Geology Reference
In-Depth Information
Fig. 4.22
Removal of reverberations
by predictive deconvolution. (a) Seismic
record dominated by strong
reverberations. (b) Same section after
spiking deconvolution. (Courtesy
Prakla Seismos GmbH.)
Practically achievable inverse filters are always
approximations to the ideal filter that would produce a
reflectivity function from a seismic trace: firstly, the ideal
filter operator would have to be infinitely long; second-
ly, predictive deconvolution makes assumptions about
the statistical nature of the seismic time series that are
only approximately true. Nevertheless, dramatic im-
provements to seismic sections, in the way of multiple
suppression and associated enhancement of vertical
resolution, are routinely achieved by predictive decon-
volution. An example of the effectiveness of predictive
deconvolution in improving the quality of a seismic
section is shown in Fig. 4.22. Deconvolution may be
carried out on individual seismic traces before stacking
(
deconvolution before stacking
: DBS) or on CMP stacked
traces (
deconvolution after stacking
: DAS), and is com-
monly employed at both these stages of data processing.
4.8.3 Velocity filtering
The use of
velocity filtering
(also known as
fan filtering
or
pie slice filtering
) is to remove coherent noise events from
seismic records on the basis of the particular angles
at which the events dip (March & Bailey 1983). The
angle of dip of an event is determined from the apparent
velocity with which it propagates across a spread of
detectors.
A seismic pulse travelling with velocity
v
at an angle
a
to the vertical will propagate across the spread with an
apparent velocity
v
a
=
v
/sin
a
(Fig. 4.23). Along the
spread direction, each individual sinusoidal component
of the pulse will have an apparent wavenumber
k
a
related
to its individual frequency
f
, where
f
=
v
a
k
a
