Seismic method - Geophysics for the Mineral Exploration Geoscientist - page 372

applied to each trace to restore the travel times to those for
an ideal survey geometry, i.e. sources and detectors at
constant topographic elevation with uniform velocities in
the shallow subsurface ( Fig. 6.18b ) . Further static correc-
tions, known as residual statics and based on the lateral
coherence of prominent re ectors, may also be required.
Poor static corrections can signi cantly degrade the
quality of seismic data recorded on land, so considerable
time is spent ensuring that the corrections are as accurate
as possible. Gibson ( 2011 ) shows an example of how
important static corrections can be.

A detailed description of deconvolution in all its forms is
provided by Yilmaz ( 2001 ).

6.5.2.3 Example of pre-stack processing
The improvement in signal-to-noise ratio of the re ections
in a shot gather can be very marked. The data in Fig. 6.19
are from the Kristineberg mining area in northern Sweden
(Ehsan et al., 2012 ) . The local geology consists of deformed
low- to medium-grade metasedimentary and metaigneous
rocks and granitoids. Note the curvilinear moveout char-
acteristic of reflections in a processed shot gather. These
arrivals are barely visible in the unprocessed data. Data
processing included: manually editing out noisy traces,
muting of the airwave, frequency filtering to suppress
surface waves, and deconvolution to compress the reflected
arrivals ( Fig. 6.19b ) ; static corrections ( Fig. 6.19c ) ; then
muting of

Scaling
Gain recovery is intended to increase the smaller ampli-
tudes of reflections originating from greater depths. These
arrivals have travelled further, suffering greater losses due
to energy partitioning, absorption and geometric spreading
(see Section 6.3.3 , and Energy partitioning in Section
6.3.4.2 ) . A time-varying gain (ampli cation) function is
applied to each trace with the largest gains being at later
times, so amplifying the later/deeper arrivals. The gain
function may be based on some mathematical relationship
or derived from the trace itself, for example using auto-
matic gain control methods (see Amplitude scaling in
Section 2.7.4.4 )

first arrivals in preparation for stacking
( Fig. 6.19d ). Another example of pre-stack processing is
the two shot gathers in Figs. 6.13 and 6.14 , the latter being
the processed data. Again there is considerable enhance-
ment of re ected arrivals in the processed data compared
with the unprocessed data.

6.5.2.4 Stacking
Stacking converts the data to an easier-to-interpret zero-
offset form and suppresses non-re ected arrivals, random
and transient noise, and most kinds of coherent noise
(usually arrivals other than P-wave reflections). It produces
a significant improvement in the signal-to-noise ratio of
the data. Furthermore, stacking provides useful informa-
tion about seismic velocities in the subsurface required for
migration and depth conversion (see Section 6.5.2.5 ). For
simplicity, we first describe stacking using a 2D
dataset although it is also routinely applied to 3D data;
the additional complications incurred with 3D data are
considered later.
The first step in stacking involves sorting the shot
gathers from a continuous pro le survey ( Fig. 6.16 ) into
common-midpoint (CMP) gathers ( Fig. 6.20a and b ) . These
consist of sets of traces whose midpoints (halfway between
the source and detector) are at the same location; the
number of traces in each gather is equal to the fold of the
stack, six in this case. The colour-coded raypaths in
Fig. 6.16 are shown, after gathering, in Fig. 6.20a . The
different source

6.5.2.2 Deconvolution
Deconvolution is a form of filtering operation and, as
described in Section 2.7.4.4 , is also called inverse filtering.
It may be applied several times during a seismic processing
sequence, both pre- and post-stack. Deconvolution is used
to undesirable characteristics of the seismic wavelet

resulting from its passage through the subsurface

' undo '

the
effects of the Earth filter. These may involve changing
its shape into one that makes interpretation of the data
easier, for example changing it to zero phase (see online
Appendix 2 ) or shortening it to improve resolution (see
Section 6.7.1 ) .
There are several different types of deconvolution,
differing in how the deconvolution operator is determined,
which itself is related to how well the source wavelet and
the nature of the Earth filter are known. Predictive decon-
volution is commonly used to suppress multiples and
works on the principle that arrivals that are multiples,
being periodic, can be predicted from earlier parts of the
trace whereas primary re ections cannot.
It is not possible to convert the seismic wavelet perfectly
to the desired form, but deconvolution is very effective and
an integral part of any seismic processing sequence.

-

-

receiver offsets means that the raypath
lengths and the travel times of the re ections vary, again
with a curvilinear moveout ( Fig. 6.20b ) .
The next stage of stacking involves analysis of CMP
gathers. The aim here is to combine their constituent traces