Three-station interferometry and tomography: coda versus direct waves

Traditional two-station ambient noise interferometry estimates the Green's function between a pair of synchronously deployed seismic stations. Three-station interferometry considers records observed three stations at a time, where two of the stations are considered receiver-stations and the third is a source station. Cross-correlations between records at the source-station with each of the receiver-stations are correlated or convolved again to estimate the Green's function between the receiver-stations, which may be deployed asynchronously. We use data from the EarthScope USArray in the western United States to compare Rayleigh wave dispersion obtained from two-station and three-station interferometry. Three three-station interferometric methods are distinguished by the data segment utilized (coda-wave or direct-wave) and whether the source-stations are constrained to lie in stationary phase zones approximately inline with the receiver-stations. The primary finding is that the three-station direct wave methods perform considerably better than the three-station coda-wave method and two-station ambient noise interferometry for obtaining surface wave dispersion measurements in terms of signal-to-noise ratio, bandwidth, and the number of measurements obtained, but possess small biases relative to two-station interferometry. We present a ray-theoretic correction method that largely removes the bias below 40 s period and reduces it at longer periods. Three-station direct-wave interferometry provides substantial value for imaging the crust and uppermost mantle, and its ability to bridge asynchronously deployed stations may impact the design of seismic networks in the future.