next up previous print clean
Next: Conclusions Up: Haines et al.: Multiple Previous: Example 2: A more

Example 3: Gulf of Mexico data

Our third example applies non-stationary PEF's to Gulf of Mexico subsalt data (Figure [*]). This Mississippi Canyon 2-D dataset was released by Western Geco in 1997 for the SEG multiples workshop. Again we find that a smoothed version of the PRT result provides a better model for signal PEF estimation than the raw PRT result, and that the raw PRT multiple model provides a satisfactory noise PEF estimation model. Two angle gathers are shown in Figures [*] and [*], from the salt area and from outside the salt body, respectively. In both cases, the PEF result produces a cleaner final gather.

 
CIG2_gom_nice
CIG2_gom_nice
Figure 5
(a) Angle gather from offset of 12000 m (see Figure [*] for reference), in the area of the salt body. (b) Multiple model from PRT approach, used as noise model for PEF approach. (c) Primary model from PRT. (d) Smoothed version of (c), used as signal model for PEF estimation. (e) Multiple model output by PEF approach. (f) Primary model output by PEF approach.
[*] view burn build edit restore

 
CIG_gom_nice
CIG_gom_nice
Figure 6
(a) Angle gather from offset of 4000 m, in the sediments outside the salt body. (b) Multiple model from PRT approach, used as noise model for PEF approach. (c) Primary model from PRT. (d) Smoothed version of (c), used as signal model for PEF estimation. (e) Multiple model output by PEF approach. (f) Primary model output by PEF approach.
[*] view burn build edit restore

The stacked results (Figures [*], [*] and [*]) illustrate the effectiveness of the PEF technique. In addition to the raw data (Figures [*]a, [*]a and [*]a), the PRT result (Figures [*]b, [*]b and [*]b), and the PEF results described above (Figures [*]d, [*]d and [*]d), we include also the result attained using ${\bf Nd}$ as the signal model (Figures [*]c, [*]c and [*]c). The PEF results are clearly better than the raw stack, and also notably superior to the PRT result. The ${\bf Nd}$ approach is quite effective in this case, particularly in terms of maintaining higher frequency content of the data.

 
gom_stack
gom_stack
Figure 7
(a) raw data. (b) Result from muting in the PRT domain. (c) PEF result, using ${\bf Nd}$ as the signal model. (d) PEF result using smoothed version of PRT primary model as signal model.
[*] view burn build edit restore

 
gom_stackz1
gom_stackz1
Figure 8
Windows of various final stack options, from beyond the margin of the salt body: (a) raw data. (b) Result from muting in the PRT domain. (c) PEF result, using ${\bf Nd}$ as the signal model. Note higher frequency content than any of the other stacks. (d) PEF result using smoothed version of PRT primary model as signal model.
[*] view burn build edit restore

 
gom_stackz2
gom_stackz2
Figure 9
Windows of various final stack options, from beneath the salt body: (a) raw data. (b) Result from muting in the PRT domain. (c) PEF result, using ${\bf Nd}$ as the signal model. (d) PEF result using smoothed version of PRT primary model as signal model.
[*] view burn build edit restore


next up previous print clean
Next: Conclusions Up: Haines et al.: Multiple Previous: Example 2: A more
Stanford Exploration Project
7/8/2003