CSUG/SPE 137748

An Unconventional Rate Decline Approach for Tight and Fracture-Dominated

Gas Wells
A.N. Duong, ConocoPhillips

Copyright 2010, Society of Petroleum Engineers

This paper was prepared for presentation at the Canadian Unconventional Resources & International Petroleum Conference held in Calgary, Alberta, Canada, 19–21 October 2010.

This paper was selected for presentation by a CSUG/SPE program committee following review of information contained in an abstract submitted by the author(s). Contents of the paper have not been
reviewed by the Society of Petroleum Engineers and are subject to correction by the author(s). The material does not necessarily reflect any position of the Society of Petroleum Engineers, its officers, or
members. Electronic reproduction, distribution, or storage of any part of this paper without the written consent of the Society of Petroleum Engineers is prohibited. Permission to reproduce in print is
restricted to an abstract of not more than 300 words; illustrations may not be copied. The abstract must contain conspicuous acknowledgment of SPE copyright.

Abstract
Traditional decline methods such as Arps’ rate-time relations and their variations do not work for super-tight or shale gas wells
where 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 (BDF), neither matrix permeability nor drainage area can be established. This indicates that matrix contribution is negligible
compared to fracture contribution and the expected ultimate recovery (EUR) can not be based on a traditional concept of drainage
area.
An alternative approach is proposed to estimate EUR from wells where fracture flow is dominant and matrix contribution is
negligible. To support these fracture flows, the connected fracture density of the fractured area must increase over time. This
increase is possible due to 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 higher than unity is normally observed due to actual field operations,
data approximation and flow regime changes. A rate-time or cumulative production-time relationship can be established based on
the intercept and slope values of this log-log plot and initial gas rate.
Field examples from several super-tight and shale gas plays for both dry and high liquids gas production were used to test the new
model. All display the predicted straight line trend, with its slope and intercept related to the type of fractured flow regimes. In
other words, a certain fractured flow regime or a combination of flow types that dominate a given area or play due to 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 based on this range if the best-three-month average or the initial production rate of the well is
already known or estimated.
The results show that this alternative 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 probability distributions of reserves in
terms of P90 (lower side) to P10 (higher side).
Introduction
Unconventional gas reservoirs, especially wet shale gas, are currently being aggressively pursued for new development in both US
and Canada. Forecasting production and estimating reserves accurately for these resource plays has become more urgent and
important than ever before.
2 CSUG/SPE 137748

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, if adequate
production data is available. Fetkovich’s type curves (1980) are not appropriate since the matched b value is usually greater than
one. 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 since it can have a “best fit” for the long transient linear flow regime
observed in shale gas wells b values greater than 1. Lee and Sidle (2010) showed that values of b equal to or greater than 1 can
cause the reserves to have physically unreasonable properties. To avoid this drawback, “stretched exponential” models have been
recently proposed (Valkό 2009; Ilk et al. 2008 & 2009).
The current simulation models that are available in the market are still employing conventional technologies to 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 indicates 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 over 25 times its saturated gas gradient or 3 times the litho-static gradient. Spencer (1987) presents some pressure gradients in
overpressured and gas bearing shale plays in the Rocky Mountain region. The interpreted pressure profile across the active zone
for the MWX site wells in the Piceance basin shows pressure gradient around 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). They 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 based on a long-term linear flow in a large number of wells in tight
and shale gas reservoirs. Based on this model, a new methodology has been developed for production analysis and forecasting of
unconventional reservoirs. This method also employs probability distributions of reserves in forecasting resource plays, to
represent any uncertainty in reserve estimation.
Methodology Development
Long-Term Linear Flow
Fig. 1 shows the production history of a vertical well producing from the Barnett shale. Production data shows a half-slope line
indicating linear flow lasting for more than 5 years of production.
PRODUCTION HISTORY FOR A SHALE GAS WELL
100 100
Press. Corrected Rate
Measured Rate
Wellhead Pressure
Rate, E3 m3/d

Skin effect
WHP, MPa

10 10

Skin effect &

Surface choking

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

Fig. 1: A log-log plot of field production and pressure data for a shale gas
well. Production data shows a haft-slope line indicating linear flow
lasting 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 causes this phenomenon. Several authors have showed that this
linear flow exists in any type of fractures whether they are infinite (Agarwal et al., 1981) or finite conductivity (Cinco-Ley et al.,
1976, 1978 & 1980), single or multi-stage (Bello and Wattenbarger, 2010), hydraulic (Arévalo-Villagrán et al., 2001) or natural
fractures (Villagrán et al., 2006). The difference among these models is the length of this transient linear flow regime.
CSUG/SPE 137748 3

To prolong the linear flow regime in any of those models, all one needs to do is enlarge the term, xf2/km where xf is the effective
haft fracture 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 fractured area must be increasing over time to support this fracture flow over
the life of a producing well. This increase is achievable due to local stress changes under fracture depletion (Warpinski and
Branagan 1989). 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. A fracture expansion theory is proposed and the subject of another paper.
If a fracture flow regime (either linear or bilinear) is prolonged over the life of a well, the gas flow rate, q, will be:

q = q1t − n …………………………………..…………………………..………………….....…................…...….…… (1)

where n is ½ for linear flow, n=¼ for bilinear flow and q1 for the flow rate at day 1.
The gas cumulative will be:
t
G p = qdt ∫
0

t1− n
G p = q1 ……………..………………………..…………………………..……………....………................…… (2)
(1 − n)
This gives
q (1 − n)
= ……………………………………………………..…………………………………….....…………… (3)
Gp t

Field Applications
Field data from several shale gas plays were used to test Eq. 3. These field data were operated under the actual field conditions that
may violate some of the ideal assumptions used to derive Eq. 3. Some of the results are shown in Fig. 2 where log-log plots for
q/Gp vs. time in days were constructed.
Constants a & n Constants a & m
Constants a & m 1 1
1

-1.1381
0.1 0.1 y = 1.2776x 0.1 -1.1487
y = 1.4105x
-1.1384
2
y = 1.3665x
R = 0.9808
q/Gp, 1/Day
q/Gp, 1/Day
q/Gp, 1/Day

2
2
R = 0.9744 R = 0.9836
0.01 0.01 0.01

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
Days Days Days

Fig. 2: A log-log plot of q/Gp vs. t for various shale gas wells with over 5 years of production. The data all forms 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.

These plots all give a log-log straight-line with a negative slope, -m and an intercept of a. That is:
q
= at −m ……………………………………………………..………………….................……………… (4)
Gp
Equations for q and Gp can be derived as shown in Appendix A:

(t a 1− m −1)
q
= t −m e1− m ……………………………..…………………………..…………..…......….………… (5)
q1
and
4 CSUG/SPE 137748

q1 1−am (t1−m −1)

Gp = a e ……………………………..…………………………..……………………...…....….…… (6)

It will be shown later that “a” can be eliminated from the above equations if they are rewritten in forms of dimensionless rate and
time. That indicates “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 versus time in days can be constructed as shown Fig. 4.

a vs. m q/q1 vs. time

3 10
qmax
Tight/Shallow Gas
2.5
Dry Shale Gas

2 Wet Shale Gas 1

a

1.5

q/q1
m~1.00
m=1.05
1 a = 0.7364 m4.7954
0.1 m=1.10
R2 = 0.9417 m=1.15
m=1.20
0.5
m=1.25
m=1.3
0
0.01
0.9 0.95 1 1.05 1.1 1.15 1.2 1.25 1.3 1 10 100 1000 10000
m Days
Fig. 3: A correlation between “a” and “m” for various gas plays Fig. 4: Type curves for q/q1 versus time in days for m>1

Fig. 4 shows there is a maximum rate flow, qmax for each line where “m” is greater than 1. Time, tmax, which gives qmax can be
estimated by:

1
m 1−m
t max = ( ) …………………………………..…………………………..……………....…..............………… (7)
a
and,

m m−a
qmax = q1 ⎛⎜ a ⎞⎟ e 1−m ……………………………….…..…………………………..……………....…………… (8)
1− m
⎝m⎠
Eq. 7 was obtained by approaching the rate derivative of Eq. 6, dq/dt, to be zero. If tmax is less than 45 days, the best-three-month
average, q3ma, will be the first three month average and can be calculated by:

a 1 − m −1)
q 1 1 − m ( 90
q 3 ma = e ……………………..…….……………………..…………...….….……… (9)
3a
Eq. 9 is obtained based on Eq. 6 by calculating the cumulative gas of the first three months of production. Gas rate and cumulative
gas based on this q3ma would be:

a (t1− m −901− m )
q = 3q3ma at −m −
e m
1 ………………………..……….…………………………....….....………… (10)

and
CSUG/SPE 137748 5

a [t1− m −901− m ]
1−
Gp = 3q3ma e m …………………………….…………….………..……………....……….......…… (11)

For any given length of production, a plot of cumulative gas against the best-three-month average of all wells with a similar “a”
and “m”, would fall on the same straight line passing through the origin. The slope of this straight line is
a [t 1− m − (90)1− m ]
x
3e1− m in which t is the prediction time such as 24 months (730 days) or 60 months (1825 days) as examples.
x

This characteristic has been reported by Frantz et al. (2005) for 2.00

Barnett shale wells. To validate Eq. 11, a sample of 25 Barnett

Best-3-Month Ave, E6 m3
wells was used. Actual best-three-month averages for both 1.50 1 year 2 years
5 years
vertical and horizontal wells were plotted against their
cumulative productions of one, two and five years as shown in 1.00
Fig. 5. A very good match is obtained. See more details on the
section of Field Examples under “Analog Type Curves”.
0.50
Dots - 25 Barnett Samples
In summary, if either value of q1, qmax or q3ma is known, reserve Solid lines - Eq. 17 (P50)
evaluations can be established for any individual well, group of 0.00
0 5 10 15 20 25 30
wells, development area or play with a given set of “a” and Cum Gas, E6 m3
“m”. The q3ma is preferred since it gives a longer well Fig. 5: Best-three-month averages for a sample of 25 Barnett wells
production performance and hence is more reliable. were plot against their cumulative productions of one, two and five
years.
Fig. 4 gives a set of type curves for a typical field data plot
since 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 ...............................................................................................................................................................……… (12a)
tm =
t max

and,

m 1− m −1)
q −m 1− m ( tm
= tm e ….……………………………..…………………………..…………..…....…………… (12b)
qmax
tmax and qmax are defined by Eq. 7 and Eq. 8. The general tm=t/tmax
1 tD=Dit
type curve is shown in Fig. 6 for an “m” range of 1 to
q1D → q @ tD=1
1.96. Note that “a” has been eliminated from the type m=1.05
qmax → q @ tm=1
curves (see Eq. 12b).
q/qmax or q/q1D

b=0.5 m=1.15
b=0 b=1
0.1
For comparison, the Arps’ rate-time relations (Arps 1945)
were reconstructed in the same plot using dimensionless
time, tD=Dit, and rate, q1D, for their normal range of “b” m=1.30
from 0 to 1. Note that q1D is defined as the dimensionless
0.01
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
m=1.96 m=1.60
Fig. 6. For b>0.5, Arps’ curves give more optimistic
0.001
forecasts compared to the new approach.
1 10 100 1000 10000
tm or tD

Fig. 6: A comparison between Arps’ and new method’s type curves for
dimensionless rate and time.
6 CSUG/SPE 137748

Field Examples

In this section, several field examples from different unconventional gas plays were used to test the new approach. Examples were
selected to illustrate the decline behaviours in various completion and production scenarios such as the cases of gas retrograde,
newly developing or developed plays, and vertical or horizontal wells. Changes in flow regime, “a” value and “m” value will be
highlighted for each example in the following discussion. A comparison of results is also made with other methods.
Gas Retrograde Case
Wet shale gas plays have become more desirable lately due to higher liquid contents and attractive liquid price. Unfortunately,
very little is known about their flow characteristics and how the liquid dropout in 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 normalized based on the highest rate from
these wells as shown in Fig. 7.
The left graph on Fig. 7 shows that the flow regime varies from linear to bilinear flow. Generally, the wells with a bilinear flow
regime will have the lowest initial flow rate. This observation is drawn not only from this example but also from other retrograde
plays where more wells and longer production time are available. The lower deliverability rates may be caused by liquid holdup in
the fractures, or fracture choking.
Values of “a” and “m” obtained from a retrograde shale gas have covered a large range of values, these vary from 0.94 to 1.91 and
from 1.02 to 1.28 respectively, as shown on the right graph in Fig. 7 for this example.

1
1
N ormalized R ates

½ slope
0.1 -1.1121
y = 1.4112x
q /G p , 1 /d a y

0.1 ¼ slope
Well 1
Well 2 -1.019
Well 3 y = 0.9403x
Well 4 Well 1
0.01 Well 2 -1.1128
Well 3 y = 1.18x
-1.2791
Well 4 y = 1.9091x

0.01
1 10 Days 100 1000 0.001
1 10 100 1000
Days
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 fracture. There is a wider range for “a” and “m” observed in this play compared to other types.

Individual Well Analysis

In this example, a step-by-step procedure is shown in Fig. 8 on how to perform decline analysis on an individual well basis. A
vertical well taken from the Barnett shale has been selected for this illustration. The well has been on production for the last 5
years and operated with minimum water production.
CSUG/SPE 137748 7

Step 1: Data Check and Correction Step 2: a & m Determinations

100 100
Press. Corrected Rate 1
Measured Rate
-1.1384
Wellhead Pressure y = 1.4105x
0.1 2
R = 0.9744

q/Gp, 1/day
Rate, E3 m3/d Skin effect

WHP, MPa
10 10 0.01

Skin effect & 0.001

Surface choking

0.0001
1 1 1 10 100 Days 1,000 10,000
1 10 Days 100 1,000 10,000

Step 3: q1 Determination Step 4: Validation and Forecasts

100
40

Rate E3 m3/d or Cum. E6 m3

Gas Rate, E3 m3/d

30 Rate Data
y = 35.18x - 0.5207 Matched Rate
Cum Gas
20 Cal. Cum Gas
10

10 y = 33.7x

0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
1
t(a,m) 0 365 730 1,095 1,460 1,825 2,190
Days

Fig. 8- Four steps for using the proposed approach

1. Step 1: Data Check and Correction. The production data histories such as flowing wellhead pressures, gas rate and
water rate are plotted as shown in the top left graph. The well was initially choked back at the surface based on the
observed flat rate and high wellhead pressures (WHP). A correction was made for the rate based on average operating
pressures. The wet gas rate should also be corrected to the dry gas rate equivalent if there is a high condensate gas ratio
(CGR).
2. Step 2: “a” & “m” Determination. A log-log plot of q/Gp is constructed to determine these values as shown in the top
right graph. Data is examined to determine which section should be used to obtain the representative “a” and “m”
parameters for the well. The “R2” value is used to determine the best fit of the data. An R2 value of over 0.95 is
recommended.
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

q = q1t (a, m) …………………………..………………………..……………....….………… (13a)

a (t1− m −1)
where t ( a, m) = t
−m
e1− m
The q-vs.-t(a,m) plot should give a straight line through the origin with slope of q1. However, this may not be the best
selection in some cases due to the current wellbore operating conditions. In such cases, Eq. 13a becomes:

q = q1t ( a , m ) + q ∞ …………………………..……………………………..…..……………....….....……… (13b)

where q∞ is the rate at infinitive time, so it can be zero, positive or negative. In either case, the abandonment rate should
be set based on economics.

Once q1 (and q∞ ) have been determined, gas rate forecasts can be done by using Eq. 13a or Eq. 13b.

4. Step 4: Reserves Estimate. Eq. 14 modified from Eq. 6 is used for reserves estimate.

(t a 1− m −1)
q
G p = a1 e1− m + q∞ t ………………………………………………..……………....……....……… (14)
8 CSUG/SPE 137748

for q > qeco , where qeco is the minimum economic rate.

Result Comparisons
To compare the results, other methods such as Arps’ hyperbolic with b greater than 1 and the recently developed Power-Law-
Exponential method (Ilk et al. 2009) were selected. Texas A&M University has developed a Rate-Time Spreadsheet on using the
Power-Law-Exponential approach. Data from Example 3 from this spreadsheet was employed here for the purpose of comparison.
The analysis results are shown in Fig. 9. The results from the Power-Law-Exponential approach and Arps’ hyperbolic with b of
1.9 are included in Fig. 9 for comparison.

a&m Rate at day 1

1 150

0.1

Gas Rate, E 3 m 3/d

q/Gp, 1/day

100

0.01
y = 140.35x - 0.4779
y = 1.1142x-1.0895
50
0.001

0.0001 0
1 10 100 1000 10000 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Days
t(a,m)

Rate Comparison Cumulative Comparison

1000 1000
Data
Power-Law Exp
Gas Rate, E3 m3/d

Hyperbolic 100
Cum Gas, E6 m3

100 New Approach

10 Data
Power-Law Exp
10 Hyperbolic
1 New Approach

1 0.1
1 10 100 Days 1000 10000 100000 1 10 100 Days 1000 10000 100000

Fig. 9- Result comparison with Power-Law Exponential method and hyperbolic with b=1.9

The reserves which were calculated based on time limit of 30 years and economical rate of 2.83 103 m3/d (100 mscf/d) are
summarized in Table 1.

TABLE 1 – RESERVE COMPARISON BETWEEN METHODS

Time limit = 30 years
3 3
Rate limit = 2.83 10 m /day
Time limit Rate limit
6 3 6 3
Model Gp, 10 m Gp, 10 m

Hyperbolic (b=1.9) 118 222

Power-Law Exponential 116 188
New Approach 111 153

Table 1 shows that the new approach provides conservative reserve estimate compared to the other two methods.
Analog Type Curves
A field example was used to illustrate how a developing play can be used as an analog for other similar plays. This analog is in
form of type curves (P10, P50 and P90) which were generated by the new technique.
A set of 25 wells that includes 8 horizontal wells and 17 vertical wells from the Barnett shale were selected for this study. In this
study, there is not much difference between the vertical and horizontal wells except their initial rate. For this reason, no attempt
was made to separate them into two different data sets.
CSUG/SPE 137748 9

Following steps are proposed to establish the type curves:

1. Establish “a” and “m” for each individual well. “a” and “m” were determined for all 25 wells as shown in Fig. 10b.

Cumulative Normal Probability Chart

0.99
Slopes
Cumulative Normal Probability

0.90 Linear (Slopes) P90

0.50
P50

0.10
P10
0.01

1 1.05 1.1 1.15 1.2 1.25 1.3

a. Probability plot for 25 wells b. “a” and “m” for all 25 wells
Fig. 10- A statistical approach to analyze production forecasts for a group of wells or a set of sample wells.

2. Construct a probability plot to obtain the P10, P50 and P90 values for m. (Figure 12.a)
3. Establish an a-m relationship such as shown in Fig. 3 if sufficient data is available. Otherwise, use the correlation from
that figure.
4. Estimate the best-three-month rate averages for both vertical and horizontal wells. The best-three-month rate averages are
0.51 106 m3/month (18 MMscf/month) and 1.10 106 m3/month (39 MMscf/month) for the vertical and horizontal wells,
respectively. For comparison, Frantz et al. (2005) obtain the best-3-month-gas averages of 0.54 106 m3/month (19
MMscf/month) and 1.25 106 m3/month (44 MMscf/month) based on a data sample of 2842 vertical wells and 246
horizontal wells from the Barnett shale.
5. Use Eq. 11 to plot q3ma against Gp for various values of “a” and “m” to construct Fig. 11 for new well EUR prediction.
The left graph is for the EUR prediction based on the time limit of 30 years regardless of flow rate whereas the right
graph is based on 30 years of production with a cut-off rate of 2.83 103 m3/d (100 Mscf/d).
EUR RANGES FOR BARNETT (30 YEARS) 3 3
EUR RANGES FOR BARNETT (2.83 10 m /d )
2.0 2.0

P90 P50 P90 P50

P10
BEST-MONTH AVE, E6 m3

P10
BEST-3-MTH AVE, E6 m 3

1.5 1.5

1.0 1.0

Rate limit with 30 years

0.5
0.5

Time limit only EUR shifted due to

0.0 rate cutoff
0.0
0 20 40 60 80 100
EUR, E6 m3 0 20 40 60 80 100
EUR, E6 m3

Fig. 11- EUR prediction based on time and rate limit constraints. Time limit only would give an optimistic
prediction for EUR if the initial best rate is low.
10 CSUG/SPE 137748

TABLE 2 – PARAMETERS FOR TYPE CURVE GENERATION FOR BARNETT SHALE

Vertical Well Horizontal Well
6 3
q3ma, 10 m /mth 0.51 1.10
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
3 3
q1, 10 m /d 12.19 19.09 32.58 26.40 41.37 70.59

6. Use Eq. 9 to estimate the q1 values (P10, P50 and P90) based on the q3ma for both vertical and horizontal type wells. The
results for this example are shown in Table 2. The type curves as shown in Fig. 12 were constructed based on these
parameters.

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

14 P90 P90
30
P50 P50
12
Gas Rate, E3 m3/d

Gas Rate, E3 m3/d

P10 25 P10
10
20
8
15
6
10
4

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
Months Months

Fig. 12- Type Curves for Barnett Shale based on a set of 25 wells for both vertical and horizontal wells

Conventional Tight Gas Wells

All examples illustrated above result in an “m” value greater than one. However, m can be less than one as shown in Fig. 3. The
cases where m is less than 1 in Fig. 3 are from a shallow and tight gas formation. A typical rate decline analysis is shown in Fig.
13.
Rate & Pressure Histories a & m Values
4 4 1
Gas Rate
3 Pressure 3 0.1
Gas Rate, E3 m3/d

Pressure, MPa

q/Gp

2 2
0.01

-0.9624
0.001 y = 0.6226x
1 1 2
R = 0.9865

0.0001
0 0
1 10 100 1000 10000
0 365 730 1095 1460 Days
Days

q1 Determination Results
2 10

y = 5.3375x - 0.0424
Rate or Cum, E3 or E6 m3/d
Gas Rate, E3 m3/d

1.5
y = 5.1312x 1
Actual Rate
Cal. Rate 1/4 slope
1
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 1000 10000
t(a,m) Days

Fig. 13- A rate decline example from a shallow and tight gas formation

Notice slightly concave-up shape of the rate plot as shown on the bottom right graph in Fig. 13. This concave-up shape which is
caused by a late time pseudoradial flow is characteristic for m <1. For m>1, a log-log plot of rate vs. time will display a concave-
down feature (Fig. 4).
CSUG/SPE 137748 11

This feature distinguishes flow characteristics of a conventional (late-time pseudoradial flow with m <1) versus unconventional
(lack of late-time pseudoradial flow or m>1) reservoir.
Discussion
Absence of Late-Time Pseudoradial Flow
Unsurprisingly, the “m” value for all shale gas wells investigated in this study is greater than 1 or a log-log plot of rate vs. time
will display a concave-down feature (Fig. 4). This flow characteristic that indicates an absence of the late-time pseudoradial flow
from shale gas reservoirs can be explained as below.
Fig. 14 was constructed to distinguish between a Pressure
Top reservoir
conventional (tight) and shale gas formation with an
assumption that point A, wellbore location, has the same A
pressure in both conventional and unconventional Unconventional
Depth
formations. If well A is located in a tight but conventional Shale Gas
formation, its pressure will be in equilibrium with its fluid 20 - 60 kPa/m
Conventional
gradients. As in the case of a gas saturated formation, the
Tight Gas
pressure gradient should be about 2 kPa/m. A hydraulically
~ 2 kPa/m
fractured well located in such a formation will be expected
to pass through several distinct flow regimes such as fracture Bottom reservoir
linear, bilinear, elliptical, pseudoradial, and boundary- Fig. 14: Pressure gradient distribution in tight and shale gas
dominated flow as discussed by Clarkson and Bierler (2010). An unconventional reservoir may not be working according to such a
model since there is no matrix transient flow here to give rise to radial or elliptical flow.
If the reservoir is shale, the pressure gradient will be over-pressured. This can be up to 63 kPa/m (3 psi/ft) as recorded in the Lower
Bossier play (Newsham and Rushing 2002) and shown in Fig. 15. Spencer (1987) shows some pressure gradients in overpressured
and gas bearing shale plays in the Rocky Mountain region. The interpreted pressure profile for the MWX site wells of the Piceance
basin shows that the pressure gradient based on measured pore pressures in the active zone is about 36 kPa/m (1.7 psi/ft). The
author of this study has also experienced pressure distributions across producing gas shales exceeding the lithostatic gradient as in
the Montney and other studied plays.
With this high pressure gradient in shale gas, the formation is thought to be in a dynamical equilibrium state where hydrocarbons
have been generated and leaked off the formation over geologic time. Under a normal production time scale, shale gas formations
can be treated as impermeable barriers that contain millions and millions of tiny isolated reservoirs such as sandy laminates, micro
cracks, isolated sand bodies, etc… To produce gas from this shale formation, these tiny reservoirs need to be connected by a
fracture network, without which they will remain isolated and unproductive.

~ 36 kPa/m pressure
~ 63 kPa/m pressure distribution (measured
distribution (acoustic pore pressure)
log response)

a. Pressure distribution across the Lower Bossier b. Pressure distribution across the Piceance
play (Newsham and Rushing 2002) Basin (Spencer 1987)
Fig. 15- Two examples of high pressure gradient distribution across the producing shale gas.

Due to such isolated characteristics of micro-reservoirs in shale gas formations, capillary pressure or relative permeability curves
should be applicable in each individual tiny reservoir, but they may not be applicable applied across the entire shale formation.
This can explain the reason why the late-time radial or elliptical flow regimes are absent from all shale gas wells studied.
12 CSUG/SPE 137748

EUR Constraints
In most cases, without the presence of BDF in shale reservoirs, drainage area can not 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 (m, q∞) with both time and economic
rate limits.
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 fracture system and wellbore. The values of m (See Fig. 7) and q∞ (Eq. 13b & Fig. 8)
are designed to account for this acceleration in rate decline.
Time Limit vs. Rate Limit
Fig. 11 shows a typical EUR prediction for an infill in the Barnett shale for both vertical and horizontal wells. The prediction based
on a time limit alone would give an optimistic value. Note that the EUR shift to the left when the rate limit was imposed on the
time limit plot as shown on the graph to the right in Fig. 11. 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 due to 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 tight and
fracture-dominated gas wells in unconventional reservoirs.
2. A new approach is also able to provide a statistical methodology to analyze production forecasts of resource plays and
establish a range of results for these forecasts, including probability distributions of reserves in forms 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 reservoir rock characteristics, 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 (m, 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 characteristics.
Nomenclature
Field Variables
a = Intercept constant defined by Eq. 4, d-1
b = Arps’ decline exponent, dimensionless
Di = Arps’ initial decline rate, d-1
kf = Fracture permeability, 10-3 μm2
km = Matrix permeability, 10-3 μm2
G = Original gas-in-place, m3
Gp = Cumulative gas production, m3
m = Slope defined by Eq. 4
n = Fracture time exponent, dimensionless
q = Rate, m3/d
q1 = Rate at day 1, m3/d
q1D = Rate at tD=1, m3/d
q3ma = Best-three-month average, m3/month
qeco = Economic cutoff rate, m3/d
qmax = Maximum rate, m3/d
q∞ = Rate at infinity time, m3/d
t = Time, d
tD = Dimensionless time defined by tD=Dit
tDxf = Dimensionless time based on fracture haft length
tm = Dimensionless time based on Eq. 12a
CSUG/SPE 137748 13

tmax = Time at maximum flow rate, d

tx = Prediction time, d
t(a,m) = Time function based on Eq. 13a
xf = Fracture haft length, m
Greek Letters
q
ε = Decline rate defined by ε (t ) = , 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 in reviewing the paper, coworkers in both
Houston and Calgary offices for peer review of the proposed methodology, and to Matt Maguire and his group for supplying the
Barnett production data.
References
Agarwal, R.G., Carter, R.D. and Pollock, C.B. 1981. Evaluation and Performance Prediction of Low-Permeability Gas
Wells Stimulated by Massive Hydraulic Fracturing. Paper SPE 6838 presented at the 1977 SPE-AIME Annual Technical
Conference and Exhibition, Denver, CO, USA, October 9-12, published in JPT, May 1979.
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 2001 SPE Annual Technical
Conference and Exhibition, New Orleans, LA, USA, 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. JCPT, Vol. 45, No. 3 (Mar., 2006) 31-37.
Arps, J. J. 1945. Analysis of Decline Curves. Published in Petroleum Transactions, AIME, 160 (1945): 228-247.
Bagnall, W.D. and Ryan, W.M. 1976. The Geology, Reserves, and Production Characteristics of the Devonian Shale in
Southwestern West Virginia. Paper presented at the Annual Appalacchian Petroleum Geology Symposium, Morgantown,
VA, USA, 1-4 March 1-4.
Bello, O.R. and Wattenbarger, A.R. 2008. Rate Transient Analysis in Naturally Fractured Shale Gas Reservoirs. Paper SPE
114591 presented at the 2008 CIPC/SPE Gas Technology Symposium, Calgary, AB, Canada, 16-19 June. doi:
10.2118/114591-MS.
Bello, O.R. and Wattenbarger, A.R. 2010. Multi-stage Hydraulically Fractured Horizontal Shale Gas Well Rate Transient
Analysis. Paper SPE 126754 presented at the 2010 SPE North Africa Technical Conference and Exhibition, Cairo, Egypt,
14-17 February. doi: 10.2118/126754-MS.
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 2010 SPE Unconventional Reservoir Conference,
Pittsburgh, PA, USA, 23-25 February. doi: 10.2118/131786-MS.
Cinco-Ley, H., Samaniego-V, F. and Dominguez, A. 1976. Transient Pressure Behavior for a Well with a Finite-
Conductivity Vertical Fracture. Paper SPE 6014 presented at the 1976 SPE-AIME Annual Technical Conference and
Exhibition, New Orleans, LA, USA, 3-6 October. Published in SPEJ, August 1978. doi: 10.2118/6014-PA.
Cinco-Ley, H. and Samaniego-V, F. 1978. Transient Pressure Analysis for Fractured Wells. Paper SPE 7490 presented at
the 1978 SPE Annual Technical Conference and Exhibition, Houston, TX, USA, 1-3 October, published in JPT, September
1981. doi: 10.2118/7490-PA.
Cinco-Ley, H. and Samaniego-V, F. 1981. Transient Pressure Analysis: Finite Conductivity Fracture Case versus Damaged
Fracture Case. Paper SPE 10179 presented at the 1981 SPE Annual Technical Conference and Exhibition, San Antonio,
TX, USA, 5-7 October. doi: 10.2118/10179-MS.
Fetkovich, M. J. 1980. Decline Curve Analysis Using Type Curves. Published in SPEJ of Petroleum Technology, June
1980.
14 CSUG/SPE 137748

Frantz Jr., J.H., William, 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 2005 SPE Annual Technical Conference and Exhibition, Dallas, TX, USA, 9-12 October.
doi: 10.2118/96917-MS.
Gringarten, A.C., Ramey, H.J and Raghavan, R. 1975. Applied Pressure Analysis for Fractured Wells. Published in JPT,
May 1975. doi: 10.2118/5496-PA.
Ilk, D., Rushing, J.A., Perego, A.D. and Blasingame, T.A. 2008. Exponential vs. Hyperbolic Decline in Tight Gas Sands -
Understanding the Origin and Implications for Reserve Estimates Using Arps’ Decline Curves. Paper SPE 116731
presented at the 2008 SPE Annual Technical Conference and Exhibition, Denver, CO, USA, 21-24 September. doi:
10.2118/116731-MS.
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 2009 SPE Annual Technical Conference and Exhibition, New Orleans,
LA, USA, 4-7 October. doi: 10.2118/125031-MS.
Lee, W.J. and Sidle, R.E. 2010. Gas Reserves Estimation in Resource Plays. Paper SPE 130102 presented at the 2010 SPE
Unconventional Reservoir Conference, Pittsburgh, PA, USA, 23-25 February. doi: 10.2118/130102-MS.
Mattar, L., Gault, B., Morad, K., Clarkson, C.R., Freeman, C.M., Ilk, D. and Blasingame, T.M. 2008. Production Analysis
and Forecasting of Shale Gas Reservoirs: Case History-Based Approach. Paper SPE 119897 presented at the 2008 SPE
Shale Gas Production Conference, Fort Worth, TX, USA, 16-18 November. doi: 10.2118/119897-MS.
Newsham, E. and Rushing, J.A., 2002. Laboratory and Field Observations of an Apparent Sub Capillary-Equilibrium Water
Saturation Distribution in a Tight Gas Sand Reservoir. Presented at the Gas Technology Symposium, Calgary, Canada, 30
April- 2 May. doi: 10.2118/16011-MS.
Spencer, W.C. 1987. Hydrocarbon Generation as a Mechanism for Overpressuring in Rocky Mountain Region. The
American Association of Petroleum Geologists Bulletin, Vol. 71, No. 4, April 1987, Pages 368-388.
Soltanzadeh, H. and Hawkes, D. C. 2008. Semi-analytical models for stress change and fault reactivation induced by
reservoir production and injection. Published in Journal of Petroleum Science and Engineering, Volume 60, Issue 2,
February 2008, Pages 71-85.
Valkό, P.P. 2009. Assigning value to stimulation in the Barnett Shale-A simultaneous analysis of 7000 plus production
histories and well completion records. Paper SPE 119639 presented at the 2009 SPE Hydraulic Fracturing Technology
Conference, The Woodlands, TX, USA, 19-21 January. doi: 10.2118/16011-MS.
Warpinski, N.R. and Branagan, P. T. 1989. Altered-Stress Fracturing. Paper SPE 17533 published in Journal of Petroleum
Technology, September 1989, Pages 990-997. doi: 10.2118/17533-PA.
Weng, X. and Siebrits, E. 2007. Effect of Production-Induced Stress Field on Refracture Propagation and Pressure
Response. Paper SPE 106043 presented at the 2007 SPE Hydraulic Fracturing Technology Conference, College Station,
TX, 29-31 January. doi: 10.2118/16011-MS.
Appendix A - General Equations for q and Gp
q
General equations are developed based on [ = ε (t ) ].
Gp

q
= Gp (A-1)
ε (t )
Taking a derivative respecting to time, we have:
d [q / ε (t )] = d Gp (A-2)
dt dt
dε (t )
Using q’=dq/dt and ε ' (t ) = , Eq. A-2 becomes
dt
CSUG/SPE 137748 15

q' / ε (t ) − qε ' (t ) / ε 2 (t ) = q (A-3)

or

dq / q = dε (t ) / ε (t ) + ε (t )dt (A-4)

Integrating both sides from t=1 to t, we have:

q ε (t ) t

ε (1) ∫1
ln = ln[ ] + ε (t )dt (A-5)
q1
where q1 is the theoretical rate at t=1 day,
t
∫ ε (t )dt
ε (t ) 1 (A-6)
q = q1. e
ε (1)
and,
t

q1 ∫ ε (t )dt
Gp = e1 (A-7)
ε (1)
q
If = at −m (A-8a)
Gp
or

ε (t ) = at − m (A-8b)

Eq. A-6 and A-7 can be written as:

(t a 1− m −1)
q
= t −m e1− m (A-9a)
q1
and

q a (t1−m −1)
G p = a1 e 1−m (A-9b)