Explosion-Earthquake Discrimination at Local Distances

Thursday, 18 December 2014
Colin T O'Rourke1, Glenn Eli Baker2, Anne F Sheehan1 and Steven H Harder3, (1)University of Colorado at Boulder, Boulder, CO, United States, (2)AFRL/RVBYE, Kirtland AFB, NM, United States, (3)University of Texas at El Paso, El Paso, TX, United States
Event discrimination research has largely focused on regional distances over the last couple of decades. Most regional discriminants have not been thoroughly tested at local distances, and a recent investigation of P/S ratios from the Source Physics Experiment (SPE) found that they were less effective locally. As discrimination thresholds are reduced, local methods will need to be validated and the physical bases for their performance will need to be understood. We are assessing new and existing methods of source discrimination at 10 to 200 km distance.

We use data from a temporary array deployed around the Bighorn Mountains in Wyoming and from nearby US Array stations. Small magnitude local earthquakes, coal mine explosions, and controlled borehole shots were recorded at over 200 broadband and short-period seismometers spaced at 5-35 km intervals across the tectonically complex area.

We assess the local discrimination performance of the P/S amplitude ratio, the presence of the fundamental Rayleigh wave Rg, and the signature of scattered Rg in the coda. P/S amplitudes are used because earthquakes typically have higher S-wave energy relative to P than shallow explosions, but the separation of populations is not always effective at local distances. Site effects have a significant impact that may need to be accounted for, along with propagation effects. Specifically, relative P and S site amplifications at basin and mountain stations vary in such a way that they limit separation of the event populations.

The Rg phase is only excited by near-surface events, so its presence indicates the likelihood of a man-made source since most earthquakes occur at too great a depth to excite Rg. We implement two methods for detecting Rg: one that uses the cross-correlation of the Hilbert transformed vertical and the radial seismograms, and one that scans a time-frequency representation for the signature of scattered Rg in the coda. These methods are effective in low signal-to-noise seismograms.