Observations of anomalous splitting and their interpretation in terms of aspherical structure

Abstract

The problem of inverting for the aspherical structure of the earth is complicated by the nonlinear dependence of low‐frequency seismic waveforms on aspherical structure. In an attempt to overcome this obstable, we report on the application of two complementary techniques. The first, a data space technique called singlet stripping, which linearly recombines seismic recordings to estimate singlet resonance functions, has been applied to 190 International Deployment of Accelerometers and Global Digital Seismographic Network recordings from five large events. More than 290 singlets from 34 low harmonic degree multiplets appear to have been resolved. A subset of these measurements has been compared with those produced from the second technique, a nonlinear regression, which iteratively estimates coefficients which are linear functionals of aspherical structure. Both techniques agree that most multiplets are normally split, with singlet frequency distributions insignificantly different from those predicted for a rotating, hydrostatic (RH) earth model. The main result of this paper is that both techniques also agree that approximately a third of the multiplets are anomalously split, some of which span frequency bands up to 2.5 times greater than predicted for an RH model. All of the anomalously split multiplets are SKS, PKP, or PKIKP equivalent. The observation of anomalously widely split multiplets is highly robust and provides compelling evidence for the existence of deep large‐scale, nonhydrostatic aspherical structure. The inverse problem for the axisymmetric part of aspherical structure has been performed in the hope of illuminating anomalous splitting. Unless a large amount of structure in the core is included, we are unable to construct a smooth axisymmetric model which accurately predicts the splitting characteristics for the anomalous multiplets while simultaneously fitting the normally split multiplets. The location and nature of this core heterogeneity are unclear, but we find that a simple outer core structure is sufficient to give a reasonable fit to the data. There are good theoretical reasons for believing that such nonhydrostatic outer core structure is geophysically unreasonable, yet differential travel time data sensitive to core structure apparently require similar large scale heterogeneity in the core. Although this dilemma remains unresolved, spectral fitting techniques like the nonlinear regression can be applied to many more multiplets than considered here and it is not unreasonable to predict that reliable large scale aspherical models of the deep earth soon will become available.