Comparison with adaptive subtraction in the image space

In addition to the pattern-based approach, the multiples are attenuated with adaptive subtraction. Instead of comparing both results in the data space, the two techniques are now analyzed in the image space after migration. A split-step DSR migration is used with three reference velocities. The results of the migrations are then studied in the angle domain on common-image gathers (ADCIG). These angle gathers are generated with the method of Sava and Fomel (2003) after migration Stolt and Weglein (1985); Weglein and Stolt (1999). Angle gathers can be useful for AVO analysis Kuhl and Sacchi (2003); Prucha et al. (1999) and can form a convenient basis for multiple attenuation as well Alvarez et al. (2004); Sava and Guitton (2005).

Figure shows the migration results in the angle domain for the input data (Figure a), the estimated signal with the pattern-based approach (Figure b) and the estimated signal with adaptive subtraction (Figure c). By looking at the constant angle section (X=11000 m), it appears that more multiples have been removed with the pattern-based approach, especially below the salt where the multiples are the strongest. The common angle gathers corroborate this: a lot more energy from the multiples is present with the adaptive subtraction. Note that some multiple energy remains in both Figure b and Figure c where the salt body is present. These multiples have strong curvature and low-frequency content. These events are probably internal multiples bouncing between the sea bottom and the top of salt. Another pass of multiple attenuation in the image space Sava and Guitton (2005) could eliminate these reflections.

 

cube.mig
Figure 23 comparison of migrated sections in the angle domain for (a) the input data with multiples, (b) the estimated primaries from the pattern-based approach, and (c) the estimated primaries with adaptive subtraction. Arrows M point to multiples.

Finally, Figure shows a comparison of ADCIGs for the input data with multiples (Figure a), the estimated primaries with the pattern-based approach (Figure b) and the estimated primaries with adaptive subtraction (Figure c) outside the salt boundaries (X=4000 m). It illustrates once again that the pattern-based approach outperform the adaptive subtraction method with a cleaner panel.

 

sedi.mig
Figure 24 Angle domain common image gathers of the Gulf of Mexico dataset for (a) the input data, (b) the estimated primaries with the pattern-based approach and, (c) the estimated primaries with adaptive subtraction. This gather position is X=4000 m, outside the salt boundaries. Primaries are flat (shown as P) and multiples are curved (shown as M).

The Gulf of Mexico example demonstrates that the pattern-based approach is an effective tool for multiple attenuation in complex geology. Although the multiple model obtained with SRMP presented some obvious flaws (short offset amplitudes), the proposed approach is able to attenuate the multiples while preserving the primaries. In addition, comparisons in the image space after migration on ADCIGs show that the pattern-based approach gives cleaner panels than adaptive subtraction.