Using Network Theory to Understand Seismic Noise in Dense Arrays

Abstract:

Dense seismic arrays offer an opportunity to study anthropogenic seismic noise sources with unprecedented detail. Man-made sources typically have high frequency, low intensity, and propagate as surface waves. As a result attenuation restricts their measurable footprint to a small subset of sensors. Medium heterogeneities can further introduce wave front perturbations that limit processing based on travel time. We demonstrate a non-parametric technique that can reliably identify very local events within the array as a function of frequency and time without using travel-times. The approach estimates the non-zero support of the array covariance matrix and then uses network analysis tools to identify clusters of sensors that are sensing a common source. We verify the method on simulated data and then apply it to the Long Beach (CA) geophone array. The method exposes a helicopter traversing the array, oil production facilities with different characteristics, and the fact that noise sources near roads tend to be around 10-20 Hz.