Adjoint Tomography of South America based on 3D Spectral-Element Seismic Wave Simulations

The great improvement in computational power experienced in the last decades combined with the ever growing seismographic coverage worldwide are allowing the Earth's internal structure to be better constrained by new velocity models. Full-waveform Inversion, the state-of the-art method for seismic imaging, migration, source inversion and tomographic studies was unfeasible in the past, but nowadays is becoming the standard method in many seismic and seismological studies.

In South America, the ray coverage has largely increased in the last years, particularly after 2011 with the installation of the Brazilian Seismographic Network. We are using CMT solutions from the global catalog (Dziewonski, Chou and Woodhouse, 1981; Ekström, Nettles and Dziewonski, 2012) for 96 earthquakes with magnitudes ranging from 6.0 to 7.0 Mw, recorded in the 2011-2016 period and carefully selected to maximize ray coverage, signal-to-noise ratio and CMT solution reliability (based on cross-correlation and misfit comparison of synthetic seismograms calculated using PREM, s40rts and S362wmani global velocity models with seismic stations records) to perform a 3D Full-waveform Adjoint Tomography of the South American continent and surrounding oceans using SPECFEM3D Global software (Komatitsch and Tromp, 2000).

Our starting model is GLAD-M15, a first-generation global adjoint tomography model (Bozdag, 2016). The time-window selection will soon be upgraded using machine learning algorithms. We believe this new tomographic model will bring many improvements in coverage, accuracy and resolution over previous models based solely on ray theory, fundamental mode group velocities of surface waves and partitioned waveform inversion, like Feng et al. (2007).