DISCLAIMERS

It is not strictly true that a properly migrated data volume should have only one plane at a given location. This is merely a model that for its validity requires many conditions including that (1) a local planar model suffices for curved interfaces, (2) the data is mainly primary reflections, and (3) faults and unconformities are a second order issue.

The idea that the weighting function for regression must be determined along with the regression itself is a nonlinear concept, accompanied by the usual difficulties. The most elementary linearization is to presume the desired events are the strongest ones and that superposing semicircles and other events are weaker. In reality, we typically start from such an oversimplified model and then hope that by iteration we can escape its limitations.