RESUMO
The interaction of ocean surface waves produces pressure fluctuations at the seafloor capable of generating seismic waves in the solid Earth. The accepted mechanism satisfactorily explains secondary microseisms of the Rayleigh type, but it does not justify the presence of transversely polarized Love waves, nevertheless widely observed. An explanation for two-thirds of the worldwide ambient wave field has been wanting for over a century. Using numerical simulations of global-scale seismic wave propagation at unprecedented high frequency, here we explain the origin of secondary microseism Love waves. A small fraction of those is generated by boundary force-splitting at bathymetric inclines, but the majority is generated by the interaction of the seismic wave field with three-dimensional heterogeneity within the Earth. We present evidence for an ergodic model that explains observed seismic wave partitioning, a requirement for full-wave field ambient-noise tomography to account for realistic source distributions.
RESUMO
Ultrasound computed tomography (USCT) is a noninvasive imaging modality that has shown its clinical relevance for breast cancer diagnostics. As opposed to traveltime inversions, waveform-based inversions can exploit the full content of ultrasound data, thereby providing increased resolution. However, this is only feasible when modeling the full physics of wave propagation, accounting for 3D effects such as refraction and diffraction, and this comes at a significant computational cost. To mitigate this cost, a crosstalk-free source encoding method for explicit time-domain solvers is proposed. The gradient computation is performed with only two numerical "super" wave simulations, independent of the number of sources and receivers. Absence of crosstalk is achieved by considering orthogonal frequencies attributed to each source. By considering "double-difference" measurements, no a priori knowledge of the source time function is required. With this method, full-physics based 3D waveform inversions can be performed within minutes using reasonable computational resources, fitting clinical requirements.