Imaging the Earth's Interior based on Seismic Full Waveform Inversion

Information about Earth's interior comes from seismograms recorded at its surface. Seismic imaging based on spectral-element and adjoint-state methods has enabled assimilation of this information for the construction of 3D (an)elastic Earth models. These methods account for the physics of wave excitation and propagation by numerically solving the equations of motion, and require the execution of complex computational procedures that challenge the most advanced high-performance computing systems. Current research is petascale; future research will require exascale capabilities. Our research addresses the long-standing challenge of imaging Earth’s interior based on full-waveform inversion. What we mean by `full-waveform inversion' is combining 3D forward simulations with Fréchet derivatives computed in 3D background models to fit complete three-component seismograms both in phase (traveltime) and amplitude.

The inverse problem consists of reconstructing the characteristics of the medium from -often noisy- observations. A nonlinear functional is minimized, which involves both the misfit to the measurements and a Tikhonov-type regularization term to tackle inherent ill-posedness. Achieving scalability for the inversion process on tens of thousands of multicore processors is a task that offers many research challenges.

We are performing global adjoint tomography using a data set of 1,480 events recorded by tens of thousands of seismographic stations. We observe a significant increase in the resolution of plume and slab features throughout the mantle. The level of detail in our current model enables us to answer some of the questions the geosciences community has been asking about the existence of plumes and hotspots. In this presentation I will present our latest model.