Building an initial model for full waveform inversion using a global optimization scheme

Full waveform inversion (FWI) is often formulated as an optimization problem to derive the best model that can minimize the difference between a field data and the simulated one. Usually, iterative gradient based methods are employed for the problem. These methods require an accurate initial model to avoid cycle-skipping, which cannot be described by the Born approximation. Many methods can be used to build such an initial model for FWI, like reflection tomography and migration-based velocity analysis. In this paper, we propose another way to build the initial model for FWI by using a global optimization method with multigrid technique. This scheme not only takes the advantage of multigrid technique to reduce the model dimension of the inversion, but also benefits from a new highly efficient global optimization method to reduce computation cost. We apply our scheme to a synthetic cross-well data. Numerical results show that this new scheme is robust to build a good initial model to be used for FWI.