Recovering physical property information from subduction plate boundaries using 3D full-waveform seismic inversion

Our understanding of subduction margin seismogenesis has been revolutionised in the last couple of decades with the discovery that the size of the seismogenic zone may not be controlled simply by temperature and a broad spectrum of seismic behaviour exists from stick-slip to stable sliding. Laboratory and numerical experiments suggest that physical properties, particularly fluid pressure may play an important role in controlling the seismic behaviour of subduction margins. Although drilling can provide information on physical properties along subduction thrust faults at point locations at relatively shallow depths, correlations between physical properties and seismic velocity using rock physics relationships are required to resolve physical properties along the margin and down-dip. Therefore, high resolution seismic velocity models are key to recovering physical property information at subduction plate boundaries away from drill sites. 3D Full waveform inversion (FWI) is a technique pioneered by the oil industry to obtain high-resolution high-fidelity models of physical properties in the sub-surface. 3D FWI involves the inversion of low-frequency (>2 to <7 Hz), early arriving (principally transmitted) seismic data, to recover the macro (intermediate to long-wavelength) velocity structure. Although 2D FWI has been used to improve velocity models of subduction plate boundaries before, 3D FWI has not yet been attempted. 3D inversions have superior convergence and accuracy, as they sample the subsurface with multi-azimuth multiply-crossing wavefields. In this contribution we perform a suite of synthetic tests to investigate if 3D FWI could be used to better resolve physical property information along subduction margin plate boundaries using conventionally collected 3D seismic data. We base our analysis on the Muroto Basin area of the Nankai margin and investigate if the acquisition parameters and geometry of the subduction margin render 3D seismic data collected across this basin in 1999 suitable for future 3D FWI. We build a 3D model of the sub-surface based on an existing velocity model that was used to migrate these data (Tsuji et al. 2000, JGR). We then add a low P-wave velocity layer along the décollement, which is supported by ODP core data but does not feature in the current seismic velocity model, to test if it could be recovered using 3D FWI. We use the same acquisition parameters as in the 1999 seismic survey (including a 6 km long streamer) to generate a fully-elastic synthetic seismic dataset, added noise and inverted the windowed transmitted arrivals only. We also ran a suite of resolution tests across the model. The results show that 3D FWI of conventionally collected 3D seismic data across the Muroto Basin would be capable of resolving variations in P-wave velocity along the décollement of the order of half the seismic wavelength at the plate boundary. This is a significant improvement on conventional travel-time tomography which resolves to the Fresnel width. In this presentation we will also postulate on the optimal 3D FWI experiment design for the next generation of 3D seismic surveys across subduction margins as a guide for those embarking on new data collection.