Fresh insight into time migration tomography

Richard Leggott and Richard Morgan of Veritas DGC provides this discussion of a tomographic algorithm application which offers the potential for an improved velocity model for seismic data analysis. In hydrocarbon exploration and development, imaging of seismic data is used for accurate geologic interpretation, estimation of rock properties, pore pressure prediction and many other purposes. All such analyses depend on a good seismic image, and this in turn requires an accurate velocity model. As more information is extracted from seismic data in areas of increasing structural complexity, the role of a good velocity model has become even more crucial. In this article an idealized tomography algorithm is described, which simultaneously images the seismic data and updates the velocity model using the same imaging algorithm. By using data from every sample of the seismic image, the tomographically updated velocity will exhibit high resolution features that other velocity analysis tools cannot resolve. Results from this tomography are shown using time migration imaging in both isotropic and anisotropic media. Conventional velocity analysis The conventional method for generating a velocity model for imaging is vertical updating (Deregowski, 1990). Here, an initial velocity model is used to image the acquired seismic data to form common image gathers (CIGs). An updated RMS velocity field is picked using normal moveout (NMO) to flatten selected seismic events on the CIGs. This velocity is often smoothed, converted to an interval velocity, re-smoothed, and clipped. A new seismic image is then formed using the updated velocity model. If required, the procedure is iterated until the primary seismic events are flat on the CIGs. Vertical updating is only suitable for simple geology. The method assumes that applying a residual moveout with NMO is a reasonable approximation to re-imaging the seismic data with an updated velocity model; this can only be a valid assumption when the geology has little structure. Additionally, care must be taken when picking RMS velocities at a high temporal resolution to avoid an unstable interval velocity. This means that the velocity for thin layers cannot be picked accurately, and that a velocity contrast between distinct lithological layers can only be identified when that boundary is identified with a significant seismic reflection. Many different methods exist for imaging seismic data (Kirchhoff time migration, Kirchhoff depth migration, wave-equation migration, etc.). Different imaging techniques will use the velocity model in different ways to form a seismic image. Hence, a velocity model suitable for one imaging algorithm cannot be assumed to be suitable for another. For example, an appropriate velocity model for a Kirchhoff time migration will not generally be adequate for a Kirchhoff depth migration. There is a coupling between the velocity model and the imaging algorithm used. In other words, seismic imaging and velocity analysis are two aspects of the same problem and this should be reflected in how the velocity model is generated.