A dual formulation of wavefield reconstruction inversion for large-scale seismic inversion

Many currently proposed seismic inversion techniques that focus on robustness with respect to the background model choice are not appropriate for large-scale 3D applications, and methods computationally feasible industrial problems, such as full-waveform inversion, notoriously limited by convergence stagnation require adequate starting models. We have developed a novel solution is scalable less sensitive models or inaccurate parameters (such anisotropy) typically kept fixed during inversion. It based dual reformulation of classic wavefield reconstruction whose empirical these issues well documented in literature. Although version suited three dimensions because it leverages expensive frequency-domain solvers wave equation, our proposal allows deployment state-of-the-art time-domain finite-difference methods, potentially sufficiently mature industrial-scale problems.