HR: 1340h
AN: S33A-0302 [Abstracts]
TI: Seismic Event Identification Using Scanning Detection: A Comparison of Denoising and Classification
Methods
AU: * Rowe, C A
EM: char@lanl.gov
AF: Los Alamos National Laboratory, EES-11, M.S. D-408, Los Alamos, NM 87545
United States
AU: MacCarthy, J K
EM: jkmacc@lanl.gov
AF: Los Alamos National Laboratory, EES-11, M.S. D-408, Los Alamos, NM 87545
United States
AU: Giudicepietro, F
EM: giudicep@ov.ingv.it
AF: Osservatorio Vesuviano - INGV, Via Diocleziano, 328, Napoli, 80124
Italy
AB:
Automatic detection and classification methods are increasingly important in observatory operations, as the volume and rate
of incoming data exceed the capacity of human analysis staff to process the data in near-real-time. We explore the success of
scanning detection for similar event identification in a variety of seismic waveform catalogs. Several waveform
pre-processing methods are applied to previously recorded events which are scanned through triggered and continuous waveform
catalogs to determine the success and false alarm rate for detections of repeating signals. Pre-processing approaches include
adaptive, cross-coherency filtering, adaptive, auto-associative neural network filtering, discrete wavelet package
decomposition and linear predictive coding as well as suites of standard bandpass filters.
Classification / detection methods for the various pre-processed signals are applied to investigate the robustness of the
individual and combined approaches. The classifiers as applied to the processed waveforms include dendrogram-based clustering
and neural network classifiers.
We will present findings for the various combinations of methods as applied to tectonic earthquakes, mine blasts and volcanic
seismicity.
DE: 3255 Spectral analysis (3205, 3280)
DE: 3270 Time series analysis (1872, 4277, 4475)
DE: 3280 Wavelet transform (3255, 4455)
DE: 7280 Volcano seismology (8419)
SC: Seismology [S]
MN: Fall Meeting 2005