Phase identification analysis using directionality and rectilinearity functions in three component seismograms

Abstract

An efficient technique for phase identification analysis is introduced using newly defined directionality and conventional rectilinearity functions obtained from three component seismograms. The directionality function,D, is defined asD(t)=1-|eb(t s )·e1(t)| where eb(t s ) and e1(t) are vectors obtained from two different short time windows at timest=t s andt=t i , respectively of a three-component seismogram set. The vector eb(t s ) is the eigenvector associated with the largest eigenvalue of the covariance matrix whose elements are values of covariance between component time series with a short time duration covering a basis phase at timet=t s . The vector e1(t) is the eigenvector associated with the largest eigenvalue of the covariance matrix computed for another phase at timet=t i . Computation of the directionality is done for a fixed basis phase at timet=t s and other phases to be compared at variable timest=t si . The first arrival phase is generally used as the basis phase, because this phase could be the most uncontaminated among all the phases in seismic records. The rectilinearity function obtained from the ratio of the largest and the second-largest eigenvalues of a covariance matrix in three component seismograms is also used in this study to increase confidence and accuracy of phase identification. The directionality and rectilinearity functions are computed along the time points of three component seismograms, and resulting time series are compared with original seismograms by inspection. These functions are not used as filters to be applied to seismograms by multiplications as used in previous researches, in order to avoid identification ambiguity caused by mixing of the characteristics of the functions and amplitudes of seismograms. This approach has improved performances in accuracy, stability, and computing time compared with the previously existing ones and is especially suitable for identification of phases in weak or moderate seismicity regions where routine identifications with large amount of data is not necessary. The technique was applied to a set of local, regional and teleseismic data, and body wave and surface wave phases were effectively identified.