Until recently, most of the seismic tomographic modeling has been addressing the question of lateral heterogeneity either in P- or S-wave velocities. The S-wave velocity variations are larger and hence provide stronger signal on long-period waveforms. The direct P travel times, being the first arrivals, on the other hand, are most frequently reported in the International Seismological Centre (ISC) Bulletins. In mineral physics experiments, the variation in bulk velocity is more often measured. To better understand the differences between δvP and δvP patterns and better link the results from mineral physics to those of seismic tomography, we formulate the inverse problem in terms of relative perturbations in the shear velocity and bulk sound velocity . We use a large data set which consists of waveforms, waveform-derived travel times and travel times from the ISC Bulletins. The earthquakes are relocated using corrections for lateral heterogeneity. The events which cannot be reliably determined are discarded. The model is defined as spherical harmonics to Degree 12 horizontally and as Chebyshev polynomials to order 13 radially, for both shear and bulk sound velocity. The inversion is performed under smoothness constraint. The resolution tests and bootstrapping analysis indicate that the model is well recovered, particularly at long wavelength.
The results indicate a much larger variability of shear than bulk sound velocity. The model explains observations well. The most intriguing result obtained in this study is that the variations in shear velocity and bulk sound velocity are negatively correlated in the lowermost mantle. The explanation is not very clear. From the mineral physics point of view, it is not unlikely that this could be explained in terms of thermal variation, even though we are unwilling to rule out the possibility of large wavelength compositional variations.