Robustness of the scalar elastic imaging condition for converted waves

For elastic reverse-time migration, one constructs elastic source and receiver wavefields using a source function and multi-component data, and then applies an imaging condition to the reconstructed wavefields. A widely used imaging condition is crosscorrelation of individual wave modes separated in both wavefields, e.g., compressional (P) waves and shear (S) waves. One difficulty in elastic migration is that the PS and SP images change sign at normal incidence. We recently proposed a scalar imaging condition that corrects for the polarity change; however, this imaging condition requires additional prior information, specifically the reflector normal. Here, we address two practical issues associated with this imaging condition. The first is the estimation of the reflector normal. Using incorrect Pand S-mode velocities causes reflectors in migrated PP, PS, and SP images to be shifted from their true positions. We show that it is more reliable to estimate reflector normals from PS and SP images, computed using conventional imaging methods, instead of from PP images. The second issue is the illumination and imaging of a reflector from its opposite sides. We demonstrate that the PS and SP images computed using waves from opposite sides of a reflector have the same polarity, and therefore they can be stacked over experiments without canceling each other at various positions in space. Several numerical examples illustrate the two issues in simple and complex models.