Ambient noise tomography across the Cascadia subduction zone using dense linear seismic arrays and double beamforming

In the summer of 2017, we deployed 174 three-component nodal geophones along a 130 km west-east line across the central Oregon forearc lasting about 40 d. Our goal was to evaluate the possibility of imaging the lithospheric structure in detail with a dense but short-duration sampling of passive seismic signals. In this study, we used passive recordings from the nodal array and the previous CASC93 broad-band array along the same line to calculate noise cross-correlations. Fundamental Rayleigh wave signals were observed in the cross-correlations between 3 and 15 s period. To enhance the signal and simultaneously measure the phase velocity, we employed a double beamforming method. At each period and location, a source beam and a receiver beam were selected and the cross-correlations between the two were shifted and stacked based on the presumed local velocities. A 2-D grid search was then used to find the best velocities at the source and receiver location. Multiple velocity measurements were obtained at each location by using different source and receiver pairs, and the final velocity and uncertainty at each location were determined using the mean and the standard deviation of the mean. All available phase velocities across the profile were then used to invert for a 2-D shear wave crustal velocity model. Well resolved shallow slow velocity anomalies are observed corresponding to the sediments within the Willamette Valley, and fast velocity anomalies are observed in the mid-to-lower crust likely associated with the Siletzia terrane. We demonstrate that the ambient noise double beamforming method is an effective tool to image detailed lithospheric structures across a dense and large-scale (>100 km) temporary seismic array.