Você está na página 1de 11

Stress-Dependent Permeability

and Porosity of Coal and Other


Geologic Formations
C.R. McKee, SPE, In-Situ Inc.

A.C. Bumb, In-Situ Inc.


R.A. Koenig, In-Situ Inc.

Summary. Fundamental relationships have been derived for permeability, porosity, and density as a function of effective stress.
The primary formation parameter is pore compressibility, and formulas are derived for both constant and variaWe pore
compressibility. The grains are assumed to be incompressible. The Carman-Kozeny equation is assumed valid, and changes in total
porosity are proportional to changes in porosity effective stress.
These relationships fit laboratory and field data. Eight laboratory core tests (four coal, one granite, two sandstone, and one clay)
are used to illustrate the applicability of the permeability and porosity theory. Additional clay and shale core samples yielded an
excellent match with the theoretical density formula. In cases in which the correlation coefficient was applicable, it was above
0.95. A number of field permeability data points from the Piceance, San Juan, and Black Warrior basins were fitted to the theory.
The correlation coefficient ranged from 0.85 to 0.92 in these cases. Mudstone porosity data from borehole lithologic logs also
yielded excellent fits to the theoretical formulation. Of the 13 examples reviewed, only 2 (which were core tests) required a
variable pore compressibility.
Further improvements to the derived relationships will be required to include the effects of structurally enhanced permeability.
Introduction
Economical coalbed methane production depends on four important coal-seam characteristics-gas pressure, gas content, coal-seam
thickness, and permeability. (Concerning gas pressure, note that
most coal seams are initially water-saturated before production. If
gas is present in solution, formation gas pressure is defined as the
pressure that is in equilibrium with the measured gas content in
solution. If free gas is present, then the gas pressure is pressure
of the gas phase. If gas is absent, gas pressure is determined by
drawing the well down until gas first appears in solution. The bottomhole water pressure on the coal seam when gas first appears
in the coal seam is equal to the formation gas pressure. In either
case, the gas pressure so determined must be in equilibrium with
the coal's desorption isotherm.) Without the necessary gas content
or coal thickness, a project obviously cannot be viable. An extensive body 'of information exists for selecting a site with some degree of confidence based on gas contents and coal thickness. 1-6
However, unexpected variations in coal gas content can occur.
Information on coal permeability has been more difficult to obtain. For most coalbed methane wells, permeability tests either have
not been conducted or, if they have, the results remain proprietary. Consequently, very few permeability values have been reported
in the literature. 7 In 1982, the Gas Research Inst. (GRI) began to
pursue extensive permeability testing and data collecting to assess
the permeabilities of coalbed methane sites. A program was developed to acquire the necessary information from operators and from
"wells of opportunity" using simple but effective slug-testing
methods. 8 Far more information was available in the form of
specific capacity data, which relate well flow rate to drawdown;
permeability can be estimated from these data by using available
information or by assuming depths to water. In this manner more
than 40 permeability data points were calculated for unstimulated
wells in the San Juan, Piceance, and Black Warrior basins. 7
While some scatter was evident, the data indicated that permeability decreased with depth. Some initial success was achieved by
drawing empirical straight-line correlations between depth and
permeability oflog-Iog plots, and led to the search for a more fundamental theory to explain the data and correlations. 7
In view of the advances in reservoir engineering research, it is
surprising that stress-dependent permeability has not been more exCopyright 1988 Society of Petroleum Engineers

SPE Formation Evaluation, March 1988

tensively studied. As natural gas reservoirs are found at ever greater


depths, understanding stress-dependent permeability will become
essential because, under large drawdown, reduced permeability can
lower the production from a stress-sensitive reservoir.
Work at the Inst. of Gas Technology (IGT) documented the strong
stress dependence of core-sample permeability in laboratory experiments, and was the basis for establishing a theoretical relationship of permeability reduction as a function of increasing stress or
depth. These and other laboratory studies have shown that coal
permeability decreases by one to two orders of magnitude when
effective stress is increased by 2,000 psi [13.8 MPa]. 9-11 Furthermore, stress-sensitive permeability and porosity have been observed
in experimental studies on various reservoirs. 12,13 As a result, a
number of empirical permeability/stress relationships have been proposed, most of which show a strong decrease in permeability with
depth. 12,14,15

Objectives
These efforts laid the foundation for the theoretical relationship between stress-dependent permeability and porosity for coals. McKee
et at. developed the first theoretical equations using matrix compressibility as a fundamental property. 16 The theory of McKee et
at. fits both laboratory and field data, and is of limited use because
maximum closure stress, a function of matrix compressibility, is
inherent and limits the depth to which the results apply. 16 New
theoretical equations are developed here for permeability and porosity in terms of pore compressibility. This formulation generally
eliminates the need for variable compressibility in the range of interest, and therefore is simpler to use and not limited by maximum
stress. The resultant formulas are shown to fit both laboratory and
field data, but their practical significance is that they provide a simple, usable tool to establish favorable areas of permeability for gas
production from coalbeds or other engineering applications where
permeability is a critical parameter. However, the theory and correlations that use the effective stress gradient can only indirectly account for structurally enhanced or reduced permeability; hence, the
role of the geologist is still paramount in selecting those geologic
factors leading to the best production prospects.
While the work in this paper was motivated by GRI's Coalbed
Methane Program, the theory is applicable to a wide range of geo81

10.----------------------------------,
100.------------------------------------,

p;: 310 psia


ko

=1.56 md at 0 =150psi
0

= 1.87 x 10 3/psi
r = 0.96

ko = 35.8 }J.d at Go = 250 psi


cp = 1.34 X 10'3/psi
r = 0.99

cp

1L-__

0.1L-____L-__~L-__~____~____~____~

100

400
200
300
Effective Stress (psi)

500

600

Fig. 1-Laboratory-measured permeability vs. effective stress


data and match with theoretical curve for Pittsburgh coal mine
sample from depth of 300 ft, Northern Appalachian basin.

logic material ranging from sand to fractured granite. This is substantiated by considerable experimental evidence.

Theory
Appendix A details the mathematics of the stress- or depth-dependent
permeability derivations; major results are given in Eqs. A-6, A-9,
A -10, and A -15. Stress-dependent pore compressibility, which is
required for some cores, is given in Eq. 1. The mathematically
oriented reader may find it convenient to review these equations
before proceeding to the discussion of results. Conceptually, the
theory is very simple. It assumes that the solid grains are incompressible and that the change in porosity is caused by compression
of the void space, which is assumed to depend only on effective
stress. Permeability is obtained by substituting the expression for
porosity into the Carman-Kozeny equation for fractures. The tacit
assumption is that the effective porosity is proportional to total
porosity. The variation of density with effective stress is a result
of the porosity relation.

Comparisons of Theory With Laboratory Data


The literature does not contain as much field data as laboratory data
on stress-dependent porosity and permeability for coal or other geologic materials. Hence, the theory first will be compared with coretest permeability data from five basins. Rose and Foh II performed
four tests on coal, Krantz et al. 17 performed experiments on granite, and Jones and Owens 18 conducted a test on sandstone. Net effective stresses ranged from 150 to 6,000 psi [1.03 to 41.4 MPa].
The theory will then be compared with core test data on porosity
(void ratio) for sand and clay from Venezuelan oil fields. 19,20

Permeability Data. Constant Pore Compressibility. In reviewing


the data from the first four samples, constant pore compressibility
fit the permeability/effective-stress data very well. The first sample shown in Fig. 1 involved laboratory permeability vs. effective
stress for a Pittsburgh seam coal sample obtained from a depth of
300 ft [91.4 m] in the U.S. Bureau of Mines test mine in the Appalachian basin. II The tabular data were reported in millidarcies
to one decimal point and ranged from 0.2 to 1.8 md. While this
does not create a problem with linear plots, on log plots it can show
as scatter, particularly at low permeability ranges. To report all
data as two significant digits, a range is shown with the data reported
to only one significant digit. Fig. 1 shows the data on a semilog
82

200

____

300

____-L____

400

____L __ _

700
600
500
Effective Stress (psi)

_ _ _ __

800

900

Fig. 2-Laboratory-measured permeability vs. effective stress


data at constant fluid pressure and match with theoretical
curve for Menefee formation (Mesaverde group) coal outcrop
sample, San Juan basin, CO.

plot with a least-squares straight line through the data. The correlation coefficient is 0.96, which is indicative of a good fit. The leastsquares resulted in a constant pore compressibility of 1.87 x 10 - 3
psi -I [0.271 MPa -I], and a permeability of 1.56 md at 150 psi
[1.03 MPa].
The second sample, in Fig. 2, used laboratory permeability vs.
effective stress for a coal sample from the top seam of the Menefee
formation in the King mine, San Juan basin. II For this core test,
fluid pressure was kept at approximately 310 psi [2.14 MPa], while
confining pressure was varied. As shown in Fig. 2, permeability
data and theoretical values obtained from least-squares matching
agreed quite closely for a constant pore compressibility of
1.34 x 10 - 3 psi - I [0.194 MPa - I ], and an initial permeability of
35.8ttd, at an arbitrarily selected initial effective stress of 250 psi
[1.72 MPa]. The permeability of this sample is more than two orders of magnitude smaller than that of the sample in Fig. 1, and
an excellent curve fit was obtained by assuming constant pore compressibility. A correlation coefficient of 0.99 justifies constant pore
compressibility.
The third sample, in Fig. 3, used laboratory permeability data
for a Cameo seam coal sample from a depth of 2,766 ft [843 m]
in Coors Well 1-23, Piceance basin, CO. 11 Effective stress was
obtained by subtracting fluid pressure from confining pressure (Eq.
A-12, with s = 1). Data were obtained by holding confining pressure approximately constant at 2,615 psi [18 MPa]. Almost all the
data points fall on a straight line, justifying use of constant pore
compressibility (Fig. 3). The least-squares match resulted in a pore
compressibility of 7.76 x 10 -4 psi -1 [0.112 MPa -1] and an initial permeability, ko' of 273 nd at an arbitrarily selected initial effective stress of 1,000 psi [6.89 MPa). Again, the correlation
coefficient of 1.00 and excellent fit to data (Fig. 3) justify constant
pore compressibility.
The fourth example comes from a laboratory permeability study
done by Krantz et al. on whole, unjointed Barre granite. I? The
data and the theoretical fit to data by the least-squares method are
shown in Fig. 4. The open symbol and the closed symbols directly
beneath it represent beginning points and endpoints, respectively,
of cyclic experiments. The fit to the theoretical relationship resulted in a correlation coefficient of 0.95. The res!1ltant pore compressibility is 6.08x 10- 5 psi -1 [8.82 x 10- 3 MPa -1). While the first
three samples were coal samples, this sample was granite. As expected, granite is less compressible and pore compressibility is
SPE Formation Evaluation, March 1988

10~-------------------------------------'

pc(bars)
170-175
o. 490
1000
1450

..

1000,-------------------------------------,
pc=2615 psi

".1

ko = 2.0 md at 0 0

ko

=273 IJd at 0 =1000 psi


0

Cp= 7.76 x 10 /psi

100

= 25 bars

cp = 8.a2x104 /bar

1.0

= 6.0a x10-s/psi
= 0.95

!.

=1.00

:c

e'"

Q;

:c

e'"

0.1,...

Q;

~
=

10 c-

I"
'"

l
I
1L-__

__

____

1000 1200 1400

__- L_ _ _ _L __ _

1600 1800 2000


Effective Stress (psi)

.01 '
0

~_ _ ~_ _ ~

2200

2400 2600

200

400

600
800
1000
Effective Stress (bars)

1200

1400

Fig. 3-laboratory-measured permeability vs. effective stress


data and match with theoretical curves for Cameo coal core
sample from depth of 2,766 ft, Piceance basin, CO.

Fig. 4-laboratory-measured permeability vs. effective stress


data and theoretical match to data for whole, unjointed Barre
granite.

roughly one order of magnitude lower than for coal samples. Table 1 compares the pore compressibilities of various samples.
Variable Compressibility. In the preceding section, four cases
were shown where constant pore compressibility was a good assumption for the range of data in each case. However, Somerton
et al. 10 and Scorer and Miller 19 have shown that compressibility
is unlikely to remain constant over the entire stress history. Instead
it will approach the rock (or grain) compressibility as effective stress
increases, because small asperities on the cleat face resist closure
and must be deformed or crushed before further reduction in porosity
is achieved. The result is a departure from linearity (constant pore

compressibility). For the first four core samples, compressibility


can essentially be assumed constant, which implies that the material is initially relatively consolidated. A laboratory study of the
relationship between relative compressibility and effective stress
may be useful in understanding this phenomenon, provided that a
size effect (small size oflaboratory core vs. large size of field core)
is incorporated. The reason for doing so will be discussed in the
following sections.
As a general rule, compressibility declines with increasing stress,
although on some samples it may increase. 19 Because a stressdependent compressibility can be sample-specific, a general func-

TABLE 1-PORE COMPRESSIBILITIES OBTAINED BY FITS TO EXPERIMENTAL


PERMEABILITYIPOROSITY AND EFFECTIVE STRESS DATA

Basin
Appalachian
San Juan
Piceance
Piceance

Rocky Mountain
region
Bolivar coast
Venezuelan
oil field
Bolivar coast
Venezuelan
oil field

Data Type

Pore
Compressibility
(psi -1)

Coal

Permeability

1.87x10- 3

Coal
Coal

Permeability
Permeability

1.34x10- 3
7.76 x 10- 4

Coal

Permeability
Permeability

6.08 x 10- 5

Frontier

Whole, unjointed
Barre granite
Sandstone

Permeability

2.31 x10- 4 -

Post-Eocene

Sand

Porosity

1.47x10- 4

Post-Eocene

Clay

Porosity

1.44x10- 4

Lithologic
Type

Formation
Pittsburgh
seam
Menefee
Cameo
(2,766 ft)
Cameo
(2,767 ft)

1.8x 10- 3 -

'Initial pore compressibility shown for variable cases.

SPE Formation Evaluation, March 1988

83

1.0 r----------------------------,

100,-----------------------------------------,

0.9

pc = variable 185 to 2751 psi

0.8
0.7
Sand (3100 tt)

<Po/l-<Po= 0.53 at zero stress


<Po = 34.6%
cp = 1.44 x l0-4 /psi

0.6
Cp= n::J (l-e- uAfJ )
Co

l'/

0.5

= 1.8xl0~'/psi
= 2.95 xl0'/psi

= 0.99

-.J.

cPo <: 10%

::: 0.4

-e-

= 40nd
(Jo = 500 psi

ko

g
III

c::

'------L _ _ ~l_

500

600

700

800

900

1000

1100

1200

"tl

1300

Effective Stress (psi)

Fig. 5-Laboratory-measured permeability vs. effective stress


data and theoretical match using variable compressibility for
a Cameo coal sample from a depth of 2,767 ft, Piceance basin, CO.

0.3

>

0.2

Clay (3104 tt)

1.0,----..------------------,

CPo/l-<po = 0.23 at zero stress

<Po = 18.7%
cp= 1.47 x l0- 4 /psi

r = 1.00_______
0.1L-______
~

o
-

Co

Cp= a60 (l-e- a 0)

~ 0.1

Co

___

1000

____

____

____

2000
3000
4000
Effective Stress (psi)

~_~

5000

6000

Fig. 6-Laboratory-measured permeability vs. net confining


stress data and theoretical match using a variable compressibility for a Frontier formation sandstone core sample.

tion is unavailable. The theoretical curves fit the laboratory values


well when these samples were assumed to have exponentially declining compressibilities. Note, however, that because compressibility is variable, average pore compressibility (cp in Eq. A-5) also
is variable.
For exponentially declining pore compressibility, cp is given by

where a defines the rate of decline of pore compressibility as effective stress increases, Co is the pore compressibility at initial effective stress, uo ' and ~u is the increase in effective stress from
uo. For variable pore compressibility, Cp is not constant but simply represents the average pore compressibility over the stress in84

______

3000
1500
Effective Stress. a (psi)

4500

Fig. 7-Laboratory-measured void ratio vs. effective stress


data and theoretical match for sand (upper curve) and clay
(lower curve) core samples from a Venezuelan oil field.

=2.31 xl0- Ipsi

a = 3.5 x l0-4 /psi


<Po =5-10% at 0 0 =1000 psi

lL-__

terval (u-u o). If u changes, the average cp over the interval will
also change. Core samples from the Piceance basin and Frontier
formation show exponentially declining pore compressibility.
Laboratory permeability data for the Piceance basin Cameo seam
coal from a depth of 2,767 ft [843.3 m] are shown in Fig. 5. 11
Using the parameter a=2.9 x 10- 3 psi -I [0.4 MPa -I] in Eq. I,
the solid line in Fig. 5 yields an excellent fit to the Piceance coal
sample data. While the sample in Fig. 5 was taken only I ft [0.3
m] lower than the sample used in Fig. 3, the samples show different characteristics. This indicates that core samples, especially in
fractured media, are not of sufficient size to constitute a relative
elementary volume. In Fig. 3, constant pore compressibility gave
an excellent approximation; however, the data in Fig. 5 require
variable pore compressibility. This may be a result of different sample orientations in the test equipment (the plug axis was perpendicular to the bedding planes for Fig. 3, while in Fig. 5 the plug axis
was perpendicular to the face cleat) or just the local nature of the
core and the effect of small sample size. Another reason for variable compressibility for the sample in Fig. 5 may be that this sample was never stressed above the in-situ stress conditions, and both
confining stress and fluid pressure were varied. II The net effective stress was calculated with s=I in Eq. A-l2, where this assumption may not be true.
A final example will illustrate the general applicability of the theory to laboratory permeability data. Permeability values for a Frontier formation sandstone sample from the Rocky Mountain region
are plotted in Fig. 6 and scaled with respect to permeability measured at I,ODO-psi [6.9-MPa] initial confining pressure. 18 Confining pressure changes more than six-fold, and it was again necessary
to use exponentially varying pore compressibility. The resultant fit
is excellent, as before. The theoretical curve can be obtained with
an initial porosity value from 5 to 10%.
SPE Formation Evaluation, March 1988

2.7
.:

2.6

>'iii
c
CI)

.:

2.5
2.4

:; 2.3
u
CI)

Co

(/)

P g 2.65
cp= 6.03 x 10-4 /ft of overburden

2.2
2.1
2.0
0

1000

2000

3000

4000

5000

Depth (ft)
Fig. 8-laboratory-measured density vs. depth data and theoretical match for shale core samples from northern Oklahoma.

Although only a few cases have been included here, this theory
.fits other permeability data. The only other way to verify the theory is to analyze porosity data; unfortunately, porosity measurements as a function of effective stress are limited. As will be
discussed in the following section, however, such additional data
are useful in verifying the theoretical soundness of these concepts.

1000
o

k
cp
r

100

Porosity Data. In this section, limited porosity data (principally


results of consolidation tests for sand and clay samples) are shown
to fit the theoretical equation for porosity (Eq. A-6) very well.
Scorer and Miller l9 reviewed limited data on the relationship of
effective stress to void ratio, which relates to porosity. From Eq.
A-6, a semilog plot of void ratio and effective stress will result in
a straight line for constant pore compressibility. As an example,
void-ratio-reduction data for adjacent sand and clay samples from
about 3, l00-ft [945-m] depth in a post-Eocene formation in the Bolivar coast oil fields, Venezuela, have been plotted in Fig. 7. 19,20
As shown, void ratio [q,/(I-q,)] for sand decreased from about 0.59
to 0.30 (or porosity decreased from 38 to 22 %) when effective stress
increased to about 4,400 psi [30.3 MPa]. Fig. 7 shows that the theoretical curve matched the sand void ratio, and thus the porosity,
data very well with constant pore compressibility. Note that the
first two data points were not included in the least-squares fit to
data because these two points may represent the sample in its unconsolidated state. The least-squares match results in 1.44 x 10- 4
psi -I [2.09 x 10- 2 MPa- l ] for pore compressibility and 34.65%
porosity (void ratio of 0.53) at zero stress. The correlation coefficient was 0.99. These values fall within the range of compressibilities described elsewhere. 19
Fig. 7 also shows the match of theoretical values and void ratio
data for the clay sample, which was taken within 4 ft [1.2 m] of
the sand. The theoretical curve fits the data quite well except at
very low effective stresses, such as the first two data points where
the sample may be unconsolidated. The least-squares fit to the rest
of the data results in a pore compressibility of 1.47 x 10 -4 psi -I
[2.1 x 10 - 2 MPa - 1], which is only 2 % higher than sand pore
compressibility. The clay sample is highly compressed and, as a
result, has stiffened to the point where its compressibility is similar to that of sand.

Piceance Basin

o San Juan Basin

=234 md at 100 ft
=9.24 l0- /psi
=0.92
X

10
~

..

:cIII

CI)

CI)
~

0.1

f
lI

0.01
0.001

'--_...l-_-!..._--'-_ _"---_--'-...-_-'-_---l

1000 2000 3000 4000 5000 60007000


Equivalent Lithostatic Depth (ft)
1000
2000
3000
Effective Stress (psi)

4000

Density Data. Athy21 has reported a general porosity relationship


with depth that is exponential and in agreement with the limiting
case of our theory. He used 200 samples to construct the exponential curve; however, no porosity data were given. Nevertheless,
density data were collected on more than 400 core samples (Fig.
8). The samples come from a small area around Garber, OK, within which approximately the same amounts of sediment were deposited with no intervening unconformity. The resultant pore
SPE Formation Evaluation, March 1988

Fig. 9-Semllog plot of fleld-test-derived and estimated


permeabilities with equivalent IIthostatic depth (and effective
stress) and match with theoretical curve for coal seams in
the Piceance and San Juan basins.

85

1000

1000

Warrior Basin
k = 186 md at 100 ft
cp= 1.87 X 10-3 /psi
r = 0.87

100

-"

10

E 10
>-

...

:c
(II

CI)

E
...
CI)

1.0
o. Piceance Basin
c

0.1 ...
j:
~

oj

A.o.

San Juan Basin


Warrior Basin

cp = 8.78 x l0 4 /psi
k 109 md at 100 ft
r = 0.85

0.01

'--_~_--I._ _....L..-_--L..._ _l..-I

400 800 1200 1600 2000


Equivalent Lithostatic Depth (ft)
I

200 400 600 800 1000 1200


Effective Stress (psi)

Fig. 10-Semllog plot of fleld-test-derlved and estimated


permeabilltles with equivalent lithostatlc depth (and effective
stress) and match with theoretical curve for coal seams in
the Black Warrior basin.

compressibility in per unit depth of overburden is 6.03 x 10- 4 ft - 1


[1.84 x 10 -4 m -1]. Specific grain density was 2.65, while specific
density at zero depth is 2.02 g/cm.3 Fig. 8 shows a good match
of theoretical curve to the data.

Comparison of Theory With Field Data


This section compares the theory of stress-dependent permeability
with field permeability and porosity measurements. The search for
such field permeability data targeted the three most active coalbed
methane areas-Piceance basin in Colorado, San Juan basin in New
Mexico and Colorado, and Black Warrior basin in Alabama. In a
recent GRI report, Way et al. 7 reported the information acquired
. from 53 contacts with private companies and state and federal agencies involved in coalbed degasification, and geologic and hydrologic investigations. The only field porosity data we could obtain
were from induction logs from Nagaoka Plain, Japan.
Lithostatic Conditions. Overburden materials have an average
specific grain density of approximately 2.65 and a porosity of about
20%, which gives an estimated stress gradient, gs' of 1.005 psi/ft
[22.73 kPa/mJ, derived by
gs=[Pg(l-cf+Pwcf>]

=[2.65(1-0.2) + 1(0.2)](62.4 Ibm/ft3)(1


=1.005 psi/ft [22.73 kPa/m].
86

ft 2 1l44

in. 2 )

E
I

o.oolf~~~~~"~1 ~~~~~I ~~~~


__

10

..

__

1000
10,000
. Equivalent Lithostatic Depth (tt)

100

Fig. 11-Log-log plot of field-test-derived and estimated


permeabilitles with equivalent lithostatic depth (and effective
stress) and match with theoretical curve for combined data
from coal seams in the Piceance, San Juan, and Black Warrior basins.

Note that .the commonly used value for this gradient is I psi/ft [22.63
kPaim] and that for fresh water it is 0.4335 psi/ft [9.81 kPa/m].
For those coal seams that are at or near hydrostatic pressure, the
effective stress gradient is 0.572 psi/ft [12.9 kPa/m] (i.e.,
1.0005 -0.4335), in which case the effective stress (in psi) at any
depth, D (in feet), for lithostatic conditions is given simply by

u=0.572D . ....................................... (2)


For some data, information about effective stress and/or water levels
was available. This information was used to calculate equivalent
lithostatic depth, the depth at which effective stress would equal
the effective stress at lithostatic conditions (Eq. 2). When information on effective stress/water level was lacking, it was assumed
that lithostatic conditions existed. If the average effective stress gradients for these points differed from 0.572 psi/ft [12.9 kPalm], compressibilities obtained by fitting Eq. A-9 to the data would change
inversely with the stress gradient. For example, with other
parameters held constant, a lower actual in-situ effective stress gradient than the lithostatic assumption of Eq. 2 would result in a higher
in-situ permeability at a given depth.
Permeability Data. Hydrologic data were sparse and few data
points were obtained from carefully conducted well tests. In most
cases, permeabilities were estimated from static water levels, water productions, and productivity index data. (As pressure in the
SPE Formation Evaluation, March 1988

TABLE 2-COMPARISON OF CURVE-FITTING PARAMETERS FOR COAL SEAMS IN THE


SAN JUAN, PICEANCE, AND BLACK WARRIOR BASINS
Initial Conditions
Basin
Piceance and
San Juan
(Fig. 8)
Black Warrior
(Fig. 9)
Piceance, San Juan,
and Black Warrior
(Fig. 10)

100

Permeability
(md)
234

Pore
Compressibility
(psi-')
9.24 x 10- 4

Coefficient of
Correlation
0.92

100

186

1.87 x 10- 3

0.87

100

109

8.78x10- 4

0.85

Depth
(ft)

reservoir is lowered, gas begins desorbing until two-phase flow occurs. To avoid relative permeability effects typical of this flow regime, only the initial production phases were analyzed.)
For consistency, only data from unstimulated wells in coal seams
have been used because hydraulically stimulated wells may yield
larger, possibly scale-dependent, permeability values, which have
been observed in fractured media by other investigators. 22 Fractured media, such as coal, exhibit small fractures (at spacings of
centimeters) between larger fractures at spacings of several to tens
of meters, which in turn are fault-bounded at intervals of hundreds
or even thousands of meters. A localized permeability measurement in an unstimulated well would have a probability of intersecting
only the more numerous small-scale fractures. Large hydraulic fractures will have a higher probability of intersecting the larger-spaced,
larger-scale natural fractures. Because permeability varies with the
square of fracture aperture, permeability can increase as larger natural fractures are encountered. Therefore, with a sufficiently large
hydraulic fracture, one can expect to attain a constant permeability
value consistent with the intersection frequency of the main fractures in the coal. Interference tests conducted at GRI's Rock Creek
site with wells spaced from 40 to 200 ft [12 to 61 m] have shown
from 0 to 75 % increase in permeability over single-well tests. 23
Hydrologic data by nature contain experimental errors. Furthermore, in low-permeability formations, several tests did not measure beyond the wellbore storage region, and hence, only upper
bounds on permeability could be inferred. The limits are indicated
in Figs. 9 through 11 by data points at the end of a range joined
by a line. Open symbols indicate assumed lithostatic conditions
where water level and/or effective stress information was unavailable. Closed (filled) symbols represent data for which waterlevel/effective-stress information was available and used to calculate the equivalent lithostatic depth. The initial curve-fitting
parameters (depth, permeability) and calculated pore compressibilities for the three basins are summarized in Table 2.
McKee et at. noted that permeabilities of the San Juan and Piceance basin coals showed similar trends and compressibility
values. 16 The agreement between permeability values for the San
Juan and Piceance basins could possibly be attributed to their similarities in geologic age and regional depositional settings. Also, because of the limited number of data points for these basins, they
are plotted on the same figure (Fig. 9), which shows that permeabilities from the San Juan and Piceance basins follow the same general trend. With depth (stress)-dependent compressibility, the rate of
change generally decreases with increasing depth; however, because
of scarcity of data and lack of information on depth dependence
of compressibility, a constant value for pore compressibility is used
for the San Juan and Piceance basins. Moreover, the porosity of
these coals is less than 10%, so Eq. A-1O is used. Fig. 9 shows
a good data fit to theory. If compressibility is varied, one could
expect a better fit to the data-Le., permeability decreasing more
gradually with increasing depth and eventually becoming constant.
Eq. A-9 accounts for this effect by incorporating the function for
exponentially declining pore compressibility in Eq. A-5.
Fig. 10 shows similar comparisons of theory and permeability
estimates for the Black Warrior basin, excluding permeability data
at known anomalies. The anomaly in the Black Warrior basin
represents a highly fractured well on the campus of the U. of Alabama; high permeability (250 md) has made this well difficult to
SPE Formation Evaluation, March 1988

1000~--------------------------~

k o = 109 md

cPo =10%
00

=100 ft

100

10

Variable pore
compressibility

S'

Co

Cp= - - (l-e- a .... O )

a~a

Co

1r
0.1 E'

0.Q1

= 4.33 x l0- 3 /psi

a =1.0xl0-3 /psi

Constant pore
compressibility
(8.78 x 10-4 /psi)

=-

0.001t~~I~'~'~'~'~"~'!~~~'~I~'~I~"~.!~~'~'~'~1'~'
10

100

1000

10,000

Equivalent Lithostatic Depth (ft)


Fig. 12-Variation in geologic log-derived porosity with depth
and match. with theoretical curve for Miocene mudstone,
Nagaoka plain, Japan.

dewater and, accordingly, gas production has been low. For the
Black Warrior basin, pore compressibility is two times that for the
San Juan and Piceance basins.
Fig. 11 shows a log-log plot for all the coal permeability results
obtained to date from the Piceance, San Juan, and Black Warrior
basins, except for known anomalies. A good correlation of theory
with the data is seen in the figure, and the pore compressibility for
all data combined is nearly the same as that for individual basins.
Again, constant pore compressibility was used as the simplest assumption pending additional data. Fig. 12 compares constant core
compressibility vs. exponentially declining pore compressibility.
The curve for constant pore compressibility is the same (same
87

0
Shiunji SK-21

.:

fj> = 50% at 500 m


cp= 1.48Xl0-3 1m

500

. 1-

, .'

. .r

.'

.
:

Uonuma
Group
Haizume
Formation

Nishiyama
Formation

i."

1000

=
o

Co 1500
CII

fj> = 25.4% at 1400 m

Shiiya
Formation

cp = 6.43x1 0-4 1m
2000

.~

-,..::..

250J

'.

':.

t ".
J"

".

ITeradomari
Formation
Nanatani
Formation

I Tuff
Shiunji

at the Deep Seam Coal Project site. * Finally, for the Black Warrior basin at Rock Creek, Diamond reported a range of approximately
1.6xlO- 6 to 2.7xlO- 6 psi- I [2.3xlO- 7 to 3.9xlO- 7
kPa -I]. **
The porosity of these coals is approximately 4 %, implying
mechanical-response pore compressibilities in the range of
2.5xlO- 5 to 9.5xlO- 5 psi- I [3.63xlO- 6 to 1.38xlO- 5
kPa - 1 ]. The arithmetic average coal core (Table 1) and basin pore
(Table 2) compressibilities determined from stress-dependent permeability theory are 1.45 X 10- 3 and 1.22 x 10 -3 psi -I [2.lOx 10- 4
and 1. 77 x 10 -4 kPa -I], respectively. These two, core and basin, pore compressibilities from permeability-response are within
19% of each other, and are 22 times higher than the average
mechanical-response pore compressibility of 6 x 10 -5 psi- I
[8.7 x 10 -6 kPa -I]. The implication is that the active porosity in
the mechanical-response tests is the total porosity, whereas only
the effective porosity to flow is respom;ling in the permeability/effective-stress tests. Assuming permeability scales are the cube
of porosity (Eq. B-7), only 22 - y, or 35.7 % of the porosity is effective or active for fluid flow. Hence, with 4 % porosity, only
1.43 % is effective for purposes of permeability response on pore
compressibility. Coal at Rock Creek at a depth of 1,450 ft [442
m] has a matrix compressibility of2xlO- 5 psi- I [2.9xlO- 6
kPa -I]; when divided by the effective porosity (0.0143), it yields
a pore compressibility of 1.4 x 10 - 3 psi -I [2.03 x 10 -4 kPa -I ],
which is in very good agreement with both the laboratory and basin permeability-response pore compressibilities. More work is required to establish the above hypothesis. If correct, however, it
suggests that stress-dependent core permeability tests would be capable of determining in-situ pore compressibility values for use in
reservoir modeling and in interpretation of well-test pressure response.

3000
0

10

50
60
20 30 40
Mudstone Porosity (%)

70

Fig. 13-Comparison of constant with variable pore compressibility correlations for stress-dependent permeability.

parameters) as in Fig. 11. For variable pore compressibility, initial compressibility was selected such that average pore compressibility between 100 and 5,000 ft [30.5 and 1524 m] (just before
permeability becomes constant) is 8.78 x 10 -4 psi -I [0.127
MPa -I], the same as in Fig. 11. The constant defining rate of decline of pore compressibility, ex, was taken as 1 x 10 - 3 psi - 1
[0.145 MPa- I ], an average for core samples in Figs. 5 and 6. Fig.
12 shows that with exponentially declining pore compressibility,
permeability will eventually become constant. However, the current field data available for these basins do not warrant use of exponentially declining pore compressibility.
Porosity Data. As noted earlier, porosity data available were very
limited. The mudstone porosity values at Shiunji SK-21 in Nagaoka plain, Japan, were estimated from induction logs and are shown
in Fig. 13, which also shows excellent theoretical curve matches
to porosity data for the Nishiyama and Shiiya formations using constant pore compressibility. 24 For the remaining formations, data
were insufficient to perform a curve match. The shallower Nishiyama formation has higher pore compressibility and initial porosity
than the deeper Shiiya formation. The pore compressibilities for
these formations are within the ranges of other formations reported herein.

Comparison of Field- and Laboratory-Measured


Core CompresslbUltles
It is noteworthy that pore compressibilities given in Tables 1 and
2 are at least one order of magnitude higher than those measured
from stress/strain response on core samples. For Fruitland formation coal from a depth of 3,085 ft [940 m] in the San Juan basin,
Jones et ai. reported core matrix compressibilities of approximately lxlO- 6 to 2.5xlO- 6 psi- I [1.45xlO- 7 to 3.63xlO- 7
kPa- I ].25 In the Piceance basin, Seccombe noted 3.8xlO- 6
psi- I [5.5xlO- 7 kPa- l ] for Coal Seam D (Mesaverde group)
88

Practica~ Application of Permeability and


Porosity Correlations
An important consideration in estimating site permeability from
these correlations is the strong influence of local geology. In particular, structural deformation-folding and faulting-frequently
causes secondary or fracture permeability. Moreover, stresses other
than lithostatic stress can enhance or reduce permeability. Although
the theory presented here does not account for fracturing as a result of such processes, some of these phenomena can be modeled
using an effective stress gradient other than lithostatic (0.572 psi/ft
[12.9 kPa/mD as shown in the section discussing field permeability data. Stresses less than lithostatic can occur for several reasons.
First, because effective stress is the difference between total stress
and fluid pressure (Eq. A-12), overpressure of gas or water can
lower it. Second, formation buckling or uplifting can dilate fractures, thereby lowering effective stress and enhancing permeability. Third, a more competent unit in the stratigraphic section can
bear a disproportionate amount of load and reduce the stress on
underlying units. Many workers have used Heim's rule, which states
that stresses in rock tend to approach lithostatic because of creep. 26
If creep is not significant, however, then the horizontal stress will
be less than the total stress by the ratio v/(1- v), where v is Poisson's ratio.
Under GRI sponsorship, Terra Tek Inc. has been working to determine optimum design conditions that will improve hydraulic fractures. One product of that study has been downhole stress gradients
of 0.264 psi/ft [5.97 kPa/m] for the Brookwood unit and 0.194 psi/ft
[4.39 kPa/m] for the Cedar Cove unit in the Black Warrior basin,
compared with an effective stress gradient of 0.572 psi/ft [12.9
kPa/m], where the measured total stress gradient approximates 1
psi/ft [22.6 kPa/m] in the Rock Creek and Oak Grove areas. Therefore, even in the same basin the critical factors determining permeability with depth may vary appreciably and it is important to obtain
in-situ stress conditions for use with these correlations.
The theory's success to this point evidently has been attributable
to two factors: (1) applicability ofthe Carman-Kozeny equation to

'Personal communication, J. Seccombe, REI Inc., Grand Junction, CO (1986).


"Personal communication, P. Diamond, USBM, Pittsburgh, PA (1986).

SPE Formation Evaluation, March 1988

fractured media and (2) continuity of preservation of fractures and


coal-pore geometries from shallow to greater depths. The theory
fits both available laboratory and field data.

Conclusions
Relationships have been developed from fundamental principles
regarding permeability and porosity changes in coal and other porous materials as a function of effective stress. The well-known
Carman-Kozeny relation can be used to describe permeability. Because solid matrix can be considered incompressible, all volume
changes resulting from compression were attributed to pore spaces.
These simple assumptions were used to develop formulas for permeability and porosity as a function of depth.
The developed relations agreed closely with permeability- and
porosity-vs.-stress data determined from laboratory core measurements. Although two of the eight examples 'required exponentially
declining compressibilities to fit the data, the other six samples could
be fitted using constant pore compressibility. Based ,on these results,
constant pore compressibilities were used in fitting field
permeability-vs.-depth data for the Piceance, San Juan, and Black
Warrior basins. Surprisingly, all three basins showed close fits using constant pore compressibility. Mean pore compressibility,
cP' for Black Warrior (1.87 X 10 -3 psi -I [0.271 Pa -I D is twice
that of the San Juan and Piceance basins (9.24 x 10- 4 psi -I [0.134
Pa-1D.
Although the theory is capable of using either a variable matrix
or pore compressibility, more data must be obtained to define variable compressibility more precisely. The effective stress gradient
also was identified as being critical to determining permeability at
a given depth. The theory predicts that a lower effective stress gradient will result in higher permeability.
The reader is cautioned that the theory discussed does not account for structurally enhanced permeability; when applying the
theory to field cases, local effective stress gradients must be used
rather than assuming lithostatic conditions.

Acknowledgments
We wish to express our thanks to Chuck Brandenburg, D' Arcy
Homer, Leo Rogers, and Dick Schraufnagel of GRI for their comments and assistance. We are grateful to Phil Randolph ofIGT for
bringing this problem to our attention; his review of the concepts
was very helpful. We also acknowledge work done by Randolph
and Bob Rose at IGT. We express appreciation to Fred Schwerer,
formerly of US X, Vello Kuuskraa of Levin and Assocs., and Ted
Way of In-Situ Inc. for their thoughtful and constructive criticisms.
We also thank Arfon Jones of Terra Tek Inc. for bringing J.B.
Walsh's work to our attention, Greg Bell of Terra Tek Inc. for helpful discussions, and Jane Reverand and Lindsay Holichek for their
editing and preparation of the manuscript. This research was supported by GRI Contract No. 5082-214-0729.
References

Subscripts
g = grain
o = value at initial conditions
w = water

I. "Coal Seams," Unconventional Gas Sources, Natl. Petroleum Council (1980) 2, 99.
2. Wicks, D.E., McFalls, K.S., and Kuuskraa, V.A.: "A Geologic Assessment of Natural Gas From Coal Seams in the Warrior Basin, Alabama," Gas Research Inst. topical report No. GRI-86/0272, Chicago
(1987).
3. Lent, J. "San Juan Basin Report-Early Tertiary Geology, Coal, and
the PoteJJ1iatIor Methane Recovery From Coalbeds in Colorado and
New Mexico," report, u.s. DOE Morgantown Energy Technology
Center, TRW Inc., Denver, CO (1979, 1982).
4. Wicks, D.E., McFalls, K.S., and Kuuskraa, V.A.: "A Geologic Assessment of Natural Gas from Coal Seams in the Piceance Basin, Colorado," Gas Research Inst. topical report No. GRI-87/0060 (1987).
5. Tremain, C.M. and Toomey, J.J.: "Coalbed Methane Desorption
Data," Colorado Geological Survey, Open"File Report 81-4 (1983) 514.
6. Coalbed Methane Resources of the United States, C.T. Rightmire, G.E.
Eddy, and J.N. Kirr (eds.), AAPG Studies in Geology, Series 17 (1984)
378.
7. Way, S.C. etal.: "Hydrologic Characterization of Coal Seams for Methane Recovery, Activity 1 Topical Report-Hydrologic Data Base for
Piceance, San Juan, and Warrior Basins," Gas Research Ins!., final
report GRI-83-0064 (1984) 59.
8. Way, S.C. et al.: "Hydrologic Characterization of Coal Seams for Methane Recovery, Activity 2 Topical Report-Hydrologic Constraints and
Test Procedures: Selected Methods for Well Test Analysis," Gas Research Inst., final report GRI-83/0065.2 (1984) 2, 108.
9. Dabbous, M.K. etal.: "The Permeability of Coal to Gas and Water,"
SPEI (Dec. 1974) 563-72.
10. Somerton, W.H., Soylemezoglu, I.M., and Dudley, R.C.: "Effect of
Stress on Permeability of Coal," USBM, Open-File Report 45-74, contract H0122017 (1974) 56.
11. Rose, R.E. and Foh, S.E.: "Liquid Permeability of Coal as a Function of Net Stress," paper SPE 12856 presented at the 1984
SPEIDOE/GRI Unconventional Gas Recovery Symposium, Pittsburgh,
May 13-15.
12. Jones, F.O.: "A Laboratory Study of the Effects of Confining Pressure on Fracture Flow and Storage Capacity in Carbonate Rocks," JPT
(Jan. 1975) 21-27.
13. Ward, D.C. and Thomas, R.D.: "Effect of Overburden Pressure and
Water Saturation on Gas Permeability of Tight Sandstone Cores," JPT
(Feb. 1972) 120-24.
14. Ostensen, R. W.: "Microcrack Permeability in Tight Gas Sandstone, "
SPEI (Dec. 1983) 919-27.
15. Schwerer, F.C. and Pavone, A.M.: "Effect of Pressure-Dependent
Permeability on Well-Test Analyses and Long-Term Production of Methane from Coal Seams," paper SPE 12857 presented at the 1984
SPE/DOE/GRI Unconventional Gas Recovery Symposium, Pittsburgh,
May 13-15.
16. McKee, C.R. et al.: "Using Permeability-Vs-Depth Correlations to
Assess the Potential for Producing Gas from Coal Seams," Quarterly
Review of Methane from Coal Seams Technology (July 1986) 4,
No.1, 15-26.
17. Krantz, R.L. et al.: "The Permeability of Whole and Jointed Barre
Granite," IntI. J. Rock Mech. Min. Sci. & Geomech. Abstr. (1979) 16,
225-34.
18. Jones, F.O. and Owens, W.W.: "A Laboratory Study of LowPermeability Gas Sands," JPT(Sept. 1980) 1631-40.
19. Scorer, J.D.T. and Miller, F.G.: "A Review of Reservoir Rock Compressibility and Its Relationship to Oil and Gas Recovery," paper IP
74-003, Inst. of Petroleum, London (1974) 29.
20. van der Knapp, W. and van der Viis, A.C.: "On the Cause of Subsidence on Oil-Producing Area," Proc., Seventh World Pet. Cong., Elsevier Science Publishing Co., Mexico City (1967) 3, 85-95.

SPE Formation Evaluation, March 1988

89

Nomenclature
a = aperture, ft [m]
A = cross-sectional area, ft2 [m 2]
c m = bulk matrix compressibility, psi - 1 [kPa - 1]
~P = pore volume compressibility, psi - 1 [kPa - 1]
c P = average pore compressibility, psi - 1 [kPa - 1]
D = depth, ft [m]
gs = stress gradient, psi/ft [kPa/m]
h = solid matrix thickness between two fractures, ft [m]
k = permeability, md
n = number of fractures per unit cross-sectional area
p = pore (fluid) pressure, psi [kPa]
/j.p = pressure drop, psi [kPa]
r = correlation coefficient
s = constant relating change in pore pressure to change
in effective stress
Vs = volume of solids, ft3 [m 3]
w = fracture height, ft [m]
jj.z = change in vertical direction, ft [m]
ex = rate constant defining rate of decline of
compressibility as effective stress increases, psi- I
[kPa- I ]
E = void ratio
II = Poisson's ratio
p = specific density
(J = effective stress, psi [kPa]
(Jt = total stress, psi [kPa]
/j.(J = increase in effective stress, psi [kPa]
T = tortuosity
cP = porosity

21. Athy, L.F.: "Density, Porosity, and Compaction of Sedimentary


Rocks," AAPG Bulletin (1930) 14, No. I, 1-24.
22. Hsieh, P.A., Neuman, S.P., and Simpson, E.S.: "Pressure Testing
of Fractured Rocks-A Methodology Employing Three-Dimensional
Crosshole Tests," U.S. Nuclear Regulatory Commission, Report
NUREG/CR-3213 (1983) 176.
23. Koenig, R.A. and Stubbs, P.B.: "Interference Testing of a Coalbed
Methane Reservoir," paper SPE 15225 presented at the 1986 SPE/DOE
Unconventional Gas Symposium, Louisville, May 18-21.
24. Magara, K.: "Compaction and Migration of Fluids in Miocene Mudstone, Nagaoka Plain, Japan," AAPG Bulletin (Dec. 1968) 52, No. 12,
2466-2501.
25. Jones, A.H. et al.: "Methane Production Characteristics of Deeply Buried Coalbed Reservoirs," Gas Research Inst., annual report,
GRI-83/0017 (1983).
26. Jaeger, J.C. and Cook, N.G.W.: Fundamentals of Rock Mechanics,
third edition, Halstead Press, New York City (1979) 111.
27. Hantush, M.S.: "Hydraulics of Wells," Advances in Hydroscience,
V.T. Chow (ed.), Academic Press, New York City (1964) 1, 281-432.
28. McKee, C.R. and Hanson, M.E.: "Explosively Created Permeability
From Single Charges," SPEJ (Dec. 1975) 495-501; Trans., AIME, 259.
29. Reiss, L.H.: The Reservoir Engineering Aspects of Fractured Formations, Gulf Publishing Co., Houston (1980).
30. Walsh, J.B.: "Effect of Pore Pressure and Confining Pressure on Fracture Permeability," Inti. J. Rock Mech. Min. Sci. Geomech. Abstracts
(1981) 18, No.5, 429-35.
31. Brandt, H.: "Compressional Wave Velocity and Compressibility of Aggregates of Particles of Different Materials, " ASME J. Applied Mechanics (Dec. 1967) 34, Series E, No.4, 866-72.
32. Carman, P.C.: "Fluid Flow Through Granular Beds," Trans., Inst.
Chern. Eng. (1937) 15, 150-66.

Appendix A-Poroslty/PermeablIlty/Effectlve
Stress Relationships
When the compressibility of individual grains in solids is negligible compared with the change in porosity, the following relationship holds for the change in porosity, cf>, with respect to change
in compressive (effective) stress, (J, which is derived by considering the volume of solid grains [VS =ALlz(I-cf] constant 27 :
dcf>= -cm(l-cfd(J, .............................. (A-I)

where Cm' bulk matrix compressibility, is related to pore compressibility, Cp' by porosity as follows:
C m =cf>cp '

........................ (A-7)

cf>=cf>o 1-cf>o(l-e-cpt:.<J)

Based on the Carman-Kozeny equation (see Appendix B), the following relationship between permeability and porosity can be
written:
cf>3

ko:

................................... (A-8)
(1-cf2

It is unnecessary for the fracture surface to be smooth because the


tortuosity factor, T, is included in Eq. A-8 as shown in Appendix
B (Eq. B-7). This equation also was used by McKee and Hanson
to explain explosively created permeability. 28 Then, combining
Eqs. A-7 and A-8,

......................... (A-9)

k=ko I-cf>o(l-e -cpt:.<J)

For most deeply buried fractured media, porosity is on the order


of I %, and the denominator term can be approximated to unity to
give 29
k=koe -3cpt:.<J . ................................. (A-IO)

This simplification introduces an error less than the scatter in the


experimental data. Schwerer and Pavone observed this simplification. 15 Eq. A-1O shows that, with small porosity, a semilog plot
of permeability vs. effective stress (log k vs. (J) results in a straight
line if pore compressibility is nearly constant. The derivation for
matrix compressibility as the fundamental parameter is given by
McKee et at. 16

...................................... (A-2)

Eq. A-I assumes that all the stress relief is the result of pore space
comprising the effective interconnected porosity, cf>, which is the
basis of the commonly accepted storage coefficient used in hydrology and petroleum reservoir engineering. Combining Eqs. A-I and
A-2 yields
dcf>= -cf>(I-cfcpd(J . ............................. (A-3)

Integrating Eq. A-3 over the stress range yields the following porosity / effective-stress relationship:

-J

cf> o_exp ( ' <Jcpd(J) , ................. (A-4)


= _cf>_ = __
I-cf> I-cf> o
<Jo

where is void ratio. Somerton et at. 10 and Scorer and Miller 19


observed that bulk and pore matrix compressibilities are stress dependent. Average pore compressibility, cp' introduced to reduce
the complexity of the following equations, is defined as

cp = -I - \'
(J-(Jo

where ~(J, the change in effective stress, equals (J-(Jo. The remaining derivations in this section are still valid for variable compressibility using Eq. A-6, because p can vary as a function of
stress. Eq. A-6 shows that if pore compressibility is constant, then
a plot of log and void ratio vs. effective stress should result in a
straight line. Solving for porosity in Eq. A-6 results in

<J

Dependence of Permeability on Fluid Pressure. If the water pressure is assumed to act effectively throughout the elemental volume,
Hantush notes that a change in fluid pressure results in an equal
and opposite change in effective stress, 27
dp= -d(J . ..................................... (A-II)

Walsh 30 then gave the following general relationship between effective stress,. (J, total stress, (JI' and pore (fluid) pressure, p:
(J=(JI-SP. . ................................... (A-12)

He assumed and justified that s was a constant for linearly elastic


materials, and also rocks. A similar factor was also used by
Brandt. 31 Integrating Eq. A-12 over the stress range gives
~(J=(J-(Jo =s(Po -p) =s~p, ...................... (A-l3)

where ~p is pressure drop, while ~(J is the increase in effective


stress, and (JI is taken as constant caused by the overburden. Substituting Eq. A-l3 into Eqs. A-6 and A-9 yields the desired pore
pressure porosity and permeability relationship:

cpd(J, ............................. . (A-5)

~o

_cf>_

I-cf>

so that Eq. A-4 then becomes

=~e-scpt:.P

.......................... (A-14)

l-cf>o

and

=~=~e-Cpt:.<J,
I-cf>

90

l-cf>o

........................ (A-6)

e -3scpt:.p
....................... (A-15)

k=k
o I-cf>o(l-e -scpt:.P)

SPE Formation Evaluation, March 1988

or, in the simplified form of Eq. A-lO,


k=koe -3scpf!.p. . ............................... (A-16)

Dependence of Density on Effective Stress. The specific density, P, of a formation can be written in terms of specific grain density, Pg' and formation porosity as

Unit_-+~

Area

p=pg(l-cf . .................................. (A-17)

Porosity, cf>, can be substituted from Eq. A-7 to yield specific density as a function of effective stress:
pg(l_cf>o)
p=

l-cf>o(l-e c pf!.lJ)

............................ (A-lS)

Solving Eq. A-17 for cf> as a function of P gives


Pg-P
cf>=-- . .................................... (A-19)
Pg

a = Aperture
= Crack Width

Specific
Discharge

Fig. B-1-Diagrammatlc representation of unit model used In


Carman-Kozeny calculation of fracture permeability.

Therefore,
cf>
Pg-P
_
--=--=e-cpf!.lJ
l-cf>
P

........................ (A-20)

where (a+h) is the cleat spacing, and h is the solid matrix thickness between the two fractures. Solving Eq. B-4 for a as a function of cf>,
'cf>h

Eq. A-20 indicates that a semilog plot of (p g -p)/p vs. effective


stress would yield a straight line for constant pore compressibility .

Appendix B-Derlvatlon of Permeability/Porosity


Relationship for Fractures
The Carman-Kozeny permeability equation for fractures 32 as derived by McKee and Hanson is 28

................................. (B-5)

a=
w(I-cf>/w)

Substituting Eq. B-5 into Eq. B-3 yields


cf>3
........................ (B-6)

k=(h2/12TW2)
(1-cf>/w)2

For continuous fractures, fracture height, w, is unity, which gives


k= 127(nw)2' .................................. (B-1)
k=(h 2 /127)

cf>3

(l_cf2

............................ (B-7)

where 7 is tortuosity, n is the number of fractures per unit area,


and w is fracture height. The unit model used in calculating fracture permeability is shown in Fig. B-1. Because porosity can be
expressed as

Because pore compressibility is much higher than the solid matrix


compressibility, h can be assumed to be constant. Thus,

cf>=naw, ....................................... (B-2)

krxcf>3 /(1_cf2. . ................................ (B-S)

where a is aperture, therefore,


cf>a2
k = - . ....................................... (B-3)

127

However, porosity can also be expressed as


cf>= (a:h )w, .................................. (B-4)

SPE Formation Evaluation, March 1988

SI Metric Conversion Factors


ft x 3.04S*
E-Ol
psi x 6.S94 757
E+OO
psi- 1 x 1.450377
E-Ol
Conversion factor is exact.

m
kPa
kPa- 1
SPEFE

Original SPE manuscript received for review May 16, 1984. Paper accepted for publica
tion June 29, 1987. Revised manuscript received Aug. 3, 1987. Paper (SPE 12858) first
presented at the 1984 SPE/DOE/GRI Unconventional Gas Recovery Symposium held in
Pittsburgh, May 13-15.

91

Você também pode gostar