3D Seismic Survey Design

Theres more to designing a seismic survey than just choosing sources and receivers and shooting away. To
get the best signal at the lowest cost, geophysicists are tapping an arsenal of technology from integration of
borehole data to survey simulation in 3D.
C. Peter Ashton
Mrsk Olie og Gas AS
Copenhagen, Denmark

Christian Dplant
Elf Aquitaine
Pau, France

nCost of marine 3D

40

seismic surveys for

one oil company.
Since 1990, the cost
of a marine 3D survey has decreased
by more than 50%.

35

Dollars, in thousands

Brad Bacon
Angus Mann
Nick Moldoveanu
Houston, Texas, USA

Cost of Marine 3D Seismic Survey per km2

30
25

(Courtesy of Ian Jack,

BP Exploration, Stockley Park, England.)

20
15
10
5

DickiIreson
Thor Sinclair
Gatwick, England
Glen Redekop
Maersk Oil Qatar AS
Doha, Qatar
For help in preparation of this article, thanks to Jack
Caldwell and Greg Leriger, Geco-Prakla, Houston, USA;
Mandy Coxon and Dominique Pajot, Geco-Prakla,
Gatwick, England; Jacques Estival, Elf Petroleum Nigeria,
Lagos, Nigeria; Dietmar Kluge, Geco-Prakla, Hannover,
Germany; Lloyd Peardon, Schlumberger Cambridge
Research, England; Lars Sonneland, Geco-Prakla,
Stavanger, Norway; and Tim Spencer, British Gas,
Reading, England.
Appreciation is expressed to Qatar General Petroleum
Corporation (QGPC) for its consent to the release of data.
QUAD-QUAD is a mark of Geco-Prakla. TWST
(Through-Tubing Well Seismic Tool) is a mark of Schlumberger.
1. For the most recent worldwide figures:
Riley DC: Special Report Geophysical Activity in
1991, The Leading Edge 12, no. 11 (November
1993): 1094-1117.
2. Personal communication: Thor Sinclair.

April 1994

0
1990

1991

1992

1993

Year

Increased efficiency has brought the cost of

marine three-dimensional (3D) seismic data
to its lowest level ever, expanding the popularity of 3D surveys (above ). In the past five
years, oil companies have increased expenditures on seismic surveys by almost 60%,
to $2.2 billion. 1 However, an estimated
10% of surveys fail to achieve their primary
objectivesome because the technology
does not exist to process the data, some
because the surveys are improperly
planned. 2 Careful planning can result in
more cost-effective acquisition and processing, and in data of sufficient quality to benefit from the most advanced processing.
But before the first shot is fired or the first
trace recorded, survey designers must determine the best way to reveal the subsurface
target. As basics, they consider locations
and types of sources and receivers, and the
time and labor required for acquisition.
Many additional factors, including health,
safety and environmental issues, must be

taken into account. This article investigates

the objectives and methods of seismic survey design and reviews field examples of
state-of-the-art techniques.
The ideal 3D survey serves multiple purposes. Initially, the data may be used to
enhance a structural interpretation based on
two-dimensional (2D) data, yielding new
drilling locations. Later in the life of a field,
seismic data may be revisited to answer
questions about fine-scale reservoir architecture or fluid contacts, or may be compared
with a later monitor survey to infer fluid-front
movement. All these stages of interpretation
rely on satisfactory processing, which in turn
relies on adequate seismic signal to process.
The greatest processing in the world cannot
fix flawed signal acquisition.

19

Temporal Aliasing
50 Hz

nTemporal and

200 Hz

16

24

spatial aliasing
caused by sampling less than
twice per cycle.
Temporal aliasing
(top) occurs when
insufficient sampling renders a 50Hz signal and a
200-Hz signal indistinguishable
(arrows represent
sample points). The
50-Hz signal is adequately sampled,
but not the 200-Hz.

32

Time, msec

Extreme Aliasing

(Adapted from Sheriff,

reference 4.) Spatial

Two-way time

Minor Aliasing

Increasing offset

Increasing offset

aliasing (bottom)
occurs when
receiver spacing is
more than half the
spatial wavelength.
With minor aliasing
(left) arrivals can be
tracked at near offsets as time
increases, but
become difficult to
follow at far offsets.
With extreme aliasing (right) arrivals
even appear to be
traveling backwards, toward near
offsets as time
increases. (Adapted
from Claerbout, reference 6.)

Two-way time

Offset

Offset

Hyperbolic
moveout

+ Stacking
velocity =
CMP gather

Corrected
CMP gather

=
Stacked
CMP trace

nBetter stacking from a wide and evenly spaced set of offsets.

Reflection arrival times from different offsets are assumed to follow a hyperbola. The shape of the hyperbola is computed from
the arrivals. Traces are aligned by flattening the best-fitting
hyperbola into a straight line, then summed, or stacked. Perfect
alignment should yield maximum signal amplitude at the time
corresponding to zero offset. A wide range of evenly spaced offsets gives a better-fitting hyperbola, and so a better stack.

20

Elements of a Good Signal

What makes a good seismic signal? Processing specialists list three vital requirementsgood signal-to-noise ratio (S/N),
high resolving power and adequate spatial
coverage of the target. These basic elements,
along with some geophysical guidelines (see
Guidelines from Geophysics, page 22 ),
form the foundation of survey design.
High S/N means the seismic trace has
high amplitudes at times that correspond to
reflections, and little or no amplitude at
other times. During acquisition, high S/N is
achieved by maximizing signal with a seismic source of sufficient power and directivity, and by minimizing noise.3 Noise can
either be generated by the sourceshotgenerated or coherent noise, sometimes
orders of magnitude stronger than deep seismic reflectionsor be random. Limitations
in the dynamic range of acquisition equipment require that shot-generated noise be
minimized with proper source and receiver
geometry. Proper geometry avoids spatial
aliasing of the signal, attenuates noise and
obtains signals that can benefit from subsequent processing. Aliasing is the ambiguity
that arises when a signal is sampled less
than twice per cycle (left ). Noise and signal
cannot be distinguished when their sampling is aliased.
A common type of coherent noise that
can be aliased comes from low-frequency
waves trapped near the surface, called surface waves. On land, these are known as
ground roll, and create major problems for
processors. They pass the receivers at a
much slower velocity than the signal, and
so need closer receiver spacing to be properly sampled. Planners always try to design
surveys so that surface waves do not contaminate the signal. But if this is not possible, the surface waves must be adequately
sampled spatially so they can be removed.
During processing, S/N is enhanced
through filters that suppress noise. Coherent
noise is reduced by removing temporal and
spatial frequencies different from those of
the desired signal, if known. Both coherent
and random noise are suppressed by stackingsumming traces from a set of sourcereceiver pairs associated with reflections at
a common midpoint, or CMP.4 The sourcereceiver spacing is called offset. To be
stacked, every CMP set needs a wide and
evenly sampled range of offsets to define the
reflection travel-time curve, known as the
normal moveout curve. Flattening that
curve, called normal moveout correction,
will make reflections from different offsets
arrive at the time of the zero-offset reflection. They are then summed to produce a
stack trace (left ). In 3D surveys, with the

3. Directivity is the property of some sources whereby

seismic wave amplitude varies with direction.
4. For a full description of terms used in seismic data
processing see Sheriff RE: Encyclopedic Dictionary of
Exploration Geophysics. Tulsa, Oklahoma, USA: Society of Exploration Geophysicists, 1991.
5. Streamers are cables equipped with hydrophone
receivers. Multistreamer vessels tow more than one
receiver cable to multiply the amount of data acquired
in one pass. For a review of marine seismic acquisition
and processing see Boreham D, Kingston J, Shaw P
and van Zeelst J: 3D Marine Seismic Data Processing, Oilfield Review 3, no. 1 (January 1991): 41-55.

April 1994

Offsets and Azimuths in a CMP Bin

Source
Bin

Receiver
Offset Distribution
1

nReflections from source-receiver pairs bounce in a bin, a rectan-

gular, horizontal area defined during planning. In a 3D survey a

CMP trace is formed by stacking traces that arrive from a range of
azimuths and offsets (top). The distribution of offsets is displayed in
a histogram within each bin (bottom). The vertical axis of the histogram shows the amount of offset, and the horizontal axis indicates the position of the trace in offset.
Shotpoint number
130

160

190

220

130
40
36
160
33
29
190

25
22

220

Fold

advent of multielement marine acquisitionmultistreamer, multisource seismic

vesselsand complex land acquisition
geometries, reflections at a CMP come from
a range of azimuths as well as a range of
offsets (right ).5 A 3D CMP trace is formed by
stacking traces from source-receiver pairs
whose midpoints share a more or less common position in a rectangular horizontal
area defined during planning, called a bin.
The number of traces stacked is called
foldin 24-fold data every stack trace represents the average of 24 traces. Theoretically, the S/N of a survey increases as the
square root of the fold, provided the noise is
random. Experience has shown, however,
that for a given target time, there is an optimum fold, beyond which almost no S/N
improvement can be made.
Many survey designers use rules of thumb
and previous experience from 2D data to
choose an optimal fold for certain targets or
certain conditions. A fringecalled the fold
taper or haloaround the edge of the survey will have partial fold, thus lower S/N,
because several of the first and last shots do
not reach as many receivers as in the central
part of the survey (below, right ). Getting full
fold over the whole target means expanding
the survey area beyond the dimensions of
the target, sometimes by 100% or more.
Many experts believe that 3D surveys do not
require the level of fold of 2D surveys. This
is because 3D processing correctly positions
energy coming from outside the plane containing the source and receiver, which in the
2D case would be noise. The density of data
in a 3D survey also permits the use of noisereduction processing, which performs better
on 3D data than on 2D.
Filtering and stacking go a long way
toward reducing noise, but one kind of
noise that often remains is caused by multiple reflections, multiples for short. Multiples are particularly problematic where
there is a high contrast in seismic properties
near the surface. Typical multiples are reverberations within a low-velocity zone, such
as between the sea surface and sea bottom,

18
14
11

250

7
3

nA fold plot showing 40-fold coverage over the heart of the survey.

The edge of the survey has partial fold because several of the first
and last shots do not reach as many receivers as in the central part
of the survey.

21

a
0

Seafloor reflection

1.0

Time, sec

Seafloor multiple

2.0

Seafloor multiple

Primary reflection

3.0

Multiple

Multiple

4.0

nSeismic section with strong multiple noise. Multiples can

appear as a repetition of a shallower or deeper portion of the

seismic image. [Adapted from Morley L and Claerbout JF: Predic-

tive Deconvolution in Shot-Receiver Space, Geophysics 48 (May 1983):

515-531.]

Primary
reflection

Ghost

Near-surface
multiples

Long-path
multiple

nMultiple reflec-

tions. After leaving the source,

seismic energy
can be reflected a
number of times
before arriving at
the receiver.

or between the earths surface and the bottom of a layer of unconsolidated rock
(below, left ). Multiples can appear as later
arrivals on a seismic section, and are easy to
confuse with deep reflections (left ).6 And
because they can have the same characteristics as the desired signalsame frequency
content and similar velocitiesthey are
often difficult to suppress through filtering
and stacking. Sometimes they can be
removed through other processing techniques, called demultiple processing, but
researchers continue to look for better ways
to treat multiples.
The second characteristic of a good seismic signal is high resolution, or resolving
powerthe ability to detect reflectors and
quantify the strength of the reflection. This is
achieved by recording a high bandwidth, or
wide range of frequencies. The greater the
bandwidth, the greater the resolving power
of the seismic wave. A common objective of
seismic surveys is to distinguish the top and
bottom of the target. The target thickness
determines the minimum wavelength
required in the survey, generally considered
to be four times the thickness.7 That wavelength is used to calculate the maximum
required frequency in the bandwidth
average seismic velocity to the target
divided by minimum wavelength equals
maximum frequency. The minimum frequency is related to the depth of the target.
Lower frequencies can travel deeper. Some
seismic sources are designed to emit energy
in particular frequency bands, and receivers
normally operate over a wider band. Ideally,
sources that operate in the optimum frequency band are selected during survey
design. More often, however, surveys are
shot with whatever equipment is proposed
by the lowest bidder.

Guidelines from Geophysics

Many of the rules that guide 3D survey design are

can reach a receiver within a half-wavelength of

where R f is the radius of the Fresnel zone and dx

simple geometric formulas derived for a single

the first reflected energy. For a dipping reflector

is the CMP interval in the 2D data.

plane layer over a half-space: the equation

x =

describing the hyperbola used in normal moveout

correction is one example. Others are approxima-

Maximum offset, Xmax , is chosen after consid-

V rms

4f max sin

ering conflicting factorsvelocity resolution,

normal moveout stretch and multiple

tions from signal processing theory. Sometimes

where Vr ms is the root mean square average of

attenuation.1 For a velocity resolution v/v

survey parameters are achieved through trial and

velocities down to the target, fmax is the maxi-

required to distinguish velocities at time T,

error. The following formulas hold for some sim-

mum nonaliased frequency required to resolve

ple 3D surveys:

the target, and is the structural dip. Normally

Bin size, xy, is calculated to satisfy vertical

and lateral resolution requirements. For a flat

y = x.

3D fold is determined from estimated S/N of

X max =

2Tv 2
f v
v

where f is fmax fmin , or the bandwidth. As Xmax

reflector, bin length, x, can be the radius of the

previous seismic data, usually 2D. 3D fold must

increases, v/v increases, or improves. But with

Fresnel zone or larger. The Fresnel zone is the

be greater than or equal to

long offsets, normal moveout stretch increases

area on a reflector from which reflected energy

2D fold

22

xy
2Rf dx

and multiples can become worse.

1. Normal moveout stretch is the distortion in waveshape caused by normal moveout correction.

a
Horizontal Reflector

Another variable influencing resolution is

source and receiver depthon land, the
depth of the hole containing the explosive
source (receivers are usually on the surface),
and at sea, how far below the surface the
sources and receivers are towed. The
source-receiver geometry may produce
short-path multiples between the sources,
receivers, and the earth or sea surface. If the
path of the multiple is short enough, the
multiplesometimes called a ghostwill
closely trail the direct signal, affecting the
signals frequency content. The two-way
travel time of the ghost is associated with a
frequency, called the ghost notch, at which
signals cancel out. This leaves the seismic
record virtually devoid of signal amplitude
at the notch frequency. The shorter the distance between the source or receiver and
the reflector generating the multiple, the
higher the notch frequency. It is important to
choose a source and receiver depth that
places the notch outside the desired bandwidth. It would seem desirable to plan a
survey with the shallowest possible sources
and receivers, but this is not always optimal,
especially for deep targets. On land, shortpath multiples can reflect off near-surface
layers, making deeper sources preferable. In
marine surveys, waves add noise and instability, necessitating deeper placement of
both sources and receivers. In both cases,
survey design helps reach a compromise.
The third requirement for good seismic
data is adequate subsurface coverage. The
lateral distance between CMPs at the target
is the bin length (for computation of bin
length, see Guidelines from Geophysics,
previous page ). Assuming a smooth horizontal reflector, the minimum source spacing and receiver spacing on the surface
must be twice the CMP spacing at the target. If the reflector dips, reflection points are
not CMPs (above, right ). Reflected waves
may be spatially aliased if the receiver spacing is incorrect. A survey designed with good
spatial coverage but assuming flat layers
might fail in complex structure. To record
reflections from a dipping layer involves
more distant sources and receivers than
reflections from a flat layer, requiring expan-

April 1994

Dipping Reflector

Shotpoint
Receiver

nEffect of reflector dip on the reflection point. When the reflector is

flat (top) the CMP is a common reflection point. When the reflector
dips (bottom) there is no CMP. A dipping reflector may require
changes in survey parameters, because reflections may involve
more distant sources and receivers than reflections from a flat layer.

sion of the survey areacalled migration

apertureto obtain full fold over the target.
In general, survey planners use simple
trigonometric formulas to estimate optimal
CMP spacing and maximum source-receiver
offset on dipping targets. As geophysicists
seek more information from seismic data,
making the technique more cost-effective,
simple rules of thumb will no longer provide optimum results. Forward modeling of
seismic raypaths, sometimes called raytrace
modeling, provides a better estimate of subsurface coverage, but is not done routinely
during survey planning because of cost and
time constraints. An exception is a recent
evaluation by Geco-Prakla for a survey in
the Ship Shoal South Addition area of the
Gulf of Mexico (page 31).

6. Claerbout JF: Imaging the Earths Interior. Boston,

Massachusetts, USA: Blackwell Scientific Publications
(1985): 356.
7. This is the criterion for resolving target thickness visually. By studying other attributes of a seismic trace
such as amplitude or signal phase, thinner layers can
be resolved.
8. Survey design and survey planning are sometimes
used interchangeably, but most specialists prefer to
think of planning as the part of the design process that
considers cost constraints and logistics.

Balancing Geophysics with Other

Constraints

Acquiring good seismic signal is expensive.

On land or at sea, hardware and labor costs
constrain the survey size and acquisition
time. The job of the survey planner is to balance geophysics and economy, achieving
the best possible signal at the lowest possible cost. 8 On land, source lines can be
aligned with receiver lines, or they can be at
angles to each other. Different sourcereceiver patterns have different cost and signal advantages, and the planner must

23

Theoretical Grid

Final Grid

choose the one that best suits the survey

(right ). Once a survey pattern is selected,
subsurface coverage can be computed in
terms of fold and distribution of offset and
azimuth. If the coverage has systematic
holes, the pattern must be modified.iIn complex terrain, planned and actual surveys
may differ significantly (left ).9
Land acquisition hardware can cost $5
million to $10 million for recording equipment and sourcesusually vibrating trucks
or dynamitebut labor is the major survey
cost. Cost can be controlled by limiting the
number of vibrator points or shotpoints, or
the number of receivers. But limiting
receivers limits the area that can be shot at
one time. If a greater area is required,
receivers must be picked up and moved,
increasing labor costs. The most efficient
surveys balance source and receiver
requirements so that most of the time is
spent recording seismic data and not waiting for equipment to be moved. Land preparation, such as surveying source and
receiver locations and cutting paths through
vegetation or topography, must be included
on the cost side of the planning equation. In
countries where mineral rights and land surRandom Technique

Checkerboard Pattern

Brick Pattern

Zigzag Pattern

Source
Receiver

nCommon source-receiver layouts for land

acquisition. The checkerboard pattern

(top), sometimes called the straight-line or

cross-array pattern, is preferred when the

source is a vibrator truck, because it
requires the least maneuvering. The brick
pattern, (middle) sometimes called staggered-line, can provide better coverage at
short offsets than the checkerboard, but is
more time-consuming, and so costlier. The
zigzag pattern (bottom) is highly efficient in
areas of excellent access, such as deserts,
where vibrator trucks can zigzag between
receiver lines.
Source line
Receiver line

Source
Receiver

nPlanned versus actual surveys. A survey planned in West Texas,

USA (top, left) calls for a checkerboard of receiver lines (blue) and
source lines (red). The actual survey shot (bottom, left) came very
close to plan. Other cities present acquisition challenges. A survey
in Milan,iItaly (right) used a random arrangement of sources and
receivers. (Adapted from Bertelli et al, reference 9.)

24

face rights are separately and privately held,

such as in the US, landowners must give
permission and can charge an access fee.
Other constraints that can affect survey
planning include hunting seasons, permafrost, population centers, breeding seasons, animals migrating or chewing cables,
and crops that limit vibrator source trucks to
farm roads.
Marine survey planners consider different
constraints. Hardware is a major cost;
sources and recording equipment are a sizable expense, but additionally, seismic vessels cost $35 to $40 million to build, and

Oilfield Review

clock starts ticking once acreage is licensed.

Exploration and development contracts
require oil companies to drill a certain
number of wells, spend a certain amount of
money, or shoot a certain amount of seismic data before a given date. There is often
little time between gaining approval to
explore or develop an area and having to
drill. In some cases, oil companies plan
every detail of the acquisition before putting
the job out to bid. In other cases, to
increase efficiency, oil companies and seismic service companies share the planning.
In many cases, service companies plan the
survey from beginning to end based on
what the oil company wishes to achieve. In
the quest for cost savings, however, seismic
signal is often compromised.

Cost-Effective Seismic Planning

How would 3D seismic acquisition, processing and interpretation be different if a

little more emphasis were given to survey
design? Geco-Praklas Survey Evaluation
and Design team in Gatwick, England, has
shown that by taking a bit more care, signal
can be improved, quality assured and cost
optimized simultaneously. There are three
parts to the process as practiced by GecoPraklaspecification, evaluation and design
(next page ). Specification defines the survey
objectives in terms of a particular depth or
target formation, and the level of interpretation and resolution required. The level of
interpretation must be defined early; data to
be used solely for structural interpretation
can be of lesser quality, leading to lower

nMarine acquisi-

tion geometry
showing seismic
vessels looping in
oblong circuits.
The length of
straight segments is
calculated from
fold plots, and must
include additional
lengthrun in
and run outto
allow cable to
straighten after
each turn.

tens of thousands of dollars per day to operate. Sources are clusters of air guns of different volumes and receivers are hydrophones
strung 0.5 m [1.6 ft] apart in groups of up to
48, on cables up to 6000 m [19,680 ft]
long. Sources and receivers are almost
always towed in straight lines across the target (below, right ), although other geometries
are possible. Circular surveys have been
acquired with sources and receivers towed
by vessels running in spirals or concentric
circles.10 Geco-Praklas QUAD-QUAD system tows four receiver cables and four
source arrays simultaneously, acquiring 16
lines at a time. Currents and tides can cause
the long receiver cables to deviate by calculable amountsup to 30from the towing
direction. Spacing between shotpoints is a
function of vessel speed, and can be limited
by how quickly the air guns can recover full
pressure and fire again. Access is usually
limited only by water depth, but drilling
rigs, production platforms and shipping
lanes can present navigational obstacles.
Environmental constraints also influence
marine surveys: the commercial fishing
industry is imposing limits on location of,
and seasons for, marine acquisition.11 For
example, planning in the Caspian Sea must
avoid the sturgeon breeding season or seismic surveys would wipe out caviar production for the year.
Transition zonesshallow water areas
have their own problems, and require specialized equipment and creative planning.12
Transition zones are complex, involving
shorelines, river mouths, coral reefs and
swamps. They present a sensitive environment and are influenced by ship traffic,
commercial fishing and bottom obstructions. Survey planners have to contend with
varying water depths, high environmental
noise, complex geology, wind, surf and
multiple receiver typesoften a combination of hydrophones and geophones.
One thing all surveys have in common is
that planning must be done quickly. The
9. Bertelli L, Mascarin B and Salvador L: Planning and
Field Techniques for 3D Land Acquisition in Highly
Tilled and Populated AreasTodays Results and
Future Trends, First Break 11, no. 1 (January 1993):
23-32.
10. Hird GA, Karwatowski J, Jenkerson MR and Eyres A:
3D Concentric Circle SurveyThe Art of Going in
Circles, EAEG 55th Meeting and Technical Exhibition, Stavanger, Norway, June 7-11, 1993.
11. Gausland I: Impact of Offshore Seismic on Marine
Life, EAEG 55th Meeting and Technical Exhibition,
Stavanger, Norway, June 7-11, 1993.
12. Petersen C, Brakensiek H and Papaterpos M:
Mixed-Terrain 3D Seismics in the Netherlands,
Oilfield Review 4, no. 3 (July 1992): 33-44.

April 1994

Bin

25

Specification
Define survey
objectives

Required
geophysical
parameters

Resolution
analysis

Design
No Required Yes
equal
obtainable?

Source,
template
and array
design

Planning
Preferred
survey
parameters

No
Objectives Yes Final survey
design
achieved?

Evaluation
Analysis of
existing data

Data
Type

Resolution, noise
and coverage
analysis

Parameters
to be Determined
Maximum frequencies
attainable

Prospect
description

Obtainable
geophysical
parameters

Means to
Determine Parameters
VSP processing

Reflection response
of target

VSP

Identification
of multiples origin
Source peak amplitude

Source modeling

Source volume
Resolution attainable

Process or Output

Loss modeling
Frequency
dependent losses

Peak-to-bubble ratio
Source depth

Operational,
cost and safety
constraints

Source signature
for various depths
Bandwidth at target

Apply losses to
source signatures

Target wavelet

Noise levels

Logs or
1D Models

Estimate spatial and

temporal resolution
Shooting direction
Primary/multiple
velocity discrimination
Stack fold, offset and
group length for
optimum multiple
moveout discrimination

Synthetic
CMP gathers
Synthetic shots
Migration aperture

Required streamer length

Analysis of
2D synthetic
CMP gathers

Long-offset analysis
Normal incidence
stacks
Statics model

Signal-to-noise ratio

FK plots, filter tests

Establish noise
mechanisms

Refraction velocities
(near surface)

Near trace offset

2D or 3D
Surface Seismic

Modeled section
Build geological
2D model and
apply appropriate
target wavelet

Useful offset with

time, stack and
S/N relationship

Analysis of existing
surface seismic

Amplitude versus
time plots

Group interval

Mute, stack,
fold tests

Crossline spacing
Spatial frequency
Spatial resolution
Shotpoint interval

Noise records

costs, compared to data used for stratigraphic interpretation, analysis of amplitude

variation with offset (AVO) or seismic monitoring.13 Specification quantifies the geophysical parameters needed to meet the
interpretation objectives: frequency content
and signal-to-noise ratio of the recorded signals, and spatial sampling intervalthe
familiar requirements for good signal.
Evaluation of existing data, which can be
done independently and concurrently, tells
which geophysical parameters are obtainablesometimes different from those stipulated by specification. The types of data
evaluated include logs, vertical seismic profiles (VSPs) and 2D or existing 3D data.
Existing data can provide models for simulating the effects of the geophysical parameters on new seismic data. If the required
parameters are not obtainable, the survey
objectives are reexamined, or respecified.
The loop is repeated until a set of geophysical parameters is found that is both desired
and obtainable.
In the third step, design, the geophysical
parameters are weighed against other constraints. Keeping in mind the understanding
gained from evaluation of existing data, survey planners select the source and receiver
configuration and choose the shooting
sequence and type of seismic source. These
preferred survey parameters are tempered by
cost, safety and environmental constraints.

Analysis of
migration
requirements

Migration aperture
Shooting direction
Record length

Migration of
synthetic
zero-offset data
Migration of
existing 2D data
Ambient noise
estimation

nSurvey evaluation and design scheme.

26

Oilfield Review

Shot Depth 28 m

Putting Planning into Practice

13. For a review of AVO see:

Chiburis E, Franck C, Leaney S, McHugo S and Skidmore C: Hydrocarbon Detection With AVO, Oilfield Review 5, no. 1 (January 1993): 42-50.
For more on seismic monitoring see:
Albright J, Cassell B, Dangerfield J, Deflandre J-P,
Johnstad S and Withers R: Seismic Surveillance for
Monitoring Reservoir Changes, Oilfield Review 6,
no. 1 (January 1994): 4-14.
14. Source patterns are groups of dynamite charges in
separate holes at the same depth, fired simultaneously. The goal is to cancel low-frequency noise that
travels laterally, called ground roll.

April 1994

nTests with dyna-

mite sources at different depths.

Traces recorded
from the shot at 28
m [92 ft] (left) show
less low-frequency
noiseground
rollthan from the
shot at 9 m (right).
In general, the
deeper the source,
the less ground roll
generated.

1.0

Time, sec

In 1991 Elf Petroleum Nigeria Limited put

out for tender a 160-km2 [62-sq mile] land
survey in the Niger Delta. Working with the
Seismic Acquisition Service of Elf Aquitaine
Production in Pau, France, the Survey Evaluation and Design group evaluated sources
and geometries for optimal acquisition. The
primary target is the structure of the Ibewa
oilfield at 3500 m [11,480 ft], at or below 3
sec two-way time, with secondary deeper
objectives. Signal-to-noise requirements,
based on previous experience, suggested
the data should be 24-fold. Resolution of
the target required signal bandwidth of 10 to
60 Hz and 25 m by 25 m [82 ft by 82 ft]
bins. The source was specified to be dynamite, which would be fired in shotholes
drilled and cased or lined to 25 m, again
based on previous experience. Constraints
on the survey included the high population
density, potential damage to personal property and the many oil pipelines that cross
the area. A roll-along acquisition pattern
similar to a checkerboard was suggested in
the bid, with four receiver lines to be moved
as the survey progressed (below, right ).
Evaluation of existing data2D seismic
lines and results from seismic source
testswarned of potential problem areas.
Source tests compared single-source dynamite shots to source patterns, and tested several source depths.14 The tests indicated the
presence of ghost notches at certain depths,
leading to a reduction in signal energy
within the desired frequency band of 10 to
60 Hz (above, right ). The source tests also
indicated source patterns were ineffective in
controlling ground roll in this prospect area.
Deployment of the source at 9 m [30 ft]
gave a good S/N ratio at 25 to 60 Hz, but
produced very high levels of ground roll.
Deployment of the source below 40 m [130
ft] gave a good S/N ratio from 10 to 60 Hz
and low levels of ground roll. However,
such deep holes might be unacceptably
time-consuming and costly.
Evaluation of existing 2D lines revealed
the frequency content that could be

Shot Depth 9 m

2.0

3.0

4.0

Minimal ground roll

Offset

Ground roll
Offset

300 m
Source
Receiver

nOriginal roll-along geometry proposed for

Elf Petroleum Nigeria survey. Four receiver

lines would be laid at 300-m intervals. Each
line would have 144 receivers with 50-m
spacing. Shots would be fired at 50-m intervals in a line perpendicular to the receiver
lines, and then the four receiver lines would
be rolled along to the next position.

27

0-10 Hz

10-20 Hz

20-30 Hz

30-40 Hz

nFiltered 2D data

showing frequency
content variation
with depth. Each
panel has been filtered to allow a different band of frequencies, called the
passband, to pass.
As the passband
rises, the maximum
depth of penetration
of seismic energy
decreases. Lower
frequencies (left)
penetrate deeper.
Higher frequencies
(right) do not propagate to deeper levels. At the target
level of 3.0 sec there
is still some 50 Hz
energy left.

1.0

Time, sec

40-50 Hz

2.0

3.0

4.0

expected from seismic data in the area

(above ). Resampling along the 2D line at
the sampling interval planned for the 3D
survey confirmed that the 50-m [165-ft]
receiver and shot spacings initially recommended were appropriate. Fold-reduction
simulations performed on the 2D sections
showed that 24-fold would be appropriate
for the survey. However, a brick pattern
would give better fold and offset distribution
than the roll-along pattern, potentially
improving the survey results. The brick pattern would also reduce the lateral offset
between source and receiver line, thus
reducing the potential for ground roll arriving at the same time as the reflection from
the target and making the ground roll easier
to handle in processing.

28

The complete survey evaluation and

design took two months and reached the
following conclusions.
1. A target bandwidth of 10 to 60 Hz is a
reasonable acquisition objective.
2. Placement of sources deeper than 40 m
would avoid complex processing problems and high levels of ground roll in the
3D data set. If logistics prevent locating
the sources at this depth, then a fallback
deployment of sources at 9 m would
meet the target bandwidth criterion with
minimal notching but higher levels of
ground roll. Field quality control should
verify there is no notch between 10 and
60 Hz.
3. A 144-trace brick pattern with 300-m
[984-ft] receiver line spacing and 300-m
shot line spacing would give the best offset distribution.
4. Shot and receiver intervals should be no
more than 50 m.
Drilling 40-m holes for each source location
was deemed impractical. Optimizing costs
and logistics, the company obtained satisfactory results with a 24-m [79-ft] source
depth, single-shot dynamite, and brickwork
acquisition pattern.
Evaluation and design can be different in
the marine setting. A case in point is the Al
Shaheen location in offshore Qatar, under
development appraisal by Maersk Oil Qatar
AS, according to an agreement with Qatar
General Petroleum Corporation (QGPC).
Maersk Oil had only eight months to design

and acquire a 3D survey that would provide

a 25 km 2 [9.6 sq mile] image, requiring
about 49 km2 [18.8 sq mile] of full fold data,
and to spud a vertical development
appraisal well. Given the tight schedule
processing alone normally takes a year
Maersk Oil contracted a survey evaluation
and design study based on existing VSPs and
2D surveys. This study was more extensive
than the previous example, with more preexisting data, particularly well data.
The objective of the 3D survey was to
produce a stratigraphic image of the
Kharaib limestones and a thin 13- to 15-ft
[4- to 4.6-m] thick overlying oil-filled sand.
The seismic data were to be analyzed for
porosity-related amplitude variations along
with small-scale faulting and fracturing to
help in planning the trajectory of future horizontal wells. The acquisition vessel had
already been contracted, limiting the seismic source to a 1360- or 1580-in.3 [22,290or 25,900-cm3] air gun.
Evaluation of existing data indicated areas
where special care had to be taken to
ensure a successful survey. For example,
high-velocity beds at the seafloor promised
to cause strong multiples, reducing the
energy transmitted to deeper layers and

Oilfield Review

5-10 Hz

10-20 Hz

20-40 Hz

40-60 Hz

60-80 Hz

80-100 Hz

Time, sec

0.5

1.0

1.5

nBandpass filters on VSP data showing energy present up to 80 to 100 Hz at target.

April 1994

2400

90

3325

80

4250

70

60
5175

Level number

leading to strong reverberations in the water

layer. A bandwidth of 10 to 90 Hz was
required to resolve the thin sands above the
target and the small faults within it.
Evaluation of existing borehole data
offered valuable insight into the transmission properties of the earth layers above the
target and the geophysical parameters that
could be obtained at the target. Comparison
of formation tops inferred from acoustic
impedance logs with reflection depths on
the two VSPs allowed geophysicists to differentiate real reflections from multiples.
Identification of the origin of multiples
allowed the acquisition and processing
parameters to be designed to minimize their
effect. Analysis of the amplitude decrease of
the VSP downgoing first arrivals quantified
transmission losses (right ). Bandwidth studies on the VSPs showed that frequencies in
the 80- to 100-Hz range were present and
being reflected at the depth of the target
(above ). This meant the frequencies required
for thin-bed resolution might be obtainable
by the 3D survey.
The study also looked into quantifying the
seismic resolution of small-scale faulting
(next page, top) and analyzed five different

Measured depth, ft

Each panel passes a different band of frequencies. Coherent energy up to 80 to

100 Hz reflects from the survey objective at 0.8 sec.

6100

50

40

7025
30
7950

20

8875
10
9800

0
0

0.5

Time, sec

1.0

80

70

60

50

Amplitude loss from surface, dB

nVertical seismic profile (VSP) traces (left) analyzed for amplitude loss with depth (right).
Amplitudes of first arrivals recorded in a 92-level VSP are calibrated with amplitudes of a
surface reference signal to account for changes in source amplitude from level to level.
The amplitude ratio from one level to the next is plotted in decibels (dB). One dB is 20
times the log of the amplitude ratio. An amplitude ratio of 100 is equivalent to 40 dB.
Amplitudes expected from a surface seismic survey would normally be 3 dB less than
those from a VSP, and scaled by a reflection coefficient.

29

3-m throw

6-m throw

12-m throw

24-m throw

0.3

Two-way time, sec

0.4

0.5

12-m
thick
0.6

24-m
thick

0.7

36-m
thick

0.8

nResolution of thin beds and small-scale faulting. Each panel shows the modeled

response of a seismic wave of 48-m [160-ft] wavelength ( ) to a different vertical fault displacing a series of thin beds of thicknesses 12 m, 24 m and 36 m. From left to right, faults
with 3-m [10-ft], 6-m [20-ft], 12-m [40-ft] and 24-m [80-ft] throws correspond to /16, /8,
/4 and /2, respectively. A fault throw of at least 12 m, corresponding to /4, can be
resolved quantitatively. At less than that, existence of a fault can be detected, but its
throw resolved only qualitatively.

nA time slice from

Maersk Oil Qatar

3D cube showing
fractures and faults.

-128

Amplitude

30

127

miles

km

3.2

energy sources, source and streamer depth,

spatial sampling and minimum and maximum offsets. Some of the early 2D lines
were reprocessed to evaluate migration
requirements and techniques for removing
multiples. 15 Five recommendations were
offered for survey acquisition:
1. A target frequency of 90 Hz is a reasonable objective and can achieve the
desired resolution.
2. Multiples reverberating in the water will
create severe problems. Offsets longer
than about 1000 m [3280 ft] may not be
useable because they will contain multiples indistinguishable from the target
signal.
3. Of the available sources, the 1580-in.3
source would be preferred to the 1360in.3 source because of its higher energy
output at the important higher frequencies. This, however, is subject to the
ability of the larger source to be cycled
at a 12.5-m [41-ft] shotpoint interval.
4. Receiver intervals of 12.5 m and shotpoint intervals of 12.5 m should sufficiently sample the signal and the
expected noise, allowing further reduction of noise during processing. These
intervals provide sufficient fold to
achieve the desired S/N within the 10- to
90-Hz bandwidth.
5. Because the primary reflection and multiples cannot be discriminated by differences in their velocities, stacking may
not adequately attenuate multiples.
Additional demultiple processing may
be necessary.
All the survey design recommendations
were implemented except the larger source,
which for technical reasons could not be
towed as planned.
The survey acquired superb data. Maersk
Oil Qatar drilled the vertical well on time
and based on interpretation of the new seismic data, spudded two horizontal wells
one with a 10,200-ft [3120-m] long horizontal section. The 3D data show fine-scale
faulting and two fracture sets (left ). Fault
location prediction based on interpretation
of the 3D data was confirmed during

Oilfield Review

0
1000

2000

Depth, m

drilling. Faults with throws as little as 8 to 10

ft [2.5 to 3 m] interpreted in the seismic data
were verified by on-site biostratigraphic
evaluation of the reservoir limestones.
In addition to these two surveys, GecoPrakla has conducted more than 30 other
survey evaluation and design studies, sometimes with surprising results. In one case,
analysis of tidal currents led the team to propose a change of 120 in shooting direction,
which would add $150,000 to the processing cost, but cut 45 days and $1,500,000 off
the acquisition cost, for a savings of $1.35
million. In another study, analysis of previous seismic data showed that coherent shotgenerated noise was aliased at shot intervals
of 37.5 m [123 ft]. Although it would
increase acquisition and processing costs, a
denser shot interval of 25 m would sample
the noise sufficiently to allow removal during processing. The 37.5-m shot spacing
was used in the survey, giving data that
required extra prestack processing costs,
which did not entirely eradicate the noise.
In a study with Schlumberger Technical
Services in Dubai, UAE, data from a VSP
acquired just before a marine 3D survey
helped optimize planning.16 In a deviated
production well near the center of the survey, a slimhole TWST Through-Tubing Well
Seismic Tool was run through tubing to the
reservoir to record shots fired from the seismic source to be used in the 3D survey. The
shot records allowed geophysicists to determine the effects at the depth of the target of
source parameters such as air-gun volume,
depth and pressure. The records also
showed that at far offsets, high amplitude
shear waves contaminate the traces. With a
shorter receiver cable, a better survey was
acquired in less time, and so for lower cost,
than originally planned.

Salt

3000

4000

5000

6000
0

4000

8000

12,000

16,000

20,000

24,000

28,000

Distance, m

nRaytrace modeling showing strong changes in reflection paths through salt. Traces that

would have a common midpoint in a flat-layered earth no longer bounce in the same
bin. Salt, with its ability to deform and its high seismic velocity, creates complex structure
and strong refraction, or ray bending.

TEXAS

LOUISIANA

Ship Shoal
South Addition

G U L F
0
0

miles
km

O F

M E X
I C O

100
161

nShip Shoal South Addition in the Gulf of

Mexico.

ture, survey designers recognize the need to

model raypaths, and some are beginning to
do this. Geco-Prakla has used raytrace
modeling to determine coverage in a survey
to image below salt in the Ship Shoal South
Addition in the Gulf of Mexico (left ).
Salt introduces large contrasts in seismic
velocity, bending and distorting seismic rays
along complex paths (top). Survey designers
anticipated that a super-long receiver cable
would be required to provide adequate coverage of the subsalt layers. They tested various cable lengths by shooting raypaths
through a geologic model derived from 2D

For the Future

Some of the advances to be made in 3D survey design have origins in other fields. VSP
design routinely models seismic raypaths
through complex subsurface structure, but
rarely does surface seismic design account
for structure. Despite considerable sophistication in 3D data processing, most 3D survey design assumes plane layer geometry in
the subsurface to calculate midpoints and
target coverage. But to estimate subsurface
coverage adequately in complicated struc15. Migration, sometimes called imaging, is a processing
step that rearranges recorded seismic energy back to
the position from which it was reflected, producing
an image of the reflector.
16. Poster C: Taking the Pulse of 3D Seismics, Middle
East Well Evaluation Review, no. 13 (1992): 6-9.

April 1994

31

0
1000

Depth, m

2000

3000

4000

5000

6000
0

4000

8000

12,000

16,000

20,000

24,000

28,000

20,000

24,000

28,000

Distance, m
0
1000

Depth, m

2000

3000

4000

5000

6000
0

4000

8000

12,000

16,000

Distance, m

nRaytrace modeling to optimize cable length. Refraction through salt may mean a longer
cable is required to image structure below. Two cable lengths, 8075 m (top) and 5425 m
(bottom) were tested using the model on the previous page. Surprisingly, in this case both
cables give similar coverage of subsalt horizons.

seismic data (above ). Surprisingly, a standard 5425-m [17,794-ft] cable provides

coverage similar to that of the proposed
8075-m [26,500-ft] cable.
Another advance may come through integration of survey design with acquisition,
processing and interpretation into a single
quality-assured operation. The aim is to
maximize cost-effectiveness of the overall
seismic survey, to supply quality-assured
processed data with minimum turnaround
time and optimal cost. Within Geco-Prakla,
this idea is called Total Quality 3D, or
TQ3D. Such surveys may be acquired on a
proprietary (exclusive) or a speculative
(nonexclusive) basis, or a combination of
the two. For example, 75% of a 700-km2
[271-sq mile] TQ3D survey in the southern
UK continental shelf will be delivered as

32

proprietary data to three oil companies. The

remaining 25% is nonexclusive, and
although sponsored in part by the current
players in this area, the data will also be
available to new players.
Defining the objectives of a TQ3D survey
can be a difficult process. Rather than hazarding a guess at which reflectors in an area
are the sought-after targets, Geco-Prakla
planners involve proprietary and nonexclusive clients at early stages of the project.
Over open acreage they examine a data
base of nonexclusive 2D seismic surveys to
learn about the targets.
Choosing acquisition parameters that will
be optimal over the entire survey is also a
challenge. It is not always practical to follow
all the recommendations proposed by a survey evaluation and design study, but a judgment can be made of the impact that any
decision will have on the quality of the data.
Then, other options can be explored. For

example, in a recent TQ3D survey, steeply

dipping reflectors in 20% of the area would
have been optimally sampled if the receiver
spacing had been reduced from 25 m to 20
m [66 ft], but the 25% additional cost was
unacceptable to clients. Having flagged this
as an area where data quality could be
improved, attention will be paid to processing that may help imaging of steep dips.
As oil companies and service companies
strive for efficiency and acquisition of highquality, cost-effective seismic data, more
emphasis is being placed on survey design.
The other pieces of the seismic puzzle
acquisition, processing and interpretationhave all benefited from advances in
technology, and survey design is following
the trend. Through powerful modeling and
integration of log, VSP and surface seismic
data, 3D survey design will become the
foundation for all that follows.
LS

Oilfield Review