Conclusions

Lateral velocity deviations as small as three percent from the background can cause visible focusing. The FEAVA detector performs well both in the presence and in the absence of multiples. FEAVA removal by migration works when the migration operator is the adjoint of the modeling one and when no multiples are present. When internal multiples are present and imaging is performed with an algorithm of a lower order than the one used for modeling, FEAVA is removed only partially by migration. This is most likely caused by multiples not being defocused by a migration with the velocity of the primary reflections, regardless of the order of the algorithm. To verify this conjecture, one would need to model with an amplitude-preserving two-way algorithm two similarly complex datasets - one multiple-free and one multiple-affected, and then migrate each of them with the operator adjoint to the one used in modeling. If the conjecture is true, FEAVA will be removed completely from the multiple-free dataset, but only partially from the multiple-affected one. I also conjecture that, for a synthetic dataset, FEAVA removal is possible if the algorithm used for migration has the same accuracy or greater than the one used for modeling, and that exact adjointness of migration and modeling operators is not important. To verify this I will need to image with a higher-order algorithm a multiple-free dataset generated with a lower-order algorithm. I plan to verify these assertions in the near future.