Reconciling discrepancies among estimates of small-scale mantle heterogeneity from PKP precursors

We stack amplitudes of over 10 000 high-frequency (similar to 1 Hz) PKP precursor waveforms, amassed from broad-band global seismic data with source-receiver distances between 120 degrees and 145 degrees recorded from 1990 to 2012. We forward model the stacked precursor envelope with an energy-conserving, multiple-scattering algorithm to find that an rms velocity perturbation of similar to 0.1 per cent fits the data reasonably well, in agreement with Margerin & Nolet. Similar results can be obtained using single-scattering (Born) theory, given the relatively weak scattering produced by our preferred model. The ramp-like increase in PKP precursor amplitudes with time is best fit with whole mantle scattering rather than models where scattering is restricted to the core-mantle boundary. Correctly modelling the relative amplitude of PKP precursor amplitudes compared to PKPdf requires taking into account the pulse broadening and coda of PKPdf, which can be done either empirically or by including a strongly scattering lithospheric layer in the multiple-scattering code. Several mantle scattering models proposed to explain other scattered seismic phases predict PKP precursor amplitudes much larger than those observed.