Hybrid frequency domain full waveform inversion using born sensitivity kernels

Full Waveform inversion (FWI) using the scattering integral (SI) approach is an explicit formulation of the inversion optimization problem. The inversion equations are straightforward and the dependence on the data residuals and model parameters is clear. However, the biggest limitation with this approach is the huge computational cost for exploration seismology applications. To deal with this issue, we propose a hybrid implementation of the frequency domain FWI using the Born sensitivity kernels. Specifically, we use the sensitivity kernels computed from dynamic ray-tracing to build the gradient. We compute also the truncated Gauss-Newton update direction using the kernels without extra wavefield modeling steps. Considering that in FWI long-to-intermediate wavelengths are updated during the first iterations, using a transmission experiment we obtain accurate inverted models. The inversion managed to develop the anomaly embedded in the homogeneous background medium. The truncated Gauss-Newton updates helped in the fast convergence. With this approach we managed to reduce the computational cost and the memory requirements. For more complex models, the hybrid inversion method help improving the initial model with little cost compared to conventional SI inversion.