An adaptive seismic deconvolution algorithm based on and constrained by a
physical model of attenuation is proposed as an alternative to more
conventional, time-varying deconvolution methods. Q-adaptive deconvolution
(QAD) requires an iterative application of conventional prediction error
filtering and inverse Q-filtering, with the former process aimed at source,
receiver, and near-surface reverberations and the latter compensating for
attenuation effects. With special consideration given to the possible over-
amplification of high-fequency noise, a "clipped" inverse Q-filter is described
as particularly appropriate for QAD.
We compare QAD with conventional deconvolutions through application to field
recorded seismograms. QAD more effectively compensates for the attenuation
of high-frequenies and dispersion effects while yielding estimates of the
quality factor Q and avoiding the windowing or weighting of seismograms
required by conventional methods.
STANFORD