Effects of Source Scaling on Local-Distance P/S Amplitude Ratios for Seismic-Source Discrimination

Methods to discriminate between earthquakes and explosions using observations from teleseismic and regional distances have been widely investigated and employed. These earlier studies explored discriminant performance using observations from high yield (generally greater than 1Kt) explosions that were recorded at regional and teleseismic distances. Many discriminants rely on the amplitude ratio of P and S phases, which vary systematically with source type as a result of the strong difference in shear-wave energy radiated by earthquakes and explosions. Recent studies have used a similar strategy to investigate P/S discriminant performance on smaller, low yield events observed at local distances (<200km). Shorter distances and higher frequency observations require more scrutiny of path effects, site responses, and source corrections. We use local distance observations (<35km) from the Klerksdorp gold mining region of South Africa to investigate P/S discriminant performance on low-magnitude mining-related events. The dataset includes observations from 226 mining-related events (pillar collapses, rock bursts, etc.), the August 5th, 2014 M 5.5 Orkney earthquake, and 21 suspected aftershocks. Events extend to depths of 10km but do not include known explosions, so our focus is on the comparison of P/S ratios for earthquakes and mining-related events. Ground motions were recorded at roughly 25 surface accelerometers operating from 2013 to 2014. Preliminary network-averaged P/S amplitude ratios show little difference between the mine-related events and suspected earthquakes in the frequency band from 1 to 25 Hz. Most of the events are small and the amplitude observations correspond to frequencies below the expected source corner frequencies. We discuss the results of our investigation of spectral fitting (model-guided) and linear regression (data-guided) to reduce the scatter in the observed amplitude ratios in an effort to characterize the sources in the available event population.