SRME can effectively attenuate all multiples with a bounce at the water surface, but only if all necessary data is collected. Virtually in all instances of real 3D data acquisition, the cross-line sampling is too coarse, the cross-line aperture is too small, the short offsets are missing, and acquisition obstacles and cable feathering produce irregular geometry. The data need to be interpolated and extrapolated to satisfy the requirements of 3D SRME. This is not a trivial endeavor and the performance of SRME greatly depends on it. Moreover, diffracted multiples and specular multiples from an interface with steep cross-line dip, may have bounce points well outside the cross-line aperture making them hard or impossible to predict.
Attenuating the multiples in image space solves the problem of the complex wave propagation of primaries and multiples. Prestack wave-equation migration takes care of the complexity of the wavefield propagation and makes the primaries very likely to be flat in ADCIGs and therefore more easily separable from the multiples in the Radon domain. No data interpolation or extrapolation is necessary because no multiples are predicted. Since very accurate migration velocities are always necessary to get a good depth image, however, postponing the multiple attenuation step until after migration does not come without a price. Computation of an accurate migration velocity field may be compromised by the presence of the multiples.