Maximum-Likelihood Multichannel Deconvolution of P Waves at Seismic Arrays

Abstract : The results of maximum-likelihood multichannel deconvolution applied to array recordings and three component station networks for teleseismic P waves, are presented and interpreted in terms of possible surface reflections and other arrivals from explosions conducted at many of the world's major nuclear test sites. The deconvolution method utilizes the well known fact that P wave spectra can be decomposed into source and receiver spectral factors. The source functions obtained in the deconvolution process provide a better picture of the nature of explosion source time functions, and in particular of the presence or lack of secondary arrivals following the P wave such as pP or spall. The presence of such secondary arrivals and their effects on the first cycle of the P wave are very important in yield estimation. Significant variations in the deconvolved source time function between test sites may be associated with topography and testing practices. All of the source functions show complexity in addition to the initial P arrival and pP arrival, if present. There is also a great deal of variation between different source time functions for events at the same test site. Often, but by no means always, events occurring near each other look particularly similar. The site functions are also complex in most cases and azimuthal variations are significant in both source and receiver regions. Site and source effects contribute about equally to the energy observed in the P codes of the events analyzed. Deconvolved source time functions should be especially useful for improving estimates in the m sub b bias between test sites and to improve yield estimates since site as well at t*, instrument, and any known source spectra are removed.