Stratigraphic filtering and Q estimation

Gary F. Margrave

The long-standing prediction that a seismic wave propagating in a finely layered earth model displays an apparent attenuation is investigated. Called stratigraphic filtering, this effect looks much like constant-Q attenuation and adds to intrinsic attenuation to produce effective attenuation. Using a 1D synthetic seismogram algorithm, this paper calculates the effective attenuation in a sequence of finely layered models derived from well logs. The models all have a finely-layered Q structure, representing intrinsic attenuation, derived from measured density and sonic logs by an empirical relation. The model properties are all sampled at 0.5 m intervals and averaged into constant thickness layers. Using 1m layers, when the Q value is carefully measured using the spectral-ratio technique, the measured Q is always lower than that expected from the specified model. In a series of experiments in which various physical effects are turned off and on again, it is demonstrated conclusively that this Q bias (measured Q - Q model ) is due to internal multiples. Using a series of models derived from the same logs but with progressively thicker layers (each model has constant thickness layers and each is sampled a 0.5m) it is demonstrated that there is significant Q bias for layer thicknesses less than 20m but for thicknesses greater than 20m the Q bias disappears. The feasibility of estimating stratigraphic Q from such experiments and using these measurements to correct measurements from field data is discussed.