next up previous [pdf]

Next: Discussion Up: Examples Previous: Land field data example

Marine field data example

For the final example, we adopt a field data example from the Gulf of Mexico (Claerbout, 1996) used previously by Li and Fomel (2015) to additionally test the proposed method and compare its performance with previous methods. In this dataset, the maximum recording time is 4 $ s$ with the maximum offset of 3.48 $ km$ . We estimate the initial $ w_{dr}(x,z)$ automatically using the method of velocity continuation (Fomel, 2003b) followed by 1D Dix inversion to depth. We use the central trace of $ w_{dr}(x,z)$ as the reference $ w_r(z)$ model and the remaining inputs to the proposed method are shown in Figure 14.

A comparison of the final estimated interval velocity from the proposed method and the optimization-based method (Li and Fomel, 2015) is shown in Figure 15 with a good general agreement despite using only about one-tenth of the computational time. Due to the availability of larger-offset data, we compare the the final seismic image after the proposed time-to-depth conversion process and that from prestack Kirchhoff depth migration (PSDM) using the estimated interval $ w(x,z) = w_r(z) + \Delta w (x,z)$ in Figure 16. The results are comparable verifying the effectiveness of the proposed method. We further investigate the common-image gathers (CIGs) generated from PSDM based on the conventional Dix velocity squared $ w_{dr}(x,z)$ and the estimated $ w(x,z)$ . We observe a noticeable improvement in the flatness of the CIGs from the estimated $ w(x,z)$ especially in the deeper sections, where the effects from lateral variations become more prominent (Figure 17). The results from this example are comparable with those of Li and Fomel (2015) but achieved with approximately one-tenth of the cost. They further attest the validity of the proposed method.

input-field
input-field
Figure 14.
The inputs of the proposed conversion method for the GOM field data example: Dix velocity squared $ w_{dr}$ and its gradients.
[pdf] [png] [scons]

finalvcompare
finalvcompare
Figure 15.
The difference between the estimated velocity squared using the proposed method and the Dix-inverted velocity squared (top). A comparison of the estimated interval $ w(x,z)$ from the proposed method (middle) and from the optimization approach (bottom) for the GOM field data example.
[pdf] [png] [scons]

finalcompare
finalcompare
Figure 16.
A comparison of the final converted seismic images from the proposed time-to-depth conversion method (left) and from the PSDM using the estimated interval velocity for the GOM field data example.
[pdf] [png] [scons]

cig
cig
Figure 17.
A comparison of CIGs generated from PSDM at 8 $ km$ and 11.25 $ km$ using the conventional Dix-inverted velocity and the estimated interval velocity from the proposed method. In deeper sections, where there is prominent lateral variations, we can observe an improvement in flatness of the CIGs from the estimated $ w(x,z)$ .
[pdf] [png] [scons]


next up previous [pdf]

Next: Discussion Up: Examples Previous: Land field data example

2018-11-16