The datasets were migrated using a shot-profile migration algorithm (an adjoint to the forward-modeling operation). By using exactly the same processing parameters, we preclude frequency, bandwidth and phase differences. Migrated images -- using the correct model velocities -- are shown in Figures and . Since the actual velocity models are not usually known, it is common practice to migrate the datasets with the same velocity (e.g. baseline) model. Raw difference images obtained using different velocity fields are shown in Figures and . For exactly the same acquisition geometry, the time-lapse response in the presence of the salt wedge (Figure b) is relatively weaker and much more complicated than when no salt is present (Figure a). Also, as seen in Figures c and d, a change in the acquisition geometry introduces more artifacts into time-lapse image when the salt wedge is present. These effects result partly from the poor illumination close to the salt flank, unequal illumination of the salt flank (due to the difference in survey direction) and contamination by differences in the relatively much stronger migration artifacts. As shown in Figure , using an incorrect velocity field introduces a different pattern of artifacts when the salt wedge is present. Thus, since exact velocities are not used in practice, any assumptions made with regards to the migration velocities may be difficult to account for. In the absence of the salt wedge, the results are much less complicated and simple cross-equalization of the surveys should suffice. While opposite acquisition directions may be an extreme, surveys over very complex salt bodies acquired with slightly different geometries may result in highly deteriorated time-lapse images even after careful cross-equalization.