HR: 0800h
AN: SM51A-0272    [Abstracts]
TI: Mutual Information based clustering of Pi 2 magneic pulsations observed at CPMN stations
AU: * Tokunaga, T
EM: tokunaga@geo.kyushu-u.ac.jp
AF: Department of Earth and Planetary Sciences, Kyushu University, Japan, 6-10-1 Hakozaki, Higashi-ku, Fukuoka, 812-8581, Japan
AU: Yoshikawa, A
EM: yoshi@geo.kyushu-u.ac.jp
AF: Department of Earth and Planetary Sciences, Kyushu University, Japan, 6-10-1 Hakozaki, Higashi-ku, Fukuoka, 812-8581, Japan
AU: Uozumi, T
EM: uozumi@serc.kyushu-u.ac.jp
AF: Space Environment Research Center, Kyushu University, Fukuoka, Japan, 6-10-1 Hakozaki, Higashi-ku, Fukuoka, 812-8581, Japan
AU: Yumoto, K
EM: yumoto@serc.kyushu-u.ac.jp
AF: Space Environment Research Center, Kyushu University, Fukuoka, Japan, 6-10-1 Hakozaki, Higashi-ku, Fukuoka, 812-8581, Japan
AU: Group, C
AB: Pi 2 magnetic pulsations are observed on the ground as mixed signals of several independent components that are reflecting (1) propagations of fast and shear Alfven wave, (2) resonances of plasmaspheric/magnetospheric cavity and magnetic field lines, and (3) transformations to ionospheric current systems [e.g., Yumoto et al., 2001]. However, it has been unclear how they coupled with each other and how their signals are distributed at different latitudes. We have attempted to separate mathematically ground-observed Pi 2 pulsations by applying Independent Component Analysis (ICA). ICA is one of the multivariate statistical techniques that started to be used in the 1990s in the field of signal processing [e.g., Common, 1994]. With ICA, source signals are assumed to be non- Gaussian and statistically independent of each other and estimated by maximizing their statistical independence. It has been successful in resolving observed mixed signals including brain imaging data and voice signals into source signals. As an initial stage of this study, we applied FastICA algorithm suggested by Hyvarinen and Oja [1997] to an isolated Pi 2 event on a quiet day observed at CPMN (Circum-pan Pacific Magnetometer Network) stations and successfully decomposed them into two components. One was the global oscillation that occurs from nightside high to equatorial latitudes with the common waveform and has an amplitude maximum at nightside high latitude. Another component was localized at nightside high latitudes. Its amplitudes were quite weak at low latitudes, but were enhanced near dayside dip equator [Tokunaga et al., 2007, GRL]. As a second stage of this study, we have attempted to classify ground-observed Pi 2 pulsations into some groups systematically. In this paper, MILCA (mutual information based least-dependent component analysis) suggested by Stogbauer et al., [2004] have been introduced, which are based on crude approximations for MI (mutual information). The numerical values of the MI can be used for (i) estimating residual dependencies between the output components; (ii) estimating the reliability of the output by comparing the pairwise MIs with those of remixed components; and (iii) clustering the output according to the residual interdependencies.
DE: 2752 MHD waves and instabilities (2149, 6050, 7836)
DE: 2790 Substorms
DE: 3270 Time series analysis (1872, 4277, 4475)
DE: 3285 Wave propagation (0689, 2487, 4275, 4455, 6934)
SC: SPA-Magnetospheric Physics [SM]
MN: 2007 Fall Meeting