Earthquake extraction and correlation energy at Long Beach, California seismic survey

Method

Prior to performing any cross-correlations, we searched for time windows that were clear of significant seismic events. This is because one of the primary conditions for seismic interferometry to be effective is that the ambient seismic field satisfies the principle of equipartition. Because earthquakes produce high amounts of directed energy, they would potentially compromise our correlation results.

Once clear of major seismic events, we formed our time windows. We chose three windows spanning from $1\colon30$  pm, January $21$ to $2\colon00$  am, January $22$ ($12.5$  hours); $5\colon00$  pm, February $20$ to $5\colon00$  am, February $21$ ($12$  hours); and $2\colon00$  pm, February $27$ to $4\colon00$  am, February $28$ ($14$  hours). We received each data file as a time series for a single station, which meant that we had to re-sort our data according to recording time. We chose to synthesize $32.5$ -minute patches with $2.5$ -minute tapers at both ends, leading to a $2.5$ -minute overlap from one window in time to the next.

Prior to cross-correlating, we bandpass all the traces. de Ridder and Dellinger (2011) have had success generating virtual low-frequency ($0.35$ -$1.75$  Hz), omnidirectional Scholte waves along the ocean floor, so we examine signals that have been similarly bandpassed between $0.175$  Hz and $1.75$  Hz. Because we want to compare correlation results over time, for each time window we cross-correlate all stations with the same station location. Finally, we stack each time patch within each time window to improve the signal-to-noise ratio of the correlations.

Earthquake extraction and correlation energy at Long Beach, California seismic survey