← Back to Sessions

Differential Seismic Phase Detection Probability as a Potential Attribute for Discrimination of Explosions and Earthquakes

Description: 

Deep learning models trained to estimate the probability of seismic P and S phases are rapidly expanding the scale of event detections. Here, we evaluate the potential for phase detection probabilities to contribute to event type classification, particularly discrimination of single-fired borehole explosions and earthquakes at local distances of <300 km. Motivated by the empirical success of P/S amplitude ratios for explosion discrimination, we consider the difference between P and S pick probability, P_prob - S_prob, as a discriminant. Test data include M_L~1-4 earthquakes and explosions observed by common seismographs in ten geologically diverse localities. PhaseNet (Zhu and Beroza, 2019) trained with STEAD (Mousavi et al., 2019) shows higher P_prob - S_prob for explosions than for earthquakes. Binary classification with PhaseNet P_prob - S_prob with at least 3 stations yields a Receiver Operating Characteristic area under the curve of 0.85, compared to 0.88 for P/S amplitude ratios. Although the performance is slightly lower, P_prob - S_prob is efficient as an automated byproduct of event detection and use of phase probability avoids the binary choice of picking or not picking weakly visible S waves that are common to explosions. The results generally suggest that phase pick probabilities may be useful event classification attributes.

Session: Advancements in Forensic Seismology and Explosion Monitoring - II

Type: Oral

Date: 4/17/2025

Presentation Time: 11:00 AM (local time)

Presenting Author: Brandon

Student Presenter: No

Invited Presentation: 

Poster Number:

Authors