Joint imaging with streamer and ocean bottom data

Synthetic Example

Figures 4 (a) and 4 (b) shows the velocity and reflectivity models used. The model is 2.8 km deep and 8 km wide with a spacing of 10 m. To simulate a monitoring survey, we assume there is an existing production structure that spans the location from x=3950-4000 m. The water depth is about 1000 m. Figure 3 shows a plane map view of the streamer survey and the OBN survey.

plainview
Figure 3. (a) Plane map view of streamer survey during undershooting. Streamer must be at least 500m from the platform. (b) Plane map view of OBN survey. The shot carpet (red circles) is much denser than the receiver grid (green open cicles).

Figure 3 a shows a snap shot of the streamer survey during undershooting. In our 2D synthetic study, we create synthetic data along with shots and receiveres lined along the crossline direction, cutting through the production platform. Shots run from 0 to 8000 m on the sea surface at an interval of 350 m along the 2D line. There are 8 receivers to the left and to the right of each shot, spaced at 50 m apart. This geometry captures the same sub-surface illumination range as a dual-shot towed-streamer survey along the crossline direction. The towed-streamer cannot be within 500 m from the platform, which leads to illumination gap. We create two additional shots undershooting near the platform.

For the OBN survey, receivers are placed from x=2000 to 6300 m at the ocean-bottom with a 400 m spacing. The shot carpet spans the entire model along the sea surface except in the area within 50m of the platform. For the OBN data, we forward-modeled the down-going mirror signal only. We added 10 percents of the up-going signal as noise to the data to simulate imperfect up-down separation.

Figure 4 (c) and 4 (d) show the reverse-time migration (RTM) image of the streamer survey and the OBN survey, respectively. By comparing with the original model, we can see different kinds of artifacts in the two results. The streamer image looks noisy in general because of the sparse shot spacing. The area beneath the obstruction is poorly illuminated. The OBN image is less noisy because of the dense shot carpet. However, there is a spurious event at depth z=1200 m. This spurious reflector is caused by imperfect up-down separation. In addition, the dipping reflector near x=5500 m and z=2400 m is not well illuminated in the OBN image.

panel1
Figure 4. (a) The velocity model, (b) the reflectivity model, (c) the streamer RTM image, and (d) the OBN RTM image with mirror imaging. Even with undershooting, the streamer image shows some artifacts (x) near the production structure (y). The residual up-going OBN energy in the down-going OBN data create some artifacts (z).

For comparison, we individually perform linearized inversion on the two data sets. The resulting images are shown in (a) and (b) of Figure 5. We can see an overall improvement from the migration image in (c) and (d) of Figure 4 to the inversion image in (a) and (b) of Figure 5. In general, we can see that the inverted images contain higher resolution and better amplitude information than the migrated images. Linearized inversion can discrimmiate certain artifacts and remove them from the image. For example, the annotate artifacts in Figure 4 c) are suppressed in 5 (a). For the OBN survey, one noticeable improvement is the wider illumination. In the next section, we will explore the effect of joint inversion on the output image.

Joint imaging with streamer and ocean bottom data