Surface wave phase-velocity tomography based on multichannel cross-correlation

We have developed a new method to retrieve seismic surface wave phase velocity using dense seismic arrays. The method measures phase variations between nearby stations based on waveform cross-correlation. The coherence in waveforms between adjacent stations results in highly precise relative phase estimates. Frequency-dependent phase variations are then inverted for spatial variations in apparent phase velocity via the Eikonal equation. Frequency-dependent surface wave amplitudes measured on individual stations are used to correct the apparent phase velocity to account for multipathing via the Helmholtz equation. By using coherence and other data selection criteria, we construct an automated system that retrieves structural phase-velocity maps directly from raw seismic waveforms for individual earthquakes without human intervention. The system is applied to broad-band seismic data from over 800 events recorded on EarthScope's USArray from 2006 to 2014, systematically building up Rayleigh-wave phase-velocity maps between the periods of 20 and 100 s for the entire continental United States. At the highest frequencies, the resulting maps are highly correlated with phase-velocity maps derived from ambient noise tomography. At all frequencies, we observe a significant contrast in Rayleigh-wave phase velocity between the tectonically active western US and the stable eastern US, with the phase velocity variations in the western US being 1-2 times greater. The Love wave phase-velocity maps are also calculated. We find that overtone contamination may produce systemic bias for the Love-wave phase-velocity measurements.