The imaging of the interior of the Earth using ground motion data, or seismic tomography, has been a subject of great interest for over a century. The full elastic wave equations are not typically used in standard tomography codes. Instead, the elastic waves are idealized as rays and only phase velocity and travel times are considered as input data. This results in the inability to resolve features which are on the order of one wavelength in scale. To overcome this problem, models which use the full elastic wave equation and consider total seismograms as input data have recently been developed. Unfortunately, those methods are much more computationally expensive and are only in their infancy. While the finite element method is very popular in many applications in solid mechanics, it is still not the method of choice in many seismic applications due to high pollution error. The pollution effect creates an increasing ratio of discretization to best approximation error for problems with increasing wave numbers. It has been shown that standard finite element methods cannot overcome this issue. To compensate, the meshes for solving high wave number problems in seismology must be increasingly refined, and are computationally infeasible due to the large scale requirements. A new generalized least squares method was recently introduced. The main idea is to select test spaces such that the discrete problem inherits the stability of the continuous problem. In this dissertation, a discontinuous Petrov-Galerkin method with optimal test functions for 2D time-harmonic seismic tomography problems is developed. First, the abstract DPG framework and key results are reviewed. 2D DPG methods for both static and time-harmonic elasticity problems are then introduced and results indicating the low-pollution property are shown. Finally, a matrix-free inexact-Newton method for the seismic inverse problem is developed. To conclude, results obtained from both DPG and standard continuous Galerkin discretization schemes are compared and the potential effectiveness of DPG as a practical seismic inversion tool is discussed. / text

This paper presents results of a seismic tomography experiment carried out at the Bullseye cold vent site offshore Vancouver Island. In the experiment, a seismic air gun survey was recorded on an array of five ocean bottom seismometers (OBS) deployed around the vent. The locations of the shots and the OBSs were determined to high accuracy by an inversion based on the shot travel times. A three-dimensional tomographic inversion was then carried out to determine the velocity structure around the vent, using the localized source and receiver positions. The inversion indicates a relatively uniform velocity field around and inside the vent. The velocities are close to the values expected for sediments containing no hydrate, which supports previous claims that the bulk concentrations of gas hydrates are low at the site. However, the largest resolved velocity anomalies of + 25 m/s are spatially within the limits of the acoustic blank zone seen in multichannel seismic data near the Bullseye vent. The velocity inversion is consistent with zones of high concentration (15-20 % of the pore space) in the top 50-100 m of sediment.

Thesis (Ph.D.)--University of Western Australia, 2007.

- / -

Bien que de nombreuses études géodynamiques et tectoniques aient été effectuées à partir l'activité sismique en Equateur, il n'existait pas à ce jour une tomographie complète utilisant l'ensemble des données du réseau sismologique Equatorien (RENSIG), mise à part une étude prélimaire sur la partie centrale de l'Equateur menée en 1994 par Prévot et coll. et de plusieurs profils sismiques déterminés à la suite des campagnes marines SALIERI et SISTEUR. Inverser les centaines de millier de temps d'arrivées d'ondes P et S, de qualité inégale, formant le catalogue RENSIG était le défi qu'a constitué le sujet de cette thèse.Nous décrivons comment nous avons complété le catalogue RENSIG par des données provenant du Nord du Pérou et comment nous avons homogénéisé et filtré l'ensemble de données résultant, comportant plus de 800 000 temps d'arrivée correspondant à plus de 50 000 séismes. Pour inverser ces données nous avons adopté une approche Bayésienne. Nous montrons comment le problème peut être reformulé dans un contexte Gaussien par un changement de variables, tout en imposant une statistique robuste aux données, qui conduit à un problème de moindre carrés non linéaire. Nous détaillons particulièrement la régularisation du problème au travers des noyaux de covariance qui conduit à définir des paramètres de contrôle fort utils pour l'inversion. Nous montrons également qu'inverser des différences de données revient à introduire des termes spécifiques de corrélation dans la matrice de covariance des données, tout en conservant les données brutes. Nous indiquons finalement comment le calcul de l'indice de restitution permet de définir une zone de confiance du modèle résultant de l'inversion.L'inversion a été menée pratiquement en utilisant les codes informatiques (en Fortran 2003 par B. Potin, B. Valette, V. Monteiller): LOCIN (localisation) et INSIGHT (tomographie). La région finale d'étude est constituée par une boite parallélipipédique de dimension 590$times$770 km$^2$ de base et de 244 km de hauteur qui contient la topographie de la surface. Le modèle est constitué d'une part des valeurs de $v_P$ et $v_P/v_S$ sur une grille ayant 5 km de pas horizontal et 2 km de pas vertical et, d'autre part, des paramètres d'identification spatiale et temporelle des séismes. Un ensemble de tests nous a permis de déterminer des valeurs raisonables de ces paramètres au travers d'un analyse de type courbe en L.Nous avons obtenu une amélioration de la localisation de la sismicité, qui nous a permis de mieux décrire les essaims superficiels comme ceux de Pisayambo, Macas et du Reventador et d'identifier des linéaments en relation avec la Tectonique. Nous avons également obtenu une image de la sismicité à profondeur intermédiaire qui est dominée par la présence de 4 nids sismiques, ceux de Madonaldo, La Man'a et de Guayaquil à des profondeurs entre 75 et 115 km et celui de Puyo à de plus grandes profondeurs. La zone de Wadati-Benioff nous a permis de définir la profondeur du slab jusqu'à des profondeurs de 100-150 km en fonction de la latitude et d'observer la décroissance du pendage de 25° environ au nord et au centre de l'Equateur jusqu'à environ 10° au sud puis au nord du Pérou. Par ailleurs, l'analyse du champ de vitesse des ondes P suggère fortement que le slab est coupé en deux morceaux, le morceau sud passant sous le morceau nord au niveau du nid sismique de Puyo. Le modèle $v_P/v_S$ présente une forte anomalie positive de ce rapport le long de la cordillère occidentale à des profondeurs entre 30 et 50 km qui caractérise des matériaux partiellement fondus et correspond au réservoir d'alimentation profond de l'arc volcanique. Enfin, nous avons déduit de notre modèle un modèle de profondeur de Moho en prenant la profondeur de maximum de la norme du gradient de vitesse entre les vitesses de 7.2 et 7.4 km/s et en incorporant l'information sur la profondeur de Moho provenant des campagnes SALIERI et SISTEUR dans la marge active. / Although there have been numerous studies on the geodynamics and the tectonics in Ecuador based on the seismic activity, there has not been to date a comprehensive tomography study using the entire database of the National Seismic Network (RENSIG). Only a preliminary limited study was performed by Prevot et al. to infer a simple P velocity model in central Ecuador, and several profiles in the South-Colombian-Ecuador margin were also investigated by using travel time inversion of wide-angle seismic data obtained during the two marine experiments SISTEUR and SALIERI. Inverting the hundreds of thousands of arrival times of P and S waves of uneven quality that constitutes the RENSIG catalogue is the challenging subject of this thesis.We describe how we complemented the RENSIG catalogue with data from the Northern Peru network and how we homogenized and filtered the resulting dataset of more than 800 000 first arrival times of P and S waves corresponding to more than 50 000 earthquakes. To invert these data for both the velocity models and the event locations we adopted a Bayesian approach. We show how the problem can be recast in the Gaussian framework by changes of variable while imposing a robust statistics to the data, and how it leads to a generalized nonlinear least squares problem. We detail in particular the regularization of the models through the smoothing and damping properties of the covariance kernels. We also show that inverting differences in data instead of the raw data amounts to the introduction of specific correlation terms in the data covariance matrix, while keeping the same set of data. We finally indicate how the computation of the averaging index allows the delimitation of a confidence region for the resulting model.The practical inversion has been carried out by using the two Fortran 2003 codes (B. Potin, B. Valette, V. Monteiller): LOCIN (prior localization) and INSIGHT (tomography). The final study region is a parallelepipedic box of 590$times$770 km$^2$ area and 244 km height that contains the topography of the surface. The models consist of the $v_P$ and $v_P/v_S$ fields discretized over a grid, the spacing of which is 5 km in the horizontal directions and 2 km in the vertical one, and of the spatial and temporal parameters of the seismic events. A battery of tests allowed us to set reasonable values for these tuning parameters through an L-curve analysis.We obtained the spatial distribution of the seismicity with an improved accuracy which allows us to describe with more details the shallow seismic clusters, as those of Pisayambo, Macas, Reventador, and to identify lineaments in the seismicity in relation with tectonics. We obtained also a clear image of the intermediate depth seismicity wich is dominated by 4 nests, namely the Maldonado, La Man'a, and Guayaquil nests, at depths ranging between 75 km and 115 km, and the Puyo nest at much deeper depths. The Wadati-Benioff zone allowed us to clearly defined the topography of the slab only to a depth to about 110-150 km, depending on the latitude, and to observe the decrease of the dip angle from about 25° in northern and central Ecuador down to about 10° in southern Ecuador and northern Peru. On the other hand, the analysis of the P velocity clearly suggests that the slab is broken in two pieces, the southern one passing under the northern at the level of the Puyo nest. The $v_P/v_S$ model presents a high anomaly of the ratio along the western cordillera at a depth ranging between 30 km and 50 km that characterized partially melted rocks and corresponds to the feeding reservoir of the volcanic arc. Finally, we deduced the Moho depth from our model by taking the depth for which the norm of the velocity gradient is maximum between 7.2 and 7.4 km/s and by incorporating information on the Moho depth provided by the SISTEUR and SALIERI experiments in the convergent margin.

xv, 151 p. : ill. (some col.) A print copy of this thesis is available through the UO Libraries. Search the library catalog for the location and call number. / To explain the origin of several distinct aspects of the Galápagos volcanic hotspot, such as the broad geographical extent of recent volcanism and the unusual pattern of geochemical anomalies, we conducted seismic tomography studies of the upper mantle and crust beneath the Galápagos Archipelago. The studies combine measurements of group and phase velocities of surface waves and delay times of body waves. We find that upper mantle seismic velocities are lower than those beneath other regions of comparable age in the Pacific and consistent with an excess temperature of 30 to 150°C and ∼0.5% melt. We attribute the excess temperature and presence of melt to an upwelling thermal mantle plume. Crustal seismic velocity is up to 25% lower than that of very young crust at the East Pacific Rise (EPR) and is comparable to that of Hawaii, which we attribute to heating by increased intrusive activity above the Galápagos plume and the construction of a highly porous volcanic platform. In addition, we find that the Galápagos hotspot is underlain by a high-velocity region whose thickness varies from 40 to 100 km. The tomographic images reveal that the upwelling mantle plume tilts northward (towards the nearby Galápagos Spreading Center) as it rises and then spreads laterally when it reaches the bottom the lid. The lid, which we attribute to residuum from melting, is thickest where it is farthest from the spreading center, suggesting that ridge processes may affect the generation and amount of thinning of the residuum layer. In addition, the thickness of the lid correlates well with the geographical pattern of geochemical anomalies of erupted lavas, suggesting that the lid may control the final depth of decompression melting. We conclude that many of the distinct characteristics of the Galápagos can be attributed to the interaction of the upwelling plume with the lid and the nearby ridge. We further suggest that the ridge affects the geometry of plume upwelling in the upper mantle and also the pattern of lateral spreading of the plume due to its effect on the thickness of the residuum layer. This dissertation includes previously published co-authored material. / Committee in charge: Dr. Douglas R. Toomey, Chairperson; Dr. Eugene Humphreys, Member; Dr. Emilie Hooft Toomey, Member; Dr. Paul Wallace, Member; Dr. John Conery, Outside Member

This thesis applies three-dimensional (3-D) non-linear seismic tomography to image crustal/upper mantle structure of S.W. British Columbia and N.W. Washington. Two tomographic inversions are carried out including high-resolution imaging of upper crustal structure using controlled source data, and deeper imaging by simultaneous inversion of controlled source and earthquake data. Non-linear first arrival travel-time tomography is applied to controlled source data from the Seismic Hazards Investigation of Puget Sound (SHIPS) experiment conducted in 1998. Nearly 175,000 first arrival travel-times are inverted to obtain a minimum structure upper crustal velocity model to a depth of 12 km with a cubical cell size of 1 km. Results from checker-board tests for this velocity model indicate a lateral resolution of 20 km and above. The main geological and structural features in the study area are well defined by this velocity model. The structural outline of the sedimentary basins in the Straits of Georgia and Juan de Fuca are distinctly mapped. The Crescent Terrane is mapped beneath southern Vancouver Island with velocities up to 7 km/s that correlate well with the presence of gabbro in the subsurface. The northwest-southeast structural trend observed in the Strait of Georgia correlates with the observed seismicity. Shallow seismicity observed at the southern tip of Vancouver Island correlates with the location of the Leech River Fault. An earthquake tomography algorithm was developed for joint estimation of hypocentral and velocity parameters, and tested on a synthetic data set. Using this algorithm, tomographic inversion was performed simultaneously on earthquake and controlled source data from southwestern British Columbia and northwestern Washington. Approximately 15,000 first arrivals from 1,400 earthquakes and 40,000 first arrivals from the SHIPS experiment were simultaneously inverted for hypocentral parameters and velocity structure. Model resolution studies indicate a lateral resolution of 30 km and above. Upper-crustal earthquakes close to southern Vancouver Island correlate with the velocity contrasts associated with the Leech River, Southern Whidbey Island, and Darrington-Devils Mountain faults. Three mafic to ultramafic high velocity units are identified at approximately 25 km depth beneath the Crescent Terrane and above the subducting Juan de Fuca crust. The continental crust and subducting Juan de Fuca crust and mantle are well mapped. The transition zone to continental mantle occurs at 35 km depth beneath the eastern Strait of Georgia. The slab seismicity beneath the Strait of Georgia at depths >65 km lies below a low velocity zone mapped in the mantle wedge at depths of about 45–55 km. This low velocity zone may be indicative of the presence of fluids released during the phase change from basalt/gabbro to eclogite in the subducting slab. / Graduate

La collision alpine a créé des structures complexes comme des chaînes de montagnes très arqués et des interactions compliquées entre les slabs subduits. La polarité de subduction est inversée à la transition entre les Alpes et les Apennins et les Alpes et les Dinarides. Le fait que la plaque Adria subducte en même temps vers l'ouest et vers l'est avec un fort pendage, presque verticalement, suggèrent une flexion importante de cette plaque. Notamment, si on considère de plus la proposition qu'Adria subducte aussi vers le nord sous les Alpes de l'est, ce qui est toujours sujet de discussion. Des déchirures dans le slab adriatique sous les Dinarides du nord à plus de 150 km de profondeur et sous les Apennins à moins de 200 km, pourraient être des signes d'une forte tension et, en conséquence, un détachement de la plaque adriatique. La plaque européenne pourrait aussi avoir subi plusieurs déchirures le long des Alpes. Cette hypothèse sujette à débat nécessite de nouveaux modèles tomographiques. Le modèle tomographique présenté dans cette thèse se base sur les vitesses de phase des ondes de surface pour donner un modèle 3-D à haute résolution des vitesses de cisaillement de la surface jusqu'à 200 km de profondeur. Ce modèle est unique de par sa haute résolution dans le manteau lithosphérique où des modèles antérieurs montrent de fortes incertitudes. Afin d'imager la croûte et le manteau supérieur en même temps, une combinaison des données de vitesses de phase des ondes de surface mesurés à partir des bruits ambiants ainsi que des séismes est utilisée dans cette thèse.Pour tester la validité de cette procédure, une comparaison détaillée des mesures de vitesses de phases et des structures imagées avec les deux méthodes est présentée. De l’analyse résulte un faible biais qui montre des vitesses plus élevées avec les données se basant sur des séismes par rapport aux données se basant sur le bruit ambiant. En comparant avec des travaux antérieurs, il est apparu que ce biais est dû à une différence méthodologique. Plusieurs paramètres qui pourraient influencer les mesures du bruit ambiant sont testés numériquement. Une cause unique n'a pu être identifiée. L'explication la plus probable pour le biais est une combinaison entre différentes sensibilités des méthodes aux structures et l'influence des modes supérieurs. Néanmoins, l'écart est suffisamment faible par rapport aux variations structurales pour être négligé.Un modèle final de vitesse de cisaillement de la région alpine est obtenu avec une résolution latérale d'environ 25 km dans la croûte peu profonde. Les tests synthétiques donnent une résolution approximative de profondeur estimée à 2 km près de la surface et de 5 km à la profondeur du Moho. Dans le manteau supérieur, la résolution baisse rapidement mais les structures principales des panneaux plongeants restent bien imagées jusqu'à une profondeur de 200 km le long des Alpes et des Apennins.La partie crustale du modèle donne des informations à haute résolution sur la taille et la profondeur des bassins sédimentaires et du corps d’Ivrée ainsi que sur la profondeur et la structure du Moho. Ce modèle de vitesses de cisaillement est le premier montrant autant de détails et couvrant les Alpes entières, il est proposé que le modèle pourrait servir comme référence pour la région.Le modèle montre les limites des zones de subduction et les régions de basses vitesses asthénosphèriques montants sous les bassins Ligure et pannonien. Des structures connues comme les déchirures de slabs sous les Apennins et les Dinarides sont imagés. Des découvertes supplémentaires ont été mises en évidence : une petite zone de faible vitesse qui coupe la lithosphère au nord des Dinarides est interprété comme l'expression d'une grande faille décrochante... / The plate collision in the Alps and adjacent orogens has created a complex picture of highly arcuate mountain belts and complicated interactions of subduction slabs. The subduction polarity is reversed from European to Adriatic subduction in the transition of the Alps to the Apennines and to the Dinarides. The subduction of Adria both to the west and east and the almost vertical dip of the slabs implies an important flexure of this plate. Even more so if one considers the proposed subduction of Adria also to the north under the eastern Alps, which is still a matter of discussion. Gaps in the Adriatic slab under the northern Dinarides, below 150~km depth and in the southern Apennines above 200~km may be signs of the stresses and the consequent tearing that the Adriatic plate is exposed to.Also the European plate has supposedly undergone one or several break-offs all along the Alpine arc. Especially in the eastern and western Alps it is still an open question whether the European slab is detached below the lithosphere. New tomographic models are thus needed.The herein presented tomographic model is based on surface-wave phase velocities and gives a picture of the shear-velocity structure from the surface to 200 km depth. It is the first high-resolution shear-velocity model of the entire Alpine crust and upper mantle. It is also unique in its good resolution in the lithospheric mantle, where previous body-wave models are subject to high uncertainties. In order to be able to image both crust and upper mantle, a combination of ambient-noise and earthquake-based phase-velocity measurements is used in the present thesis.The validity of this approach is tested by a detailed comparison of the phase-velocity measurements and the structures that are imaged from each method individually. A small bias between the methods results in slightly elevated velocities from earthquake measurements. By comparison with earlier works it appears that this bias is due to methodological differences. Several effects that may influence the ambient-noise records are tested with synthetic experiments, but no unique cause is found. The most likely explanation for the bias between the two methods is a combination of different structural sensitivities and the influence of higher modes. Nevertheless, the discrepancy is sufficiently small with respect to the structural variations that the bias can be neglected.A final shear-velocity model of the Alpine region is obtained which has a lateral resolution in the shallow crust of approximately 25 km. From synthetic tests, the average depth resolution is estimated to be 2~km close to the surface and 5 km for the Moho depth. In the upper mantle the resolution decreases significantly, but main slab structures are well imaged in the central Alps and the Apennines down to the bottom of the model at 200 km depth.Highlights of the crustal part of the model are size and depth of sedimentary basins, the Ivrea body and the Moho structure. Being the first shear-velocity model of this detail and extend it is proposed to serve as reference for the Alps...

The Carboneras fault zone (CFZ, Almería Province, SE Spain) is a major NE-SW trending tectonic lineament that marks part of the diffuse plate boundary between Iberia and Africa. Developed within a basement terrain dominated by mica schist, the fault system comprises two main strands within a complex zone up to 1 km wide. Between these two strands is a braided network of left-lateral strike-slip, phyllosilicate-rich fault gouge bands, ranging between 1 and 20 m in thickness, passively exhumed from up to 3 km depth. The excellent exposure in a semi-arid environment, the wide range of rock types and fault structures represented and the practicality of carrying out in-situ geophysical studies makes this fault zone particularly well suited to verifying and interpreting the results of in-situ seismic investigations. Integration of elements of field study, laboratory analysis and modelling has aided interpretation of the internal structure of the fault zone. Ultrasonic measurements were made using standard equipment over confining and pore pressure ranges appropriate to the upper 10 km of the continental crust. Seismic velocities have also been approximated from modal analysis and mineral phase elastic properties and adjusted for the effects of porosity. In-situ seismic investigations recorded P-wave velocities 40-60% lower than those measured in the laboratory under corresponding pressures and at ambient temperatures for hard rock samples. Fault gouge velocities measured in the laboratory, however, are comparable to those measured in the field because, unlike the host rocks, fault gouges are only pervasively micro-fractured and lack the populations of long cracks (larger than the sample size) that cause slowing of the velocities measured in the field. By modelling the effect of fractures on seismic velocity (by superimposing upon the laboratory seismic data the effects of crack damage) the gap between field- and laboratory-scale seismic investigations has been bridged. Densities of macroscopic cracks were assessed by measuring outcrop lengths on planar rock exposures. Assuming crack length follows a power law relation to frequency, this fixes a portion of the power spectrum, which is then extrapolated to cover the likely full range of crack sizes. The equations of Budiansky and O'Connell (1976), linking crack density to elastic moduli, were used to calculate modified acoustic velocities, and the effects of the wide range of crack sizes were incorporated by breaking the distribution down into small sub-populations of limited range of crack density. Finally, the effect of overburden pressure causing progressively smaller cracks to close was incorporated to predict velocity versus depth of burial (i.e. pressure). Determination of rock physical properties from laboratory analysis and sections constructed from geological mapping provides a representation of velocity from selected parts of the Carboneras fault zone. First break tomography images show particularly well the location of steeply-inclined fault cores, and these correlate generally well with geological mapping and laboratory velocity measurements corrected for the effect of cracks. The decoration of the fault zone with intrusive igneous material is well correlated with the results of geological observations. Comparisons made between the field (seismic) inversion model and laboratory forward velocity model in El Saltador valley show the laboratory and field velocity measurements made within the fault zone can be reconciled by accounting for the effects of crack damage in field data.

Lateral perturbations in anelasticity (Q) and wave speed together provide important constraints on thermal and chemical structures in the mantle. In present-day tomography studies of global wave speed and anelasticity, the significance of 3-D wave speed and 3-D Q structures on surface wave travel times and amplitudes has not been well understood. In this dissertation, the effects of lateral perturbations in anelasticity (Q) and wave speed on surface wave observables are quantified based upon wave propagation simulations in 3-D earth models using a Spectral Element Method. Comparison between phase delays caused by 3-D wave speed structures and those caused by 3-D Q variations show that anelastic dispersion due to lateral perturbation in Q is important in long-period surface wave and can account for 15-20% observed phase delays. For amplitude perturbations, elastic focusing/defocusing effects associated with 3-D wave speed structures are dominant while energy dissipation is important in short-period (~ 50 s) surface waves but decreases quickly with increasing wave period. Anelastic focusing/defocusing associated with 3-D anelastic dispersion becomes more important than wave attenuation in longer period surface waves. In tomography studies, ray theory breaks down and finite frequency effects become important when the length scale of heterogenities are smaller than seismic wavelength. Finite frequency effects in 3-D earth models are investigated by comparing theoretical predictions of travel times and amplitudes with "ground truth" measurements made on synthetic seismograms generated in SEM simulations. The comparisons show that finite frequency effects are stronger in amplitudes than in phases, especially at long periods. / Ph. D.