Paper Summary
Over the past few years, many techniques have been presented for updating the velocity depth model required for full-volume 3-D prestack depth migration (preSDM). These techniques (many of which are listed in “Suggestions for further reading”) have specific theoretical or practical limitations. Furthermore, a universal limitation to date is the spatial sampling of the information used to perform the velocity estimation. Prestack-migrated velocity information (usually in the form of CRP gathers) is typically output on a coarse grid, often 500 500 m.
This article presents a technique for increasing the statistical reliability of the velocity information to be input to the chosen velocity update scheme. The automated nature of this technique greatly reduces the unreasonably high number of man-hours needed to manually pick very dense velocity grids—the factor that has really limited us in the past in obtaining dense velocity grids.
Thus, this paper does not present a new method for obtaining velocity information for preSDM model updating; it demonstrates a technique for analyzing information produced continuously. We do not in any way improve on the limitations of the underlying technique; we merely make the best possible use of the information already available by looking at a very dense sampling of information.
In other words, when we estimate the velocity with many values, we only improve the precision of that estimate—not the accuracy. Thus, if the values coming from our velocity estimator were all erroneous, but consistently erroneous, then we would simply have a very precise estimate of that inaccurate result.