next up previous [pdf]

Next: Discussion and Conclusions Up: Almomin et al.: WEMVA Previous: Method

Synthetic Examples

In this synthetic example, we will compare three inversion results after 20 iterations: low-frequency back-scattered FWI, correlation-based WEMVA (Almomin, 2011), and the combined gradient of WEMVA and the back-scattered FWI weighted by a scalar. The receiver spacing is 20 m, the source spacing is 80 m, and the temporal sampling is 2 ms. Figure 6(a) shows the true slowness anomaly we are inverting for. There is one reflector at the bottom of the slowness model. Figure 6(b) shows the results of low-frequency back-scattered FWI. Although it has the correct anomaly shape, the result has strong side-lobes and the amplitude is still very weak, because convergence is very slow.

wemva30tv wemva30bp
wemva30tv,wemva30bp
Figure 6.
The true slowness model with a cosine perturbation (a) and the inversion results using low-frequency back-scattered FWI (b).
[pdf] [pdf] [png] [png]

Figure 7(a) shows the results of correlation-based WEMVA. As shown earlier, the gradient has low vertical resolution. Also, the anomaly has negative side-lobes. However, the amplitude of the anomaly is much stronger than with back-scattered FWI, and the kinematics of the reflector below converges much faster. Figure 7(b) shows the results of the combined gradient. The amplitude of the anomaly is even stronger and is closer to the true anomaly than is the WEMVA inversion. Furthermore, the side-lobes are much weaker, and the model has much better vertical resolution. Notice that figures 6(b), 7(a) and 7(b) have different color scales.

wemva30fp wemva30fbp
wemva30fp,wemva30fbp
Figure 7.
Slowness inversion results using (a) correlation-based WEMVA inversion and (b) the combined gradient.
[pdf] [pdf] [png] [png]


next up previous [pdf]

Next: Discussion and Conclusions Up: Almomin et al.: WEMVA Previous: Method

2011-05-24