next up previous [pdf]

Next: discussion and conclusions Up: Ayeni et al.: Inversion Previous: Cascaded covariance-based preconditioning

Numerical Example

We demonstrate J-PLSI using a modified section of 2D Marmousi model (Figure 1). The objective is to image the seismic amplitude changes at the reservoir using simultaneous source data sets. We assume seismic amplitude changes only within the reservoir and neglect geomechanical changes.

The data consist of two sets of 29 encoded shot records over the 8x8 m grid model at different production stages. In this example, because we use a known background baseline velocity model, both data sets are fully encoded. The random encoding function, with a maximum delay of 1 s, is different for each data set. Shot positions vary randomly between surveys with a maximum displacement of $ 32$ m, whereas the receiver array is the same for both surveys and fixed for all shots. Dips were computed from a single source migrated baseline image (not shown).

marm-vel
marm-vel
Figure 1.
2D Marmousi velocity model.[ER]
[pdf] [png]

Figures 2 shows the baseline and monitor data sets. The downgoing baseline and monitor source wavefields at time 0.52 s are shown Figure 3. The single and simultaneous source migrated monitor images are shown in Figures 4(a) and 4(b), respectively. The inverted simultaneous source monitor image obtained from J-PLSI is shown in Figure 4(c). Figure 5 shows the migrated and inverted time-lapse images. Note that the cross-term artifacts in the migrated time-lapse image (Figure 5(b)) are significantly attenuated in the inverted image (Figure 5(c)). The weighted RMS difference (non-repeatability) between the migrated and inverted images are shown in Figure 6.


next up previous [pdf]

Next: discussion and conclusions Up: Ayeni et al.: Inversion Previous: Cascaded covariance-based preconditioning

2009-09-25