Inference of core-mantle boundary topography from ISCPcPandPKPtraveltimes

Abstract

We invert ISC PcP and PKP absolute and differential traveltimes in an attempt to infer the long-wavelength topography of the core-mantle boundary (CMB). The data selection and processing methods are described and evaluated. These travel-time data are very noisy and the geographic distribution of the data is highly non-uniform, inhibiting reliable inference of CMB topography. Spatial averaging enhances the coherent component of the residual variance (related to heterogeneity), however, the random component of the variance is much larger than the coherent component. We show that for PcP data the coherent signal due to mantle heterogeneity overshadows that arising from the CMB, and that the effects of mantle heterogeneity are mapped into our inferred CMB solutions. The PcP data are not correlated across the spatial averaging bins and seem to have a strong bias due to small-scale structure and/or noise. The non-uniform geographic sampling of the data plays a role in the mapping of mantle heterogeneity onto the CMB. Spatial patterns of CMB models inferred from different phases do not agree. Amplitudes of seismically inferred CMB undulations vary greatly. The sensitivity of inferred CMB models to the processing, spatial averaging procedure, and inversion techniques are investigated. Topographic amplitudes increase strongly with increasing input residual variance. The power spectrum of inferred topography indicates that there are unmodelled heterogeneities that must be described with spherical harmonics of degree 6 and higher. Based on this work, we conclude that reliable inference of long-wavelength CMB topography is not likely with the current ISC data set or with a spherical harmonic expansion truncated to degree and order 6.