SPE 159801

Pressure-Dependent Natural-Fracture Permeability in Shale and its Effect

on Shale-Gas Well Production
Y. Cho, SPE, Colorado School of Mines, O. G. Apaydin, SPE, EOG Resources, Inc., and E. Ozkan, SPE,
Colorado School of Mines
Copyright 2012, Society of Petroleum Engineers
This paper was prepared for presentation at the SPE Annual Technical Conference and Exhibition held in San Antonio, Texas, USA, 8-10 October 2012.
This paper was selected for presentation by an 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
This paper presents the results of an experimental study of pressure-dependent natural-fracture permeability in tight,
unconventional reservoirs. Bakken cores are used in the experiments. For the purpose of this paper, pressure-related
permeability losses in hydraulic fractures and matrix system are not considered. Experimental data are used to screen the
stress-dependent matrix-permeability correlations available in the literature for application to shale fractures. Selected
correlations are matched with the data to delineate the reasonable ranges of the correlation coefficients for shale fractures.
The applications of the correlations over practical ranges of pressure drop in shale reservoirs indicate over 80% reduction in
fracture permeability, with most of the permeability loss occurring during the initial pressure drop. To appraise the effect of
pressure-dependent natural-fracture permeability on shale-gas production, experimentally developed correlations are
incorporated in an analytical model of a fractured horizontal well surrounded by a stimulated reservoir volume. The model is
used to history match the performances of two wells in the Barnett and Haynesville formations. It is shown that the effect of
pressure-dependent natural-fracture permeability on shale-gas-well production is a function of the permeability of the matrix
system. If the matrix system is too tight, then the retained permeability of the natural fractures may still be sufficient for the
available volume of the fluid when the system pressure drops.
Introduction
Production performances of shale reservoirs depend strongly on the existence of a dense and conductive network of fractures
in the drainage volume of the well, known as stimulated reservoir volume (SRV). The natural-fracture network in the SRV is
usually generated or rejuvenated as a result of the stress and strain changes induced by hydraulic fracturing. Because they are
poorly propped, conductivity of natural fractures is highly susceptible to the changes of stress and strain during production
also. Therefore, incorporating stress-dependency of natural fracture permeabilities is a significant step toward more realistic
assessments of the production potential and the economic outlook of shale reservoirs. The main objective of the research
presented in this paper is to study the effect of stress-dependent natural-fracture permeability on the productivity of shale
reservoirs. Although the effect of stress on matrix (primary) permeability and hydraulic fracture conductivity has been
studied in more detail in the literature, information on the change of natural-fracture (secondary) permeability when the fluid
pressure declines is limited (e.g., Pedrosa, 1986, Celis et al., 1994, Best and Katsube, 1995, Gutierrez et al., 1999, Chen et al.,
2000a and b, Raghavan and Chen, 2004, Kwon et al., 2004, and Tao et al., 2010).
The objective of the research presented in this paper is not to model the stress dependency of shale natural fractures in its
full complexity. Considering our abilities to measure and characterize the properties of shale natural fractures, our intention
in this paper is limited to finding a first approximation with minimum data requirement. The use of uniaxial stress and
isotropic permeability assumptions allows us to represent the effect of effective stress in terms of pressure. Therefore, under
the assumptions of this research, pressure-dependent permeability correlations can be used to incorporate the effect of stressdependent permeability into dual-porosity models of shale reservoirs. It is also presumed that the pressure-dependency of the
matrix (primary) permeability is negligible compared with the natural-fracture (secondary) permeability in dual-porosity
systems. With these, assumptions, we have compiled the existing correlations in the literature and screened for the use in our
research. Because the existing correlations were developed mainly for matrix permeability, they were to be verified for
fracture permeability in shale.
For the verification of the selected correlations, we have used middle Bakken core samples taken at 9,026 ft in the

SPE 159801

Williston Basin of North Dakota. (Production in the Bakken formation is mainly from the dolomitized siltstone intervals; our
samples, however, are from the shale intervals. The selection of the Bakken cores was mainly because they were readily
available to us. Moreover, the success of hydraulic fracturing and horizontal drilling in the Bakken field made these cores a
good candidate for the general purposes of this research.) We matched the experimental fracture-permeability vs. pressure
data with the selected correlations and assessed the success of each correlation. CMS-300 Automated Permeameter was used
in the measurements (Core Labs, 2012).
Assessment of the impact of fracture-permeability reduction with production requires incorporating the experimental
correlations into a shale-reservoir performance model. In this research, we used an analytical model for a fractured horizontal
well surrounded by an SRV (Brown et al., 2011, and Ozkan et al., 2011). The model used here includes the standard
assumptions of the common fractured horizontal well models for shale gas production. It was selected for its availability and
convenience in modifications; any other analytical or numerical model, which permits incorporation of pressure-dependent
fracture-permeability, could be used for the same purpose.
Background
Fractures exist in many oil and gas formations and are the principle source of flow capacity in many of these reservoirs.
Changes in fracture flow capacity and conductivity are expected to have an important influence on reservoir performance.
Fractures are characterized mainly by their length, aperture (width), orientation, density, spacing, and porosity. Some of these
characteristics, such as the aperture and porosity, which determine the conductivity of the fracture, are functions of effective
stress. Thus, the conductivity of natural fractures is affected by the stress changes in the formation.
The mechanical response of the porous and fractured media to changes in temperature and effective stress and strain
include changes in void volumes with accompanying changes in permeability (Rutqvist et al., 2002). Normal stress-induced
closure/opening and shear-induced dilation are two basic mechanisms that change fracture apertures and fracture
transmissivity. The normal stress is generally the main cause for the aperture change. In addition, the changes in the apertures
are nearly uniform along the length of the fracture. The shear dilation is not significant with a low horizontal to vertical stress
ratio in most cases (Min et al., 2004). The stresses can be redistributed as a result of the closure of fracture apertures.
Furthermore, a varying aperture is likely to exist in a fracture network (Baghbanan and Jing, 2007). Thus, it is important to
understand and model the influence of stress-induced deformation (closure) on natural fracture permeability.
When the stress-dependent permeability is modeled, the next step is to incorporate it into flow models to assess its
impact on reservoir performance. Pedrosa (1986) investigated pressure-transient responses in stress-sensitive formations by
defining a permeability modulus, which is the fractional change of permeability with unit change in pressure. The
permeability modulus concept assumed an exponential variation of permeability with pressure, which was simplified by
using the Taylor expansion of the diffusivity and the permeability moduli. Celis et al. (1994) used the permeability modulus
concept to analytically model unsteady- and pseudosteady-state flow of oil in naturally fractured, stress-sensitive reservoirs.
Best and Katsube (1995), however, focused specifically on shale permeability and noted that there was no agreement on
whether the permeabilitypressure relationship in shale could be represented by a single mathematical expression.
Gutierrez et al. (1999) conducted an experiment to study stress-dependent permeability of de-mineralized fractures in
shale. They showed that the fracture permeability decreased significantly by increasing effective normal stress across the
fracture, and after shearing of the fracture under high constant normal stress. However, they noted that the fractures never
completely closed under normal stresses; the only condition for the fractures to lose their conductivity completely was
cementation.
Raghavan and Chin (2004) presented a fully coupled geomechanical and fluid-flow model to analyze pressure-transient
responses of wells in stress-sensitive reservoirs with nonlinear elastic and plastic constitutive behavior. Their model
incorporated three basic principles, linear momentum, mass conservation, and Darcys law, to compute pressure responses
using finite-element discretization. The model was tested by problems with known solutions to verify the numerical accuracy
and ability to retain correct physical principles. In addition to the model development, Raghavan and Chin (2004) presented
correlations to evaluate productivity changes in stress-sensitive reservoirs that may be used for production forecasts. Using
the developed numerical model, the stress impact on physical and mechanical properties was analyzed extensively.
Kwon et al. (2004) experimentally investigated anisotropy and effects of clay content and loading. They developed a
correlation using a cubic law to represent permeability reduction at different pressures. The permeability measurement
showed different results based on flow direction (parallel and perpendicular to bedding) and clay content. The permeability
measurement parallel to the bedding was about one order of magnitude greater than that perpendicular to the bedding.
Furthermore, the degree of permeability anisotropy was reduced at higher pressures.
Franquet et al. (2004) considered the effect of pressure-dependent permeability on transient radial flow in tight-gas
reservoirs by defining a modified real-gas pseudopressure. They demonstrated that neglecting the variation of rock
compressibility under various states of stress might result in performance predictions with substantial errors. Xiangjiao et al.
(2009) investigated pressure-transient behavior of stress-sensitive gas reservoirs by using a finite-element method. They also
concluded that the consequences of not accounting for the stress-dependent permeability could be significant. Wang et al.
(2009) discussed that pressure-dependent permeability and the Klinkenberg slippage effect are only substantial when the pore
pressure is less than 500 psi. Nevertheless, they suggested that the permeability reduction should be accounted for due to the
considerably high permeability reduction in shale as compared to consolidated sandstones or carbonates.

SPE 159801

Tao et al. (2010) presented an indirect boundary element method to determine the fracture aperture changes by pressuretransients in a naturally fractured medium. They focused on fracture deformation because the fractures dominate the flow
behavior and are more sensitive to pressure and stress change than the matrix. Their model combined a fully coupled
displacement discontinuity method and the Barton-Bandis (1983) joint deformation model. Tao et al. found that the fracture
permeability decreases as the pore pressure depletes under isotropic stress conditions. However, under highly anisotropic
stress conditions, the fracture permeability increases at intermediate times due to shear dilation. Therefore, additional
geomechanical data should be incorporated into the model to accurately evaluate the effect of stress-dependent fracture
permeability on transient pressure responses of fractured reservoirs.
Experimental Study
Although several pressure-dependent permeability correlations have been discussed in the literature, most of the discussions
consider the primary permeability in the porous matrix. Experimental data for permeability-stress relationship in natural
fractures have not been sufficiently documented. We used core samples from middle Bakken formation, taken approximately
at 9,026 ft, in the Williston basin of North Dakota, to obtain experimental data for the stress-dependent permeability of
natural fractures in extremely tight shales and carbonates.
Fig. 1 shows some core samples used in the experiments. All cores were cut cylindrically into 1 diameter, and the
length of the cores varied from 0.8 to 1.5. The average length was approximately 1. The cores were cut mostly
perpendicular to the bedding plane to represent the reservoir, except a few, which were cut parallel to the bedding plane to
analyze the effect of fracture closure at different conditions. Then, the cores were cleaned with toluene for several days in a
soxhlet extraction set up. After the cores were cleaned, the matrix permeabilities and porosities were measured using the
CMS-300 Automated Permeameter (Core Labs, 2012).

Figure 1 - Bakken core sample used in the experiments: (A) Original sample with 1 scale; (B) Top view of the sample; (C) 1 Bakken
Core Plugs

The CMS-300 is an automated core measurement system that measures the porosity and permeability of the reservoir
rock samples at each confining stress in the range of 500 to 9,800 psi. The grain volume is measured using Boyles Law by
the CMS-300 system to calculate the pore volume and porosity. The grain volume is considered to be incompressible for the
experiment. CMS-300 can measure pore volumes between 0.02 and 25 cm3 and the permeability between 0.1 d and 5000
md. In the experiments, matrix permeability could not be measured due to the tightness of the cores. Thus, it was concluded
that the matrix permeability was below 0.1 d. To simulate the natural fractures, cores were cut vertically at the center. The
fractured cores were then held together with Teflon tapes. Fig. 2 shows the geometry of the cores used in the CMS-300
experiments.

SPE 159801

Figure 2 - Schematic of the core samples used in the CMS-300 measurements.

During the experiments, the confining stress was increased gradually from 1,000 psi to 5,000 psi by 1,000-psi increments
to study the effect of stress on fracture closure. The confining stress was created around the core by nitrogen gas. This created
the biaxial conditions during the experiment. Helium was injected at a constant rate at 245 psi to the core. Effective porosities
and permeabilities of the fractured cores were measured repeatedly using CMS-300. The flowchart in Fig. 3 summarizes the
procedure used in the CMS-300 experiments.

Start

Create fracture

Collect core sample

Measure effective fracture

permeability & porosity
Yes

Cut a core into 1" in diameter

& in desirable length

Repeat?
No

Soxhlet extractor with

toluene

End
No

Clean?
Yes

Measure matrix
permeability
Figure 3 - The CMS-300 experimental procedure.

The effective permeability of the fracture was measured repeatedly at three different confining stresses. Based on the
three effective permeability measurements, the fracture permeability, porosity, and width were calculated using the following
equations:

keffective = km + f k f ,

(1)

SPE 159801

k f = 10 3

f =

w 2f
12

4w f
10 4 d

(2)

(3)

w f = 3 30 d(keffective km ) .

(4)

The experimental results for Cores 1 and 3 are presented in Figs. 4 and 5 and summarized in Tables 1 and 2, respectively.

Figure 4 - Bakken Core 1 measurements by CMS-300: (A) Fracture permeability vs. confining stress; (B) Fracture porosity vs.
confining stress; (C) Fracture width vs. confining stress.

Figure 4 - Bakken Core 3 measurements by CMS-300: (A) Fracture permeability vs. confining stress; (B) Fracture porosity vs.
confining stress; (C) Fracture width vs. confining stress.

SPE 159801

TABLE 1 Summary of CMS-300 Measurements

Unfractured Core
Confining
Stress, psia

Bakken
1

Effective
*Effective
Matrix
Confining **Effective
Fracture
Stress, Permeability, Stress,
Stress,
Permeability,
psia
km, md
psia
psia
keff, md

Bakken
1

951

851

7.24E-04

1,043

993

4.87

9.23

10.52

5.28E-04

1,626

4.10E-04

1,992

1,942

3.77

7.79

9.67

4.85E-04

2,435

2,335

3.26E-04

2,992

2,942

3.00

6.68

8.95

4.49E-04

4,007

3,957

2.43

5.80

8.34

4.18E-04

4,478

4,428

2.36

5.69

8.26

4.14E-04

995

895

1.31E-03

989

939

3.58E

7.52

9.50

4.76E-04

1,717

1,617

6.06E-04

1,988

1,938

2.37

5.71

8.28

4.15E-04

2,443

2,343

3.92E-04

3,009

2,959

1.60

4.40

7.27

3.64E-04

3,989

3,939

1.29

3.81

6.76

3.39E-04

4,517

4,467

1.15

3.53

6.51

3.26E-04

951

851

7.24E-04

1,021

971

4.54

8.81

10.28

5.15E-04

1,726

1,626

4.10E-04

1,982

1,932

3.64

7.60

9.55

4.79E-04

2,435

2,335

3.26E-04

2,986

2,936

3.07

6.78

9.02

4.52E-04

3,964

3,914

2.64

6.14

8.58

4.30E-04

4,499

4,449

2.42

5.78

8.33

4.18E-04

4,806

4,756

2.32

5.62

8.22

4.12E-04

Bakken
1

Bakken
3

Bakken
1

995

895

1.31E-03

1,051

1,001

3.42

7.29

9.35

4.69E-04

1,717

1,617

6.06E-04

2,024

1,974

2.12

5.31

7.98

4.00E-04

2,443

2,343

3.92E-04

3,015

2,965

1.57

4.34

7.21

3.62E-04

4,033

3,983

1.28

3.78

6.73

3.37E-04

4,496

4,446

1.15

3.53

6.50

3.26E-04

4,823

4,773

1.10

3.41

6.40

3.21E-04

951

851

7.24E-04

1,000

950

4.64

8.93

10.35

5.19E-04

1,726

1,626

4.10E-04

1,793

1,743

4.00

8.10

9.86

4.94E-04

2,435

2,335

3.26E-04

2,520

2,470

3.37

7.22

9.31

4.67E-04

3,982

3,932

2.62

6.11

8.56

4.29E-04

4,488

4,438

2.39

5.74

8.30

4.16E-04

4,832

4,782

2.31

5.62

8.21

4.12E-04

Bakken
1

Bakken
3

Fracture
Fracture
Fracture
Permeability,
Width,
Porosity,
kf , D
wf, micron f, fraction

1,726

Bakken
3

Fractured Core

995

895

1.31E-03

1,051

1,001

2.56

6.01

8.50

4.26E-04

1,717

1,617

6.06E-04

1,832

1,782

1.89

4.92

7.68

3.85E-04

2,443

2,343

3.92E-04

2,492

2,442

1.52

4.25

7.14

3.58E-04

3,905

3,855

1.15

3.52

6.50

3.26E-04

4,464

4,414

1.05

3.31

6.30

3.16E-04

4,794

4,744

1.00

3.21

6.21

3.11E-04

*Unfracture pore pressure 100 psi

**Fractured pore pressure 50 psi

We also used the experimental data obtained by Alamdari (2011) from two carbonate core-plugs (including a Silurian
dolomite outcrop) to test the applicability of the correlations presented in the literature and to obtain correlation coefficients
for natural fractures in tight carbonates. Fig. 5 shows the carbonate permeability-confining pressure relationship for
Alamdaris data.

SPE 159801

Figure 5 Fracture permeability-confining pressure relationship for two carbonate cores (Alamdari, 2011).

Discussion on the Effect of Fracture Surface Roughness

Fracture permeability is not a unique function of effective stress; it also depends on the fracture geometry and the roughness
of the fracture surface. Fractures in nature can have complex geometry and it is extremely difficult to model the impact of
geometry on permeability. In the above described fracture-permeability estimations, planar fracture geometry has been
assumed. Similarly, the surface roughness of natural fractures is extremely difficult to determine in the field. To assess the
relationship between fracture permeability and fracture surface roughness, however, we measured the surface roughness by
using a profilometer prior to stress deformation.
The profilometer used in the surface roughness measurements was capable of measuring cross-sectional view of the
roughness below 10 and waviness parameters up to 14 width and 2.25 thickness. The Bakken cores used in this research
were cut into two halves using an IsoMet low-speed saw to minimize the variations on the surface. Surface roughness was
measured on each half approximately 1 cm along the axis of the cores. The cores were scanned at 200 m/s and the data were
recorded at 50 Hz. The surface roughness was measured at every 4 m. (It must be noted that the measurements were not
completely satisfactory due to the machine limitations. First, the surface roughness varied significantly along the fracture but
the measurements could only be made in a 1-cm section along the center of the cores. Second, the machine was designed to
measure much smoother surfaces. Third, some exterior effects, such as the mechanical, acoustic, thermal, and electrical
conditions of the operations, could not be controlled and interfered with the integrity of the measurements. Therefore,
although the results of the experiments are reported here, they are for information only and should be taken cautiously.
Additional studies with more appropriate experimental set-up and measurement procedures are recommended.)
The measured surface roughness distributions for four core samples are presented in Fig. 6 and the average roughness,
Ra, and the maximum peak to minimum roughness, Rt, values are shown in Table 2. This information was used to compute
the effective fracture width by

weff = w f Ra

(5)

and then the effective fracture permeability was computed from

keff = 10 3

2
weff

12

(6)

Figure 6 Measured fracture surface roughnesses of the Bakken core samples: (A) Sample 1A; (B) Sample 1B; (C) Sample 3A; (D)
Sample 3B.

SPE 159801

TABLE 2 - Roughness data measured by the profilometer

SAMPLE CORE

AVERAGE ROUGHNESS, m

MAXIMUM PEAK TO MINIMUM ROUGHNESS, m

Bakken 1A

11.64

153.76

Bakken 1B

40.51

144.51

Bakken 3A

3.58

19.92

Bakken 3B

4.37

30.09

Table 3 indicates that, for the surface roughness values in Table 2, the average variation in permeability (the difference
between the permeabilities computed with and without consideration of the surface roughness) is less than 13%. Therefore, it
may be concluded that the effect of fracture surface roughness is relatively insignificant for the purposes of this research.
This result may be justified on the basis of the macroscopic definition of permeability used in this study. In other words, an
average fracture aperture is sufficient to define the permeability in this case. This approach is akin to neglecting the
individual effects of tortuosity and pore-throat size distribution in the definition of rock-matrix permeability at core scale. For
the study of flow at the pore (or microscopic) scale, the effect of surface roughness may not be lumped with the other
correlation parameters. However, from a practical viewpoint, it is not possible to infer the surface roughness characteristics of
natural fractures from any conventional measurement. Therefore, this brief discussion of the effect of surface roughness on
fracture permeability has been provided only for completeness and will not be further pursued in this paper; it will be
assumed that the effect of fracture-surface roughness on permeability could be included as a lumped parameter in the
correlation constants and coefficients.

TABLE 3 - Effective width and permeability data

Sample
Core
Bakken 1A
Bakken 1B
Bakken 3A
Bakken 3B

Fracture Width,
m

Effective Fracture Width,

m

Fracture Permeability,
D

Fracture Effective
Permeability, D

458
458
254
254

435
377
237
235

17,480
17,480
5,376
5,376

15,748
11,843
4,674
4,612

!
Pressure-Dependent Permeability Correlations
As noted by Best and Katsube (1995), there is no sufficient support for a firm statement on how fracture conductivity
changes as a result of stress-induced deformation or closure. The main challenges are to model the complex fracture
geometry resulting from different orientations and sizes of fractures and to express the complex mechanical deformation as a
function of the interactions between individual fractures. Thus, different correlations have been proposed to model complex
fracture systems with deformation of fractures.
We have screened the correlations proposed in the literature for stress- or pressure-dependent permeability and initially
selected the nine correlations shown in Table 4 for further verification with the experimental data. The correlations have been
chosen based on the practical applicability (that is, the convenience in incorporating into existing fractured horizontal-well
models in shale formations) and data availability. Some correlations presented in the literature require specific experimental
data and/or coefficients, which are not the focus of this research. Therefore, these correlations have been disregarded in this
research.
In general, the correlations in Table 4 display an exponential relationship between the permeability and the pressure
change. The input data required for most of these correlations are similar and include the initial fracture permeability, kfi,
initial fracture porosity, fi, and fracture compressibility, cfi. Some of the correlations require additional or alternative
information to be applicable. The input variables have to be measured experimentally. However, in reality, the data required
to use the correlations in Table 4 may not be available. Therefore, a desirable feature of the correlation to be used in
applications is sufficient accuracy with minimum input variables.
The fracture compressibility, cf, required by some of the correlations in Table 4 is a key variable to determine the effect
of stress on long-term productivity. For this research, the compressibility is calculated based on experimental porosity data
using the following equation:

cf =

1 v f
1 f
=
v f p
f p

(7)

SPE 159801

Similarly, Youngs modulus, E, and Poissons ratio, v, are important to determine the stress-sensitive properties; but these are
usually difficult to measure due to apparatus limitations. These geomechanical properties are obtained by matching the
experimental data with the correlation. The coefficients a1 and c1 in Correlation 1 in Table 4 are also determined by matching
the experimental data. The pressure-dependent permeability is correlated as a function of the changes in fracture width in
Correlations 2 and 7.

TABLE 4 Selected Correlations from the Literature

Correlation

k f = k fi e

c1

Reference

( 1) with = + ( )ea1p
r1
i
r1

Rutqvist et al. (2002)

'
'
d '
d '
b
(e 2 y e 2 yi ) + bmax 2 (ed2 z ed2 zi )
max
2

k = ki Fk with Fk =
3
3
bi2 + bi2

k f = k fi e

Raghavan and Chin (2004)

(Rock Type I)

d f 3p

Raghavan and Chin (2004)

Rock Type II

k f = k fi (1 m f 4 p)

k f = k fi (

f n
1 nf
p(1+ v5 )(2v5 1)
) with n+1
=
1
and v = 5
f
fi
E5 (1 v5 )
e v

Raghavan and Chin (2004)

Rock Type III

f n
) with f = fi ed f 6 p
fi

Raghavan and Chin (2004)

and Celis et al. (1994)

k f = k fi (

k f = k fi (

f n
) with f = fi F , F = b1 + b2 + b3 , and
fi
b1i + b2i + b3i
,

b = bi 7 + bmax 7 e

k f = k fi (

d f 7 n'

'
d f 7 ni

Raghavan and Chin (2004)

and Rutqvist et al. (2002)

f n
) with f = fi 1 m f 8 p
fi

Raghavan and Chin (2004)

and Minkoff et al. (2003)

f n
(1 )
) with = 1 i and v = 9 p(1+ v9 )(2v9 1)
fi
E9 (1 v9 )
ev

Raghavan and Chin (2004)

and Minkoff et al. (2003)

k f = k fi (

Rutqvist et al. (2002)

Correlations 3, 4, 6, and 8 in Table 4 are derived by integrating the compressibility equation (Eq. 7) as follows:
v2

dv f

v1

p2

= c f dp .

(8)

p1

Assuming cf is constant with respect to pressure, we obtain from Eq. 8

v f 2 = v f 1e

c f p

(9)

Correlations 4 and 8 in Table 4 assume that cf is small and approximate Eq. 9 as follows:

v f 2 v f 1 (1 c f p) .

(10)

Correlations 5 and 9 are derived by using the isotropic, linear-elastic model to simulate the constitutive behavior. Two
geomechanical variables of the constitutive model are the Youngs modulus, E, and Poissons ratio, v. The constitutive
equation is presented in the following matrix form:

10

SPE 159801

'x
'y
'z
'xy
'yz
'zx

E
C
=
(1+ v)(1 2v)

x
y
z
xy
yz
zx

(11)

where

1 v
v
v
0
0
0
v 1 v
v
0
0
0

v
v
1
v
0
0
0
C=
0
0
0
0.5 v
0
0

0
0
0
0
0.5

v
0

0
0
0
0
0
0.5 v

(12)

The uniaxial 1-D strain is obtained from Eq. 11 as follows

v = z = p

(1+ v)(1 2v)

,
E(1 v)

(13)

and applied to porosity functions. Then, a power-law relationship is applied to relate permeability changes and rock porosity.
Correlations 1, 2, and 7 are empirical correlations that were developed using experimental statistical data.
The correlations presented in Table 4 have been proposed in the literature mainly for pressure-dependent rock-matrix
(primary) permeability. Furthermore, the constants and coefficients used in these equations are available mainly for
applications to rock-matrix permeability. In this work, new experimental data have been used to match the correlations in
Table 4 to verify their applicability to fracture permeability in shale and carbonate and sensitivity analyses have been
performed to estimate the correlation constants and coefficients. Figure 7 demonstrates, as an example, matching the
experimental data from two carbonate (Fig. 7A) and two shale (Fig. 7B) cores by Correlation 3 (Raghavan and Chin, 2004) in
Table 4. Similar sensitivity analyses have been performed for all correlations given in Table 4; a complete set of the results is
available in the work of Cho (2012).

Figure 7 Matching the experimental data with Correlation # 3 in Table 4: (A) Carbonate Cores # 22 and 32; (B) Shale (Bakken) Cores
# 1-2 and 3-1.

SPE 159801

11

Table 5 summarizes the constants and coefficients obtained by matching the experimental data in this work with the
correlations in Table 4. The data given in Table 5 are specific to the core samples used in this study; they are not presented
for general use. However, they are useful to provide a sense of magnitude and range when the correlations in Table 4 are used
to model pressure-dependent fracture permeability in tight shale and carbonate formations.

!
PARAMETER!
OR!
CONSTANT!
a1#

c1#
r1!
bi2#
bmax2#
df2#
df3#
mf4#
5#
v5#
E5#
N5#
df6#
N6#
bi7#
bmax7#
df7#
N7#
mf8#
N8#
9#
v9#
E9#
N9#

TABLE!5!!Correlation!Parameters!
SHALE!
CARBONATE!
CORE!1!

CORE!3!

CORE!22!

CORE!32!

3.19E@4!

3.20E@4!

4.20E@4!

4.50E@4!

2.15!
3.70E@4!

2.35!
2.64E@4!

2.50!
4.86E@4!

2.60!
5.21E@4!

7.38!
12.23!

5.26!
10.11!

14.55!
30.18!

15.60!
42.57!

1.10E@4!
2.02E@4!

1.00E@4!
2.61E@4!

9.00E@5!
2.92E@4!

7.00E@5!
4.02E@4!

1.60E@4!
0.30!

2.00E@4!
0.30!

2.0E@4!
0.68!

2.0E@4!
0.65!

0.47!
1.0E6!

0.47!
1.0E6!

0.47!
1.0E6!

0.44!
1.0E6!

2!
1.01E@4!

2!
1.31E@4!

2!
1.46E@4!

2!
2.01E@4!

2!
7.38!

2!
5.26!

2!
14.55!

2!
15.60!

12.23!
6.00E@5!

10.11!
6.00E@5!

30.18!
6.00E@5!

42.57!
6.00E@5!

2!
9.00E@4!

2!
1.10E@4!

2!
1.15E@4!

2!
1.40E@4!

2!
0.35!

2!
0.35!

2!
0.85!

2!
0.65!

0.47!
1.00E6!

0.47!
1.00E6!

0.48!
1.00E6!

0.44!
1.00E6!

2!

2!

2!

2!

According to the sensitivity analysis, among the correlations presented in Table 4, Correlation 3 presented by Raghavan
and Chin (2004) has been identified as the most practical correlation (it is simple to use and does not require extensive data)
that demonstrates the best fit for the Bakken-shale and carbonate-core data. Figure 8, taken from Ozkan et al. (2010),
demonstrates the effect of pressure on fracture permeability (modeled by Correlation 3 in Table 4). It is shown that
approximately 75% (~ 1500 md) of the initial fracture permeability (2000 md at 5000 psi) is lost in the first 2000-psi pressure
drop. This is an important observation because most shale-gas reservoirs are overpressured and display a sharp pressure drop
initially. Also, considering a reasonable bottomhole flowing pressure of 500 psi for a fractured horizontal shale-gas well, the
fracture permeability around the well is expected to be approximately 200 psi (~ 10% of the initial permeability).

Figure 8 Change in fracture permeability as a function of pressure as predicted by Correlation 3 in Table 4 (Ozkan et al, 2010).

12

SPE 159801

The results shown in Fig. 8 may seem to support the arguments suggesting proppant use to prevent the closure of natural
fractures. However, it should be noted that the magnitude of the fracture permeability needed for a given system depends on
the volume of fluid available to flow in the fracture network (if there is not enough fluid to flow, then there is no need for
high fracture permeability). Therefore, without consideration of the flow from matrix into the fractures, the interpretation of
the results in Fig. 8 may be deceiving. In the following section, we incorporate the pressure-dependent permeability
represented by Correlation 3 into an analytical flow model to have a more realistic assessment of the effect of fracture
permeability reduction on productivity.
Effect on Pressure-Dependent Fracture Permeability on Well Performance
To demonstrate the effect of the reduction in natural-fracture permeability on shale-gas well performance, we use the
analytical, trilinear model of a fractured horizontal well with a stimulated reservoir volume introduced by Brown et al., 2011,
and Ozkan et al., 2011. The choice of the model is simply because of its availability and the ease in modifying; otherwise,
any other analytical or numerical model, capable of incorporating a pressure-dependent permeability, could be used for the
purposes of this section. The details of the trilinear flow model are given by Brown et al., 2011, and Ozkan et al., 2011 and
will not be repeated here. It is only important to note that the pressure-dependency of the fracture-network permeability
makes the model highly non-linear. To permit an analytical solution, the pressure-dependent part of the permeability is
included in the definition of the real-gas pseudopressure (Al-Husseiny et al. 1966) and then an iterative approach is used for
the solution. The details of the model and the solution pressure are not essential for the main objectives of this paper and will
not be discussed further. The interested readers are referred to the work by Cho (2012), Ozkan et al. (2010), and Apaydin
(2012).
In order to investigate the effect of pressure-dependent fracture permeability on fractured horizontal-well performances
in shale-gas reservoirs, first, we discuss the history matching of the performances of two gas wells in Haynesville and Barnett
formations. For both examples, we use the production history in the trilinear model to generate the corresponding bottomhole
pressure responses. We iterate on input model parameters until the model predicts the pressure history reasonably well.
Although we have matched the data by using all correlations in Table 4, we will only show the results for Correlation 3 here
because all, except Correlation 2, yielded similar results. History matching results with the other correlations can be found in
Cho (2012). To demonstrate the effect of the pressure-dependent fracture permeability, we run the model first with pressuredependent fracture permeability. When we obtain a good match, we fix all the other input parameters and remove the
pressure-dependency of the fracture permeability (that is, we run the model with a constant natural-fracture permeability
equal to the initial permeability of the fracture system).
After the discussion of the two field examples, we also present a synthetic example to address some of the issues noted
in the discussion of the field data. This example is intended to discuss the effect of pressure-dependent permeability on well
performance under controlled conditions to make sure that the only change in the performance is due to the change in the
fracture permeability.
Haynesville Shale-Gas Well
The first example is from a multiply fractured horizontal gas well in Haynesville shale and was taken from Wang and Liu
(2011). The horizontal well in the example was drilled 10,600 ft vertically and extended 3,200 ft laterally. In addition,
hydraulic fracturing was performed at 10 stages for 4 perforations per stage. Fig. 9 shows the gas production for the
Haynesville well for over 3 months. The flowing bottom-hole pressure was calculated from the tubing-head pressure using
two-phase (water-gas) flow correlations and is presented in Fig. 10. In the early-portion of the production, the well was shut
in for approximately 10 days. Before the shut-in period, a rapid decline in the initial pressure was observed which is
characteristic to overpressured reservoirs. Because the effect of overpressuring is not considered in the trilinear model, the
history match focused on the production after the shut-in period.

Figure 9 Production history of the Haynesville shale-gas well (Wang and Liu, 2011).

SPE 159801

13

Figure 10 Pressure data and the history match of the Hayneswille shale-gas well; Trilinear model with Correlation 3.

Figure 10 shows the history match of the pressure responses by the trilinear model including Correlation 3. Match values
of the reservoir properties and the correlation coefficients used for the pressure-dependent permeability are given in Table 6.
The transient pressure responses from the trilinear model with pressure-dependent permeability match the actual pressure
data better than the model with constant fracture permeability. As expected, the constant fracture permeability model matches
the pressure history at early times; at later times, however, it predicts higher pressures because it does not take into account
the loss of conductivity in the fracture network as pressure drops.

Barnett Shale-Gas Well

The Barnett shale-gas well data for the second history match example are from a hydraulically fractured horizontal well. The
daily pressure and gas production history over three years were reproduced from Anderson et al. (2010). (The data included
some ambient noise, which required some corrections.) The production and pressure histories for the Barnett shale-gas well
are presented in Figs. 11 and 12, respectively. The history matches of the pressure data, with pressure-dependent and constant
natural-fracture permeabilities, are shown in Fig. 12. Table 7 summarizes the match parameters.

14

SPE 159801

Figure 11 Production history of the Barnett shale-gas well (Anderson et al., 2010).

Figure 12 Pressure data and the history match of the Barnett shale-gas well; Trilinear model with Correlation 3.

TABLE 7 Well, reservoir, and fluid data for the Barnett Well example
Specific gravity of gas, SG

0. 8

Initial reservoir pressure, pi, psi

o
Reservoir temperature, T, F
Formation thickness, h, ft

1,600
106
300

Wellbore radius, rw, ft

Reservoir size perpendicular to well axis, xe, ft
Half-distance between hydraulic fractures, ye, ft
Constant matrix permeability, km, md
Initial matrix porosity,!mi!
Initial matrix compressibility, ctmi, psi-1
Number of fractures in pay thickness

90.3
1.0E-6
0.036
3.17E-4
36

Initial fracture permeability, kfi, md

2.0E4

Correlation 3 constant, df3

Initial fracture porosity,!fi!

6.0E-5
0.08

Initial fracture compressibility, ctfi, psi-1

3.17E-4

Fracture thickness, hf, ft

Hydraulic fracture permeability, kF, md

5.0E-4
1.0E5

Hydraulic fracture porosity,!F!

0.38
-1

Hydraulic fracture compressibility, ctF, psi

Hydraulic fracture half-length, xF, ft

0.23
275

53.9E-2
100

Hydraulic fracture width, wF, ft

0.001

Hydraulic fracture flow rate, qF, Mscf/D

1,000

SPE 159801

15

Unlike the Haynesville well example shown in Fig. 10, the history match of the pressure responses of the Bakken well
shown in Fig. 12 does not indicate a significant effect of fracture-permeability loss on the well performance. Comparing the
match parameters in Tables 6 and 7, the difference between the two examples may be attributed to the different initial
pressures. As mentioned earlier, both formations were initially overpressured and, thus, the early portion of the data was not
used in history matching. Because the initial pressure was not known, it was estimated in the process of history matching.
The estimated initial pressure for the Barnett well is significantly lower than that for the Haynesville well and, as discussed
earlier (Fig. 8), the exponential nature of the pressure-dependent fracture permeability correlation predicts much larger
change in permeability if the initial permeability is taken at a higher initial pressure. We could proceed with similar lines of
discussion to justify the differences between the two examples; however, a more relevant discussion here is the large number
of matching parameters used in these examples.
In most real-life applications of shale-gas-well history matching, very little information is available about the formation
(by measurements or other independent means) to reduce the number of regression parameters. This exacerbates the
nonuniqueness of the results obtained by history matching. In the two examples presented above, we assumed similar
conditions and obtained a history match by using a large number of parameters in addition to the initial value of the naturalfracture permeability and the correlation constant df3. Therefore, if the nonuniqueness issues caused by the large number of
parameters used in history matching is ignored, examples similar to the ones given above may be used to derive different
conclusions above the effect of pressure-dependent fracture permeability on shale-gas well performances. To alleviate the
nonuniqueness problem, below we discuss a synthetic example where we fix all other properties and observe the effect of
pressure-dependent fracture permeability only.
Synthetic Example
The data used to generate the synthetic example by using the trilinear model is shown in Table 8. This example considers a
horizontal well with uniform spacing between hydraulic fractures. Due to the symmetry of the problem, only one fracture
with its associated stimulated reservoir volume is considered in the example (Apaydin, 2012). Pressure drop at the wellbore is
shown in Fig. 13 with pressure-dependent and constant natural-fracture permeability cases.
TABLE 8 Data used in the synthetic example
Specific gravity of gas, SG
0.55 Initial fracture permeability, kfi, md
Molecular weight of gas, M, lbm/lbm-mol
16 Correlation*3*constant,*df3*
Initial reservoir pressure, pi, psi
2300 Initial fracture porosity,* fi
o
-1
Reservoir temperature, T, F
109 Initial fracture compressibility, ctfi, psi
Formation thickness, h, ft
250 Fracture thickness, hf, ft
Wellbore radius, rw , ft
0.25 Number of fractures in pay thickness*
Reservoir size perpendicular to well axis, xe, ft
500 Hydraulic fracture porosity,*F
Half-distance between hydraulic fractures, ye, ft
250 Hydraulic fracture permeability, kF, md
Initial viscosity,*i, cp*
0.0184 Hydraulic fracture compressibility, ctF, psi-1
Constant matrix permeability, km, md
1.E-08 Hydraulic fracture half-length, xF, ft
Initial matrix porosity,*mi*
0.05 Hydraulic fracture width, wF, ft
-1
Initial matrix compressibility, ctmi, psi
9.E-04
Hydraulic fracture flow rate, qF, Mscf/D

2000
5.0E-4
0.45
9.E-04
0.001
20
0.38
1.E+05
9.E-04
250
0.01
94

Figure 13 Synthetic example to examine the effect of pressure-dependent natural-fracture permeability on shale-gas well
performance (Apaydin, 2012).

16

SPE 159801

Contrary to the expectations from the results presented in the previous sections, the effect of natural-fracture
permeability reduction is negligible in Fig. 13. This result can be explained by noting that the first upward bend of the
pressure responses at around 10-2 hr in Fig. 13 indicates the start of the depletion of the fracture network. The pressure drop
in the fractures at this time is around 2.2 psi, which does not cause noticeable change in fracture permeability. By 0.5 hr, the
fluid stored in the fracture network is mostly depleted and if the support of the matrix is slow, there is not much fluid to be
transmitted by the fractures. As a result, the reduced permeability is still sufficient to maintain the flow of fluids to the
wellbore. Apaydin (2012) has shown that if the matrix support is stronger (e.g., due to higher matrix permeability or
contribution of slip flow), then the pressure drop in the fracture network is slower, which, in turn, yields less reduction in
fracture permeability. Again, at late times, the pressure drop becomes large enough to considerably reduce the permeability;
but, at that time, there is not much fluid to flow in the fractures (due to the decline in the support of matrix blocks).
It must be emphasized that the conclusions drawn from Fig. 13 are a result of the data used for this synthetic case.
Different combinations of the matrix and fracture properties and well-hydraulic fracture configurations may yield different
pressure decline trends, which might cause more noticeable effect of pressure-dependent fracture permeability. The results
presented in this paper emphasize the importance of evaluating the effect of pressure-dependent natural-fracture permeability
in the context of the interactions of the matrix and fractures.
Conclusions
In this paper, the effect of pressure-dependent natural-fracture permeability on the performances of shale-gas wells has been
discussed. Pressure-dependent matrix-permeability correlations from the literature have been delineated and screened for
applications to natural-fracture permeability. Experimental data have been presented for use in the verification and calibration
of the existing pressure-dependent permeability correlations. Field examples, as well as synthetic data, have been used to
demonstrate the applications of the correlations. The results of this work indicate that unpropped natural fractures lose a
significant portion of their initial permeability under pressure depletion. However, the data demonstrating fracture closure
with pressure drop should not be used without considering the complex interactions between the natural fractures and the
tight shale matrix. In general, very small fracture permeabilities create an infinite-conductivity fracture effect when compared
with the nano-Darcy shale-matrix permeabilities. Therefore, the permeability which is retained when the fractures close may
still be sufficient to transmit the limited volume of fluid available to flow.
Acknowledgement
This work has been conducted under Marathon Center of Excellence for Reservoir Studies at Colorado School of Mines.
Portions of the work have been conducted to fulfill the requirements for the MSc degree of Younki Cho and the PhD degree
of Osman G. Apaydin. The support of EOG Resources, Inc. for the PhD study of Osman G. Apaydin is gratefully
acknowledged.
Nomenclature
a
Characteristic parameter of rock [psi-1]
bi
Initial fracture width [m]
bmax
Maximum fracture width [m]
c
Compressibility [psi-1]
c1
Characteristic parameter of rock [dimensionless]
d
Characteristic parameter of rock [psi-1]
E
Elasticity
F
Parameter used in Correlation 7 [-]
h
Thickness [ft]
k
Permeability [md]
m
Pseudopressure [psi2/cp]
mf
Characteristic parameter of rock [psi-1]
M
Molecular weight [lbm/lbm-mol]
p
Pressure [psi]
q
Production rate [Mscf/D]
Q
Cumulative production [Mscf]
r
Radial distance [ft]
T
Reservoir temperature [oR]
t
Time [D]
v
Volumetric velocity [ft/D]
v
Poissons ratio [-]
x
Length in direction parallel to horizontal well axis [ft]
y
Length in direction perpendicular to horizontal well axis [ft]

SPE 159801

Greek

17

Biot number [-]

Difference operator
Density [lbm/ft3]
Effective stress [psi]
Volumetric strain
Porosity [fraction]
Viscosity [cp]

Subscripts
D
Dimensionless
f
Natural fracture
F
Hydraulic fracture
g
Gas
h
Horizontal
i
Initial
m
Matrix
sc
Standard condition
t
Total
References
Alamdari, B. 2011 Viability of Matrix-Fracture Transfer Functions for dilute Surfactant-Augmented Waterflooding in Fractured Carbonate
Reservoirs. PhD dissertation. Colorado School of Mines, Golden, CO.
Al-Hussainy, R., Ramey, H. J., and Crawford, P. B. 1966. The Flow of Real Gases through Porous Media, JPT 18 (5): 624-636. SPE 1243A.
Anderson, D. M., Nobakht M., Moghadam S., and Mattar L. 2010. Analysis of Production Data from Fractured Shale Gas Wells, SPE
131787, SPE Unconventional Gas Conference, February 23-25, Pittsburgh, PA.
Apaydin, O. G. 2012. New Coupling Considerations Between Matrix and Multiscale Natural Fractures In Unconventional Resource
Reservoirs, PhD dissertation. Colorado School of Mines, Golden, CO.
Baghbanan, A., and Jing, L. 2008. Stress effects on permeability in a fractured rock mass with correlated fracture length and aperture, Int J
Rock Mech Min Sci (45): 1320-1334.
Barton, N., Bandis, S., and Bakhtar, K. 1985. Strength, Deformation and Conductivity Coupling of Rock Joints, Int J Rock Mech Min Sci
(22): 121-140
Best, M.E., and Katsube , T.J. 1995. Shale permeability and its significance in hydrocarbon exploration, Geological Survey of Canada
14(3): 165-170
Brown, M., Ozkan, E., Raghavan, R., and Kazemi, H. 2011. Practical Solutions for Pressure Transient Responses of Fractured Horizontal
Wells in Unconventional Shale Reservoirs, SPEREE, 14 (6): DOI 10.2118/125043-PA
Celis, V., Guerra, J., and Prat, G. Da, 1994, A New Model for Pressure Transient Analysis in Stress Sensitive Naturally Fractured
Reservoirs, SPE Advanced Technology Series 2 (1): 126-135.
Chin, L. Y., Raghavan, R., and Thomas, L. K. 2000. Fully Coupled Analysis of Well Responses in Stress-Sensitive Reservoirs, Paper SPE
66222 presented at SPE Annual Technical Conference and Exhibition, 27-30 September, New Orleans, LA.
Chin, L. Y., Raghavan, R., and Thomas, L. K. 2000. Fully Coupled Geomechanics and Fluid-Flow Analysis of Wells With StressDependent Permeability, Paper SPE 58968 presented SPE International Conference and Exhibition, 2-6 November, Beijing, China.
Core Labs. 2012. http://www.corelab.com/rd/instruments/routine/routine.aspx?ename=cms300&mnuSect=routMeas, accessed on June 10,
2012.
Franquet, M., Ibrahim, M., Wattenbarger, R.A., and Maggard, J.B. 2004. Effect of Pressure-Dependent Permeability in Tight Gas
Reservoirs, Transient Radial Flow. Paper SPE 2004-089 presented Canadian International Petroleum Conference, 1-8 June, Calgary,
Alberta.
Gutierrez, M., ino, L. E., and Nygrd, R. 2000. Stress-Dependent Permeability of a De-Mineralised Fracture in Shale, Marine and
Petroleum Geology 17, 895-907.
Kwon, O., Kronenberg, A. K., Gangi, A.F., Johnson, B., and Herbert, B.E. 2004. Permeability of illite-bearing shale: 1. Anisotropy and
effects of clay content and loading. Journal of Geophysical Research (109): B10205
Min, K., Rutqvist, J., Tsang, C., and Lanru, J. 2004. Stress-dependent permeability of fractured rock masses: a numerical study, Int J Rock
Mech Min Sci (41): 1191-1210.
Minkoff, S.E., Stone, C.M., Bryant, S., Peszynska, M., Wheeler, M.F. 2003. Coupled fluid flow and geomechanical deformation modeling,
Int J Rock Mech Min Sci (38): 37-56
Ozkan, E., Brown, M., Raghavan, R., and Kazemi, H. 2011. Comparison of Fractured Horizontal-Well Performance in Conventional and
Unconventional Reservoirs, SPEREE (April 2011)
Ozkan, E., Raghavan, R., and Apaydin, O. G. 2010. Modeling of Fluid Transfer from Shale Matrix to Fracture Network, Paper SPE 134830
presented SPE Annual Technical Conference and Exhibition, 19-22 September, Florence, Italy.
Pedrosa, Jr. 1986. Pressure Transient Response in Stress-Sensitive Formatoins, Paper SPE 15115 presented 56th California Regional
Meeting of SPE, 2-4 April, Oakland, CA.
Raghavan, R., and Chin, L. Y. 2004. Productivity Changes in Reservoirs With Stress-Dependent Permeability, Paper SPE 88870 presented
SPE Annual Technical Conference and Exhibition, 29 September-2 October, San Antonio, TX.

18

SPE 159801

Rutqvist, J., Wu, Y.S., Tsang, C.F., Bodvarsson, G. 2002. A modeling approach for analysis of coupled multiphase fluid flow, heat transfer,
and deformation in fractured porous rock, Int J Rock Mech Min Sci (39): 429-442
Tao, Q., Ghassemi, A., and Ehlig-Economides, C.A. 2010. Pressure Transient Behavior for Stress-Dependent Fracture Permeability in
Naturally Fractured Reservoirs. Paper SPE 131666 presented CPS/SPE International Oil and Gas Conference and Exhibition, 8-10
June, Beijing, China.
Wang, F.P., Reed, R.M., A., John, and G., Katherine. 2009. Pore Networks and Fluid Flow in Gas Shales, Paper SPE 124253 presented
SPE Annual Technical Conference and Exhibition, 4-7 Octoboer, New Orleans, LA.
Wang, J., Liu, Y. 2011. Simulation Based Well Performance Modeling in Haynesville Shale Reservoir. SPE 142740, SPE Production and
Operations Symposium, March 27-29 2011, Oklahama City, OK
Xiangjiao, X., Hedong, S., and JianPing, Y. 2009. Dynamic Characteristic Evaluation Methods of Stress Sensitive Abnormal High Pressure
Gas Reservoir, Paper SPE 124415 presented SPE Annual Technical Conference and Exhibition, 4-7 October, New Orleans, LA.