The Mobil AVO dataset is a marine dataset collected from the North Sea. The dataset contains strong water-bottom multiples. Before AVO analysis, some processing procedures have been applied to the data to remove the multiple energy. All together, there are 952 CMP gathers with 25m sampling intervals. Each gather contains 60 traces, the offset sampling interval is 50m and the near offset is 288m. The trace length is 1000 samples (sampling rate = 4ms). There are two well logs available at CMP-809 and CMP-1571. In well CMP-809, the density, Vp, and Vs were recorded from 1km to 3.15km. On the basis of this well's information, Dong and Keys 1997 built up a 12-layer (some with vertical gradient) background velocity model for the inversion.
Since our new approach can output common-image gathers (CIG), we initially use this model in our inversion and then check the accuracy of this layered model. As shown in 10, the events from the old velocity model bend upwards, which means the interval velocity in the old model is lower than the correct one. We then conducted a conventional velocity analysis. After converting the RMS velocity model into an interval velocity model, we applied the new velocity model to the dataset and produced the new common-image gather at the same CMP location. It is clear that the new velocity model is better for imaging and inversion (Figure 10).
By stacking the common-offset inversion result, we got a R and section (Figure 11). After obtaining the Kirchhoff inversion result, we estimated the cosine of the specular angle . According to the elastic AVO approximation theory, we estimated of intercept A and slope B, as shown in Figure 12. Furthermore, we combined the intercept and slope sections and produced a fluid-line section, which shows the Vp/Vs anomaly (Figure 13).