An analysis of large‐scale variations in small‐scale mantle heterogeneity using Global Seismographic Network recordings of precursors to PKP

Abstract

High‐frequency precursors to the core phase PKP are caused by scattering off heterogeneities in the lowermost mantle and D″ regions, and they provide a unique window into the small‐scale structure of the deep Earth. We study lower mantle scattering by analyzing 412 high‐quality PKP precursor records at ranges between 120° and 137.5° as obtained from the global seismic networks during the last 10 years. To examine regional variations in scattering strength, we compare individual records with the globally averaged PKP precursor stack of Hedlin et al.. [1997]. We identify strong differences in apparent scattering strength among specific source‐receiver paths. Inversion of these data for scattering source regions is complicated by ambiguity between source‐ and receiver‐side scattering and the sparse and uneven data coverage. Synthetic tests, however, suggest that inversions with applied smoothness constraints can resolve large‐scale differences in scattering strength over significant parts of the lower mantle. We use a conjugate gradient method based on an approximation to Rayleigh‐Born scattering theory to image differences in the average strength of scattering within the lowermost 1000 km of the mantle. Our results indicate particularly strong scattering beneath central Africa, parts of North America, and just north of India, whereas weaker scattering is seen beneath South and Central America, eastern Europe, and Indonesia. Some regions of strong scattering correlate roughly with large‐scale anomalies revealed by seismic tomography including the African plume and the Tethys trench. These correlations are tentative rather than definitive because bootstrap resampling tests show that many details in our model are not reliably resolved and the network data alone do not permit complete resolution of the source‐receiver ambiguity in all areas. Further progress in this area will require integration of available network recordings with data collected by regional networks and arrays and consideration of the phase velocity of the precursors as well as their temporal variations.