Professional Documents
Culture Documents
Chapter 1
2-D symmetric sampling
This chapter is modified from Vermeer (1991). FIG. 1.1. Prestack data coordinate systems. (a) The four
1
With this notation, the offset vector points from receiver spatial coordinates in relation to the seismic line. (b)
to shot. In the next chapter the more logical notation is used Shot/receiver coordinate system. (c) Midpoint/offset coor-
Downloaded 27 Feb 2012 to 198.3.68.20. Redistribution subject to SEG license or copyright; Terms of Use: http://segdl.org/
in which the offset vector points from shot to receiver. dinate system.
5
Chapter 01_v2.qxd 8/28/02 1:37 PM Page 6
shot/receiver coordinate system and by an oblique line in in the 3-D space of the prestack seismic data set. The
the midpoint/offset coordinate system. three dimensions of the prestack data should not be con-
By keeping the time coordinate constant, a time slice fused with the three dimensions of the subsurface. In
is generated from the prestack seismic data set. In a time prestack data, offset is the third dimension. Take, for ex-
slice, the spatial coordinates vary so that the surface and ample, the reflection traveltime surface of a dipping
subsurface diagrams could also be regarded as a descrip- plane as shown in Figure 1.3. This figure illustrates the
tion of the data points in a time slice. three orthogonal cross-sections: common midpoint,
We are inclined to think of reflections in prestack data common offset, and common time. The shape of the trav-
as hyperbolas in the common midpoints. However, it is eltime surface is a hyperbola in the common midpoint
important to realize that each event represents a surface and an ellipse in the time slice.
xs
xs = C
xr = C
(xr + xs )/2 = C
xs xr = C
xr
xo xm + / xo = C
1
2
xm / xo = C
1
2
xm = C
xo = C
xm
FIG. 1.2. Descriptions of prestack seismic data set in (a) a shot/receiver coordinate system and (b) a midpoint/offset coor-
dinate system. The former is also called a surface diagram or surface stacking chart and the latter a subsurface diagram or
Downloaded 27 Feb 2012 to 198.3.68.20. Redistribution subject to SEG license or copyright; Terms of Use: http://segdl.org/
subsurface stacking chart.
Chapter 01_v2.qxd 8/28/02 1:38 PM Page 7
W(t, ks , kr ) = s
W(t, xs , xr ) exp [2i(ks xs + kr xr )]dxs dxr , (1.2)
A field data example is given in Figure 1.4. Of course, where s is the range of shots and receivers included in
now there is a multitude of events, all having their own the integration. Equation (1.2) represents the double
spatial and temporal relationships. Actually, the com- wavenumber spectrum of a time slice, whereas the triple
mon-offset gather in this example is a stack, which is ba- Fourier transform of W(t, xs , xr ) can be written as
sically a zero-offset section with a relatively high signal-
to-noise ratio. It is possible to follow reflections from
dipping layer boundaries through all three cross-sec- midpoint
time xm
tions.
et
t
xo offs
Creating time slices from the prestack data of a 2-D
line can be a very rewarding exercise. Time slices in-
crease insight in the characteristics of the data and allow
useful diagnostics-at-a-glance of the whole data set.
Time slices created after NMO correction allow a quick
quality control of the chosen velocities for the level of in-
terest.
For proper sampling of the temporal coordinate, it is
important to know the maximum frequency of the data to
be sampled. Likewise, for spatial sampling, the maxi-
mum wavenumber of the spatial coordinates must be
known. A discussion of spatial sampling requires the in-
troduction of four different wavenumbers (ks , kr , km , ko ) FIG. 1.3. Dipping event in a midpoint/offset coordinate sys-
corresponding to the four spatial coordinates (xs , xr , xm , tem. The event is a hyperbola in the common-offset pan-
xo ). For instance, ks and kr can be defined by the forward els (a straight line for zero offset), a hyperbola in the CMP,
Fourier transform and an ellipse in the time slice.
xo xm
xo
Downloaded 27 Feb 2012 to 198.3.68.20. Redistribution subject to SEG license or copyright; Terms of Use: http://segdl.org/
FIG. 1.4. Three cross-sections through a prestack data set. Note that each event is a surface in 3-D (t, xm , xo) space.
Chapter 01_v2.qxd 8/28/02 1:38 PM Page 8
W(f, ks , kr ) = t
W(t, ks , kr )exp[-2ift]dt, (1.3) xs
Symmetric sampling 9
curate reconstruction of the underlying continuous wave- and the signal may lose some of its high-frequency con-
field, especially close to Nyquist (Niland, 1989). tent. Some further remarks on array length are made in
The basic sampling intervals are much smaller than Section 1.6.
considered practical or affordable (e.g., for fmax = 75 Hz, Figure 1.9 illustrates the concept of symmetric sam-
and a ground-roll velocity Vmin = 300 m/s, the shot and re- pling. (In these figures the xs, xr coordinate system has
ceiver intervals should be 2 m). As a compromise, seis- been rotated 45 for ease of display.) To keep the pic-
mic field arrays are to be used which act as spatial antialias tures simple, I used only three elements per array (when
filters and as resampling operators. As resampling opera- there is an array). Figure 1.9a shows an asymmetric con-
tors they allow the use of more affordable shot and receiv- figuration with no shot array and three geophones per
er intervals. As spatial antialias filters, they aim to attenu- geophone array. Each recorded trace is the sum of three
ate all energy above the Nyquist wavenumber. Spatial an- elemental traces registered by the three geophones of
tialias filtering must be applied when sampling both spa- the geophone array. The groups of three elemental traces
tial coordinates. In other words, shot arrays are as neces- are represented by alternating between three open cir-
sary as receiver arrays and, for optimal results, shot arrays cles and three closed circles. Note that the three ele-
should be identical to receiver arrays. This reasoning leads
to the concept of symmetric sampling as a prerequisite of ks,r
consistent data handling:
Shot interval equal to receiver interval kmax
Shot arrays equal to receiver arrays
I call this technique symmetric sampling because it
utilizes the symmetry property of reciprocity and it pre-
serves the inherent symmetry of the prestack wavefield.
A consequence of symmetric sampling is that there will
be as many traces in the common-receiver gather as in max
the common-shot gather.
In case of a linear nonweighted array with equal in-
tervals between array elements, the dimension of the
array should be such as to achieve a regular uninterrupt-
ed sampling by the array elements of the whole length of FIG. 1.7. Regions with and without energy in (f,ks ) and
the receiver line. Basically, this means that the length of (f,kr ). Maximum frequency of event with minimum-phase
the array should be equal to the station interval; if short- velocity determines maximum wavenumber and it is the
same for shot and receiver coordinates.
er, the wavefield is undersampled, and if longer, the
wavefield is oversampled, intra-array statics are larger,
ks
common receiver
at xr = xp
xs kr
0
=
xo
xp common shot
at xs = xp
xp xr
xs
xr
a)
xs
xr
b)
xs
xr
c)
FIG. 1.9. Various center-spread shooting geometries with three elements per array, when there is an array. Each symbol rep-
resents an elemental shot/receiver pair; each group of equal symbols (either open circles or closed circles) represents one
recorded trace. (a) Asymmetric configuration with shot interval equal to receiver station interval, and a geophone array but
no shot array. (b) Symmetric configuration with shot interval equal to receiver-station interval, and both geophone and shot
arrays. (c) Asymmetric configuration with a shot interval three times the receiver-station interval. Note the large unsampled
area of shot/receiver space.
Downloaded 27 Feb 2012 to 198.3.68.20. Redistribution subject to SEG license or copyright; Terms of Use: http://segdl.org/
Chapter 01_v2.qxd 8/28/02 1:38 PM Page 11
mental traces for a given recorded trace have different sponding sampling interval is called basic signal sam-
shot-to-geophone offsets and different midpoint posi- pling interval.
tions. Summing these elemental traces causes some
damage to the signal, but this is the price to be paid for 1.4 Symmetric sampling versus
the antialias effect of the geophone array. In Figure 1.9b,
asymmetric sampling
a symmetric sampling configuration is shown. Now
there is also a shot array consisting of three elements so Having established that symmetric sampling is neces-
that each recorded trace corresponds to nine elemental sary to honor the properties of the prestack wavefield, I
traces. Note that, again, each of the elemental traces oc- shall now discuss some effects of asymmetric sampling
cupies a different position in the shot/receiver coordi- and enumerate advantages of symmetric sampling.
nate system. Together, all the elemental traces provide a The effects of asymmetric sampling are different for
regular two-dimensional sampling of the shot/receiver end-on shooting and center-spread shooting. For the for-
plane. Compare this with Figure 1.9a, where whole mer, asymmetric sampling leads to differences between
areas of the plane are not sampled. These empty areas updip and downdip shooting. Figure 1.11 illustrates there
may lead to spatial aliasing in the common-receiver do- is less difference between the arrival times of the reflec-
main and also in the common-midpoint and common- tions over the length of the receiver array for updip
offset domains. shooting than there is for downdip shooting. Therefore,
Another, perhaps even more common, example of with asymmetric sampling (for instance, shot array being
asymmetric sampling is shown in Figure 1.9c. Now the different from receiver array), the reflection character is
shot interval is three times the group interval. Here even less affected by updip shooting than by downdip shoot-
more of the shot/receiver plane is not sampled. ing. This effect would only be visible by careful inspec-
An interesting illustration of the need for arrays with tion of two neighboring parallel lines acquired in oppo-
length equal to the station interval is given as Figure 3 in
Newman (2000) and reproduced here as Figure 1.10. The
shorter length arrays used on the left have a wide pass-
band in the wavenumber domain and do not suppress
much of the aliased ground-roll energy. The arrays with
length equal to station interval used on the right have
suppressed the ground-roll energy better. There is still re-
maining ground-roll energy which exhibits an odd/even
effect (odd traces look more similar to each other than to
even traces) in Figure 1.10b, because the first notch of
the array response occurs at twice the Nyquist wavenum-
ber kN corresponding to the station spacing [cf. discus-
sion in Section 1.6 and equations (1.5) and (1.8)]. How-
ever, if necessary, the energy passed above kN may be
further suppressed by a two-trace running mix (a convo-
lution with a two-point spatial filter with equal coeffi-
cients) in processing.
Whether or not spatial aliasing occurs for a particular
shooting geometry (symmetric or asymmetric) and how
large the effect is depends on the distance between the
array elements and on the shot and receiver intervals.
Symmetrically sampled data may still be aliased if the
sampling intervals are too large, and asymmetrically
sampled data may not show aliasing if the sampling in-
tervals are small enough. A nice compromise to aim for
FIG. 1.10. (a) Three-D spread of 60 channels of six phones
is to use shot and receiver intervals that would record the over 100 ft. Deep hole dynamite source (15 pounds at 60
desired wavefield without aliasing up to the frequency of ft). (b) Parallel line [1320 ft away from line in (a)] of 60
interest. Then arrays are only necessary to suppress noise channels with 24 phones laid out over 220 ft from the
and to average out sampling irregularities. This tech- same shot. The source is between the two lines (from
Downloaded 27 Feb 2012 to 198.3.68.20. Redistribution subject to SEG license or copyright; Terms of Use: http://segdl.org/
nique is called full-resolution recording and the corre- Newman, 2000).
Chapter 01_v2.qxd 8/28/02 1:38 PM Page 12
site directions. On the other hand, in a single 2-D line, cussed earlier, these lines are ellipses.) The lines are
asymmetric sampling will lead to asymmetries on either symmetric with respect to the diagonal, which is the
side of the apex of diffractions. zero-offset line. The rectangles represent traces of one
In center-spread shooting, the effect of asymmetric common midpoint with each trace formed by a 25 m shot
sampling is less visible in the stack, as the asymmetry of array and a 75 m receiver array. In the top left corner, the
the sampling is hidden by the symmetry of the spread. rectangle averages across the time lines; in the lower
However, the differences between updip and downdip right corner, the rectangles run more or less parallel to
shooting, as discussed for end-on shooting, now occur in the time lines. This difference in averaging leads to a dif-
the recording of one and the same line. Now the effect ferent character between positive and negative offsets.
becomes visible in the common-midpoint gathers. I have Similar effects can be observed with single-hole dyna-
simulated the effect in the example shown in Figure 1.12. mite shooting. I am convinced that many seismic proces-
Figure 1.12a is a CMP with equal shot and receiver ar- sors have noticed those effects. Obviously, it will lead to
rays, whereas in Figure 1.12b the receiver arrays are a suboptimal stack for center-spread shooting.
three times as long as the shot arrays (75 m versus 25 m). The severity of asymmetric sampling depends on a
I constructed the right panel using a three-trace running number of factors, such as spatial sampling intervals, de-
mix in the common-shot gathers, followed by CMP sort. gree of asymmetry, relative strength of coherent noise,
In both CMPs, the traces are sorted according to increas- dip (stronger dip, larger effects), and geologic complexi-
ing absolute offset. The right panel now shows jitter in ty. With some further analysis, some of those effects may
many reflection events. The jitter occurs for events that be quantified (for instance, the increasing severity with
dip in the common-offset gather. The explanation of the increasing dip).
jitter follows from the difference in averaging effects of On the other hand, symmetric sampling has numerous
the arrays on either side of zero offset. advantages:
This averaging effect is illustrated in Figure 1.13 in
Symmetry of wavefield is preserved
which the two curved lines represent constant time lines
of a dipping event in the shot/receiver plane. (As dis-
a) b)
m offset range, hence the first notches of the three arrays certainly not uniform although it is symmetric. There are
occur at ks = 1/50, kr = 1/50, and k0 = 1/1200 m-1. [This areas of very good suppression where all three arrays are
discussion does not take NMO into account; see effective, and there are also areas of less good suppres-
Ongkiehong and Askin (1988) for the effect of NMO on sion. (The parameters of this example should not be
the array responses.] taken as recommended symmetric sampling field
In displays of array and stack responses usually the parameters; usually, smaller intervals are necessary for
absolute value of equation (1.8) is taken, thus neglecting good results.)
the phase of the response. For a 2-D seismic line, the off- How much unwanted energy is left after application
set distribution tends to be regular with an equal number of the three arrays depends on
of traces and a constant offset interval between traces in
all midpoints, but with a slightly shifted range of offsets energy distribution of the prestack wavefield,
between neighboring midpoints. For such data the ab- choice of field parameters (shot and receiver inter-
solute value of the stack response is the same for all mid- val, and fold), and
points, whereas the range shift leads to a phase variation choice of pre- and poststack processing parame-
in p(k0 ) causing the checkerboard or odd/even effect ters.
(Vermeer, 1990, Section 5.11.4). Leaving out the shot array has a dramatic effect on the
Although the stack response is formulated mathemat- total stack response (as shown in Figure 1.16). The sever-
ically in the same way as the array response, the desired ity of not using a shot array or any other form of asym-
response is quite different. The ideal array response metric sampling depends on the energy distribution of
looks like an antialias filter response with a passband and the original continuous wavefield in the (km , k0 )-
a cut-off wavenumber. The ideal stack response passes wavenumber domain. If there are many rapid variations
all energy at k0 = 0, and rejects all energy with k0 0. as a function of midpoint, asymmetric sampling will do
Figures 1.14, 15, and 16 represent displays of equa- more harm than if the geology varied more slowly.
tion (1.6) in the midpoint/offset wavenumber domain for Finer sampling (shorter shot and receiver intervals
different situations. Figure 1.14 shows the response of a with array lengths equal to those intervals) pushes the fil-
50-m shot array combined with a 50-m receiver array ter notches out toward larger wavenumbers. As a conse-
and no CMP array. Lines of constant shot and receiver quence, a larger part of the original wavefield will fall in
wavenumber run obliquely in the midpoint/offset the passband of the combined field arrays. In the pass-
wavenumber domain. Note the diamond-shaped central band, more of the suppression of the unwanted events is
passband of the two arrays in this domain. then left to the stack and to other digital processes. Dig-
A common simplification is to compute the product of ital processes such as (f,k)-filtering are usually required
the two array responses as a function of only one to compensate for the reduced effect of the two field ar-
wavenumber. This product describes the effect of the ar- rays.
rays on a horizontal earth with no midpoint dependence.
The horizontal earth response is found for km = 0, i.e.,
along the vertical axis of Figure 1.14. However, any dip- 1.7 Concluding remarks
ping events will contain energy away from the vertical This chapter serves as a summary of 2-D symmetric
axis and will be affected differently by the field arrays. sampling as described in more detail in Vermeer (1990).
So, the correct representation uses the double wavenum- Symmetric sampling is the preferred recording technique
ber domain. for 2-D seismic surveys, so it should be high on the
Figure 1.15 shows the total stack response for a sym- wish list of every interpreter. In case an asymmetric
metric sampling technique with center-spread shooting technique has been used in the past, for instance with a
and an offset range -1200 to 1200 m. Taking reciprocity shot interval that is larger than the receiver interval, there
into account, this configuration effectively produces a is always scope for improvement by repeating the survey
50-m offset interval in each CMP leading to a first alias with symmetric sampling parameters.
in the stack response at k0 = 1/50 m-1. Note that the Yet, it should be realized that the parameters of sym-
stack produces notches parallel to the horizontal km-axis. metric sampling are still a compromise and need to be
The diamond-shaped passband of the field arrays has established after an evaluation of the geologic and geo-
now been reduced to a narrow passband centered on the physical problems at hand. In particular, time and again
midpoint wavenumber axis. Everywhere else the combi- improvements have been achieved by using smaller sta-
nation of field arrays and stack is supposed to suppress tion spacings leading to noisier field records, but allow-
Downloaded 27 Feb 2012 to 198.3.68.20. Redistribution subject to SEG license or copyright; Terms of Use: http://segdl.org/
all energy. As is clear from the picture, the suppression is ing better noise removal in processing.
Chapter 01_v2.qxd 8/28/02 1:41 PM Page 15
Concluding remarks 15
0.03
ko in m-1
0.03
0.06 0 0.06
km in m-1
FIG. 1.14. Combined response of shot and receiver arrays in midpoint/offset wavenumber domain. Oblique dark blue lines
represent notches in the shot-array and receiver-array responses.
0.03
ko in m-1
0.03
0.06 0 0.06
km in m-1
FIG. 1.15. Total stack response for a symmetric sampling technique. The notches of the CMP array run parallel to the km
axis.
0.03
ko in m-1
0.03
0.06
Downloaded 0 or copyright; Terms of Use: http://segdl.org/
27 Feb 2012 to 198.3.68.20. Redistribution subject to SEG license 0.06
km in m-1
FIG. 1.16. Total stack response for an asymmetric sampling technique without a shot array.
Chapter 01_v2.qxd 11/29/05 7:17 AM Page 16
A better understanding and knowledge of the energy suppression by the stackarray: Geophysics, 54,
distribution in (f, km , k0 ) would help predict the effect of 290301.
any choice of the acquisition parameters. Noise spreads Newman, B. J., 2000, Spatial aliasing and 3-D bin size:
and very densely sampled multiple-coverage data can be The quest for cleaner, cheaper data: The Leading
used to help gain such information. Edge, 19, 158161.
Niland, R. A., 1989, Optimum oversampling: J. Acoust.
References Soc. Am., 86, 18051811.
Ongkiehong, L., and Askin, H. J., 1988, Towards the uni-
Ak, M. A., 1990, How effective is the stack-array?: Pre- versal seismic acquisition technique: First Break,
sented at the 52nd Eur. Assn. Expl. Geophys. Con- 6, 4663.
ference. Taner, M. T., Koehler, F., and Alhilali, K. A., 1974, Esti-
Anstey, N., 1986, Whatever happened to ground roll?: mation and correction of near-surface time anom-
The Leading Edge, 5, No. 3, 4045. alies: Geophysics, 39, 441463.
Fokkema, J. T., and van den Berg, P. M., 1993, Seismic Vermeer, G. J. O., 1990, Seismic wavefield sampling:
applications of acoustic reciprocity: Elsevier Sci- Soc. Expl. Geophys.
ence Publ. Co., Inc. 1991, Symmetric sampling: The Leading Edge,
Morse, P. F., and Hildebrandt, G. F., 1989, Ground-roll 10, No. 11, 2127.
Downloaded 27 Feb 2012 to 198.3.68.20. Redistribution subject to SEG license or copyright; Terms of Use: http://segdl.org/