As a preprocessing step, I notch-filtered the data to remove monochromatic spikes (e.g. 60 Hz noise). Although I later low-pass filtered the data to 30 Hz, high amplitude spikes need to be removed before crosscorrelation/factorization to avoid numerical underflow. Figure shows 5 seconds of passive data after notch-filtering. Two types of event are clearly visible in Figure : slow linear events that appear to be traveling in the direction of decreasing channel number, and hyperbolic events with a higher apparent velocity. I speculate that these events were generated by the drill truck and power lines, respectively.
Following the processing steps applied to the helioseismic data, I crosscorrelated (or spectrally factorized) the data, stacked over the record axis, and then stacked over radius (distance from zero spatial lag). I then applied a (30 Hz) low-pass filter to the data to remove aliased ground-roll energy, and finally applied an offset dependent gain correction xpow=1 for display purposes.
Figure 7 Five seconds of passive seismic data from the Long Beach experiment after notch-filtering.