Full waveform inversion (FWI) usually gives a poor update of the sediment velocity when in the close vicinity of the top of salt (TOS) reflection. This phenomenon is a common practical challenge and is due to the strong velocity contrast between the sediment and salt—although its exact cause is not yet well understood. We investigated the relationship between FWI’s sediment velocity update and the role of salt insertion, namely the accuracy of the salt interface used in the FWI input velocity model. The results indicated that the initial model with salt inserted helps update the sediment model, and furthermore, using a more accurate TOS (i.e., the TOS interpretation used to insert the salt is closer to the true TOS) provides a better sediment velocity update. However, an accurate TOS is not available unless the sediment velocity, particularly directly above the TOS, has been correctly updated. Therefore, we propose a workflow to obtain a more accurate TOS for the FWI sediment velocity update using iterative FWI and salt interpretation. Using 2D synthetic data and 3D real data, we demonstrate that our workflow yields a better FWI update above the salt compared to using the sediment model as the initial model.
FWI aims to minimize the misfit of phase and amplitude between real shot gathers and synthetic shot gathers (Lailly, 1983; Tarantola, 1984; Sirgue and Pratt, 2004; Virieux and Operto, 2009). Updating the velocity in areas of high impedence and/or velocity contrast, such as the sedimentsalt boundary, is challenging for FWI. Two approaches are commonly used to address this overburden velocity update problem. Kapoor et al. (2012) used sediment models created from ray-tracing tomography as FWI input. Chen et al. (2014) used a salt model approach that ran an initial pass of FWI with the sediment model to update the sediment velocity before creating a salt model for the second pass of FWI to update the sediment velocity again. However, few studies have been performed to understand the impact of the TOS interpretation on the sediment velocity update using FWI. By using band limited data and conventional FWI, the TOS singularities are not automatically updated at each iteration.