Focal spot imaging on USArray records

The spatial zero-lag amplitude distribution of correlations obtained from vertical component dense array records of diffuse seismic wave fields is characterized by a large-amplitude feature around the origin referred to as focal spot. In the context of time-reversed surface waves it can be understood as the collapse of a converging wavefront. The analogy to the SPAC method implies that the nine-component solutions that describe the spectral features can readily be applied to the time-domain focal spot shape to estimate local phase velocity, which connects this method to established elastographic medical imaging. In contrast to sparse SPAC arrays, modern dense arrays allow a properly resolved focal spot at near-field distances for an inversion-free sensor-by-sensor image compilation, with intriguing implications for vertical and lateral resolution enhancement. We demonstrate the applicability of this method on the basis of Rayleigh wave focal spots in the 60 s to 200 s period range that are obtained from ambient field correlations using USArray data between -125 and -90 degrees west. The 1000 s long noise correlations are computed using standard techniques, Gaussian filtered around the central target frequency, and the spatial zero-lag distribution fitted with the SPAC Bessel functions model to distances of 1.2 wavelengths. The effectiveness and accuracy of this approach is demonstrated by the impressive similarity between the obtained “instantaneous image” at 60 s and surface wave tomography results from the literature. The stark velocity contrast between the western and central U.S. is clearly resolved, but the similarity extends to well resolved details including the Sierra Nevada, the Snake River Plain feature, the circular low-velocity rim around the Colorado Plateau, and part of the Mississippi Embayment. Based on this benchmark result obtained with vertical-component data we explore the internal consistency of the obtained maps towards longer periods and the associated extension of dispersion measurements; we probe the limits of the near-field approach by systematically lowering the fitting distance to sub-wavelength scales; and we quantify the similarity of vertical-radial results. The zero-lag amplitude distributions in the wavenumber domain show signatures of near-vertically incident energy associated with global body wave reverberations. We mute this energy by neglecting time windows from the correlation data after global large earthquakes. Systematic tests of the window length, and again the comparison to the benchmark observations, inform about the efficiency of this approach. We conclude that time-domain dense array near-field imaging yields accurate distributions of the velocity structure. We emphasize the disadvantageous low randomization of the long period ambient field, and the sub-array shapes resulting from the rolling USArray deployment. Imaging on smaller scales should therefore work better.