How to make robust splitting measurements for single-station analysis and three-dimensional imaging of seismic anisotropy

Abstract

SUMMARY We quantify errors on splitting intensity measurements produced by incoherent noise on horizontal components through a detailed analysis of synthetic seismograms and of SKS and SKKS waves recorded by four permanent broad-band stations. We find that these errors are quite significant, even on high-quality records, because the measurement procedure involves the comparison of radial and transverse components, which are both contaminated by noise. To decrease the level of noise in the data, it is thus recommended to average splitting intensities from waves coming from the same backazimuth. Alternatively, one can analyse stacked radial and transverse components after applying a Wiener filter, which standardizes the waveforms. This is found to be equivalent because the measure of splitting intensity is a linear process. The utilization of Wiener filters reduces variations of apparent splitting resulting from differences in frequency content of the waveforms. It is possible to determine splitting parameters (splitting delays and fast directions) from azimuthal variations of splitting intensity at a particular station. Tests on synthetic seismograms demonstrate that this approach gives unbiased and robust estimates of splitting parameters, in contrast to the Silver and Chan method, which only works on the rare records that have very large signals on their transverse components. The Wolfe and Silver method does not suffer from this limitation and gives the same results as the multichannel splitting intensity method. Analysis of splitting intensities at stations BDFB, BGCA, CAN and ERM gives fast directions that are generally in excellent agreement with those determined with the Wolfe and Silver method. However, at some stations, azimuthal variations of splitting intensities reveal features that are not well captured by the sinusoidal variation representing the average anisotropy beneath the station. This suggests that smallscale lateral variations of anisotropy are present in the subcontinental mantle, which remain to be imaged in detail.