With 3-D data, attenuating surface-related multiples with wavefield techniques such as SRME is particularly challenging. The difficulty stems from the lack of source/receiver coverage, the lack of crossline offsets, and the coarse source/receiver spacing at the surface for the prediction of multiples. Making SRME work with 3-D data can be achieved by interpolating/extrapolating the missing data and designing subtraction techniques that are less sensitive to modeling errors due to the acquisition geometry. However, most of today's efforts are focused on the prediction step relying on the adaptive subtraction to optimally estimate the primaries. This Chapter presented a multiple removal strategy on one streamer from one sail line of a 3-D survey in the Gulf of Mexico that focuses on the subtraction step only.
Due to the acquisition geometry, the shots and the near offset traces are first interpolated in 2-D with a radon-based technique with and without sparseness constraints. Interpolation is a vital element of the whole multiple attenuation procedure because it affects directly the quality of the multiple prediction. For this reason, interpolation with the sparseness constraint gives the best results by recovering most of the steep dips present in the data and by giving the fewest reconstruction artifacts.
Then, from the interpolated shots, a multiple model is estimated. Because some dips were not well reconstructed during the interpolation process, some aliasing artifacts are present in the multiple model. Again, interpolation with the sparseness constraint yields the best multiple model. However, important discrepancies exist between the actual and modeled multiples. For instance, some diffracted multiples are either poorly or not recovered by the 2-D prediction. This result should be expected wherever the geology is complex and the acquisition geometry is irregular. This observation makes a strong case for 3-D prediction.
Finally, the estimated multiple model from the 2-D prediction is subtracted from the data using adaptive subtraction and the pattern-based approach. Because this model has some flaws, the adaptive subtraction does not remove the multiples as well as the pattern-based method. In particular, the migrated images of the estimated primaries exemplify the weaknesses of adaptive subtraction where the prediction is the least accurate, e.g., below the salt.
Therefore with 3-D data, because it remains very difficult to obtain an exact multiple model from the recorded data, both the prediction and the subtraction should be looked at and improved to obtain better attenuation results. One one hand, deriving interpolation strategies in order to recover the missing traces is vital to obtain accurate multiple models. The interpolation schemes must be carefully tuned to leave as few artifacts as possible. The radon-based reconstruction with sparseness constraints proved to be quite effective. On the other hand, deriving intelligent and flexible subtraction techniques is also fundamental. Here, the pattern-based approach offers an alternative to the more conventional adaptive subtraction technique by being robust to modeling inadequacies.