2:45 PM - 3:00 PM
[SSS06-04] Upper mantle discontinuities estimated from Bayesian inversions using receiver functions and surface wave dispersions
Keywords:S receiver functions, Receiver functions, Surface wave, Upper mantle discontinuities, Trans-dimensional Bayesian inversion
Seismic discontinuities in the upper mantle reflect the evolutional history of tectonic plates and dynamics in the mantle. They have been investigated using receiver functions (RFs) and surface wave dispersion (SWD). Recent seismological studies have used RFs (P-RF for P-wave; S-RF for S-wave) to effectively detect the upper mantle discontinuities such as the Lithosphere-Asthenosphere Boundary (LAB) and the enigmatic Mid-Lithospheric Discontinuities (MLDs). Surface wave tomography based on SWD data has also been widely used to investigate the three-dimensional distribution of S-wave speeds in the upper mantle.
In recent years, trans-dimensional hierarchical Bayesian inversion has become a popular approach in seismology. Such a statistical approach enables us to treat the number of model parameters and data noises as unknown variables, which have been empirically determined in traditional methods. Some earlier studies based on the Bayesian inversion have used both P-RF and SWD to estimate the S-wave velocity structure, including internal boundaries in the upper mantle. However, such conventional RF analysis has employed seismograms in limited ranges of epicentral distances and/or back-azimuths to avoid severe moveout effects. Therefore, the number of employed data tends to be limited in earlier studies. Also, such analysis can be challenging to be applied to temporary/transportable seismic arrays, whose observation periods are generally limited to one or two years. Furthermore, S-RF has rarely been used in RF inversions due to the difficulty in its analysis despite its good sensitivity to the upper mantle discontinuities.
In this study, we develop an improved RF analysis method with moveout corrections to eliminate the limitation of distance and azimuth, which can be applied to both P-RFs and S-RFs. Then, the RF data and multimode SWD are jointly used to estimate 1-D radially anisotropic S-wave speed models in the upper mantle. Results from various synthetic experiments indicated that the joint inversions using P-RFs and S-RFs and multimode SWD data of Rayleigh and Love waves could be preferred for recovering 1-D S-wave speed models, including the upper mantle discontinuities and radial anisotropy.
We applied the joint inversion using P-RF and/or S-RF and multimode SWD to a permanent seismic station (MBWA) in western Australia. Several resultant models from P-RF (with little restrictions in data-selection criteria) and multimode SWD suggested the dependency on the P-wave incident-angle, which can be a critical point in using the RF data with moveout corrections. A suitably adjusted incident-angle enables us to obtain reasonable models and find major upper mantle discontinuities such as LAB and Lehmann discontinuity. These results suggest that the joint inversions using the P-RFs with moveout corrections and multimode SWD can be applied to temporary and/or transportable seismic stations. On the other hand, the joint inversions incorporating S-RF, which is expected to provide a better model from our synthetic experiments, unexpectedly resulted in unstable S-velocity models with significant uncertainties compared to those from P-RF. It may be attributed to the fact that S-RF does not reflect the seismic structure right beneath the seismic station due to the largely inclined ray-paths of converted P-phases from the incident S-wave, violating the implicit assumptions of the layered structure in the RF analysis.
In recent years, trans-dimensional hierarchical Bayesian inversion has become a popular approach in seismology. Such a statistical approach enables us to treat the number of model parameters and data noises as unknown variables, which have been empirically determined in traditional methods. Some earlier studies based on the Bayesian inversion have used both P-RF and SWD to estimate the S-wave velocity structure, including internal boundaries in the upper mantle. However, such conventional RF analysis has employed seismograms in limited ranges of epicentral distances and/or back-azimuths to avoid severe moveout effects. Therefore, the number of employed data tends to be limited in earlier studies. Also, such analysis can be challenging to be applied to temporary/transportable seismic arrays, whose observation periods are generally limited to one or two years. Furthermore, S-RF has rarely been used in RF inversions due to the difficulty in its analysis despite its good sensitivity to the upper mantle discontinuities.
In this study, we develop an improved RF analysis method with moveout corrections to eliminate the limitation of distance and azimuth, which can be applied to both P-RFs and S-RFs. Then, the RF data and multimode SWD are jointly used to estimate 1-D radially anisotropic S-wave speed models in the upper mantle. Results from various synthetic experiments indicated that the joint inversions using P-RFs and S-RFs and multimode SWD data of Rayleigh and Love waves could be preferred for recovering 1-D S-wave speed models, including the upper mantle discontinuities and radial anisotropy.
We applied the joint inversion using P-RF and/or S-RF and multimode SWD to a permanent seismic station (MBWA) in western Australia. Several resultant models from P-RF (with little restrictions in data-selection criteria) and multimode SWD suggested the dependency on the P-wave incident-angle, which can be a critical point in using the RF data with moveout corrections. A suitably adjusted incident-angle enables us to obtain reasonable models and find major upper mantle discontinuities such as LAB and Lehmann discontinuity. These results suggest that the joint inversions using the P-RFs with moveout corrections and multimode SWD can be applied to temporary and/or transportable seismic stations. On the other hand, the joint inversions incorporating S-RF, which is expected to provide a better model from our synthetic experiments, unexpectedly resulted in unstable S-velocity models with significant uncertainties compared to those from P-RF. It may be attributed to the fact that S-RF does not reflect the seismic structure right beneath the seismic station due to the largely inclined ray-paths of converted P-phases from the incident S-wave, violating the implicit assumptions of the layered structure in the RF analysis.