Direct P-wave anisotropy measurements at HomestakeMine: implications for wave propagation in continental crust

We measured anisotropic seismic properties of schists of the Homestake Formation located at a depth of 1478 m in the Sanford Underground Research Facility (SURF) in the Black Hills of South Dakota, USA. We deployed a 24-element linear array of three-component geophones in an area in the Homestake Mine called 19-ledge. An airless jackhammer source was used to shoot two profiles: (1) a walkaway survey to appraise any distance dependence and (2) a fan shot profile to measure variations with azimuth. Slowness estimates from the fan shot profile show a statistically significant deviation with azimuth with the expected 180 degrees variation with azimuth. We measured P-wave particle motion deviations from data rotated to ray coordinates using three methods: (1) a conventional principal component method, (2) a novel grid search method that maximized longitudinal motion over a range of search angles and (3) the multiwavelet method. The multiwavelet results were computed in two frequency bands of 200-600 and 100-300 Hz. Results were binned by azimuth and averaged with a robust estimation method with error bars estimated by a bootstrap method. The particle motion results show large, statistically significant variations with azimuth with a 180 degrees cyclicity. We modelled the azimuthal variations in compressional wave speed and angular deviation from purely longitudinal particle motion of P-waves using an elastic tensor method to appraise the relative importance of crystalline fabric relative to fracturing parallel to foliation. The model used bulk averages of crystal fabric measured for an analogous schist sample from southeast Vermont rotated to the Homestake Formation foliation directions supplied by SURF from old mine records. We found with average crustal crack densities crack induced anisotropy had only a small effect on the observables. We found strong agreement in the traveltime data. The observed amplitudes of deviations of P particle motion showed significantly larger variation than the model predictions and a 20 degrees phase shift in azimuth. We attribute the inadequacies of the model fit to the particle motion data to inadequacies in the analogue rock and/or near receiver distortions from smaller scale heterogeneity. We discuss the surprising variability of signals recorded in this experimental data. We show clear examples of unexplained resonances and unexpected variations on a scale much smaller than a wavelength that has broad implications for wave propagation in real rocks.