High-resolution imaging of crustal melts using 3D full-waveform seismic inversion

A newly practical seismic imaging technique, 3D full-waveform inversion (FWI), now has the ability to image zones of melt and melt pathways throughout the crust with a better resolution than any other geophysical method. 3D FWI has recently changed practice within the petroleum industry where it is used to obtain high-resolution high-fidelity models of physical properties in the sub-surface that are both interpreted directly and used to improve the migration of deeper reflections. This technology has been spectacularly successful in improving the imaging of reservoirs beneath shallow heterogeneities produced by, for example, gas clouds, buried fluvial channels, carbonate reefs and salt bodies. During FWI, the sub-surface model is recovered principally by using the low-frequency transmitted, refracted portion of the wavefield which is most sensitive to the macro-velocity structure. In the petroleum industry, these inversions are now routinely performed using long-offset surface-streamer and ocean-bottom data to maximum source-receiver offsets of about 15 km, leading to a maximum penetration depth of around 5 km. Using longer offsets, it is possible to extend this technology to image deeper crustal targets. Localised zones of partial melt produce large changes in p-wave and s-wave properties that are restricted in their spatial extent, and that therefore form ideal targets for 3D FWI. We have performed a suite of tests to explore the use of 3D FWI in imaging melt distribution beneath the active volcano of Montserrat. We built a model of the subsurface using a 3D travel-time tomographic model obtained from the SEA CALIPSO experiment. We added two magma chambers in accordance with a model obtained using surface-elevation changes and geochemical data. We used a wide-angle, wide-azimuth acquisition geometry to generate a fully-elastic synthetic seismic dataset, added noise, and inverted the windowed transmitted arrivals only. We used an elastic code for the forward modelling but an acoustic code for the inversion. We are able to incorporate both anisotropy and attenuation into our acoustic code, but we did not do that here. We ran a suite of 3D FWI tests, using a range of frequencies, length scales, amplitudes and aspect ratios, in order to explore which experimental geometries and inversion strategies are best able to recover fine-scale crustal heterogeneity. We also ran a suite of resolution tests across the entire model. Our results demonstrate that 3D FWI of wide-angle seismic data can recover sub-surface variations in p-wave velocity on a scale of about half the seismic wavelength in the crust, and can detect isolated features that are two to three times finer than this. For the frequencies that we expect to be able to record, and that we have previously inverted successfully in field data, we are able to detect features that are only a few tens of metres in lateral extent, and to resolve these into distinct features on a vertical and lateral scale of about 100 m. 3D FWI is a maturing technology that will open a new era of seismic experiments that can detect melt, recover its geometry, and determine its physical properties, with much greater resolution and fidelity than has previously been possible.