HR: 1340h
AN: S33B-1095    [Abstracts]
TI: Travel time estimation from a transfer function in frequency domain: the revised Sompi event analysis
AU: * Hasada, Y
EM: hasada@tono.jnc.go.jp
AF: Tono Geoscience Center, Japan Nuclear Cycle Development Institute, 1-63 Yamanouchi Akeyocho, Mizunami, 509-6132 Japan
AU: Kumazawa, M
EM: mkz@tono.jnc.go.jp
AF: Tono Geoscience Center, Japan Nuclear Cycle Development Institute, 1-63 Yamanouchi Akeyocho, Mizunami, 509-6132 Japan
AU: Tsuruga, K
EM: tsuru@tono.jnc.go.jp
AF: Tono Geoscience Center, Japan Nuclear Cycle Development Institute, 1-63 Yamanouchi Akeyocho, Mizunami, 509-6132 Japan
AU: Kunitomo, T
EM: kunitomo@tono.jnc.go.jp
AF: Tono Geoscience Center, Japan Nuclear Cycle Development Institute, 1-63 Yamanouchi Akeyocho, Mizunami, 509-6132 Japan
AB: We have developed a method to extract "events" localized in time domain from a transfer function in frequency domain, which is a basic analysis method of ACROSS (Accurately Controlled Routinely Operated Signal System) and referred to as the Sompi event analysis (Hasada et al., 2001, Earth Planets Space). In this method we assume that the complex frequency sequence to be analyzed is a transfer function between the source input and the receiver output sampled at discrete frequencies. Through a kind of autoregressive modeling in frequency domain, we obtain a set of "events" characterized by complex travel time and complex amplitude, where the former expresses travel time and attenuation, and the latter amplitude and phase angle. Though the validity of the Sompi event analysis is confirmed through various numerical tests especially for analysis of dispersive wave, there are some problems in practical application; (1) Identification of wave elements among the different data sets is difficult, (2) Error estimation is not easy. Here we propose a revised procedure by means of weighted least squares fitting, which is based on maximum likelihood estimation. We attempt to improve parameters by weighted least squares fitting, through linearization of model about small variation of parameters. By using this procedure, Maximum likelihood estimation of complex travel times is realized and the problems (1) and (2) are partly solved. On the other hand, there is serious instability in the calculation of least squares. To overcome this problem, some procedure of non-linear optimization may be available.
DE: 7260 Theory and modeling
DE: 7200 SEISMOLOGY
SC: Seismology [S]
MN: 2004 AGU Fall Meeting