Crustal-scale Vp velocity model building for a sparse regional OBS survey using full-waveform inversion: a case study from Japan Trench

Ocean Bottom Seismometers (OBS), in conjunction with an active source, have for a longtime been used to investigate the crust and upper mantle. Conventionally a number of Seismometers are deployed along a line with relatively large separation (several kilometers) between each instrument. A large air-gun array is then used to create a number of recordings with distance between the receiver and source varying from zero up to several hundred kilometers.

The data is usually analyzed with tomographic methods leading to a P-wave velocity model. Tomographic methods tend to produce velocity models with low resolution and lack of details. To increase resolution the Full Waveform Inversion (FWI) can be used. Velocity models estimated using FWI shows better resolution and are better constrained than tomographic models.

Our work sets out to develop a FWI workflow for a sparsely sampled OBS data. We demonstrate high resolution crustal-scale velocity model building workflow based FWI for a deep water and sparse (6km) sampled 300 km long wide-angle OBS data. We use an initial model traveltime tomography inversion. We apply waveform inversion to constrain crustal and upper mantle layers more confidently and with increased resolution compared to conventional traveltime tomography with layer-based parameterization. FWI allows to reduce significantly data-fit error (i.e. rms error) and explains better larger offsets of an observed data compared to traveltime tomography.

The workflow is developed and applied to data acquired in 2009 in the Japan Trench where the 130-150 Myrs old oceanic plate is subducting under Eurasia. Subduction zones plays an important role in Earth tectonics . Traveltme tomography inversion was used the 2009 Japan Trench dataset to construct velocity models and showed that P-wave velocities close to the trench axis, where the plate bends downward, are systematically lower than the P-wave velocities at larger seaward distances. This is interpreted as water penetration into the plate from the seafloor, causing serpentinization which decreases P-wave velocities. However, tomographic models does not reveal details of the velocity model, only smoothed averages. We hope that a velocity model generated by FWI has sufficiently increased resolution to reveal more details of the water penetration and geological setting along the Japan Trench.