While low-wavenumber components of an Earth model can be constructed using reflection tomography, full-waveform inversion (FWI) is capable of deriving a model with high accuracy that covers the full range of the spatial spectrum, from low to high wavenumbers, supported by the seismic data. This data-domain algorithm minimizes the mismatch in both amplitude and phase between the recorded data set and a numerically simulated one. The inversion is commonly solved using a local gradient-based iterative scheme, like the steepest decent or conjugate-gradient method. The model update is proportional to the gradient of the objective function and is computed by crosscorrelating a forward extrapolated source wavefield with a backward propagated data residual, which is computationally equivalent to reverse time migration of the data residual.

You do not currently have access to this article.