The Economics of Enhanced Oil Recovery:

Estimating Incremental Oil Supply and CO2

Demand in the Powder River Basin
Klaas van t Veld* and Owen R. Phillips**

Expanding the use of CO2-enhanced oil recovery (EOR) promises to both

significantly increase recovery from existing U.S. oil reserves and possibly form
a bridge to large-scale CO2 capture and sequestration. An important input into
planning for such expansion are estimates of how both the supply of incremental
oil and the derived CO2 demand from EOR are likely to vary with the prices of
oil and CO2. We demonstrate how the analog method of predicting oil and
CO2 flows can be used to readily generate such estimates, and apply the method
to Wyomings Powder River Basin.
1. INTRODUCTION
As the policy issues of energy security and climate change have taken
center stage in recent years, the technique of CO2-enhanced oil recovery (CO2EOR) has received increasing attention from industry and government.1 One reason is that the technique promises significant increases in oil recovery from existing, mature oil fields. Typically, the primary phase of oil extraction from a new
1. See for example the statistics on CO2-EOR growth in the biennial Oil & Gas Journal surveys
of EOR, the royalty relief and tax credits for CO2-EOR in the Energy Policy Act of 2005, the
requirement to study pipeline construction for CO2-EOR in the Lieberman-Warner Climate Security
Act of 2008, and the further tax credits for CO2-EOR in the Emergency Economic Stabilization Act
of 2008 (the $700 billion federal bailout package for the financial industry). For a discussion of CO2EORs role in overall energy policy see Griffin (2009).
The Energy Journal, Vol. 31, No. 3, Copyright 2010 by the IAEE. All rights reserved.
*
**

Corresponding author. Department of Economics & Finance, University of Wyoming, Dept

3985, 1000 E. University Ave., Laramie, Wyoming 82071-3985. E-mail: klaas@uwyo.edu.
Department of Economics & Finance and Enhanced Oil Recovery Institute, University of Wyoming.

We thank J. Michael Boyles for laying much of the groundwork for this study, and Brian F. Towler
and Vladimir Alvarado for helpful discussions. We also thank the editor, associate editor, and three
anonymous referees for many comments that improved our paper.

31

32 / The Energy Journal

field, which utilizes the reservoirs natural pressure to bring oil to the surface,
extracts about 520% of the estimated original oil in place. The secondary phase,
which usually involves injection of water into the reservoir to augment or maintain its pressure, extracts another 1020%. After the primary and secondary
phases of extraction, about two thirds of the original oil is left stranded. According to a recent Department of Energy study (DOE, 2008), the tertiary method
of CO2-EOR is technically able to recover about a third of this stranded oil, or
an additional 20% of the original oil in place. For all U.S. reservoirs combined,
this amounts to 87.1 billion barrels, of which the study estimates that (at an oil
price of $70 per barrel and a CO2 price of $45 per metric ton) 45 billion barrels
are economically recoverable.2
CO2-EOR recovers this additional oil by injecting slugs of CO2 at high
pressure into the reservoir, usually alternated with slugs of water. The injected
CO2 mixes with the reservoir oil, thereby reducing capillary forces that trap the
oil in pores of the rock and allowing oil that would otherwise remain stranded to
flow towards production wells.3 Most of the CO2 resurfaces with the recovered
oil and is separated, recompressed, and reinjected. In every pass through the
reservoir, however, a fraction of the CO2 remains sequestered underground. In
order to maintain a given CO2 injection rate, operators of CO2-EOR projects4
therefore need reliable sources of CO2 over extended periods of timeit is common for an EOR project to take 20 years or longer. As a result, EOR creates a
derived demand for relatively pure CO2 gas.
This steady demand for, and ultimately sequestration of, CO2 provides
the second reason for the recent interest in EOR. Realistically, the total amount
of CO2 that EOR projects might be able to sequester is limited: Dahowski et al.
(2005) estimate the sequestration capacity of depleted U.S. oil reservoirs (including those depleted through EOR) at 10 GtCO2, which amounts to just over two
years worth of current U.S. CO2 emissions (EPA, 2008). Nevertheless, EOR may
2. To put these figures in perspective, technically recoverable reserves in the Arctic National
Wildlife Refuge (ANWR) are estimated by the U.S. Geological Survey at 10.4 billion barrels. See
the USGS National Assessment of Oil and Gas Resources Update (December, 2007) at http://
certmapper.cr.usgs.gov/data/noga00/natl/tabular/2007/ summary_07.pdf.
3. To avoid fracturing the caprock overlying the reservoir, a small number of CO2-EOR projects
are operated at pressures too low for the CO2 to mix with the oil. These so-called immiscible CO2
floods generally recover significantly less of the stranded oil.
4. Hereafter, we drop the qualifier CO2 as understood. The term EOR is more generally used to
denote a variety of processes that enhance oil recovery beyond levels attained through primary and
secondary methods, including injection of steam, liquid chemicals, and gases other than CO2. According to the Oil & Gas Journals most recent biennial survey of EOR (Moritis, 2008), CO2-EOR is
the fastest-growing EOR technique in the U.S., generating 250,000 barrels per day (bo/d) from 105
projects spread throughout the country. This amounts to about 5% of total U.S. oil production, and
is up from just 30,000 bo/d in 1986. So-called thermal EOR, which uses injections of mostly steam,
is slightly more prevalent in terms of production volume, generating 293,000 bo/d from currently 45
projects. Use of this method is almost entirely limited to heavy-oil fields in California, however, and
the aggregate production volume has steadily declined from a peak of 469,000 bo/d in 1986.

The Economics of Enhanced Oil Recovery / 33

jump-start the building of pipelines and other infrastructure required for ultimately much larger-scale sequestration in unmineable coal seams and saline aquifers.
If indeed EOR is to play a major role in expanding U.S. oil output as
well as providing a bridge to large-scale geological sequestration of CO2, an
important input into planning for the required infrastructure are estimates of how
both the supply of incremental oil and the derived CO2 demand from EOR are
likely to vary with the prices of oil and CO2. To our knowledge, very few such
estimates are currently available. Holtz et al.s (2001) study of EOR potential in
Texas, for example, uses only physical screening criteria to identify reservoirs
suitable for EOR, and then applies rule-of-thumb multipliers to estimate the incremental oil that can be recovered and the CO2 that can be sequestered in these
reservoirs, without any reference to economics. Similar methods also are used to
estimate EORs sequestration potential (without accompanying oil recovery estimates) in the DOEs Carbon Sequestration Atlas of the United States and Canada (DOE, 2007). The above-cited study by Dahowski et al. (2005) does estimate
a cost curve for CO2 sequestration in 220 oil plays5 in the U.S., but does so for
just three oil prices, namely $15, $23, and $38/bo. It is also again based on ruleof-thumb multipliers, adjusted only for API gravity6 and average depth of each
play. Finally, the above-cited DOE (2008) study uses physical screening criteria
to identify 1,111 large oil reservoirs amenable to EOR, and then uses reservoirsimulation software to predict oil and CO2 flows for each reservoir. In principle,
this approach could be used to generate full oil supply and CO2 demand curves,
but the study in fact examines only four price scenarios.7
In this paper, we introduce a procedure for estimating incremental oil
supply and CO2 demand curves that is based on the so-called analog method
of predicting oil and CO2 flows for a given reservoir. The method is more sophisticated than simple rule-of-thumb multipliers, while avoiding a key drawback
of using reservoir-simulation software. This drawback is that reservoir simulations require reservoir-specific relative permeability curves as inputs, to predict
the rates at which different fluids (oil, water, CO2) will move through a given
reservoirs rock as their concentration levels in the reservoir change over time.
Data required to determine these curves are rarely available.
The analog method, explained in detail in the appendix to this paper, is
not new to reservoir engineers. Jarrell et al. (2002) discuss it, for example, in
their monograph on CO2 flooding published by the Society of Petroleum Engi5. An oil play is a grouping of geologically similar reservoirs in a given oil-producing region.
6. A standard measure of oil density introduced by the American Petroleum Institute.
7. A now dated study by the National Petroleum Council (NPC, 1984) used very similar procedures. The main difference lies in the software used to predict CO2 and oil flows: whereas the NPC
study used the CO2PM package developed by Scientific Software-Intercomp, the DOE study uses
the more versatile CO2Prophet package later developed by the Texaco Exploration and Production
Technology Department. (Both packages are available from the DOEs National Energy Technology
Laboratory website, http://www.netl.doe.gov.)

34 / The Energy Journal

neers. The method also underlies a spreadsheet model made available by Kinder
Morgan, Inc., which allows oil-field operators to estimate the likely profitability
of applying EOR to a given reservoir.8 The method is not widely known by energy
economists, however. Nor has it, to our knowledge, ever been used to estimate
incremental oil supply curves and derived CO2 demand curves for all reservoirs
in an entire region or basin, as we do in this paper.
In essence, the analog method scales the historical production and injection flows observed at some existing, mature EOR project (the analog) to
predict those of a new, proposed EOR project. The validity of doing so relies on
a central assumption of the method, namely that all the various dimensions across
which reservoirs may differlithology, area, thickness, porosity, permeability,
etc.are relevant to incremental oil and CO2 production only insofar as they
affect two key scaling factors: (i) per-pattern hydrocarbon pore volume and (ii)
injectivity. The term pattern refers to a (typically square) sub-area of a reservoir
centered on a single injection well and bordered by production wells;9 hydrocarbon pore volume (HCPV) is the space originally occupied by oil in that sub-area
of the reservoir before any of the oil was produced; and injectivity is the rate at
which fluids can be injected into the reservoir, expressed in units of HCPV per
unit time.
Specifically, the analog method predicts that if the proposed EOR project
happens to have the same per-pattern HCPV and injectivity as the EOR analog
project, each of its patterns will generate roughly the same incremental oil and
CO2 flows over time as the analog project did historically. More generally, the
proposed project will differ from the analog project in terms of either HCPV or
injectivity, in which case the predicted flows are scaled accordingly. If, for example, the proposed project has twice the per-pattern HCPV but the same injectivity, it is predicted to cumulatively produce twice as much from each pattern as
the analog project did at any given time after switching to EOR; if, on the other
hand, the proposed project has the same per-pattern HCPV but twice the injec-

8. The model is available at www.kindermorgan.com/business/co2/tech.cfm. It requires the operator to enter engineering parameters for a proposed EOR project (e.g., the reservoir dimensions,
the current rate and decline rate of oil production, the number of existing and planned injection and
production wells) as well as economic parameters (e.g., the price of oil and CO2 anticipated by the
operator, royalty and tax rates, discount rate). The model then uses the analog method to project
incremental oil and CO2 flows for a single pattern (an injection well surrounded by production
wells) of the proposed project, multiplies these flows by the number of planned patterns, and combines
the result with the economic parameters to predict the projects NPV. Although the model uses a very
similar approach to ours, its implementation as a spreadsheet limits its application to a single project
at a time. The model is also cruder than ours in several respects. For example, it terminates the project
at an exogenously determined time, rather than optimally as in our model.
9. Common patterns are the five-spot, which has four production wells at the corners of the
square, and the nine-spot, which has four additional production wells at the square sides. These
patterns are typically repeated more or less regularly to cover the entire reservoir area, whereby
neighboring injection wells share the production wells on their common pattern borders.

The Economics of Enhanced Oil Recovery / 35

tivity, it is predicted to cumulatively produce as much as the analog project did,
but in half the time; etc.
Note that if the methods central assumption held exactly, all EOR projects would literally trace out the same normalized production and injection paths.
A single analog project would therefore suffice to predict EOR flows at any and
all proposed projects, regardless of any differences between project reservoirs
besides per-pattern HCPV and injectivity. In practice, of course, the assumption
holds only approximately. For example, even if two project reservoirs have similarly high injectivity levels, injectivity for project A may be uniformly high
throughout its reservoir, whereas that for project B may be concentrated in highly
permeable streaks or zones. If so, then in project A, injected CO2 is likely to push
oil uniformly towards production wells, whereas in project B, CO2 may flow
preferentially through the permeable zones, bypassing oil elsewhere in the reservoir. As a result, incremental oil recovery from project B is likely to be lower,
and would be overpredicted by a model using project A as an analog.10 For reasons
such as these, the analog method is considered more reliable the more closely the
reservoir characteristics of a proposed project match those of the analog used.11
By way of illustration, we apply our procedure to the Wyoming portion
of the Powder River Basin (PRB). This basin, which covers the northeast corner
of the state, is a major oil-producing region in the US, with currently about 500
actively producing fields or (since many fields produce from several reservoirs)
over 700 actively producing field-reservoir combinations (FRCs). To date, 1.9
billion barrels of oil have been extracted from these fields, almost all through
primary and secondary recovery. Enhanced oil recovery is just getting underway
in the PRB. On the western edge of the basin, in the Salt Creek Field, one operator
has been applying EOR since 2004. Several other oil-field operators have plans
to begin EOR projects in the near future.
Not all FRCs are suitable for EOR, however. For a given FRC, the size
and geological properties of the reservoir are factors that decide the profitability
of EOR, along of course with expected revenues from incremental oil production
and costs related to CO2 injection and recycling. As oil prices increase or CO2
prices decline, more FRCs become profitable for EOR, giving rise to the incremental oil supply and derived CO2 demand schedules that we map out for the
basin.

10. Similarly, even if two project reservoirs have the same original HCPV, they may experience
different degrees of compaction over time as fluids are removed during the various recovery phases.
This too might differentially impact EOR performance, although the effect would likely be small.
11. Consistent with this, Kinder Morgan provides two different versions of its spreadsheet model:
one uses the Denver Unit project in the San Andres formation of West Texas as its analog, while the
other uses an unspecified project in the Morrow formation of western Kansas and the Oklahoma
Panhandle. Unfortunately, the Morrow projects history is quite short, which reduces its usefulness
for predicting the lifetime performance of candidate EOR projects.

36 / The Energy Journal

2. THE DATA AND MODEL
Extensive data were collected on all FRCs in the Wyoming portion of
the Powder River Basin. These data describe the geology, oil composition, and
production and injection history of identified FRCs. Our data were collected from
numerous primary sources, including the Wyoming Geological Association, the
Wyoming Oil and Gas Conservation Commission, and the proprietary IHS data
bank. Journal descriptions of a number of FRCs were consulted to fill gaps.
Considerable work continues in updating and checking these data against different
source materials.
To estimate the potential CO2 demand for enhanced oil recovery in the
PRB and the corresponding supply of incremental oil, we examined all FRCs that
met two criteria. First, given the large up-front capital costs of EOR projects, we
required the fields to be large. The cutoff chosen was an FRC that had cumulative production of at least 5 million barrels of oil (MMbo) through the end of
2005. Smaller reservoirs were included if another reservoir in the same field met
the 5-MMbo cumulative production criterion. This is because reservoirs in the
same field can share capital facilities required for EOR. A total of 138 FRCs met
the first criterion. The second criterion is that a complete set of data had to be
available for the demand analysis. Key data were unavailable for 38 of the 138
FRCs, leaving 100 FRCs.
For each of these 100 FRCs, we first determined whether the reservoir
passed a key physical hurdle, namely the capability to be pressured to a level at
which injected CO2 mixes with the oil. If this so-called minimum miscibility
pressure (MMP), which depends on the reservoirs temperature and the oils API
gravity, exceeds the maximum pressure that the reservoirs caprock can withstand,
then using CO2 for EOR becomes far less attractive.12 This was found to be the
case for 3 FRCs.
Table 1 lists the remaining 97 FRCs, together with their original oil in
place (OOIP)the estimated total amount of oil originally present in the reservoir
before any extraction took placeand their cumulative production up to mid2009. The accompanying Figure 1 shows the FRCs locations.

12. As pointed out in footnote 1, CO2-based EOR projects can be operated at pressures below
MMP, but their performance drops significantly.

The Economics of Enhanced Oil Recovery / 37

Table 1. Field-reservoir combinations evaluated, with estimated original oil
in place (OOIP, from various sources including McDaniel (1991),
the DOEs TORIS database, and estimates of HCPV) and
cumulative oil production up to mid-2009 (Cum., from the IHS
PI/Dwights PLUS database)
OOIP

Cum.

FieldReservoir Combination

AlphaMinnelusa C

13.2

5.7

Ash CreekShannon

17.3

4.6

Barber CreekFerguson

19.6

1.5

Big HandMinnelusa

9.0

6.1

Big MuddyDakota

11.7

7.3

Bone PileMinnelusa B

15.6

8.9

Buck Draw NorthDakota

36.0

24.4

Camp CreekMinnelusa B

10.5

4.9

Cellars RanchTensleep

10

ClaretonMuddy

11

(MMbo)

34.1

6.2

149.0

3.2

Cole CreekDakota

30.0

1.1

12

Cole Creek SouthDakota

47.3

1.5

13

Cole Creek SouthLakota

34.3

5.4

14

CollumsMuddy

23.6

4.4

15

Coyote CreekDakota

54.4

13.3

16

Coyote Creek SouthDakota

11.4

5.4

17

Coyote Creek SouthTurner

9.2

1.2

18

Culp & Heldt DrawShannon

28.3

13.8

19

Dead Horse CreekFerguson

16.1

1.5

20

Dead Horse CreekParkman

26.4

2.0

21

Dillinger RanchMinnelusa A

24.1

8.0

22

Donkey CreekDakota

23.9

3.8

23

Donkey CreekMinnelusa

10.4

4.7

24

Dry GulchMinnelusa A

25

Duvall RanchMinnelusa A

26

EdselMinnelusa B

27
28

9.1

5.2

24.9

14.6

9.5

5.6

Fiddler CreekMuddy

25.9

2.9

Fiddler CreekNewcastle

55.9

1.2
(continued)

38 / The Energy Journal

Table 1. Field-reservoir combinations evaluated, with estimated original oil
in place (OOIP, from various sources including McDaniel (1991),
the DOEs TORIS database, and estimates of HCPV) and
cumulative oil production up to mid-2009 (Cum., from the IHS
PI/Dwights PLUS database) (continued)
OOIP

Cum.

FieldReservoir Combination

(MMbo)

29

Finn-ShurleyTurner

250.0

16.1

30

Finn-ShurleyWall Creek

15.5

1.4

31

Gas DrawMuddy

50.6

23.2

32

Glenrock SouthDakota

80.0

15.1

33

Glenrock SouthMuddy

62.1

19.4

34

GutheryMinnelusa B Upper

12.5

3.8

35

HalversonMinnelusa A

40.7

8.3

36

HammMinnelusa B Lower

20.3

8.1

37

Hartzog DrawShannon

353.3

114.9

38

HilightMuddy

110.0

74.3

39

House CreekSussex

67.0

38.2

40

Jepson-Holler DrawShannon

49.7

6.0

41

KayeTeapot

86.1

9.3

42

KittyMuddy

133.4

16.5

43

KummerfeldDakota

12.8

3.3

44

KummerfeldMinnelusa B

15.0

6.4

45

Lance CreekLeo

121.0

15.7

46

Lance Creek EastDakota

21.1

1.4

47

Little Mitchell Creek Minn. B

13.0

8.3

48

M-DMinnelusa B

12.0

5.8

49

MaysdorfMinnelusa A

11.5

5.4

50

Meadow CreekFrontier

7.2

2.3

51

Meadow CreekLakota

9.4

2.1

52

Meadow CreekShannon

34.0

5.4

53

Meadow CreekTensleep

40.5

13.8

54

Mellott RanchMinnelusa

19.1

5.0

55

Mikes DrawTeapot

23.0

14.6

56

Miller CreekDakota

17.0

4.4
(continued)

The Economics of Enhanced Oil Recovery / 39

Table 1. Field-reservoir combinations evaluated, with estimated original oil
in place (OOIP, from various sources including McDaniel (1991),
the DOEs TORIS database, and estimates of HCPV) and
cumulative oil production up to mid-2009 (Cum., from the IHS
PI/Dwights PLUS database) (continued)
OOIP

Cum.

FieldReservoir Combination

(MMbo)

57

Moorcroft WestDakota

28.9

4.5

58

Moorcroft WestMinn. A

0.7

0.2

59

Moorcroft WestNewcastle

28.9

1.7

60

Mule CreekLakota

10.0

1.0

61

Mush CreekNewcastle

35.0

2.8

62

North ForkTensleep

55.6

21.5

63

OsageNewcastle

69.0

17.2

64

Pine TreeShannon

14.5

9.5

65

Poison DrawTeckla

13.4

7.6

66

Prong CreekMinnelusa

14.0

6.4

67

Raven CreekMinnelusa

73.8

43.0

68

RecluseMuddy

64.5

13.6

69

ReelMinnelusa

20.0

7.4

70

RenoMinnelusa

41.2

6.4

71

Robinson RanchMinnelusa

14.7

5.8

72

RozetMinnelusa

44.9

9.1

73

RozetMuddy

71.9

13.6

74

Rozet WestMinnelusa

22.7

9.8

75

Sand DunesFrontier

2.5

0.9

76

Sandbar EastMinnelusa B

32.1

9.0

77

ScottParkman

250.0

17.6

78

ScottTeapot

51.4

0.4

79

SemlekMinnelusa B

11.6

5.4

80

Semlek WestMinnelusa B

21.0

8.3

81

Skull CreekNewcastle

30.0

5.2

82

SlatteryMinnelusa

28.5

11.9

83

SlatteryMuddy

1.4

0.4

84

Springen RanchMuddy

22.8

8.8
(continued)

40 / The Energy Journal

Table 1. Field-reservoir combinations evaluated, with estimated original oil
in place (OOIP, from various sources including McDaniel (1991),
the DOEs TORIS database, and estimates of HCPV) and
cumulative oil production up to mid-2009 (Cum., from the IHS
PI/Dwights PLUS database) (continued)
OOIP

Cum.

FieldReservoir Combination

(MMbo)

85

StewartMinnelusa B

40.9

11.1

86

SussexFrontier

5.4

0.3

87

SussexShannon

4.6

2.1

88

SussexSussex

11.7

3.5

89

SussexTensleep

25.0

5.4

90

Sussex WestShannon

35.0

13.4

91

TerraceMinnelusa B

13.8

6.5

92

Timber CreekMinnelusa

34.1

11.7

93

Timber CreekMuddy

6.4

0.2

94

UteMuddy

43.9

9.7

95

WallaceMinnelusa B

18.3

7.9

96

Well DrawTeapot

95.0

33.5

97

Winter DrawMinnelusa

9.2

6.5

An overview of the model

For each of the 97 FRCs, we estimated both the baseline net present
value NPV bas of continuing with secondary oil recovery using water injection,
T bas

NPV bas

p o Qop,bas
(1s R ) (1s SP ) C o ( Qtp,bas )
t
,
(1r) t
t1

and the net present value NPV eor of switching to EOR,

NPV
eor

Teor o

r
cp
o
p,eor
p Qop,eor
(1sR)(1sSP)pcQcm
)
t
t C (Qt )C (Qt
K.
t
(1r)
t1

In these expressions, po represents the price of oil (assumed constant over the
lifetime of the project), Qop,bas
, projected baseline oil recovery in period t under
t
the continued waterflood, and Qop,eor
, projected oil recovery in that same period
t
were the project to switch to a CO2 flood. To arrive at net operating profits in
each period, we subtract from pre-tax oil revenues any royalties at rate sR , as well

The Economics of Enhanced Oil Recovery / 41

Figure 1. Location of field-reservoir combinations listed in Table 1. Circle
areas are proportional to projected incremental oil supply at our
reference oil price of $100/bo and CO2 price of $3/Mcf.

as severance and property taxes at combined rate sSP . For the EOR project, we
we also subtract the projected cost of CO2 purchases, equal to the CO2 purchase
price pc (also assumed constant over the lifetime of the project) times the pror
jected quantity purchased, Qcm
t , as well as the projected cost C of recycling and
re-injecting CO2. This cost depends on the quantity Qcp
of
CO
t
2 that is produced
together with the oil. Lastly, we subtract other operating costs Co , which depend
in part on the total amount of liquids produced. For the baseline without EOR,
liquid production Qt p,bas is just the sum of water and oil produced; for the EOR
project, Qt p,eor is the sum of water, oil, and CO2 produced, since before recycling
the CO2 is mixed in with the oil.
The remaining net operating profits are discounted to the present at the
internal rate of return r required by the FRC operator. The economic lifetime of
the project, denoted Tbas for the continued waterflood and Teor for the CO2 flood,
is reached when operating profits turn negative. Switching to EOR in addition
involves an up-front investment cost K. We assume that this entire cost is incurred

42 / The Energy Journal

immediately at time 0.13 Switching is optimal if and only if NPV eor exceeds
NPV bas .14
Predicting the response to EOR
Implementing the above estimation of NPV requires estimates of an
FRCs response to EORessentially a production function that projects time
paths of incremental oil recovery and CO2 production. As noted, we use the analog
method laid out in the appendix to generate these estimates. Two analog schedules
are necessary to calculate NPV eor . One schedule predicts incremental oil production Qop,inc
, which is then added to predicted baseline production to obtain overall
t
EOR production. The other schedule predicts CO2 production Qcp
t , all of which
is assumed to be recycled and re-injected. Carbon purchases Qcm
are calculated
t
by subtracting predicted production from CO2 injection.
Completing the analysis
Completing the economic analysis (for a given oil price and CO2 price)
requires combining the predicted production, recycling, and purchase paths with
cost data. There are a number of cost categories that enter in the NPV calculation.
Investment costs are the up-front expenditures needed to get an EOR
project underway; they are the variable K in the expression for NPV eor and are
typically large. Apart from the cost of constructing spur pipelines to connect a
project with trunk pipelines for CO2 and oil (a cost not included in our analysis),
the three main cost components are those of (i) drilling new wells, (ii) reconfiguring (working over) well equipment, and (iii) constructing a CO2 recycling
plant. Costs of drilling new wells are typically very high. This is particularly true
in older fields, where many original wells may have been capped because their
secondary-phase production declined to unprofitable levels. The spacing between
the remaining, active wells may then be too large for an EOR project. Wellworkover costs include the costs of replacing tubing in existing wells with tubing
that can resist the corrosion by carbonic acid created when CO2 mixes with water,
as well as the costs of laying new pipelines in the field to pump CO2 to individual
wells. Lastly, CO2 recycling plant costs must be incurred to process the CO2 that
eventually comes back to the surface through producing wells.
Incremental operating costs for EOR projects also have three main components. They consist of (i) standard operating costs for incremental liquid pro13. Realistically, converting an oil field currently under secondary recovery to EOR may take
considerable time, and therefore the investment of K could be spread over several months, or even
years. Our analysis does not account for this possibility.
14. Our analysis does not consider possible alternative methods of EOR, such as injection of
methane or nitrogen. In principle, nothing prevents operators from trying such methods even after
completing a CO2 flood. In practice, however, this is unlikely to occur, as the CO2 flood would leave
little or no oil that these alternative (and generally more expensive) methods would be able to recover.

The Economics of Enhanced Oil Recovery / 43

duction and incremental wells, (ii) costs of purchasing CO2, and (iii) costs of
recycling CO2. Standard operating costs are those of labor, maintenance, and fuel
required for both waterflooding and EOR operations. Some of these are roughly
proportional to the amount of liquids produced, others to the number of wells
operated. CO2 purchasing costs cover the costs of CO2 production and compression at its source, and transportation from the source to the FRC. CO2 recycling
costs consist largely of the costs of energy (electricity or gas) required to fuel the
recompression pumps in the recycling plant, together with some labor and maintenance costs associated with that plant.
3. INCREMENTAL OIL SUPPLY AND CO2 DEMAND FOR THE PRB
The results of running the model to predict both the incremental oil
supply and the derived demand for CO2 from EOR in the PRB are shown in
Figures 2 and 3. More specifically, Figure 2 shows, for oil prices plotted on the
vertical axis and a CO2 price of $3 per thousand cubic feet (Mcf),15 the projected
Figure 2. Cumulative incremental oil supply at CO2 price pc $3 /Mcf
using the Lost Soldier-Tensleep analog (LST) and the Denver
Unit-San Andres analog (DSA)

15. Mcf is the standard quantity unit used by U.S. oil field operators for gases. Because the volume
of a gas depends on its temperature and pressure, the unit is defined at a reference pressure of 60

44 / The Energy Journal

Figure 3. Cumulative CO2 demand at oil price po $100 /bo using the Lost
Soldier-Tensleep analog (LST) and the Denver Unit-San Andres
analog (DSA)

cumulative incremental oil recovery, denoted Qo and measured in millions of

barrels per year (MMbo), for all FRCs that at these prices could profitably switch
to EOR. In other words, Qo represents the difference between projected cumulative oil recovery from these FRCs were they to continue waterflooding and
projected cumulative oil recovery were they to switch to CO2 flooding. Similarly,
Figure 3 shows, for CO2 prices on the vertical axis and an oil price of $100 per
barrel, the corresponding cumulative deliveries of CO2, denoted Qc and measured
in billions of cubic feet (Bcf).
Note that, because the schedules show projected cumulative volumes
rather than rates, they are not supply or demand curves in the standard sense
(although we shall refer to them as such for brevity). The projected time frame
over which the cumulative volumes are realized varies significantly across FRCs
and depends also on the oil and CO2 price, but is usually on the order of several
decades.
The main message of Figures 2 and 3 is that oil supply and CO2 demand
projections using the analog method are highly sensitive to the analog used. In
degrees Fahrenheit and a reference temperature of 14.7 psi (1 atmosphere). The actual pressure at
which pipelines deliver CO2 to oil fields is typically much higher, between 1250 and 2250 psi (McCoy
and Rubin, 2008). At these pressures and ambient temperatures, the CO2 is either a liquid or a
supercritical fluid, where the latter has properties of both a liquid and a gas.

The Economics of Enhanced Oil Recovery / 45

both figures, the curve on the left is derived using the Lost Soldier-Tensleep (LST)
project in Wyoming as analog, whereas the curve on the right is derived using
the Denver Unit-San Andres (DSA) project in West Texas as analog.
Figure 4 plots the dimensionless curves generated from the LST and
DSA projects historical EOR performance. It shows that, at the point in time
when cumulatively one HCPV worth of CO2 and water had been circulated
through the LST project after it switched to EOR, the project had produced cumulatively 0.059 HCPV worth of incremental oil. In contrast, the DSA project
had at that same point produced 0.108 HCPV, and was therefore 1.8 times more
productive.
A number of differences between the LST and DSA projects may be
relevant to explaining their quite dissimilar EOR performance. One difference
concerns timing. According to Brokmeyer et al. (1996), CO2 flooding of the LST
project commenced when 44.3% of OOIP had been recovered in the primary and
waterflooding stages, and when the waterflooding oil cut (the share of oil in
overall liquid production) had dropped to as little as 3%. In contrast, CO2 flooding
at the DSA project commenced when according to Hsu et al. (1997) only 35.5%
of OOIP had been recovered, and when according to Tanner et al. (1992) the oil
cut was still 14%. However, Jarrell et al. (2002) note that reservoir simulation
studies of west Texas reservoirs have shown that the rate of incremental oil
recovery to CO2 flooding is only slightly sensitive to the stage of waterflooding
Figure 4. Comparison of the dimensionless curves for incremental oil
production for the Lost Soldier-Tensleep analog (LST) and the
Denver Unit-San Andres analog (DSA)

46 / The Energy Journal

when CO2 flooding starts, as long as the reservoir pressure is well above the
thermodynamic MMP. Both the LST and DSA projects satisfy the latter condition.
More important, perhaps, is the difference in lithology of the project
reservoirs: whereas the Tensleep formation is a sandstone, the San Andres formation is a carbonate. Several review studies of existing EOR projects (Holtz et
al., 1999; Christensen et al., 2001; Hustad, 2004) have noted that CO2 flooding
tends to yield somewhat higher incremental recovery rates in carbonate reservoirs
than in sandstone ones. Hustad (2004), for example, notes that of 115 worldwide
CO2 floods in a database maintained by the Norwegian consulting firm SINTEF
Petroleum Research, average incremental oil recoveries for sandstone and carbonate reservoirs are 12% and 17% of OOIP, respectively.
A third difference concerns the degree of fracturing of the two reservoirs,
which is much higher for the Tensleep. Injected CO2 is more likely to flow to
production wells through such fractures, bypassing much of the oil and hence
reducing incremental oil recovery. This may explain why the LST projects ultimate recovery appears to below the sandstone average (its dimensionless curve
in Figure 4 asymptotes to roughly 9%).
Lastly, the injection history of the two projects has been quite different
as well. Whereas the LST project has consistently maintained a 1:1 ratio of water
to CO2 (commonly referred to as the water-alternating-gas or WAG ratio), the
DSA project started out injecting pure CO2, and thereafter gradually increased
the WAG ratio.
Whatever the full explanation for the differential EOR performance of
the two analog projects may be, it is evident that the choice of analog significantly
affects estimates of how much incremental oil the examined candidate projects
in the PRB will produce, how much CO2 they will demand, and thereby how
profitable they are likely to be. At the individual project level, this is illustrated
in Figure 2 by the rightward jump in both oil supply curves when CO2 flooding
of the large Finn-Shurley Turner field becomes profitable. If the LST analog is
used, this is estimated to require an oil price of at least $113, and the field is then
estimated to cumulatively produce 20.8 million incremental barrels of oil over its
economic lifetime. But if the DSA analog is used, the same field is estimated to
become profitable when the oil price is only $93, and to cumulatively produce as
much as 56.5 million barrels. At the aggregate level, Figure 2 shows that at our
reference oil price of $100 and CO2 price of $3, estimated incremental oil supply
for the basin as a whole is more than three times as high if the DSA analog is
used than if the LST analog is used (although the ratio is smaller at both lower
and higher oil prices). Figure 3 shows that estimated CO2 demand is higher with
the DSA analog as well.
As noted in the previous section, the analog method is considered more
reliable the more closely the reservoir characteristics of a proposed project match
those of the analog used. Since all fields in our study produce from sandstone

The Economics of Enhanced Oil Recovery / 47

rather than carbonate reservoirs16 and a handful in fact produce from reservoirs
that are part of the Tensleep formation, our estimates based on the LST analog
are perhaps more likely to be predictive. That said, many (about a third) of the
fields in our study produce from the Minnelusa formation, a carbonate-rich sandstone that tends to be much less fractured than the Tensleep. For these fields, the
LST analog may well turn out to underpredict EOR performance.17
As for the maturity of the fields in our study (insofar as this matters for
EOR performance, despite Jarrell et als assertion to the contrary), the median
current recovery factor for the FRCs in our study is 30% of OOIP, which is closer
to that of the DSA project when it switched to EOR, but the median oil cut is
7%, which is closer to the initial oil cut of the LST project.
More generally, the considerable heterogeneity of reservoirs in the PRB
implies that the true aggregate curves are likely to differ from both of the curves
plotted. Nevertheless, subject to these caveats and inevitable data constraints,18
the results presented here provide at least a rough, order-of-magnitude estimate
of EOR market conditions, illustrating the potential usefulness of the analog
method.
To put the incremental oil supply estimates in perspective, cumulative
oil production to date from all 97 FRCs in our study combined is 978 MMbo, or
about 25% of their combined OOIP of 3.9 Bbo. Current combined oil production
is 5.9 MMbo/yr, but is declining: at our baseline oil price of $100, we estimate
cumulative future production under continued waterflooding to be just 63 MMbo.
Figure 2 shows that at our baseline CO2 price of $3/Mcf and using the LST analog,
estimated cumulative future production from EOR is 74.5 MMbo higher, or 137.5
MMbo in total.
In Figure 1, FRCs that can profitably switch to EOR at these reference
prices are plotted as circles, with the circle areas proportional to each FRCs
projected incremental oil supply. Note that similar maps showing each FRCs
CO2 demand under various price scenarios could be used, for example, to plan
the trajectory of CO2 pipelines.
4. SENSITIVITY ANALYSIS
In practice, many of the geological and engineering parameters used in
the analog method (as listed in Table 3 of the appendix) are measured quite
imprecisely. For example, measures such as reservoir depth, thickness, and permeability are typically only available as averages for a given FRC, ignoring pos16. Some fields in the PRB do produce from carbonate reservoirs, but none of these met the
cumulative production hurdle for being part of our study.
17. Personal communication with J. Michael Boyles, a geologist previously with the Enhanced
Oil Recovery Institute at the University of Wyoming.
18. Better data will not be forthcoming until individual operators actively contemplate switching
to EOR, at which point they will want to invest in detailed geological studies and modeling exercises
for their specific FRC.

48 / The Energy Journal

sibly significant heterogeneity in these measures across different sections of the
reservoir. The key scaling parameter of injectivity is usually estimated from historical per-well water injection rates, even though these rates may be highly variable both across wells and over time. As for HCPV, the other key scaling parameter, this is sometimes based on published estimates of an FRCs original oil in
place; sometimes on a volumetric calculation combining estimates of the reservoirs area, thickness, porosity, and residual oil saturation; and in rare cases on
simple extrapolation from cumulative extraction to date. All three methods have
drawbacks, and the resulting estimate should not be taken as more than a rough
guess. Fortunately, for most of these parameters there is no reason to suspect a
systematic bias in one direction or another across all FRCs in a basin. As a result,
when it comes to estimating aggregate oil supply or CO2 demand for all FRCs
combined, it is reasonable to expect that errors will tend to cancel out.
To investigate how sensitive our estimates are to variations in key geological and engineering parameters, we performed a Monte Carlo analysis, the
results of which are presented in the top panel of Table 2. Each of the parameters
listed in the first column was randomized by selecting, independently for each
FRC, 200 values from a symmetric beta distribution with mean li equal to our
central estimate of the parameter for the i-th FRC, support (0,2 li ) , and 95%
Table 2. Results of sensitivity analysis using the Lost Soldier-Tensleep
analog. Geological/engineering parameters were varied
independently across FRCs to generate 95% confidence intervals
for oil supply and CO2 demand. Cost and other parameters were
varied by 50% in the same direction for all FRCs simultaneously.

Geological/engineering parameters
Injectivity
Original oil in place
Reservoir area
Cost parameters
Royalties, severance & property taxes
Well drilling costs
Well conversion costs
Gas processing costs (fixed)
Gas processing costs (variable)
Other operating costs
Other parameters
Reference CO2 price
Reference oil price
Internal rate of return
Maximum pattern size

Change in
oil supply

Change in CO2
demand

95% conf. intvl.

95% conf. intvl.

7.4%
9.4%
2.7%

9.1%
11.0%
12.9%

12.9%
12.5%
8.3%

11.9%
11.1%
36.9%

f 50%

F 50%

f 50%

F 50%

23.5%
20.9%
15.5%
5.6%
7.4%
7.6%

21.1%
11.7%
9.7%
3.7%
5.4%
4.1%

53.1%
38.1%
36.2%
14.0%
19.1%
20.6%

41.3%
15.8%
17.2%
9.2%
12.2%
13.2%

f 50%

F 50%

f 50%

F 50%

18.4%

21.0%

39.5%
34.3%

19.1%
35.5%

74.9%
92.0%
57.2%

62.4%
33.2%
81.9%

The Economics of Enhanced Oil Recovery / 49

confidence interval (0.5 li ,1.5 li ) . The second and third columns show the resulting 95% confidence intervals for incremental oil supply. The intervals are
expressed as percentage deviations from our QoLST estimate shown in Figure 2,
after averaging these deviations over the oil-price range shown (but dropping
values at which supply is zero). The fourth and fifth columns show the analogous
95% confidence intervals for CO2 demand. The first row of the Table therefore
shows, for example, that if the uncertainty about injectivity rates is represented
by the above-described beta distributions, then incremental oil supply lies between 7.4% below and 9.1% above the QoLST estimate shown in Figure 2 for 95%
of the simulation trials.
Error independence across FRCs is of course not a reasonable assumption for the cost parameters discussed in Section 2; deviations from our central
estimates for these parameters will clearly affect all FRCs in the same way. The
second panel of Table 2 therefore shows, for each of the cost categories listed in
the first column, the average effects on incremental oil supply and CO2 demand
of simply reducing or increasing the relevant per-unit costs by 50%. The first row
of the panel shows, for example, that if the royalty rate as well as severance and
property tax rates were to be simultaneously reduced (increased) by 50%, the
resulting incremental oil supply would on average increase by 23.5% (drop by
21.1%) relative to the QoLST estimate shown in Figure 2. Clearly, the relative
magnitudes of the induced changes are reflective of the different cost categories
share of overall EOR costs. Next to taxes, the up-front investment costs associated
with drilling new wells are the most important cost category, followed by the upfront costs associated with converting and equipping wells for EOR use.
The third panel of Table 2 shows the effects of varying four other key
parameters by 50% in either direction. Doing so for our reference CO2 price of
$3/Mcf has an effect on oil supply comparable in magnitude to that of varying
taxes or drilling costs. Dropping our reference oil price from $100/bo to $50/bo
shifts in the CO2 demand curve by 74.9% on average, whereas increasing the
reference price to $150/bo increases demand by 62.4% on average.
A further parameter of particular interest is the discount rate applied by
oil-field operators in calculating a proposed projects NPV, i.e., their internal rate
of return (IRR). According to industry contacts, a common IRR applied to profits
net of royalties, severance taxes, and property taxes but gross of income taxes is
20%; this is the rate used in the simulations underlying Figures 2 and 3. Note
that the rate is considerably above conventional riskless rates of return, reflecting
the high level of project failures in the oil industry. The next-to-last row of Table
2 reports the average effects of reducing the IRR for all FRCs to 10%, or raising
it to 30%.
A final parameter considered in our sensitivity analysis is the maximum
pattern size applied to EOR floods. Because many oil fields in the PRB are approaching the economic limit of their secondary recovery phase, it is not uncommon for operators to either temporarily or permanently shut down wells of underperforming patterns. Were these operators to switch to EOR, however, they

50 / The Energy Journal

may well bring temporarily abandoned wells back into operation, or even drill
new wells to achieve a desired well configuration. Our baseline simulations assume that operators will do so up to the point where the average EOR pattern
size is at most 80 acres.19 Reducing this maximum pattern size has two opposing
effects on EOR profitability. On the one hand, it reduces profitability, by increasing the number of new wells that must be drilled and thereby the up-front capital
cost of switching to EOR. On the other hand, reducing the pattern size also
increases the rate at which CO2 cycles through a reservoir, thereby speeding up
incremental oil recovery and increasing profitability.20 We find, however, that the
former effect dominates over the price ranges for oil and CO2 considered: reducing the maximum pattern size to 40 acres shifts in the oil supply curve by 34.3%
on average, while increasing the pattern size to 120 acres shifts out supply by
35.5% on average.
5. CONCLUDING DISCUSSION
In this paper, we have shown how, given sufficiently detailed data on
the geology and extraction history of FRCs, along with the documented experiences of mature or completed EOR projects, it is possible to forecast which FRCs
in an oil-producing region can profitably implement EOR at given prices for oil
and CO2. Aggregating these forecasts generates region-wide incremental oil production and CO2 demand schedules.
The availability of such forecasts has clear benefits to EOR development.
Knowing the location of other candidate EOR projects facilitates cost sharing and
coordination of CO2 and oil shipments by projects in proximity to each other.
Knowing the derived demand for pure CO2 allows emitters to consider locating
closer to points that will demand CO2 as a commodity. It may also help pipeline
authorities to more efficiently design delivery systems connecting emitters to EOR
projects.
The analog method used to generate the forecasts essentially extrapolates experience at an existing, mature EOR projectthe analogto new, candidate projects. Not surprisingly, we find that the forecasts are quite sensitive to
the specific choice of analog. Ideally, this choice should be driven by the geological properties of each candidate project, which should match those of the analog
as closely as possible. Currently, however, only a small number of analogs for
both incremental oil production and CO2 circulation are available. This problem
should diminish over time, as EOR becomes more widely applied.

19. For the common five-spot pattern (see footnote 9) this corresponds to an average well spacing
of 40 acres.
20. In terms of Figure 4, halving the pattern size halves the time by which, for example, 1 HCPV
worth of CO2 and water can be injected into the reservoir, thereby halving the time by with 0.059
HCPV worth of incremental oil is recovered.

The Economics of Enhanced Oil Recovery / 51

An important set of questions we leave for future work is how potential
subsidies for CO2 sequestration might affect market conditions for EOR, and how
the analog method might be adapted to forecast these effects. As noted for example by Jessen et al. (2005), EOR operators may turn to co-optimization of oil
recovery and CO2 sequestration, by adjusting the CO2 content of their injection
stream. They may also choose to continue CO2 injection even after incremental
oil has been extracted. Although such changes would not invalidate the analog
method used in this paper, they would imply a need for adjustments. Specifically,
given that existing EOR projects have been operated if anything with the aim of
minimizing CO2 sequestration, dimensionless curves based on such projects experience are likely to become less accurate as predictors of future projects performance. Instead, new dimensionless curves will have to be developed, possibly
based on reservoir simulations.
REFERENCES
Brokmeyer, R. J., D. C. Borling, and W. T. Pierson (1996). Lost Soldier Tensleep CO2 Tertiary
Project, Performance Case History: Bairoil, Wyoming. Paper # SPE 35191.
Christensen, J. R., E. H. Stenby, and A. Skauge (2001). Review of WAG Field Experience. SPE
Reservoir Evaluation & Engineering, 4(2): 97106.
Dahowski, R. T., J. J. Dooley, C. L. Davidson, S. Bachu, and N. Gupta (2005). Building the Cost
Curves for CO2 Storage: North America. Report 2005/3, IEA Greenhouse Gas R&D Programme.
DOE (2006). Basin Oriented Strategies for CO2 Enhanced Oil Recovery: Rocky Mountain Region.
Washington, DC: Department of Energy, Office of Fossil Energy, Office of Oil and Natural Gas,
Prepared by Advanced Resources International.
DOE (2007). Carbon Sequestration Atlas of the United States and Canada. Washington, DC: Department of Energy, Office of Fossil Energy, National Energy Technology Laboratory.
DOE (2008). Storing CO2 with Enhanced Oil Recovery. Washington, DC: Department of Energy,
National Energy Technology Laboratory, Prepared by V. Kuuskraa, R. Ferguson, Advanced Resources International, DOE/NETL-402/1312/02-07-08.
EIA (2006). Oil and Gas Lease Equipment and Operating Costs 1988 Through 2006. Energy Information Administration. Document available online at http://www.eia.doe.gov/pub/oil_gas/ natural_gas/data_publications/cost_indices_equipment_production/current/coststudy.html.
Emera, Mohammed K. and Hemanta K. Sarma (2005). Use of Genetic Algorithm to Estimate CO2Oil Minimum Miscibility PressureA Key Parameter in Design of CO2 Miscible Flood. Journal
of Petroleum Science and Engineering, 46(12): 3752.
EPA (2008). Inventory of U.S. greenhouse gas emissions and sinks: 19902006. Washington, DC:
Environmental Protection Agency, EPA 430-R-08-005.
Fox, Charles E. (1995). Cost Estimation Parameters. University of Texas of the Permian Basins
Center for Energy & Economic Diversification (CEED) CO2 Flooding Shortcourse No. 2, September 12, 1995, Section F.
Griffin, James M. (2009). A Smart Energy Policy: An Economists Rx for Balancing Cheap, Clean,
and Secure Energy. New Haven, CT: Yale University Press.
Holtz, Mark H., Peter K. Nance, and Robert J. Finley (1999) Reduction of Greenhouse Gas Emissions
through Underground CO2 Sequestration in Texas Oil and Gas Reservoirs. Digital Publication
Series, 99-01, Gulf Coast Carbon Center (GCCC).
Holtz, Mark H., Peter K. Nance, and Robert J. Finley (2001). Reduction of Greenhouse Gas Emissions through CO2 EOR in Texas. Environmental Geosciences, 8(3): 187199.
Hsu, C-F., J. I. Morell, and A. H. Falls (1997). Field-scale CO2-Flood simulations and their impact
on the performance of the Wasson Denver Unit. SPE Reservoir Engineering, 12(1): 411.

52 / The Energy Journal

Hustad, Carl-W. (2004). Large-Scale CO2 Sequestration on the Norwegian Continental Shelf: A
Technical, Economic, Legal and Institutional Assessment. Project No., 151393/210, Norwegian
Research Council.
Jarrell, P. M., C. E. Fox, M. H. Stein, and S. L. Webb (2002). Practical Aspects of CO2 Flooding.
SPE Monograph Series. Society of Petroleum Engineers.
Jessen, Kristian, Anthony R. Kovscek, and Franklin M. Orr, Jr. (2005). Increasing CO2 storage in
oil recovery. Energy Conversion and Management, 46(2): 293311.
Lucken, J. E. (1969). Raven Creek. WGA Earth Science Bulletin, 2(4): 2427.
Mack, J. C. and M. L. Duvall (1984). Performance and Economics of Minnelusa Polymer Floods.
Paper # SPE 12929.
McCoy, Sean T. and Edward S. Rubin (2008). An Engineering-Economic Model of Pipeline Transport of CO2 with Application to Carbon Capture and Storage. International Journal of Greenhouse
Gas Control, 2(2): 219229.
McPherson, Brian J. O. L., Weon Shik Han, and Barret S. Cole (2008). Two Equations of State
Assembled for Basic Analysis of Multiphase CO2 Flow and in Deep Sedimentary Basin Conditions. Computers & Geosciences, 34(5): 427444.
Moritis, Guntis (2008). More US EOR projects start but EOR production continues decline. Oil &
Gas Journal, 106(15): 4146.
NPC (1984). Enhanced Oil Recovery. Washington, DC: National Petroleum Council.
Tanner, C. S., P. T. Baxley, J. G. Crump, III, and W. C. Miller (1992). Production Performance of
the Wasson Denver Unit CO2 Flood. Paper # SPE 24156.

APPENDIX
In this appendix, we illustrate the analog method using a medium-sized
field in our study area, Raven Creek-Minnelusa (RCM), as a worked example.21
Table 3 lists the data required for applying the analog method, although strictly
speaking the list contains some redundancies. Reservoir HCPV, for example, can
be estimated either as OOIPBoi or as Ah Soi 43,560(ft2/acre)/5.615(ft3/rb). Similarly, injectivity can be estimated either from past water injection rates or from
reservoir thickness h and permeability k.
Step 1: Estimate per-pattern HCPV. Mack and Duvall (1984) provide a
direct estimate of RCMs OOIP, so we estimate HCPV for the reservoir as a whole
as OOIPBoi 81,169,000 rb. (The alternative formula Ah Soi 43,560/5.615
would yield a somewhat higher estimate of 85,052,000 rb.) Assuming that 80%
of the total reservoir area will be covered with 80-acre patterns (our baseline
maximum pattern size) the CO2 flood will have nceil(0.82,975/80)30 patterns. Letting H denote per-pattern HCPV, we therefore have H 2,706,000 rb.
Step 2: Estimate water and CO2 injection rates. We assume that, averaged over the water and CO2 injection cycles, per-well injectivity for the EOR
project is equal to that for the waterflood, so qi, eor q0wi 36,833 bl/ptn-mnth.
Assuming equal-sized alternating slugs of water and CO2 will be injected (i.e., a

21. Readers interested in further detail are referred to a presentation available from the University
of Wyomings Enhanced Oil Recovery Institute (EORI), at http://eori.gg.uwyo. edu/downloads/
CO2_Conf_2009/Presentation%20PDF/ProfitableCO2ProWyo092909.pdf. The results shown in that
presentation are based on less recent production/injection data and cost figures, however.

The Economics of Enhanced Oil Recovery / 53

Table 3. Data requirements for the analog method, with example values
for the Raven Creek-Minnelusa field
Variable
OOIP
Boi
Boc
API
T
MMP
BCO2
pf
d
h
A

k
Soi
npc
nic
nta
Q0op
d
Q0wp
q0wi
R

Value
Original oil in place
Oil formation vol. factor (initial)
Oil formation vol. factor (current)
Oil gravity
Temperature
Minimum miscibility pressure
CO2 formation vol. factor
Fracture pressure
Depth
Thickness
Area
Porosity
Permeability
Oil saturation (initial)
Existing producer wells
Existing injector wells
Temporarily abandoned wells
Oil production (end of waterflood)
Oil production decline rate (e.o.w.)
Water production (e.o.w.)
Water injection per well
Gas-oil ratio

73,790,000
1.1
1.089
33
200
3,082
0.7001
5,548
8,380
37
2,975
0.12
50
0.83
15
13
12
6,833
0.45
396,536
36,833
0

stba
rbb/stb
rb/stb
API
F
psi
rb/Mcf
psi
ft
ft
acres
(fraction)
md
(fraction)

stb/mo
%/mo
bl/mo
bl/wl-mo
Mcf/bl

Source
Mack and Duvall (1984)
TORISc
TORIS
TORIS
TORIS
Derivedd
Derivede
WOGCCf
Lucken (1969)
Lucken (1969)
Lucken (1969)
Lucken (1969)
Lucken (1969)
TORIS
WOGCC
WOGCC
WOGCC
IHSg (estimated) h
IHS (estimated) i
IHS (estimated) j
IHS (estimated) k
IHS (estimated) l

Stock-tank barrels (42 gallons U.S. at 60 F and 14.7 psi).

Reservoir barrels (42 gallons U.S. at reservoir temperature and pressure).
c
Total Oil Recovery Information System database, U.S. Department of Energy, National Petroleum
Technology Office.http://www.netl.doe.gov/technologies/oil-gas/Software/database.html.
d
0.87022
Derived from API and T using DOE (2006) correlation MWC
combined
5 4247.98641API
1.2785
with Emera and Sarma (2005) correlation MMP0.00726538 T1.164 (MWC
.
5 )
e
Derived from T and MMP using sw_SPECIFIC_DENSITY.m Matlab subroutine provided by
McPherson et al. (2008) at http://www.iamg.org/documents/oldftp/VOL34/v34-05-01.zip
f
Wyoming Oil and Gas Conservation Commission. http://wogcc.state.wy.us.
g
IHS PI/Dwights PLUS database. http://energy.ihs.com.
h
0op from linear regression logQop
Predicted end value Q
t dt of log monthly oil production on a
time trend t35,34,...0 for the most recent 36 months of production data.
i
Estimated coefficient on t in above regression.
j
Estimated using same procedure as that for oil production, but using log monthly water production.
k
Median water injection rate per actively injecting well since 1985.
l
Median ratio of monthly hydrocarbon gas production to oil production for the most recent 36 months
of production data.
a
b

1:1 WAG ratio), the average rate of water injection will then be qwi, eor
0.5 qi, eor 18,416 bl/ptn-mnth, while that of CO2 will be qci 0.5 qi, eor / BCO2
41,788 Mcf/ptn-mnth.

54 / The Energy Journal

Step 3: Estimate time paths of incremental oil and CO2 production using
the dimensionless curves. Dividing qi, eor by H gives a total injectivity of 0.01361
HCPV/ptn-mnth. Cumulating this for any given number of months after starting
the CO2 flood yields a point on the horizontal axis of the dimensionless curve for
incremental oil production (see Figure 4). Reading off the corresponding point
on the vertical axis yields cumulative incremental oil production by that same
number of months in HCPV/ptn, or, after multiplying by H/ Boc , in stb/ptn. Differencing the cumulative values yields the corresponding time path of incremental
oil production rates, qop,inc
, in stb/ptn-mnth. Lastly, multiplying by the number of
t
patterns n yields the predicted time path of incremental oil production for the
inc
field as a whole, Qop,
, in stb/mnth. Applying the same steps to the dimensionless
t
curve for CO2 production (but multiplying cumulative production in HCPV/ptn
on the vertical axis by H/ BCO2 ) yields the predicted time path of CO2 production,
Qcp
t , in Mcf/mnth.
Step 4: Estimate time paths of baseline and EOR oil production, water
production, non-CO2 (hydrocarbon) gas production, and EOR purchases of CO2.
bas
Baseline oil production Qop,
from a continued waterflood is assumed to cont
bas
tinue declining at rate d, so Qop,
Q0op edt . EOR oil production is then
t
op, eor
op, bas
op, inc
Qt
Qt
Qt
. Overall water production is assumed to change in a
manner that keeps overall liquids (oil plus water plus possibly CO2) production
at reservoir conditions constant over time, and equal to overall liquids injection.
bas
Baseline water production is therefore Qwp,
Q0wp Q0op (1edt ) Boc . EOR
t
wp, eor
i, eor
op, eor
water production is Qt
nq Qt
Boc Qcp
t B CO2 . Hydrocarbon gas
production is assumed to maintain a constant ratio to oil production, so
bas
bas
and Qtgp, eor RQtop, eor . For RCM, no gas production is reQgp,
RQop,
t
t
ported, however, which we interpret to imply that R0 . Lastly, CO2 purchases
are calculated as the difference between CO2 production (all of which is
Qcm
t
cp
ci
recycled) and injection, so Qcm
t Q t nq .
Step 5: Estimate up-front costs of well drilling, conversion, and equipment. With on average one injector and one producer for each of n30 patterns,
the CO2 flood will require 2 nnpc nic nta 20 new wells to be drilled, at
(given RCMs depth in the PRB) $1.31 million per well.22 Converting and equipping any kind of well (newly drilled, existing injector, existing producer, or temporarily abandoned) for use as a CO2-flood injector cost $152,000 plus $19/ft;
converting and equipping wells for use as CO2-flood producers cost $142,000 for
existing producers, and $325,000 plus $19/ft for newly drilled wells or existing
non-producers.23 Total costs sum to about $45.0 million.
Step 6: Estimate the up-front capital cost of gas recycling. The required
capacity of the gas recycling plant is determined by the peak of projected com22. Based on personal communication with Bob King, engineer at Wold Oil Properties, Inc., in
Casper, Wyoming. All costs are for the year 2006, the most recent year for which oil-industry cost
indices are available from the Energy Information Administration (EIA, 2007).
23. Based on personal communication with Charles E. Fox, Vice President of Operations and

The Economics of Enhanced Oil Recovery / 55

bined production of CO2 and (in this case zero) hydrocarbon gas, which for RCM
is 36.9 thousand Mcf/day. We assume the gas plant operates at inlet pressure
ps 225 psi, and that the CO2 flood is operated at a reservoir injection pressure
of max( pf 200,MMP)5,348 psi. Subtracting from this the pressure gain of
0.28(psi/ft)8,3802,346 psi obtained from the CO2 column in the injection
well yields the plants required discharge pressure, pd 3,001 psi. Based on a
rule-of-thumb formula provided by Fox (1995), the gas throughput capacity and
pd / ps ratio combined yield an estimated power requirement for the plant of 6,361
hp, which at a cost per hp of $1,52024 yields a capital cost for the recycling plant
of about $9.7 million.
Step 7: Estimate operating costs of gas recycling. Labor and maintenance
operating costs for gas recycling are $80 per hp-year. Electricity costs, at 7,000
kWh/hp-year and Wyomings electricity price for industrial users of 4.03 cents/
kWh, amount to an additional $282 per hp-year. The overall gas recycling operating cost is about $2.30 million/year.
Step 8: Estimate other operating costs. Based on data in EIA (2007),
operating costs were the waterflood to continue are estimated at $0.233 per barrel
of total liquids produced plus $31,000 per well-year for each of RCMs current
28 wells. Based on a rule of thumb given in Fox (1995), the corresponding costs
for the CO2 flood are assumed to be 10% higher, whereby the per-well costs apply
to the CO2 floods 60 wells.
Step 10: Determine the terminal times for the continued waterflood and
CO2 flood. At our reference oil price of $100, oil revenues net of 16% royalty,
6% severance tax, and 6% property tax drop below operating costs after Tbas
20.5 years of continued waterflooding. At the same oil price and our reference
CO2 price of $3/Mcf, net revenues from the CO2 floods total oil production (i.e.,
baseline and incremental oil production combined) drop below operating costs,
including the cost of CO2 purchases, after Teor 13.5 years.
Step 11: Compare the NPV of the continued waterflood and CO2 flood.
At the reference prices and our baseline internal rate of return of r0.20 , the
NPV bas of continuing the waterflood is $15.1 million, while the NPV eor of switching to a CO2 flood, net of initial capital costs, is substantially higher, at $89.6
million. Switching is therefore optimal.
Step 12: Calculate cumulative incremental oil production and CO2 demand. Cumulative oil production under the continued waterflood, up to Tbas , is
1.016 million barrels, while that under the CO2 flood, up to Teor , is 6.870 million
barrels. Incremental production is therefore 5.854 million barrels, or 7.9% of
OOIP. Cumulative CO2 demand is 50.4 Bcf.

Technology at Kinder Morgan CO2 Company, LP in Houston, Texas, updating figures given in Fox
(1995).
24. Based on personal communication with Mark Nicholas, President of Nicholas Consulting
Group in Midland, Texas.

56 / The Energy Journal