The effects of the theoretical formalism and data selection on mantle models derived from waveform tomography

Abstract

Summary Most global tomographic models to date are derived using a combination of surface wave (or normal-mode) data and body wave traveltime data. The traveltime approach limits the number of phases available for inversion by requiring them to be isolated on the seismogram. This may ultimately result in limiting the resolution of 3-D structure, at least in some depth ranges in the mantle. In a previous study, we successfully derived a degree 12 whole-mantle SH-velocity tomographic model (SAW12D) using exclusively waveform data. In that inversion, a normal-mode formalism suitable for body waveforms, the non-linear asymptotic coupling theory (NACT), was combined with a body wave windowing scheme, referred to as the ‘individual wavepacket’ (IW) technique, which allows one to assign individual weights to different body wave energy packets. We here compare the relative merits of this choice of theoretical formalism and windowing scheme at different depth ranges in the mantle. Choosing as the reference a model obtained using 7500 transverse-component body wave and 8000 surface wave seismograms and the NACT and IW approaches, we discuss the relative performance of the path average approximation (PAVA), a zeroth-order theoretical approximation appropriate for single-mode surface waves, relative to NACT, and compare the IW windowing scheme with a more standard ‘full window’ (FW) approach, in which a single time window is considered from the first body wave arrival to the fundamental-mode surface waves. The combination PAVA/FW is often used in global tomography to supplement the traveltime data. We show that although the quality of the image derived under the PAVA/FW formalism is very similar to that derived under NACT/IW in the first 300 km of the upper mantle, where the resolution is dominated by surface waves, it deteriorates at greater depths. Images of the lower mantle are shown to be strongly sensitive to the theoretical formalism. In contrast, the resolution of structure near the core–mantle boundary depends mostly on the windowing scheme. This is because this resolution is controlled by low-amplitude phases such as Sdiff, which are downweighted in the FW scheme. Whilst the image obtained in D″ using the combination NACT/IW is in good agreement with images obtained by other authors using both waveforms and traveltimes, we show that, when using FW, uppermost mantle structure can be mapped into D″. This result is confirmed by synthetic tests performed on a composite of the upper-mantle geodynamic model 3SMAC. We also show, based on synthetic tests, that for structures in the upper mantle with sharp boundaries, differences are observed between NACT and PAVA. Whilst a combination of traveltimes and surface wave data is adequate for resolving relatively smooth features in the mantle, our results show that by potentially increasing the achievable sampling, the waveform approach shows great promise for future high-resolution tomographic modelling of mantle structure, if cast in an appropriate theoretical framework.