Improvements to Reservoir
Material-Balance Methods
J.L. Pletcher,* SPE, Marathon Oil Co.
Summary
Experience with material-balance data sets from the field and from
simulation has revealed some procedures that can be used to im-
prove analysis of both oil and gas reservoirs:
• Failure to account for a weak waterdrive can result in sig-
nificant material-balance errors.
• The assertion of previous authors that weak waterdrive ex-
hibits a negative slope on the Cole (gas) and Campbell (oil) plots
has been confirmed. A weak waterdrive is much more unambigu-
ous on these plots than on commonly used plots, such as the p/z
plot for gas.
• A modified version of the Cole plot is proposed to account
for formation compressibility.
• The reservoir drive indices are a useful tool for determining
the correctness of the material-balance solution because they must
sum to unity. The drive indices should never be normalized to sum
to unity because this obscures their usefulness and leads to a false
sense of security.
• A modified version of the Roach plot (for gas) is proposed
that improves interpretation in some waterdrive situations.
• Material balance has not been replaced by reservoir simula-
tion; rather, it is complementary to simulation and can provide
valuable insights to reservoir performance that cannot be obtained
by simulation.
Introduction
Classical material balance is one of the fundamental tools of res-
ervoir engineering. Many authors have addressed the difficult
problem of solving the material balance in the presence of a wa-
terdrive (Refs. 1 through 5 are just a few of the more significant
ones). The emphasis in the literature has been on strong and mod-
erate waterdrives. In this paper, examples of weak waterdrives
are shown in which the effects on the material balance are signifi-
cant. All aquifers studied here are of the “pot aquifer” type, which
is time-independent.
In gas reservoirs, the plot of p/z vs. cumulative gas production,
Gp, is a widely accepted method for solving the gas material bal-
ance1 under depletion-drive conditions. Extrapolation of the plot to
atmospheric pressure provides a reliable estimate of original gas in
place (OGIP). If a waterdrive is present, the plot often appears to
be linear, but the extrapolation will give an erroneously high value
for OGIP. Many authors have addressed this problem (including
those in Refs. 2 and 5 through 8), especially in cases of strong or
moderate waterdrives. The p/z plot is actually more ambiguous in
weak waterdrives than in strong or moderate ones.
The Cole plot7,9 has proven to be a valuable diagnostic tool for
distinguishing between depletion-drive gas reservoirs and those
that are producing under a waterdrive. The analogous plot for oil
reservoirs is the Campbell plot.10 The literature has emphasized
strong and moderate waterdrives, the signature shapes of which are
a positive slope and a hump-shaped curve, respectively, on these
plots. Previous authors have recognized that weak waterdrives can
produce negative slopes on these two diagnostic plots, but this
* Now retired.
Copyright © 2002 Society of Petroleum Engineers
This paper (SPE 75354) was revised for publication from paper SPE 62882, first presented
at the 2000 SPE Annual Technical Conference and Exhibition, Dallas, 1–4 October. Original
manuscript received for review 7 December 2000. Revised manuscript received 10
September 2001. Paper peer approved 1 October 2001.
February 2002 SPE Reservoir Evaluation & Engineering
author is not aware of any example plots in the literature. This
paper shows examples, using simulation and actual field data,
wherein a negative slope clearly reveals a weak waterdrive. These
plots are much more diagnostic than the p/z plot. Once a weak
waterdrive has been diagnosed, the appropriate steps can be taken
in the material-balance equations to yield more accurate results.
The Cole plot assumes that formation compressibility can be
neglected, which is frequently the case with gas. However, in those
reservoirs in which formation compressibility is significant, a
modification to the Cole plot is presented that incorporates forma-
tion compressibility and gives more accurate results.
The reservoir drive indices have been used to quantify the
relative magnitude of the various energy sources active in a res-
ervoir. It is shown here that the drive indices are also a useful
diagnostic tool for determining the correctness of a material-
balance solution because they must sum to unity. If they do not
sum to unity, a correct solution has not been obtained. In some
commercial material-balance software, the drive indices are auto-
matically normalized to sum to unity, which not only obscures
their usefulness but also leads to the false impression of having
achieved a correct solution.
The Roach plot has been presented11 as a tool for solving the gas
material balance when formation compressibility is unknown, with or
without the presence of waterdrive. This paper shows that for water-
drives that fit the small pot aquifer model, incorporating cumulative
water production into the x-axis plotting term improves the linear-
ity of the Roach plot and gives more accurate values for OGIP.
Finally, it is argued that even in those reservoirs for which a
simulation study is performed, classical material-balance evalua-
tion should be performed on a stand-alone basis. Simulation should
not be viewed as a replacement for material balance because the
latter can yield valuable insights that can be obscured during simu-
lation. Performing a separate material balance study usually will
improve overall reservoir understanding and enhance any subse-
quent simulation study. Material balance should be viewed as a
complement to simulation, not as a competing approach.
In this paper, formation compressibility, cf, is assumed to be
constant and unchanging over the reservoir life under investiga-
tion. References are given for recommended methods to be used in
those cases in which cf is variable.
Gas Reservoirs
Cole Plot. The Cole plot7,9 is a useful tool for distinguishing be-
tween waterdrive and depletion-drive gas reservoirs. The plot is de-
rived from the general material-balance equation for gas reservoirs:
F = G共Eg + Efw兲 + We , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ( 1)
where F⳱cumulative reservoir voidage,
F = Gp Bg + Wp Bw ; . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ( 2)
Eg⳱cumulative gas expansion,
Eg = Bg − Bgi ; . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ( 3)
and Efw⳱cumulative formation and water expansion,
Swi cw + cf
Efw = Bgi 共pi − p兲. . . . . . . . . . . . . . . . . . . . . . . . . . . . . ( 4)
1 − Swi
In Eq. 1, G⳱OGIP, and We⳱cumulative water influx. Often in
gas reservoirs, Efw is negligible compared to Eg and can therefore
49
be ignored. Then, by substitution and rearranging, Eq. 1 can be
expressed as
Gp Bg We − Wp Bw
= G+ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ( 5)
Bg − Bgi Bg − Bgi
Gp Bg
Cole proposed plotting the left side of Eq. 5, , on the
Bg − Bgi
y-axis vs. cumulative gas production on the x-axis. If the reservoir
is depletion drive (i.e., no water influx), the term on the far right
side of Eq. 5 goes to zero and the points plot in a horizontal line
with the y-intercept equal to G, the OGIP. If a waterdrive is
present, the far right-side term is not zero, and the points will plot
above the depletion-drive line with some type of slope. In other
words, the existence of a sloping line vs. a horizontal line is a
valuable diagnostic tool for distinguishing between depletion drive
and waterdrive.
Cole and others have suggested that the sloping waterdrive line
can be extrapolated back to the y-intercept to obtain the OGIP.
However, the slope usually changes with each plotted point; thus,
the correct slope for extrapolation is very difficult, if not impos-
sible, to establish, so this method for estimating OGIP is not rec-
ommended. This does not, however, detract from the plot’s quali-
tative value in establishing that the reservoir is under waterdrive,
as opposed to depletion drive.
Dake7 showed two types of curving Cole plots in his Fig. 6.6,
a strong waterdrive curve and a moderate waterdrive curve, de-
picted here in Fig. 1. (Actually, Dake’s plot is a slightly modified
version of Cole’s plot because Dake incorporated water production
into the y-axis plotting term; that is, he in effect moved the water-
production term to the left side of Eq. 5. The net effect on the curve
shapes is negligible.)
Wang and Teasdale12 stated that in the presence of a weak
We − Wp Bw
waterdrive, the far right-side term in Eq. 5, , would
Bg − Bgi
decrease with time because the denominator (gas expansion)
would increase faster than the numerator (net water influx). There-
fore, the plotted points will exhibit a negative slope, as shown in
Fig. 1; indeed, this has been observed in practice, as will be shown
later in this paper. As reservoir depletion progresses, the points
migrate down and to the right toward the true OGIP; the smaller
the aquifer, the closer the plot will approach the true OGIP.
Note that the negative slope of the weak waterdrive curve rep-
Gp Bg
resents an unexpected anomaly. The y-axis plotting term
Bg − Bgi
amounts to the apparent OGIP that would be calculated, assuming
no waterdrive is present. Therefore, under a weak waterdrive, the
apparent OGIP decreases with time, contrary to that for a strong or
moderate waterdrive.
Actually, before developing the signature negative slope, the
weak waterdrive curve begins with a positive slope in the very
early stages of reservoir depletion, as shown in Fig. 1. The very
Strong
GpBg / (B gÐBgi ) , Mscf
Waterdrive
Moderate
Waterdrive
Weak
Waterdrive
OGIP
Depletion drive
Gp , Mscf
Fig. 1—Cole plot curve shapes as a function of aquifer strength.
50
early points are difficult to use for determining OGIP, however,
because they frequently exhibit a great deal of scatter that is in-
troduced by even small errors in pressure measurement early in the
reservoir life. Technically, then, the curve is hump-shaped like
Dake’s moderate waterdrive curve in Fig. 1, except that the posi-
tive-slope portion of the hump is over with very early and in
practice will not show up at all unless frequent and accurate very
early time data are obtained.
Modified Cole Plot. In some gas reservoirs, formation compress-
ibility is not negligible, in which case Efw should not be ignored
and Eq. 5 should be written:
F We
= G+ , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ( 6)
Et Et
where Et⳱total reservoir expansion,
Et = Eg + Efw . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ( 7)
F
The left side of Eq. 6, , now incorporates in the denominator
Et
the energy contribution from formation (and water) compressibil-
ity, as well as from gas expansion. The modified Cole plot consists
F
of plotting on the y-axis vs. Gp on the x-axis. Curve shapes will
Et
be the same as in Fig. 1. Vertically, the points will lie closer to the
true value of OGIP than the original Cole plot.
In reservoirs in which formation compressibility (cf) is a sig-
nificant contributor to reservoir energy, such as abnormally pres-
sured reservoirs, the original Cole plot will exhibit a negative
slope, even if no waterdrive is present. The modified Cole plot,
however, will plot in a horizontal line, assuming the correct value
F
of cf is used in calculating the term. Thus, constructing both
Et
the original and modified Cole plots will distinguish between those
reservoirs that are subject to both a weak aquifer and significant
formation compressibility and those reservoirs in which formation
compressibility is significant but there is no aquifer attached; for
the former, both plots will have a negative slope, and for the latter,
the original Cole plot will have a negative slope while the modified
plot will be horizontal. This assumes, of course, that formation com-
pressibility is known with certainty, which is often problematical.
Actually, negative slopes on the original and modified Cole
plots can result from any unaccounted-for source of energy that is
decreasing with time relative to gas expansion. This could include,
for example, communication with other depleting reservoirs.
Drive Indices. Drive indices have been defined for oil reservoirs13
to indicate the relative magnitude of the various energy forces
contributing to the reservoir. Similarly, drive indices can be de-
fined for gas reservoirs as follows.
Gas drive index:
GEg
IGD = . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ( 8)
Gp Bg
Formation and connate water compressibility drive index:
GEfw
ICD = . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ( 9)
Gp Bg
Waterdrive index:
We − Wp Bw
IWD = . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ( 10)
Gp Bg
The numerators of these three dimensionless fractions represent
the cumulative gas expansion, cumulative rock and connate water
expansion, and cumulative net water influx, respectively, all at
reservoir conditions. The common denominator is the cumulative
hydrocarbon voidage at reservoir conditions. If the material bal-
February 2002 SPE Reservoir Evaluation & Engineering
 TABLE 1ÑPROPERTIES OF TWO-CELL TABLE 2ÑPERFORMANCE HISTORY OF TWO-CELL
GAS-SIMULATION MODEL GAS-SIMULATION MODEL

Node area 640 acres Cumulative Cumulative

Node thickness = net pay thickness 200 ft Gas Water Cumulative

Porosity 15% Pressure Produced Produced Water Influx

Year (psia) (Bscf) (STB) (STB)
Gas reservoir pore volume 74.5×10 6 res bbl
Aquifer original water in place 74.5×10 6 res bbl 0 6,411 0.000 0 0

Sw 15% 1 5,947 5.475 378 273,294

OGIP 100.8 Bcf 2 5,509 10.950 1,434 552,946

Permeability 100 md 3 5,093 16.425 3,056 817,481

6×10Ð6 psi Ð1
4 4,697 21.900 5,284 1,068,632
cf
cw 3×10Ð6 psi Ð1 5 4,319 27.375 8,183 1,307,702

Reservoir temperature 239¼F 6

7
3,957

3,610
32.850

38.325
11,864

16,425
1,535,212

1,752,942

8 3,276 43.800 22,019 1,962,268

ance has been solved correctly, the sum of these three fractions 9 2,953 49.275 28,860 2,163,712

equals unity. 10 2,638 54.750 37,256 2,359,460

IGD + ICD + IWD = 1. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ( 11)

If the drive indices do not sum to unity, a correct material-balance
solution has not been obtained.
In practice, drive indices calculated from actual field data rarely
sum exactly to unity because the data are not perfect. The summed
drive indices typically vary between values somewhat larger than
unity and somewhat smaller, with the degree of deviation from
unity a function of data quality. However, if the summed drive
indices are consistently greater than or less than unity, or show a
consistent increasing or decreasing trend, this is an indication that
a correct solution to the material balance has not been obtained.
Gas-Simulation Model. A simple two-cell gas model was con-
structed with the Eclipse14 reservoir simulator to study the effects
of weak water influx on gas reservoir material balance. One cell
contained gas at irreducible water saturation (i.e., a “tank” model
ideally suited to material-balance analysis), and the other cell con-
tained an equal pore volume containing 100% water saturation.
OGIP was approximately 101 Bcf. A single well was produced at
a rate of 15 MMscf/D for 10 years, recovering a little more than
one half (54.3%) of the OGIP. Other properties of the model are
found in Table 1.
The simulator output at 1-year intervals was used to perform a
material-balance evaluation of the reservoir. Production and pres-
sure histories used in the material balance are given in Table 2, and
pressure/volume/temperature (PVT) properties are given in Table 3.
The p/z plot is shown in Fig. 2, where each point represents year-end
conditions for Years 1 through 10. Because formation compress-
TABLE 3ÑPVT DATA FOR TWO-CELL
GAS-SIMULATION MODEL
ibility in this case is significant, the p/z plotting term was modified
to account for the energy contribution from rock compressibility
with a method equivalent to that of Ramagost and Farshad.15
The plotted p/z points in Fig. 2 appear to lie on an almost
perfect straight line (R2⳱0.9998 after 10 years), giving the im-
pression that an extrapolation to OGIP could be made with con-
fidence. However, an extrapolation of the points made after 2
years, when 11% of the true OGIP had been produced, would yield
a value for OGIP of 109 Bcf, or 8.2% too high. After 5 years, the
error would be +6.5%. Even after 10 years and recovery of 54% of
the OGIP, the error would still be +4.0%. Errors of this magnitude
are not insignificant, even though this aquifer is very small.
Existence of a waterdrive would be practically impossible to
detect from well performance because even after 10 years, the well
made only 1.5 STB of water per MMscf of gas. In the simulation,
the well produced water only because the encroached water is
dispersed uniformly throughout the single cell. In an actual reser-
voir, the well would likely produce less water because of saturation
gradients, depending on the proximity of the well with regard to
the original gas/water contact.
Cole and Modified Cole Plots. The Cole plot (Fig. 3) for this
weak aquifer data set exhibits a negative slope. The plot corrobo-
rates Wang and Teasdale’s contention that the Cole plot clearly
indicates the presence of even a weak waterdrive, whereas the p/z
6,000 After 2 yrs.
(Cum.=11%)
5,000 G=109 Bcf
p/z (modified), psia

Pressure Gas Deviation After 5 yrs.

Year (psia) Factor, z Bg (RB/Mscf) Bw (RB/STB) 4,000 (Cum.=27%)
G=107 Bcf
0 6,411 1.1192 0.6279 1.0452
3,000 After 10 yrs.
1 5,947 1.0890 0.6587 1.0467 (Cum.=54%)
2 5,509 1.0618 0.6933 1.0480 G=105 Bcf
2,000
3 5,093 1.0374 0.7327 1.0493

4 4,697 1.0156 0.7778 1.0506

1,000
5 4,319 0.9966 0.8300 1.0517

6 3,957 0.9801 0.8910 1.0529 0

7 3,610 0.9663 0.9628 1.0540
0 20 40 60 80 100
8 3,276 0.9551 1.0487 1.0551 G p , Bcf
9 2,953 0.9467 1.1532 1.0560

10 2,638 0.9409 1.2829 1.0571 Fig. 2—Modified p/z plot from output of two-cell gas simulation.
Actual OGIP=101 Bcf.

February 2002 SPE Reservoir Evaluation & Engineering 51

 120
1.18×108 After 2 yrs.
(Cum.=11%)
Original =105 Bcf

GpBg /(Bg ÐBgi ) or F/E t , Bcf

G
1
After 5 yrs. 2
115
(Cum.=27%) 3
Modified 1.12×108 G =102 Bcf 4
After 10 yrs.

Mscf
5
(Cum.=54%) 6
=101 Bcf
110
7

F/Eg ,
G
8
1.06×108 10
9

105

G = End of
True OGIP = 100.8 Bcf
y -intercept year
100 1.00×108
0 4,000 8,000 12,000 16,000
psi-Mscf/RB
0 20 40 60 80
∆P/E ,
G p , Bcf
g

Fig. 4—Pot aquifer plot for two-cell gas simulation.

Fig. 3—Cole plot, original and modified, for two-cell gas simulation.

pressibility, or even initial water saturation. The sequence of plot-

plot is completely ambiguous. The negative slope distinguishes the
ted points will be from right to left.
weak waterdrive system from the strong waterdrive (positive
The slope of the pot aquifer plot is given by the term in brackets
slope), moderate waterdrive (hump-shaped), and depletion-drive
in Eq. 13. The water in place in the aquifer, W, can be calculated
(horizontal line) systems (Fig. 1).
from the slope if, in fact, cf is known with some degree of confi-
Note that the ordinate values plotted in Fig. 3 appear to be
dence. Rearranging the slope term,
migrating toward the true OGIP value, 101 Bcf, as reservoir deple-
tion proceeds. Thus, the most recent plotted point on the Cole plot GBgi
could be taken as the maximum possible value of OGIP, approxi- A− 共S c + c 兲
mately 107 Bcf after 10 years (Gp⳱54% of OGIP). 1 − Swi wi w f
W= , . . . . . . . . . . . . . . . . . . . . . . . . . . ( 14)
Because formation compressibility is significant in this ex- cw + cf
ample, the modified Cole plot should be used. As expected, the
points lie closer to the true value of OGIP than the original Cole where G and the slope are obtained from the least-square fit
plot, Fig. 3. The ordinate value after 10 years is 104.4 Bcf, more straight line.
nearly approaching the true OGIP than the original Cole plot. Fig. 4 shows the pot aquifer plot for the two-cell gas-simulation
Included in Fig. 3 are values from the first year of production example. As before, each plotted point (large symbols) represents
at 1-month intervals, plotted with smaller symbols. The early time conditions at the end of each year. Also shown are straight lines
points exhibit a steep positive slope. The negative slope develops fitted to the data using the least-squares method, assuming that
after approximately 10 months, when 4.5% of the true OGIP has analyses had been performed at several times during the reser-
been produced. To obtain this early-time portion of the plot in an voir’s history (after 2, 5, and 10 years). Values of OGIP are ob-
actual reservoir, it would be necessary to obtain frequent and very tained from extrapolation of those straight lines to the y-intercept.
accurate pressure measurements. Even then, the points do not plot Typically with this plot, the early-time points fall below the true
with a constant slope, rendering impractical the extrapolation back straight line that eventually develops, and such is the case with this
to Gp⳱0 for the purpose of obtaining OGIP. data. After 2 years of performance, an analysis would consist only
of points from Years 1 and 2, and the true straight line would not
Pot Aquifer Plot. If the aquifer is relatively small and in good yet be apparent, giving a value for OGIP approximately 4% too
communication with the hydrocarbon reservoir, and permeabilities high. Analyses conducted after 5 and 10 years would likely have
are sufficiently high, the aquifer can be represented with the pot excluded the Year 1 data from the least-square fit. In all cases, the
aquifer model, and original hydrocarbons in place (OHIP) can be OGIP values are significantly closer to the actual value of 101 Bcf
obtained from the pot aquifer plot1,16 or by using Tehrani’s than the corresponding values obtained from the p/z plot (Fig. 2).
method.4 This type of aquifer should apply in high-permeability In Fig. 4, data points during the first year are plotted at 1-month
reservoirs having a hydrocarbon/water contact where the “water intervals with the smaller symbols. These points have a negative
leg” is isolated from large regional aquifers by permeability pinch- slope and do not start “turning over” toward the correct positive
out or faulting. Examples would be found in the U.S. Gulf Coast, slope until approximately three-quarters of the way through the
where high-permeability sands typically are broken up into rela- year. This is typical of the pot aquifer plot; therefore, the plot may
tively small reservoirs by faulting. An example from the U.S. not be usable in the very early life of the reservoir.
midcontinent is shown later in the paper. Table 4 summarizes the OGIP values obtained using the three
For the pot aquifer model, any drop in reservoir pressure is instanta- evaluation methods (modified p/z, modified Cole, and pot aquifer),
neously transmitted throughout the entire aquifer. Mathematically, as well as the percent errors. Even the modified Cole plot solution
is closer to the true OGIP than the p/z plot. The reason the modi-
We = 共cw + cf兲W共pi − p兲, . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ( 12) fied Cole plot is so near the true OGIP is that the aquifer is so small
for this example. For larger aquifers, neither the original nor the
where W⳱aquifer original water in place (OWIP), res bbl.
Substituting Eq. 12 for We and Eq. 4 for Efw in Eq. 1 and then
rearranging yields an equation of a straight line: TABLE 4ÑMATERIAL-BALANCE RESULTS ON TWO-CELL

冋 册
GAS-SIMULATION MODEL
F pi − p GBgi 共Swicw + cf兲
= G+
1 − Swi
+ 共cw + cf兲W . . . . . . . . . . ( 13)
% of Modified
for Modified Cole Plot Pot Aquifer Plot
p/z
Eg Eg
OGIP cf

F pi − p Produced % error G G % error G % error

Plotting on the y-axis vs. on the x-axis yields a straight
Eg Eg 11 109.0 8.2 <108.9 <8.0 105.3 4.5
line with the y-intercept equal to G. This is the pot aquifer plot. 27 107.3 6.5 <107.2 <6.3 101.6 0.8
The value of this plot is that it permits determination of OGIP 54 104.8 4.0 <104.4 <3.6 101.0 0.2
without any prior knowledge of aquifer size, rock or water com-

52 February 2002 SPE Reservoir Evaluation & Engineering

 TABLE 5ÑDRIVE INDICES AFTER 5 YEARS, TWO-CELL TABLE 6ÑDRIVE INDICES AFTER 10 YEARS, TWO-CELL
GAS-SIMULATION MODEL GAS-SIMULATION MODEL

Modified p/z Solution Pot Aquifer Solution Modified p/z Solution Pot Aquifer Solution

Year IGD ICD Total IGD IWD ICD Total* Year IGD ICD Total IGD IWD ICD Total

1 0.916 0.066 0.982 0.868 0.073 0.062 1.003

1 0.895 0.064 0.959 0.863 0.080 0.062 1.005*
2 0.924 0.061 0.985 0.875 0.068 0.057 1.000
2 0.903 0.059 0.962 0.870 0.074 0.057 1.001
3 0.934 0.056 0.990 0.885 0.062 0.053 1.000
3 0.912 0.055 0.967 0.880 0.068 0.053 1.000
4 0.944 0.051 0.996 0.894 0.057 0.049 1.000
4 0.922 0.050 0.972 0.889 0.062 0.048 0.999
5 0.954 0.047 1.001 0.904 0.052 0.045 1.000
5 0.932 0.046 0.978 0.898 0.057 0.044 1.000
*Excluded from least-square fit.
6 0.942 0.042 0.984 0.908 0.052 0.040 1.000
7 0.951 0.038 0.989 0.917 0.047 0.037 1.000
8 0.960 0.034 0.994 0.925 0.042 0.033 1.000
modified Cole plot will give a value so close to the true OGIP as 9 0.969 0.030 0.999 0.934 0.037 0.029 1.000
in this example.
10 0.977 0.027 1.004 0.942 0.033 0.026 1.001
The slope of the solution line in Fig. 4 after 10 years is 1,103
RB/psi, giving a calculated W of 69.1 million res bbl using Eq. 14, *Excluded from least-square fit.

some 7% low compared to the true value of 74.5 million res bbl.
Cumulative water influx can be calculated from Eq. 12 as
2,346,000 res bbl after 10 years, approximately 6% less than the
2,494,000 res bbl from the simulation. Accuracy of the calculated
W and We would be improved by excluding from the least-squares
fit additional early data points after Years 2 and 3 that deviate
slightly from the true straight-line trend. However, when analyzing
actual field data, such subtle deviations are difficult to detect ow-
ing to normal data scatter.
Note that if in fact there is no aquifer, the pot aquifer plot still
applies. In this case, W goes to zero in Eq. 13. The formation
compressibility can then be calculated from the slope:
1 − Swi
cf = A − Swi cw . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ( 15)
GBgi
If for this data set it had been assumed that no aquifer were present,
a cf of 14.3×10−6 psi−1 would have been calculated from Eq. 15,
significantly larger than the “known” value of 6×10−6 psi−1. In a
real-world setting, this would be another indication that an unac-
counted-for energy source is present. Case 1 of Wang and Teas-
dale12 shows an application of this method to an actual reservoir
believed to have no waterdrive.
Drive Indices. Drive indices were calculated for the two-cell
simulation model, assuming that the OGIP obtained from the modi-
fied p/z solution was correct, and compared with drive indices calcu-
lated with the more accurate pot aquifer solution. Table 5 com-
pares the two calculations after 5 years of performance, and Table 6
compares the two after 10 years of performance. Drive indices for
the pot aquifer solution add up to unity as expected, except for
Year 1, which was excluded from the least-square solution fit.
For the incorrect p/z solution that does not account for the
aquifer, drive indices do not add up to unity until later in the
respective time periods. This would be an indication to the engi-
neer making the analysis that his solution is incorrect. Therefore,
the criterion of whether the drive indices sum to unity is an indi-
cator of the correctness of the material-balance solution. This point
is made because some commercial material-balance computer pro-
grams normalize the drive indices, which forces them to sum to
unity. This practice is counterproductive because it deprives the
engineer of a tool for evaluating the correctness of his solution and
gives the false impression that a valid solution has been obtained.
Only the raw calculated drive indices should be reported and
summed; they should never be normalized. This applies regardless
of the aquifer model being fitted to the reservoir.
greater than unity and sometimes less than unity, as opposed to a
consistently increasing trend.
To summarize, evaluation of this reservoir taking the common
approach of considering only the p/z method (modified to include
cf effects) would, on the surface, give every indication that a cor-
rect material-balance solution had been obtained for depletion
drive. Yet OGIP would be erroneously high, with the error ranging
from approximately 4 to 8%, depending on the stage of reservoir
depletion considered. Constructing the modified Cole plot or cal-
culating drive indices would signal that the solution was, in fact,
not correct. The Cole plot, original or modified, indicates unam-
biguously that a weak waterdrive exists, in which case the pot aquifer
plot should be used to calculate the most accurate value of OGIP.
Oklahoma Morrow Gas Reservoir. Production history and
other data for an Oklahoma Morrow sand gas reservoir are given
in Table 7. The lack of water production, together with the decline
in reservoir pressure, suggested that no aquifer was present. The
p/z plot, Fig. 5, also gives no hint of aquifer support. The modified
p/z extrapolation gives G⳱6.02 Bcf. (Note that even though cf is
only 3×10−6 psi−1, extrapolation of the conventional p/z that ig-
nores cf gives G⳱6.32 Bcf, some 5% greater.)
The Cole and modified Cole plots are shown in Fig. 6 and
exhibit the characteristic negative slope of a weak waterdrive sys-
tem. (Note that the maximum possible value of OGIP from the
modified Cole plot is slightly less than OGIP from the modified
p/z.) Therefore, the pot aquifer plot was used to determine OGIP
and aquifer size (Fig. 7). OGIP of 5.44 Bcf results from the ex-
trapolation of a line fit to the three data points using the least-
squares method (R2⳱0.934). Thus the p/z extrapolation gave a
value nearly 11% too high, even after being modified to account
for formation compressibility.
The slope of Fig. 7, 58 RB/psi, was used with Eq. 15 to cal-
culate a value for cf of 12×10−6 psi−1, much greater than the esti-
mated value of 3×10−6 psi−1 and too high for “hard rock country.”
Therefore, the estimated cf⳱3×10−6 psi−1 was used with Eq. 14 to
TABLE 7ÑOKLAHOMA MORROW GAS RESERVOIR
PERFORMANCE
Pressure Bg
Days (psia) z (RB/Mscf) Gp (Mscf) p/z
0 5,482 1.0471 0.5770 0 5,235

In Tables 5 and 6, observe that not only do the drive indices for 72 5,099 0.9960 0.5901 157,000 5,119
the incorrect solution using modified p/z fail to sum to unity, but
they also show a consistent trend of increasing with time. This 237 3,818 0.8286 0.6556 814,000 4,608

trend is typical of incorrect solutions and can be used to distinguish 332 3,016 0.7341 0.7353 1,350,000 4,108
incorrect solutions from those solutions that are correct, yet have Ð6 Ð1 Ð6 Ð1
Other data: reservoir temperature = 140¼F; cf = 3 x 10 psi ; cw = 3 x 10 psi ,
drive-index sums that deviate from unity owing to normal data and Sw = .3.
scatter. That is, the latter will exhibit sums that are sometimes

February 2002 SPE Reservoir Evaluation & Engineering 53

 7,500,000
6,000
Original

GpBg / (Bg ÐBgi ) or F/Et , Mscf

5,000
Conventional
7,000,000 Modified
Modified

Conventional

4,000 Modified 6,500,000

p/z , psia

3,000 6,000,000

2,000 5,500,000

1,000
5,000,000

0 500,000 1,000,000 1,500,000 2,000,000

0
G p
, Mscf
0 2,000,000 4,000,000 6,000,000

G p
, Mscf Fig. 6—Original and modified Cole plots, Oklahoma Morrow gas
reservoir.

Fig. 5—p/z plot, Oklahoma Morrow gas reservoir.

of unknown magnitude, and no waterdrive is present. Poston and
calculate W of 6.74 million res bbl. Aquifer size can be compared coworkers11 expanded Roach’s solution to incorporate water in-
with reservoir size by first calculating the original pore volume of flux. (Poston and coworkers described this approach as the “solu-
the hydrocarbon reservoir from: tion plot” method, but the “Roach plot” terminology is retained in
this paper as a more distinctive title.) Equation 6.10 from Ref. 11
GBgi can be expressed in modified form as
PV =
共1 − Swi兲
共p Ⲑ z兲i Ⲑ 共p Ⲑ z兲 − 1 1 共p Ⲑ z兲i Ⲑ 共p Ⲑ z兲 ⭈ Gp
5,440,000 Mscf × 0.5770 RB Ⲑ Mscf = ⭈
= pi − p G pi − p
共1 − .3兲
= 4.84 million res bbl. . . . . . . . . . . . . . . . . . . . . . . . . . . . . ( 16) − 冋 Swi cw + cf We − Wp Bw
1 − Swi
+
共pi − p兲GBgi
册
. . . . . . . . . . . ( 17)
Then, the aquifer is 6.74 million res bbl/4.48 million res bbl⳱1.5
times as large as the gas reservoir. Cumulative water influx of The Roach plot consists of plotting the left side of Eq. 17,
99,800 res bbl after 332 days is calculated with Eq. 12. This 共p Ⲑ z兲i Ⲑ 共p Ⲑ z兲 − 1 共p Ⲑ z兲i Ⲑ 共p Ⲑ z兲 ⭈ Gp
equates to only 3% of the original hydrocarbon pore volume , on the y-axis vs. on the x-axis.
pi − p pi − p
(HCPV) of approximately 3,139,000 res bbl, yet it represents ap-
proximately 10% of the cumulative hydrocarbon voidage The slope of this plot is 1/G, so G is equal to the reciprocal of the
(Gp×Bg⳱1,350,000 Mscf×0.7353 RB/Mscf⳱992,700 res bbl). slope. The y-intercept is the term in brackets on the right side of
Drive indices are shown in Table 8 for the modified p/z solu- Eq. 17 and incorporates formation and water compressibility, as
tion and the pot aquifer solution. Drive-index sums based on the well as water influx and water production.
OGIP obtained from p/z show a trend from too low at early time The difficulty in interpreting the plot in the presence of a wa-
to near unity at late time. Had the drive indices been normalized to terdrive is that the y-intercept is not constant because the water-
sum to unity, the fact that a problem existed with the p/z solution influx and water-production terms in brackets do not remain con-
would have been obscured. Drive indices based on the OGIP ob- stant. Thus, the correct slope is difficult to ascertain, and signifi-
tained from the pot aquifer solution fluctuate around unity, exhib- cant errors in OGIP can easily result.
iting scatter typical of field data. Modified Roach Plot. The problem can be solved, provided
that the aquifer is of the pot aquifer type. Eq. 12 is substituted for
Roach Plot. Roach17 rearranged the p/z relationship to solve for We in Eq. 17, which is then rearranged to move the water-
the correct OGIP when formation compressibility is significant but production term into the x-axis plotting term, resulting in:

Wp Bw
7,500,000 共p Ⲑ z兲i Ⲑ 共p Ⲑ z兲 ⭈ Gp +
共p Ⲑ z兲i Ⲑ 共p Ⲑ z兲 − 1 1 Bgi
= ⭈
pi − p G pi − p

冋 册
7,000,000

Swi cw + cf 共cw + cf兲 W

− + . . . . . . . . . . . . . ( 18)
1 − Swi
F/Eg , Mscf

6,500,000 GBgi

6,000,000
TABLE 8ÑDRIVE INDICES, OKLAHOMA MORROW

GAS RESERVOIR

5,500,000
OGIP=5,440,000 Mscf
Days Modified p/z Solution Pot Aquifer Solution

IGD ICD Total IGD IWD ICD Total

5,000,000

0 5,000 10,000 15,000 20,000 25,000 30,000 72 0.849 0.080 0.929 0.767 0.167 0.072 1.007

∆P/E g, psi-Mscf/RB 237 0.886 0.060 0.946 0.801 0.126 0.055 0.982

332 0.959 0.048 1.007 0.868 0.101 0.043 1.012

Fig. 7—Pot aquifer plot, Oklahoma Morrow gas reservoir.

54 February 2002 SPE Reservoir Evaluation & Engineering

 TABLE 9ÑPERFORMANCE HISTORY AND PVT DATA FOR ONE-CELL
GAS-SIMULATION MODEL WITH FETKOVICH AQUIFER

Cumulative Cumulative
Gas Water Gas
Pressure Produced Produced Deviation Bg Bw
Year (psia) (Bscf) (STB) Factor, z (RB/Mscf) (RB/STB)

0 6,411 0.000 0 1.1192 0.6279 1.0452

1 6,130 5.475 2,163 1.1008 0.6459 1.0460

2 5,849 10.950 9,293 1.0828 0.6659 1.0470

3 5,565 16.425 22,286 1.0652 0.6885 1.0478

4 5,280 21.900 43,807 1.0482 0.7141 1.0488

5 4,992 27.375 78,152 1.0316 0.7434 1.0496

6 4,700 32.850 132,011 1.0158 0.7774 1.0505

7 4,403 38.325 219,211 1.0005 0.8174 1.0515

8 4,101 43.800 358,536 0.9865 0.8653 1.0524

9 3,787 49.275 607,252 0.9731 0.9243 1.0534

10 3,459 54.750 1,034,275 0.9610 0.9994 1.0544

Wp Bw original Roach plot to deviate from a straight line can realistically

共p Ⲑ z兲i Ⲑ 共p Ⲑ z兲 ⭈ Gp + be produced before the wells load up. The method might at least
Bgi
The x-axis plotting term is now , and the find application in enhanced recovery projects in which gas res-
pi − p ervoirs are aggressively dewatered.
y-axis term is the same as before. The y-intercept, the term in
brackets, is now constant; thus, the plotted points will have a Oil Reservoirs
constant and correct slope. The method is demonstrated with a Campbell Plot. For oil reservoirs, the Campbell plot10 is the coun-
simulation example. terpart to the modified Cole plot for gas. It is based on an equation
Simulation Model. The gas model described previously was analogous to Eq. 5 for gas:
modified to give a much stronger aquifer. The water-filled cell was F We
removed, and aquifer strength was provided by attaching a Fetk- = N+ , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ( 19)
ovich-type aquifer18 to the single-cell gas reservoir. Aquifer OWIP Et Et
was 633 million res bbl, or 10 times the HCPV. Aquifer produc- where N⳱OOIP in STB and F⳱cumulative reservoir voidage,
tivity index (PI) was set to a high value, 485 RB/D/psi, and aquifer
F = Np 关Bt + Bg共Rp − Rsi兲兴 + Wp Bw . . . . . . . . . . . . . . . . . . . . . . ( 20)
compressibility (sum of cw and cf) was set to 9×10−6 psi−1. The
model was run for 10 years, as before. Simulation results and PVT Et⳱cumulative total expansion,
data are given in Table 9, and the conventional and modified Et = Eo + mEg + Efw ; . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ( 21)
Roach plots are shown in Fig. 8. Plotted points migrate from left
to right with time. Eo⳱cumulative oil expansion,
Examining Fig. 8, the conventional plot appears to be linear. In Eo = Bt − Bti ; . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ( 22)
reality, however, the points are deviating slightly to the left with
increasing time because water production causes the y-intercept (in Eg⳱cumulative gas expansion,
brackets in Eq. 17) to migrate upward with time. The “slope” of Bti
the conventional plot is 1.042×10−5 MMscf −1, giving G⳱1/ Eg = 共B − Bgi兲; . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ( 23)
Bgi g
(1.042×10−5)⳱96.0 Bcf, almost 5% low to the true OGIP⳱100.8
Bcf. The modified version straightens out the later time points and Efw⳱cumulative formation and water expansion,
gives G⳱1/(0.9853×10−5)⳱101.5 Bcf, less than 1% high com- Swi cw + cf
pared to the true value. Efw = Bti 共1 + m兲 共pi − p兲; . . . . . . . . . . . . . . . . . . . . . ( 24)
1 − Swi
The y-intercept in Eq. 18 can be rearranged to solve for W,
provided that cw and cf are known. A value of 629 million res bbl and m⳱the ratio of initial gas-cap volume to initial oil-zone vol-
is calculated from the modified plot, only 0.7% too low. ume, at reservoir conditions. Bt is the total formation volume factor:
The modified Roach plot has been tested for varying Fetkovich Bt = Bo + Bg共Rsi − Rs兲. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ( 25)
aquifer volumes with this model, using values of W larger and
smaller than the 10×HCPV used in the example. As W increases,
Conventional:
the deviation of the late-time points on the conventional plot be- y = 1.042×10 –5 x–1.082×10 –4
–1

comes more visible; for W of 1 billion res bbl, it is clearly notice- 0.00020
[(p/z) i /(p/z) – 1] / (p i –p) , psi

R 2 = 9.997×10 –1
able, and the late-time points are excluded from the least-square
fit. For W⳱5×HCPV, the conventional plot gives essentially the
correct G because water production is not too great. In application, 0.00016 Modified:
both the conventional and modified plots could be constructed as y = 9.853×10 –6 x–9.645×10 –5
R = 9.999×10 –1
2
in Fig. 8 and compared to determine the amount of deviation. If
only one plot is to be constructed, it should be the modified plot, 0.00012
Conventional
to be on the safe side.
A word of caution: the modified Roach plot has not been veri- Modified
fied with actual field data because suitable field data have not 0.00008
become available. Two questions come to mind when considering 18 22 26 30
field cases: first, whether an actual aquifer could be as large as that x plotting term, MMcf/psi
used in the simulations and still perform like a pot aquifer, and
second, whether water volumes sufficiently large to cause the Fig. 8—Conventional and modified Roach plots, one-cell simulation.

February 2002 SPE Reservoir Evaluation & Engineering 55

 F / Et , STB
TABLE 10ÑPROPERTIES OF OIL-SIMULATION MODEL WITH
POT AQUIFER
Strong
Waterdrive Sw 20.8%
φ 28%
Moderate cf 26×10Ð6 psi Ð1
Waterdrive
cw 2.28×10Ð6 psi Ð1
Weak Reservoir temperature 158¼F
Waterdrive
Bubblepoint pressure 2,648 psia
OOIP

Depletion drive
Bobp 1.2697 RB/STB
OOIP ≈ 20 million STB
Aquifer OWIP ≈80 million STB

F , RB

Fig. 9—Campbell plot curve shapes. The decline in reservoir pressure and lack of significant water
production for 8 years could lead to the interpretation that no
aquifer is present. The recommended method16,19 for solving the
F material balance for an undersaturated oil reservoir without water
Plotting on the y-axis vs. F on the x-axis will yield a plot influx is the plot of F vs. Et, which should be a straight line with
Et
OOIP equal to the slope. Fig. 10 is the plot for these data. Least-
with one of the characteristic curve shapes shown in Fig. 9, which
square straight lines were fit to the data, assuming that evaluations
is analogous to Fig. 1 for gas. In other words, like the Cole plot, the
were performed at various stages in the life of the reservoir, after
Campbell plot is useful in a qualitative sense for distinguishing
3, 7, and 20% of the true OOIP had been produced (after 700,
between depletion-drive reservoirs and strong, moderate, and weak
waterdrives. If the reservoir is depletion drive, the plot can be used 1,285, and 3,595 days, respectively). Calculated values of N
quantitatively because the y-value of the plotted points equals the (shown in the legend of Fig. 10) are in error by +160%, +90%, and
OOIP. But if a waterdrive exists, the slope of the plot is changing +50%, respectively. For this perfect data set, it is obvious that the
continuously, so extrapolation back to the OOIP is hazardous and points do not lie in a straight line, but for real field data, the
is not recommended. curvature could be obscured easily within normal data scatter,
As with the Cole plot for gas, the weak aquifer curve on the leading to the false conclusion that no aquifer is present.
Campbell plot again exhibits a negative slope except for a brief early The Campbell plot for these data, Fig. 11, clearly shows the
period of steep positive slope. Thus, the apparent OOIP, calculated signature negative slope of a weak waterdrive, even after just the
assuming no waterdrive (i.e., N⳱F / Et), exhibits the counterintuitive first two or three data points (700 and 1,285 days, respectively). As
trait of decreasing with time in the presence of weak waterdrive. with the modified Cole plot for gas reservoirs, the points migrate
This was recognized at least as early as 1963.16 The negative slope toward the true OOIP with time.
has been observed in field data and in data from simulation. Because a weak waterdrive is present, the correct material-
balance solution for this case is obtained from the pot aquifer plot
Oil-Simulation Model. A multicell simulation model of an under- that has been derived for oil,16 similar to that for gas. Because the
saturated oil reservoir with an attached pot aquifer was constructed F ⌬p
with the Eclipse14 reservoir simulator. The model used PVT and oil is undersaturated, is plotted on the y-axis vs. on the x-
Eo Eo
other properties similar to those encountered in U.S. Gulf Coast
sandstones: high permeability, porosity, and formation compress- axis (see Ref. 16 for derivation). The y-intercept gives the OOIP.
ibility. Reservoir properties are shown in Table 10. The plot for this case is shown in Fig. 12; the sequence of plotted
Pressures and produced volumes from the simulator output points is from right to left.
were used to perform a material-balance evaluation of the reser- Several solutions were obtained from the pot aquifer plot at the
voir. Performance data are given in Table 11, and PVT data are same point in the reservoir’s life as before. The initial data point at
given in Table 12. 305 days lies below the correct straight-line trend that has become
apparent after 1,285 days (third plotted point) and so is excluded
from that least-square fit. The solution at 1,285 days gives a value
TABLE 11ÑPERFORMANCE HISTORY OF OIL-SIMULATION of N of 21.7 million STB, within <10% of the true value. Some-
MODEL WITH POT AQUIFER time after 1,285 days (that is, after the third plotted point), it
Cumulative Cumulative
becomes apparent that the second data point at 700 days is off
Cumulative
Oil Water Gas
trend as well. Therefore, the second point is excluded from sub-
Pressure Produced Produced Produced
sequent fits, giving increasingly accurate answers.
Days (psia) (STB) (STB) (Mscf) Aquifer OWIP is calculated from the slope of the pot aquifer
plot, using the oil version of Eq. 14 (i.e., N replaces G, and Bti
0 2,855 0 0 0
replaces Bgi; formation compressibility is known in this simulation
305 2,779 192,821 0 94,513 example). After 3,595 days, the slope is 3,090 RB/psi, from which
700 2,627 633,942 0 312,064 W of approximately 79 million res bbl is calculated, very close to
the known value of approximately 80 million res bbl. Oil reservoir
1,285 2,457 1,314,880 4 710,670
pore volume is approximately 35.7 million res bbl, so the aquifer
1,465 2,402 1,524,400 7 850,934 is about 2.2 times as large as the reservoir.
2,005 2,223 2,152,960 26 1,355,720 Drive Indices. Drive indices for oil reservoirs as defined in Ref.
13 are presented here in modified form.
2,365 2,080 2,572,000 60 1,823,250

2,905 1,833 3,200,560 822 2,732,860

Depletion-drive index:
3,235 1,665 3,584,680 11,135 3,397,740
NEo
3,595 1,460 4,003,720 97,443 4,216,120 IDD = . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ( 26)
F − Wp Bw

56 February 2002 SPE Reservoir Evaluation & Engineering

 TABLE 12ÑPVT PROPERTIES OF OIL-SIMULATION MODEL WITH POT AQUIFER

Days Pressure (psia) Bo (RB/STB) Rs (Mscf/STB) Bg (RB/Mscf) Bt (RB/Mscf) Bw (RB/STB)

0 2,855 1.2665 0.5010 0.9201 1.2665 1.0222

305 2,779 1.2677 0.5010 0.9637 1.2677 1.0224

700 2,627 1.2681 0.4973 1.0502 1.2720 1.0228

1,285 2,457 1.2554 0.4671 1.0977 1.2926 1.0232

1,465 2,402 1.2512 0.4574 1.1146 1.2998 1.0233

2,005 2,223 1.2383 0.4269 1.2010 1.3273 1.0237

2,365 2,080 1.2278 0.4024 1.2825 1.3543 1.0240

2,905 1,833 1.2074 0.3579 1.4584 1.4161 1.0246

3,235 1,665 1.1949 0.3277 1.6112 1.4741 1.0250

3,595 1,460 1.1802 0.2908 1.8526 1.5696 1.0254

Segregation (gas cap) drive index:
NmEg
ISD = . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ( 27)
F − Wp Bw
Waterdrive index:
We − Wp Bw
IWD = . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ( 28)
F − Wp Bw
In addition, when cf is significant, as it is in this example, the
formation and connate water compressibility drive index is defined
as follows:
N共1 + m兲Efw
ICD = . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ( 29)
F − Wp Bw
The common denominator in Eqs. 26 through 29 is the hydrocar-
bon voidage. If the material balance has been solved correctly, the
sum of the four drive-index fractions equals unity; that is,
IDD + ISD + ICD + IWD = 1. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ( 30)
Table 13 shows drive indices for the oil-simulation case for the
depletion-drive solution and the pot aquifer solution after 3,595
days. Drive indices for the depletion-drive solution do not add to
unity. Had the values been normalized to add to unity as in some
commercial software, the fact that the depletion-drive solution is
incorrect would have been obscured. Indices for the waterdrive
solution add to unity as expected, excluding the first two points at
305 and 700 days, which were not used in the least-squares solu-
tion fit; thus, their indices would not be expected to add to unity.
In conclusion, the presence of a weak waterdrive acting on this
oil reservoir would likely be overlooked without examining the
Campbell plot or the raw (not normalized) drive indices, just as
with the gas reservoir cases discussed earlier. Significant errors in
12,000,000
the calculated values of OOIP would be obtained using the deple-
tion-drive solution (compare the calculated values of N in Fig. 10
with those in Fig. 12).
Carlson20 pointed out that even when material-balance results
are ambiguous or do not provide very accurate quantitative an-
swers, valuable qualitative insights may still be obtained. For this
oil-simulation case, the pot aquifer material-balance solution after
700 days is considerably in error, and even after 1,285 days, it is
not particularly accurate (Fig. 12). However, the negative slope of
the Campbell plot (Fig. 11) clearly shows the presence of a weak
waterdrive even after only 700 days (first two data points), a valuable
piece of information obtained early in the life of the reservoir.
Other Considerations
In this paper, water compressibility, cw, considers only the liquid
phase. That is, the energy contribution from gas dissolved in the
water, coming out of solution as reservoir pressure declines, is
ignored. Fetkovich et al.21 examined this problem for high-
pressure gas reservoirs and concluded that the energy contribution
from gas dissolved in the water is usually important only late in the
reservoir life (below approximately 1,500 psia). To account for
this additional energy, they defined water total formation volume
factor, Btw, analogous to oil total formation volume factor:
Btw = Bw + Bg共Rswi − Rsw兲. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ( 31)
They also defined water total compressibility, ctw:
Btw − Btwi
ctw = . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ( 32)
Btwi 共pi − p兲
The energy contribution from gas dissolved in the water can be
incorporated in the equations presented in this paper by substitut-
ing Btw (Eq. 31) for Bw and ctw (Eq. 32) for cw. The Campbell and
modified Cole plots would be affected, but not until later in the
reservoir life when pressure has declined. The pot aquifer plot no
longer applies because the slope is no longer constant, and the
Roach plot no longer applies because the y-intercept is no longer

3,595 60,000,000
days
8,000,000 50,000,000
RB

F / E t , RBL
F,

After 700 days (Cum.=3%)

40,000,000

4,000,000 N=52 million STB

After 1,285 days (Cum.=7%)
=38 million STB
30,000,000
N
1,285 days
700 days After 3,595 days (Cum.=20%)
0 305 days N=30 million STB 20,000,000

0.0 0.1 0.2 0.3 0.4 0 5,000,000 10,000,000 15,000,000

Et, RB/STB F , RBL

Fig. 10—Solution plot for oil-simulation case, assuming no waterdrive. Fig. 11—Campbell plot for oil-simulation case.

February 2002 SPE Reservoir Evaluation & Engineering 57

 210,000,000 305 TABLE 13ÑDRIVE INDICES AFTER 3,595 DAYS,
days OIL-SIMULATION MODEL WITH POT AQUIFER
170,000,000
Modified p/z Solution Pot Aquifer Solution
700 days
130,000,000
STB
Days IGD ICD Total IGD IWD ICD Total
F / Eo ,

90,000,000 After 700 days (Cum.=3%) 305 0.151 0.405 0.556 0.100 0.700 0.269 1.069*
N=39 million STB 700 0.209 0.368 0.577 0.139 0.636 0.244 1.019*
1,285 days After 1,285 days (Cum.=7%)
50,000,000 N=21.7 million STB 1,285 0.454 0.293 0.747 0.301 0.506 0.194 1.001
N =
-intercept After 3,595 days (Cum.=20%) 1,465 0.489 0.282 0.771 0.325 0.486 0.187 0.998
y

10,000,000
3,595
days N=20.3 million STB 2,005 0.582 0.256 0.838 0.386 0.442 0.170 0.998

0 20,000 40,000 60,000 2,365 0.643 0.240 0.883 0.427 0.415 0.160 1.001
∆ p/ E, , o psi-STB/RB 2,905 0.739 0.214 0.953 0.491 0.369 0.142 1.002
3,235 0.806 0.196 1.001 0.535 0.336 0.130 1.001
Fig. 12—Pot aquifer plot for oil-simulation case.
3,595 0.892 0.174 1.065 0.592 0.290 0.115 0.998

*Excluded from least-square fit.

constant. Fetkovich et al. presented a method for evaluating gas
reservoirs under these conditions.
In this paper, formation compressibility, cf, is assumed to be
constant and unchanging over the reservoir life being investigated.
Fetkovich et al. presented a method to account for changing for-
mation compressibility in gas reservoirs, and Yale et al.22 pre-
sented a method for oil reservoirs.
Various workers have investigated the effect of errors in mea-
sured reservoir pressure on material-balance results. However, to
this author’s knowledge, such an evaluation has not been per-
formed on weak aquifer material balance such as presented in this
paper. Further study is needed of the sensitivity of these relation-
ships to errors in pressure.
Material Balance and Reservoir Simulation
The perception exists among some that classical material-balance
methods have been rendered obsolete by reservoir simulation. Be-
cause simulation incorporates material balance on a cell-by-cell
basis, it may be argued that stand-alone material balance is super-
fluous and therefore serves no utility on those reservoirs that are
subject to a simulation study.
In response, it is argued that material balance and simulation
are complementary rather than competing tools. Material balance
can provide valuable insights into reservoir mechanisms and pro-
cesses that may be obscured by the multitude of parameters that go
into simulation.
Consider the cases shown in this paper in which weak water-
drives are not apparent from performance data. Simulations per-
formed on these reservoirs without benefit of a prior material-
balance study might well have resulted in rock and fluid param-
eters being adjusted to achieve matches on the wrong values of
OHIP. If the waterdrive is of the pot aquifer type, as in this paper,
material balance can solve for OHIP and aquifer size simulta-
neously and unambiguously, without resorting to trial and error
(provided that sufficient reservoir history is available). Even in
cases in which the material-balance solution is more ambiguous,
the analysis often yields qualitative insights that are as valuable as
quantitative results.
Material balance should be performed before a simulation
study to help narrow the range of the many parameters that can be
adjusted during simulation as well as the magnitude of adjustments
that are considered reasonable. And, of course, it is impractical to
perform a simulation study on every reservoir.
Dake provided an especially cogent discussion of this issue in
Ref. 7. He summarized the situation appropriately: “…numerical
simulation and material balance must not be regarded as competi-
tive techniques: we have too few tools in reservoir engineering to
discard any of them.”
Conclusions
1. The Cole plot (gas) and Campbell plot (oil) diagnose the pres-
ence of a weak waterdrive unambiguously. Depletion-drive
plots, such as the p/z, are ambiguous in the presence of a weak
waterdrive and can give OHIP values that are erroneously high
by a significant amount. As suggested by previous authors, the
weak waterdrive signature on the Cole and Campbell plots is
shown to be a negative slope.
2. The negative slope of the Cole and Campbell plots amounts to
the counterintuitive characteristic of decreasing apparent OHIP
with time.
3. The modified version of the Cole plot should be used in cases in
which formation compressibility is not negligible compared to
gas compressibility, such as abnormally pressured reservoirs.
4. If a correct solution to the material balance has been obtained, the
drive indices will sum to unity (allowing for normal scatter). If the
drive indices do not sum to unity, a correct solution has not been
obtained. The drive indices should never be normalized to sum
to unity because this obscures their usefulness as a criterion for
determining the validity of the solution and gives a false sense
of security. Only the raw calculated values should be reported.
5. The Roach plot can be modified to improve gas reservoir inter-
pretation in the presence of a pot aquifer by incorporating cu-
mulative water production in the x-axis plotting term. This pro-
cedure has not been tested on field data, however.
6. Reservoir simulation does not eliminate the need for classical ma-
terial-balance analysis. Material balance can reveal insights into
reservoir performance that cannot be obtained from simulation,
such as the presence of a weak aquifer that is not otherwise obvi-
ous, as in examples presented in this paper. Material balance is
complementary to, not competitive with, reservoir simulation.
Nomenclature
A ⳱ slope
Bg ⳱ gas formation volume factor, L3/ L3, RB/Mscf
Bo ⳱ oil formation volume factor, L3/ L3, RB/STB
Bt ⳱ total or two-phase oil formation volume factor
(Eq. 25), L3/ L3, RB/STB
Bw ⳱ water formation volume factor, L3/L3, RB/STB
Btw ⳱ total or two-phase water formation volume factor
(Eq. 31), L3/L3, RB/STB
cf ⳱ formation compressibility, L3/L3/(m/Lt2), vol/vol/psi
cw ⳱ water compressibility, L3/L3/(m/Lt2), vol/vol/psi
ctw ⳱ total or two-phase water compressibility (Eq. 32),
L3/L3/(m/Lt2), vol/vol/psi
Eg ⳱ cumulative gas expansion, L3/L3, RB/STB in oil
reservoirs, RB/Mscf in gas reservoirs
Efw ⳱ cumulative formation and water expansion, L3/L3,
RB/STB in oil reservoirs, RB/Mscf in gas reservoirs
Eo ⳱ cumulative oil expansion, including original
complement of solution gas, L3/L3, RB/STB
Et ⳱ cumulative total expansion, L3/L3, RB/STB in oil
reservoirs, RB/Mscf in gas reservoirs
F ⳱ cumulative reservoir voidage, L3, res bbl

58 February 2002 SPE Reservoir Evaluation & Engineering

 G ⳱ original gas in place (OGIP), L3, Mscf
Gp ⳱ cumulative gas production, L3, Mscf
ICD ⳱ formation and connate water compressibility drive
index, L3/L3, fraction
IDD ⳱ depletion-drive index, L3/L3, fraction
IGD ⳱ gas drive index, L3/L3, fraction
ISD ⳱ segregation (gas cap) drive index, L3/L3, fraction
IWD ⳱ waterdrive index, L3/L3, fraction
m ⳱ ratio of gas cap OGIP to oil zone OOIP at reservoir
conditions, L3/L3, dimensionless
N ⳱ original oil in place (OOIP), L3, STB
Np ⳱ cumulative oil production, L3, STB
p ⳱ pressure, m/Lt2, psia
Rp ⳱ cumulative produced gas/oil ratio, L3/L3, Mscf/STB
Rs ⳱ solution gas/oil ratio, L3/L3, Mscf/STB
Rsw ⳱ solution gas/water ratio, L3/L3, Mscf/STB
Swi ⳱ initial water saturation, fraction
W ⳱ aquifer original water in place, L3, res bbl
We ⳱ cumulative water influx, L3, res bbl
Wp ⳱ cumulative water production, L3, STB
z ⳱ gas deviation factor, or compressibility factor, dimensionless
Subscripts
bp ⳱ bubblepoint
f ⳱ formation
fw ⳱ formation and water
g ⳱ gas
i ⳱ initial
o ⳱ oil
p ⳱ cumulative produced
s ⳱ solution
t ⳱ total
w ⳱ water
Acknowledgments
I thank Marathon Oil Co. for permission to publish this paper
following my retirement, particularly Jim Gilman for his special
efforts. Teresa Schaller ran the oil-simulation case presented in the
paper. Stuart Cox provided the Morrow Gas data and consulted on
the interpretation. Lois Fitzpatrick provided valuable help format-
ting this paper.
References
1. Dake, L.P.: Fundamentals of Reservoir Engineering, Elsevier Scien-
tific Publishing Co., Amsterdam (1978) 27–29, 303–337.
2. Bruns, J.R., Fetkovich, M.J., and Meitzen, V.C.: “The Effect of Water
Influx on p/z-Cumulative Gas Production Curves,” JPT (March 1965) 287.
3. Chierici, G.L., Pizzi, G., and Ciucci, G.M.: “Water Drive Gas Reser-
voirs: Uncertainty in Reserves From Past History,” JPT (February
1967) 237; Trans., AIME, 240.
4. Tehrani, D.H.: “An Analysis of a Volumetric Balance Equation for Cal-
culation of Oil-in-Place and Water Influx,” JPT (September 1985) 1664.
5. Vega, L. and Wattenbarger, R.A.: “New Approach for Simultaneous
Determination of the OGIP and Aquifer Performance With No Prior
Knowledge of Aquifer Properties and Geometry,” paper SPE 59781
presented at the 2000 SPE/CERI Gas Technology Symposium, Cal-
gary, 3–5 April.
6. Agarwal, R.G., Al-Hussainy, R., and Ramey, H.J.: “The Importance of
Water Influx in Gas Reservoirs,” JPT (November 1965) 1336; Trans.,
AIME, 234.
7. Dake, L.P.: The Practice of Reservoir Engineering, Elsevier, Amster-
dam (1994) 73, 82–84, 97, 133–134, 472–476.
February 2002 SPE Reservoir Evaluation & Engineering
8. Lee, J. and Wattenbarger, R.A.: Gas Reservoir Engineering, Textbook
Series, SPE, Richardson, Texas (1996) 5, 236.
9. Cole, F.W.: Reservoir Engineering Manual, Gulf Publishing Co.,
Houston (1969) 285.
10. Campbell, R.A. and Campbell, J.M. Sr.: Mineral Property Economics,
Vol. 3: Petroleum Property Evaluation, Campbell Petroleum Series,
Norman, Oklahoma (1978) 26.
11. Poston, S.W. and Berg, R.R.: Overpressured Gas Reservoirs, SPE,
Richardson, Texas (1997) 105–106.
12. Wang, B. and Teasdale, T.S.: “GASWAT-PC: A Microcomputer Pro-
gram for Gas Material Balance With Water Influx,” paper SPE 16484
presented at the 1987 SPE Petroleum Industry Applications of Micro-
computers, Del Lago on Lake Conroe, Montgomery, Texas, 23–26 June.
13. Craft, B.C. and Hawkins, M.F., revised by Terry, R.E.: Applied
Petroleum Reservoir Engineering, second edition, Prentice-Hall Inc.,
Englewood Cliffs, NJ (1991) 63.
14. Eclipse 100 Reference Manual, © 1982–1999, Schlumberger.
15. Ramagost, B.P. and Farshad, F.F.: “P/z Abnormally Pressured Gas
Reservoirs,” paper SPE 10125 presented at the 1981 SPE Annual Tech-
nical Conference and Exhibition, San Antonio, Texas, 5–7 October.
16. Havlena, D. and Odeh, A.S.: “The Material Balance as an Equation of
a Straight Line,” JPT (August 1963) 896; Trans., AIME, 228.
17. Roach, R.H.: “Analyzing Geopressured Reservoirs–A Material-
Balance Technique,” paper SPE 9968 available from SPE, Richardson,
Texas (1981).
18. Fetkovich, M.J.: “A Simplified Approach to Water Influx Calcula-
tions—Finite Aquifer Systems,” JPT (July 1971) 814.
19. Wang, B., Litvak, B.L., and Bowman, G.W.: “OILWAT: Microcom-
puter Program for Oil Material Balance With Gascap and Water In-
flux,” paper SPE 24437 presented at the 1992 SPE Petroleum Com-
puter Conference, Houston, 19–22 July.
20. Carlson, M.R.: “Tips, Tricks, and Traps for Oil Material Balance Cal-
culations,” paper 95–07 presented at the 1995 Annual Technical Meet-
ing of the Petroleum Society of CIM, Banff, Alberta, 14–17 May.
21. Fetkovich, M.J., Reese, D.E., and Whitson, C.H.: “Application of a
General Material Balance for High-Pressure Gas Reservoirs,” paper
SPE 22921 presented at the 1991 SPE Annual Technical Conference
and Exhibition, Dallas, 6–9 October.
22. Yale, D.P. et al.: “Application of Variable Formation Compressibility for
Improved Reservoir Analysis,” paper SPE 26647 presented at the 1993
SPE Annual Technical Conference and Exhibition, Houston, 3–6 October.
SI Metric Conversion Factors
acre × 4.046 873 E–01 ⳱ ha
bbl × 1.589 873 E–01 ⳱ m3
ft × 3.048* E–01 ⳱ m
ft3 × 2.831 685 E–02 ⳱ m3
°F (°F–32)/1.8 ⳱ °C
psi × 6.894 757 E+00 ⳱ kPa
psi−1 × 1.450 377 E–01 ⳱ kPa−1
scf/bbl × 1.801 175 E–01 ⳱ m3/m3 (st)
*Conversion factor is exact.
Jeff Pletcher retired in 1999 from Marathon Oil Co.’s Petroleum
Technology Center in Littleton, Colorado. He was an ad-
vanced senior engineer working in the areas of reservoir evalu-
ation and reservoir engineering training. His career spanned
more than 30 years, all with Marathon. Previous assignments
were in production and reservoir engineering in Illinois, Texas,
Louisiana, and the general office in Findlay, Ohio. Pletcher holds
a BS degree in petroleum engineering from Marietta College.
59