Automatic Detection and Recognition of the First Arrival Phase of Seismic Event Signals Contaminated by Noise. The Curious Case of the Missing Explosion

摘要

Both of the papers are based upon data collected at the Lajitas Seismic Station. In the first paper a historical perspective on the methods that have been used to automatically detect event signals and pick first arrival times is developed. Then the data set used to characterize and test the detection techniques is described and the features which discriminate seismic signals effectively from the background noise are characterized. Tow automatic detection methods are investigated: 1) The AR (8)-Spectral estimates of the signal and noise are used to develop the AR(p)-Spectral estimate for a synthetic waveform; and 2) Several non-parametric tests are employed in a time domain detector to discriminate event signals from background. The non-parametric detector chosen, RANK 2700, employs a modified rank sum test to locate the seismic event and pick its first arrival time. Errors in the automatic first arrival picks for 152 of the event traces in the data set are used to analyze the performance. The second paper presents a scenario for a clandestine, decoupled nuclear explosion in southeastern New Mexico based upon data recorded at Lajitas and other stations at regional distances from the hypothetical clandestine test. Keywords: Seismic data; Seismic detection; Nuclear explosion detection; Pattern recognition; Seismometers; Texas seismological stations; Seismic wave spectral analysis; Signal to noise ratio. (edc)