A Study of the Adaptive Method for Decoupling Overlapping Seismic Records

Signal extraction from overlapping seismic records is a common problem in geophysical data analysis. Identification and separation of multiple seismic arrivals, analysis of large earthquakes as multiple point sources, and calculation of the true yield of a large nuclear explosion from interfering small explosion, all hinge on our ability to effectively decouple two interfering wave signals. This paper presents a method for signal separation based on an adaptive filtering technique. We apply a semi-deconvolution algorithm to overlapping explosion records and S/SKS phase groups, and then perform noise reduction and signal decoupling under different a priori conditions and assess the stabilities using a variance reduction approach. We demonstrate, through numerical experiments and analysis of seismic station records, that the adaptive method can be both robust and practical for regional and teleseismic applications.