The objectives of this study are to model the observed seismic spectra from large industrial explosions using information obtained from blaster's logs and to compare the explosion spectra with those of small earthquake signals from the same source region. The data set consists of digital waveforms from four mining explosions (200,000 + lb. of explosives each) and two earthquakes (M = 3.5 and 4.0) in eastern Kentucky. The data were recorded on a short-period regional network at distances ranging from 180 to 400 km and have good signal-to-noise ratios at frequencies from 0.5 to 10 Hz.
The explosion amplitude spectra differ markedly from those of the earthquakes, by exhibiting strong time-independent amplitude modulations. This spectral modulation is directly attributable to the explosive charge geometry and firing sequence and is largely independent of source-station path and recording site.
Modeling of the explosion source spectra shows that the major contributor to the modulated character of the spectra are amplitude minima at frequencies related to the total duration of the explosion sequence. Another important effect is amplitude reinforcement at low frequencies (e.g., 5 Hz) due to the comparatively long delay (0.2 sec) between the firing of individual rows of explosives. These features dominate both Pg and Lg amplitude spectra at frequencies less than 7 Hz. Accurate modeling of the observed spectra at frequencies greater than a few Hertz requires that the azimuth of the recording site be taken into account. Also, the spectra at higher frequencies become sensitive to random variations in the firing times of any of the various subexplosions.