WEMVA eliminates synthetic FEAVO

As described by Sava and Symes (2002), the image perturbation to be fed into WEMVA can be created by Target Image Fitting (TIF), or by Differential Semblance Optimization (DSO, Symes and Carazzone (1991), Chauris and Noble (1998)). We chose the TIF approach because the purpose of this test was to determine conclusively whether WEMVA-produced velocity models can eliminate realistic FEAVO effects independent of image perturbation extraction. Since the dataset is synthetic and the correct velocity model known, we were able to generate an optimal image perturbation by subtracting the image migrated with the current velocity from the one migrated with the correct velocity.

We used the ``WEMVA works'' dataset. The starting estimate for the first WEMVA iteration was the 2000 m/s background velocity. FEAVO effects in a slice from the angle-domain image migrated with the background velocity are shown in the upper panel of Figure .

We ran 10 conjugate-gradient solver iterations, then updated the velocity model. The resulting model is shown in the upper panel of Figure . Its peak anomalies as departures from the background, in m/s, from left to right, are: -74, -84, +99. We migrated with this velocity model. FEAVO effects in a slice from this image are shown in the middle panel of Figure . They are weaker now, but still visible.

We performed another WEMVA inversion loop, with 10 solver iterations, starting from the updated velocity model. The resulting velocity model is shown in the bottom panel of Figure . Its peak anomalies as departures from the background, in m/s, from left to right, are: -105, -123, +149. The angle-domain image obtained by migrating with this new velocity model are shown in the lower panel of Figure . The FEAVO effects are no longer recognizable. When WEMVA's assumptions, discussed in the next section, are satisfied and the image perturbation can be extracted in a satisfactory manner, the inversion process converges and the resulting velocity field is accurate enough to reliably eliminate FEAVO effects.

 

f3
Figure 3 Upper panel: velocity model updated after one WEMVA iteration. Lower panel: velocity model updated after a second WEMVA iteration. Both panels are represented in the same color scale.

 

f4
Figure 4 Angle-domain slices at a depth of 155m. Upper panel: from the image obtained with the constant background velocity; Middle panel: from the image obtained with the velocity in the upper panel of Figure ; Lower panel: from the image obtained with the velocity in the lower panel of Figure .