CAN THE DIFFERENTIAL SENSITIVITY OF BODY WAVE, MANTLE WAVE, AND NORMAL-MODE DATA RESOLVE THE TRADE-OFF BETWEEN TRANSITION ZONE STRUCTURE AND BOUNDARY TOPOGRAPHY

Large-scale seismic models of the three-dimensional (3D) variations in elastic properties will be biased by topography on mantle boundaries to the extent that volumetric and topographic structures produce simil ar effects in the data. To date, seismic inversions for global-scale 3 D elastic models of the mantle have largely ignored the effect that to pography on the major mantle discontinuities would have on estimating these models. In this paper we address three questions: (1) to what ex tent does unmodeled structure on the 410 and 660 km boundaries bias vo lumetric structure in inversions based on normal mode-mantle wave stru cture coefficients, absolute S-wave travel times, and differential SS- S travel times? (2) Can the differences in the sensitivity of S waves, SS-S phase pairs, and normal mode-mantle wave data be exploited to es timate transition zone volumetric models that are relatively unbiased by topographic structure? (3) Have current volumetric models resolved the trade-off that exists between volumetric and topographic structure s? To address question (1), synthetic experiments were performed which show that volumetric models inferred from normal mode-mantle wave dat a can be biased by an average value of 20-25% in r.m.s. amplitude in t he transition zone relative to current aspherical Earth models, depend ing on the model parametrization employed. The average transition zone volumetric bias of models inverted from absolute travel times from bo th transition zone and lower-mantle bottoming S rays that are imprinte d with the same topographic signatures is reduced by at least by a fac tor of five relative to models inverted from normal mode-mantle wave d ata alone. A reduction in bias by a factor of about four is obtained u sing SS-S phase pairs in which the SS legs bottom in both the transiti on zone and lower mantle. The use of lower-mantle bottoming S rays or SS-S phase pairs with the SS legs bottoming in the lower mantle reduce s the bias only by an average factor of about two to three relative to the normal mode-mantle wave inversion. These estimates of bias reduct ion can vary with the type of damping or smoothness constraints that a re applied in the inversion. With respect to question (2), these resul ts suggest that, in principle, absolute S-wave travel time data can be used to desensitize volumetric inversions to the bias caused by topog raphy on the transition zone boundaries. In practice, however, near-so urce structure that contaminates S-wave travel times would diminish th is capability. The use of SS-S differential times for SS waves which b ottom in the transition zone and mantle waves sensitive to transition zone structure but insensitive to the 660 km boundary, should be the m ost effective means of overcoming the trade-off. To address question ( 3), we adopt the criterion that the combination of an unbiased volumet ric model with an accurate topographic model should provide a better f it to normal mode structure coefficients than the volumetric model alo ne. The boundary model Topo660a when added to the volumetric model S12 _WM13 fits the normal mode structure coefficients significantly better than the volumetric model alone. However, more recent models of 410 a nd 660 km boundary topography degrade the fit of the volumetric models S12_WM13 and SH.10c.17 to the normal mode structure coefficients, sug gesting that there is not yet a conclusive answer to this question.