| Popis: |
Summary Wave equation migration plays an increasingly important role in seismic data processing. For our purposes we are only considering the one-way wave equation, because of its efficiency. Currently the predominant one-way propagators are FFD and its variations (e.g., GSP, GOFFD), FD + Li’s filters, FD, and PSPI, etc. However, even within the FFD propagator class, there are still many different approximation schemes and implementations to achieve wide-angle phase-correction part by FD. The problem is how can one benchmark an algorithm as being accurate and reliable? From the published papers, many algorithm developers prefer to show the impulse response in a constant velocity model with a reference velocity that gives a three-to-one, or two-to-one velocity contrast. This approach gives the impression that if the impulse response obtains high-dip accuracy (say, 60° for all azimuths), the algorithm usually will work well for other complex models. This conclusion may not be justified in all cases. We will show examples to demonstrate that two different approximation schemes of FFD could result in huge differences in imaging a complex heterogeneous model even though they both imaged very well for a constant velocity model and the SEG/EAGE 3D salt model. The significance of these examples tells us that a constant, or nearly constant, velocity perturbation model (although the velocity contrast may be big) is not sufficient to benchmark a propagator’s performance. |
