Statistical deconvolutions such as predictive deconvolution, Minimum Entropy Deconvolution, etc. perform beest when applied to stationary data. The typical seismic trace, however, is highly non-stationary; and one persistent source of non-stationarity is attenuation. Toward the objective of enhancing stationarity, a mathematical model of attenuation is described which is consistent with currently popular models and, most importantly, leads to a computationally efficient method for "backing out" attenuation.
STANFORD