Inversion of OBC seismic data for P- and S-wave velocities.
Hall, Guy R
This thesis describes an new method for deriving a shared-earth velocity model for P-P ans P-SV reflections measured with ocean bottom cable (OBC) data.<p> The data have the potential to reveal lithological and fluid information about the rocks in the subsurface. The S waves recorded on OBCs are usually SV waves that have been converted on refraction of downgoing P-waves from the source. Conventional preocessing of OBC data separates the P-waves and SV-waves on the basis of particle motion: P-waves on the vertical component and SV-waves on the horizontal components. The P-waves are then processed in the conventional way, using procedures based on common mid point (CMP) gathers that have been well established for decades and are very sucessfulin determining subsurface structure. The SV waves are conventionally processed in a similar proceedure, based on common- conversion point (CCP) gathers, that requires the P-wave to S-wave velocity ratio, gamma, to be knowna priori; initiallythis must be guessed. the result of processing these two data sets is two seismic time sections: one a P-wave section and the other a converted wave section. By subjectively correlating events in these two sections it is possible to estimate the S-wave velocities. This may lead to further iterations in the converted wave processing.<p> my aim is to remove the need for any guesswork in the estimate of gamma and to eliminate the subjective correlation step.<p> The basic earth model underlying my approach is of discrete homogenous isotropic elastic layers separted by interfaces at which reflections occur. I invert the reflection travel times of the common-shot gathers or common recievers gathers to find the layer velocities and the positions of the interfaces in depth, working from the top downwards. I start with a travel time inversion scheme developed by Guangpin Li that assumes the interfaces are plane, but locally dipping. This gives an initial estimate of the ray paths and the interfaces. We thwn assume that the interfaces can be described as locally parabolic which gives better inversion results. For the P-wave data, the P-wave velocities and the interface geometries are the output of the inversion. thes P-wave velocities, but not he interface geometry, are used for the converted wave inversion, the output of which is S wave velocities and interface geometry. The interface geometry must be the same for both inversions: this shared-earth modal is the criterion for determining which converted-wave refractions correspond with the P-wave reflections. <p> I have developed a processing flow based aroud this inversion scheme, that requires a number of new steps, including separation of P-waves and S-waves, manually picking travel time curves in shot gatheres, and parameterising the picked data using cubic polynomials. The output of the processing flow are P- and S- wave interval-velocity in depth models that can be used for pre-depth stack migration. The processing scheme I develop is very simple compared with traditional schemes for generating interval-velocity depth models. Tests of the model building flow on both P-P and P-SV synthetic data yield good results. However, the initial model affects the final result and it is clear that, in some cases, the inversion drives the solution to a local, rather than a global,eroor minimum. I propose a brute-force solution for this problem: give the inversion a range of velocities in any layer and for each velocity find the minimum-error interface; then choose the velocity that gives the least error.<p> I apply the new processing flow to real data provided by Shell from the Guillemot field in the North Sea. the results are good, but there are still small eroors in the velocity model which i attribute to limitations in the way we have chosen to parameterise the earth. These errors can be reduced by updating the velocity field based on the residual moveout of reflection events in the migrated common image gathers. The method needs to be extended to layers that may have vertical and horizontal gradients.