EAEG/SEG Summer Workshop - Construction of 3-D Macro Velocity-Depth Models

A technique of layer velocity estimation tbrough semblance analysis was reported in Hadleyet al. (1988) and Landa et al. (1989). lts basic premise is that a coherency measure on unstacked seismograms reaches a maximum along reflection time trajectories that correspond to the correctly estimated velocity-depth model.

An interactive method has been developed for carrying out seismic velocity analysis that results in a three dimensionally tied and consistent velocity field for either stacking or time migrating 3-D seismic data (also grids of 2-D data). This method uses interactive horizon-based velocity picking and statistical data fitting as the primary model building tools. The motivation for this development comes from difficulties we experience in picking seismic velocities from data which is either poor in its signal to noise ratio characteristics or is structurally complex and requires complicated velocities to image the subsurface.

Practical application of traveltime inversion on multi 2D seismic surveys is limited by difficulties to be confident of the final velocity-depth model reliability due to a lot of mis-ties. An integrated approach combining seismic interpretation techniques and seismic processing like traveltime inversion and depth migration is developed in order to improve the geological and geophysical consistency of results. This approach is characterized by the use of coherence criteria firstly applied to each 2D seismic profile and secondly extended to 3D structural analysis. Consistency control procedures are described using examples deduced from multi 2D dataset study recorded in the Paris Basin. The objective is then to obtain accurate layer velocity estimates and reliable time to depth conversion of seismic sections.

With the advent of steep-dip non-dispersive Isotropic 3D time and depth migration algorithms, the ability to image complex structures has entered a new phase. However, the most important issue in subsurface imaging is nor the migration algorithm, but the elaboration of the subsurface velocity model.

Exploration targets in structurally complicated areas with strong velocity contrasts require interpretation of seismic data imaged in depth. In areas where conventional processing which includes dip-moveout correction yields CMP-stacked data with good signal strength and continuity, 3-D poststack depth migration offers an attractive solution to the imaging problem.

Summary not available

Time migration algorithms prove to be inadequate when trying to correctly image seismic sections in areas of complex geology. In such media, velocities vary both vertically and laterally giving rise to non-hyperbolicity in the observed moveout of the seismic reflections. To deal with this situation, a depth migration must be employed. However, a dilemma arises in that the performance of the pre-stack depth migration is sensitive to errors in the velocity model supplied by the user, that is the more accurate the underlying velocity model, the better the migrated image. This sensitivity provides an excellent tool for the iterative building of depth models (Figures 1 & 2). We present a method that combines pre-stack depth migration with velocity analysis and mapmigration.

Prestack migration is a successful tool for imaging complex structures and has recently become quite popular. Experience with prestack migration indicates that the quality of the image is very sensitive to the velocity model, and therefore prestack migration is also used as a velocity analysis tooI. The migrated image is usually obtained in an iterative procedure and the velocity model is updated between iterations. In other words, prestack migration and velocity estimation are related procedures. The velocity model is required so a good depth image can be obtained while the migration process itself is the tool for deriving the velocity models.

Migration velocities can be determined by performing several prestack migrations with a series of test velocity models. The velocity model that produces a consistent image for all source-receiver offsets is taken to be the correct migration velocity.

We use an irregular tetrahedral mesh to represent 3-D velocity models. The mesh is generated by individually inserting data points into the model with refinement of the mesh to include the newly added point. When isolated data points are to be interpolated, a Delauney tessellation is often a good choice since it partitions the model into tetrahedra connecting nearest-neighbor data points by its empty-circumsphere criterion. We use Delauney tessellation to import isolated velocity picks to form a 3-D velocity field and to import isolated surface picks to form a simply connected 3-D surface (Wiggins et. al, 1993). Velocity data comes to the explorationist in small parts: first, velocity picks for a shallow layer; then, a shallow horizon; then a deeper layer velocities and surfaces. A modeling system must represent these parts and combine them lnto a single model.