| Summary: | Full-waveform inversion (FWI) is a powerful nonlinear tool for quantitative estimation of high-resolution high-fidelity models of subsurface seismic parameters, typically P-wave velocity. A solution is obtained via a series of iterative local linearised updates to a start model, requiring this model to lie within the basin of attraction of the solution space's global minimum. The consideration of seismic anisotropy during FWI is vital, as it holds influence over both the kinematics and dynamics of seismic waveforms. If not appropriately taken into account, then inadequacies in the anisotropy model are likely to manifest as significant error in the recovered velocity model. Conventionally, anisotropic FWI either employs an a priori anisotropy model, held fixed during FWI, or uses a local inversion scheme to recover anisotropy as part of FWI; both of these methods can be problematic. Constructing an anisotropy model prior to FWI often involves intensive (and hence expensive) iterative procedures. On the other hand, introducing multiple parameters to FWI itself increases the complexity of what is already an underdetermined problem. As an alternative I propose here a novel approach referred to as combined FWI. This uses a global inversion for long-wavelength acoustic anisotropy, involving no start model, while simultaneously updating P-wave velocity using mono-parameter local FWI. Combined FWI is then followed by multi-parameter local FWI to recover the detailed final model. To validate the combined FWI scheme, I evaluate its performance with several 2D synthetic datasets, and apply it to a full 3D field dataset. The synthetic results establish the combined FWI, as part of a two-stage workflow, as more accurate than an equivalent conventional workflow. The solution obtained from the field data reconciles well with in situ borehole measurements. Although combined FWI includes a global inversion, I demonstrate that it is nonetheless affordable and commercially practical for 3D field data.
|
|---|
