When studying the particle energy spectra, particles with different energies are counted within the respective energy intervals. Also, in this case counting is performed during long time intervals so that large enough count numbers are obtained. A considerable fraction of temporal variations and thus the important information contained in them will be lost. For example, the variance pattern contains information about subpopulations present in the data. In order to preserve the information about the covariance of different spectral channels, high counting rates must be guaranteed by detector characteristics and not by long measuring intervals. In the present report a linear decomposition technique using the principal components is used to separate the subpopulations of particles present in the data. The described method is applied to particle energy spectra measured onboard the Swedish satellite Viking (Hultqvist 1988).
IRF Scientific Report 237
January 1997.