You are on page 1of 9

Downloaded 07/07/15 to 201.76.175.58. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.

org/

GEOPHYSICS, VOL. 66, NO. 6 (NOVEMBER-DECEMBER 2001); P. 18181826, 11 FIGS., 2 TABLES.

Optimizing operations in 3-D land seismic surveys

Douglas J. Morrice , Astrid S. Kenyon , and Christian J. Beckett


In this paper, we formulate and solve an optimization model
for operations planning in 3-D land seismic surveys. The requirements of seismic surveying can be subdivided into two categories: geophysical and operational. As Ashton et al. (1994)
point out, the job of the survey planner is to balance geophysics and economy, achieving the best possible signal at the
lowest possible cost. A great deal of the formal research on
the subject has focused on improving the geophysical aspect of
seismic surveys. Much less emphasis has been placed on techniques for improving operations. Based on experience, field
personnel have developed the best operational strategies available today. Since domain specific knowledge is critical to improved operational efficiency, the purpose of this paper is not
to replace input from experienced personnel. It is, however, intended to complement these field heuristics with a formal procedure for optimizing survey operations while balancing the
geophysical and operational requirements. More specifically,
we will formulate and solve an optimization problem for a simple orthogonal acquisition geometry using formal techniques
referred to as mathematical programming from the field of operations research [for an introduction to operations research,
see Hillier and Lieberman (1990)]. The model we develop provides a formal procedure for selecting a 3-D land seismic survey design optimized to reduce cost. It is not intended to be an
all-encompassing model; rather it is intended to illustrate how
mathematical programming can be used to solve problems in
survey design.
We are aware of only one other published paper on this
topic: Liner et al. (1998). Our paper differs from that work in
two ways. First, we explicitly incorporate operations decisions
and costs into our model. This approach affords the decision
maker an additional degree of control and the ability to incorporate realistic cost figures directly into the model. Second, we
illustrate the implementation of our model in a spreadsheet.
The main advantages of the spreadsheet environment include
a high degree of user familiarity and ready access to existing
models and data.
The remainder of our paper is organized into the following sections: 3-D land seismic surveying provides some

ABSTRACT

We formulate and solve a mathematical programming


optimization model to find a minimum cost solution for
an orthogonal split-spread design in a 3-D land seismic
survey. The model contains decision variables on source
and receiver location spacings, the amount of receiver
equipment, and the production rate of the seismic crews.
The model includes operational constraints for source
and receiver movements. It also includes geophysical
constraints for fold coverage, offset, and azimuth. To
demonstrate the efficacy of the model, we include an
example and solve it using the nonlinear optimization
solver in Microsoft Excel. The model results demonstrate the classic trade-off between source and receiver
points to satisfy the geophysical requirements. In addition, we conduct sensitivity analysis on an important
production parameter: the maximum number of source
points that can be shot per day. We show that although
changes to this parameter do not impact the decision
variables, such changes do have a significant effect on
the total cost of the survey.

INTRODUCTION

One of the key technologies in geophysical prospecting is


seismic surveying (Dobrin and Savit, 1988). Although this technology has been used commercially since the 1940s (DeGolyer,
1947), over the years increasingly sophisticated survey designs
and data processing techniques have been developed. More
sophisticated designs have led to more complex field acquisition operations. For example, until the early 1980s, most land
seismic surveys were two dimensional (2-D) surveys conducted
along a single line of source and receiver points. Today, most
surveys are three-dimensional (3-D) designs where source and
receiver points are distributed on a region on the earths surface. The design of the source and receiver geometry is defined
by the requirements of the survey.

Manuscript received by the Editor August 2, 1999; revised manuscript received March 16, 2001.

University of Texas at Austin, MSIS Department, CBA 5.202, Austin, Texas 78712-1175. E-mail: morrice@mail.utexas.edu.
Siemens AG, CT SE 6, Otto-Han-Ring 6, 81730 Munich, Germany. E-mail: astrid kenyon@mchp.siemens.de.

Formerly Schlumberger Reservoir Evaluation Seismic, 200 Gillingham Lane, Sugarland, Texas 77478; presently Rice University, Jones Graduate
School of Management, 270 Herring HallMS531, 6100 Main Street, Houston, Texas 77005-1892. E-mail: beckettc@swbell.net.

c 2001 Society of Exploration Geophysicists. All rights reserved.


1818

Downloaded 07/07/15 to 201.76.175.58. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

Optimizing Seismic Surveys

background information on 3-D seismic surveys; Mathematical programming and optimization includes a discussion on
formulating and solving mathematical programming models;
An optimization model for operations planning presents the
optimization of an orthogonal split-spread survey design to
minimize the costs, subject to both geophysical and operational
constraints (this section also includes an example and sensitivity analysis results using the Microsoft Excel nonlinear optimization solver); Conclusions contains some final remarks
and future research directions.
3-D LAND SEISMIC SURVEYING

This section provides the information required to motivate


the model considered later in this paper. It also includes terminology used throughout the paper. Additionally, it illustrates
the operational perspective from which we view 3-D land seismic surveying. Ashton et al. (1994) provide an excellent introduction to 3-D seismic survey design. Stone (1994) provides a
thorough historic perspective of seismic survey designs in two
and three dimensions. For a review of modern design issues,
see Horman et al. (2000).
Seismic surveying is primarily concerned with reflection seismology. A source on the surface of the earth generates a signal
that propagates through the earth. The underground geological
structures attenuate, reflect, and refract the signal. Receivers
on the earths surface monitor the reflected signal. Using this
and other information, seismologists construct a picture of the
earths subsurface. For other information used in geophysical
prospecting, see Dobrin and Savit (1988).
Figure 1 illustrates a highly simplified picture of reflection
seismology. A signal is generated from a source point, follows
along a raypath, and is reflected off a subsurface contrast in
elastic impedance, a reflector. The reflected energy returns to
a receiver point along a defined raypath. This reflected signal
provides information about the subsurface structure, in this
case, a flat layer. For a flat layer, the point of reflection is called
the common midpoint (CMP). In seismic surveying, multiple
receivers detect the signal sent from a single source point. This
permits the simultaneous sampling of several CMPs (Figure 2).
If the process is repeated for several sources, then the same
CMPs get sampled several times (Figure 3) by independent
raypaths. Each raypath is represented by a single trace in the
field data. Sampling of the same point multiple times and stacking of the results improves the signal-to-noise ratio (S/N) of the
data.
Figures 13 illustrate a 2-D arrangement of sources and receivers along a line. In 3-D surveys, sources and receivers are

FIG. 1. Reflection seismology. CMP = common midpoint.

1819

distributed on a plane over the earths surface. This distribution


typically forms a grid pattern similar to that shown in Figure 4.
The vertical lines in Figure 4 contain source locations (represented by an X) and the horizontal lines contain receiver
locations (represented by a ). This 3-D survey design is called
an orthogonal design. Design considerations for an orthogonal design include the spacing of the source and receiver lines,
and the spacing of the source and receiver locations along the
lines. A number of different designs exist for 3-D surveys with
more complicated design criteria than the one presented in
Figure 4. For some examples, see Ashton et al. (1994). The signal sent from any given source point is monitored by multiple
receivers on different lines (see Figure 5). The set of receivers
that monitor a single source signal are often referred to as the
active patch or spread. The term rack usually refers to the set

FIG. 2. Simultaneous sampling of several CMPs.

FIG. 3. Simultaneous sampling of several CMPs with several


sources.

FIG. 4. Orthogonal survey design.

Downloaded 07/07/15 to 201.76.175.58. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

1820

Morrice et al.

of source locations on the same source line between two adjacent receiver lines. In general, all sources in a single rack shoot
into the same patch. The survey design in Figure 5 is called
an orthogonal, split-spread design because there are an equal
number of active receiver lines on each side of the source rack.
CMPs that share more or less the same position are grouped
into predefined rectangular bins. The desired bin size is determined by the nature of the exploration target and depends
on the horizontal and vertical spatial resolution required, the
subsurface velocity profile and dip at the target, recoverable
frequency content, and economics. The economic constraint is
a result of the direct relationship between the natural bin size
and the source and receiver spacing. For an orthogonal design,
the natural bin dimensions are one half the source and receiver
spacing. Smaller bins, therefore, typically require tighter source
and receiver spacing and consequently more sources and receivers, thereby raising costs. During processing, the data in
each bin are averaged (or stacked) in order to improve the
S/N. The number of CMPs in each bin is called the fold of coverage. The source/receiver pairs for each bin have a range of
offsets (distance from shot point to receiver) and azimuths (direction or compass angle from shot point to receiver). Bin size,
fold of coverage, range of offsets, and range of azimuths are
the main measures by which the geophysical requirements of
a 3-D survey are assessed.
Operations on a land seismic survey involve the coordination of signal generation (shooting), placement of equipment
to receive the reflected signal, and recording of the results.
Land operations are performed by a seismic crew, and the logistics center of a survey is called the base camp. Line crews
continually move equipment from the back of the spread to
the front and place the equipment (channels) on the ground
in preparation for recording. Their vehicles are referred to as
line trucks in this paper. The equipment consists of cables, receivers, and boxes. In a modern acquisition system, receivers
collect the data which is then collated in the boxes and converted from an analog to a digital signal. Batteries and solar
panels power the boxes. The time required to prepare a single
receiver for recording and the number of receivers required is
typically referred to as the receiver effort.

Source crews (sets of vibrator or dynamite-shooting crews)


progress through a preplanned sequence of source locations.
The time required to acquire a single source is typically referred
to as the source effort. A central recording system, located at
a recording platform, turns different receivers on and off so
that the required patch is used for each source measurement.
The central recording system also records the results from each
shot.
Line crews move equipment in the direction of the progression of the survey. Typically, there is not enough equipment
available to cover the entire survey area. Thus, equipment
movement is required to enable recording of all the sources into
the correct active patch. A group of sources will be recorded
into the available spread and then recording operations may
be suspended while the receivers are moved to new locations
required for the next set of sources. Measurements from a single source point typically take less than one minute. Movement of receiving equipment from one patch to the next one
in the shooting sequence can take several minutes or even
hours. Therefore, to ensure efficient recording operations, extra
equipment is used to cover several patches ahead of the current source location. In order to avoid suspension of recording
operations, equipment must be removed from the back of the
spread once all the sources requiring a receiver in that location have been recorded and placed at the front of the spread
before receivers at the new location are required.
If the line crews regularly have to wait to remove equipment
from the back of the spread because the equipment is still in
use and they have already set out all the available equipment,
the seismic crew is said to be source limited. In this case, the
time required to complete the survey is determined by the time
required to acquire the sources. If a seismic crew must regularly suspend recording operations and wait for equipment to
be moved, it is said to be receiver limited. In this case, the time
required to complete the survey is determined by the time required to move the equipment.
The economics of any survey are ultimately controlled by
the time required to acquire it and the fixed daily costs of
the equipment used. An optimized survey meets the predetermined geophysical constraints at the lowest cost. This requires
the balancing of source and receiver effort to avoid a crew
being source or receiver limited, while maintaining the lowest possible fixed daily cost. The time required to record the
sources and move the equipment is determined by the source
effort and the logistics of the survey. Many factors influence
the logistics of a survey. Some factors such as size of the survey,
the design of the survey, the amount of equipment available,
and the number of line and source crews available are controllable. Other factors such as weather and terrain are not. It is
the controllable factors that provide the decision maker with
the opportunity to improve or even optimize survey design and
operations.
MATHEMATICAL PROGRAMMING AND OPTIMIZATION

FIG. 5. Patch and racks.

Mathematical programming techniques can be used when a


problem is formulated as a set of decision variables, a function to be minimized or maximized (called an objective function), and a set of constraints (Hillier and Lieberman, 1990).
Decision variables are quantities under the control of the decision maker. Their values are computed such that the objective

Downloaded 07/07/15 to 201.76.175.58. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

Optimizing Seismic Surveys

function is optimized and the set of constraints satisfied. Other


quantities in a mathematical program are not computed but set
to constant values. They are called problem data or parameters.
Parameters must be quantifiable. Mathematical programming
has been widely applied in operations problems in manufacturing, transportation and logistics, finance, and marketing.
Problems amenable to mathematical programming techniques abound in land seismic operations. Vehicle routing
problem formulations used in transportation could be applied
to the vehicles used in both moving the equipment and shooting operations. Models used in logistics for facility locations
could be used to determine the location of the base camp and
trucks carrying the recording system. Other problems include
crew shuttling to and from remote locations, line-crew scheduling, battery replacement strategies in the data recording boxes,
and the balancing of workloads between line crews, line trucks,
and source crews. Ultimately optimization modeling might be
used to design seismic survey strategies that would differ from
those specified by geophysical criteria alone. For example, a
seismic survey design might be developed by incorporating all
costs (including cost of operations) that would ensure a certain
fold coverage, offset distribution, and azimuthal distribution.
Consider the problem statement for survey design given in
Figure 6. The objective function can be broken down into several components: permitting costs, line clearance costs, drilling
costs, surveying costs, crew costs, and equipment costs. Geophysical constraints cover bin size, fold, offset, and azimuth
requirements. Examples of operational constraints include a
limit on the equipment available and on the maximum number
of shots acquired per day. Examples of decision variables include source and receiver density, and size of the active patch.
Problem data are quantities such as crew cost per day, equipment cost per channel, etc.
Models become harder to solve when the decision variables
are integer, the objective function and/or the constraints are
nonlinear, and the problem data are stochastically defined. For
integer programming problems, the computational complexity
of the solution grows exponentially in the number of integer
variables. No algorithm can solve all nonlinear programming
problems to optimality. The best any algorithm can do in the
general case is to guarantee local optimality (i.e., guarantee
the best solution over a local region of possibilities). In certain
special cases when the objective function has special shape or
structure (e.g., convexity), global optimality (the best solution
over all possibilities) can also be guaranteed. Accounting for
uncertainty in the data is a challenge for mathematical programming. Making assumptions to eliminate the uncertainty
may yield solutions far from optimality. Stochastic programming uses the structure of the mathematical model to furnish
exact solutions or simulation (Monte-Carlo sampling) to bound
the optimal solution.
A plethora of software exists for solving mathematical programming problems, including built-in optimizers available

in the most popular spreadsheet packages: Excel (Microsoft


Corporation, 1993), Lotus 1-2-3 (Lotus Development Corporation, 1994), and Quattro Pro (Campbell, 1993). For more details
on spreadsheet optimizers, see Fylstra et al. (1998). Spreadsheet optimizers are sufficient for small problems (less than
200 variables). For larger problems, packages such as the General Algebraic Modeling System (GAMS) (Brooke et al., 1992)
or A Mathematical Programming Language (AMPL) (Fourer
et al., 1993) are required. These packages are available for all
major commercial computing environments.
AN OPTIMIZATION MODEL FOR OPERATIONS PLANNING

Costing of 3-D land seismic survey operations is discussed


in Stone (1994, chapter 7). Some commercial software exists
for costing survey design and operations (see, for example,
Green Mountain Geophysics, Inc., 1995). In practice, spreadsheet costing models are commonly used. In this section, we
formulate a 3-D survey costing model as a mathematical program as stated in Figure 6. The objective is to minimize the total
acquisition costs in the design and the operations of the survey.
Source and receiver points are arranged in an orthogonal, splitspread design (see Figure 5). We solve the problem using the
Microsoft Excel optimizer.

Model formulation
The first step in formulating a mathematical programming
model is to decide what decisions can be made. On a land
seismic survey, the main decisions involve how to shoot the
survey and how to place the receiving equipment. These decisions must be made in the most cost-effective manner subject
to the geophysical and operations requirements of the survey
(see Figure 6). Table 1 contains definitions for the decision
variables in this problem.
Table 2 defines the symbols used for problem data in the
formulation of this problem. These quantities are assumed to

Table 1.

Decision variable definitions for the 3-D land seismic


survey operations optimization model.
Definition

Variable
SLS
RLS
RS
SS
NR
NL
NS
ND
N RD
N SB
N RB
N RL
HNL

FIG. 6. A mathematical programming problem statement for


optimizing operations in a 3-D land seismic survey.

1821

Shot line spacing


Receiver line spacing
Receiver spacing in a line
Shot spacing in a line
Number of active receiver channels
Number of active receiver lines
Number of shots/km2
Number of shots/day
Number of receiver channels moved/day
Number of shot locations between adjacent
receiver lines
Number of receiver locations between
adjacent shot lines
Number of active receiver channels/active
receiver line
One half the number of receiver lines (used
to constrain the number of receiver lines
to be even for the split-spread design)

Distances are in kilometers.

Downloaded 07/07/15 to 201.76.175.58. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

1822

Morrice et al.

be fixed and known. Unlike the decision variables, they are not
calculated by the model, rather they act as input to the model.
The next step is to define the objective function. We
start with an objective function: Minimize (total acquisition
costs/km2 ). The objective can be subdivided and restated
as: Minimize (preparation costs/km2 + shooting and recording
costs/km2 + processing costs/km2 ).
Preparation costs include permitting costs, surveying costs,
cutting and clearing costs, and drilling costs (in the case when
dynamite is used as the source). Permitting costs for 3-D surveys are generally a function of area and, hence, are constant
per square kilometer. They include the actual cost of the permits and the administrative costs to acquire them. Depending
on the complexity of the area, the administrative costs may
be higher than the permit costs, and they can therefore most
simply be modeled by an average price per unit area. In the
event that the reader wishes to use a line-mile cost basis, this
term could be added into the equation in the same way as line
clearance costs discussed below. For the sake of our example,
we have chosen to use a constant areal unit cost/km2 , CPER .
Surveying costs are a function of the number of source and
receiver locations. The survey costs/km2 can be written as

CSSV (1/SS) (1/S L S) + CRSV (1/R S) (1/R L S).


(1)
The cutting and clearing costs are functions of the length of
the source and receiver lines. This quantity may be written as

CSCC (1/S L S) + CRCC (1/R L S).

(2)

drilling environment are constant. Therefore, drilling costs/km2


may be written as

CDR (1/SS) (1/S L S).

(3)

For a survey on which the source is not dynamite, CDR is set


equal to zero.
Shooting and recording costs can be separated into two components: crew operating costs and equipment costs. We assume
that crew operating costs can be stated as an average daily rate.
Thus, crew operating costs/km2 are

CCR (NS/ND).

(4)

We assume that an average daily equipment cost can be obtained which varies with the number of receiver channels used
on the job. This can be stated as

CEQ (NR (1 + I )) (NS/ND).

(5)

The quantity I is an inflation factor to calculate the total number of channels used by the crew (active patch + spares). For
example, if I = 0.5, the total number of channels is 1.5 times
the number needed to cover the active patch.
The final cost component, the processing costs, are a function of the processing techniques used and the total number
of traces involved in the survey. The processing technique to
be used is not a variable in this equation, but determines the
per trace cost. The total cost of processing can, therefore, be
represented as a function of the size of the survey, the bin size,
and the fold of coverage. Hence, processing costs/km2 equals

F (1/Bx ) (1/BY ) CPCT .

Drilling costs are a function of the number of source locations assuming that the depth of the hole, size of the charge, and

(6)

By combining equations (1) through (6), the objective function can be written as
Table 2.

Problem data definitions for the 3-D land seismic


survey operations optimization model.

Parameter
CPER
CSSV
CRSV
CSCC
CRCC
CDR
CCR
CEQ
I
CPCT
BX
BY
F
OIN
OCR
OMAX
OMIN
SMAX
RMAX
CMAX

Definition
Permitting costs/km2
Surveying cost/source point
Surveying cost/receiver point
Cutting and clearing cost/km for source
line
Cutting and clearing cost/km for receiver
line
Drilling cost/source location
Average daily cost of the crew
Equipment cost/receiver channel per day
Inflation factor to calculate the total
number of receiver channels
Processing costs/trace
In-line bin dimension
Cross-line bin dimension
Required fold coverage
Required in-line offset
Required cross-line offset
Required maximum offset
Required maximum minimum offset
Maximum number of shots/day
Maximum number of receiver channels
moved/day
Maximum number of receiver channels
available

Costs are in U.S. dollars; distances are in kilometers.

Minimize CPER + (CSSV + CDR ) (1/SS) (1/S L S)


+ CRSV (1/RS) (1/R L S) + CSCC (1/S L S)
+ CRCC (1/R L S) + (CCR + CEQ (NR (1 + I )))
(NS/ND) + F (1/Bx ) (1/BY ) CPCT .

(7)

The constraints set includes constraints for the geophysical


requirements of fold, offset (maximum, and maximum minimum), and azimuth. The constraint for fold (see Stone, 1994,
125) is

NR NS Bx BY F 0.

(8)

In order to constrain azimuth, maximum in-line and crossline offset will be used. Although in-line and cross-line offsets
constrain azimuth in a crude manner, this approach is common
in practice in the field. These constraints are

Maximum in-line offset :


(NR/NL)/2 RS RS/2 OIN

(9)

and

Maximum cross-line offset: NL/2 R L S SS/2 OCR .


(10)

Optimizing Seismic Surveys

A constraint to ensure the same number of channels per active


receiver line is

The constraints for the offsets (see Stone, 1994, 129) are

Maximum: ((NR/N L)/2 R S RS/2)

Downloaded 07/07/15 to 201.76.175.58. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

+ (N L/2 R L S SS/2)2 (OMAX )2

(11)

and

Maximum minimum: (S L S)2 + (R L S)2 (OMIN )2 .


(12)
The following operational constraints are required:

Maximum number of shots/day: ND SMAX ,

(13)

Maximum number of receiver channels moved/day:


NR D RMAX ,

(14)

Shots and receiver channels moved/day:

(ND S L S SS)/(NR D R L S R S) = 1,

(15)

Maximum number of channels: NR(1 + I ) CMAX .


(16)
SMAX is controlled by the number of source crews. If ND =
SMAX , then the survey is source limited. Additionally, in vibroseis acquisition, SMAX is controlled by the sweep effort used
and the time required to move between successive sources and
adjacent source lines. RMAX is controlled by the number of
line crews and the time required by each crew to move the
equipment. If NRD = RMAX , then the survey is receiver limited. Therefore, changes in the value of constraints (13) and
(14) must be matched by a change in the value of CCR , the
average daily crew operating cost.
Equation (15) requires further explanation. This equation
provides an approximate relationship that specifies the number of receiver channels that need to be moved to support a
given number of shots/day. It links the number of shots/day to
the number of receivers channels moved/day using a relative
comparison of the shot density/km2 to the receiver density/km2 .
In addition, constraints are included to ensure that various
decision variables are consistently and practically defined. For
a symmetric split-spread design using natural bins, the relationship between shot and receiver spacing and the dimensions of
the natural bins are characterized by

R S = 2 Bx ,

(17)

SS = 2 BY .

(18)

and

The following constraint relates NS, SS, and S L S:

(19)

The distance between each receiver line must be an integer


number of shot locations:

R L S N S B SS = 0.

(20)

The distance between each shot line must be an integer number


of receiver locations:

S L S N R B R S = 0.

NR N L N R L = 0.

(21)

(22)

To constrain the number of receiver lines to be even for the


split-spread design, the following constraint is necessary:

N L 2 H N L = 0.

(23)

For other geometries, equations (17) through (23) may be


changed or omitted to suit the specific design. Additionally,
other constraints can be added. As long as a constraint can
be stated as a mathematical function of the decision variables,
then it can be added to the constraint set of a mathematical
program.
Finally, all decision variables are constrained to be positive,
and some are constrained to be integer valued, i.e.,

S L S, R L S, RS, SS, NR, N L , NS, ND,


N S B, N R B, N R L, H N L 0,

and

NS SS S L S = 1.

1823

(24)

NR, N L , NS, ND, N S B, N R B, N R L,


and H N L are integer valued.

(25)

The complete statement of the mathematical program for


this problem is minimize expression (7) subject to the constraints in expressions (8) through (25). This mathematical program is an integer, nonlinear program. As mentioned above,
various algorithms exist for finding local optimal solutions to a
problem of this form, but these algorithms do not guarantee a
global optimal solution. We demonstrate solving this problem
using the Solver in Excel.
Solution using Excel
Figure 7 contains an Excel spreadsheet of the survey design
and operations model. Notice that it is subdivided into four
parts: problem data, decision variables, objective function, and
constraints. Immediately below the objective function, we report the production rate in km2 /day since, in addition to cost,
this performance measure is often of interest to decision makers. The problem data and the initial values of the decision variables are typical for a dynamite job. The corresponding spreadsheet formulas are given in Figure 8. The notation LHS and
RHS stands for left-hand side and right-hand side, respectively, of the constraint expressions given in the complete statement of the mathematical program given above.
To optimize the survey design and operations problem in
Excel, one must invoke the Solver from the Tools menu. Once
invoked, a Solver Parameters dialog box appears (see Figure 9)
that permits the user to specify the objective function (target
cell), decision variables (changing cells), and the constraints.
Figure 10 illustrates the Excel spreadsheet after running the
Excel Solver. Notice how the decision variables have changed,
resulting in a decrease in the objective function value by almost $1300/km2 . The most significant change is due to the fact
that spacing between the receiver lines has decreased and the
spacing between the shot lines has increased, resulting in more
receiver channels and fewer shots/km2 . Adding extra receiver
channels to compensate for a reduction in shot points is a classic trade-off in land seismic operations. In light of the problem
data, substituting receiver channels for shot points makes sense

Downloaded 07/07/15 to 201.76.175.58. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

1824

Morrice et al.

data because each additional receiver channel costs $3, whereas


each shot hole costs $100 to drill.
In practice, trial and error or scenario generation would be
used to move from the initial design in Figure 7 to the improved
results in Figure 10. Depending on the initial decision variable
values and the number of alternatives, such an analysis might
take several minutes or hours, perhaps longer, making it infeasible. Using the Solver, the trial and error effort is greatly reduced although not eliminated because the mathematical program is nonlinear and converges to a local optimal value from
an initial solution. The results in Figure 10 were generated by
multiple (in this case, three) runs of the Solver from the initial
solution in Figure 7 until the solution converged. Each run took
a few seconds of real clock time.
The maximum number of shots/day and the maximum number of receiver channels moved/day are the main production
parameters in this model controlling the length and hence the

cost of the job. Therefore, it is instructive to perform sensitivity analysis on these parameters. In the interest of brevity,
we will illustrate sensitivity analysis on the maximum number
of shots/day only. Figure 11 contains results when the maximum number of shots per day is 130 (a) and 100 (b). Although
changes in this parameter do not impact the value of the decision variables, decreasing daily production does have a significant impact on the cost of the job.
CONCLUSIONS

The model we have developed in this paper complements


and strengthens existing heuristic approaches used for operations planning in the field and the existing geophysical modeling techniques. Although the model is not all encompassing,
it illustrates the application of mathematical programming to
survey design. Sensitivity analysis provides a way to assess the
impact of changes to model parameters on the optimal solution.
The model aggregates many of the details found in a 3-D
land seismic survey and, in this way, simplifies reality. However, such aggregation is often necessary in order to make
the model solution procedure computationally tractable. Additionally, aggregated models are sufficient for planning purposes
in most cases. More details would be necessary if, for example,
one needs to model day-to-day operations.
Many of the assumptions used in the model can be relaxed
and the problem formulation extended to address more complex issues. For example, operationally, the problem could
be extended to account for the possible procurement of
more shooting and recording equipment and the associated
cost/benefit of doing so. Additionally, it would be possible to
include a calculation for the value of SMAX assuming an input vibroseis sweep effort, move-up time, line change time
and frequency, terrain and weather factors, and anticipated
equipment failure rate. Including this extension would make
the model more complicated, but it would provide a more
realistic model for production and operations than presuming a fixed production maximum. Equations of this nature are
commonplace in seismic acquisition contractors existing bidpreparation spreadsheets already. A similar argument could be
made for RMAX .
The model could also be extended to include more geophysical details like dip and velocity to determine the bin dimensions
(see Stone, 1994, 124) rather than including them as fixed problem data. Additionally, more geophysical constraints based on
a subset of the full patch could be defined to address multiple
targets. Finally, similar modeling formulations could be derived
for other types of designs, such as brick or zig-zag designs (see,
for example, Ashton et al., 1994). In fact, similar formulations
could be used for any design that has receiver and source locations that can be defined on an orthogonal grid.
ACKNOWLEDGMENTS

The authors thank Andy Peters, Peter Cantor, and Peter


Highnam of Schlumberger for their assistance and support on
this project.
REFERENCES

FIG. 7. Excel spreadsheet for dynamite job with original values.

Ashton, C. P., Bacon, B., Mann, A., Moldoveanu, N., Deplant

e,
C.,
Ireson, D., and Redekop, G., 1994, 3-D seismic survey design:
Schlumberger Oilfield Review, 6, No. 2, 1932.
Brooke, A., Kenrick, D., and Meeraus, A., 1992, GAMS: A users guide,
second edition: The Scientific Press.

Downloaded 07/07/15 to 201.76.175.58. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

Optimizing Seismic Surveys

1825

FIG. 8. Excel spreadsheet formulas for the optimization model.

FIG. 9. Solver settings for the optimization model.


Campbell, M., 1993, Quattro Pro 5.0 for Windows handbook: Borland
Press.
DeGolyer, E., 1947, Notes on the early history of applied geophysics in
the petroleum industry, in SEG Staff, Ed., Early geophysical papers,

1: Soc. Expl. Geophys., 245254.


Dobrin, M. D., and Savit, C. H., 1988, Introduction to geophysical
prospecting: McGraw-Hill Book Co.
Fourer, R., Gay, D. M., and Kernighan, B. W., 1993, AMPL: A
modeling language for mathematical programming: The Scientific
Press.
Fylstra, D., Lasdon, L., Watson, J., and Waren, A., 1998, Design and
use of the Microsoft Excel Solver: Interfaces, 28, 2955.
Green Mountain Geophysics, Inc., 1995, Mesa Geocost users manual:
Green Mountain Geophysics, Inc.
Hillier, F. S., and Lieberman, G. J., 1990, Introduction to operations
research, fifth edition: McGraw-Hill Book Co.
Horman, K., Vermeer, G., Lansley, M., Musser, J. A., Gilbraith, M.,
Meunier, J., Gillot, E., Monk, D., and Yates, M., 2000, 3-D seismic
survey design: First Break, 18, 161185.
Lotus Development Corporation, 1994, Lotus 1-2-3 spreadsheet for
Windows users guide, release 5: Lotus Development Corp.
Liner, C. L., Gobeli, R., and Underwood, W. D., 1998, 3-D seismic
survey design as an optimization problem: 68th Ann. Internat. Mtg.,
Soc. Expl. Geophys., Expanded Abstracts, 117120.
Microsoft Corporation, 1993, Microsoft Excel users guide, version 5.0:
Microsoft Corp.
Stone, D. G., 1994, Designing seismic surveys in two and three dimensions: Soc. Expl. Geophys.

Downloaded 07/07/15 to 201.76.175.58. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

1826

Morrice et al.

FIG. 11. Sensitivity analysis on maximum number of shots.


FIG. 10. Excel spreadsheet for dynamite job with optimized
values.

You might also like