140 Seismic Imaging velocity (from Figure 5.6, top): only the density model is updated. After 30 nonlinear quasi-Newton FWI inversions, the final density model is highly oscillating and tends towards either low or high values (Figure 5.13). The cemented structure cannot be interpreted. As expected, this means that a density model cannot be replaced by a velocity model. Velocity has an influence on the kinematics (event positioning) and dynamics (amplitude) of wave propagation, whereas density only plays a role with regard to amplitude. Thus, a pure density-FWI scheme cannot update an incorrect kinematic part. The objective function did not reach a value as low as before (Figures 5.8 and 5.14). Note that the value of the objective function itself is not very significant; the data residuals should be analysed as in Figure 5.10. Figure 5.13 As for Figure 5.6 bottom, except for the density model (in kg/m3). The velocity model has not been updated (Figure 5.6, top). Figure 5.14 As for Figure 5.8, but where only the density model has been updated. Towards applications on real data Before concluding, we discuss the additional steps required for applications on real data. Most of the published results are related to the marine environment. Water in the shallow part prevents energetic surface waves, which remains a challenging problem for FWI (Brossier et al., 2009). As FWI is a data misfit approach, the amplitudes play an important role. We discussed the importance of density contrasts in the synthetic example. Diving waves are sensitive to velocities but not to density. This is one of the main reasons why they are
RkJQdWJsaXNoZXIy NjA3NzQ=