ALBERT

Description: SUMMARY Subduction zones are the places on the Earth where the greatest earthquakes occur. It is now widely accepted that seismic asperities at the interface of subducting plates play a major role in whether a region of a subduction zone behaves seismically, creating strong earthquakes or exhibits aseismic slip. In the last decades, huge advances have been made to decipher the underlying processes; however, the physical parameters along the subduction zone interfaces are still not very well known due to a sparsity of high-resolution experiments and significant costs associated with amphibious seismic experiments. Therefore, synthetic tests are needed to investigate the potential of currently possible high density seismic deployments and to aid future experiment design. As standard local earthquake traveltime tomography in a subduction zone setting cannot resolve structures on a kilometre scale at depth, we explore the suitability of full-waveform inversion (FWI) to increase resolution by using amplitude and phase information in the recorded earthquake seismograms. We apply 2-D-elastic FWI to synthetic earthquake data, using vertical and horizontal receivers, and utilize a realistic model of the seismic velocities at the Ecuadorian margin. We add perturbations within the subducting plates of 4×4 km and 2×2 km in P- and S-wave velocities, respectively, such that potential crosstalk between the two models can be identified. Our results show that the location and amplitude of the perturbations can be reconstructed in high quality down to approximately 70 km depth. We find that the inversion of the S-wave velocity prior to the inversion of the P-wave velocity is necessary to guarantee a good reconstruction of both models; however, the spatial resolution of the S-wave model is superior to the P-wave model. We also show that frequencies up to 1 Hz are sufficient to achieve high resolution. Further tests demonstrate how results depend on the accuracy of the estimated source orientation. Resulting models do not suffer in quality as artefacts near the source positions compensate for the inaccuracy of source orientation. If sources are located within the subducted plate instead of beneath, resulting models are comparable and the convergence of the inversion scheme is sped up. The accuracy of the source position within the model compared to the true earthquake location is critical and implies that earthquake relocation during the inversion process is necessary, in a similar way as in local earthquake traveltime tomography.