Rate-Decline Analysis for
Fracture-Dominated Shale Reservoirs
Anh N. Duong, SPE, ConocoPhillips
Summary
Traditional decline methods such as Arps’ rate/time relations and
their variations do not work for wells producing from supertight
or shale reservoirs in which fracture flow is dominant. Most of the
production data from these wells exhibit fracture-dominated flow
regimes and rarely reach late-time flow regimes, even over several
years of production. Without the presence of pseudoradial and
boundary-dominated flows (BDFs), neither matrix permeability
nor drainage area can be established. This indicates that matrix
contribution is negligible compared with fracture contribution,
and the expected ultimate recovery (EUR) cannot be based on a
traditional concept of drainage area.
An alternative approach is proposed to estimate EUR from
wells in which fracture flow is dominant and matrix contribu-
tion is negligible. To support these fracture flows, the connected
fracture density of the fractured area must increase over time.
This increase is possible because of local stress changes under
fracture depletion. Pressure depletion within fracture networks
would reactivate the existing faults or fractures, which may breach
the hydraulic integrity of the shale that seals these features. If
these faults or fractures are reactivated, their permeabilities will
increase, facilitating enhanced fluid migration. For fracture flows
at a constant flowing bottomhole pressure, a log-log plot of rate
over cumulative production vs. time will yield a straight line with
a unity slope regardless of fracture types. In practice, a slope of
greater than unity is normally observed because of actual field
operations, data approximation, and flow-regime changes. A rate/
time or cumulative production/time relationship can be established
on the basis of the intercept and slope values of this log-log plot
and initial gas rate.
Field examples from several supertight and shale gas plays for
both dry and high-liquid gas production, and for oil production
were used to test the new model. All display the predicted straight-
line trend, with its slope and intercept related to reservoir types.
In other words, a certain fractured flow regime or a combination
of flow types that dominate a given area or play because of its
reservoir-rock characteristics and/or fracture-stimulation practices
all produce a narrow range of intercepts and slopes. An individual-
well performance or EUR can be derived that is based on this range
if the best 3-month average or the initial production rate of the well
is already known or estimated. The results show that this alterna-
tive approach is easier to use, gives a reliable EUR, and can be
used to replace the traditional decline methods for unconventional
reservoirs. The new approach is also able to provide statistical
methods to analyze production forecasts of resource plays and
to establish a range of results of these forecasts, including prob-
ability distributions of reserves in terms of P90 (lower side) to
P10 (higher side).
Introduction
Unconventional reservoirs, especially wet shale gas and oil, are
currently being pursued aggressively for new development in
both the US and Canada. Forecasting production and estimating
Copyright © 2011 Society of Petroleum Engineers
This paper (SPE 137748) was accepted for presentation at the Canadian Unconventional
Resources and International Petroleum Conference, Calgary, 19–21 October 2010 and
revised for publication. Original manuscript received for review 29 September 2010. Revised
manuscript received for review 27 January 2011. Paper peer approved 23 March 2011.
June 2011 SPE Reservoir Evaluation & Engineering
reserves accurately for these resource plays have become more
urgent and important than ever before.
Mattar et al. (2008) have discussed various techniques for
production analysis and forecasting of shale-gas reservoirs, while
Lee and Sidle (2010) have analyzed some of the commonly used
procedures for forecasting and identifying key strengths and
limitations of these techniques. The commonly used method for
decline-curve analysis is Arps’ hyperbolic rate decline (Arps 1945),
if adequate production data are available. Fetkovich type curves
(1980) are not appropriate because the matched b value is usually
greater than unity. For horizontal wells with multiple transverse
fractures, numerical-simulation modeling is a highly preferred
option. There are, however, some validity and applicability issues
with these techniques.
The hyperbolic decline equation is conveniently used because
it can have a “best fit” for the long transient linear-flow regime
observed in shale-gas wells with b values greater than unity. Lee
and Sidle (2010) showed that values of b equal to or greater than
unity can cause the reserves to have physically unreasonable
properties. To avoid this drawback, “stretched-exponential” models
have been proposed recently (Valkó 2009; Ilk et al. 2008, 2009).
The current simulation models that are available in the mar-
ket are still using conventional technologies for unconventional
reservoirs. Some of the assumptions used in simulation modeling
are inconsistent with the field-data observations such as pressure
initialization and radial (or elliptical) transient flow. Field data
indicate that the pressure transition through a shale-gas zone is at
a disequilibrium state, while pressure initialization in simulation
modeling is based on an equilibrium state with its fluid gradient.
Newsham and Rushing (2002) showed that pressure gradients
measured in the Lower Bossier can be as high as 63 kPa/m
(3 psi/ft) or more than 25 times its saturated gas gradient or three
times the lithostatic gradient. Spencer (1987) presents some pres-
sure gradients in overpressured and gas-bearing shale plays in the
Rocky Mountain region. The interpreted pressure profile across
the active zone for wells in the Piceance basin shows a pressure
gradient of approximately 36 kPa/m.
The concept of stimulated rock volume (SRV) that is supported
by microseismic monitoring of hydraulic-fracture treatments is not
confirmed by the observed radial flow from a numerical simulator,
as pointed out by Mattar et al. (2008). These authors claim that
the radial flow of fractured shale wells observed in a numerical
simulator is more likely to be a false radial flow and may not be
representative of the enhanced permeability of the SRV.
This paper introduces an empirically derived decline model that
is based on a long-term linear flow in a large number of wells in
tight and shale-gas reservoirs. On the basis of this model, a new
method has been developed for production analysis and forecasting
of unconventional reservoirs. This method also uses probability
distributions of reserves in forecasting resource plays, to represent
any uncertainty in reserves estimation.
Method Development
Long-Term Linear Flow. Fig. 1 shows the production history of
a vertical well producing from the Barnett shale. Production data
show a half-slope line, indicating linear flow lasting for more than
5 years of production.
Long-term linear flow in a large number of wells in tight or
shale gas has been observed as early as 1976 (Bagnall and Ryan
1976). Since then, many researchers have tried to explain what
377
 PRODUCTION HISTORY FOR A SHALE GAS WELL
100 100
Press. Corrected Rate
Measured Rate
Wellhead Pressure

Rate, 103 m3/d

Skin effect

WHP, MPa
10 10

Skin effect and

Surface choking

1 1
1 10 100 1,000 10,000
Days

Fig. 1—A log-log plot of field production and pressure data for a shale-gas well. Production data show a half-slope line, indicat-
ing linear flow lasting for more than 5 years of production.

causes this phenomenon. Several authors have shown that this The gas cumulative will be
linear flow exists in any type of fractures whether they are infi-
t 1− n
G p = ∫0 qdt = q1
nite- (Agarwal et al. 1979) or finite-conductivity (Cinco-Ley et al. t
. . . . . . . . . . . . . . . . . . . . . . . . . . . (2)
1978; Cinco-Ley and Samaniego 1981a, 1981b), single or multi- (1 − n)
stage (Bello and Wattenbarger 2010), hydraulic (Arévalo-Villagrán
et al. 2001), or natural fractures (Arévalo-Villagrán et al. 2006). This gives
The difference among these models is the length of this transient
linear-flow regime.
To prolong the linear-flow regime in any of those models, q (1− n) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (3)
=
all one needs to do is enlarge the term xf2/km in which xf is the Gp t
effective fracture half-length of the fracture system and km is the
matrix permeability. For any given shale-gas reservoir, matrix
permeability is fixed. Thus, the connected fracture density of the Field Applications. Field data from several shale-gas plays were
fractured area must be increasing over time to support this fracture used to test Eq. 3. These field data were operated under the actual
flow over the life of a producing well. This increase is achievable field conditions that may violate some of the ideal assumptions
because of local stress changes under fracture depletion (Warpinski used to derive Eq. 3. Some of the results are shown in Fig. 2 in
and Branagan 1989). Pressure depletion within fracture networks which log-log plots for q/Gp vs. time in days were constructed.
would reactivate the existing faults or fractures, which may breach These plots all give a log-log straight line with a negative slope,
the hydraulic integrity of the shale that seals these features. If −m, and an intercept of a. That is,
these faults or fractures are reactivated, their permeabilities will
increase, facilitating enhanced fluid migration. A fractured shale q
model that imitates this long-term linear flow and nonequivalent = at − m. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (4)
pressure distribution will be briefly discussed later and is the Gp
subject of another paper.
If a fracture flow regime (either linear or bilinear) is prolonged Note that the slope is negative but m is always positive. As shown
over the life of a well, the gas-flow rate q will be later, m is always greater than unity for shale reservoirs. If m is
less than unity, it may indicate a conventional tight well.
q = q1t − n , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (1) Equations for q and Gp can be derived as shown in Appendix A:
a 1− m
where n is one-half for linear flow, n is one-quarter for bilinear q (t −1)
= t − m e 1− m . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (5)
flow, and q1 is the flow rate at Day 1. q1

1 1 1

y = 1.1437x –1.1054 y = 1.2776x –1.1381 y = 1.3665x –1.1487

0.1 R 2 = 0.9861 0.1 R 2 = 0.9808 R 2 = 0.9836
q/Gp , 1/Day

q/Gp , 1/Day

0.1
q/Gp, 1/Day

0.01 0.01 0.01

a =1.1437 a =1.2776 a =1.3665

m =1.1054 m =1.381 m =1.1487
0.001 0.001 0.001

0.0001 0.0001 0.0001

1 10 100 1,000 10,000 1 10 100 1,000 10,000 1 10 100 1,000 10,000

(a) Days (b) Days (c) Days

Fig. 2—A log-log plot of q/Gp vs. t for various shale-gas wells with more than 5 years of production. The data all form a straight
line with a slope m and an intercept a. A slope of higher than unity is normally obtained, to account for any deviations from the
ideal cases.

378 June 2011 SPE Reservoir Evaluation & Engineering

 a vs. m
3

Tight/Shallow Gas
2.5
Dry Shale Gas
2 Wet Shale Gas

1.5

a
1 a = 0.7364 m 4.7954
R 2 = 0.9417
0.5

0
0.9 0.95 1 1.05 1.1 1.15 1.2 1.25 1.3
m

Fig. 3—A correlation between a and m for various gas plays.

and 30 days of production, q3ma will be equal to a monthly average of

a 1− m the first 3 months or first 90 days. The slope of the straight line in
q1 (t −1)
Gp = e 1− m
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (6) a ⎡ 1− m
t −( 90 ) ⎤
1− m

a this case will be 3e 1− m ⎣ x ⎦

It will be shown later that a can be eliminated from Eqs. 5 and 6
if they are rewritten in forms of dimensionless rate and time. This
indicates that a and m are related variables, as shown in Fig. 3,
obtained from various gas plays. Using this correlation to convert
a into m, type curves for q/q1 vs. time in days can be constructed,
as shown Fig. 4.
Fig. 4 shows that there is a maximum flow rate, qmax and cor-
responding time tmax, for each m value of greater than unity. The
parameters qmax and tmax as a function of a and m are found in
Appendix B. Fig. 4 also indicates that tmax is normally less than
30 days. Thus, the well will flow at the maximum flow rate within
the first month of production if there are no well constraints such
as rate, WHP, or bottomhole-flowing-pressure limits.
During production, a well may deviate from fracture flow and
show high decline rate because of factors such as water production
or liquid dropout/holdup in the fracture system and wellbore. The
values of q1 and q∞(the rate at infinite time, see the Individual Well
Analysis subsection) are designed to account for this acceleration
in rate decline.
Eq. 6 indicates that Gp is proportional to q1 for any given time
t. Thus, a plot of cumulative gas against the gas rate at Day 1 of
all wells with similar a and m would fall on the same straight
line passing through the origin. The slope of this straight line is
1 1−am ( t1x− m −1)
e in which tx is the prediction time such as 24 months (730
a qmax
days) or 60 months (1,825 days), as examples. To reduce the data
uncertainties, the best 3-month average rate q3ma is used instead
of q1. Because the rate reaches maximum value within the first
.
The characteristic of a linear relation between the best 3-month
average and cumulative production has been reported by Frantz
et al. (2005) for Barnett shale wells. To validate this characteristic,
a sample of 25 Barnett wells was used. Actual best 3-month aver-
ages for both vertical and horizontal wells were plotted against their
cumulative productions of 1, 2, and 5 years, as shown in Fig. 5.
A very good match is obtained. See more details in the Analog
Type Curves subsection of the Field Examples section.
In summary, if the value any one of the three parameters 3q1,
qmax, or q3ma is known, reserves evaluations can be established for
any individual well, group of wells, development area, or play with
a given set of a and m. The q3ma is preferred because it uses a longer
production period and, hence, reduces data uncertainties.
Fig. 4 gives a set of type curves for a typical field-data plot
because they are based on the real time (days) and a real relation
between a and m. To construct type curves in a dimensionless
form, the curves in Fig. 4 were normalized using tmax and qmax. The
dimensionless time tm and dimensionless rate q/qmax will be
t
tm = . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (7)
tmax
and
m ⎛ t 1− m −1⎞
q ⎜ m ⎠⎟
= tm− m e 1− m ⎝ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (8)
The general type curve is shown in Fig. 6 for an m range of 1 to 1.96.
Note that a has been eliminated from the type curves (see Eq. 8).

q/q 1 vs. time 2.00

Best-3-Month Ave, 106 m3

10
q max
1.50 1 year 2 years
5 years

1
1.00
q/q 1

m~1.00
m=1.05 0.50
0.1 m=1.10 Dots - 25 Barnett Samples
m=1.15 Solid lines - Eq. B-5 (P50)
m=1.20
m=1.25 0.00
m=1.3 0 5 10 15 20 25 30
0.01 6 3
Cum Gas, 10 m
1 10 100 1,000 10,000
Days Fig. 5—Best 3-month averages for a sample of 25 Barnett wells
were plotted against their cumulative productions of 1, 2, and
Fig. 4—Type curves for q/q1 vs. time in days for m>1. 5 years.

June 2011 SPE Reservoir Evaluation & Engineering 379

 tm = t/tmax
td = Dit
1 q1D → q at td = 1
qmax → q at tm = 1
m=1.05

b=0 b=0.5 b=1 m=1.15

q/qmax or q/q1D 0.1

m=1.30

0.01

m=1.96 m=1.60
0.001
1 10 100 1,000 10,000
tm or tD

Fig. 6—A comparison between Arps’ and the new method’s type curves for dimensionless rate and time.

For comparison, the Arps rate/time relations (Arps 1945) were
reconstructed in the same plot using dimensionless time tD=Dit
and dimensionless rate q1D for their normal range of b from 0 to 1.
Note that q1D is defined as the dimensionless rate at tD=1, not the
same as normally based on qi. The results show that Arps’ curve of
b=0.5 is equivalent to the new curve of m=1.96 at late time (tD>10),
as shown in Fig. 6. For b>0.5, Arps curves give more-optimistic
forecasts compared to the new approach.
Field Examples
In this section, several field examples from different unconven-
tional gas plays were used to test the new approach. Examples were
selected to illustrate the decline behaviors in various completion
and production scenarios such as the cases of retrograde gas, newly
developing or developed plays, and vertical or horizontal wells. For
changes in flow regime, a and m values will be highlighted for each
example in the following discussion. A comparison of results is
also made with other methods.
Retrograde-Gas Case. Wet-shale-gas plays have become more
desirable lately because of higher liquid contents and attractive
liquid price. Unfortunately, very little is known about their flow
characteristics and how the liquid dropout in the formation and
fracture network behaves.
The following example contains production data from four
horizontal wells from a new emerging play. The production rates
from both gas and liquid were recombined and converted into
gas-rate equivalent and finally were normalized on the basis of the
highest rate from these wells, as shown in Fig. 7. Rates on Wells 2
and 3 were reinitialized because of some significant downtimes.
Fig. 7a shows that the flow regime varies from bilinear to linear
flow. Generally, wells with bilinear flow regime will have lower
maximum flow rate compared to wells with linear flow. The lower
deliverability rates and/or bilinear flow may be caused by liquid
holdup in fractures or by fracture choking.
Values of a and m obtained from a retrograde shale gas have
covered a large range of values; these vary from 1.02 to 1.67
and from 1.04 to 1.21, respectively, as shown in Fig. 7b for this
example.
Bakken Oil. An abandoned vertical well that produced from the
Bakken formation in the Viewfield pool, Saskatchewan, Canada,
for more than 26 years is shown here as an example for oil pro-
duction from a fractured shale or tight formation. Fig. 8a shows
the well was under a transient linear flow (half slope) for all of its
production life except for a few months at the start of production.
The late-time flow that indicates a big rate drop in oil rate can
be mistaken as BDF, but in reality it is caused by an unexpected
increase in water production.
The normalized decline-trend plots for the well are shown in
Fig. 8b. If q/Np is based on the total liquid produced, the plot gives
a good fitted straight line (R2 of 0.9824) with a and m values of
1.7894 and 1.1477, respectively. The values are slightly higher if
only the oil stream is used in the calculation. Thus, an increase

1
1
¼ slope ½ slope
Normalized Rates

0.1
q/G p , 1/day

y = 1.4112x –1.1121
0.1 Well 1 y = 1.0164x –1.0409
Well 2
Well 3 Well 1
0.01 Well 2
Well 4 Well 3 y = 1.18x–1.1128
Well 4 –1.2139
y = 1.6655x

0.01 0.001
1 10 100 1,000 1 10 100 1,000
Days Days
(a) (b)

Fig. 7—A condensate/shale-gas example. Flow regimes that change from linear to bilinear in this retrograde play may be caused
by liquid dropout in the fracture. There is a wider range for a and m observed in this play compared to other types.

380 June 2011 SPE Reservoir Evaluation & Engineering

 100
1
¼ slope ½ slope

Liquid Rate, m3 /d
0.1
y = 1.8316x –1.1488
10
R 2 = 0.9963

q/N p , 1/day
0.01 y = 1.7894x –1.1477
R 2 = 0.9824
Oil+Water 0.001
1
Oil Only
0.0001 Oil+Water
Oil Only
0 0.00001
1 10 100 1,000 10,000 1 10 100 1,000 10,000
(a) Days (b) Days

Fig. 8—An oil example of fractured shale with 26 years of monthly production data. Fracture-dominated flow regimes lasted for
all of its production life. Rate drop at late time is caused by an increase in water production. (a) Linear flow lasted more than 26
years (9,500 days). (b) Normalized decline-trend plots for different fluid streams.

in water production will cause both a and m to increase in value used to obtain the representative a and m parameters for the well.
(Fig. 8b). Because of the unexpected increase in water production The R2 value is used to determine the best fit of the data. An R2
at the late time, an overprediction of only 8% in the well’s EUR value of more than 0.95 is recommended.
is projected. 3. Step 3: Rate Forecast. q1 Determination. To obtain q1, gas-
flow rate should be plotted against t(a, m). From Eq. 5, we have
Individual-Well Analysis. In this example, a step-by-step pro-
cedure is shown in Fig. 9 of how to perform decline analysis on
an individual-well basis. A vertical well taken from the Barnett q = q1t (a, m), . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (9)
shale has been selected for this illustration. The well has been in
production for the last 5 years and operated with minimum water a
(t1− m −1)
production. where t (a, m) = t − m e 1− m . The q-vs.-t(a,m) plot should give
1. Step 1: Data Check and Correction. The production-data .
a straight line through the origin with slope of q1. However, this
histories such as flowing WHPs, gas rate, and water rate are plot- may not be the best selection in some cases because of the current
ted as shown in Fig. 9a. The well was initially choked back at the wellbore operating conditions. In such cases, Eq. 9 becomes
surface on the basis of the observed flat rate and high WHPs. A
correction was made for the rate on the basis of average operating
pressures. The wet-gas rate should also be corrected to the dry-gas q = q1t (a, m) + q∞, . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (10)
rate equivalent if there is a high condensate/gas ratio.
2. Step 2: a and m Determination. A log-log plot of q/Gp (or where q∞ is the rate at infinite time, so it can be zero, positive, or
q/Np for oil) is constructed to determine these values, as shown in negative. In either case, the abandonment rate should be set on the
Fig. 9b. Data are examined to determine which section should be basis of economics.

Step 1: Data Check and Correction

100 100 Step 2: a and m Determinations
Press. Corrected Rate 1
Measured Rate
Wellhead Pressure y = 1.4105x –1.1384
Rate, 103 m3/d

0.1 R 2 = 0.9744
WHP, MPa

q/G p , 1/day

Skin effect
10 10 0.01 a =1.4105–1 d
m =1.1384
Skin effect and 0.001
Surface choking

1 1 0.0001
1 10 100 1,000 10,000 1 10 100 1,000 10,000
Days Days
(a) (b)
m3/d

Step 3: q 1 Determination 100 Step 4: Validation and Forecasts

40
106
Gas Rate, 103 m3/d

Rate Data
or Cum.

30 Matched Rate
y = 35.18x – 0.5207 Cum Gas
Cal. Cum Gas
20 10
m3/d

10 = 33.7x103
q1y=35.18 m3/d
q =–0.52 103 m3/d
103

0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1
Rate

t(a,m) 0 365 730 1,095 1,460 1,825 2,190

Days
(c) (d)

Fig. 9—Four steps for using the proposed approach.

June 2011 SPE Reservoir Evaluation & Engineering 381

 a and m Rate at Day 1
1 150

Gas Rate, 103 m3d

0.1 y = 1.1142x –1.0895

q/Gp, 1/day
100
a =1.1142
0.01 m =1.0895 y = 140.35x – 0.4779
50
0.001 q 1=140 103 m3d
q∞=–0.48 103
0.0001 0
1 10 100 1,000 10,000 0 0.2 0.4 0.6 0.8 1
(a) Days (b) t(a,m)

Rate Comparison Cumulative Comparison

1,000 1,000
Data
Gas Rate, 103 m3d

Power-Law Exp

Cum Gas, 106 m3d

Hyperbolic 100
100 New Approach
10 Data
10 Power-Law Exp
1 Hyperbolic
New Approach

1 0.1
1 10 100 1,000 10,000 100,000 1 10 100 1,000 10,000 100,000
(c) Days (d) Days

Fig. 10—Result comparison with power-law-exponential method and Arps hyperbolic with b=1.9.

Once q1 (and q∞) have been determined, gas-rate forecasts can This analog is in the form of type curves (P10, P50, and P90),
be made by using Eq. 9 or 10. which were generated by the new technique.
⎛ q ⎞
4. Step 4: Reserves Estimate. Eq. 4 ⎜ G p = t m ⎟ is used for A set of 25 wells that includes eight horizontal wells and 17
⎝ a ⎠
cumulative calculation with q>qeco, whereas EUR is estimated by vertical wells from the Barnett shale was selected for this study.
Eq. 11: In this study, there is not much difference between the vertical and
horizontal wells except for their initial rate. For this reason, no
q m attempt was made to separate them into two different data sets.
EUR = eco teco, . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (11)
a The following steps are proposed to establish the type curves:
where qeco is the minimum economic rate, and teco is at q= qeco. 1. Establish a and m for each individual well. a and m were
determined for all 25 wells, as shown in Fig. 10b.
Analysis Comparisons. To compare the results, other meth- 2. Construct a probability plot to obtain the P10, P50, and P90
ods such as Arps’ hyperbolic with b greater than unity and the values for m (Fig. 11a).
recently developed power-law-exponential method (Ilk et al. 2009) 3. Establish an a–m relationship such (as shown in Fig. 3) if
were selected. Texas A&M University has developed a rate/time sufficient data are available. Otherwise, use the correlation from
spreadsheet on using the power-law-exponential approach (Ilk that figure.
et al. 2008, 2009). Data from Example 3 from this spreadsheet was 4. Estimate the best 3-month rate averages for both vertical
employed here for the purpose of comparison. The analysis results and horizontal wells. The best 3-month rate averages are 0.51×106
are shown in Fig. 10. The results from the power-law-exponential m3/month (18 MMscf/month) and 1.10×106 m3/month (39 MMscf/
approach and Arps’ hyperbolic with b of 1.9 are included in Fig. 10 month) for the vertical and horizontal wells, respectively. For com-
for comparison. parison, Frantz et al. (2005) obtain the best 3-month gas averages
The reserves that were calculated on the basis of time limit of of 0.54×106 m3/month (19 MMscf/month) and 1.25×106 m3/month
30 years and economical rate of 2.83×103 m3/d (100 Mscf/D) are (44 MMscf/month) that are based on a data sample of 2,842 verti-
summarized in Table 1. cal wells and 246 horizontal wells from the Barnett shale.
Table 1 shows that the new approach provides conservative 5. Use Eq. B-5 to plot q3ma against Gp for various values of a and
reserves estimate compared with the other two methods. m to construct Fig. 12 for new well EUR prediction. Fig. 12a is for
the EUR prediction based on the time limit of 30 years regardless
Analog Type Curves. A field example was used to illustrate how of flow rate, whereas Fig. 12b is based on 30 years of production
a developing play can be used as an analog for other similar plays. with a cutoff rate of 2.83×103 m3/d (100 Mscf/D).
6. Use Eq. B-3 to estimate the q1 values (P10, P50, and P90)
based on the q3ma for both vertical and horizontal wells. The results
for this example are shown in Table 2. The type curves, as shown
TABLE 1—RESERVES COMPARISON BETWEEN METHODS in Fig. 13, were constructed on the basis of these parameters.
Time limit 30 Years
Rate limit 2.83
3
10 m /d
3 Conventional Tight Gas Wells. All examples illustrated thus far
Time limit Rate limit
result in an m value greater than 1. However, m can be less than 1,
as shown in Fig. 3. The cases in which m is less than unity in Fig.
Gp, 10 m Gp, 10 m
6 3 6 3
Model 3 are from a shallow and tight gas formation. A typical rate-decline
Hyperbolic (b=1.9) 118 222 analysis is shown in Fig. 14.
Power-law 116 188
Notice the slightly concave-up shape of the rate plot, as shown
exponential in Fig. 14d. This concave-up shape, which is caused by a late time
pseudoradial flow is characteristic for m<1. For m>1, a log-log
New approach 106 141
plot of rate vs. time will display a concave-down feature (Fig. 4).

382 June 2011 SPE Reservoir Evaluation & Engineering

 Cumulative Normal Probability Chart Constants a and m for 25 Barnett wells
0.99 1

Cumulative Normal Probability

Slopes R a m
0.90 Linear (Slopes) 0.1 P90 1.88 1.210
P90 P50 1.41 1.140

q/Gp, 1/day
P10 1.00 1.075
0.50 0.01
P50

0.001
0.10
P10

0.01 0.0001
1 10 100 1,000 10,000
1 1.05 1.1 1.15 1.2 1.25 1.3 Days
m
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 19 20 21 22 23
25 24 26 27 28
(a) (b)

Fig. 11—A statistical approach to analyze production forecasts for a group of wells or a set of sample wells: (a) a probability
plot for 25 wells and (b) a and m for all 25 wells.

EUR RANGES FOR BARNETT (30 YEARS) EUR RANGES FOR BARNETT (2.83 103 m3/d)
2.0 2.0

BEST-3-MONTH AVE, 106 m3

BEST-3-MONTH AVE, 106 m3

P90 P50 P90 P50

P10 P10
1.5 1.5

1.0 1.0
Rate limit with 30 years
0.5 0.5
Time limit only EUR shifted because
0.0 of rate cutoff
0.0
0 20 40 60 80 100 0 20 40 60 80 100
EUR, E6 m 3 EUR, E6 m3
(a) (b)

Fig. 12—EUR prediction based on time- and rate-limit constraints. Time limit would only give an optimistic prediction for EUR
if the initial best rate is low.

This feature distinguishes flow characteristics of a conventional its fluid gradients. As in the case of a gas-saturated formation, the
(late-time pseudoradial flow with m<1) vs. unconventional (lack pressure gradient should be approximately 2 kPa/m. A hydrauli-
of late-time pseudoradial flow or m>1) reservoir. cally fractured well located in such a formation will be expected to
pass through several distinct flow regimes such as fracture linear,
bilinear, elliptical, and pseudoradial flows, and BDF, as discussed
Discussion by Clarkson and Beierle (2010). An unconventional reservoir may
Absence of Late-Time Pseudoradial Flow. Unsurprisingly, the not be working according to such a model because there is no
m value for all shale-gas wells investigated in this study is greater matrix transient flow here to give rise to radial or elliptical flow.
than unity, or a log-log plot of rate vs. time will display a concave- If the reservoir is shale, the pressure gradient will be overpres-
down feature (Fig. 4). This flow characteristic, which indicates an sured. This can be up to 63 kPa/m (3 psi/ft), as recorded in the
absence of the late-time pseudoradial flow from shale gas reser- Lower Bossier play (Newsham and Rushing 2002) and shown
voirs, can be explained as follows. in Fig. 16. Spencer (1987) shows some pressure gradients in
Fig. 15 was constructed to distinguish between a conventional overpressured and gas-bearing shale plays in the Rocky Mountain
(tight) and shale-gas formation with an assumption that Point A, region. The interpreted pressure profile for the MWX site wells of
the wellbore location, has the same pressure in both conventional the Piceance basin shows that the pressure gradient based on mea-
and unconventional formations. If Well A is located in a tight but sured pore pressures in the active zone is approximately 36 kPa/m
conventional formation, its pressure will be in equilibrium with (1.7 psi/ft). The author of this study has also experienced pressure

TABLE 2—PARAMETERS FOR TYPE-CURVE GENERATION FOR BARNETT SHALE

Vertical well Horizontal well
q3ma, 10
6

3 0.51 1.10
m /month
P90 P50 P10 P90 P50 P10
a 1.88 1.42 1.00 1.88 1.42 1.00
m 1.21 1.14 1.07 1.21 1.14 1.07
q1, 10
3
12.19 19.09 32.58 26.40 41.37 70.59
3
m /d

June 2011 SPE Reservoir Evaluation & Engineering 383

 BARNETT VERTICAL WELL TYPE CURVES BARNETT HORIZONTAL WELL TYPE CURVES
16 35
14 P90 P90

Gas Rate, 103 m3/d

Gas Rate, 103 m3/d
30
P50 P50
12 P10 P10
25
10
20
8
15
6
4 10
2 5
0 0
0 12 24 36 48 60 72 84 96 108 120 0 12 24 36 48 60 72 84 96 108 120
(a) Months (b) Months

Fig. 13—Type curves for Barnett shale based on a set of 25 wells, for both vertical and horizontal wells.

Rate and Pressure Histories a and m Values

4 4 1
Gas Rate
Gas Rate, 103 m3/d

Pressure, MPa
3 Pressure 3 0.1

q/Gp
2 2 0.01

0.001 y = 0.6226x –0.9624

1 1 R 2 = 0.9865
0.0001
0 0 1 10 100 1,000 10,000
0 365 730 1095 1460 Days
(a) Days (b)
Rate or Cum, 103 or 106 m3/d

q1 Determination Results
2 10
Gas Rate, 103 m3/d

y = 5.3375x – 0.0424
1.5
y = 5.1312x 1 Actual Rate
1 Cal. Rate 1/4 slope
Actual Cum.
0.1 Cal. Cum
0.5

0 0.01
0 0.1 0.2 0.3 0.4 0.5 1 10 100 1,000 10,000
t(a,m) Days
(c) (d)

Fig. 14—A rate-decline example from a shallow and tight gas formation.

distributions across producing gas shales exceeding the lithostatic from this shale formation, these tiny reservoirs need to be con-
gradient, as in the Montney and other studied plays. nected by a fracture network, without which they will remain
With this high pressure gradient in shale gas, the formation is isolated and unproductive.
thought to be in a dynamical equilibrium state in which hydro- Because of such isolated characteristics of microreservoirs in
carbons have been generated and leaked off the formation over shale-gas formations, capillary pressure or relative permeability
geologic time. Under a normal production time scale, shale-gas curves should be applicable in each individual tiny reservoir, but
formations can be treated as impermeable barriers that contain they may not be applicable applied across the entire shale forma-
millions and millions of tiny isolated reservoirs (such as sandy tion. This can explain why the late-time radial- or elliptical-flow
laminates, microcracks, or isolated sand bodies); to produce gas regimes are absent from all shale-gas wells studied.

Numerical-Model Comparison. To use the current conventional

Pressure modeling for fractured shale, some assumptions or modifications
Top reservoir must be made. This section proposes and briefly discusses a frac-
ture model consisting of a macrofracture located inside an SRV.
A To overcome the nonequilibrium issue, permeability within the
Depth Unconventional SRV is assumed to have been anisotropically enhanced by stimula-
Shale Gas tion so that the pressure within it is in a state of equilibrium and
Conventional 20–60 kPa/m can be matched to the post-stimulation transient data. To prolong
Tight Gas the linear-flow regime, the drainage area or SRV should be large
≈ 2 kPa/m enough to imitate the gradual increase in the SRV as fractures
increase the connectivity.
Bottom reservoir
A commercial reservoir simulator is used here for numerical-
modeling work so that various fracture models in both conventional
Fig. 15—Pressure-gradient distribution in tight and shale gas. and unconventional reservoirs can be compared. Adsorbed gas

384 June 2011 SPE Reservoir Evaluation & Engineering

 Mcadmsa7
ft ft M EXPLANATION
Sheliya1 .D45 MEASURED PORE PRESSURE
1,000 MWX-1 MUD-WEIGHT PRESSURE
MWX-3 MUD-WEIGHT PRESSURE
Mcadmsa7 .D45 2 INTERPRETED PRESSURE GRADIENT
2,000 FRACTURE GRADIENT IN SHALE AND MUDSTONE
FRACTURE GRADIENT IN SANDSTONE
FRACTURE GRADIENT IN COAL

3,000 3

DEPTH (X 1,000)
4,000 0.7% RO
1.0
≈ 63 kPa/m pressure
5,000 4 0.8% RO ≈ 36 kPa/m pressure
distribution (acoustic
6,000 distribution (measured

0.8
0 .6
log response)

PS
5 1.5

PS
7,000 pore pressure)

I/F
I/F

T
165°F (74°C)

T
8,000
6 INACTIVE ZONE
9,000
2.0
10,000 200°F (93°C)
4

0.4
11,000 ACTIVE ZONE

33
12,000

PSI/
8 2.5

FT
13,000
10 20 30 40 50 60 kPa
9
2 3 4 5 6 7 8 9 PSI
0 2,000 4,000 6,000 8,000 10,000 12,000 0 120 200 50
Pressure, psig 1,000/div GR Acoustic PRESSURE (X 1,000)
(a) (b)

Fig. 16—Two examples of high pressure-gradient distribution across the producing shale gas: (a) the Lower Bossier play (New-
sham and Rushing 2002) and (b) the Piceance basin (Spencer 1987).

is not considered in the study but can be included in the model.
Fig. 17 is a summary of the results in the form of normalized
decline-trend plots. The q/Gp ratio of the proposed fracture model
is shown as the blue line. If the stimulated region is much smaller,
the red or BDF curve proposed by Bello and Wattenbarger (2008)
is obtained.
To compare with the proposed analytical method, the thin
white line with the matched a and m factors of 0.7785 and 1.0481,
respectively, is produced. This line lies exactly on top of the blue
line from the numerical simulation result.
The yellow curve shows one of the results of a normal reser-
voir model where the virgin area surrounding the SRV is also in
pressure communication. This yellow curve, which includes a late-
time linear flow from the virgin region into the SRV, can also be
obtained using the approach from Anderson et al. (2010). Another
result is shown as a green line where both stimulated and virgin
regions are the same or under a homogeneous state. Pseudoradial
flow will be achievable to make m a value of less than unity. Fig.
14 illustrates a field example of a gas well producing from a tight
reservoir.
Method Limitations. Downtime Effect. Because of some opera-
tional conditions, a well may have to be shut in for some periods
of time during its production life. If the downtime of each shut-in
is significant, rate initialization is required as in traditional decline-
curve-analysis methods. To use q/Gp, both rate and cumulative
production should be initialized with pressure correction. Fig. 7
shows the example plots of rate initializations for Wells 2 and 3.
EUR Constraints. In most cases, without the presence of BDF
in shale reservoirs, drainage area cannot be established for EUR
prediction. In this method, the drainage area will be defined only
when the fracture-expansion process is brought to an end. Thus,
EUR is not based on the traditional concept of drainage area
(BDF) but on the constraints of the latest trends of (q1, q∞) with
both time and economic rate limits. Because of this, results are
valid only where analog wells are similar in both interwell and
fracture spacings.
In case of spacing differences, a numerical reservoir modeling
(see the Numerical Model Comparison subsection) should be used
for any EUR adjustment.
Latest Trends. During a well production, many factors can
contribute to the deviation from a fracture flow to accelerate the
decline rate such as water production or liquid dropout/holdup in
the fracture system and wellbore. The values of q1 and q∞ (Eq. 10
and Fig. 9) are designed to account for this acceleration in rate
decline in late time. An increase in water production will also cause
both a and m to increase in value (Fig. 8).
Time Limit vs. Rate Limit. Fig. 12 shows a typical EUR predic-
tion for an infill in the Barnett shale for both vertical and horizontal
wells. The prediction based on a time limit alone would give an

Normalized Decline Trend Plots Normalized Decline Trend Plots

1 1.E-03
Conventional
–1.0481 Proposed Model Pseudoradial m<1
y = 0.7787x -1.0481
0.1 Expanding
y = 0.7787xSRV
2
R = 0.9997 and RAnisotropic
2
= 0.9997 k
q/Gp, day –1
q/Gp, day –1

0.01
Anderson et al. 2010
1.E-04
Contribution from
0.001 Undisturbed Shale

0.0001 Bello et al. 2008

Bounded SRV or BDF
0.00001 1.E-05
1 10 100 1,000 10,000 100 1,000 10,000
Days Days
(a) (b)

Fig. 17—Normalized decline trend plots for various simulated fracture models in shale or tight reservoirs: (a) for all production
data and (b) at late time.

June 2011 SPE Reservoir Evaluation & Engineering 385

optimistic value. Note that the EUR shifts to the left when the rate
limit was imposed on the time-limit plot, as shown in Fig. 12b.
The lower the initial rate is, the higher and more optimistic the
value of the predicted EUR will be. This observation is especially
applicable to vertical wells because of their lower initial rates. The
recommendation is to use both time and economic rate limits for
predicting EUR.
Conclusions
1. A new approach that is easy and simple to use for predicting the
future rate and EUR has been developed for fracture-dominated
wells in unconventional reservoirs.
2. The new approach is also able to provide a statistical method
to analyze production forecasts of resource plays and establish
a range of results for these forecasts, including probability dis-
tributions of reserves in the form of P90 to P10.
3. A log-log plot of rate over cumulative production vs. time is
observed to fit a straight line in all unconventional-reservoir
cases studied so far. The slope and intercept are related to res-
ervoir-rock characteristics (tight or shale), fracture-stimulation
practice, operational conditions, and (possibly) liquids content.
4. EUR for the new approach is not based on the traditional concept
of drainage area (BDF) but on the constraints of the latest trends
(q1, q∞ ) with both time and economic rate limits.
5. Pressure initialization used in numerical modeling based on
fluid gradients may be incorrect. Results from such numerical
modeling may not be representative of the shale-gas flow char-
acteristics. Some modifications of flow characteristics in both
stimulated and undisturbed regions are needed.
Nomenclature
a = intercept constant defined by Eq. 4, d−1
b = Arps’ decline exponent, dimensionless
Di = Arps’ initial decline rate, d−1
G = original gas in place, m3
Gp = cumulative gas production, m3
kf = fracture permeability, 10−3 µm2
km = matrix permeability, 10−3 µm2
m = slope defined by Eq. 4
n = fracture time exponent, dimensionless
Np = cumulative oil production, m3
q = rate, m3/d
q1 = rate at Day 1, m3/d
q1D = rate at tD=1, m3/d
q3ma = best 3-month average rate, m3/month
qeco = economic cutoff rate, m3/d
qmax = maximum rate, m3/d
q∞ = rate at infinite time, m3/d
t = time, days
tD = dimensionless time defined by tD=Dit
tDxf = dimensionless time based on fracture half-length
tm = dimensionless time based on Eq. 7
tmax = time at maximum flow rate, days
tx = prediction time, days
t(a,m) = time function based on Eq. 9
xf = fracture half-length, m
ε = decline rate defined by ε (t ) = q , d−1
Gp
Acknowledgments
The author would like to thank the management of ConocoPhillips
for their support and permission to publish this study. He is also
grateful to Dev Mukherjee, Chris Clarkson, Sheila Reader, and
Kevin Raterman for reviewing the paper; to coworkers in both
the Houston and Calgary offices for peer review of the proposed
method; and to Matt Maguire and his group for supplying the
Barnett production data.
386
References
Agarwal, R.G., Carter, R.D., and Pollock, C.B. 1979. Evaluation and
Performance Prediction of Low-Permeability Gas Wells Stimulated by
Massive Hydraulic Fracturing. J Pet Technol 31 (3): 362–372; Trans.,
AIME, 267. SPE-6838-PA. doi: 10.2118-6838-PA.
Anderson, D.M., Nobakht, M., Moghadam, S., and Mattar, L. 2010. Analysis
of Production Data from Fractured Shale Gas Wells. Paper SPE 131787
presented at the SPE Unconventional Gas Conference, Pittsburgh, Penn-
sylvania, USA, 23–25 February. doi: 10.2118/131787-MS.
Arévalo-Villagrán, J.A., Wattenbarger, R.A., Samaniego-Verduzco, F., and
Pham, T.T. 2001. Production Analysis of Long-Term Linear Flow in
Tight Gas Reservoirs: Case Histories. Paper SPE 71516 presented at
the SPE Annual Technical Conference and Exhibition, New Orleans,
30 September–3 October. doi: 10.2118/71516-MS.
Arévalo-Villagrán, J.A., Wattenbarger, R.A., and Samaniego-Verduzco, F.
2006. Some History Cases of Long-Term Linear Flow in Tight Gas
Wells. J Can Pet Technol 45 (3): 31–37. JCPT Paper No. 06-03-01.
doi: 10.2118/06-03-01.
Arps, J.J. 1945. Analysis of Decline Curves. SPE-945228-G. Trans., AIME,
160: 228–247.
Bagnall, W.D. and Ryan, W.M. 1976. The geology, reserves, and production
characteristics of the Devonian Shale in southwestern West Virginia. In
Devonian Shale Production and Potential: Proceedings of the Seventh
Appalachian Petroleum Geology Symposium held at Morganstown,
West Virginia, March 1–4, ed. R.C. Shumaker and W.K. Overbey Jr.,
MERC/SP-76/2, 41–53. Alexandria, Virginia, USA: National Technical
Information Service.
Bello, R.O. and Wattenbarger, R.A. 2008. Rate Transient Analysis in Natu-
rally Fractured Shale Gas Reservoirs. Paper SPE 114591 presented at
the CIPC/SPE Gas Technology Symposium, Calgary, 16–19 June. doi:
10.2118/114591-MS.
Bello, R.O. and Wattenbarger, R.A. 2010. Multi-stage Hydraulically Frac-
tured Horizontal Shale Gas Well Rate Transient Analysis. Paper SPE
126754 presented at the North Africa Technical Conference and Exhibi-
tion, Cairo, 14–17 February. doi: 10.2118/126754-MS.
Cinco-Ley, H. and Samaniego-V., F. 1981a. Transient Pressure Analysis
for Fractured Wells. J Pet Technol 33 (9): 1749–1766. SPE-7490-PA.
doi: 10.2118/7490-PA.
Cinco-Ley, H. and Samaniego-V., F. 1981b. Transient Pressure Analysis:
Finite Conductivity Fracture Versus Damaged Fracture Case. Paper SPE
10179 presented at the SPE Annual Technical Conference and Exhibi-
tion, San Antonio, Texas, USA, 4–7 October. doi: 10.2118/10179-MS.
Cinco-Ley, H., Samaniego-V., F., and Dominguez, A.N. 1978. Transient
Pressure Behavior for a Well With a Finite-Conductivity Vertical Frac-
ture. SPE J. 18 (4): 253–264. SPE-6014-PA. doi: 10.2118/6014-PA.
Clarkson, C.R. and Beierle, J.J. 2010. Integration of Microseismic and
Other Post-Fracture Surveillance with Production Analysis: A Tight
Gas Study. Paper SPE 131786 presented at the SPE Unconventional
Gas Conference, Pittsburgh, Pennsylvania, USA, 23–25 February. doi:
10.2118/131786-MS.
Fetkovich, M.J. 1980. Decline Curve Analysis Using Type Curves. J Pet
Technol 32 (6): 1065–1077. SPE 4629-PA. doi: 10.2118/4629-PA.
Frantz, J.H. Jr., Williamson, J.R., Sawyer, W.K., Johnston, D., Waters, G.,
Moore, L.P., Macdonald, R.J., Pearcy, M., Ganpule, S.V., and March,
K.S. 2005. Evaluating Barnett Shale Production Performance Using an
Integrated Approach. Paper SPE 96917 presented at the SPE Annual
Technical Conference and Exhibition, Dallas, 9–12 October. doi:
10.2118/96917-MS.
Gringarten, A.C., Ramey, H.J. Jr., and Raghavan, R. 1975. Applied Pressure
Analysis for Fractured Wells. J Pet Technol 27 (7): 887–892. SPE-5496-
PA. doi: 10.2118/5496-PA.
Ilk, D., Rushing, J.A., and Blasingame, T.A. 2009. Decline Curve Analysis
for HP/HT Gas Wells: Theory and Applications. Paper SPE 125031
presented at the SPE Annual Technical Conference and Exhibition,
New Orleans, 4–7 October. doi: 10.2118/125031-MS.
Ilk, D., Rushing, J.A., Perego, A.D., and Blasingame, T.A. 2008. Expo-
nential vs. Hyperbolic Decline in Tight Gas Sands—Understand-
ing the Origin and Implications for Reserve Estimates Using Arps’
Decline Curves. Paper SPE 116731 presented at the SPE Annual
Technical Conference and Exhibition, Denver, 21–24 September. doi:
10.2118/116731-MS.
June 2011 SPE Reservoir Evaluation & Engineering
Lee, W.J. and Sidle, R.E. 2010. Gas Reserves Estimation in Resource If
Plays. Paper SPE 130102 presented at the SPE Unconventional Gas
q
Conference, Pittsburgh, Pennsylvania, USA, 23–25 February. doi: = at − m . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (A-8a)
10.2118/130102-MS. Gp
Mattar, L., Gault, B., Morad, K., Clarkson, C.R., Freeman, C.M., Ilk, or
D., and Blasingame, T.M. 2008. Production Analysis and Forecasting
of Shale Gas Reservoirs: Case History-Based Approach. Paper SPE ε (t ) = at − m . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (A-8b)
119897 presented at the SPE Shale Gas Production Conference, Fort
Worth, Texas, USA, 16–18 November. doi: 10.2118/119897-MS. Eqs. A-6 and A-7 can be written as:
Newsham, K.E. and Rushing, J.A. 2002. Laboratory and Field Obser-
vations of an Apparent Sub-Capillary-Equilibrium Water Saturation q (
a t1− m −1
)
Distribution in a Tight Gas Sand Reservoir. Presented at the SPE Gas = t − m e 1− m . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (A-9a)
Technology Symposium, Calgary, 30 April–2 May. Search and Dis-
q1
covery Article #40400.
Soltanzadeh, H. and Hawkes, D.C. 2008. Semi-analytical models for stress
and
change and fault reactivation induced by reservoir production and injec-
q1 1−am ( t1− m −1)
tion. J. Pet. Sci. Eng. 60 (2): 71–85. doi: 10.1016/j.petrol.2007.05.006. Gp = e . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (A-9b)
a
Spencer, C.W. 1987. Hydrocarbon Generation as a Mechanism for Over-
pressuring in Rocky Mountain Region. AAPG Bulletin 71 (4): 368–
388. Appendix B—Specific Equations for t and q
Valkó, P.P. 2009. Assigning Value to Stimulation in the Barnett Shale: A tmax and qmax. Fig. 4 shows there is a maximum rate flow qmax for
Simultaneous Analysis of 7000 Plus Production Histories and Well each line where m is greater than unity. Time tmax, which gives qmax,
Completion Records. Paper SPE 119639 presented at the SPE Hydrau- can be estimated by
lic Fracturing Technology Conference, The Woodlands, Texas, USA, 1
19–21 January. doi: 10.2118/119639-MS. 1− m
⎛ m⎞
Warpinski, N.R. and Branagan, P.T. 1989. Altered-Stress Fracturing. J Pet tmax = ⎜ ⎟ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (B-1)
Technol 41 (9): 990–997. SPE-17533-PA. doi: 10.2118/17533-PA. ⎝ a⎠
Weng, X. and Siebrits, E. 2007. Effect of Production-Induced Stress Field
on Refracture Propagation and Pressure Response. Paper SPE 106043
and

( )
presented at the SPE Hydraulic Fracturing Technology Conference, m
m−a
1− m
College Station, Texas, USA, 29–31 January. doi: 10.2118/106043- qmax = q1 a e 1− m . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (B-2)
MS. m

Appendix A—General Equations for q and Gp Eq. B-1 was obtained by allowing the rate derivative of Eq. 5,
dq/dt, to approach zero.
⎡ ⎤
⎢ q = ε (t ) ⎥⎥ .
General equations are developed on the basis of ⎢
⎢⎣ G p ⎥⎦
Best 3-Month Average, q3ma. Since tmax is less than 45 days (Fig.
q 4), the best 3-month average (q3ma) will be equal to the average of
= G p . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (A-1)
ε (t ) the first 3 months of production and can be calculated by
Taking a derivative with respect to time, we have
q1 1−am (901− m −1)
q3 ma = e . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (B-3)
d [ q / ε (t ) ] = d G . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (A-2) 3a
dt dt p
Eq. B-3 is obtained on the basis of Eq. 6 by calculating the cumula-
Using q’= dq/dt and ε’(t) = dε (t ) , Eq. A-2 becomes tive gas of the first 3 months of production. Gas rate and cumulative
dt gas based on this q3ma would be
q '/ ε (t ) − qε '(t ) / ε 2 (t ) = q , . . . . . . . . . . . . . . . . . . . . . . . . (A-3)
a
(t 1− m
− 901− m ) . . . . . . . . . . . . . . . . . . . . . . . . (B-4)
or q = 3q3 ma at − m e 1− m

and
dq / q = dε (t ) / ε (t ) + ε (t )dt . . . . . . . . . . . . . . . . . . . . . . . . (A-4)
a
(t 1− m
− 901− m ). . . . . . . . . . . . . . . . . . . . . . . . . . . (B-5)
Integrating both sides from t=1 to t, we have: G p = 3q3 mae 1− m

q ⎡ ε (t ) ⎤ t Eq. B-5 is used to construct the solid lines in Fig. 3.

ln = ln ⎢ ⎥ + ∫ ε (t )dt , . . . . . . . . . . . . . . . . . . . . . . . . (A-5)
q1 ⎣ ε (1) ⎦ 1
Anh N. Duong is a principal reservoir engineer and holds the
where q1 is the theoretical rate at t=1 day, title Reservoir Modeling Director at ConocoPhillips Canada,
where he has worked since 2001. E-mail: anhnduong@yahoo.
t com or anh.n.duong@conocophillips.com. He heads a subsur-
ε (t ) 1∫ ε (t )dt , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (A-6) face team that provides technical expertise in reservoir engi-
q = q1 . e neering and modeling to all ConocoPhillips Canada assets.
ε (1) These include oil sands, shale gas, and tight oil and gas res-
ervoirs. Before joining ConocoPhillips, Duong worked 21 years
and for Gulf Canada Resources as a reservoir engineer. He holds a
t BSc degree (1978) in petroleum engineering from Petroleum
q ∫ ε (t )dt. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (A-7)
University of Technology (Abadan Faculty). In late 1980s, Duong
Gp = 1 e1 pursued but did not complete an ME degree in reservoir engi-
ε (1) neering at the University of Calgary.

June 2011 SPE Reservoir Evaluation & Engineering 387

 REE137748 DOI: 10.2118/137748-PA Date: 16-April-14 Stage: Page: 1 Total Pages: 2

Discussion of Rate-Decline Analysis for

Fracture-Dominated Shale Reservoirs
Martin Wolff, SPE, Occidental Petroleum Corporation

Duong (2011) presents a new forecasting approach using log-log Linear-Flow Plot
rate/cumulative production (q/Gp) vs. time (t) plots for wells in
unconventional reservoirs which often exhibit long-term linear y = 1.2325×104 – 8.0794×104

Cumulative Gas, Mscf

flow. He also relates parameters of his new method to Arps curves R 2 = 9.9942×10–1
and discusses both linear and bilinear flow. Kupchenko et al.

Gas Rate, Mscf/D

EUR = 1.2 Bscf
(2008) also discuss linear and bilinear flow and how they relate to
Arps parameters but note that rate and/or rate-derivative plots may
require sophisticated filtering to become interpretable because of
the noisiness of field data. The following discussion presents an-
other approach linked to Arps parameters which complements the
methods proposed by Duong and Kupchenko et al.
A classic decline method, the linear-flow plot of cumulative
production vs. the square root of time (Nott and Hara 1991), sug-
gests some advantages of using cumulative production with noisy
field data. These include the inherent filtering involved with cu-
mulative rather than rate or derivative plots, as well as the ease of 0.0 5.0 10.0 15.0 20.0
fitting straight lines to Cartesian plots. Although such plots are Square root, Days
not as diagnostic as sensitive derivative plots, they may have
some advantages in allowing relatively robust and repeatable Fig. 1—Field example of b52 linear-flow plot (time0.5). EUR5esti-
interpretations once suitable time exponents are selected [possibly mated ultimate recovery.
from plots of q/Gp vs. 1/t as suggested by Eq. 3 in Duong (2011)].
The following derivation shows how to construct generalized Car-
tesian cumulative production vs. time plots for arbitrary b factors
(=1). When a range of hyperbolic decline exponents is used to
Eqs. 1 and 2 define cumulative production Gp and instantane- define a range of forecasts for a particular play or field, a set of
ous rate q at time t for production declines following the Arps Cartesian plots with varying n can be used to fit the data more eas-
hyperbolic form with initial rate qi and initial decline, Di. ily, rather than fitting both Di and b as required by Arps hyper-
bolic forms. Figs. 1 and 2 show sample field data plotted with fits
qi b for b¼2 and b¼1.1 (which could represent a more-conservative
Gp ¼ ðq1b  q1b Þ: . . . . . . . . . . . . . . . . . ð1Þ projection, possibly for proven reserves). In this case, the b¼2
ð1  bÞDi i
plot shows a better fit from earlier time, suggesting some form of
linear flow; however, extended production histories are needed to
q ¼ qi ð1 þ b Di tÞ1=b : . . . . . . . . . . . . . . . . . . . . . . . .ð2Þ confirm how long this transient-flow regime will actually last in
different situations.
Substituting for q in Eq. 1 and rearranging, we can rewrite Eq. 2 As both Duong (2011) and Kupchenko et al. (2008) note, the
as Arps hyperbolic form, with all its flaws but also with its well-
qi understood parameters, remains the most common forecasting
Gp ¼  qi ð1 þ b Di tÞðb1Þ=b : . . . . . . . . . . . . ð3Þ form used by the industry. Therefore, these Cartesian square-root-
ð1  bÞDi of-time plots may have application as simple, repeatable methods
When ðb Di tÞ  1, the equation simplifies to the linearized for forecasting with relationships to Arps decline factors.
equation

Gp ¼ A tðb1Þ=b þ B; . . . . . . . . . . . . . . . . . . . . . . . . . ð4Þ b = 1.1 Plot

where y = 6.5763×105x – 9.7189×105

2
R = 9.9751×10–1
qi
Cumulative Gas, Mscf

ðb Di Þðb1Þ=b : . . . . . . . . . . . . . . . . . . . . ð5Þ
Gas Rate, Mscf/D

A¼
ð1  bÞDi EUR = 0.6 Bscf
and
qi
B¼ : . . . . . . . . . . . . . . . . . . . . . . . . . . . . ð6Þ
ð1  bÞDi
Therefore, we can construct a Cartesian plot of Gp against tn
where n ¼ (b–1)/b. This allows easily fitting a straight line which
corresponds to Slope A. For the special case of b ¼ 2, n ¼ 0.5
which is the classic linear-flow plot with Gp plotted against the
square root of time. For bilinear flow with b ¼ 4, n ¼ 0.75. 1.0 1.1 1.2 1.3 1.4 1.5 1.6 1.7
t 0.091, days

Copyright V
C 2014 Society of Petroleum Engineers Fig. 2—Field example of b51.1 plot (time0.091).

2014 SPE Reservoir Evaluation & Engineering 1

ID: komathi.k Time: 09:44 I Path: S:/3B2/REE#/Vol00000/140019/APPFile/SA-REE#140019

 REE137748 DOI: 10.2118/137748-PA Date: 16-April-14 Stage: Page: 2 Total Pages: 2

References Technology Symposium, Calgary, Alberta, Canada, 16–19 June. SPE-

Duong, A.N. 2011. Rate-Decline Analysis for Fracture-Dominated Shale
Reservoirs. SPE Reservoir Evaluation & Engineering 14 (3): 377–387.
SPE-137748-PA. http://dx.doi.org/10.2118/137748-PA.
Kupchenko, C.L., Gault, B.W., and Mattar, L. 2008. Tight Gas Production
Performance Using Decline Curves. Presented at the CIPC/SPE Gas
114991-MS. http://dx.doi.org/10.2118/114991-MS.
Nott, D.C. and Hara, S.K. 1991. Fracture Half-Length and Linear Flow in
the South Belridge Diatomite. Presented at the SPE Western Regional
Meeting, Long Beach, California, 20–22 March. SPE-21778-MS.
http://dx.doi.org/10.2118/21778-MS.

2 2014 SPE Reservoir Evaluation & Engineering

ID: komathi.k Time: 09:44 I Path: S:/3B2/REE#/Vol00000/140019/APPFile/SA-REE#140019