Beruflich Dokumente
Kultur Dokumente
Show/Hide Bookmarks
Search
Close
Navigation
Show Hide
3.1 Introduction
Overview
This chapter reviews aspects of seismic data acquisition and processing that are relevant to a seismic
interpreter. These topics are covered extensively by other authors and, because the focus of this text
is on interpretation, the interested reader is referred elsewhere (see below) for in-depth treatment.
The primary purposes of this chapter are to:
1. Briefly explain how seismic data are acquired and processed, including a short
discussion of aspects of 2-D and 3-D seismic survey design.
2. Ensure the reader understands that seismic data acquisition and processing parameters
will have a significant impact on the interpretability of the data (i.e. the appearance of
structural and stratigraphic features).
The chapter ends with a discussion of coherency (also referred to as semblance and other names)
processing, a processing step that is, at least in the petroleum industry, almost routinely applied to
3-D seismic data in order to enhance faults and stratigraphic features.
There is a consensus that a seismic interpretation begins at the survey design phase. As described
below, the target depth and dimensions, structural dip, rate at which high frequencies are attenuated
with depth, and other factors will influence survey design. If these parameters are improperly known
or planned for, the data acquisition effort is unlikely to yield useful results. Similarly, although
data processing may appear to have a cookbook character (follow the recipe for predictable results
every time), it is in fact full of interpretive choices made by the data processors. It is possible
to process seismic data such that they are optimized for some types of analyses but not others.
Although there may be cases for which processing can minimize problems caused by survey design
flaws, no amount of processing can make data interpretable if the survey design is too flawed, or the
acquisition is poorly executed.
Formerly it was common for seismic acquisition, processing and interpretation to be undertaken
Page 1
Show/Hide Bookmarks
Search
Close
Navigation
Show Hide
by separate groups, with little communication between those groups1. For example, the people
processing the seismic data may not have known what the data were to be used for, and the
interpreters may have had little knowledge of how the data were acquired or processed. This is not
a recipe for success. Seismic acquisition, processing and interpretation groups need to work closely
together in order to ensure that: 1) the data quality is optimized for the interpretation task at hand,
and 2) the interpreters understand the limitations of the data being analyzed.
As discussed below, seismic data collected in the field need to be processed to generate images that
come close to representing the subsurface stratigraphy and structure. Seismic processing has several
objectives that include:
Enhancing resolution
The seismic processing canon is Yilmaz (2001), a two-volume publication that presents an extensive,
mathematically focused treatment of the topic but also shows many real-data examples of the
effects of different processing steps on the appearance of seismic data. Duncans (1992) summary
is much shorter and somewhat dated, but still a useful starting point. General aspects of seismic
data acquisition were covered by Evans (1997) and, for marine surveying, Dessler (1992). Further
discussions of acquisition and processing were presented by Sheriff and Geldart (1995), Henry
(1997), Gadallah and Fisher (2005), Veeken (2007) and others. Aspects of data acquisition and
processing particular to 3-D seismic data were presented by Hardage (1997), Cordsen et al. (2000),
Galbraith (2001), Vermeer (2002) and Liner (2004). Sheriff (2002) presented definitions and short
explanations of many terms used in seismic data acquisition and processing. Mosher and Simpkin
(1999) reviewed sources and techniques for the acquisition of high-resolution marine seismic
profiles. Steeples and Miller (1998), Burger et al. (2006) and Schuck and Lange (2008) discussed
aspects of shallow land-based reflection profiling.
Stacking
We generally need or want to increase the signal-to-noise ratio of our seismic data in order to better
define structural and stratigraphic features, and rock properties. Recall from CHAPTER 2 how the
seismic method works. Very simplistically, we make a bang at the surface, the sound travels down
through the subsurface, reflects off various horizons, and we record the energy that was reflected
to the surface (FIGURE 2.1). Those reflections can be weak, depending on the strength of the
seismic source and the distance traveled by the sound. Furthermore, in addition to recording the
reflections, we also record noise. For example, at sea we will record acoustic energy generated by
breaking waves and the engines of the survey ship, in addition to the reflected energy. On land, our
receivers might record acoustic energy generated by vehicles driving by our survey site or electrical
noise generated by power lines passing overhead. Whatever its source, noise will be present in
our data and the noise can be quite strong compared to the reflections that come from horizons
several kilometers below the surface. The ratio of the amplitude (strength) of the reflections to the
__________________________________________________________________
1
Page 2
Show/Hide Bookmarks
Search
Close
Navigation
Show Hide
Page 3
Back to Chapter
The trace at left is a noise-free signal, perhaps the reflection from the top of a bed.
In the middle, the signal is contaminated by random noise on each of the traces,
with the strength of the noise being approximately equal to the strength of the
signal. It is difficult to distinguish that there is indeed a signal. The traces at right
show the effect of stacking (adding) different numbers of noisy traces together. The
signal is still difficult to see when the stacking fold is five, but it becomes clearer for
stacking folds of 10 or higher. The more traces that are stacked together, the more
the random noise will tend to cancel itself out, making the signal easier to detect.
FIGURE 3.1:
Stacking to improve signalto-noise ratio
Back to Chapter
FIGURE 3.2:
Theoretical increases in
signal-to-noise ratio as the
stacking fold is increased
Back to Chapter
FIGURE 3.3:
Reflections from a marine
seismic shot
Show/Hide Bookmarks
Search
Close
Navigation
Show Hide
Required penetration. All else being equal, the deeper the imaging target, the more
energy will be needed to image it.
Availability and cost. Seismic crews or particular types of equipment may not always
be available. This can be particularly true of areas far removed from traditional
petroleum exploration and development activities. Some seasons are preferred
for seismic data acquisition. For example, seismic data in northern Canada is
traditionally collected in the winter, when frozen ground makes these muskeg areas
accessible to vehicular traffic. Crews may wish to avoid marine seismic acquisition
Page 4
Show/Hide Bookmarks
Search
Close
Navigation
Weight drops. These include sledgehammers striking a metal plate on the ground,
and weights (e.g., leather bags filled with birdshot) that are dropped from a height
of 2 m or more. These types of sources are generally used in shallow seismic work.
Burger et al. (2006) suggested that under optimal conditions sledgehammer sources
can be used to detect the contact between overburden and bedrock at depths up to 50
m. Gendzwill et al. (1994) used a sledgehammer source in a mine dug into a granitic
batholith and were able to image fractures and other features over 200 m below the
mine floor.
Explosive sources include dynamite and firearms. Dynamite charges are commonly
used onshore in the petroleum industry. They need to be buried, both for safety
reasons and to ensure that the blast energy is effectively transmitted to the ground,
rather than dissipated as an air blast. Shotguns (FIGURE 3.4A) and other firearms
are sometimes used in shallow seismic work (e.g., engineering and hydrogeology
studies), and they provide more energy than sledgehammer sources. Seeber and
Steeples (1986) described the use of a .50-caliber machine gun as a seismic source
that produced frequencies in the range of 30 to 170 Hz. Even automobile spark plugs
have been tested as seismic sources for very shallow surveying; they generate high
frequencies but low energy levels (Don Steeples, personal communication, 2008).
Show Hide
Show/Hide Bookmarks
Search
Close
Navigation
Show Hide
Back to Chapter
Page 6
FIGURE 3.4:
Selected seismic sources
Back to Chapter
FIGURE 3.5:
Schematic illustrations of
seismic receivers
Back to Chapter
ANIMATION 5:
Various types of equipment
and activities associated with
both marine and land-based
seismic acquisition
Show/Hide Bookmarks
Search
Close
Navigation
Show Hide
ANIMATION 5 (courtesy Global Geophysical Services) shows various types of equipment and
activities associated with both marine and land-based seismic acquisition. It shows seismic crews
engaged in various types of activities, including drilling shot holes (for dynamite), collecting
airgun seismic data with relatively small vessels, having safety meetings, surveying the locations of
geophones and shotpoints for land acquisition, and generating seismic waves with vibroseis trucks.
Receivers
Like the choice of sources, the type of receiver used depends on whether the seismic data is being
collected on land or at sea.
On land, the primary type of receiver is the geophone. A geophone is a motion detector, and a
schematic geophone design is shown in FIGURE 3.5A. The spike of a geophone is planted into
the ground to ensure good coupling between the ground motion and the device itself. Inside the
geophone is a hollow magnet that is attached to the body of the geophone. Reflected P-waves cause
the ground surface, and therefore the geophone and the magnet, to vibrate up and down. A wire
coil is suspended within the opening of the magnet. Because of inertia, the wire coil will tend not
to move at the same rate as the geophone housing. This differential movement causes the coil to
pass through the magnetic field generated by the magnet and therefore, in accordance with Faradays
Law, generating an electrical current in the coil. This electrical signal is the recorded seismic trace,
which is then transmitted to the recording truck.
At least in most petroleum applications, more than one geophone is planted at a receiver location,
forming what is known as a geophone array. Designing the geophone array (e.g., should they be
arranged in a line, in a cross or some other shape?) can be important. The location of each geophone
group needs to be accurately known through surveying. The recordings of each geophone array are
stacked prior to recording in an effort to enhance the signal-to-noise ratio. With newly designed
Q-Technology (by WesternGeco) all active geophones or hydrophones (perhaps more than 30,000
for a single shot on land) are recorded separately. This acquisition effort allows all sorts of to noise
be more effectively removed, preserves signal fidelity and high frequencies in the pre-stack data.
Shabrawi et al (2005) described the first application of this technology to improve seismic imaging
of a carbonate reservoir in Kuwait.
When the objective is to record shear waves, the geophone needs to be designed to detect horizontal
ground motions. FIGURE 3.6 compares the vertical motions recorded by P-wave geophones to the
horizontal motions (in two mutually orthogonal directions) measured by S-wave geophones. Threecomponent geophones have two S-wave detectors arranged at right angles and a P-wave detector.
New geophones have been designed that capture the full wavefield motion generated as the reflected
energy returns to the surface. Subsequent processing is used to decompose this energy into the
constituent P-wave and S-wave signals.
During normal land-based operations, the geophones are planted into the ground and left there for
most of the acquisition program. Only the source is moved, although the receivers can be selectively
turned on or off depending on whether they are needed to record the reflections coming from
below. Manually planting and removing the geophones can be a time-consuming, and therefore
costly, process. One innovative design, used to date in shallow geophysical work, involves a land
Page 7
Show/Hide Bookmarks
Search
Close
Navigation
Show Hide
Page 8
Back to Chapter
FIGURE 3.6:
Comparison of the vertical
motions recorded by P-wave
geophones to the horizontal
motions (in two mutually
orthogonal directions)
measured by S-wave
geophones
Back to Chapter
FIGURE 3.7:
A small vibroseis truck
towing a land streamer, a
series of geophones in a cable
that is dragged along the
ground surface
Show/Hide Bookmarks
Search
Close
Navigation
Show Hide
surveying, the location of each receiver group (x, y and z coordinates) needs to be
known precisely.
Show/Hide Bookmarks
Search
Close
Navigation
Show Hide
Back to Chapter
FIGURE 3.8:
Schematic illustration of
raypaths generated by a shot
from a source point into a
line of receivers during 2-D
seismic acquisition
Back to Chapter
FIGURE 3.9:
Comparison of an end-on 2-D
seismic acquisition geometry
and a split-spread 2-D
seismic acquisition geometry
Back to Chapter
FIGURE 3.10:
The common midpoint
method
__________________________________________________________________
2
Recall however from SECTION 3.2 that the receiver groups are usually not physically moved between shots on land.
Instead all the receivers may be laid out prior to surveying and different combinations of receiver groups are activated
for each shot.
Page 10
Show/Hide Bookmarks
Search
Close
Navigation
Show Hide
t t0
h2
V 2 NMO
Back to Chapter
[3.1]
Page 11
Although the reflections all come from the same midpoint, the distances traveled by
the raypaths are different because of the differences in distance (offset) between the
sources and receivers. NMO corrections level out the reflection, making it possible to
stack the traces to produce one trace that represents the geology at the CMP location.
FIGURE 3.11:
Schematic representation of
the common midpoint gather
derived in Figure 3.10 before
and after normal moveout
(NMO) corrections
Back to Chapter
FIGURE 3.12:
Hyperbolae approximating
arrival times versus offset
Back to Chapter
Primary reflections
form hyperbolae (red
lines) in part A that
are leveled out after
NMO corrections.
All traces in part B
will subsequently be
stacked to generate
a single trace in a
2-D seismic line or
3-D seismic volume.
From Duncan (1992).
FIGURE 3.13:
A common midpoint
gather before, and after
velocity analyses and NMO
corrections
Show/Hide Bookmarks
Search
Close
Navigation
Show Hide
Vint
t 2Vrms 2 t1Vrms1
t 2 t1
[3.2]
Back to Chapter
FIGURE 3.14:
Sample 1980s vintage 2-D
seismic profile showing
locations of velocity analyses
used for stacking
Back to Chapter
FIGURE 3.15:
Application of the Dix
equation (Equation 3.2) to
calculate interval velocities
Back to Chapter
Figure 3.16: Reflections from all depths where all horizons are planar
and horizontal, and there are no lateral velocity
variations within layers, still line up at the midpoint
Previous images (e.g., FIGURE 3.8) showed only one reflecting horizon,
but midpoints should line up vertically for different levels if the
stratigraphy does not dip and there are no lateral velocity variations.
FIGURE 3.16:
Reflections from all depths
where all horizons are planar
and horizontal, and there are
no lateral velocity variations
within layers, still line up at
the midpoint
Show/Hide Bookmarks
Search
Close
Navigation
Show Hide
Back to Chapter
Correct stacking velocities are used in part A, but velocities in part B have been
deliberately modified by approximately 10% to show how the image would be
degraded. From Henry (1997a). Reproduced with permission from Editions Technip.
FIGURE 3.17:
Importance of correctly
picking stacking velocities
on the quality of the stacked
seismic image
Back to Chapter
FIGURE 3.18:
Decimation test to illustrate
the importance of stacking
fold on image quality
Show/Hide Bookmarks
Search
Close
Navigation
Show Hide
The stacking fold needed to generate a useable or good seismic image varies from basin to
basin, and also varies according to parameters such as source energy, etc. A stacking fold of 30
might generate an acceptable image in one basin whereas a stacking fold of 120 might be needed
somewhere else to generate an equivalent image. Experience, gained from examination of existing
data, can help guide acquisition efforts. The stacking fold for a 3-D seismic survey does not have to
be as high as the fold in a 2-D survey to generate a comparable image. Krey (1987) suggested that
to obtain comparable data quality, the stacking fold of the 3-D data Nf3 can be calculated using:
Nf3 Nf2
f
100
[3.3]
where Nf2 is the 2-D stacking fold and f is the frequency of interest. Lansley (2004) suggested
however that for 3-D seismic data the important quantity is not the stacking fold but rather the trace
density, defined as the traces per square kilometer.
In addition to improvements to signal to noise ratio, another benefit of the NMO correction and
stacking combination is that it tends to remove long-period multiples. Multiples take different
paths through the subsurface than primary reflections (FIGURE 3.3). As such, NMO corrections
that cause primaries to line up horizontally do not cause multiples to line up. The energy from
multiples does not stack constructively and therefore the multiples tend to be attenuated in the
stacked 2-D data.
One disadvantage of the stacking process is that it reduces the frequency content of the seismic data
somewhat. Reflections that are imperfectly lined up will stack together to generate a broader, i.e.
lower frequency, stacked trace (FIGURE 3.19).
In some cases the sources and receivers cannot be laid out in a straight line, perhaps because the
seismic acquisition crew needs to follow a bending road. These types of 2-D acquisition geometries
are known as crooked line geometries and a variety of problems can be encountered when processing
and interpreting these data. Wu (1992) illustrated these problems using data acquired from a mining
camp, but indicated that similar issues would be encountered when collecting seismic data in any
type of environment. Similar problems can be encountered during 2-D marine surveying if cross
currents cause receiver cable drift, such that the cable does not truly form a straight line behind the
ship (a problem known as cable feathering). Nedimovi et al. (2003) studied this latter problem and
proposed a way for imaging 3-D structure in these cases.
Page 14
Show/Hide Bookmarks
Search
Close
Navigation
Show Hide
Back to Chapter
FIGURE 3.19:
The high-frequency content
of the seismic data is reduced
if reflections are imperfectly
aligned prior to stacking
Back to Chapter
FIGURE 3.20:
Schematic illustration of
a simple 3-D acquisition
geometry on land
Back to Chapter
The seismic trace at each midpoint location will be used to represent an area known as a bin,
the x-y dimensions of which are equal to the trace spacing. Binning the data in this way gives
the appearance of spatial continuity when viewing 3-D seismic data (CHAPTER 4).
FIGURE 3.21:
Plan-view image of midpoints
(blue dots) from a 3-D
seismic survey
Show/Hide Bookmarks
Search
Close
Navigation
Show Hide
Back to Chapter
Figure 3.22: Horizontal slices through two different 3-D seismic cubes collected
over the same area but with different acquisition geometries
High amplitudes in orange/red and low amplitudes in blue. Curvilinear blue areas/lines show the locations of faults.
The acquisition geometry used for the survey on the right allowed the faults to be more sharply defined (compare
arrowed faults between the two images) and is therefore the preferred acquisition geometry in this setting.
Modified from Vermeer (2002) and reproduced with permission.
FIGURE 3.22:
Horizontal slices through two
different 3-D seismic cubes
collected over the same area
but with different acquisition
geometries
Back to Chapter
Marine 3-D seismic data acquisition differs from landbased acquisition. In early programs, a ship towed an
airgun array and two streamers. Each shot generated two
lines of midpoints, one for each streamer. As the ship
sailed to the new shotpoint location, new shots would
generate new midpoints, some of which would be the
same as for previous shots. Unlike land-based acquisition
where geophones commonly need to be planted manually
and dynamite or vibroseis trucks moved from shotpoint
to shotpoint, vessels designed for 3-D seismic acquisition
continuously sail from shotpoint to shotpoint and move the
source and receivers as they sail. As such marine seismic
acquisition costs are lower than land-based surveying
and, as a consequence, trace spacing in marine surveys is
typically much closer than in land-based 3-D surveys.
Currently, two airgun arrays and two or more streamers
(perhaps as many as 12) are usually deployed behind a
ship during 3-D acquisition at sea. Liner (2004) discussed
the advantages of these recording geometries. Marinewide azimuth seismic data are proving to be particularly
beneficial for problems such as sub-salt imaging surveys.
A parallel geometry is the most commonly used marine
3-D acquisition design. The survey ship sails a series of
parallel lines. In some cases it may be desirable to use two
ships, one pulling the airgun array(s) and the other towing
Page 16
FIGURE 3.23:
Highly simplified traditional
seismic processing flow
showing the order in which
selected processing steps are
completed
Show/Hide Bookmarks
Search
Close
Navigation
Amplitude Recovery
The acoustic pulse loses energy as it spreads away from the source (i.e. the farther one is from a
source of sound, the quieter the sound will seem). As a result, horizons in the shallow part of the
section will appear to have strong amplitudes and reflections from deeper horizons will be weaker.
Spherical divergence is a term that refers to the apparent loss of energy due to geometric spreading
of a wavefront. The energy at a point on an expanding wavefront decreases with the square of
the distance from the energy source. We need to compensate for this loss of signal amplitude as
it travels through the earth because we want the strength of a reflection to be proportional to the
acoustic impedance change, and not be affected by the depth of the reflector.
Automatic gain control (AGC) is one type of process, used in many electronics applications, that
compensates for loss of signal strength with distance. AGC is an adaptive system that attempts to
maintain a constant signal level. In the seismic processing realm it will produce a seismic image
in which reflection amplitudes will remain relatively constant with depth. Although this type of
processing can help enhance the appearance of structural and stratigraphic features deep in the
section, information about seismic amplitudes that is associated with acoustic impedance changes
is lost. As such, AGC is no longer in favor in seismic processing situations where amplitude
preservation is critical.
Other approaches exist to recover amplitudes, including application of a correction for spherical
divergence. FIGURE 3.24 shows a shot gather before and after amplitude recovery. Note how the
amplitude recovery enhances the appearance of deep reflections. These events will be visible in the
stacked data following the amplitude recovery, but would not be visible otherwise.
True-amplitude recovery is a processing approach that attempts to compensate for attenuation,
spherical divergence and other effects. The objective is to produce seismic images in which the
reflection amplitudes are directly proportional to changes in acoustic impedance (i.e. reflectivity) at
all levels of the seismic data (both before and after stacking). Compensating for energy losses due
to spherical divergence is relatively straightforward, but other factors are more problematic. Recall
__________________________________________________________________
3
This technique is known as undershooting. Undershooting is also used in land-based acquisition to collect data
beneath surface obstructions such as small lakes or land held by uncooperative landowners.Instead all the receivers
may be laid out prior to surveying and different combinations of receiver groups are activated for each shot.
Page 17
Show Hide
Show/Hide Bookmarks
Search
Close
Navigation
Show Hide
Static Corrections
Seismic data record the two-way traveltime from the
surface to a reflecting horizon and back up to the surface.
By collecting a 2-D profile or a 3-D volume we seek to
image the structure at depth, taking into account velocity
problems like those illustrated in FIGURE 2.1. However
we will have an imaging problem if the surface over which
we move our sources and receivers is not horizontal.
FIGURE 3.25 illustrates this problem in a simple
conceptual example. In the illustration, the surface over
which the seismic data are to be collected has a hill, and
the target horizon at depth is planar and horizontal. If all
we know is the two-way traveltime, our seismic image will
show a false structure, because it takes longer for the sound
to reach the horizon beneath the hill. Clearly we have to
account for changes in surface elevation along the length
(or area) of our seismic survey. In order to do so, we need
to specify an elevation, known as the seismic reference
datum, which corresponds to 0 s two-way traveltime in our
seismic image. At sea, or in low-lying coastal areas, mean
sea level is the obvious choice. In areas on land which are
considerably above sea level, it is necessary to arbitrarily
pick a seismic reference datum.
Changes in ground elevation are but one type of problem
(FIGURE 3.26). In some areas unconsolidated materials
in the shallow subsurface (e.g., marsh deposits, glacial
till, sand dunes) can have velocities that are much lower
than the underlying rocks. In some cases, high-porosity,
uncemented sediments in the vadose zone (not all pore
space filled with water) can have velocities of nearly 100
m/s, less than the velocity of sound in air (Schuck and
Lange, 2008). This surface layer is sometimes referred to
as the low-velocity layer or the weathered layer. Changes
in thickness of this layer, if not properly accounted for, can
cause imaging problems for the underlying strata.
Page 18
Back to Chapter
FIGURE 3.24:
Comparison of a source
gather before and after
amplitude recovery
Back to Chapter
Figure 3.25: Illustration of imaging problem that occurs when the surface over
which we move our sources and receivers is not horizontal
FIGURE 3.25:
Illustration of imaging
problem that occurs when the
surface over which we move
our sources and receivers is
not horizontal
Back to Chapter
FIGURE 3.26:
Statics Corrections
Show/Hide Bookmarks
Search
Close
Navigation
Show Hide
Back to Chapter
FIGURE 3.27:
Common midpoint gather
before and after statics
corrections, and prior to any
NMO corrections
Back to Chapter
Figure 3.28: Stacked seismic image A) without and B) with statics corrections
Deconvolution
Deconvolution is a processing operation, generally applied
before stacking, that attempts to:
FIGURE 3.28:
Stacked seismic image
without and with statics
corrections
Back to Chapter
FIGURE 3.29:
Comparison of stacked
seismic section generated
without and with
deconvolution prior to
stacking
Show/Hide Bookmarks
Search
Close
Navigation
Show Hide
the version of the data without deconvolution, whereas only two are visible when deconvolution has
been applied. One of the peaks was probably either a reverberation or a short-period multiple and
deconvolution has removed it. The resulting image more truly represents changes in amplitude that
are due to changes in acoustic impedance (i.e. the stratigraphy) in the image. Similar changes are
visible in other parts of the image.
Henry (2001) divided most deconvolution methods into two categories: deterministic deconvolution
and statistical deconvolution. Deterministic deconvolution can be applied when the source wavelet
is known, either because it is measured or it can be modeled. In this case, the deconvolution can
produce a seismic section that truly is zero phase. Generally, the source wavelet is an unknown
and so a statistical approach is taken. Two key assumptions that need to be made when working
with statistical deconvolution are that: a) the reflectivity series is random (sometimes referred to as
having a white spectrum), and b) that the input wavelet was minimum phase. Neither of these
assumptions is usually true, and so statistical deconvolution will not produce a truly zero-phase
wavelet. Although in CHAPTER 2 we noted that the wavelets generated by airgun and dynamite
sources are approximately minimum phase, the differences between the actual character and an ideal
minimum phase wavelet are significant enough to affect the deconvolution result.
FIGURE 3.30 compares the results of two different deconvolution methods applied to the same
dataset. In part A, a statistical deconvolution has been applied that resulted in a mixed-phase wavelet
being embedded in the seismic data. In part B, deterministic deconvolution was applied to the same
dataset for purposes of comparison (the source wavelet was known). The wavelet embedded in the
statistical deconvolution result is not zero phase, whereas the wavelet embedded in the deterministic
deconvolution result is truly zero phase. The deterministic deconvolution (i.e. zero phase) image has
better definition of structural and stratigraphic features.
A consequence of the shortcomings of statistical deconvolution methods is that we are never sure
of the seismic data phase even if, in principle, the data were processed to zero phase. We need to
determine the phase of the seismic data independently, and methods for doing so are described in
CHAPTER 5 when discussing synthetic seismograms.
Deconvolution is commonly not applied in shallow seismic work (Don Steeples, Personal
Communication, 2008). Reasons include:
The reflectivity profile is not random. Commonly only 3 or 4 reflections may be real,
and the statistical basis of some deconvolution methods is not valid in these cases.
Low signal-to-noise ratios. Deconvolution does not work properly in these cases.
Migration/Seismic Imaging
In the simple cases examined so far, horizons in the subsurface have been planar and horizontal. In
this case, and assuming that there are no statics-related problems at the surface, reflections are from
the midpoint between a source and a receiver. We now consider a scenario where a seismic transect
is collected above a dipping horizon.
Page 20
Show/Hide Bookmarks
Search
Close
Navigation
Show Hide
Page 21
Back to Chapter
Seismic image on left was generated using a statistical deconvolution approach that left
a mixed-phase wavelet in the data. Seismic image on the right was generated using a
deterministic deconvolution that generated a truly zero-phase version of the data. Arrows point
to improvements in image quality in the zero-phase version of the data. From Henry (2001).
FIGURE 3.30:
Comparison of two different
deconvolution results on the
same data
Back to Chapter
When beds dip the reflection points are not located at midpoints because the angle of incidence
(1) is equal to the angle of reflection (2) for the dipping bed. However, and in the absence of
any other knowledge of subsurface structure, the reflection is still assumed to come from the
midpoint location. In other words, the reflection will be misplaced spatially from its true location.
FIGURE 3.31:
A source and receiver
combination located at Points
A and C recording a reflection
from Point B on the dipping
surface
Back to Chapter
FIGURE 3.32:
Surface with a concave-up
feature showing a source
location and a series of
receiver locations
Show/Hide Bookmarks
Search
Close
Navigation
Show Hide
Back to Chapter
In this schematic
sch
example, the source
location is directly
above the buried body
energy is
and diffracted
ra
by receivers on
received b
sides.
both sides
FIGURE 3.33:
A single point reflector
generates a diffraction in a
stacked seismic image
Back to Chapter
FIGURE 3.34:
Comparison of stacked,
unmigrated seismic image
with migrated version of the
same data
Back to Chapter
Collapsing diffractions.
Page 22
FIGURE 3.35:
Comparison of an unmigrated
and migrated seismic image
of a fault
Show/Hide Bookmarks
Search
Close
Navigation
Show Hide
migration), and a discussion of these methods is beyond the scope of this book. Interested readers
are referred to the texts listed in the Introduction to this chapter for more details on migration. Here,
and following Sheriff (2002), Liner (2004) and others, we distinguish between migration approaches
according to two principal factors:
Is the migration performed before or after stacking the data (pre- or post-stack
migration respectively)?
Does the migration properly handle lateral velocity variations and account for ray
bending (depth migration) or does it not (time migration)?
All migration methods require some sort of velocity model in order to undertake the mathematical
operation of migration. Velocity information needed to build that model can be obtained in a
variety of ways. The derivation of stacking velocities was discussed in SECTION 3.3. Checkshot
surveys, vertical seismic profiling and sonic logs are described in CHAPTER 5. Seismic traveltime
tomography uses seismic recordings to construct a velocity model that minimizes the error between
the measured and theoretical traveltimes. Different approaches are possible to tomography, and
Rawlinson et al. (2003), Accaino et al. (2005), and Lehmann (2007) discuss various applications
ranging from engineering to lithospheric imaging.
Bednar (2005) discussed the history of seismic migration. Originally, i.e. before digital recording,
migration was performed manually (on paper). Even once digital recording became commonplace
(e.g., 1960s and 1970s) much seismic data was processed and displayed without being migrated.
Subsequently, migration was applied after stacking, and stacking velocities, usually derived from
NMO corrections, were used as a first guess. These velocities were varied by perhaps 5 or 10%
to examine the effects of changing the migration velocity on the seismic image. If the migration
velocities are too low, the data are said to be under-migrated and concave-up patterns known as
migration smiles are apparent. Migrated noise in the deep part of seismic images can generate
prominent migration smiles. If the migration velocities are too high, i.e. the data are over-migrated,
concave-down patterns known as frowns may be visible. The velocity field that provided the
sharpest image and best removed the diffractions and other problems was selected as the appropriate
velocity model for migration. This approach is known as post-stack time migration (Post-STM)
because the data are migrated after stacking with the vertical axis in time.
Post-stack time migration is an acceptable, and relatively cheap, way of migrating seismic data
when velocities vary vertically in a more-or-less uniform manner, and lateral velocity variations
are minimal or non-existent (e.g., areas with a structurally undisturbed layer-cake stratigraphy).
However this type of migration is inadequate in the presence of abrupt lateral and vertical velocity
variations such as those that might be present adjacent to salt diapirs or in thrusted areas (e.g.,
Paleozoic carbonates thrust over Tertiary clastics). In these cases, raypaths (or, more correctly,
wavefronts) can follow complicated paths as they are refracted (according to Snells law, which is
not honored in time imaging) at dipping interfaces having high velocity contrasts (e.g., Figures 2.24,
2.26). Reflections do not necessarily come from common midpoints in these areas (Figures 3.36,
3.37), and stacking then migrating these reflections will not produce a clear or geometrically
accurate image.
In cases of steep dips and significant vertical and lateral velocity variations the data need to be
Page 23
Show/Hide Bookmarks
Search
Close
Navigation
Show Hide
Back to Chapter
Although these two source-receiver combinations share the same midpoint, the reflections do not
come from the same subsurface locations because of the dipping bed. (Raypaths bend according
to Snells Law and the angle of incidence is equal to the angle of reflection). Stacking these two
traces will not help to generate a good-quality, or geometrically accurate, seismic image.
FIGURE 3.36:
Source-receiver combinations
with a common midpoint
Back to Chapter
Page 24
Raypaths for reflections from a single horizon for a common midpoint gather in the presence of a
salt body and associated structural deformation. Note that reflection points do not coincide. Inset at
lower right shows CMP gather. Reflections from this one interface do not fall along a simple hyperbola
(e.g., FIGURE 3.12) and NMO corrections followed by migration will not be adequate to generate an
acceptable image. Pre-stack depth migration is needed in this case. From Abriel et al. (2004).
FIGURE 3.37:
Reflections below a salt body
do not come from a common
midpoint
Back to Chapter
FIGURE 3.38:
Comparison of post-stack
time-migrated seismic image
and pre-stack depth migrated
seismic image of a salt body
in the Gulf of Mexico
Show/Hide Bookmarks
Search
Close
Navigation
Show Hide
Back to Chapter
FIGURE 3.39:
Comparison of three
different pre-stack migration
algorithms: Kirchoff
migration, wave-equation
migration, and controlledbeam migration
Back to Chapter
FIGURE 3.40:
Schematic illustration
showing how sideswipe is
generated
FIGURE 3.41:
Modeling experiment
undertaken by Fred Hilterman
in the early 1970s to illustrate
the appearance of sideswipe
from a carbonate buildup on
2-D seismic profiles
Page 25
Back to Chapter
Show/Hide Bookmarks
Search
Close
Navigation
Show Hide
to the bedding and perpendicular to the bedding can be as high as 30% (Lawton et al., 2001) the
assumption of an isotropic velocity field can lead to misplaced and poorly imaged reflections.
3-D Migration
One of the biggest pitfalls of 2-D seismic methods is that acoustic energy does not simply follow
2-D raypaths such as those shown previously. Sound expands outwards from the source as a 3-D
wavefront in the subsurface. As such, geological bodies located off to the side of the line can
generate reflections that are recorded by receivers on the line (FIGURE 3.40). These out-of-plane
reflections are known as sideswipe. Sideswipe, or seeing off to the side on a 2-D profile, needs to
be acknowledged as a potential problem when working with that type of data. Conventional 2-D
migration cannot account for this problem. Instead, and as recognized at least in the 1970s (e.g.,
French, 1974), 3-D migration can be used to eliminate sideswipe.
The image shown in FIGURE 3.41 illustrates this point. The upper part of the figure shows a map
view (top) and transect (below) through a geological model of a carbonate buildup. The locations
of six modeled 2-D seismic lines are also shown. The lower images show these modeled seismic
profiles. Note that the carbonate buildup is imaged even on lines that do not cross the feature (e.g.,
1500). That is because the sound expands out in three dimensions away from the seismic source.
Some of that sound may be reflected from points located below the line of receivers, but some of the
sound may travel out away from the line, then bounce off a surface back into the line of geophones.
There is commonly no way of knowing whether a feature on a 2-D profile lies directly under the
surface survey line or off to the side. Note that no migration was applied to the modeled seismic
profiles shown in that figure because it was not common to apply even 2-D migration in the early
1970s. As such, the flanks of the buildup are poorly imaged in the seismic profiles. Although it
would improve the seismic definition of the flanks of the buildups, 2-D seismic migration would not
solve the sideswipe problem illustrated here.
The modeling shown in FIGURE 3.41 was designed to illustrate the effects of sideswipe on the
exploration for carbonate buildups in the Michigan Basin, but has proven to be useful elsewhere.
For example, Crawley Stewart (1995) described problems encountered by companies exploring for
Pennsylvanian algal mounds in southeast Utah using 2-D seismic-based mapping. There would be
times that the seismic profiles indicated the presence of a buildup and drilling would confirm the
seismic prognosis. However, there would be times that a buildup would be clearly observed on
a 2-D seismic profile, but drilling did not find one. Only 3-D seismic, and the application of 3-D
migration, could reliably indicate the presence or absence of productive mounds. This particular
exploration problem is illustrated in FIGURE 3.42.
FIGURE 3.43 illustrates how 3-D migration eliminates sideswipe. Part A of that figure shows the
same image as shown in FIGURE 3.39, but this time with the Fresnel zone illustrated. The diameter
of the Fresnel zone is given by EQUATION 2.19. FIGURE 3.43B shows how 2-D migration
shrinks the Fresnel zone. The diameter is shrunk to 4 but only in the direction of the 2-D line. The
Fresnel zone diameter is unchanged perpendicular to the line, and so sideswipe problems remain.
Because of the way that sources and receivers image midpoints from a variety of azimuths with a
3-D acquisition geometry, the Fresnel zone is shrunk to 4 in all directions following 3-D migration
Page 26
Show/Hide Bookmarks
Search
Close
Navigation
Show Hide
Processing Summary
In an ideal world, the seismic data represent geometrically
accurate images of the subsurface after processing. The
seismic traces simulate data that would have been collected
using waves that traveled vertically up and down from
the surface to the reflecting horizons and back up at the
common-midpoint location. Stacking, migration and statics
corrections are all processing steps that are intended to
produce this type of image. Multiples and other sources of
noise have been removed through deconvolution, stacking
and potentially other processing steps for noise or multiple
suppression that were discussed by Yilmaz (2001) and
Gadallah and Fisher (2005). We have a true image of the
geology, within the resolution limits of the seismic data.
Unfortunately, in practice most seismic datasets have
varying degrees of imaging problems and various types of
noise, and an experienced seismic interpreter needs to be
able to be aware of these limitations.
Page 27
Back to Chapter
Figure 3.42: 2-D seismic transect (top) and corresponding transect from a 3-D
seismic volume (below) showing the appearance of a sideswipe
FIGURE 3.42:
2-D seismic transect and
corresponding transect from a
3-D seismic volume showing
the appearance of a sideswipe
Back to Chapter
FIGURE 3.43:
3-D migration removes
sideswipe
Back to Chapter
The 2-D line was in fact extracted from the 3-D acquisition program and processed
using 2-D migration that did not remove sideswipe. Notice the presence of the
bump in part A (red arrow) that is not seen in the image with the 3-D migration.
That feature is due to sideswipe on the 2-D image. Data courtesy Western Geco.
FIGURE 3.44:
Comparison of seismic
transect with 2-D migration
with the same transect
through a 3-D volume
Show/Hide Bookmarks
Search
Close
Navigation
Show Hide
V
4 F sin D
[3.4]
Back to Chapter
The inset quote reminds us of the difficulties associated with simplifying complex workflows.
FIGURE 3.45:
Conceptual diagram for
seismic acquisition
Back to Chapter
FIGURE 3.46:
Spatial aliasing
Show/Hide Bookmarks
Search
Close
Navigation
Show Hide
Bin size in 3-D seismic data. As per the discussion of 2-D seismic trace spacing, the
bin size of a 3-D survey needs to be small enough to avoid spatial aliasing problems.
The bin size also helps to define the lateral resolution of a 3-D seismic survey. One
rule of thumb is that the bin size should be small enough so that three bins will fall
into the narrowest feature that needs to be resolved. For example, if an interpretation
team needs to be able to consistently map channels that are 50 m wide, then the
50
desired maximum bin size would be 3 18 m. In reality, the true lateral resolution
of a 3-D survey is defined by whatever is greatest, 3 times the bin size or the postmigration Fresnel zone. Consider this example: A 3-D survey has a bin size of 15 m,
a dominant frequency of 50 Hz, and an interval velocity of 2400 m/s at the level of a
particular horizon. The wavelength is determined to be 48 m (using
Size of survey. The survey should be designed so that the entire area of the target is
covered by full-fold data (FIGURE 3.47). This area is sometimes referred to as the
sweet spot and, in the petroleum industry, it should be big enough for the structural
or stratigraphic limits of a hydrocarbon reservoir to be defined. At the end of a 2-D
seismic line or around the margins of a 3-D survey will be an area that has midpoints
but the fold and range of offsets will not be adequate to properly stack or migrate the
data. This poorly imaged area, commonly trimmed off by processors and sometimes
referred to as the migration fringe, combines with the sweet spot to constitute the
image area, i.e. the area for which seismic traces are recorded. Outside of the image
area is the acquisition fringe, an area at the surface where receivers (and possibly
sources) need to be located but for which no CMP locations (i.e. no traces) will be
generated4. The dimensions of the acquisition area are a function of the depth to the
target and the structural dip.
Once data collection begins, it is advisable to monitor the acquisition effort in order to ensure
that hardware and crews are properly located and performing properly. Field records need to be
monitored to ensure that the data quality will be adequate. If not, acquisition might need to be
terminated.
__________________________________________________________________
4
Recall from FIGURE 3.8 that the area of midpoint coverage does not extend to include the entire length of the spread.
Page 29
Show/Hide Bookmarks
Search
Close
Navigation
Show Hide
Page 30
Back to Chapter
Figure 3.47: Illustration of the sweet spot for 3-D seismic imaging
The entire seismic target should be covered by a full-fold area known as the
sweet spot. The image area includes the sweet spot and the adjacent, poorly
imaged areas that cannot be properly stacked or migrated because not enough
traces are available. The acquisition area includes areas outside of the image
area where receivers will be located, but for which no data will be collected.
FIGURE 3.47:
Illustration of the sweet spot
for 3-D seismic imaging
Back to Chapter
FIGURE 3.48:
Examples of how
reprocessing was used to
improve seismic image
quality
Back to Chapter
FIGURE 3.49:
Timeslice through some
marine seismic data showing
strong acquisition footprint
and corresponding timeslice
through a reprocessed version
of the data
Show/Hide Bookmarks
Search
Close
Navigation
Show Hide
Back to Chapter
FIGURE 3.50:
Factors affecting reflection
coherency across a fault
Back to Chapter
FIGURE 3.51:
Comparison of seismic
transect through a 3-D
seismic amplitude volume,
and corresponding transect
through a coherency
(semblance) volume
Back to Chapter
FIGURE 3.52:
Sample coherency cube
showing the locations of
faults (black)
__________________________________________________________________
5
For simplicity, I will use the term coherency to refer to these processing products, even though there are differences
in the way the attributes are derived mathematically.
Page 31
Show/Hide Bookmarks
Search
Close
Navigation
Show Hide
Back to Chapter
FIGURE 3.53:
Comparison of A) amplitude
timeslice and B) coherency
(semblance) timeslice for a
structurally complex area
Page 32
FIGURE 3.54:
Comparison of timeslices
through coherency volumes
before and after post-stack
processing using principal
component filtering
Show/Hide Bookmarks
Search
Close
Navigation
Show Hide
3.9 References
Abriel W.L., J.P. Stefani, R.D. Shank, and D.C. Bartel, 2004, 3-D depth image interpretation, in A.R.
Brown, Interpretation of 3-D seismic data (6th ed.). AAPG Memoir 42, p. 449-475.
Accaino, F., G. Bhm, and U. Tinivella, 2005, Tomographic inversion of common image gathers:
First Break, v. 23, p. 39-44.
Albertin, U., M. Woodward, J. Kapoor, W. Chang, S. Charles, D. Nichols, P. Kitchenside, and W.
Mao, 2001, Depth imaging examples and methodology in the Gulf of Mexico: The Leading
Edge, v. 20, p. 498513.
Bahorich, M., and S. Farmer, 1995, 3-D seismic discontinuity for faults and stratigraphic features:
The coherence cube: The Leading Edge, v. 14, p. 1053-1058.
Bednar, J.B., 2005, A brief history of seismic migration: Geophysics, v. 70, p. 3MJ-20MJ.
Burger, H.R., Sheehan, A.F., and Jones, C.H., 2006, Introduction to applied geophysics, exploring
the shallow subsurface. W.W. Norton and Company, 554 p.
Chopra, S. and K.J. Marfurt, 2008, Gleaning meaningful information from seismic attributes: First
Break, v. 26, p. 43-53.
Cordsen, A., M. Galbraith, and J. Peirce, 2000, Planning land 3-D seismic surveys: SEG Geophysical
Development Series, 9, 204 p.
Crawley Stewart, C.L., 1995, 3-D solution to a 2-D pitfall: seismic detection of carbonate buildups,
Kiva Field, Paradox Basin, San Juan County, Utah, in High-definition seismic 2-D, 2-D swath,
and 3-D case histories (R.R. Ray, ed.), Rocky Mountain Association of Geologists Guidebook,
p.177-183.
Cummings, D.I. and H.A.J. Russell, 2007, The Vars-Winchester esker aquifer, South Nation River
watershed: Geological Survey of Canada, Open File 5624, 68 p.
Dessler, J. F., 1992, Marine seismic data acquisition in Development Geology Reference Manual
(Ms. Morton-Thompson and A. M. Woods, eds.): AAPG Methods in Exploration 10, p. 361-363.
Dix, C.H., 1955, Seismic velocities from surface measurements: Geophysics, v. 20, p. 68-86.
Duncan, P.M., 1992, Basic seismic processing in Development Geology Reference Manual (M.
Morton-Thompson and A.M. Woods, eds.): AAPG Methods in Exploration 10, p. 364-271.
Evans, B.J., 1997, A handbook for seismic data acquisition in exploration. SEG Geophysical
Monographs Series, 7, 305 p.
French, W.S., 1974, Two-dimensional and three-dimensional migration of model-experiment
reflection profiles: Geophysics, v. 39, p. 265-277.
Page 33
Show/Hide Bookmarks
Search
Close
Navigation
Show Hide
Gadallah, M.R., and R.L. Fisher, 2005, Applied seismology. PennWell, 473 p.
Galbraith, M., 2001, 3D seismic surveys past, present and future: CSEG Reservoir, p. 9-12,
available at: http://www.cseg.ca/publications/recorder/2001/06jun/jun01-3d-seismic.pdf.
Gendzwill, D.J., M.H. Serzu, and G.S. Lodha, 1994, High-resolution seismic reflection surveys
to detect fracture zones at the AECL Underground Research Laboratory: Canadian Journal of
Exploration Geophysis, v. 30, p. 28-38.
Hardage, B. A., 1997, Principles of onshore 3-D seismic design: The University of Texas at Austin,
Bureau of Economic Geology Geological Circular 97-5, 23 p.
Hart, B.S., 1999, Definition of subsurface stratigraphy, structure and rock properties from 3-D
seismic data: Earth-Science Reviews, v. 47, p. 189-218.
Henry, G., 1997, La sismique rflexion, principes et dveloppements. ditions Technip, 172 p.
Henry, S., 2001, Understanding the seismic wavelet, AAPG Search and Discovery Article #40028,
http://www.searchanddiscovery.net/documents/geophysical/henry/index.htm.
Hill, S.J., K. Marfurt, and S. Chopra, 2006, Searching for similarity in a slab of seismic data: The
Leading Edge, v. 25, p. 168-177.
Krey, T., 1987, Attenuation of random noise by 2D and 3D CDP stacking and Kirchhoff migration:
Geophysical Prospecting, v. 35, p. 135-147.
Lansley, R.M., 2004, CMP fold: a meaningless number?: The Leading Edge, v. 23, p. 1038-1041.
Lawton, D.C., J.H. Isaac, R.W. Vestrum, and J.M. Leslie, 2001, Slip-slidin awaysome practical
implications of seismic velocity anisotropy on depth imaging: The Leading Edge, v. 20, p. 70-73.
Leetaru, H.E., 1996, Seismic stratigraphy, a technique for improved oil recovery planning at King
Field, Jefferson County, Illinois. Illinois State Geological Survey, Illinois Petroleum 151, 37 p.
Lehmann, B., 2007, Seismic traveltime tomography for engineering and exploration applications.
EAGE Publications BV, 273 p.
Liner, C.L., 2004, Elements of 3-D Seismology. PennWell, 608 p.
Marfurt, K.J., R.M. Scheet, J.A. Sharp, and M.G. Harper, 1998, Suppression of the acquisition
footprint for seismic sequence attribute mapping: Geophysics, v. 63, p. 1024-1035.
Mosher, D.C. and P.G. Simpkin, 1999, Status and trends of marine high-resolution seismic profiling:
data acquisition, Geoscience Canada, v. 26, p. 174-188.
Nedimovi, M.R., S. Mazzotti, and R.D. Hyndman, 2003, Three-dimensional structure from
feathered two-dimensional marine seismic reflection data: The eastern Nankai Trough: Journal of
Geophysical Research, v. 108, p. doi:10.1029/2002JB001959
Ogiesoba O.C., and Hart B.S., 2009, Fault imaging in hydrothermal dolomite reservoirs: A case
study. Geophysics, v. 74 , p. B71-B82.
Page 34
Show/Hide Bookmarks
Search
Close
Navigation
Show Hide
Ray, R.R., 1995, Utilizing seismic for higher definition of geologic prospects: 2-D, 2-D swath, and
3-D seismic in High-definition seismic 2-D, 2-D swath and 3-D case histories (R.R. Ray, ed.):
Rocky Mountain Association of Geologists Guidebook, p. 1-6.
Rawlinson, N., and M. Sambridge, 2003, Seismic traveltime tomography of the crust and
lithosphere: Advances in Geophysics, v. 46, p. 81-198.
Roberts, G., K. Rutherford, and C. OBrien, 2008, Observations on the petroleum potential of deep
offshore west coast India from newly reprocessed 2D seismic data: First Break, v. 26, p. 77-86.
Schuck, A., and G. Lange, 2008, Chapter 4.6 Seismic Methods, in K. Kndel, G. Lange and H.-J.
Voigt (eds.), Environmental Geology, Springer, p. 337-402.
Seeber, M.D., and D. Steeples, 1986, Seismic data obtained using .50-caliber machine gun highresolution seismic source. AAPG Bulletin, v. 70, p. 970-976.
Shabrawi, A., A. Smart, B. Anderson, G. Rached, and A. El-Emam, 2005, How single-sensor seismic
improved image of Kuwaits Minagish Field: First Break, v. 23, p. 63-69.
Sheriff, R.E., 2002, Encyclopedic dictionary of applied geophysics. Society of Exploration
Geophysics, Geophysical References Series, 13, 429 p.
Sheriff, R.E., and L.P. Geldart, 1995, Exploration Seismology (2nd Ed.). Cambridge University
Press, 592 p.
Steeples, D.W., and R.D. Miller, 1998, Avoiding pitfalls in shallow seismic reflection surveys:
Geophysics, v. 63, p. 1213-1224.
van der Veen, M., and A.G. Green, 1998, Land streamer for shallow seismic data acquisition:
evaluation of gimbal-mounted geophones: Geophysics, v. 63, p. p. 1408-1413.
Veeken, P.C.H., 2007, Seismic stratigraphy, basin analysis and reservoir characterization: Elsevier,
509 p.
Vermeer, G., 2002, 3-D seismic survey design, SEG Geophysical References Series 12, 205 p.
Wu, J., 1992, Potential pitfalls of crooked-line seismic reflection surveys: Geophysics, v. 61, p. 227281.
Yilmaz, O., 2001, Seismic Data Analysis. SEG Investigations in Geophysics, 10, 2027 p.
Page 35