélène LYON-CAEN (ENS, Paris) President of the jury Reporter Reporter Examiner Examiner Examiner Co-director of thesis Co-director of thesis Invited University Pierre and Marie Curie – Paris VI, France, Doctoral School “Geosciences and Natural Resources” (ED 398) Speciality: Geosciences and Bulgarian Academy of Sciences, Sofia, Bulgaria, National Institute of Geophysics, Geodesy and Geography Speciality: Surveying, Geodesy and Applied Geodesy (cipher 02.16.01) Crustal deformations of shallow earthquakes in the Eastern Mediterranean studied by radar interferometry and seismology Maya ILIEVA, Meng ABSTRACT of dissertation for the degree of Doctor Co-directors of the thesis: Assoc. Prof. Dr. Dimitar Dimitrov and Dr. Pierre Briole Paris 2011 Introduction 2 Introduction Methods The geodetic techniques for co-seismic surveying of crustal deformation are divided into two main groups (Feigl 2002): ground-based or classical and space techniques. The first group includes triangulation, trilateration and first-order spirit levelling. While the first two methods are applicable to measure horizontal displacement, the levelling is more suitable for detecting vertical displacements and especially those which are larger than about a centimetre. The most frequent usage of spirit levelling is in the cases of normal faulting caused of earthquakes with magnitude of 6 and higher. On the other hand, the usage of triangulation and trilateration is limited by distance range as well as by the necessity of clear line of sight between measured points. Most of the triangulating sites are situated on mountain tops which are very hard to be approached. Also, the measurements require team of several members, a lot of time, efforts and expenses. Nevertheless, all these data acquired with the classical techniques are the only available “historical” metric information with high accuracy and they are very useful for completing bigger time span set of data. The second group of techniques consists of VLBI, SLR, GNSS, SAR interferometry, etc. The very-long baseline interferometry (VLBI) provides sub-millimetre precision and is probably the most accurate of these techniques, but is also the most rarely applied one in coseismic investigations because of the sparsely located antennas on the Earth’s surface. At present, the European VLBI Network (EVN) consists of 18 individual antennas. This technique is more applicable for mapping tectonic plate movements. The International Laser Ranging Service (ILRS), on turn, supports 28 Satellite Laser Ranging (SLR) stations in Europe. These instruments could provide monitoring of three dimensional deformations of some Earth’s points, but in the recent years is substituted by the usage of lighter and mobile GNSS equipments. Both VLBI and SLR have another indirect role in earthquake investigations like definition of geodetic reference systems and precise orbit calibration of satellites used in the surface deformations studies. Not long ago the most popular space technique for co-seismic investigations in the scientific society were the Global Navigation Satellite Systems (GNSS) and in particular the Global Positioning System (GPS). Originally developed as a military navigation system, many other applications were gradually found and one of them is the Earth’s crust monitoring. There are two approaches in GPS application for seismic studies. The first one is consisted in establishment of a network of permanent stations, which provides time series of station positions in 3 directions – east, west and up, and the displacements are detectable as “jumps” in the series. The other method consists in re-measuring of set of points, in most cases more densely located around the epicentre, in pre- and post-event campaigns. The biggest disadvantage here is the lower precision of vertical position determination of the measured points. The geodetic receivers of this system could provide even sub-centimetre estimation of horizontal position for a sufficiently long period of measurements while the vertical accuracy is about 3 to 5 times less. This fact is a precondition for applying of GPS crust deformation measurements mainly in the cases of strike-slip faulting. Another powerful technique for surface deformation detection was developed in the recent decades. The Synthetic Aperture Radar Interferometry (InSAR) is based on the interference pattern calculated from the phase difference between two radar images of the affected area acquired before and after the seismic event. The accuracy is defined as a fraction of the wavelength in millimetre range. All methods mentioned are characterised by measuring the spatial dependence in the location between particular points of the Earth’s surface or measuring the location of a point Introduction 3 with respect to a space object. With re-measuring the position of these points, angles or distances between them, after occurrence of a seismic event, the deformation of the crust could be estimated as a difference from the initial state. The only exception of this principle is the InSAR technique. In that case, the methodology is more or less similar to the already described but there is a very important difference. Here the deformation data are collected for all the covered area, not only for several points. Thus, the field of deformation of the affected area could be formulated directly and not by interpolation, which in all other cases is highly dependent on the proper network configuration, network density or the distance from the epicentre to the equipment. Another advantage of the SAR Interferometry is the possibility to use it in inaccessible areas, where placing measuring equipment is difficult. Also, the costs for such field measurements are economised. The only expenses needed for radar interferometry “measurements” are for receiving images and partly software (some free of charge processing tools are available). Of course, the method has its own shortcomings, some of which will be presented in the following chapters. Nevertheless, the scientific community works very intensively on improving the processing algorithms and on the possibilities for elimination of some errors, so this method can be more reliable for future researches. All mentioned advantages over the other geodetic methods prompted the SAR interferometry to be chosen for the current investigation. Furthermore, for some of the studied cases comparisons or complex examinations were made in conjunction with other methods like GPS and seismology if such are available. The first Chapter of the thesis gives more details and definitions for the InSAR technique and interferometric processing, advantages and limitations. Some preliminary tests were made to define the most suitable parameters of the processing for the studies in this thesis. These tests are described in the end of Chapter I. As an introduction for the second Chapter the InSAR capabilities for detection of crustal deformations caused by moderate-size earthquakes is reviewed. Objectives of the dissertation and objects of investigation The main objective of the present dissertation is the determination of the crustal deformations caused by four significant moderate-to-strong earthquakes which occurred in the Eastern Mediterranean using InSAR technique and for one of the events – seismological investigation based on four-month aftershock sequence recorded by a local temporal seismological network. The Eastern Mediterranean is one of the most seismically active regions in Europe. It is located at the boundary between the African and Eurasian plate where ocean-continental subduction dominates. The area also falls under the influence of the northward movement of the Apulian plate which collides with Eurasia, causing transform faulting on the western coasts of the Balkan Peninsula and forming the vast DinaridesAlbanides-Hellenides thrust and fold chain. The transition between the two types of geotectonic processes takes place in the area of the Central Ionian Islands along the Cephalonia Transform Zone. High tectonic activity exists also as a result of western drifting of the Anatolian platform due to the squeezing pressure of the Arabian plate. Besides, the Eastern Mediterranean lithosphere is being subducted beneath the Aegean block along the Hellenic Arc (Lagios et al. 2007). A number of smaller blocks complement the complex fault system in the Eastern Mediterranean. The present study is focused mainly on four moderate-to-strong earthquakes occurred in the region – Konitsa, Valandovo, Lefkada and Movri (Northwestern Peloponnesus) events (Fig.1). In attempt to determine the field of deformations caused by these seismic events an InSAR detection of the crustal deformations is implemented. The first investigated case is a sequence of moderate earthquakes with magnitude in the range of 5.0-5.3 which hit the vicinity of Konitsa Town in the summer of 1996 (Chapter II). The moderate size of this group Introduction 4 of events is a serious challenge for the crustal deformation detection with the InSAR technique. For comparison, a complementary study is conducted for an earthquake occurred close to Dojran Lake in the area of Valandovo town which had similar magnitude and parameters as the previous sequence of events. The third studied event occurred offshore the island of Lefkada – the northernmost of the Central Ionian Islands situated at the transition zone between the two main tectonic regimes in the region – subduction to the south and collision to the north. This earthquake with magnitude 6.3 occurred in August 2003 causing remarkable damages on the island (Chapter III) and is remarkably stronger then the previous mentioned, but the location of the main shock offshore the island is another specific issue for the InSAR technique. In the area of Northern Peloponnesus in 2008 another strong event happened with a magnitude of 6.4 but without exhibiting significant surface ruptures (Chapter IV). Since no definite results are yield with InSAR study in the last case, analysis of the microseismic aftershock series is applied. Fig.1. Seismicity in the Eastern Mediterranean with M>4 (1973-2008) according to USGSNEIC and the earthquakes investigated in this thesis: Konitsa and Valandovo in Chapter II; Lefkada in Chapter III; Movri (Northwestern Peloponnesus) in Chapter IV. The second objective of this work is to model the ground deformations with the assumption for elastic medium using the information gained from the InSAR investigation. The proposed models of dislocation for the four studied earthquakes are analysed jointly with the available geodetic, seismological and tectonic information. The modelling strategy is based on the Okada (1992) formalism for dislocation model of a rectangular source in uniform elastic half-space where this source is described with nine parameters (Fig.2): - depth (D) of the upper edge of the fault (DUEF); - 2 coordinates (x,y) of the centre of the upper edge of the fault; - 2 angles: strike ( φ ) and dip (δ); - 2 dimensions: length (L) and width (W); - 2 slip (Δu) components: dip-slip and strike-slip (or slip (Δu) and rake (λ)). Introduction 5 Fig.2. Parameters of the dislocation model of a rectangular source after Okada (1992) The third main goal of the thesis is to gain knowledge about the seismicity in the investigated region and to study the potential of the InSAR technique for investigation of shallow and moderate (5 < M < 6) earthquakes in the Mediterranean. Discussion on these subjects is presented in the “Conclusions and perspectives” chapter. This chapter completes the dissertation and gives the general conclusions for the entire work as well as a discussion on the existence of a smaller tectonic feature in the area of the Central Ionian Islands. In addition some ideas for future work are pointed out. I. Background of InSAR processing 6 1. Chapter I: Background of InSAR processing 1.1. Remote sensing Remote sensing (RS) is the science for acquiring information (spectral, spatial, temporal and polarization) about the Earth’s surface without coming into physical contact with the object under investigation (source: CCRS). The information is transferred by the use of electromagnetic radiation (EMR) which is reflected or emitted from an object (source: GIS development). In RS technology a wide range of EMS (Electromagnetic spectrum) is in use, from a very short wave “Gamma ray” to a very long “Radio Wave”. Remote Sensing has some remarkable advantages over the other techniques for observation of the Earth’s crust like (source: Geoscience Australia): • continuous acquisition of data • regular revisit capabilities (resulting in up-to-date information) • broad regional coverage • good spectral resolution (including infra-red bands) • good spatial resolution • ability to manipulate/enhance digital data • ability to combine satellite digital data with other digital data • cost effective data • map-accurate data • possibility of stereo viewing • large archive of historical data 1.2. SAR Interferometry 1.2.1. Definition Synthetic Aperture Radar Interferometry (InSAR, IFSAR or ISAR) is a microwave remote sensing technique for measuring the topography of the Earth’s surface and its changes over time. InSAR is the synthesis of conventional SAR and interferometry techniques. It has developed in recent years to improve some of the limitations in conventional SAR systems, like twodimensional registration of the ground targets. SAR interferometry subsequently has opened entirely new application areas in earth system science studies (Rosen et al. 2000). Conventional SAR measures only the location of a target in a two-dimensional coordinate system, with one axis along the flight track (“along-track direction” or azimuth) and the other axis defined as the distance from the SAR to the target (“cross-track direction” or range) (Rosen et al. 2000). By combining two SAR images in an interferogram, the information about the third dimension – elevation of the target, can be revealed (Bürgmann et al. 2000). The basis for the interferogram generation is the superposition theory for two coherent waves which have same frequency. By their combination, new phase pattern – interferogram, is created. This pattern is determined by the phase difference between the two waves – transmitted and backscattered. 1.2.2. InSAR concept and principles InSAR uses the phase information in two complex images consisting of amplitude and phase information, covering the same area, acquired either by two different antennas or using repeated acquisitions, to determine the phase difference between each pair of corresponding image points, thus producing an interferogram (Fig.1.1). The phase difference of two images taken from slightly different locations, but at different times, can precisely measure any shifts of the returned phase. Thus, if the Earth’s surface moved towards or away from the radar between the two imaging passes, phase changes that result in the range (distance) from the satellite to the ground target could be measured as a fraction of the radar wavelength in the radar line of sight direction with a precision corresponding to millimetre-level displacements. The InSAR I. Background of InSAR processing 7 effectiveness is based on its coherence – both amplitude and phase are directly available by controlled coherent radar waves of the satellites (Bürgmann et al. 2000; Hanssen 2001; Madsen and Zebker 1998). Fig.1.1. Example demonstrating the interferogram formation (source: GMTSAR) Unlike the conventional SAR where the location of the target is received only in two dimensions (Fig.1.2.a), the InSAR coordinate system is 3 dimensional. Besides the two coordinates, one along the flight track (azimuth direction) and the other along the range from the antenna to the target (cross-track), a third dimension exists here – phase change (Fig.1.2.b), which could be used to derive a topographic map. If the phase component of the first complex radar image is Φ 1 (x , r ) , the phase component of the second image is Φ 2 (x , r ) and the phase delay is Φ= 2πR λ (1.1) where R is the slant range and λ is the wavelength, so subtracting the first phase from the second will give the difference which form the interferogram φ(x , r ) or φ = ΔΦ = Φ 2 − Φ 1 = 2πQδR λ (1.2) where Q = 1 when the two antennas are located on one and the same platform (Hanssen 2001). I. Background of InSAR processing 8 Fig.1.2. Radar imaging geometry: (a) SAR principle, (b) InSAR principle 1.2.3. Advantages The biggest advantage of InSAR over the conventional and GPS geodetic techniques is that it gives the opportunity to obtain dense and continuous information about the behaviour of large areas of the surface. Also equipment on the ground or expensive field campaigns are not required and for that reason it could be used in unapproachable or distant areas of Earth’s surface. The advantages of this methodology come from its characteristics. Because the system uses signals in the microwave spectrum it has cloud-penetrating potentiality. On the other hand, it has day and night operational capabilities since it provides its own illumination. The facile, cheap and quick access to the data and the processing softwares is another advantage of InSAR. Also the big archive of data gives the opportunity for longer monitoring of the areas of interest. 1.2.4. Limitations If the spectrum and thermal character of the ground surface changes between image acquisitions a temporal decorrelation will appear. Bare rock and man-made structures often remain coherent for long periods of time, but C-band (wavelength of 5.6 cm) interferograms of forested areas, for example, can be incoherent even if the images were acquired with only one day separation. This is due to multiple scattering of leaves and small branches whose positions are unstable. The use of images provided by L-band radar signals ensures coherence, even in vegetated areas (Wright 2002). Another limitation for the technique is the Earth’s changing atmosphere: water vapour concentrations, in particular, distort the phase ruler, causing phase shifts that can be confused with deformation. A thundercloud, for example, can cause phase changes equivalent to ground motions of up to 10 cm. In addition, InSAR can only measure surface deformation in the line of sight (LOS) of the satellite, so it is insensitive to motions parallel to the satellite track. The true, three-dimensional character of the motion is lost and only information from a single dimension is received. This can cause ambiguities in interferogram interpretation, which could be overcame through combinations of data taken from different look directions (Bürgmann et al. 2000; Wright 2002). I. Background of InSAR processing 9 Among all interferometric data there is always possibility of lack of baseline correlation for some of the pairs. The critical value for the perpendicular baseline B c⊥ for a pair defines the limit beyond which loss of all coherence will occur: B c⊥ = λR 2Δr cos 2 θ , (1.3) where λ is the wavelength, R is the range, Δr is the ground resolution cell in the range, θ is the look angle. 1.2.5. Errors The main error sources that influence the interferometric results are the contribution of atmospheric signal delay, interferometric decorrelation due to temporal and geometric scattering characteristics, the unknown phase ambiguity number, and errors in supplementary information such as the reference satellite orbits and elevation models (Hanssen 2001). These errors are included in the interferometric phase signal as: φint = φ flat + φ topo + φ def + φ orb + φ atm + φ noise (1.4) where φ flat is the phase delay (known) associated with the contribution of the flat Earth, φ topo is the phase delay (known) associated with the topographic phase, φ def is the phase delay (unknown) which represent surface displacement at LOS, φorb is the phase delay (known) due to orbit errors, φ atm is the delay (unknown) associated with the atmospheric (ionospheric and tropospheric) contribution and φ noise is the phase delay (unknown) associated with the noise due to the loss of coherence between the two scenes. 1.2.6. Applications The geodetic applications of spaceborne repeat-pass SAR Interferometry can be categorized in roughly four disciplines (Fig.1.3): topographic mapping with a relative accuracy of 10-50 m, deformation mapping with mm-cm accuracy, thematic mapping based on change detection, and atmospheric delay mapping with mm-cm accuracy in terms of the excess path length (Hanssen 2001). 1.3. Differential SAR Interferometry One of the InSAR technique variations, basic for geodynamics, is the Differential SAR Interferometry (DInSAR), which goal is to detect changes of the Earth’s surface. For this purpose the influence of the topography should be removed by multiple imaging of the surface. The interval of the repetitive imaging may be from a second to year spanning a surface displacement event. The main principle includes comparison between an interferogram representing the topography before the event and a second interferogram covering or following the geo-event (Fig.1.4). The fact that the displacement could be measured in the size of a pixel is a crucial distinction with respect to other techniques where the motion could be detected only when it is more than a significant fraction of the resolution of the imagery (Rosen et al. 2000). Another designation of this method is the along-track interferometry (ATI) stemming from the arrangement of two apertures on a single platform or the succession of acquisitions by different satellites. But the case seldom is ideal and that is why errors (sensitivity to topography) caused by cross-track separation of the antennas or big cross component of the baseline between two positions of the satellites occurred. I. Background of InSAR processing 10 Fig.1.3. InSAR applications Fig.1.4. Radar imaging geometry for deformation measurements – DInSAR (after Chen et al. 2000) II. July – August 1996 Konitsa earthquakes 11 2. Chapter II: Crustal deformation after July – August 1996 Konitsa (Greece) earthquake sequence 2.1. July – August 1996 Konitsa earthquake sequence 2.1.1. Introduction At the end of July 1996 in the area of Konitsa Town close to the Greek-Albanian border an earthquake with magnitude 5.3 was felt. The seismic activity continued during the following month causing a sequence of shocks, including two events with magnitudes similar to the first strong shock. First major shock struck the valley on 26th July (22:30 local time) and was followed by another, according to some sources, stronger event (Mw 5.7) on 5th August. The last significant earthquake from the series occurred on 20th August and had a magnitude of around 4.9 – 5.3. Various seismological agencies and some authors report for the hypocenter of the three big earthquakes under the so-called Sarantaporos fault (Fig.2.1) at a depth from 5 to 15 km. The main damages were observed in the town of Konitsa where 22% of the houses became nonhabitable and 29% – temporarily non-habitable (Galanakis et al. 2007). Some neighbouring settlements were also affected by the quakes. No human injures were reported. Many soil cracks with N 40° – 50° orientation, an aperture of 1 – 2 cm and some metres length were observed in the city of Konitsa and south of Aoos River. On the contrary, no vertical fractures were discovered. Intensive rockfalls and spring discharges were also reported at the slopes of Tymfi Mountain. The main goal of the current study is to determine the field of deformation caused by the seismic sequence of 1996 applying the InSAR approach and to define the best fit model for the activated fault based on the Okada formalism. Fig.2.1. The main earthquakes from the July-August 1996 Konitsa sequence according to different sources. Aftershocks from Papanastassiou (2001); fault system after Doutsos and Koukouvelas (1998); LOU – Louvari et al. (2001); PN – Papanastassiou (2001) The size of the main earthquakes, as well as the depth of the hypocenters, is reported with big variations by the seismological agencies. The main reason for this could be the very shallow occurrence of the events which leads to some difficulties in the hypocenter calculations. The II. July – August 1996 Konitsa earthquakes 12 mechanisms, reported by the seismological agencies, together with the aftershock sequence calculated by Papanastassiou (2001) are plotted in Fig.2.1. The mechanisms show similarity, but the Harvard CMT solution (shown in red in the figure) will be used as primary in the further investigation as more reliable. 2.1.2. InSAR processing and analysis During the occurrence of the earthquake sequence from 1996, the ERS 1 and 2 satellites were operative. The chosen tracks and frames for ascending and descending flight directions are track 186, frame 803 and track 322, frame 2794, respectively (Fig.2.2). Fig.2.2. The Northern Epirus vicinity with ERS ascending footprint - red dashed line, and ERS descending footprint - blue dashed line; black rectangle – the investigated area; in the inlet – location of the Northern Epirus in the Balkan peninsula. The raw images are provided by ESA, while for removing the orbital effects the precise DEOS Delft orbit files are used. The interferograms are processed using the open source package ROI_PAC 3.0 (Rosen et al. 2004) developed by JPL/Caltech team. The standard two-pass procedure (differential radar interferometry) is performed. To eliminate the topographic effect, a simulated interferogram is generated from the SRTM Digital elevation model (DEM) version 4 (Jarvis et al. 2008). This DEM has 90 m resolution and a nominal vertical error of less than 16 m. Nevertheless, thorough studies conducted by Mouratidis et al. (2010b) indicate that the actual absolute and relative errors (Std. Dev.) are well bellow 10m. In order to increase the quality of the interferometric pattern, two step-filtering is performed. As a first step, a multilooking (4 looks) operation is used to reduce the speckle noise. With this procedure the original interferogram pixel size of 8 m in range and 22.5 m in azimuth is averaged to 32 m in range and 90 m in azimuth, taking into account that the actual azimuth looks are 20 since the azimuth pixel dimension in the original raw image is 5 times smaller. This operation does not decrease the detection quality since the averaged pixel size corresponds to the DEM resolution. At the second step of the filtering the adaptive Goldstein-Werner filter (Goldstein and Werner 1998), which is included as an optional in the processing tool, is applied. The filtered interferograms are geocoded into the DEM coordinate system – GCS/WGS84. II. July – August 1996 Konitsa earthquakes 13 Fig.2.3. Left: Ascending differential co-seismic interferogram – ifg8 (Bp 95m). The star shows the main shock from 26.07.1996; faults – after Doutsos and Koukouvelаs (1998). Inset: the observed fringes used for the inverse modelling. Right: The same interferogram but unwrapped. Fig.2.4. Left: Descending differential co-seismic interferogram – ifg14 (Bp = 83 m). The arrows indicate the direction of descending satellite flight and side of looking of the satellite (LOS). Inset: the observed fringes used for the inverse modelling. Right: The same interferogram but unwrapped. The ascending co-seismic interferogram with the best quality (ifg8: 21.07.1996-25.08.1996) has time span of only one month and clearly shows two fringes which corresponds to at least 5.6 cm deformation in LOS (Fig.2.3). The interferogram is unwrapped through the SNAPHU software (Chen and Zebker 2002) and thus displacement in the range of 80 mm in LOS is revealed. The perpendicular baseline between the master and the slave images is 94 m. Such a distance between the two acquisitions could lead to a 4 mm error due to the topography which is only 1/7 part of the fringe and is not expected to bring significant changes in the interferometric pattern. The descending interferogram 3.01.1996-22.01.1997 (ifg14) is generated following the same processing strategy as that used for the ascending interferogram formation. The resulting interferometric pattern is shown in Fig.2.4. Three fringes are observed, which corresponds to about 8.4 cm crust deformation in the line of sight. II. July – August 1996 Konitsa earthquakes 14 2.1.3. Modelling The two fringes from the best ascending interferogram (ifg8) are sampled to 125 discrete points distributed approximately 400 m apart. The phase values corresponding to the chosen fringe points are used as reference in the inversion algorithm developed by Briole et al. (1986), based on the least-square minimisation approach proposed by Tarantola and Valette (1982). Rectangular-plane in uniform elastic half-space (Okada 1992) is assumed for the source of the dislocations. The Harvard CMT solution is used as initial parameters for the fault by fixing the strike to 215° and the dip to 36°. A forward model of synthetic interferogram is presented using the estimated fault model and taking into account the ascending direction of the satellite flight and its incidence angle. The complexity of the field of deformation appears as an obstacle for the model determination. Since the real interferometric deformation pattern is distorted, probably on account of the influence of the atmospheric disturbance and the occurrence of all three strongest events from the sequence within a short time span with similar magnitudes, two models of synthetic interferograms are proposed (Fig.2.5). The first model exhibits very good correspondence to the inner fringe (Fig.2.5 – left), but does not present the second circle of deformations. Fig.2.5. Synthetic ascending interferograms, compared with the observed fringes (white) from the real interferogram (Fig.2.3). Left: model 1, which describes the inner fringe better, but does not represent the second fringe. Right: model 2, which fits the outer fringe better. The rectangle represents the surface projection of the fault model. On the other hand the second model describes better the outer fringe (Fig.2.5 – right), while the inner fringe is extended. This model proposes a fault twice larger and a kilometre deeper than the one proposed by model 1. A conclusion can be made that both proposed models fit well the observed data and have a similar quality. Despite the fact that the first model fits very well the inner fringe, it does not describe the existence of the outer fringe. In contrast, the second model conforms both interferometric circles. Therefore, the model 2 is assumed best fitting forward model for the field of deformation based on the ascending pass of satellite. The deformation pattern in the descending interferogram is noisier, since the interferogram comprises one year time span. Three interferometric fringes could be distinguished here. The parts of the fringes closest to the Tymfi Mountain slope are distorted most probably because of the satellite side looking geometry which is directed from the mountain side toward the valley. Following the procedure executed for the ascending modelling, the descending fringes are sampled at approximately 700 m to 73 discrete points. The same initial and fixed parameters are used in the inversion procedure, but in the forward modelling the descending incidence angle is II. July – August 1996 Konitsa earthquakes 15 applied. As a result a descending synthetic interferogram is produced. The parameters behind this model are introduced in Table 2.1. Table 2.1. Preferable parameters for the ascending and descending modelled fault model name lat* (N) long (E) MA2 MD 40.034 40.029 20.713 20.719 depth of the UE (km) 2.2 2.0 strike (°) dip (°) rake (°) L (km) W (km) u (cm) Mo (Nm) Mw 215 215 36 36 -85 -84 10.0 12.7 5.5 6.2 19 20 2.6e+17 4.0e+17 5.6 5.7 * lat, long – coordinates of the centre of the upper edge (UE) of the fault model; L – surface rupture length; W – down-dip rupture width, u – average slip, MA2 – model proposed for the ascending pass (model 2), MD – model proposed for the descending pass. A theoretical determination of the horizontal and vertical components of the field of deformations is performed using the ascending and descending unwrapped interferogram. To define two of the three components of the LOS displacement, an assumption for the third one should be adopted. Here, the northern component is equalised for both satellite viewings, thus the displacement upwards and eastwards is extracted using the parameters of the satellite geometry (Hanssen 2001). The resulting pattern (Fig.2.6) shows a maximum vertical displacement of the Earth’s crust as a subsidence in the range of ~8 cm and westward displacements of the area to the west of the fault in the maximum range of ~4 cm. Fig.2.6. Theoretical displacements up- and eastwards The current investigation significantly supplements the existing knowledge on the geological and seismological data of the area of Konitsa. GPS or other type of geodetic measurements are not conducted in this area. Thus, the accomplished InSAR study provides lacking information about the field of the deformations caused by the 1996 Konitsa seismic sequence. 2.2. Similar case – 24th May 2009 Valandovo earthquake The area of Dojran Lake belongs to the region between Vardar and Struma Rivers (Fig.2.7) which is characterized by extensional tectonics, with ~N-S direction and dominated by a series of grabens with E-W to NW-SE directions forming the normal faulting regime in the area (Kiratzi 2009). A sharp morphological unit with E-W orientation gives the boundary between the Valandovo valley and the mountains along the southern slope of the Valandovo graben (Dumurdzanov et al. 2005). A major earthquake with magnitude 6.5–6.8 occurred in 1932 in this area. II. July – August 1996 Konitsa earthquakes 16 In the spring of 2009 northwestward from the Dojran Lake an earthquake with Mw 5.3 occurred. The fault solution reported by Harvard CMT is similar to the 26 July 1996 Konitsa earthquake parameters. The depth of the Valandovo event is reported at 23 km by the NOA catalogue. Fig.2.7. Part of the region between Vardar and Struma Rivers. The Valandovo earthquake is showed with the Harvard CMT mechanism and NOA aftershocks. The coverage of the ENVISAT track 279, frame 2767, is presented with red dashed line. Four co-seismic interferograms are generated from five descending ENVISAT/ASAR images, provided by ESA. Only one post-seismic image is used in the analysis. None of the processed interferograms does show clear surface deformation (Fig.2.8). One of the reasons for the latter could be the deeper location of the fault. Therefore several prediction models of deformations at different depths of the upper edge of the fault (DUEF) are proposed (Fig.2.9). Some similarity between the 20.07.2008-31.05.2009 interferogram (with the shorter perpendicular baseline) and the forward models with DUEF at ~5-10 km could be found. II. July – August 1996 Konitsa earthquakes 17 Fig.2.8. Differential co-seismic descending interferograms. The black area is the non-coherent signal of the Dojran Lake Fig.2.9. Modelled interferograms with depths of the upper edge of the fault 2 km, 5 km, 10 km or 15 km. The yellow triangle represents the location of the VALA GPS permanent station. Although there is a big similarity between the Konitsa and Valandovo events, no surface deformations could be detected by InSAR for the latter event. Possible reason for this could be the local geology structure. The Valandovo area is composed mainly of sedimentary rock, while the Konitsa valley consists of alluvium deposits, which are more sensitive to the seismic waves. Comparison of the perpendicular and temporal baselines (Table 2.2) of the used co-seismic II. July – August 1996 Konitsa earthquakes 18 interferograms shows that these characteristics are more or less in the same range for both cases, so that temporal or spatial decorrelation could not be pointed as an eventual reason for the absence of fringes in the Valandovo interferometric patterns. Moreover, the vegetation coverage of the investigated sites is similar. The Valandovo 20.07.2008-31.05.2009 interferogram, which seems to contain some deformational signal, is composed by images acquired in closer seasonal time. The most reasonable cause for the impossibility of detection of the crust displacement after the 2009 Valandovo earthquake is the deeper source of the event. Another fact supporting this statement is the absence of definite “jump” related with the earthquake in the time series of the Valandovo permanent GPS station, maintained by the Department of Geodesy at NIGGG-BAS, Sofia. In this case the results that are received by the InSAR and GPS measurements, namely the absence of detected deformation, give a very valuable information about the location of the fault – buried at more than 10 km. No matter if this earthquake was felt in wide area, no surface deformations are recorded by reason of the moderate magnitude of the event. This case is an example for the restrictions of the InSAR application regarding the source depth. It also shows the uncertainty of the information reported in the seismological catalogues. The Konitsa event was reported by Harvard CMT to have depth of 15 km, but we proved that the depth of the source is significantly less, while the Valandovo earthquake was reported to have a depth of 12.8 which is confirmed as possible by the present InSAR investigation. It should be pointed that most of the events with similar magnitude in the period 1992-2003 have the exact depth of 15 km in the Harvard CMT catalogue which could be an effect of the applied velocity model in the hypocenter estimation for this period. Table 2.2. Comparison of the Konitsa and Valandovo earthquake cases. Bperp and Btemp represent the range of the perpendicular and temporal baselines, respectively, of the interferograms used in the analysis. event Mw strike° dip° rake° DUEF (km) L (km) W (km) Slip (cm) Bperp (km) Btemp (days) Valandovo 5.3 285 54 -90 > 10 ~7* ~3.5* ~22.5* 50-253 140-455 Konitsa 5.3 215 36 -84-85 2-2.5 10-13 5.5-6.2 19-20 5-363 1-420 * calculated with M0 and half duration time reported by the Harvard catalogue III. The Lefkada Earthquake, 14th of August 2003 19 3. Chapter III: The 14th August, 2003 Lefkada earthquake 3.1. Motivation of the investigation The region of the Central Ionian Islands (Lefkada, Cephalonia, Zakynthos and Ithaki), Greece is unique from the plate tectonics point of view (Sachpazi et al. 2000), as it is a multiple junction, where all four types of plate boundaries (collision, subduction, transform and spreading) connect at a 100 km-wide area. More specifically, the InSAR approach which is applied for the 14th August, 2003 earthquake investigation represents a special challenge, because of the particular case of a strong seismic event (Mw = 6.3) occurring offshore of a small island and the vast water areas around as decorrelative factor. It is a very useful occasion to examine capacity of the method in such conditions. 3.2. InSAR investigation and fault plane modelling 3.2.1. Introduction Intensive tectonic processes in the vicinity of the western coasts of the Balkan Peninsula are associated with the seismic activity in this area which is the highest in Europe. The main crustal deformation occurs along the Hellenic Arc, as a result of the subduction of the African - Nubian plate beneath the Aegean plate (Le Pichon and Angelier, 1979; Ganas and Parsons, 2009). The Central Ionian Islands lie on the Aegean Plate between the Apulia – Eurasia continental collision to the north and oceanic subduction to the south, and suffer from very intensive surface deformations with earthquake magnitudes up to 7.4 (Papazachos and Papazacou, 1997). The plate boundary in the Central Ionian Sea is defined by a major, right-lateral strike-slip fault – the Cephalonia Transform Fault (western termination of the Hellenic subduction zone) and it is characterised by a deep bathymetric trough exceeding up to 3000m (Fig.3.2). Strike-slip faulting prevails in a large area to the east of Cephalonia transform reaching the western boundaries of the Gulf of Corinth (Kiratzi et al. 2008; Ganas et al. 2009). The Central Ionian region is considered as the most active area of shallow seismicity in the Aegean Sea and the surrounding landmass (Benetatos et al. 2007). This complex geotectonic regime predetermines the high seismic hazard in the area of Lefkada Island. This region was affected indeed by several significant events in the past. The most recent strong earthquake, with magnitude of Mw 6.3, occurred on 14th August 2003 offshore the western coast of Lefkada Island, causing severe damages around the whole island. This contribution aims to investigate the pattern of co-seismic deformation including the location and dimensions of the seismic fault of the 14.8.2003 strong, shallow earthquake. 3.2.2. The 14th August 2003 earthquake On 14th August 2003 (05:15:08.31 GMT) a strong earthquake (Mw 6.3) occurred offshore the western coast of Lefkada Island causing considerable damages in some settlements, all harbours and the road network, as well as extensive landslides, rock falls, soil liquefaction, subsidence, densification and ground cracks, especially along the northwestern part of the island (Katopodi and Iosifidou 2004; Papathanassiou et al. 2005; Pavlides et al. 2004). Maximum intensity of VIII in the town of Lefkada was observed, while in the remaining part of the island intensity ranged form V to VII (Papadopoulos et al. 2003; Papathanassiou et al. 2005). Up to 31st August 2003 more than 470 aftershocks (M > 1.5), concentrated in two clusters, were recorded by Patras Seismological Laboratory (Fig.3.1), the strongest of which (Ms 5.25) occurred at 16:18:03.47, several hours after the main shock. Papadopoulos et al. (2003), Karakostas et al. (2004) and Papadimitriou et al. (2006) consider the southern cluster, located close to the northwestern coast of Cephalonia Island, rather representing a triggered seismicity, due to stress increase from the mainshock, than typical aftershock activity. They assume that the second cluster appears due to activation of Cephalonia segment. On the contrary, Benetatos et al. III. The Lefkada Earthquake, 14th of August 2003 20 (2005; 2007) and Zahradník et al. (2005) adopt the idea of complicated fault geometry of Lefkada segment, which produced two main sub-events and additional segmentation of the rupture, which is responsible for double-clustering. According to the latter group of authors, the multiple source character of the mainshock is represented by two main sub-events along a ~N12°E (or N20°E) line that ruptured Lefkada segment from north to south (Fig.3.1). The two sub-events are separated in time and space at 14s and ~40km, respectively, have strike-slip focal mechanisms and provoked concentration of aftershocks in two clusters with mainly strike-slip character along the NE-SW trending fault segment (Fig.3.1). Thrust and reverse faulting mechanisms also detected, mostly in the north and mainland parts of the island, indicate probable activation of adjacent areas, as the Ionian Thrust Zone. Fig.3.1. The 14th August 2003 Lefkada earthquake from different sources. Other fault plane solutions are also shown. Aftershock sequence from PSL is introduced with coloured circles according to the depth. In present investigation the Harvard solution (in red) is taken as a basic in inversion process. The final solution of this study is shown in brown (IL11) referred to the centre of the fault predicted by the fault model. Since only minor damages have been reported at the northwestern part of the Cephalonia Island, the present investigation is concentrated on the crustal deformations in the area of the main shock (or first sub-event) and the following sequence of aftershocks on the Lefkada Island (Fig.3.1). 3.2.3. InSAR processing and analysis According to the European Space Agency’s (ESA) catalogue (EOLi), in the period between January 2003 and January 2008, there are 13 ascending and 20 descending ENVISAT/ASAR (Cband) passes from track 458 and track 322 respectively, covering the study area (Fig.3.2). In both cases of satellite viewing, only one of these scenes is pre-seismic – 21.03.2003 for the ascending and 25.06.2003 for descending, and this fact imposes limitation to generate co-seismic pairs with only one master image. On this way the eventual flaw of this image could bring a systematic error in the whole group of interferograms. In spite of this limitation, the choice of the scenes used in the present investigation is made in such a way that the image pairs formed have perpendicular baselines less than 200 m with III. The Lefkada Earthquake, 14th of August 2003 21 respect to the pre-seismic image. This requirement is valid for six ascending images, which form six ascending co-seismic and 15 ascending post-seismic interferometric pairs, and eight descending scenes, forming eight co-seismic and 23 post-seismic interferograms. Standard processing flow of differential (2-pass) method was followed to generate the interferograms, using ROI_PAC 3.0. The raw images were provided by ESA as well as the Doris precise orbit files, used for removing the orbital effects. To eliminate the topographic effect, a simulated interferogram is generated from the SRTM DEM (version 4). Multilooking (4 looks) was applied to minimize the loss of coherence and to reduce the speckle noise in the produced interferogram. The Goldstein-Werner filter was applied in the final stage of processing to improve the quality of the interferogram. The results are georeferenced to the coordinate system of the used DEM (GCS/WGS84). Fig.3.2. ENVISAT/ASAR ascending footprint represented by red dashed rectangle, descending footprint – by blue dashed rectangle. Location of 14th August 2003 earthquake from NOA is shown with red star and aftershock sequence by PSL with red dots. Black rectangle – the area of investigation; orange dashed line – generalized view of Cephalonia Fault Zone. The ascending co-seismic interferogram with the best quality (ifg5), covering the time period between 21st March 2003 and 5th November 2004 (Fig.3.3 left), shows at least 2 fringes in the western part of the island. Since one fringe is equal to half the wavelength, which for ENVISAT C-band is 2.8cm, then these two fringes are equal to 5.6 cm of deformation of the crust in the satellite line of sight (LOS). In the case of descending co-seismic set of data, the deformations after the occurrence of the strong earthquake are detected with significantly lower quality, mainly because of the look direction of the satellite towards the highest mountains, so that the deformation pattern appeared distorted. 3.2.4. Modelling The observed fringes in the best ascending interferogram were sampled to 34 discrete points at approximately 1 km apart (Fig.3.3 left – inset). The descending interferograms are not used in the analysis, due to low quality and excessively deformed fringe pattern. The Harvard CMT solution is used as initial parameters in inversion procedure fixing the strike angle to 18° and dip III. The Lefkada Earthquake, 14th of August 2003 22 angle to 59°. The applied inversion algorithm is those developed by Briole et al (1986) which is based on a least-square minimization approach proposed by Tarantola and Valette (1982). The Okada (1992) formulation for rectangular-plane source in uniform elastic half-space is adopted for the dislocation source. The forward model of interferogram (Fig.3.3 right) is generated, taking into account the ascending direction of the satellite flight and with the new modelled parameters of the fault: top centre of the fault 20.608°E, 38.839°N, strike 18°, dip 59°, strike-slip 75cm, length 16km, width 10km, depth to the upper edge 2.5 km, seismic moment 3.6e+18Nm, corresponding to Mw = 6.3. Fig.3.3. SAR interferometry results for the Lefkada earthquake. (Left) Differential ascending interferogram – ifg5 (BL 88m). The star shows the main shock according to NOA. Inset: The observed fringes. (Right) Synthetic ascending interferogram obtained from inversion of the received observed fringes (left). The rectangle represents the approximate model of the fault segment. In order to confirm the reliability of the proposed model of the fault, a series of inversions are conducted with varying dip or length of the fault, taking into account the values proposed by previous studies (Fig.3.4). In both cases, for the dip or the length, all the rest parameters are fixed. This investigation of the confidence of these parameters shows that the best fitting model would be achieved at dip of 60° and length of the faulting segment of 16 km. The sensitivity of the model to the dip alteration is expressed by “pressing” of the fringe’s shape. In the case investigated here, exactly the mean part of the outer fringe is not very clear outlined. The averaged descending unwrapped interferogram will be used, together with the ascending result, in a theoretical determination of the horizontal and vertical components of the field of deformations. To define two of the three components of the LOS displacement, an assumption for the third one should be adopted. Here, the northern component is equalised for both satellite viewings, thus the displacement in up and east direction is extracted using the parameters of the satellite geometry (Hanssen 2001). The resulting pattern shows a maximum vertical displacement of the Earth’s crust in the range of ~5 cm in the western-central part of the island and souhward displacements in the western coast in the maximum range of ~6 cm. Fig.3.4. Dip and length confidence study by comparison with the values proposed by previous studies. To the right: rms misfits plot for the dip and the length respectively, inverted by the InSAR data. All other parameters are fixed (see the left of the figure) III. The Lefkada Earthquake, 14th of August 2003 23 III. The Lefkada Earthquake, 14th of August 2003 24 Fig.3.5. Theoretical displacements in Up and North directions. 3.2.5. Discussion and conclusions We examine fault size, geometry and its relation to earthquake magnitude and how this relates to the historical seismicity record. Our geodetic solution (Fig.3.3) provides the first published data on active fault geometry offshore Lefkada, suggesting that the rupture length may not exceed 16 km, about 40% greater than the empirical relationship of Wells and Coppersmith (1994) (see Table 3.1). The 2003 Lefkada event occurred along one fault segment plane dipping to the east as suggested by the linear inversion of the InSAR fringes. The earthquake was followed by numerous aftershocks on neighbouring fault planes that are characterised as capable to produce moderate events (Karakostas and Papadimitriou, 2010). Our results indicate that, the 2003 earthquake ruptured only the northern portion of the 50-km Lefkada segment. This segmentation model predicts rupture lengths of the order of 15-20 km and maximum earthquake magnitudes in the range 6.2-6.5. This suggestion has important consequences for the seismic hazard of the area, as the maximum earthquake expected offshore Lefkada may not exceed Mw=6.5. Table 3.1. Comparison between Length-Magnitude empirical relations and calculated fault dimensions Source of empirical relationship L* (km) W (km) Konstantinou et al. (2005) log L = -1.49 + 0.47 Mw L = 29 log W = -1.07 + 0.34 Mw W = 11 Papazachos and Papazachou (1997) log L = -1.85 + 0.51 Mw L = 23 log W = -0.13 + 0.19 Mw W = 12 u (cm) log u = -3.71 + 0.82 Mw u = 24 Well and log L = -3.55 + 0.74 Mw log W = -0.76 + 0.27 Mw Coppersmith (1994) L = 13 W=9 this study L = 16 W = 10 u = 75 * L – surface rupture length; W – down-dip rupture width; u – averaged displacement 3.3. Additional investigations Hollenstein at al. 2008b carried out co-seismic GPS measurements of several points on and around the Lefkada Island. The comparison between the measured co-seismic GPS displacements of the points on the island with the modelled displacements for the corresponding points shows some inconsistency in the results for the two southern points (Table 3.2). III. The Lefkada Earthquake, 14th of August 2003 25 Table 3.2. Comparison between the displacements obtained from the GPS measurements and the synthetic radar interferometric model for the same GPS sites. d N* dE dU (mm) (mm) (mm) GPS InSAR GPS InSAR GPS InSAR VONW 38.91 20.85 -21 -21 -23 -47 4 14 1KVL 38.97 20.66 -69 -117 -21 3 52 47 VASI 38.61 20.57 -40 -19 -6 4 20 3 DUKA 38.56 20.54 -39 -10 -40 3 -26 1 * dN, dE, dU are the displacements in the north, east, up directions, respectively. Site lat (°) long (°) According to the authors, the conflicting direction and velocity of DUKA site could be explained by the existence of local geological structure. This idea is supported here, since the rest of the GPS vectors have a relatively good compatibility with those extracted from the modelled interferogram. Fig.3.6. Horizontal displacement of GPS points by Hollenstein et al. (2008b) compared with surface slip according to the predicted model (see Table 3.2) IV. The Movri earthquake (NW Peloponnesus), 8th of June 2008 26 4. Chapter IV: The Movri earthquake (NW Peloponnesus), 8th June 2008 4.1. Introduction On 8th June 2008 a strong earthquake of magnitude Mw = 6.4 occurred 30 km south of city of Patras in North-western Peloponnesus (South Greece), and was felt in almost the whole country as well as northernmost in Tirana, Albania, and westernmost in Cosenza, Italy. The shock caused the death of two people and injury of about 265. Ground failures like landslides, coastal subsidence and rockfalls were provoked in 15-kilometre wide area around the presumable fault and structural damages at approximately 2000 km2 were caused. About 10’000 houses suffered small to severe damage, some rail lines were deformed and liquefaction appeared at many spots. According to several different sources the main seismic event is reported as a shallow shock with depth between 10 and 30 km (Fig.4.3) and the fault plane solution represent nearly pure dextral NE-SW strike-slip motion with strike of approximately 210°, dip in the range of 81° to 89° and small reverse component. 4.2. Multidisciplinary investigation International team composed of researchers from France, Greece, Bulgaria, Lebanon and Italy carried out geological, geodetic and seismological investigation in the area stricken by the earthquake. 4.2.1. Geological field survey No fault rupture was found but damaged buildings, mostly old houses and churches, were observed all over the visited sites. Numerous landslides and rockfalls were also observed throughout the region and especially in the area surrounding the Skolis Mountain. The only place where the team found some evidence of a possible fault surface rupture is located in the area of Kato Achaia. There, distributed faulting and fissuring has been observed, aligned along a roughly N-S direction over a distance of approximately 1 km. 4.2.2. Permanent seismic Network HUSN data study The mainshock and the first month aftershock sequence were recorded by the Hellenic Unified Seismological Network (HUSN) and were processed and studied by the team of UPSL (Serpetsidaki et al. 2010). The aftershock profiles show grouping of events in two clusters – one to north-east and the other to south-west of the epicentre. 4.2.3. InSAR investigation 37 SAR images in total were processed – 33 ENVISAT/ASAR with track numbers 50, 186, 279, 415 (Serpetsidaki et al. 2010). No significant co-seismic signal could be detected. The area of Peloponnesus is highly vegetated and this is a cause for signal decorrelation, which is a very significant obstacle to the ENVISAT interferogram formation despite of the short temporal and perpendicular baselines. For the best interferogram the time difference between master (11 May 2008) and slave (16 June 2008) images is about one month and the spatial separation is 139 m. The coherence of this interferogram is low but the noise is not sufficient to hide fringes if there were any. The only detected deformation is a small anomaly (1 fringe or less of subsidence) in one of the interferograms beneath the town of Kato Achaia (Fig.4.1). This coincides with the highest intensity of VIII degree which is surveyed namely in the area of Kato Achaia, as well as the most extensive liquefactions (Papadopoulos et al. 2010). IV. The Movri earthquake (NW Peloponnesus), 8th of June 2008 27 Fig.4.1. ENVISAT/ASAR co-seismic interferogram (5 August 2007 – 24 August 2008) from the area of Kato Achaia, Bp = 252 m (Serpetsidaki et al. 2010). 4.2.4. GPS measurements During a two-weeks campaign after the main shock, 25 points from a network around Gulf of Patras (Fig.4.2.a) which were measured in 2003 and 2006 were re-surveyed (Charara 2010). The velocity of the GPS sites is calculated relative to stable Eurasia (Fig.4.2.b). A displacement that could be related with the earthquake from 8 June 2008 (about 6 mm to the north) was noted only in the horizontal component in the time series of the NOA permanent RLS station – the closest to the epicentre (Fig.4.2.c). The horizontal displacements for all other points do not exceed the error and could not be assumed as a co-seismic deformation of the crust. Fig.4.2. GPS observations around Gulf of Patras and NW Peloponnesus (Charara 2010). (a) Map of the investigation network. (b) Velocities of the same points in respect to stable Eurasia. (c) Time series of the permanent NOA station RLS0. In the current investigation the modelled displacement field fits correctly the amplitude of the observed GPS results (Briole et al. 2008). The predicted horizontal vector in RLS station is about 5 mm (dE = -2.57, dN = 4.17, dU = 6.74 mm), while the time series indicate a horizontal displacement of about 6 mm (dE = 0.0 ± 2, dN = 6 ± 2, dU = 6 ± 2 mm) (Charara 2010). The absence of a significant GPS displacement is explained by activation of a fault with a limit of the upper edge at about 22.5 km. IV. The Movri earthquake (NW Peloponnesus), 8th of June 2008 28 4.2.5. Microseismicity 4.2.5.1. Investigation network Several days after the main shock starting from the 14th June, seven seismological stations have been installed for almost 4 months to record the aftershocks sequence (Briole et al. 2008). The network was established around the supposed northern termination of the main fault (Fig.4.3), supporting the northward propagation of the seismicity in the days following the main shock. The geometry of the network was prompted also by the importance of a careful monitoring of the seismicity near the town of Patras which is the third largest city in Greece with more than 220’000 inhabitants. The inter-station spacing is about 10 km. The aim of the array is to correctly resolve the depth of the seismicity, and to detect possible post-seismic activation of secondary faults in the shallow crust (Briole et al. 2008). Fig.4.3. Temporary and permanent seismic network. Orange triangles represent the position of the seismometers from the local network, while with yellow signs the permanent seismological stations from different networks (see the legend) used in the investigation are marked. The red stars represent the epicentre of the main shock according to different catalogues and corresponding focal mechanisms. In the inset: location of the area covered by the networks. 4.2.5.2. Processing and evaluation criteria I and several colleagues from the Patras Seismological Laboratory used the software SeisGram2K v5.3.4 developed by Anthony Lomax for manually picking the P and S phases from the recorded seismograms. The picked arrival times were inverted with the software HYPO71 (Lee and Lahr 1975) for hypocenter determination of the aftershocks. The crustal model applied in earthquake location software is the one used by CRL in the investigations to the east, around Gulf of Corinth. The records of several stations from different permanent networks (Fig.4.3) – ATHENET (UOA), BBNET (NOA), CRLNET (CRL) and PSLNET (UPSL) are used in the processing. The phases of the closest station to the epicentre of the main shock – RLS, are included in the processing for almost all of the aftershocks. Location of the aftershocks is calculated by the travel time of P and S waves from the source to the stations. For the proper 4D definition of a point location, at least 4 arrivals are necessary, so the assumption for the minimum number of arrivals is 4 P readings supported by at least 3 S IV. The Movri earthquake (NW Peloponnesus), 8th of June 2008 29 readings. Magnitudes are determined from duration (FMAG) or amplitude (XMAG) readings. Main criteria for evaluation of the accuracy of estimated hypocenters are the errors in position (ERH and ERZ) and root mean square (rms) error. 4.2.5.3. Results and statistical analysis The whole number of 1’412 processed events for the time span 14.06.2008 – 09.10.2008 is presented in Fig.4.5.a. About 70% of the data were processed by me and the rest ~30% by several students in PSL. The aftershock locations are distributed in NE-SW direction which is similar to one of the two nodal fault plane strike component of main shock focal mechanisms suggested by the seismological sources (Fig.4.3). For 32 of the processed earthquakes, ERH and ERZ could not be calculated because of insufficient data, and for 7 and 77 events the horizontal and vertical errors, respectively, are bigger than 10 km. These earthquakes are eliminated from further investigation (see Fig.4.4.a). Another 43 earthquakes are excluded from the analyses due to insufficient wave readings – less than 7 (Fig.4.4.b). Two attempts for reducing the resulting series of 1’256 aftershocks (initial group) on the base of the accuracy are proposed. The first one consists of the precisely located events named from now on group1 embraces 1’108 shocks and represents 88% of the initial group (Fig.4.5.b). Group 2 has 584 records represents – only 46% of the initial group (see Fig.4.4 a and c – solid line) but with higher precision. In both cases the linear trend of the aftershock sequence is very well outlined. In spite of that the absence of the aftershock cluster close to the epicentre from group 2 leads to the decision that, from now on, all analyses and conclusions will be made only for group 1 which is assumed to be more representative and with sufficient precision. For better understanding of fault geometry several cross-sections are made through the investigated area (Fig.4.6). Two clusters are detected. One of the them is in the area of the epicentre and the other one, the most impressive, is in the northeastern part of the investigated zone. It is detached from the deeper cluster above it which is situated at a depth of 20-25 km (profile E’-E”). Another characteristic of this shallower group of events is that the aftershock propagation changes its direction in this area a bit to the east compared to the main tendency. The location and the orientation of this cluster coincides with the location of a fault determined by Doutsos and Poulimenos (1992) in Margaris et al. (2008) which is named North Peloponnesus Major Fault (NPMF). In the present investigation the aftershock formation outlines the seismic fault with approximately 24° NE-SW orientation and ~ 25km length and 10 km width (Fig.4.6), which is in very good agreement with formula (5.5). The defined dimensions of the fault are in very good accordance in respect to the empirical deduced values. The upper limit of the aftershock distribution is at about 5 km depth which is, as it is mention in Papadopoulos et al. (2010), the borderline of the sedimentary layer of the crust at which the blind fault was activated. The fault dipping is revealed as approximately vertical – see Fig.4.6 (profiles B’B” and C’C”). If the zone IL_1 is assumed as an approximation of the activated fault then the averaged slip could be calculated (61 cm) taking into account the seismic moment reported by Harvard CMT. Fig.4.5. Maps of recorded aftershock sequence for the period 14 June 2008 – 9 October 2008. (a) The whole series of processed events. (b) Previous sequence reduced to group 1 – events with ERH/ERZ less than 2 km and rms less than 0.3 sec. (c) Series of events from (a) reduced to group 2 – the most precise determined events with ERH/ERZ less than 1 km and rms less than 0.2 sec Fig.4.4. Aftershock grouping after accuracy criteria. (a) Distribution of horizontal/vertical (ERH/ERZ) errors. The hatched zone represents the events with errors more than 10 km which are excluded from further investigations. (b) P&S wave readings distribution. The events calculated from less than 7 readings are excluded. (c) Distribution of rms residuals. Dotted-line rectangle in (a) and (b) limits group 1; solid-line square in (a) and (c) limits group 2. IV. The Movri earthquake (NW Peloponnesus), 8th of June 2008 30 Fig.4.6. Cross-sections through the investigated fault. Red stars represent the main shock location according to different sources. IV. The Movri earthquake (NW Peloponnesus), 8th of June 2008 31 IV. The Movri earthquake (NW Peloponnesus), 8th of June 2008 32 Fig.4.7. Cross-section along the strike of the fault showing different approximations for the fault geometry (PD10 – after Papadopoulos et al. 2010, F10 – after Feng et al. 2010, JCR – after J.C. Ruegg) and several clusters of the aftershocks (IL_1 – the main cluster, IL_2 – events associated with an activation of the northern system of faults, IL_3 and IL_4 – separate groups of events). The so determined fault lies shallower than the one proposed by J.C. Ruegg (Fig.4.7 – JCR). His representation of the fault is more compatible with the GPS and InSAR data than the other available models (PD10 and F10 in Fig.4.7) that suggest the upper edge of the fault at ~ 5 km which is in discrepancy with the GPS and InSAR data. These models do not explain and do not refer to the geodetic results. A small vertical shift of the seismological data compared with the model from J.C. Ruegg exists but his model is a preliminary and is based only on the Harvard fault plane solution. The second (IL_2) and the third (IL_3) clusters appeared at shallower depths with upper limit at ~ 5 km. While for the IL_3 no clear explanation could be found, the appearance of the IL_2 cluster is explained as an activation of neighbouring structure northward from the main cluster. To confirm this statement a cross-section through the group of the aftershocks is build using the most precise events (group 2). The forth cluster is located in the area of the Peneus Dam at relatively the same depth as the main one, but it is detached. To give more adequate explanation for this activation, an additional investigation of the geology in the area is needed. 4.3. Conclusions The aftershock locations estimated by the microseismic investigation define a rupture zone (seismic fault) of about 25 km length and 10 km width with approximately 25º orientation, the upper depth limit at around 15 km and lower limit at ~25 km which confirms the supposition for a buried fault. Compared with the GPS and InSAR studies which imply upper edge of the fault at around 22.5-25 km the aftershock fault is located shallower at 15 km, which could be a result of the velocity model applied in the processing. The Okada (1992) formulation for rectangular source is applied using the dimensions of the assumed seismic fault (Table 4.1). The strike of 209º is in concordance with the aftershock distribution and the Harvard CMT solution. The dip is assumed to be 90º since the transverse profiles (Fig.4.6) showed approximately vertical distribution of the aftershocks. Models of the vertical and horizontal displacement fields are presented in Fig.4.8. A synthetic descending interferogram which presents the line of sight (LOS) displacements is also shown in the figure. IV. The Movri earthquake (NW Peloponnesus), 8th of June 2008 33 Table 4.1. Parameters for the model of the fault lat* (N) long (E) 40.034 20.713 depth of the UE (km) 15 strike (°) dip (°) rake (°) L (km) W (km) u (cm) Mo (Nm) Mw 209 90 164 25 10 61 4.56e+18 6.4 * lat, long – coordinates of the centre of the upper edge (UE) of the fault model; L – surface rupture length; W – down-dip rupture width, u – average slip. The displacement components for the NOA GPS permanent station RLS gained from this model are as follows: dE = 1, dN = 6, dU = 6 mm, which is in good accordance with the results received from the GPS measured time series (dE = 0±2, dN = 6±2, dU = 6±2 mm). The synthetic interferogram contains one incomplete fringe of LOS displacement. Such deformation is not detected in the ENVISAT descending interferogram which could be explained with the dense vegetation in the area as an obstacle for the C-band sensor of the ENVISAT. The aftershock distribution reveals a possible post-seismic activation (around 20 July 2008) of a secondary fault at a shallower depth. An additional cluster of aftershocks at upper limit of the depth at ~10km with ~E-W orientation, north-eastward of the main group of events is located in the area of North Peloponnesous Major Fault (NPMF). Fig.4.8. Elastic dislocation model (Okada 1992) of the vertical and horizontal (east and north components) displacements and descending synthetic interferogram calculated with the assumed fault approximation after the microseismic data analysis. The white line represents the fault location. Conclusions and perspectives 34 5. Conclusions and perspectives The main objective of the present work was to determine the crustal deformations caused by several moderate-to-strong earthquakes occurred in the last fifteen years in the area of the Eastern Mediterranean. To detect the surface deformations the InSAR method is applied and subsequent modelling deformation fields is performed. When other types of data and information were available, a comparative analysis was made. In the course of the work some limitations of the methods used were pointed out. Since the InSAR technique did not give clear information about the source of the last investigated earthquake, additional seismological study and analysis of the aftershocks is performed. The first target of the investigation was the Konitsa moderate earthquakes sequence (Chapter II) occurred in 1996 with a magnitude of the three main shocks of about Mw 5.0-5.3. The area was imaged by the ERS sensor through two consecutive ascending passes separated by a period of 35 days when the main earthquakes occurred. In the resulting interferogram the deformation pattern of the surface has been revealed as consisting of two fringes corresponding to a 5.6 cm deformation in the line of the satellite direction. Similar results appeared in the descending interferometric pattern gained from the descending images separated by approximately one year. The high level of coherence of the ascending interferogram allowed performing a proper modelling of the fault responsible for the earthquake sequence. The information from the descending pass confirmed the quality of the fault plane solution. The field of deformation predicted by the model has a very good congruence with the observed deformations. The maximum of the vertical deformations is assessed as a subsidence in the range of ~8 cm. More detailed study of the influence of the atmospheric artefacts on the deformation pattern could be accomplished in the future. The study of this seismic sequence contributes to the knowledge of the potentialities of the InSAR method for measuring the crustal deformations caused by moderate-size earthquakes. Comparison with the similar case of the Valandovo earthquake showed that the depth of the hypocenter and the local geological conditions could strongly influence the capability of the method. In the latter case no deformations have been detected mainly because of the deeper location of the seismic source. The Lefkada earthquake from 2003 which occurred with larger magnitude (Mw 6.2) was the second subject of the dissertation (Chapter III). For detection of the crustal deformation, ENVISAT/ASAR images are used. The received interferograms have a high level of decoherence mainly due to the dependence of the radar wavelength (C-band) on the vegetation coverage, since almost the entire island is densely vegetated. Nevertheless, a modelling of the fault is performed using the available fringe sections and the fault plane solution of the earthquake reported by the Harvard CMT. In order to improve the quality of the interferometric pattern and to determine the field of deformations after the Lefkada seismic event, an averaging of the calculated interferograms was made. The stacked interferogram delineated the range of the maximum crust displacement as an uplift of ~6 cm in the central-west part of the Lefkada Island. The large InSAR data set (ascending and descending images) available for the 2003 Lefkada event can be used in future to investigate eventual post-seismic displacements. Although, such displacements are not visually determined the construction of time-series of proper points in the area could confirm the absence of post-seismic relaxation of the crust. More detailed investigation could be performed for the mainland area north of the island, where some evidences for subsidence processes were revealed in the processed interferograms but were not discussed in this dissertation. The depth of the seismic source as a limitation to the InSAR method appeared also in the case of the earthquake from 2008 in the Movri Mountain (the Northwestern Peloponnesous) (Chapter IV). This event occurred in an area free of significant historical seismic event. There was no information about major fault structures with the orientation of the 2008 aftershocks. Despite that the event was strong (Mw 6.3) no significant surface deformations were detected by Conclusions and perspectives 35 the ENVISAT/ASAR and ALOS/PALSAR radars, as well as no surface ruptures with the orientation of the main shock’s strike and the aftershock distribution were found by in situ geological investigations. The modelling of the fault based on the Harvard solution and the geodetic data proposes a depth of the upper edge of the fault at ~22.5 km. To complement the study for the event, the microseismic activity following the main shock was analysed. Clear clustering of the aftershocks was differentiated at a depth between 15 and 25 km thus confirming the activation of a buried fault. The main group of aftershocks spreads in length of around 25 km. These dimensions were used to define a fault approximation according the Okada formalism. This model predicts displacements which are compatible with the geodetic results. Furthermore, an activation of a neighbouring fault structure at a shallower depth is found. Involving the microseismic analysis, we demonstrated the power of multidisciplinary investigations in geodynamics. On the other hand, the regional view of the major geotectonic structures and the knowledge received for the Lefkada Segment from CTFZ (Chapter III) supports the idea of a separate small tectonic structure with the eastern border in the area of Movri earthquake (Chapter IV). The existence of a Central Ionian Islands Block comprising also NW Peloponnesus is proposed by Le Pichon et al. (1995). Later this statement is confirmed by Cocard et al. (1999), Kiratzi and Louvari (2003), and Konstantinou et al. (2009), where this structure is described as a small clockwise rotating tectonic block. The examined block comprises the Central Ionian Islands, part of the Greek mainland (the Akarnania region) and NW Peloponnesus (Fig.5.1). The Akarnania region is suggested by Haslinger et al. (1999), supported by Cocard et al. (1999) and Hollenstein et al. (2008a), to be a separate block unit bordered by the Gulf of Arta to the north, the Katouna fault zone to the east and the Lefkada Island to the west. In the present study, the idea that the Akarnania region is part of the bigger Central Ionian Islands Block is adopted and its boundaries are outlined as following the Cephalonia fault zone to the west, the Amvrakikos fault at the homonymous gulf (or the Gulf of Arta) to the north, and the Katouna fault to the northeast. Fig.5.1. An approximation of the idea about the Central Ionian Block. The Red stars represent the 2008 Movri Mountain and 2003 Lefkada Island earthquakes. The belonging of the shaded area to this block is disputable Conclusions and perspectives 36 The main attainment of the present work is the determination of the field of crustal deformation caused by the 1996 Konitsa earthquake sequence that occurred in three main events with a moderate magnitude of 5.0-5.3. The high quality of the results obtained in the study increases the interest to the possibilities of the method for detecting similar events in the area of the Eastern Mediterranean. As it was mentioned in the Introduction, the Eastern Mediterranean and the Balkans in particular suffer from an intensive shallow seismicity. If the larger events are put aside, the number of moderate-sized earthquakes that occurred in the region is still very high. When these events are limited by some criteria potential candidates for an InSAR investigation could be found. By defining the limits of the magnitude between 5 and 6, the maximum depth of hypocentre to 15 km and the time of earthquake occurrence in the time span of the SAR satellites, a set of suitable for such analysis earthquakes could be chosen. According to the Harvard CMT database ~ 90 events satisfy these terms (Fig.5.2). This is 60% of the earthquakes selected by the same criteria for the whole Europe. It should be emphasised that many of the events in the Eastern Mediterranean region occurred far in the sea and the crust deformations cannot to be detected by InSAR. Fig.5.2. Review of the moderate-size (5 < M < 6) earthquakes in the Balkans for the period 1992-2010 (after the start of ERS mission till now) according to the Harvard CMT catalogue General information 37 Information about the scientific contributions of the dissertation 1.By applying a new for Bulgaria geodetic method (InSAR) the fields of deformations of the crust are determined for the earthquakes from 1996 close to town of Konitsa, Greece and from 2003 offshore Lefkada Island, Greece. 2.On the base of the gained geodetic and the available seismological information, dislocation models for the same earthquakes are proposed. 3.The parameters of the fault which corresponds to the earthquake and the following aftershock series from 2008 in the region of Northwestern Peloponnesous, Greece are determined, using microseismic data and analysis of geodetic and geological investigations. 4.New facts in confirmation of the hypothesis of the existence of Central Ionian block as separate tectonic structure are deduced. Papers in the topic of the dissertation Ilieva, M., P. Briole, D. Dimitrov. InSAR detection of crustal deformation caused by the JulyAugust 1996 Konitsa, Greece, earthquake sequence. (in preparation) Ilieva, M., P. Briole, A. Ganas, D. Dimitrov, P. Elias, A. Mouratidis, R. Charara. InSAR investigation and fault plane modelling of 14th of August 2003 Lefkada Island (Greece) earthquake. (in preparation). Ilieva, M. (2009). SAR Interferometry application for geodynamic investigations. Annual of the University of Architecture, Civil Engineering and Geodesy, Sofia, International Conference UACEG2009: Science & Practice., 29-31 October 2009, Vol. XLIV, Fasc. III, 2009. Ilieva, M. (2008). InSAR application for geodynamic studies – the 14 August 2003 Lefkada earthquake. Geodesy 21, Special issue: 60 anniversary of Central laboratory of geodesy, CLG – BAS, Sofia, 121-127. Reports and posters at scientific conferences in the topic of the dissertation Ilieva, M., R. Charara, P. Briole (2008). Analysis of the deformations of Western Corinth Rift by satellite geodesy. Journées des Doctorants, Ecole Doctorale: Géosciences et Ressources Naturelles. Ecole des Mines de Paris, Paris, France, 6 et 7 mai 2008 (poster). Ilieva, M., R. Charara, P. Briole (2008). Analysis of the deformations of Western Corinth Rift by satellite geodesy. 4th Earth Observation Summer School – Earth System Monitoring and Modelling, European Space Agency, Frascati (Rome), Italy, 4-14 August 2008 (poster). Briole, P., Armijo, R., Avallone, A., Bernard, P., Charara, R., Deschamps, A., Dimitrov, D., Elias, P., Grandin, R., Ilieva, M., Lambotte, S., Lyon-Caen, H., Meyer, B., Mouratidis, A., Neecessian, A., Papanastassiou, D., Ruegg, J.-C., Sokos, E., Sykioti, O. (2008). Multidisciplinary study of the June 8, 2008, Mw=6.4 Andravida earthquake. 31st General Assembly of the European Seismological Commission in Hersonissos, Crete, Greece, 7-12 September, 2008 (poster). Elias P., P. Briole, M. Ilieva, A. Avallone, R. Charara, A. Belehaki, O. Sykioti, A. Mouratidis, T. Herekakis (2008). PALSAR/ALOS interferometry in the Gulf of Corinth and Partas (Greece). Comparison with ERS and ASAR/ENVISAT results and GPS measurements. ALOS 2008 Symposium, ESA, 3-7 November 2008, Rhodos, Greece (poster). General information 38 Ilieva, M. (2008). InSAR application for geodynamic studies – the 14 August 2003 Lefkada earthquake. 60 години Централна лаборатория по висша геодезия. ЦЛВГ – БАН, София, 12 ноември 2008. Ilieva, M., P. Elias, P. Briole, D. Dimitrov (2009). InSAR application for geodynamic study of the 14 August 2003 Lefkada, Greece earthquake. European Geosciences Union (EGU) General Assembly 2009. Vienna, Austria, 19 – 24 April 2009 (poster). First BALGEOS-II (BALkan GEodetic Observing System) workshop. University of Ss Cyril and Methodius, Skopje, Macedonia, 28-30 September 2009. Ilieva, M. (2009). SAR Interferometry application for geodynamic investigations. International Conference UACEG2009: Science & Practice. UACEG, Sofia, Bulgaria, 29-31 October 2009. Ilieva M., P. Elias, P. Briole, D. Dimitrov (2009). InSAR investigation of shallow seismicity in the Balkans. Colloque 2009 du Groupement de Recherche G2 ("Géodésie et Géophysique"). Institute of Earth Physics, Strasbourg, France, 18-20 November 2009 (poster). Ilieva M., P. Elias, P. Briole, D. Dimitrov (2009). InSAR investigation of shallow seismicity in Greece and the Balkans. FRINGE 2009 workshop: Advances in the Science and Applications of SAR Interferometry. ESA ESRIN, Frascati, Italy, 30th November - 4th December 2009 (poster). Serpetsidaki, A., M. Ilieva, E. Sokos, P. Elias, P. Bernard, P. Briole, H. Lyon-Caen, A. Tselentis (2010). The 8 June 2008 Andravida earthquake, Ms=6.3 and its kinematics and tectonic settings. European Seismological Commission (ESC) 32nd General Assembly, September 6-10, Montpellier, France (poster). Abstract in Bulgarian language 39 Резюме Главните цели на дисертацията са определяне, чрез радарна интерферометрия, на полетата на деформация на земната кора, причинени от няколко специфични плитки земетресения в Източното Средиземноморие, извеждане на модели на преместването, използвайки формулировката на Окада и очертаване на някои тектонски структури в изследвания район. За случаите, когато са налични допълнителни данни, като GPS или други, са представени сравнения и комплексни анализи. Две от изследваните събития, а именно групата от земетресения от юли-август 1996 край Коница и Валандовското земетресение от май 2009, са с по-малка сила с магнитуд между 5.0 и 5.3 и тяхното детектиране чрез InSAR е сериозно предизвикателство. В случая на земетресението при Коница (на границата между Гърция и Албания), деформацията на земната кора e много ясно видима в получената интерферограма и се проявява като два интерферометрични кръга, отговарящи на преместване от поне 5.6 cm. В сравнение с този случай, събитието край Валандово (северно от Дойранското езеро) има подобни магнитуд и параметри, но не е засечено от интерферометричния радар, главно поради по-дълбокото положение на ралома и може би заради раличната геоложка структура в района. От друга страна, другите два изследвани шока, земетресението край остров Лефкада от 2003 и това в Северозападен Пелопонес през 2008, имат по-голям магнитуд, съответно 6.3 и 6.4. Въпреки по-голямата сила на тези събития, те имат свои собствени особености, поради които будят интерес. Специфичната ситуация, свързана с Лефкадското събитие е, че основният шок е в морето близо до крайбрежието на острова и предизвикателството идва от обстоятелството, че трябва да се създаде и интерпретира интерферограма за малък остров заобиколен от големи повърхности с ниска отражателна способност. Получените интерферограми са повлияни от специфичните условия и съдържат големи декорелирани области. Въпреки това, се отчита повърхностна деформация от около 5.6 cm. Последното от изследваните земетресения, това от юни 2008 в Северозападен Пелопонес (в района на планината Моври), независимо от големия си магнитуд, не е причинило значителни разкъсвания на земната повърхност според полевите проучвания. С оглед на факта, че InSAR резултатите не дават задоволителни доказателства за източника на силното земетресение, основната част на проучването е пренасочено към анализиране на сеизмичната активност, записана за четирите месеца последвали основния шок. Резултатите потвърждават по-дълбокото разположение на основния разлом при дълбочина на горния край на разлома на около 15 km. В допълнение се откриват данни за възможно активизиране на съседна структура. Полученото ново познание за разломните структури в района на Лефкада и Моври потвърждава хипотезата за съществуване на отделна тектонска структура – блок на Централните Йонийски острови, обхващаща Северозападен Пелополес, района на Акарнания в континенталната част от Гърция и Централните Йонийски острови. Ключови думи: деформации на земната кора, InSAR, моделиране на разломи, Източно Средиземноморие Acknowledgments 40 Acknowledgments I would like to express my gratitude to my supervisors, Dr. Pierre Briole and Assoc. Prof. Dimitar Dimitrov, for their guidance and support during my study. Special thanks to Associate Professor Vassil Vulchinov from the University of Architecture, Civil Engineering and Geodesy, Sofia, for encouraging me to pursue scientific career. I would like to thank the entire staff of the Laboratory of Geology, Department of Geosciences, Ecole Normale Supérieure, Paris, France, the Department of Geodesy, National Institute of Geophysics, Geodesy and Geography – Bulgarian Academy of Sciences, Sofia, Bulgaria and the Seismological Laboratory, University of Patras, Greece, for their help and advices for my work, and especially Dr. Hélène Lyon-Caen, Dr. Marie-Pierre Doin, Assist. Prof. Efthimios Sokos and Assoc. Prof. Ivan Georgiev. In addition my thanks go to Pierpaolo Dubernet and Baptiste Mulot for providing the so important for my work IT support, Françoise Larincq, Dovy Tristani and Patricia Zizzo for the administrative assistance, Dr. Dragan Chobanov, Dr. Antonios Mouratidis and Dr. Stefan Patchedjiev for the English corrections, reviewing and remarks, Eugénie Perouse for the French translation. I gratefully acknowledge the scholarship provided by the Ministry of National Education (France) and the Ministry of Education, Youth and Science (Bulgaria), as well as the bilateral project between the two ministries RILA (3/29, 2007-2008). Thanks are also due to Dr. Antonios Mouratidis, Panagiotis Elias, Krassimir Matev, Dr. Rana Charara, Dr. Penelope Lopez-Quiroz, Assist. Prof. Ivaylo Radev, Dr. Georgi Marinov, Dimitar Georgiev, Carlo Barbieri, Dr. Olivier Cavaliй, Dr. Aurore Franco, Jean Roger for their friendship, encouragement and inspiring discussions throughout the hardship of the doctorate. I cannot forget to thank Vassil Gramov for his hospitality and help during my stay in Paris. Finally, I would like to thank all my friends for their continued support and patience, and whole my family – my parents, Blagoy Iliev and Lyubka Ilieva, Hristo, Ralitsa, Boyan, Sonya, Zoya, also Peter Chobanov and Regina Chobanova, Inna, Lili, Miro, and especially to the biggest helper and supporter, Dr. Dragan Chobanov. Thank you all for helping me and supporting me throughout this adventure. Благодаря! Merci! Thank you! Ευχαριστίες! Gracias! Grazie! !شكر