Three-dimensional slant stacks simulate the response of an elastic medium for both an incident (ray parameter q) and a scattered (ray parameter p) plane wave and are implemented by following a slant stack of common shot gathers by a second slant stack of common ray parameter gathers. Three-dimensional localised slant stacks are developed, and applied to analyse reverse vertical seismic profiles. The application of slant stacks to borehole data differs from the surface seismic case in that the velocities measured from plane wave decomposition are horizontal and vertical phase velocities. Velocity and dip estimates for a model with iso-velocity layers and plane interfaces with any dip are obtained using the slant stacks without explicitly utilizing traveltimes. The estimation scheme is based on a layer stripping approach. The direct arrivals are used to determine the velocity and the reflections are used to determine the dip of the reflector.

A new technique for imaging 3-D prestack seismic reflection data has been developed. This method approximates the Kirchhoff one-pass 3-D prestack depth migration and is based on directly splitting the complex one-pass process into a simpler 2-D x 2-D operation. The first step is 2-D migration in the crossline direction. This is performed as a prestack time migration and creates a "true" 2-D line, provided that the crossline structure is not very complex. The second step, inline migration, is used to determine the velocity function and to obtain the depth image. The kinematics and amplitude behavior of this new scheme closely approximate that of the one-pass process. It is a significantly faster process, and offers a practical alternative for adjusting migration velocities, and building a velocity model. When applied to a field data example, the image created by our new technique compares quite favorably to other available 3-D migration schemes.

Reverse-time migration has proven to be successful for structures with steep dips and strong velocity contrasts. Applying this algorithm to a large scale seismic model requires significant computational expense, particularly if strong velocity contrasts are present in the model. Here I present a layer-stripping migration technique, in which I use the reverse-time method to migrate seismic sections through constant or smoothly varying velocity layers, one layer at a time. As part of the migration in a given layer, the bottom boundary of the layer is defined and a seismic section is collected along it. This new section serves as the boundary condition for migration in the next layer. This procedure is repeated layer by layer. The final migration result is composited from the individual layers images. The layer-stripping migration algorithm can be summarized as three steps: (1) model definition, (2) wavefield extrapolation and imaging, and (3) boundary determination. The migration scheme posed in this way is similar to datuming with an imaging condition. The advantages of the layer-stripping method are: (1) it preserves the benefits of the reverse-time method, i.e., it handles strong velocity contrasts between layers and steeply dipping structures; (2) it eliminates artificial interlayer multiples; (3) it reduces computational expense in high velocity layers; and (4) it allows interpretational constraint during image formation. The method has been implemented with both an explicit 4th-order time, 10th-order space, finite-difference approximation to the scalar wave equation, and an implicit 2nd-order time, 4th-order space finite-difference scheme applied to the linearly transformed wave equation (Li, 1986). The capability of post-stack layer-stripping reverse-time migration is illustrated on a synthetic CMP data and a CMP data from a survey over a faulted anticline in a fold and thrust belt. For pre-stack layer-stripping reverse-time migration, I present migrations of two synthetic data sets and a field data example from part of the marine seismic reflection profile RU-3, crossing the Hosgri fault offshore southern central California. The Hosgri fault appears as a northeast dipping high angle fault with a thrust component.

Pre-stack imaging of 3D seismic reflection data is at present a major challenge of seismic data processing. The main difficulty arises because of the limited power of today's computers. Manipulating volumes of 3D multi-fold seismic data and, especially, executing the large number of mathematical operations required for 3D migration is presently a very complicated and time consuming task. More than that, 3D seismic reflection surveys are characterized by very irregular acquisition geometries. This requires special migration algorithms and even new thinking. In this thesis we present a new approach to imaging 3D pre-stack seismic reflection data. It is a practical alternative to the full 3D pre-stack depth migration and takes special care of irregular acquisition patterns. The objective of the method is three fold: (a) to perform migration before stack, (b) to correct for three-dimensional seismic effects, (c) to provide a three-dimensional velocity model and a depth image. A 3D processing scheme is presented here. It is composed of three procedures which are different approximations to the full 3D pre-stack migration operation. The interaction between those schemes is simple, and each procedure serves as a building block for the next module. That is due to the special derivation of the algorithms in the frequency domain. The three modules that compose this work are: (1) Reorganization of multi-fold data on a regular pre-stack grid, based on Dip-Moveout and the inverse transform (DMO$\sp{-1}$). (2) 3D pre-stack time migration, using the PSI technique. (3) Two-pass 3D pre-stack depth migration, performed by splitting the full 3D calculation into a succession of 2D operations.

For the purpose of simulating large-scale seafloor scattering for general heterogeneous media using an elastic finite-difference scheme, I develop new tools for modeling explosive point sources and pressure plane waves traveling at given angles of incidence and for efficiently simulating a free surface on a staggered grid. The absorbing boundary conditions and finite-difference scheme were chosen to minimize memory requirements, a vital consideration for three-dimensional simulations. In addition, the algorithms are designed with large amounts of parallelism, making them suitable for modern supercomputers and feasible tools for Monte Carlo investigations. Using these tools, I simulate scattering from a Goff-Jordan seafloor model with self-similar bathymetry. The results underscore the importance of fully elastic modeling in understanding the origin and strength of the reverberatory coda not present for smooth, planar seafloor models.

Northern Venezuela consists of seven east-west trending tectonostratigraphic belts which are from north to south: Dutch and Venezuelan islands (DVI), Venezuelan platform, Cordillera de la Costa-Margarita terrane (CCMT), Cordillera de la Costa belt (CCB), Caucagua-El Tinaco belt (CTB), Paracotos belt (PB), and Villa de Cura belt (VCB). Geochemical, metamorphic, and structural data were collected in several transects through these belts. Major- and trace-element abundances in mafic and felsic rocks were determined to find the tectonic affinity of these rocks. The Precambrian basement of the CCB appears to consist of within-plate granites. The Paleozoic basement of the CTB consists of MORB's, Andean-type tholeiites, and arc granites. The mafic rocks of Mesozoic age have variable affinities: the DVI, CCMT, and VCB are underlain by MORB's, but part of the VCB has volcanic arc affinity. All felsic rocks of Mesozoic age in the DVI and CCMT are arc granites. Several periods of metamorphism have occurred in Venezuela in Precambrian and Paleozoic time. An important metamorphic event occurred in the Cretaceous. The CCMT and VCB underwent high P/low T metamorphism and subsequently low to intermediate P/T metamorphism. These events may be related to subduction and subsequent uplift, respectively. The CCB underwent an intermediate P/T metamorphism, because it probably has never been buried as deeply. The PB underwent very low-grade metamorphism possibly related to the obduction of ophiolitic thrust sheets. Four generations of folds and three generations of faults were recognized. The first two phases of folding occurred probably in the Permian and the last two during a Late Cretaceous-Tertiary orogeny. Kinematic indicators suggest that the entire system underwent dextral shear parallel to the belts during the Late Cretaceous-Tertiary. Based on the available geologic data, a plate tectonic model is proposed. In the model the east-west trending belts of northern South America are allochthonous and deformed as a result of collision of a microcontinent (Sebastopol block) with northwestern South America. The belts were initially coupled to the overriding South American plate, but because of the large obliquity of plate convergence were transported easterly since the Late Cretaceous.

Rifting between Newfoundland and Iberia occurred in two distinct phases--the first late Triassic to early Jurassic, the second late Jurassic to early Cretaceous--culminating in the creation of the North Atlantic Ocean. A dynamic modelling method was used to examine the implications of multiple phases of rifting on the development of the Newfoundland-Iberia conjugate margins. The models predicted a lack of magmatism on these margins, and suggested that extension was significantly greater in the second rifting phase than in the first; these predictions agree with geological observations. The models could not predict the existence of highly thinned continental crust on both conjugate margins, however. In addition, a set of generic models roughly based on the Newfoundland and Iberia margins suggested that, where two rift phases occur, the site of the original rift usually will not be favored for extension when stretching resumes.

3D AVO (Amplitude Variation with Offset) analysis has evolved as an important interpretation method of 3D seismic data, in the last decade. A new approach is proposed in the present work for quantitative AVO analysis of 3D seismic data. The method is based on fitting the observed seismic data, in a least-squares approach, to a function of average angle (average of angle of incidence and angle of transmittance) or offset. The aim of the method is to obtain the estimates of changes in the elastic parameters (P-wave modulus, S-wave modulus and density) at a particular output location, using a combination of pre-stack Kirchhoff migration and matrix inversion. The method includes ways to do migration velocity analysis and estimation of dip. As a further extension, the method is modified to find out the orientation of vertical fractures and the extent of anisotropy, for the case when an isotropic medium overlies a vertically fractured medium.

Migration results of two multi-channel vertical incidence reflection seismic dataset offshore Virginia are presented. A new velocity model derived from multi-channel and wide-angle reflection/refraction data is developed and fit into regional geology context. The validity and limitation of the new model is discussed by means of ray-tracing travel time forward modeling and 2-D acoustic wave-equation modeling methods. Our new results show that there is a sharp boundary separating normal continental crust from crust produced by the sea-floor spreading process in this region and that the Moho appears to be continuous across the boundary. The zone of so-called "magmatic underplating" proposed previously (Sheridan et al., 1993) may well be magmas produced during the initial stage of sea-floor spreading. There may be some fragments of continental crust entrained in the region immediately seaward of the boundary, but there is no evidence of the existence of large volume of continental crust and we believe the crust in this area is also primarily product of the Mesozoic sea-floor spreading. (Abstract shortened by UMI.)

Standard crustal seismic modeling obtains deterministic velocity models which ignore the effects of wavelength-scale heterogeneity, known to exist within the Earth's crust. Stochastic velocity models are a means to include wavelength-scale heterogeneity in the modeling. These models are defined by statistical parameters obtained from geologic maps of exposed crystalline rock, and are thus tied to actual geologic structures. Combining both deterministic and stochastic velocity models into a single model allows a realistic full wavefield (2-D) to be computed. By comparing these simulations to recorded seismic data, the effects of wavelength-scale heterogeneity can be investigated. Combined deterministic and stochastic velocity models are created for two datasets, the 1992 RISC seismic experiment in southeastern California and the 1986 PASSCAL seismic experiment in northern Nevada. The RISC experiment was located in the transition zone between the Salton Trough and the southern Basin and Range province. A high-velocity body previously identified beneath the Salton Trough is constrained to pinch out beneath the Chocolate Mountains to the northeast. The lateral extent of this body is evidence for the ephemeral nature of rifting loci as a continent is initially rifted. Stochastic modeling of wavelength-scale structures above this body indicate that little more than 5% mafic intrusion into a more felsic continental crust is responsible for the observed reflectivity. Modeling of the wide-angle RISC data indicates that coda waves following PmP are initially dominated by diffusion of energy out of the near-surface basin as the wavefield reverberates within this low-velocity layer. At later times, this coda consists of scattered body waves and P to S conversions. Surface waves do not play a significant role in this coda. Modeling of the PASSCAL dataset indicates that a high-gradient crust-mantle transition zone or a rough Moho interface is necessary to reduce precritical PmP energy. Possibly related, inconsistencies in published velocity models are rectified by hypothesizing the existence of large, elongate, high-velocity bodies at the base of the crust oriented to and of similar scale as the basins and ranges at the surface. This structure would result in an anisotropic lower crust.