Secondary Phase Validation - Phase Classification By Polarization.

A long standing problem in observational seismology is that of confidentially picking secondary phases in earthquake recordings. In practice, the analyst efforts are often rendered useless because phase pickings are not coupled to a phase validation scheme. With the latter is meant that that a phase picking , say a pP-phase candidate should comply with preceding P-polarization and P-slowness requirements. Obviously, picking secondary P-phases in the P-wave coda is foremost a scientific problem; how many phase candidates may be picked and what screening procedure should be used ? The general problem formulation is that of many phase candidates from a network of stations and then introduce a maximum likelihood procedure for recognize them properly and jointly with focal hypocenter coordinate estimates. In this presentation we outline a Bayesean approach for screening secondary phases arriving in the P-coda between Pn/Pg and Sg/Lg using single station 3-comp. recordings. We assume that any P-wave onset is a plain sum of linerarly polarized signal wavelets and scattered non-polarized wavelets and noise and furthermore that the covariance matrix contain most of the information in the waveforms. We also presume that phases in question have smaller incidence angles than Pg, better polarized than random wavelets and use a truncated Gaussian p.d.f. for the azimuth angle derived from a preliminary epicenter location. We then use a Bayesean approach for estimating the probability of pP, sP and PmP presences in the P-coda. We have tested this phase picking & identification scheme on 3-comp. records of pure noise segments, P- coda from surface explosions and naturally from local earthquakes in Norway and Karelia (NW Russia). Seismographs are Cossack Ranger II operating in these areas. The outcomes of these experiments were highly satisfactorily; in ambient noise no phase triggering as expected. For explosion records the only secondary phase identified was PmP while for earthquake records we also found pP- and sP-phases. The mentioned phases were visible but without the polarization results at hand would hardly be interpreted as depth phases. We are planning to extend the scheme to phase candidates for a network of stations which is a far more extensive undertaking.