Path-specific, dispersion-based velocity models and moment tensors of moderate events recorded at few distant stations: Examples from Brazil and Greece

•Waveform inversion at large distances needs specific velocity models which can be derived from surface-wave dispersion.•Moment tensors (MT) from waveform inversion should be checked against polarities.•Waveform simulations using well constrained MT's can validate velocity models.•Single-station MT inversion is risky even with high-quality velocity models.•Frequency range test to check the stability of the solution.

Centroid moment tensor (CMT) determination in intraplate regions like Brazil can be very difficult, because earthquakes are often recorded just at few and distant stations. This paper introduces a methodology for datasets like that. The methodology is based on waveform inversion in which each source-station path has its own velocity model. The 1-D path-specific velocity models are derived from the Rayleigh- and Love-wave dispersion curves. The waveform inversion is accompanied by posterior check of numerous P-wave first-motion polarities. An important innovation is the use of so-called frequency range test. The test basically consists in calculating CMT's for many different frequency ranges to assess the stability and uncertainty of the solution. The method is validated on two Brazilian earthquakes and a well-known Greek event. An offshore event (mb 5.2) in SE Brazil is inverted with four stations, at epicentral distances 300-400 km. The other Brazilian earthquake (mb 4.8 in Central Brazil) is even more challenging - only two broadband stations at 800-1300 km are at disposal for waveform inversion. The paper unambiguously demonstrates that the path-specific velocity models significantly increase the reliability of the CMT's. While standard models (e.g. IASP91) typically allow waveform modeling up to epicentral distances of the order of a few (∼10) minimum shear wavelengths (MSW), using the path-specific velocity models we successfully inverted waveforms up to > 20 MSW. Single-station waveform inversions are thoroughly tested, but multi-station joint inversions are shown to be preferable. The new methodology of this paper, providing a reasonable estimate of focal mechanisms and their uncertainties in case of highly limited waveform data, may find broad applicability in Brazil and elsewhere.