HR: 09:25h
AN: S31C-06    [Abstracts]
TI: Time derivatives of the spectrum: Relaxing the stationarity assumption
AU: * Prieto, G A
EM: gprieto@ucsd.edu
AF: Scripps Institution of Oceanography, 9500 Gilman Drive, La Jolla, CA 92122 United States
AU: Thomson, D J
EM: djt@mast.queensu.edu
AF: Math and Statistics Department, Queen's University, Kingston, ON K7L3N6 Canada
AU: Vernon, F L
EM: flv@ucsd.edu
AF: Scripps Institution of Oceanography, 9500 Gilman Drive, La Jolla, CA 92122 United States
AB: Spectrum analysis of seismic waveforms has played a significant role towards the understanding of multiple aspects of Earth structure and earthquake source physics. In recent years the multitaper spectrum estimation approach (Thomson, 1982) has been applied to geophysical problems providing not only reliable estimates of the spectrum, but also estimates of spectral uncertainties (Thomson and Chave, 1991). However, these improved spectral estimates were developed under the assumption of local stationarity and provide an incomplete description of the observed process. It is obvious that due to the intrinsic attenuation of the Earth, the amplitudes, and thus the frequency contents are changing with time as waves pass through a seismic station. There have been incredible improvements in different techniques to analyze non-stationary signals, including wavelet decomposition, Wigner-Ville spectrum and the dual-frequency spectrum. We apply one of the recently developed techniques, the Quadratic Inverse Theory (Thomson, 1990, 1994), combined with the multitaper technique to look at the time derivatives of the spectrum. If the spectrum is reasonably white in a certain bandwidth, using QI theory, we can estimate the derivatives of the spectrum at each frequency. We test synthetic signals to corroborate the approach and apply it the records of small earthquakes at local distances. This is a first approach to try and combine the classical spectrum analysis without the assumption of stationarity that is generally taken.
DE: 0910 Data processing
DE: 3255 Spectral analysis (3205, 3280)
DE: 3270 Time series analysis (1872, 4277, 4475)
DE: 5200 PLANETARY SCIENCES: ASTROBIOLOGY
DE: 7290 Computational seismology
SC: Seismology [S]
MN: Fall Meeting 2005