3D Elastic FWI with a Non-Linear Model Constraint: Application to a Real Complex Onshore Dataset
P. Trinh, R. Brossier, L. Lemaistre, L. Métivier and J. Virieux
Event name: 81st EAGE Conference and Exhibition 2019
Session: Poster: Full Waveform Inversion B
Publication date: 03 June 2019
Info: Extended abstract, PDF ( 1.68Mb )
Price: € 20
Full waveform inversion (FWI) of 3D onshore targets is very challenging due to the complex free-surface-related and heterogeneities effects. In such areas, the seismic wavefield displays converted and back-scattering energy, with a strong contribution of surface waves. A time-domain spectral-element-based approach is used for accurate elastic wavefield simulation in mountain areas. The challenges of the elastic multi-parameter FWI in complex land areas are highlighted through a real foothill dataset. Various practical issues such as a significant amount of noise in the data, sparse acquisition and topography representation uncertainty must be considered. We illustrate how additional prior information such as well logs and geological knowledge can be used to assist the FWI workflow. First, a non-linear model constraint on the ratio Vp/Vs, relying on the logs information, may better constrain the Vs based on the stable Vp estimation. Second, appropriate gradient preconditioning design is also crucial for meaningful model parameters reconstruction. At the current stage, the inversion mainly focuses on the early-body waves, which are gradually incorporated through a progressive time-windowing. The strategy helps to mitigate the cycle-skipping issue at the later arrivals and maximize the considering data quantity, and eventually improve the model estimation.