Fundamentals of 3-D seismic survey design

Gijs J.O. Vermeer

SUMMARY
Three-dimensional (3-D) seismic surveys have become a major tool in the exploration and exploitation of
hydrocarbons. The first few 3-D seismic surveys were acquired in the late 1970s, but it took until the early 1990s
before they gained general acceptance throughout the industry. Until then, the subsurface was being mapped using
two-dimensional (2-D) seismic surveys.
Theories on the best way of sampling 2-D seismic lines were not published until the late 1980s, notably by
Anstey, Ongkiehong and Askin, and Vermeer. These theories were all based on the insight that offset forms a third
dimension, for which sampling rules must be given.
The design of the first 3-D surveys was severely limited by what technology could offer. Gradually, the number
of channels that could be used increased, leading to discussions on what constitutes a good 3-D acquisition
geometry. The general philosophy was to expand lessons learnt from 2-D acquisition to 3-D. This approach led to
much emphasis on the properties of the CMP gather (or bin), because good sampling of offsets in a CMP gather was
the main criterion in 2-D design. 3-D design programs were developed, which mainly concentrated on analysis of
bin attributes, and in particular on offset sampling (regularity, effective fold, azimuth distribution, etc.).
This conventional approach to 3-D survey design is limited by an incomplete understanding of the differing
properties of the many geometries that can be used in 3-D seismic surveys. In particular, the sampling requirements
for optimal prestack imaging were not properly taken into account. This dissertation addresses these problems and
provides a new methodology for the design of 3-D seismic surveys.

The approach used in this dissertation is the same as employed in my "Seismic Wavefield Sampling", a book on
2-D seismic survey design published in 1990: before the sampling problem can be addressed, it is essential to
develop a good understanding of the continuous wavefield to be sampled. In 2-D acquisition, only a 3-D wavefield
has to be studied, consisting of temporal coordinate t, and two spatial coordinates, shot coordinate xs, and receiver
coordinate xr. In 3-D acquisition, the prestack wavefield is 5-D with two extra spatial coordinates, shot coordinate ys,
and receiver coordinate yr.
In practice, not all four spatial coordinates of the prestack wavefield can be properly sampled (proper sampling is
defined as a sampling technique which allows the faithful reconstruction of the underlying continuous wavefield).
Instead, it is possible to define three-dimensional subsets of the 5-D prestack wavefield which can be properly
sampled. In fact, the 2-D seismic line is but one example of such 3-D subsets.
The 2-D seismic line is a multi-fold data set with midpoints on a single line only. However, in 3-D acquisition
there are many possible 3-D subsets which are single-fold and whose midpoints extend across a certain area. These
subsets are called minimal data sets. A minimal data set represents a volume of data (sometimes called a 3-D cube),
which has illuminated part of the subsurface. If there was no noise, a single minimal data set would be sufficient to
create an image of the illuminated subsurface volume.
Most acquisition geometries used in practice generate data that can be considered as a collection of sampled
minimal data sets. Therefore, the properties of the minimal data sets need to be studied for a better understanding of
the acquisition geometries as a whole. This allows an optimal choice of the acquisition geometry (if there is a
choice, often the geometry type is dictated by economic or environmental constraints) and of the parameters of the
geometry.
The continuous wavefield to be sampled can be reduced to the wavefield of the characteristic minimal data set of
the chosen geometry. Proper sampling of that wavefield means that at least two of the four spatial coordinates of the
5-D prestack wavefield will be properly sampled. Next, it is recommended to maximize the useful extent of each
minimal data set. Together, these two recommendations ensure a minimum of spatial discontinuities in the total data
set. Spatial continuity is maximized and the migrated minimal data sets contain a minimum of artifacts. Other
parameters of each acquisition geometry need to be chosen so that requirements of resolution, noise suppression and
illumination are satisfied as well.
2

Based on these basic principles, this dissertation addresses a wide variety of issues. It starts with a short
summary of 2-D symmetric sampling, which is a recipe for optimal sampling of the 2-D seismic line. 2-D symmetric
sampling is based on a corollary of the reciprocity theorem, which affirms that the properties of the common-
receiver gather are the same as the properties of the common-shot gather. As a consequence, sampling requirements
of shots and receivers are identical.
3-D seismic surveys can be acquired using a number of different acquisition geometries. The most important
geometries are areal geometry, parallel geometry and orthogonal geometry. Each geometry has its characteristic 3-D
basic subset. If the basic subset is single fold, it is also a minimal data set. In areal geometry either shots or receivers
are acquired in a dense areal grid. If shots are dense, receivers are sparse or vice versa. In the first case 3-D
common-receiver gathers are acquired. These gathers form the basic subset or minimal data set of this particular
areal geometry.
Parallel geometry and orthogonal geometry are examples of line geometries, in which sources and receivers are
arranged along straight acquisition lines, which are more or less widely separated. In parallel geometry the (parallel)
shot lines are parallel to the (parallel) receiver lines, whereas in orthogonal geometry shot and receiver lines are
orthogonal. The basic subset of parallel geometry is the midpoint line, which runs halfway between the shot line and
each active receiver line. The basic subset of orthogonal geometry is the cross-spread, which encompasses all
receivers in a single receiver line which are listening to a range of shots in a single shot line (Figure 1). The cross-
spread is a minimal data set with limited extent. The difference in properties of the various acquisition geometries is
illustrated by the difference in diffraction traveltime surface of the same diffractor for the basic subsets of those
geometries.

shots for template

/ cross-spread

maximum cross-
line offset

receivers for
template / cross-
spread
midpoint coverage
template / cross-
spread
maximum in-line
offset
Fig. 1. The same orthogonal geometry with template (left) and with cross-spread (right). Horizontal lines are
receiver lines; vertical lines are source lines. The template represents the way in which the data is acquired in
the field; in this case there are 8 receiver lines with a number of shots in the center of the template. The cross-
spread gathers all data for receivers that have listened to a range of shots along the same source line.

2-D symmetric sampling can be readily expanded to 3-D symmetric sampling after recognition of the existence
of the basic subsets of each geometry.
For imaging, it would be ideal to have single-fold data sets that extend across the whole survey area, but which
possess a minimum of spatial discontinuities so that they would produce a minimum amount of migration artifacts.
These data sets are called pseudo-minimal data sets and can be constructed from so-called offset-vector tiles. In
orthogonal geometry the size of the offset-vector tile is determined by the area between two adjacent shot lines and
two adjacent receiver lines. The cross-spread can be split into M disjoint offset-vector tiles (OVTs, M is fold-of-
coverage), in which the x- and the y-component of the offset vector vary over a limited range (Figure 2).
Sampling in 3-D acquisition is usually not dense enough to record low-velocity noise without aliasing. To reduce
aliasing effects, shot and receiver arrays may be used. The arrays may be linear or areal. For a proper choice of
arrays, the properties of the noise need be known. An analysis of the energy distribution of low-velocity scatterers
shows that in the cross-spread most energy is concentrated on the flanks of the traveltime surface and there is less
energy around the apex. Linear arrays are sufficient to suppress the energy in the flanks. If there is much undesirable
energy coming from all directions, circular arrays can be constructed with a circular response.
 3

Fig. 3. Offset-range gather in orthogonal geometry.

Each ring represents traces in midpoint domain
with a narrow range of absolute offsets.
Fig. 2. Cross-spread with its OVTs. For each OVT, Often, one of the aims in 3-D survey design is to
offset and azimuth of the central trace are achieve a regular offset distribution. This is based on
indicated. Anstey's stack-array approach for 2-D data, which states
that ground roll is best suppressed in the stack by a regular
sampling of offsets in each CMP gather. However, this requires high-fold data; if the data are low fold, a random
offset distribution tends to have a better stack response. This applies in particular to 3-D data where fold tends to be
low, especially if measured in separate azimuth ranges. A wide orthogonal geometry (maximum cross-line offset
close to maximum in-line offset) tends to produce irregular offset sampling in each CMP gather, hence tends to have
a better stack response than a narrow geometry.

The theoretical considerations and observations of the first part of the dissertation are translated into practical
guidelines for choice of geometry and selection of parameters for orthogonal geometry. Parallel geometry looks
most like 2-D acquisition, the stack response is similar, and processing can use many of the techniques already
developed for 2-D processing. It is not suitable for analysis of azimuth-dependent effects. Parallel geometry can be
acquired efficiently in marine environment using streamers, but on land parallel geometry is less efficient than
orthogonal geometry. Orthogonal geometry is suitable for analysis of azimuth-dependent effects. It is also used for
sea-bed acquisition using bottom cables. Processing of orthogonal geometry is much more complex than processing
of parallel geometry. For instance, offset-range gathers have very irregular fold-of-coverage (Figure 3). Zigzag
geometry is a geometry devised for efficient acquisition in desert environments. Slanted geometry is similar to
orthogonal but the shot lines cross the receiver lines at an oblique angle. The basic subsets of zigzag and slanted
geometry are less suitable for dual-domain processing than the basic subset of orthogonal geometry. Areal geometry
is also suitable for analysis of azimuth-dependent effects. It is applied mainly in deep waters in case very expensive
receiver units are used, such as vertical-hydrophone cables and 4-C receiver units (3-component geophone plus
hydrophone).
The main parameters of orthogonal geometry are station intervals, line intervals, maximum in-line and cross-line
offsets, and fold. These have to be selected such that requirements of spatial continuity, resolution, mapping of
shallowest and deepest horizons of interest, and noise suppression are satisfied. The survey area is always larger
than the area to be mapped due to the fold-taper zone and the radius of the migration operator.
In practice, a one-line roll of a nearly square template tends to be quite inefficient. Without compromising the
desired acquisition geometry, it is often more efficient to use a full-swath roll. A multi-line roll is also more
efficient, but it will create strong spatial discontinuities.
Obstacles often prevent acquisition of straight acquisition lines. Spatial continuity then requires the acquisition
lines to be smooth. Common practice of moving shots an integer multiple of the receiver interval to the right or to
the left produces discontinuities in the receiver gathers leading to migration artifacts (Figure 4). A general
requirement in acquisition of parallel and orthogonal geometry is that the receiver gathers should look as good as the
shot gathers.
4

x shotline x shotline
x x
x x
x x
offset shots x offset shots x
x x
obstacle obstacle
x x
x x
x x
x x
x receiver line x receiver line
o o o o o o o o o o o o o o o o o o o o o o
x x
x x
x x
x x
(a) (b)

Fig. 4. Procedures for dealing with obstacles. (a) Midpoint-centering solution, (b) Smooth solution. The
smooth solution preserves spatial continuity.

In marine seismic data acquisition, the designer has to choose between streamers and stationary-receiver
systems. With streamers, multisource multistreamer configurations are used in a parallel geometry. With stationary-
receiver systems there is flexibility in the choice of geometry. Streamers are most efficient in deep water without
any obstacles. Adjacent boat passes should be acquired antiparallel to minimize illumination irregularities (Figure
5). However, illumination irregularities caused by differential feathering are inevitable. Acquisition with stationary-

(a) (b)
Fig. 5. Two adjacent boat passes in 2/8 geometry. Solid lines indicate streamers, stipled lines indicate source
tracks. Four of the eight streamers follow the same track in the two boat passes to achieve single cross-line
fold. Lines connecting sources with the farthest receiver groups indicate the shot-to-receiver offsets for the
outer midpoint lines of each boat pass. The vertical row of dots indicates midpoint positions. (a) Parallel
acquisition. The adjacent midpoints in the center of the picture have opposite cross-line offset, hence different
shot-to-receiver azimuths. (b) Antiparallel acquisition. The adjacent midpoints in the center of the picture
have opposite cross-line offset and opposite in-line offsets, hence identical shot-to-receiver azimuths.
 5

receiver systems tends to be more expensive than with streamers. Systems in use are vertical hydrophone cable,
ocean-bottom cable and node. Nodes are single 4-C units, whereas ocean-bottom cables can be used with a dual-
sensor technique as well as a 4-C technique. Repeatability of stationary-receiver systems is better than repeatability
of streamer acquisition.
Survey design for PS-waves is different from P-wave acquisition, owing to the asymmetry of the PS ray path.
Differences in PS-illumination by minimal data sets of different geometries are much larger than P-illumination
differences. For instance, a cross-spread with a square midpoint area produces an illumination area with rectangular
shape even for a horizontal reflector. The ray-path asymmetry leads to asymmetric sampling requirements for shots
and receivers. Shot sampling interval is determined by P-wave velocity; receiver sampling interval by S-wave
velocity. Parallel geometry tends to suffer least from asymmetry effects whereas orthogonal geometry tends to suffer
most. For analysis of azimuth-dependent effects, areal geometry might be the best choice.
Noise spreads or microspreads are acquired with very dense spatial sampling for an analysis of low-velocity
events. A cross-spread with very dense spatial sampling was acquired in The Netherlands. Timeslices and cross-
sections illustrate the 3-D behavior of the ground-roll cone and of the scatterers inside the cone.
The theory of 3-D symmetric sampling was put to the test in Nigeria, where a cross-spread test geometry was
compared with the standard brick-wall geometry. The test geometry produced better results (higher resolution and
better continuity) at target level than the standard geometry. The improvement can be attributed to larger width
(maximum cross-line offset) of the test geometry and (most likely) to its better spatial continuity.
The same Nigerian data set is used to demonstrate that under favorable circumstances very low fold can be
sufficient to get acceptable 3-D prestack migration results (Figure 6).
The minimal data sets of the various acquisition geometries also have different resolution properties. The main
factor influencing the theoretically best resolution is the stretch effect caused by normal moveout. Therefore, zero-
offset data have potentially best resolution. Resolution is not improved by reducing the midpoint sampling intervals
while keeping the shot and receiver sampling intervals the same (bin-fractionation technique). Carefully selected
"random" coarse sampling may produce less migration artifacts than regular coarse sampling, but in order to
eliminate all artifacts, regular dense sampling is best.
The theory of DMO correction was developed for 2-D common-offset gathers. Initially, the success of
application of DMO correction to 3-D data was not really understood. The theory of DMO application to minimal
data sets in general and to cross-spreads in particular dispelled the mystery. The application of existing DMO
software to a single-fold data set (cross-spread) revealed serious amplitude and phase artifacts. This prompted
improvements in contractor software.
The required migration radius is often described in
terms of Fresnel zone radius. However, the Fresnel zone
radius for broadband data is not large enough for complete
imaging. It is better to define the zone of influence for
Energy

Migration radius

Fresnel zone

Zone of influence
Fig. 7. Fresnel zone and zone of influence. A
horizontal event is shown after flattening of the
diffraction traveltime. Start and end of the wavelets
(a) (b) is indicated by drawn curved lines. Data beyond the
Fig. 6. Comparison migration result with stacked zone of influence cannot contribute to the migration
data. (a) Migration result from 5 partially result in the point of stationary phase. Data beyond
overlapping cross-spreads. (b) Corresponding stack the Fresnel zone is still needed for a complete image
(mostly 2-fold). with correct phase and amplitude.
6

migration (in analogy to what previously has been done for modeling) and to use the radius of that zone in
establishing migration-apron requirements (Figure 7). Most minimal data sets have limited extent leading to edge
effects in migration. However, using pseudo-minimal data sets constructed from offset-vector tiles tends to produce
better single-fold images than other single-fold subsets of the geometry.

The ideas and results discussed in this dissertation should help to achieve a better understanding of the structure
of 3-D acquisition geometries. With this understanding, geophysical requirements can be satisfied with an optimal
choice of acquisition geometry and its parameters. Processing techniques can be adapted to honor and exploit the
specific requirements of each geometry, especially orthogonal and areal geometry, leading to a more interpretable
end product.