Using Delay-Fired Quarry Blasts to Image the Crust: A Comparison of Methods for Deconvolving Mixed-Delay Source Wavelets.

Quarry blasts can be effective, low-cost energy sources for seismic imaging, provided that one can deconvolve the extended source signatures produced by tipple firing. This study uses real and synthetic quarry-blast data to compare the performance of several methods for deconvolving mixed-delay signals. The problem considered here is the recovery of the effective source function along a given azimuth from traces in a shot gather. The tests focus particularly on design criteria for minimum-entropy filters and optimization of source-wavelet estimates derived by inversion of minimum-entropy filter coefficients. Predictive deconvolution reduces tinging but can generate coherent artifacts when the source wavelet is not minimum delay. Wiener filtering using source-wavelet estimates derived by least-squares inversion of minimum-entropy filter coefficients preserves relative amplitudes, allows the user to specify the degree of spiking, and avoids delays in the output. Of 47 blasts recorded in Georgia and Tennessee with 15- to 19-channel arrays, 37 yielded a localized wavelet estimate in which the duration of the most energetic portion of the wavelet estimate trace was close to the reported duration of the blast. In general, extraction of the source wavelet directly from quarry-site recordings is complicated by nonlinear effects, interference from S and Rayleigh waves, and the variation of the source wavelet with azimuth. In spite of those complications, waveforms observed at quarry sites were similar to wavelets derived from field traces for about 10 of the blasts studied. Wavelet estimates derived from minimum-entropy filter coefficients are not affected by static shifts between traces. Where statics have been removed and where recording arrays are long enough to resolve differences in ray parameter for overlapping events, localized slant stacks can be a useful alternative for estimating the source wavelet. Methods, such as Wiener filtering, that... [ABSTRACT FROM AUTHOR]