← Back to Sessions

An Efficient Subspace Detector for Rayleigh Waves, Demonstrated Against Explosions

Description: 

We develop and validate a Rayleigh waveform detector that we derive from a hypothesis test between two competing, approximate models for elliptically polarized, three channel seismic data. Our algorithm processes these data with sliding windows to output continuous estimates of source back azimuths and ~45 data products that it fuses from multiple frequency bands, at computational speeds comparable to energy detector algorithms. The sliding windows adapt to consume a fixed number of waveform cycles per frequency band and automatically adjust trigger thresholds to maintain fixed false alarm rates against noise. We summarize five tests that demonstrate these capabilities against both synthetic Airy phase seismograms and real, explosion-triggered waveforms sourced in Ukraine. These experiments demonstrate reliable detections against explosions and accurate estimates of source back-azimuths that we cross-validate with the Stockwell transform and an event catalog. Performance curves that measure the true positive rate of the detector against real sources indicate that our algorithm provides very reliable detection rates for cataloged explosions, beyond 50 km. Median errors and uncertainties in our back-azimuthal angles indicate a good localization capability. We provide physical, statistical, and computation justifications for our algorithms.

Session: Advancements in Forensic Seismology and Explosion Monitoring - II

Type: Oral

Date: 4/17/2025

Presentation Time: 10:45 AM (local time)

Presenting Author: Joshua

Student Presenter: No

Invited Presentation: 

Poster Number:

Authors