New Advances in Seismic Imaging for Exploration and Solid Earth Geophysics III
Presiding: A Weglein, University of Houston; G Pavlis, Indiana University
S33A-01 INVITED 13:30h
Addressing the Challenge of Seismic Depth Imaging Beneath a Complex Heterogeneous Medium: an Inverse Scattering Series Response
Depth imaging beneath an overburden where the medium and/or the boundaries within the medium are rapidly varying is a significant and outstanding problem in current exploration seismology. This problem occurs in plays beneath e.g., salt, basalt and karsted sediments, and is a technical impediment of such a magnitude that deep water E&P strategies and portfolios are being impacted and re-examined. At the heart of the problem is that all current depth imaging methods require an adequate velocity model in order to produce an acceptable depth image. That adequate velocity model is often not achievable for the cases listed above, even with the most expert application of our best velocity analysis methods. The inverse scattering series provides the opportunity to achieve all processing objectives associated with inversion without the need for an adequate velocity, and has been successfully applied to the removal of free and internal multiples on field data, with demonstrated value, especially in complex ill-defined media. This paper will describe the recent development of velocity independent depth imaging concepts and algorithms from the inverse inverse scattering series, and will provide examples, plans and open issues. Reference: Shaw, S. A. and Weglein, A.B. (2004), "A leading order imaging series for prestack data acquired over a laterally invariant acoustic medium", 74th International Meeting of the Society of Exploration Geophysicists, Denver.
S33A-02 13:45h
Common-Offset Pre-Stack Time Migration Using Curvelets
Recently, in the field of applied harmonic analysis, curvelets have been introduced in an effort to combine geometry and multiscale analysis. Roughly speaking, curvelets are 2D extensions to wavelets. It has been shown that curvelets in essence provide optimal sparsification of objects that are twice continuously differentiable (C2) with discontinuities along C2 edges. In addition, it was recently shown that curvelets in essence optimally sparsify certain Fourier Integral Operators (FIO). Because the seismic imaging operator belongs to this class of FIO, and since the singularities in reflection seismic data lie mainly along smooth curves, curvelets are plausible candidates for simultaneous sparsification of seismic data and the imaging operator. The geometry of seismic imaging is determined by map migration. Since map migration makes explicit use of the directions (or slopes) in the data, and since curvelets have directions associated with them, map migration can be used to determine the transformation from a curvelet in the data domain to a curvelet in the image domain. This transformation consists of a translation, rotation and dilation of the curvelet. We show the use of map migration to image seismic data using curvelets, and treat common-offset (CO) pre-stack time migration as an example. We present synthetic data examples of CO pre-stack time migration using curvelets, where we use the aforementioned transformation.
S33A-03 INVITED 14:00h
Pre-Stack Kirchhoff Depth Migrated Images of the Upper Mantle
We present images made from receiver function data recorded in three different upper mantle tectonic regimes: the Kaapvaal craton, the Japanese subduction zone, and the Jemez Lineament of the western U.S. orogenic plateau. The seismic data are migrated using a 2.5 D prestack Kirchhoff depth migration algorithm that uses 2-D seismic velocity models developed from tomography as the migration velocity models. The Kaapvaal craton image has been made from 9 earthquakes recorded along a 2000 km long array. Both of the transition zone discontinuities stand out very clearly. The amplitudes and thicknesses of the discontinuities estimated from the image are in agreement with predictions of global reference models. Above the transition zone are a number of events extending 100's of kilometers as slab-like structures which may form the bottom of the Kaapvaal craton. In Japan we make use of a number of earthquakes recorded at the very dense Hi-Net array, which consists of more than 500 borehole seismographs. The migrated receiver function images show the transition zone discontinuities very clearly, as well as the slab subducting beneath Hokkaido. The uplift predicted by slab penetratation of the 410 discontinuity is clearly seen in the migrated image. The slab appears to stall above the 660 discontinuity. Across the Jemez lineament, an image made from 7 earthquakes recorded at a 225 km long seismic array shows a complex set of upper mantle sill-like structures extending from the Moho to a depth of about 125 km. From active source seismic data we estimate that the upper mantle contains about 1 percent partial melt. Amplitudes in the receiver function image suggest that this is true in all the sill structures in the upper mantle. We relate the sill structures to recent volcanism and uplift along the lineament.
S33A-04 INVITED 14:15h
The Importance of Multiples in Teleseismic Scattered-Wave Imaging
Traditional imaging approaches in solid-earth, passive seismology suffer from the presence of free-surface multiples in the data, which hinder the detection of structure in certain depth ranges. Here, I discuss a recently developed teleseismic migration technique that, unlike traditional approaches, incorporates and relies in large part on free-surface multiples to improve the focusing and resolution of lithospheric structure. The technique involves multichannel inversion of scattered teleseismic body waves recorded at dense seismic arrays. The problem is posed for forward- and back-scattered wavefields generated at discontinuities in a 2D isotropic medium, with the backprojection operator cast as a generalized Radon transform (GRT). The approach allows for the treatment of incident plane waves from arbitrary backazimuths, and recovers estimates of material property perturbations about a smoothly varying reference model. The recovered properties are P and S velocities, which are independently obtained by treating P-P and P-S/S-S interactions at 2-D line-scatterers, respectively. The S-velocity image is constructed by a combined analysis of forward and back-scattered modes, with the latter affording better resolution of planar structure, whereas the P-velocity image relies exclusively back-scattered interactions. Applications to synthetic and field data from Cascadia, Alaska, and the SE Canadian Shield are presented to illustrate the benefits of including multiples in teleseismic imaging.
S33A-05 14:30h
Nonlinear Velocity Inversion via Extension - Layered Case
Migration velocity analysis, like all other aspects of migration,, is based on the Born or single-scattering approximation. It is capable of correcting large errors in a velocity model of the Earth, but cannot account directly for multiply reflected energy and other nonlinear effects. Inversion via data fitting, on the other hand, may be based on full waveform (nonlinear) modeling. Given a good enough starting estimate of seismic wave velocity, full waveform inversion can account for both single and multiple scattering. Considerable experience suggests, however, that (iterative) inversion of this type is unlikely to succeed unless provided with a relatively accurate initial model of velocity. The extended model concept provides a common framework for migration velocity analysis and nonlinear inversion, and suggests an approach to nonlinear velocity analysis integrating singly and multiply scattered energy. The extended model replaces the wave velocity with an operator. Its kernel corresponds to the image volume of prestack migration. In fact this image volume may be identified with the perturbation of the operator kernal about an ordinary acoustic or elastic model. The layered medium case is particularly simple and amenable to computational experimentation: the operator extension of the velocity is necessarily a convolution operator in the horizontal variables, and admits diagonalization via the discrete cosine transform. This talk will briefly sketch the framework of extended inversion and present some first numerical results in the layered case.
S33A-06 14:45h
Velocity Models From Frequency-Domain Waveform Tomography in Controlled Source Seismology
Recently, a number of leading research groups have presented examples of the success of the frequency-domain approach to waveform inversion. Apart from the use of the frequency-domain, a key common factor in all of these studies is the tomographic use of large-offset, transmitted, refracted arrivals. In this respect we refer to such work as waveform tomography. Large offset data are very sensitive to velocities; waveform tomography provides a way to unravel the complexities of the refracted arrivals in order to yield well constrained velocity models. Waveform tomography is an extension of waveform inversion methods first developed in the early 1980's (e.g., Lailly, 1983 and Tarontola, 1984). The frequency-domain version of waveform inversion developed in the 1990's (e.g., Pratt and Worthington, 1990) has now emerged as an efficient imaging tool, capable of being used on a production basis for large scale 2D problems. The examples of frequency-domain waveform tomography herein image the Earth's interior successfully on a wide range of scales. The results indicate the importance of the refracted wavefield and the equal importance of low frequencies; ultimately surveys should be specifically designed for waveform tomography. Major challenges remain to move the 2D implementations into the 3D world, to properly account for shear waves, mode conversions and multi-component data, and to incorporate the critical second order effects of anisotropy and attenuation.
http://geol.queensu.ca/people/pratt/&35;new