Computation and application of seismic common-focus point and angle-domain common-image gathers.

摘要

Common-focus point (CFP) gathers and angle-domain common-image gathers (ADCIGs) have been important tools for migration velocity analysis. Application of CFP gathers (velocity analysis) and computation of ADCIGs are extensively studied and developed in this dissertation.;A 2-D velocity tomography is developed using focusing operators which are obtained based on CFP technology and contain one-way traveltimes from CFPs to receivers in an aperture along the earth's surface. The new method solves for both velocity updates and focus point coordinates, using simultaneous iterative reconstruction that does not require calculation of a matrix inverse. The algorithm is fast because it involves a small number of rays, and is stable because the model parameterization is changed through layered, to gradient, to grid parameterizations.;We compute ADCIGs from reverse-time migration (RTM) using eleven different algorithms, which fall into three main categories: direction-vector-based methods (DVB), local-plane wave decomposition methods (LPWD), and local-shift imaging condition methods (LSIC).;A single velocity model is used, so that the results can be directly compared and evaluated in terms of image accuracy, computational efficiency and the presence of artifacts. Analysis of the algorithms and their results show that some algorithms that have previously been treated as different, are in fact, variations of the procedure and implementation, and reasons that were previously not known for some types of artifacts are revealed.;LPWD and LSIC are considered more accurate and stable than DVB methods, but they suffer from a trade-off when choosing the local window size that serv es as a factor in controlling smearing-effect artifacts and angle resolution. The causes of the smearing-effect artifacts in LPWD and LSIC are discovered, and two methods are developed to remove them in both LPWD and LSIC methods, respectively. The importance of high angle-resolution in ADCIGs is explained and demonstrated.;DVB is considered the most efficient method, but it has a limitation in common with with the LPWD and LSIC methods, which is a huge disk space requirement for saving source snapshots. The corresponding I/O times and I/O bottleneck diminish the advantage of the computation time efficiency of the DVB. Two algorithms are developed to avoid saving or reconstructing source snapshots for extracting ADCIGs from RTM. Multiple-excitation imaging condition is developed to pick and use only a few of the highest-ampliutde arrivals, and the time-gradient method is developed to calculate source propagation direction using excitation times, rather than full source snapshots.