Weighted stacking of seismic AVO data using hybrid AB semblance and local similarity

Field examples

Figure 9a shows a good pre-processed marine CMP gather after demultiple procedure. The deep reflectors (3.6 and 4.2 s) of the stacked trace are detected in Figure 9d by AB semblance and local-similarity-weighted stacking. But these deep reflectors are easily missed in Figure 9b by conventional semblance and conventional equal-weight stacking, and Figure 9c by conventional semblance and SNR-weighted stacking due to the low SNR after stacking.

cmpstack1
Figure 9. (a) Field CMP gather with high SNR; (b) Stacked trace by conventional semblance and conventional equal-weight stacking; (c) Stacked trace by conventional semblance and SNR-weighted stacking; (d) Stacked trace by AB semblance and local-similarity-weighted stacking.

Figure 10a is another marine CMP gather that has lower SNR compared to the data in Figure 9a. The stacked trace with the highest SNR is in Figure 9d by conventional semblance and local-similarity-weighted stacking, especially in the shallow part (0.5-2 s). The SNR in Figure 10c (SNR-weighted stacking) is higher than that in Figure 10b (equal-weight stacking) because the equal weight or averaging process in Figure 10b decreases the stacking amplitudes of the reflection signals and, in turn, emphasizes irregularity of random noise. The results in the deep part (after 2.5 s) of three stacked traces are similar because of the influence of multiple reflections. If a good demultiple technology is applied before NMO velocity analysis, it is predicted that our method still produce the best stacking. The two field examples above further demonstrate the proposed approach has the effectiveness in dealing with the stacking of AVO data independent of SNR.

cmpstack2
Figure 10. (a) Field CMP gather with low SNR; (b) Stacked trace by conventional semblance and conventional equal-weight stacking; (c) Stacked trace by conventional semblance and SNR-weighted stacking; (d) Stacked trace by AB semblance and local-similarity-weighted stacking.

At last we applied the proposed framework to a 2D field dataset from the Gulf of Mexico, showing the advantages of the new framework in artifact attenuation and seismic interpretation. Figure 11a shows stacked data by conventional semblance and conventional stacking, Figure 11b shows stacked data by conventional semblance and SNR-weighted stacking, and Figure 11c shows stacked data by AB semblance and local-similarity-weighted stacking. It is observed that our proposed approach produced the best stacking result in the perspective of event continuity and the least artifacts. Zoomed sections, further, shown in Figures 12a, 12b and 12c from the frame boxes in Figures 11a, 11b and 11c demonstrate these advantages more clear.

stack1-gulf,snrstack-gulf,simistack-gulf
Figure 11. (a) Stacked data by conventional semblance and conventional stacking; (b) Stacked data by conventional semblance and SNR-weighted stacking; (c) Stacked data by AB semblance and local-similarity-weighted stacking.

stack1-A,snrstack-A,simistack-A
Figure 12. (a) Zoomed section from Figure 11a; (b) Zoomed section from Figure 11b; (c) Zoomed section from Figure 11c.

Weighted stacking of seismic AVO data using hybrid AB semblance and local similarity