Teleseismic Full Waveform Inversion for High-Resolution Subduction Zone Imaging

  • Venue:


  • Date:

    May 28, 2024

  • Speaker:

    Li-yu Kan

    Postdoc in National Central University, Taiwan

  • Time:

    9:30 am


Thanks to the densification of seismic deployments and the advances in high-performance
computing, full waveform inversion (FWI) of teleseismic waves at the regional scale is now
feasible. Indeed, teleseismic FWI provides quantitative multi-parameter tomographic models
such as density and seismic velocities, with a resolution limit down to half a wavelength.
Previous applications of teleseismic FWI have successfully imaged the lithospheric structure of
the Pyrenees and the Alps subduction system, however, they still suffered from the different
resolution of each model parameter. In this work, we make improvements to the FWI method
that allows taking into account the correlation between the model parameters. We consider a
complete model covariance matrix whose non-diagonal terms describe the correlation between
each pair of model parameters. We first perform synthetic inversion experiments on a simple
subduction model to validate the new algorithm. As a result, with this new approach, we can
obtain better-reconstructed models with fewer numerical artifacts. Next, we apply it to
teleseismic waveforms recorded by the temporary seismic deployment in Cascadia. The final
models in density, Vp, and Vs reveal the lithospheric structures with unprecedented resolution.
We successfully image the thin low-velocity layer associated with the Juan de Fuca plate. Below
40 km depth, seismic velocities and density increase progressively to the values of a typical
mantle, which corresponds to the eclogitization of the oceanic crust. Finally, the new FWI
method is applied to teleseismic waveforms recorded by several temporary experiments
deployed in southern Peru, which covers both the normal and flat subduction segments in the
northern Central Andes. The final density, VP , and VS models provide new insights into the
structure of the Altiplano and the geometry of the subducting Nazca plate. The low velocities
observed beneath the Altiplano in the southern part of the model suggest that the lithosphere has
been removed, which could explain the recent uplift in this part of the Andes. We also observe a
low-velocity anomaly under the flat segment, which could be the maintenance of a shallow dip
for the subducted plate. The improved FWI method is able to provide high-resolution
tomographic models on a lithospheric scale, as highlighted by both the synthetic experiment and
real-data application. This is particularly important for complicated subduction systems that
involve the processes such as serpentinization, eclogitization, and partial melting.