DECONVOLUTION OF REFRACTION SEISMOGRAMS FROM LARGE UNDERWATER EXPLOSIONS

Publisher Website
Google Scholar
Abstract
The P-wave spectra of seismograms from large underwater explosions are frequently dominated by reverberations. When this is so, a simple reverberation model similar to that of Backus (1959) gives a good approximation to the source spectrum. The basic wavelet determined by this method is not necessarily minimum‐delay. A few promising deconvolutions have been carried out, revealing a sequence of arrivals of comparable amplitudes separated by short time intervals. Synthetic seismograms which have been constructed from these spike sequences differ very little from the field records. However, the technique often yields output seismograms which are not easily interpreted. A study using synthetic seismograms suggests five reasons for this: 1) low signal‐to‐noise ratios, which result in too narrow a frequency band in which signal is predominant; 2) pulses arriving at the recorder having undergone phase changes during transmission, especially in combination with 3) very close spacing of the sequence of arrivals, causing overlap of the deconvolved output pulses; 4) the presence of arrivals which are not propagated along simple ray‐theory paths, and 5) poor estimation of the source parameters. Nonlinear effects and complex geology at the source are other possible causes of complications in the deconvolved records.

This publication has 0 references indexed in Scilit: