Processing of P-SV surface-seismic data: anisotropy analysis, dip moveout, and migration

Abstract

In this dissertation, the current state of multicomponent converted-wave (P-SV) processing for surface seismic data is summarized and a number of new algorithms are developed and tested. These include: P-SV geometric spreading compensation; P-SV poststack velocity inversion; S-wave anisotropy analysis; P-SV dip moveout (DMO); and P-SV poststack migration. The problem of S-wave birefringence for the case of a singly-polarized shear source is examined, and an algorithm for the determination of the amount of S-wave splitting and the orientation of the natural coordinate system is developed. This algorithm is based upon the modeling of the crosscorrelation function between rotated radial and transverse field components. Synthetic examples indicate that it can resolve time delays of about 1 ms or greater for noise-free data with an 8-35 Hz bandwidth; the receiver azimuth relative to the natural coordinate system is accurately determined for time delays of about 2 ms or greater. A study of the method's performance versus signal-to-noise power ratio suggests that a signal-to-noise ratio of about 36 db is needed to obtain both the time-delay and rotation angle from a single receiver data set. For data with a signal-to-noise power ratio of O db, an analysis fold of about 60 is needed to determine the correct time delay and rotation angle. Application of the method to converted-wave data from Carrot Creek, Alberta, produces results that are in agreement with the difference in orientation between the two lines in the survey. An expression for common-conversion-point (CCP) dispersal for P-SV data is derived, and is found to give greater dispersal for data converted in the down-dip direction that in the up-dip direction. The apparent stacking velocity for the single-layer case is shown to also be asymmetric about reflector zero-dip, and for large positive (down-dip) values the apparent velocity can become imaginary. A dip moveout (DMO) equation for converted-wave data is derived, and is proven to properly account for P-SV CCP dispersal. A new cubic equation for determining the location to the zero-dip conversion point is also developed. The P-SVDMO equation is implemented using an integral-summation technique and applied to both synthetic and real data examples. In synthetic data, P-SV DMO gives better amplitude preservation of P-SV diffractions and dipping reflections. The application of poststack migration to converted-wave stack data is studied. It is shown that the exploding reflector model does not strictly hold for converted-wave stack data, but is a reasonable approximation if there are no large depth variations in Vp/Vs ratio. A P-SV migration velocity equation is developed, and gives velocities that are less than PSV stacking velocities by 6-11 %. Migration of synthetic P-SV diffractions demonstrate that stacking velocities produce substantial overmigration, whereas the P-SV migration velocities do not. A physical model data set simulating a buried reef is also migrated, and the P-SV data are found to give a good structural image. The processing techniques developed and discussed in this dissertation are applied to a three-component vibroseis survey from Carrot Creek, Alberta. The radial-component (P-SV) stack section is of good quality, with amplitude anomalies that correspond to the known location of oil pools. These anomalies are largely absent on the vertical-component (P-P) sections. The processing results for a two-component data set from Springbank, Alberta, are presented. The radial-component data are found to be of very poor quality; the stack section, however, has many coherent reflections that correlate well to those of the vertical-component stacked data. The results of processing a three-component dynamite survey from Casper Creek South, Wyoming, are also presented. Two-component geophone rotation produces some improvement in the rotated radial-component data signal strength. Reflections with substantial dips are imaged on both the P-P and P-SV stack sections. The converted-wave section, however, has a higher noise level and much lower bandwidth than the vertical-component (P-P) section.