Constraining the crustal velocity structure of the conterminous United States using receiver functions and the autocorrelation of earthquake-generated body-waves

To determine the discontinuity structure Earth's crust, passive seismic methods have primarily focused on differences in vertically and radially-polarized energy in the coda of earthquake-generated body waves (e.g., receiver functions). This approach effectively separates primary body waves from their conversions to other phases at discontinuities (P-to-S-wave or vice-versa). In order to transfer the timing of these body-wave conversions to the depth domain, receiver functions fundamentally rely 3 parameters: layer thickness, P-wave velocity, and S-wave velocity. The signals obtained through converted waves on receiver functions, however, can only provide 2 of these parameters. This requires making either an estimate or a priori assumption about one of the parameters (commonly Vp) in the region of interest, which can result in biased estimates of depths to boundaries and average velocities. As these models are commonly used as a starting point for more complex studies, the often unquantified uncertainty related to these assumptions can propagate much further than the initial study.

In this study, we focus not only on differences in body-waves and their coda, but also on their similarities by using an autocorrelation technique. The autocorrelation of the different components of motion over a layered media have been theoretically proven to recover the layered structure beneath a seismic station. We focus on constraining the timing of Moho-reflected P-waves from teleseismic earthquakes, which are removed through deconvolution in the receiver function technique. We enhance the coherence of the rather weak Pmp signal through phase-weight stacking and use its travel time to constrain the 3^rd parameter (Vp). This allows us to create a system of equations that we can solve for the thickness, average P- and average S-wave velocity of a layer. We can also quantify uncertainties in our results using either Bayesian or grid-search approaches. In this study, we apply waveform autocorrelation to teleseismic earthquakes recorded on >100 seismic stations that are spatially distributed throughout the United States to obtain a model of crustal variability that is unbiased by a priori assumptions of P-wave velocity.