Benchmarking 3D time- And frequency-domain solvers for FWI applications for different cluster sizes and variable number of sources

We compare the performance of two acoustic solvers in the framework of frequency-domain full-waveform inversion. Using a realistic 3D marine model, we conduct a series of numerical experiments varying the number of nodes and sources taken from a typical 3D seismic acquisition scenario. When the number of shots is small, a time-domain finite-difference (TDFD) solver performs best. For increasing number of shots, upon reaching a number defined by the “line of equal performance,” the frequency-domain finite-difference (FDFD) solver outperforms the time-domain solver and becomes progressively more efficient with further increases. Likewise, for an increasing number of nodes, FDFD initially performs better starting from a smaller number of nodes. Upon reaching the number of nodes defined by the “line,” TDFD becomes the better option. Both solvers deserve a place in a modern FWI toolkit, with the optimal solver selected based on specific configuration dependent on cluster size and number of shots.