Full-waveform inversion with quantified uncertainty in geophysics

In recent years, full-waveform inversion (FWI) has become an important imaging technique in geophysics. FWI fully exploits the information encoded in seismic waveforms to estimate physical parameters of subsurface. Most efforts are devoted to infer the spatial variations of fields such as velocity, density, anisotropy, and attenuation in large regions, facing huge computational costs. However, in some applications it would be enough to determine the boundaries enclosing regions where abrupt changes take place, allowing for a considerable cost reduction. Moreover, without crucial information about uncertainty, it is difficult to interpret inverted subsurface models. Here, we propose an object based full wave form inversion scheme that combines topological methods, optimization strategies and Bayesian approaches. We test our method localizing different types of reservoirs in 2D soil studies with quantified uncertainty about their location, size, geometry and nature