Synthetic Tests

In order to quantify the contamination of mantle models that can arise from the use of linear crustal corrections, we carry out a series of synthetic tests. Starting with PREM (Dziewonski and Anderson, 1981) and Moho topography and crustal velocities from CRUST2 (Bassin et al, 2000), we use the cSEM to generate a synthetic dataset of long-period (60 - 400 sec) transverse component waveforms for a set of 41 earthquakes and a realistic station distribution. Though restricted to first-orbit phases, the dataset provides good fundamental-mode and overtone coverage throughout Asia and the westernmost Pacific. Using the PAVA and NACT waveform modeling techniques, we correct the synthetic dataset for crustal effects of CRUST2. We then invert the residual seismograms - which would ideally be very small - for mantle structure. Any retrieved mantle structure is contamination resulting from unmodeled crustal effects.

Figure 2.60 shows variations of isotropic shear speed obtained from an inversion of fundamental mode surface waves and overtones. The model explains 90 percent of the variance in the residual seismograms. Note the strong tectonic character of the mantle contamination. Mantle structure is artificially slow beneath continents, where linear crustal corrections underpredict the effects of crustal structure. Models developed using only transverse component higher modes and only fundamental mode surface waves are nearly identical to the model in Figure 2.60. Significant contamination of mantle structure extends to 100 km depth. Beneath Tibet, structure is different from surrounding mantle at a depth of 200 km.

In order to determine whether the contamination of mantle structure is the result of vertical smearing, we increase the depth parameterization by 5 cubic splines in the upper mantle - 2 in the crust. The retrieved model is substantially similar to that obtained previously, indicating that the mantle contamination is a feature of the waveforms, and not imposed by the parameterization.

Confronted with significant artifacts in mantle structure at depths as great as 200 km, we explore ways of compensating for the inadequacies of linear crustal corrections and minimizing the resulting contamination. Inversion for Moho and seafloor topography/bathymetry has been used in the construction of several global models (e.g. Mégnin and Romanowicz, 2000). Therefore, we invert for mantle velocity structure and Moho and seafloor topography simultaneously. The resulting model shows no significant contamination in the mantle while explaining the data as well as the model shown in Figure 2.60. However, the retrieved Moho and seafloor perturbations are inherently unphysical, since they result from inadequacies of linear crustal corrections and effects of 3D propagation in a laterally inhomogeneous crust.

Figure 2.59: SEM (red) and NACT (blue) synthetic seismograms for the transverse (top) and vertical (bottom) component computed in a PREM mantle and a realistic crustal structure. The star denotes the event while stations squares are colored according to variance between the two synthetics.
\begin{figure}\begin{center}
\epsfig{file=Lekic07_1, width=8cm}\end{center}\end{figure}

Berkeley Seismological Laboratory
215 McCone Hall, UC Berkeley, Berkeley, CA 94720-4760
Questions or comments? Send e-mail: www@seismo.berkeley.edu
© 2007, The Regents of the University of California