You are on page 1of 17

CSUG/SPE 146580

A New Classification Plot for Naturally Fractured Reservoirs


James R. Gilman, Huabing Wang, Sepehr Fadaei, Michael J. Uland, iReservoir.com, Inc.
Copyright 2011, Society of Petroleum Engineers
This paper was prepared for presentation at the Canadian Unconventional Resources Conference held in Calgary, Alberta, Canada, 1517 November 2011.
This paper was selected for presentation by a CSUG/SPE program committee following review of information contained in an abstract submitted by the author(s). Contents of the paper have not
been reviewed by the Society of Petroleum Engineers and are subject to correction by the author(s). The material does not necessarily reflect any position of the Society of Petroleum Engineers,
its officers, or members. Electronic reproduction, distribution, or storage of any part of this paper without the written consent of the Society of Petroleum Engineers is prohibited. Permission to
reproduce in print is restricted to an abstract of not more than 300 words; illustrations may not be copied. The abstract must contain conspicuous acknowledgment of SPE copyright.
Abstract
The classic Nelson plot (Nelson, 2001) has been widely used to illustrate the geologic characteristics of various NFR types
ranging from matrix dominated flow with some fracture enhancement to purely fracture dominated systems (e.g. basement
reservoirs). However, for many reservoirs, there is significant overlap in model types and different parts of the reservoirs may
fit into different classifications. In fact, the choice of characterization and modeling approaches depends on other
considerations such as the fluid system, matrix geologic characteristics, and recovery processes. For example, for gas-oil
gravity drainage processes, the rate of vertical oil drainage in the matrix may dominate oil recovery and dual-permeability
may be the desired modeling approach even in a highly fractured reservoir. Alternatively in pure depletion processes, we
might be able to model a highly fractured system with an effective single-porosity representation because of rapid matrix-
fracture pressure equilibration. However, in all fractured reservoirs, of special concern is the rate of matrix-fracture fluid
transfer which is directly related to the fracture intensity of the open connected fractures.
In this paper we present a new classification plot that highlights the importance of fracture-matrix fluid transfer. This
consideration is not directly incorporated in the Nelson plot. Our new classification concept can be used to understand
differences in dynamic performance of NFR for purposes of analogue selection or screening of potential recovery processes.
In this paper, several characterized NFR reservoirs (including unconventional systems) are compared via the new
classification plot and their differences and performance are discussed. Additionally we provide some discussion of naturally
fractured reservoir (NFR) characteristics for the purpose of illustrating the controls on matrix-fracture fluid transfer.
Introduction
Characterization and simulation of naturally fractured reservoirs (NFR) is a difficult process. We must apply concepts learned
from analogue fields as well as petrophysical and seismic methods to define the 3D distribution of fracture network
characteristics (permeability, connectivity, intensity, etc). Many of the concepts discussed in the next few sections are
summaries of details provided by Gilman (2003) and those concepts were a result of collaborative efforts with Dr. H.
Kazemi, Colorado School of Mines (unpublished). These first sections are intended to provide the background for
establishing the importance of matrix-fracture transfer as a concept for comparing performance of NFR.
A common idealization for fractured reservoirs relies on the assumption that in the macroscopic sense, the inter-connected
open fractures are assumed to form a continuum. Thus, both fracture and matrix components of the reservoir are considered
separate continua in the macroscopic sense. This continuum view is analogous to that in conventional reservoirs where the
microscopic pores are very tortuous and the channel dimensions and flow properties vary drastically from point to point. In
this case, the continuum definition of the porous medium flow equations cannot be applied on the pore level scale (Lake,
1989) but must be averaged over some representative pore volume (REV). One of the most important aspects of these
systems is the fracture-matrix transfer. This transfer is a function of the amount of matrix-fracture surface area. Its relation to
fracture intensity has been frequently described and therefore will only be reviewed briefly.
Important flow characteristics of fractured reservoirs are well represented by the dual-continuum idealization such as:
High apparent permeability compared to core measurements
High initial productivity and rapid decline for limited matrix contribution
Early breakthrough of injection fluids or early gas/water coning
Highly directional flow behavior

2 CSUG/SPE 146580
Large variability in well productivities and recoveries
Potential loss of productivity with time
Flow from a very small portion of the open interval
High recoveries even for low quality matrix (e.g. as seen in unconventional reservoirs)
There are a number of issues in regard to the nature of fractures that have significant implications on reservoir
characterization and fluid-flow modeling. For example, whether fractures are a result of shear, extension, or tension, the fact
is that the open fractures generally exist in vertical or sub-vertical forms. However, this does not mean that effective vertical
permeability over the height of the reservoir will be high. The mechanical properties of rocks can vary significantly in
various stratigraphic layers of the reservoir possibly leading to limited vertical connectivity. The variation of properties, such
as Poisson ratio and shear modulus, results in different horizontal stress in reservoir layers, and thus different fracture lengths
and apertures if fracturing occurs. In many cases, fractures may cluster (Gale, 2002).
Additionally, the morphology (form and structure of fractures) is very important (Nelson, 2001). Fractures may be open with
attendant large permeability, gouge filled with low permeability along and perpendicular to fracture, slickensided with high
permeability along the fracture and low perpendicular to it, mineral-filled leading to partial or total fracture closure as well as
reduced permeability, and vuggy where vugs are imbedded in the fractures. Knowledge of morphology is a very important
piece of information in fluid flow modeling because models require estimates of fracture spacing, effective permeability,
porosity distribution, and permeability anisotropy. Diagenesis can alter the permeability and porosity of open fractures and
fully mineralized fractures could, potentially, become baffles or barriers to flow.
Dual-Continuum Modeling
For dual-media simulation, the fracture network is described as a continuum with independent properties (e.g., permeability
and porosity) defined in a manner similar to the matrix rock. For example, porosity and permeability are defined with respect
to the bulk volume of the core, not with respect to individual fractures. Most approaches for quantifying the flow properties
on the fracture networks are based on a qualitative assessment of the 3D distribution of fracture network characteristics, such
as relative fracture intensity, directional aspects, and whether the fractures are generally open or closed. These qualitative
assessments are then calibrated to historical well performance via direct approaches often referred to as continuum flow
models (CFM) or upscaled from discrete fracture models (DFM). Both methods require very similar workflows.
The characterization approaches use information such as structural data, seismic attributes, image logs, and field production
performance to estimate the fracture network properties (e.g., Gauthier et al., 2002). Deterministic and/or stochastic models
are built with these data and validated by comparing to the field performance (e.g., water and gas breakthrough). Non-
standard dynamic field tests such as tracer tests (Kazemi and Shinta, 1993) can be useful for quantifying the fracture
heterogeneity. Interference or pulse testing (Araujo et al. 1998) can also provide important information such as directional
information to compare to image logs, stress relations for calibrating with laboratory and fracture volume from rapid pressure
response between wells.
Traditional methods of estimating fracture parameters have also included single-well pressure transient testing to estimate
fracture network conductivity, and cores and image logs to estimate fracture network intensity. The theoretical dual-porosity
signature of the pressure build-up curve (Warren and Root, 1963) is often missing because of the presence of well bore
storage effects or non-uniform fracture distribution. However, comparing the pressure transient permeability to that estimated
from matrix permeability can give an indication of the connectivity of the fracture network.
Fracture Permeability
In an ideal setting, we could estimate fracture network effective permeability from aperture and porosity estimates using
parallel plate idealizations (Kazemi and Gilman, 1993) and for non-uniform fracture orientations through DFN models
(Dershowitz, et. al. 2000). However, in our experience, the effective fracture permeability for a network of fractures will
likely be much smaller than calculated from such relations because fractures have rough and touching surfaces, are tortuous,
may be constricted by mineral deposition, and may terminate abruptly. Furthermore, the vertical permeability of an
individual fracture could be quite high; however, if the fractures terminate in shale lenses or are of limited height for some
other reason, the effective vertical permeability of the network could be quite low. Therefore, permeability is best determined
from field tests and not from core/log-measured fracture widths and spacing. The static data however can be used to derive
scaling relations to distribute permeability in 3D.
Permeability in NFR can be highly directional. Considering this highly directional nature, the need for permeability tensors
has been widely discussed but not widely applied in field-scale fluid-flow modeling. Whether such a tensor approach is
required is still debated because the actual data required to uniquely define the tensors are not available. Also, the extra
computation required means that additional upscaling would be required, reducing other details in the model. Aligning the
CSUG/SPE 146580 3
grid system with the directional trends and applying the main tensor diagonals (principal permeabilities) has proven to be
adequate for many field scale systems.
Single-well flow tests, such as a pressure build-up test, cannot decipher permeability anisotropy. The most direct method to
calculate permeability tensor is from the interpretation of pressure interference tests conducted in various parts of a field.
Haws and Hurley (1992) summarized interference test results from a number of Big Horn Basin fractured reservoirs, where
they reported that the maximum to minimum permeability ratio varied from 10 to 1000. Tracer tests, compared to
interference tests, are more time consuming and expensive and can best be run only in displacement (e.g. injection
processes). However, tracer tests conducted in a naturally fractured reservoir can provide additional insight on permeability
anisotropy (Kazemi and Shinta, 1993). Finally, the most common method used to assess the directional flow tendencies is
based on fracture orientation studies of borehole image logs, oriented cores, outcrops and water or gas breakthrough
tendencies. Seismic methods can also provide a means to infer anisotropy which may be related to open natural fractures (Ata
and Michelena, 1995). These seismic methods must be calibrated to dynamic data.
Effective fracture permeability tensor can be calculated for a REV cube or a grid cell using discrete fracture network (DFN)
models based on measured fracture aperture, surface area and the inclination angle normal to the fracture surface. Only the
open fractures that contribute to fracture flow should be included (Oda, 1985). This approach is a common method for
scaling DFN models, image logs and outcrops to equivalent dual-porosity medium (Dershowitz, et. al., 2000).
If fractures close as reservoir pressure declines, then effective permeability could decrease. A number of researchers have
derived the expressions for permeability, aperture, compressibility, and porosity of smooth open fractures as functions of
pore pressure and stress state in an elastic porous media. However, fracture closure is very difficult to estimate because of the
presence of filler material, or complex stress states of the reservoir. Also at large flow rates, the onset of turbulent flow can
create additional pressure drop causing an apparent reduction in permeability. Step rate tests can indicate if turbulent flow is
important (Pereira, et. al. 2004).
One common method of assessment of fracture importance is a term called excess permeability. This is the permeability
above that expected from matrix flow. The Nelson plot incorporates this as one of the axes. As flow through the fractures
dominates, then the excess permeability ratio can be quite high (> 1000). Excess permeability can be estimated by comparing
well productivity or well-test permeability to that estimated for matrix only flow. However, such assessment alone may not
be sufficient to determine if natural fractures are important. Consider the ideal picture in Figure 1 for a theoretical case where
the effective fracture network permeability is such that the total effective permeability (k
e
) is two times larger than a matrix
only system. From Eq. 1 this means that k
f
|
f
(what we will call k
fe
) is equal to k
m
.
m f f e
k k k + = | (1)
Such a factor would be within the uncertainty of matrix permeability and thus such a reservoir could easily be interpreted to
be non-fractured. If fluid is injected into the system then one-half of the fluid will flow through the fracture; however, the
ratio of matrix fluid velocity to fracture velocity would be |
m
/|
f
. Because the fracture porosity is generally one to two orders
of magnitude lower than matrix porosity, the velocity in the fracture could be more than 10 times that of the matrix. Thus, in
the absence of matrix-fracture transfer, one-half of the fluid will move at a velocity more than 10 times greater than that
expected for matrix only flow resulting in very rapid breakthrough. This illustrates the importance of fractures to fluid
movement even when apparent permeability is not significantly greater than that of the matrix. What becomes of major
importance is the matrix-fracture fluid transfer which can significantly affect the behavior of the system.
Fracture Porosity
Fracture porosity calculations require an estimate of fracture spacing (intensity) and width, which is often estimated from
cores, image logs and outcrops. The local porosity calculated for an individual fracture would essentially be 1.0; however,
when calculated with respect to the bulk rock volume, fracture bulk porosity is quite small (<1.0%). In addition to cores and
logs, one can in theory use the effective permeability and fracture spacing results from pressure buildup tests to estimate
fracture porosities (Kazemi and Gilman, 1993).
For reservoir characterization and flow modeling, fracture porosity needs to be estimated at all well locations and distributed
in the 3-D space. Typically, we do not have image logs or cores at all wells; therefore, we need to estimate porosity from
conventional logs (e.g. Iwere, 2002) and develop empirical relations to relate fracture porosity to other static information
such as flexure, lithology, faulting, and other intrinsic and extrinsic rock mechanical characteristics. Researchers have also
estimated fracture porosity (and permeability) based on fracture-scaling fractals in terms of fracture width and/or length (e.g.
Hossain, et. al. 2002; Gale, 2002).
4 CSUG/SPE 146580
When calculated from cores, images logs or DFN models, fracture porosities are often very small (<0.01%). However, in our
experience from interference tests, tracer tests, injectant breakthrough times, and long-term performance, fracture volumes
appear greater than estimated from these methods (by up to one order of magnitude). This is because many small-scale
fractures and vugs can contribute to apparent fracture network volume and fracture compressibility may be higher than
estimated.
Shape Factor
In addition to the conventional reservoir parameters (porosity, permeability, and compressibility for both fracture and
matrix), an additional parameter, (often referred to as the shape factor), is needed for fluid transfer calculation between the
matrix and fracture network. This parameter is represented by the symbol o in units of L
-2
. This term is best understood
from the Warren and Root paper (1963) in which they idealized the system as a stack of sugar-cubes.
There has been much discussion about the physical meaning and the functional form of the shape factor. From a practical
view, it is a second order, distance-related, geometric parameter that is used to calculate the mass transfer coefficient between
matrix blocks and surrounding fractures. Shape factor is a function of fracture spacing (or intensity), and is not inherently a
time-dependent parameter. For precise modeling of short-term transients, several authors have treated the shape factor as
time-dependent by including time dependent characteristics in the flow terms. Ideally, the shape factor could be calculated
from pressure build-up data. Several expressions describing the shape factor have been presented in the literature and some
are summarized here.
The pseudo-steady state, analytically derived expression for the shape factor (Kazemi and Gilman, 1993, Chang, 1993,
Zimmerman, 1993, Lim, 1995) assuming uniform fracture spacing (L) in the x, y, and z directions is:
|
|
.
|

\
|
+ + =
2 2 2
2
1 1 1
z y x
L L L
t o (2)
The pseudo-steady state, numerically derived expression for shape factor (Kazemi, 1976, Kazemi and Gilman, 1988) has a
coefficient of 4 rather than t
2
. A transient state, analytically derived shape factor was given by Penuela, et. al. (2002), but is
not generally used in field scale simulations. Current commercial simulators allow this shape factor to vary on a cell-by-cell
basis and thus allow for complex characterization of fractured reservoirs.
A generalized shape factor based on water imbibition experiments (Kazemi, Gilman and El-Sharkawy, 1992, Zhang, et. al.,
1996) is:

=
=
I
i i
i
d
A
V
1
1
o (3)
In the above equation, A designates the area of the exposed fracture surface i for the given matrix block with a volume
V, d is the distance from the center of the block to surface i, and I is the number of exposed fracture surfaces. This equation is
also the most general form for upscaling from complex fracture patterns (e.g., DFN network models) to equivalent dual-
porosity media for both single- and multi-phase flow. Since flow between matrix and fractures is linearly dependent on o,
this equation shows how matrix-fracture surface area is an important factor in fractured reservoirs.
Rather than shape factor, geologists often refer to fracture intensity using nomenclature such as P10, P32 or P33 (Dershowitz
et. al., 2000). P33 is equivalent to fracture porosity (pore volume per rock volume). DFN models are often built using a
volumetric measure of intensity, P32, which is the fracture surface area per volume of reservoir. This measure has many
desirable scaling properties that make it mathematically convenient. However, it is not possible to measure this parameter
directly. Fortunately, it can be linearly related to other measurable properties, and this relation can be computed through
simulation (Dershowitz et. al, 2000). Once this relation is established, it is possible to calculate the desired P32 value from
the number of fractures per meter (P10) or from fault traces per unit area (P21). In fact, for some ideal fracture distributions,
Wang (2005) has developed a correlation between P10 and P32. Eq. 4 shows how P32 has a similar form to the value for o,
thus they should be closely related. (Here A refers to the fracture area between two exposed matrix surfaces, so is lower by
a factor of two from A in the previous equation for shape factor (o). In our experience, and from theoretical calculations, o
is proportional to P32 squared. Additionally fracture porosity and fracture permeability are theoretically directly proportional
to P32; however because of wide variability in fracture width, length, roughness and connectivity, we do not generally see a
simple correlation.
CSUG/SPE 146580 5

=
=
I
i
i
A
V
P
1
1
32 (4)
In practice, shape factor is often considered a history matching parameter, but modern workflows start with the above
equations as the best estimate of the fracture intensity (which can be converted to shape factor). The shape factor can vary
over the computational grid and is only considered a constant over the REV (grid volume). Note that the numerical models
do not actually require any discrete representation of matrix sugar cubes. The idealization of sugar cubes is just a
convenience for illustration similar to the bundle of capillary tubes used for discussion of a single-porosity media. The dual-
porosity models are simply two overlying porous media with a transfer term directly proportional to shape factor (i.e.
fracture-matrix surface area). The shape factor can be calculated for any geometry (Heinemann and Mittermeir, 2006; Sarma,
and Aziz, 2003).
Theoretically the shape factor can be calculated from pressure build-up curves on individual wells. Ideally, such curves
develop two parallel straight-line segments, indicative of dual-porosity behavior (Warren and Root, 1963). The build-up time
at the inflection point, on the transition segment between the early and late straight line build-up segments, is related to shape
factor and depends on two dimensionless parameters, the storativity and the inter-porosity flow, e and , which are defined
as follows:
e
|
| |
=
+
( )
( ) ( )
c
c c
t f
t f t m
(5)
e
w m
k
r k
2
o
= (6)
These relations are important for the classification plot presented in this paper. The inflection point on the transition between
early and late straight-line flow periods depends on the shape factor according to the following relation (time is given in
hours):
e
w f t
k
r n c
t

e |
0002637 . 0
) ( ) (
2
*

= A (7)
In the above equations, k
e
, is the effective permeability of the fractured system as calculated from the slope of the build-up
curve. The vertical pressure separation (op) between the two parallel lines is related to e. Often the customary two straight-
line segments do not appear on the pressure build-up curve because of well bore storage or boundary effects. The time
equation (Eq. 7) shows that the time frame for transition to occur in systems with moderate compressibility can be very short
(and thus is often masked by wellbore storage effects). Also non-uniform fracturing (variable o) complicates the transition
period. If the dual-porosity response is not seen, shape factor will need to be inferred from other information (e.g. image logs)
and calibrated via DFN modeling.
In practice, the effective fracture intensity and degree of connection is difficult to predict. This may be because fractures are
often well connected on a large scale, but a few primary pathways may dominate flow and thus many of the fractures seen in
logs and cores do not inter-connect on a large scale. Also, the apparent connectivity can depend on the well placement and
the recovery process. Thus, coarse-grid, dual-porosity models for multi-phase flow have historically been too well connected.
Secondly, the fracture network appears much more heterogeneous than first estimated and in fact the effective fracture
spacing can appear to be different for depletion versus waterflood. Under depletion, a near-by producer can deplete many of
the fractures, causing matrix fluid to expand through a large fracture-matrix surface area. However, under waterflood, water
can imbibe only through the matrix surface that is contacted by water. Fractures that are not continuous along the water flow
path will not be flooded by injection water even if they are open and permeable. Recent industry activity has been directed at
using discrete stochastic distributions of fractures based on statistical and deterministic information gathered from logs, cores
and outcrops to improve understanding of the fracture connectivity. These stochastic realizations can provide a means to
estimate effective fracture network properties for upscaling into conventional dual-porosity models.
Fluid Flow Behavior
In three-phase systems, consisting of gas, oil, and water, different portions of a given reservoir can undergo different
6 CSUG/SPE 146580
displacement processes leading to complex flow behavior. This usually happens because of the differences in field operations
schemes (i.e., water-flooding vs. gas injection) and rock characteristics such as wettability, fracture intensity, permeability
variations, layering, structural effects, capillary pressure and relative permeability hysteresis. The oil recovery mechanisms
associated with the field characteristics as well as the production schemes in the field are best studied by laboratory
experiments and fine-grid numerical simulation. The approximations made by the simulators for coarse-grid full-field models
must be verified by these fine-grid studies. In the following discussion, we present some of the basic ideas for flow behavior
under various recovery mechanisms. The idea is to illustrate the importance of various parameters on fracture-matrix transfer.
Depletion
In purely single-phase depletion, the dominant recovery mechanism is fluid expansion. In practice, this occurs normally as
the first stage of reservoir operations, and fluid recoveries are typically quite low (except for gas systems). In single-phase
depletion, oil or gas recovery rate can be much greater in NFR compared to non-fractured reservoirs because the high
permeability fracture network undergoes rapid depletion and provides large surface area for reservoir fluid in the low-
permeability matrix to expand into the fractures. This has been a means to obtain significant recovery from low-permeability
unconventional reservoirs. In multi-phase depletion, fracture flow can lead to free gas flow in the fractures, which, in turn,
can invoke gravity drainage of oil from the matrix. Efficient gravity drainage can lead to high oil recoveries if the gas-oil
gravity drainage process is managed properly. The important parameters that affect the rate of matrix-fracture fluid transfer
during depletion can be shown to be shape factor, o, matrix fluid mobility, (k
m
/), and fracture-matrix pressure difference,
(p
m
-p
f
) as can be seen from the equation for single-phase flow in the fractures. The second term in Eq. 8 represents the
matrix-fracture transfer.
( )
t
p
c q p p
k
p
k
f
tf f f m f
m
f
f
c
c
= +
|
|
.
|

\
|
V - V |

(8)
Solution of the single-phase equations for low compressibility flow shows that a double exponential flow rate decline can
develop for a closed radial reservoir, producing under constant bottom-hole pressure depletion,. Chen, et. al., (1986) showed
the exponential decline characteristics of a number of Austin Chalk wells.
Satman (1985) derived equations for the initial flow rate and for the slope of the exponential recovery lines as a function of
the inter-porosity parameter, . Because of the small volume of fluid in the fractures, the initial period may be very short and
therefore difficult to measure. For large (> 16 r
w
2
/r
e
2
) only a single straight line will develop because of the rapid pressure
equalization between the fracture and matrix. The reservoir then responds like an non-fractured reservoir with permeability
equal to k
e
and porosity-compressibility equal to that of the total system. For highly fractured systems with moderate to good
matrix permeability and close fracture spacing (e.g. ok
m
> 0.1 md-ft
-2
), only one straight line will develop.
In theory, Satmans equations could be used with decline curve analysis to determine reservoir properties for the fracture-
matrix system (Chen, et al. 1986). In practice this is very difficult because of variations in bottom-hole pressures, multiphase-
flow, offset well interference effects, complex well geometries/completions and the long production times required to obtain
both exponential decline periods. Also, numerical simulation has shown that the first line is very short lived and the slope is
affected by matrix-fracture flow. Therefore, fracture properties are very difficult to determine from decline curve analysis.
However, long-term rate and pressure decline can be used with conventional reservoir depletion analysis to determine the
effective permeability and total pore-volume of the system (fracture and matrix). The magnitude of effective permeability
compared to only the matrix (core) permeability gives an indication of the relative importance of fractures. This discussion
shows how high fracture intensity (large o) can lead to efficient depletion of fractured systems with low matrix permeability.
Water Imbibition
Water imbibition has proven to be an effective recovery mechanism in some NFR. Imbibition in reservoir rock is the process
in which water is drawn into the rock by the action of capillary forces. Mattax and Kyte (1962), through experimental
investigations, found that recovery in such systems could be scaled through the following dimensionless time:
t
L
k
t
m
m
D (

|
|
.
|

\
|
=
2

|
(9)
This equation shows that recovery is inversely proportional to the matrix block size squared, L
2
. Capillary pressure is
indirectly incorporated in the above equation through the interfacial tension, , permeability and porosity terms. The above
equation can be rewritten in terms of shape factor, o, by replacing the 1/L
2
term as shown in Kazemi, et. al. (1992) and
CSUG/SPE 146580 7
Zhang, et al. (1996). The experimental data generated by Mattax and Kyte can be described by an exponential time function.
Kazemi, Gilman and El-Sharkawy (1992) provide an analytical solution for one-dimensional flow in a fractured systems
using such exponential relations.
The important parameters affecting oil recovery from water imbibition are shape factor, capillary pressure and oil mobility,
k
m
k
ro
/
o
. Oil rate is greater as permeability increases and matrix block size decreases (fracture intensity or shape factor
increases). Recovery rate also increases as oil relative permeability increases making it easier for water to imbibe and
displace oil from the matrix. Higher capillary pressure increases rate of oil recovery through increased imbibition force. NFR
that have undergone waterflood or strong aquifer drive include the Fahud Field (ONeill, 1988), Ekofisk Field (Hallenbeck et
al., 1991), Midale Field (Beliveau, et al., 1993) Ezzaouia Field (Gilman et. al., 1996), and Ghawar Field (Phelps and Strauss,
2002). The more than 30% recovery from the Ezzaouia Field is a case that illustrates the effectiveness of water imbibition
dominated recovery in some NFR.
From a large-scale simulation perspective, some of the issues regarding recovery prediction uncertainty include incomplete
coverage of the matrix block as a result of non-uniform sweep through the fractures, variations in recovery as a result of non-
uniform block size in a grid block and uncertainty in laboratory measurements.
Gas-Oil Gravity Drainage
In some highly fractured reservoirs, gravity drainage can be a very effective recovery mechanism. If matrix drainage is the
limiting factor, then oil drainage can be approximated as a one-dimensional solution. In one-dimensional gas-oil gravity
drainage in a porous media, the time rate of change in elevation of a constant saturation oil shock front (that is, the gas-oil
frontal velocity, ft/d) is given by the following equation (Dykstra, 1978; Richardson, 1989):
c
c

|
c
c
z
t
x k
k
S
S
v og
o
ro
o
S o
o
=

7 83 10
6
. A
(10)
Assuming power-law oil relative permeability function with an exponent of n and coefficient of k
rom
, Eq. 10 can be solved
analytically for the initial oil rate as given by Eq. 11. This initial rate (q
oi
, STB/D) is sustained for a period of t
qmax
days:
0
5
6146 . 5
10 4 . 4

A
=

A k k x
q
rom v
oi
(11)
rom v
p
q
nk k x
h
t

|
A
=
5
0
max
10 4 . 4
(12)
For times greater than t
qmax
, analytical solution gives Eq. 13. This defines a hyperbolic decline equation given by the second
two equations. The hyperbolic b factor is equal to (n-1)/n. The subscript p on porosity refers to the effective oil porosity
|(1-S
wir
-S
gc
-S
orw
).
( )
) 1 /( ) 1 /( 1 5
) 1 /( ) 1 /( 1
0
) 10 4 . 4 ( 6146 . 5


A
=
n n n
rom v
n n n
p p
o
t nk k x n
h A
q

| |
(13)
|
|
.
|

\
|
=
t Ah
nq
b
D
p
oi
1 6146 . 5 1
|
(14)
( )
b
oi o
bDt q q
/ 1
1

+ = (15)
The important parameters affecting gravity drainage rate are vertical permeability, density difference, and oil mobility.
Fractured systems in which gas filled vertical fractures surround a vertically continuous matrix can be represented by this
equation. The above equation ignores capillary pressure. Capillary pressure has minimal effect on early rates, but can cause
hold-up of oil above barriers or horizontal fractures. For lab centrifuge data in which the core is spun at a rate scaled to the
8 CSUG/SPE 146580
field oil column height, the resulting relative permeability will include the capillary holdup and can be used directly in the
above equations for scale-up. Centrifuge experiments (Kyte, 1970) are a common method of assessing effectiveness of
gravity drainage which can be directly upscaled to field behavior.
Permeability variations can affect early drainage rate, but a harmonic average permeability can give a reasonable
approximation to the drainage behavior over a limited permeability range. Simulation shows that for matrix block radii of
10s of ft or less, there is no difference in recovery, showing the insensitivity of this system to areal fracture spacing. Because
the flow is controlled by matrix vertical permeability, the areal spacing of the fractures has minimal impact for relative close
fracture spacing (<10 ft). Fracture intensity will however, significantly affect the rates and coning tendency of producing
wells with regard to the fracture network as well as controlling the ability to capture draining oil at wells. Adding vertical
barriers across a matrix column would appear to initially increase drainage rate, because each isolated matrix block can feed
into the fractures rather flowing vertically through the entire matrix column. However, this draining oil will resaturate lower
matrix blocks as drainage occurs, and thus the system with barriers will drain more slowly. Also, capillary holdup above the
barriers will cause long-term recovery to be reduced.
A commonly used finite-difference approximation to calculate the gravity drainage rate, in STB/D, for the dual-porosity
idealization and negligible capillary pressure is based on the concept of vertical equilibrium as given by:
144
) ( 001127 . 0
og
cogth of om
o o
ro zm go
o
p h h
B
k Vk

o
t
A
= (16)
Where h
om
is the product of the mobile oil saturation in a given matrix grid cell and the matrix block height, L
z
. Similarly, h
of
is obtained by multiplying the mobile oil saturation in the given fracture cell and L
z
. The shape factor for gas-oil gravity
drainage, o
go
, needs to reflect the fact that the process of drainage is primarily vertical. It is given by the following equation:
o
go
z
L
=
2
2
(17)
Because the oil height terms are proportional to L
z
and o
go
is inversely proportional to L
z
2
, the drainage rate becomes
inversely proportional to L
z
as with the analytical equation. L
z
is effectively the distance between major flow barriers. This
form generally over estimates the rate of gravity drainage because of the assumption of vertical equilibrium in the matrix.
A more realistic alternative to conventional dual-porosity, pseudo-gravity simulation is to use fine grids and dual-
permeability simulation only in the vertical direction (Gilman and Kazemi, 1983). This will provide for more accurate
simulation of the gravity drainage process. Dual-permeability in the vertical direction means that all matrix blocks are
allowed to be continuous in the vertical direction. This will allow oil to drain vertically in the matrix as well as laterally into
the fractures. Of course, there is added computational expense because of the dual-permeability approach and the requirement
of additional layers.
In gravity-dominated recovery from NFR, fractures are used as gas flow conduits to invoke gravity drainage from the matrix
by controllably lowering the gas-oil contact in the fractures. This procedure is the basis for the concept of contact
management in fractured reservoirs, which can lead to one of the most effective improved-oil-recovery techniques for
fractured reservoirs. NFR that have undergone gas-oil gravity drainage include the Fahud Field (ONeill, 1988), the Yates
Field (Rothkopf and Wadleigh, 1994; Campanella, et. al., 2000), the Haft Kel Field (Saidi, 1996), and the Cantarell Field
(Arevalo, et al., 1996). In the Haft Kel Field (Saidi, 1996), the calculated water displacement efficiency, supported by field
measurements was only about 17%, whereas the calculated gas-oil drainage efficiency, at a reservoir pressure of 1512 psi,
was about 32%.
Other Modeling Considerations
The multi-phase flow equations require the use of relative permeability in both the matrix and fracture. Relative permeability
in the matrix is the same as obtained for conventional single-porosity cores (although there is an upscaling issue because the
flow rates are a function of the average saturation in large grid blocks). It is often assumed that fracture relative permeability
is a linear function of phase saturations. Laboratory experiments have shown this to be true for large fracture sizes (> 50 m).
However, for smaller fractures it has been demonstrated (Maloney, et al., 1997) that relative permeability in fractures is non-
linear and depends on the fracture flow velocity, direction of flow, and density difference. Other fracture relative
permeability concepts have also been presented by a number of authors (e.g. MacDonald, et al., 1991; Kazemi and Gilman,
1993). There have also been a number of different methods proposed to handle the effect of relative permeability for matrix-
fracture flow, although the most general method is to use conventional upstream weighting. Historical performance also
CSUG/SPE 146580 9
suggests that relative permeability in the fractures is not a linear function of phase saturations. This is because the relative
permeability for a fracture network is an average of a number of discrete fractures.
Previous discussions referred to depletion. water imbibition and gravity drainage, as the most common methods for
enhancing recovery in NFR; however, tertiary recovery, or enhanced oil recovery (EOR), are processes that produce further
additional oil economically over that which can be produced from primary and secondary recovery methods. In NFR, EOR
can be viewed as the methods that accelerate oil recovery by altering reservoir fluid and rock properties to better utilize the
reservoirs natural energy. The most promising EOR techniques in NFR (Christiansen, et al., 1989) include CO
2
(Beliveau
and Payne, 1993; Malik and Islam, 2000), heat (Reis and Miller, 1995), surfactants (Chen, et al., 2001), and polymers
(Sydansk and Moore, 1992). For some of these processes we must also be concerned about diffusion rates and /or heat
transfer rates between matrix and fractures. The effectiveness of these two transport phenomena is also strongly dependent on
fracture-matrix surface area (shape factor).
A New Classification Plot
In this section we present our new classification plot (referred to here as the Gilman Plot) to highlight the relative
importance of fracture-matrix transfer. This consideration is not directly incorporated in the Nelson plot (Figure 2). Effective
matrix-fracture transfer provides a means for micro-darcy rock to provide economic recovery of oil in unconventional plays.
This new classification concept is intended to be used to understand and illustrate differences in dynamic performance of
NFR for purposes such as analogue selection for decline forecasts, screening of potential recovery processes or making
decisions on modeling methodology. In this paper, several characterized NFR reservoirs (including unconventional systems)
are compared via the new classification plot and their differences and performance are discussed. This plot is certainly not all
encompassing with regard to fractured reservoir performance but it highlights the fact that matrix-fracture surface area is a
major aspect of understanding performance of NFR.
The first ratio we will use in the plot is the storativity ratio; as defined by Warren and Root (1963). However because we
often dont know the fracture compressibility, we will remove compressibility (c
t
) from this equation.
m f
f
| |
|
e
|
+
= (18)
The second ratio is the ratio of effective fracture permeability to matrix permeability; often referred to as excess permeability
ratio. However, as discussed earlier, the ratio does not need to be greater than one. For fracture permeability we generally use
a geometric average of the horizontal permeability diagonal tensor terms.
m
fe
exr
k
k
k = (19)
These two parameters are the basis of the Nelson plot and will remain the primary axes in our plot. These are relative easy to
comprehend in terms of magnitude as permeability and porosity are terms that we are all familiar with.
The previous discussion of depletion systems showed how Satman (1985) illustrated that the product of r
e
2
/r
w
2
is a
controlling factor in the long-term behavior of a radial, closed, single-phase, dual-porosity system. He showed, for example,
that if this term is greater than 16, the system (under pseudo-steady state single-phase depletion) will behave like a single-
porosity system with permeability k
e
and porosity |
f
+|
m
. Therefore, for use in our classification plot, a third ratio is a
redefined inter-porosity flow term,
A
, which indicates the relative contribution of fracture-matrix flow compared to flow
through the fractures within the well drainage area, A (Eq. 20). The term is dimensionless as long as consistent units are
used. In this equation, there are different possible ways to define the well drainage area (which will change with time and
with well density). Also the drainage area is complicated for horizontal wells; however, a consistent method from field-to-
field is adequate for classification purposes. In our examples we use a uniform well spacing.
(20)
The magnitude of this inter-porosity flow term is not intuitive, but in a simple sense, it is a measure of the ability of the
matrix to transfer fluids to the fractures, relative to the ability of the fractures to transport fluids to the well over a drainage
volume. If we consider Satmans analysis, a value of
A
> 50 would suggest that matrix-fracture transfer is rapid enough to
fe
m
A
k
A k o
=
10 CSUG/SPE 146580
keep up with depletion in the fractures throughout the drainage area. This is true only for single-phase depletion after pseudo-
steady state is reached, but it does give a magnitude for relative comparison. Larger values mean that the matrix-fracture
transfer can more likely keep up with fracture depletion. If we have competing mass transfer processes (e.g. pressure
depletion in fractures and diffusion from matrix to fractures), then the ratio may need to include other dimensionless
parameters such as a capillary or bond number to better estimate the effectiveness of matrix-fracture transfer relative to
fracture transport.
We now present a few examples of the new plot. In the plot examples here, we suggest that the inter-porosity parameters data
be broken down into a number of equalized bins (e.g. ranges of inter-porosity flow ratio). For our examples we used 10%
probability ranges. This is similar to the idea for carbonate classification (Lucia, 1999), where the phi-k data is divided into
different flow indicator types. All of the examples presented here are from actual NFR with 1) rigorously defined matrix
reservoir characterization, 2) fractures networks defined via DFN and/or CFM methods calibrated to seismic, logs and/or
matrix characteristics, and 3) calibrated to dynamic performance. The comparisons are all made assuming a 160-acre well
spacing to provide consistent inter-porosity flow ratio. The results shown are only for a limited subset of the main reservoir
around some of the most highly fractured areas. Figure 3 compares the first four examples.
Example 1: Carbonate, Light Oil, Pressure Depletion via Horizontal and Vertical Wells, Moderate Aquifer Influx
This reservoir has moderate matrix permeability (<10md) but is a highly productive oil reservoir as a result of natural
fracturing. Fractures were characterized via image logs to define mechanical units and controls on fracture intensity. Seismic
attributes such as distance to faults, brittleness, semblance, curvature and anisotropy were combined to provide either direct
correlation to P32 and thus permeability and porosity or calibrated via DFN models to match well P10 and account for
variability in fracture height, length orientation and transmissivity. The models were calibrated to continuous measures of
pressure, oil rate and water. The plot shows large variability in fracture-matrix porosity and permeability ratios as a result of
high variability in fracture intensity. Fracture-to-matrix permeability can vary by more than 3 orders of magnitude, typical of
many conventional fractured reservoirs. Fracturing here is strongly related to fault proximity. Where fracture permeability is
high we also see high fracture intensity and thus the inter-porosity flow ratio also shows a large range of values. Such a
system might require a dual-permeability representation to account for flow through the matrix in less fractured areas.
Example 2: Carbonate, Light Oil, Pressure Maintenance via Gas and Water Injection in Horizontal Wells
This is a low permeability carbonate (several md). Fracturing here is also strongly related to fault proximity; however, there
is significant vertical variation via stratigraphic mechanical units as well as structural relationships. Horizontal wells provide
high early productivity when intersecting open fractures. Water injection has provided significant recovery especially in the
less fractured peripheral areas. Gas injection in the crest provides additional oil recovery via-gas-oil gravity drainage.
Although excess fracture permeability is not as high as in Case 1, we still see 3 orders of magnitude variation, which has
resulted in high variability in well performance and early breakthrough of injection water and gas. This reservoir has been
modeled via dual-permeability approaches.
Example 3: Sandstone, Heavy Oil, Waterflood via Vertical Wells
This third example also shows high variability in fracture excess permeability. Matrix permeability for this sandstone
reservoir is quite good (~100 md), but oil viscosity is high resulting in similar oil mobility to the previous two cases. This is
an older field that is primarily developed on vertical wells and thus productivity is lower than the horizontal well cases. High
dip has allowed for economic oil recovery via down dip water injection. Early water breakthrough and high water cuts are
common. Again, excess permeability varies by three orders of magnitude a common theme with conventional naturally
fractured reservoirs. This reservoir has been modeled via dual-permeability approaches in order to account for slow matrix-
fracture water transfer.
Example 4: Quartzite Sand, Low Matrix Permeability, Gas Reservoir with Vertical and Horizontal Wells
This reservoir could be considered a tight gas reservoir (microdarcy perm) if it was not for the extensive fracture system. The
fracture network leads to high potential rates in both vertical and horizontal wells. The excess permeability does not appear
to be as heterogeneous as the previous examples, but this could be because the system has only undergone depletion. The true
measure of heterogeneity often reveals itself during displacement processes. The high fracture intensity and single-phase
depletion means that the system essentially acts like a high permeability single-porosity system although it has been modeled
as a dual-porosity reservoir.
Example 5: Unconventional, Light Oil, Pressure Depletion via Hydraulically Fractured Horizontal Wells
This is an unconventional oil reservoir developed via hydraulically fractured horizontal wells. The industry is only beginning
to characterize these reservoirs. We believe that many of the same conventional characterization methods can be applied to
unconventional reservoirs. This includes calibration to image logs via DFN; correlation to seismic attributes and calibration
of these attributes to well performance. Such systems can show extreme variability over a play, but looking within local
sweet spots, we see the type of characteristics shown in Figure 4(A): lower range in excess permeability. Surprisingly, the
CSUG/SPE 146580 11
inter-porosity flow values are quite high, because of the large matrix-fracture surface area. Dual-porosity or dual-permeability
models are required to match early pressure depletion in the near-well shattered rock volume, but later time declines are
indicative of a total system response with effective permeability one to two orders of magnitude greater than matrix
permeability. Of course, this is for one specific example only and is not a general rule.
Example 6: Unconventional Light Oil/Condensate with Free Gas Phase, Pressure Depletion via Hydraulically
Fractured Horizontal Wells (single-well characterization)
A recent single-well calibration in a different unconventional play is also shown in Figure 4(B). At first glance it looks to
have similar characteristics to the previous plot. However, here productivities are quite low. This is the difficulty in simply
using dimensionless parameters for comparison. There are many other factors besides excess permeability which dictate the
well performance. Other concerns include PVT characteristics, initial pressure, thickness, number of frac stages, and
horizontal well length. Thus such comparisons as provided by these dimensionless plots are only one aspect of an overall
integrated characterization and simulation. The dimensionless plots are most useful when trying to make development
decisions such as well spacing within one specific unconventional play.
Example 7: Synthetic Example of Conventional Reservoir with Strong Fault-Related Fracturing
This final short example shows the importance of shape factor (or fracture intensity). This is a synthetic example based on
concepts from the previous examples. It has moderate API oil and limited aquifer drive. It has a strong fault related fracturing
along several large faults. The Gilman plot is shown in Figure 5(A). The band running through the plot is a result of several
more highly fractured zones. The excess permeability ratio is generally higher than the other cases. A 10-year, multi-well
simulation of depletion is shown in Figure 5(B). All cases have the same properties except that shape factor is increased by a
factor of 10 or decreased by a factor of 10 compared to the base case relations. The main differences in decline are a result of
differences in water production. Of note here is the large long-term difference in oil production; however, the first 3-years
show very minimal differences. This is typical of the types of responses we see in NFR. The short term impact of fracture
spacing differences are difficult to determine from production response, but have a significant impact on long-term behavior.
Differences are seen more quickly in injection processes.
The final comparison of the average properties for the different cases is shown in Figure 6. Table 1 provides the data. These
figures are a plot of the average values from Table 1. Averages themselves are not sufficient for estimating reservoir
response; however, they do more clearly show the relative differences between different reservoirs. There is generally an
increasing trend of excess permeability with increasing storage ratio as expected. Case 4 is an outlier because of the very low
matrix porosity. Figure 6(B) shows lower inter-porosity flow ratio as excess permeability increases. This trend is not strong
because as excess permeability increases, matrix shape factor may also increase. For all these cases, the average inter-
porosity flow parameter is greater than the value of 50 that might indicate rapid matrix-fracture equilibration for single-phase
flow. However, as shown in Figure 6(B), the long-term behavior is still very sensitive to the value of
A
. This is because of
limited fluid transfer rate for capillary and gravity dominated flow relative to transport through the fracture network.
Summary
Performance evaluation, laboratory studies and modeling of fields with long histories has provided much information about
the nature of the fractured networks, fluid transfer mechanisms and indications of how the effective network properties relate
to directly measurable properties like image and production logs, pressure transients and tracer tests. However, the
characterization and modeling process still requires extensive calibration with dynamic data. Therefore we still must often
rely on analogue reservoirs and prior experiences to estimate the flow characteristics and future behavior in NFR for fields
with less production history. Much emphasis in modern workflows is being placed on improvements in characterization of
the fracture network; however, reservoir flow behavior is also very dependent on the effectiveness of matrix-fracture transfer
mechanisms.
As a general recommendation for NFR simulation, is it important to identify intrinsic (e.g. lithologic) and extrinsic (e.g.
structural) controls on fracture distribution and intensity and relate these to well performance (e.g. productivity and water
breakthrough) in order to develop algorithms for distributing the fracture network properties in a 3-D static geomodel. The
simulation model must honor the variability and should not be an overly simplified model in order to capture the
heterogeneous nature of the reservoirs. Fracture-matrix transfer must be studied via laboratory experimentation, fine-grid
modeling and analogous reservoirs.
The new classification plot illustrated here provides a means to compare characteristics of different reservoirs including the
very important aspect of matrix-fracture transfer. Inter-porosity fluid transfer is not directly incorporated in the Nelson plot.
Our new classification concept can be used to understand differences in dynamic performance of NFR for purposes of
analogue selection or screening of potential recovery processes. We compared several characterized NFR reservoirs
(including unconventional systems) via the new classification plot and their differences and performance were summarized.
12 CSUG/SPE 146580
Acknowledgements
As mentioned in the Introduction, many of the background concepts were a result of unpublished collaborative efforts with
Dr. H. Kazemi, Colorado School of Mines. We would like to acknowledge his many contributions to the ideas presented
here. We would also like to acknowledge Ron Nelson, PhD, Broken N Consulting for providing a modified schematic of his
well-known distribution of fracture reservoir types.
Nomenclature
A area, ft
2
B formation volume factor (FVF), RB/STB
b harmonic decline factor
c compressibility, psia
-1
d distance, ft
D depth, ft
D decline factor, 1/day
h height or thickness, ft
k permeability, md
L matrix block dimension, ft
ln natural logarithm
n exponent, power-law equation
p pressure, psia
P10 linear fracture intensity measure, 1/L
P32 volumetric fracture intensity factor, 1/L
P33 volume of fractures per unit volume
q flow rate, STB/D
r radius, ft
S saturation, fraction
t time, days
V volume, ft
3
w width, ft
x distance in x-direction, ft
y distance in y-direction, ft
z distance in z-direction, ft
A difference or change in values
| porosity, fraction
interfacial tension, dyne/cm
o difference in value
inter-porosity flow parameter, dimensionless
mobility, md/cp
viscosity, cp
t Pi, constant
density, lb/ft
3
o shape factor, ft
-2
t fracture-matrix flow rate, STB/D
e storativity ratio, dimensionless
c partial derivative
V gradient operator
V divergence operator for a vector
Subscripts
A area
c capillary
D dimensionless
e effective, external
exr excess ratio
f fracture
g gas
i initial
CSUG/SPE 146580 13
i,j,k indices
m matrix
o oil
p effective porosity
qmax maximum rate
r relative
t total
th threshold
v vertical
w water
x,y,z directions
References
Araujo, H., Gilman, J. R., and Kazemi, H.: "Analysis of Interference and Pulse Tests in Heterogeneous Naturally Fractured Reservoirs",
SPE 49234 presented at the 1998 SPE Annual Technical Conference and Exhibition, New Orleans, LA (Sep. 27-30, 1998).
Arevalo-V., J. A., Samaniego-V., F., Lopez-C., F. F., Urquieta-S., E.: On the Exploitation Conditions of the Akal Reservoir Considering
Gas Cap Nitrogen Injection SPE 35319, Intl. Petroleum Conference & Exhibition of Mexico, Villahermosa, Mexico, (March 5-7.
1996).
Ata, E. and Michelena, R. J.: Mapping distribution of fractures in a reservoir with P-S converted waves: The Leading Edge, 14, p.664-673
(1995)
Beliveau, D., Payne, D. A. and Mundry, M.: Waterflood and CO2 flood of the Fractured Midale Field, J. Pet. Tech, 881-887,
(September, 1993).
Campanella, J. D., Wadleigh, E. E., and Gilman, J. R.: Flow CharacterizationCritical for Efficiency of Field Operations and IOR, SPE
58996 presented at the 2000 SPE International Petroleum Conference and Exhibition in Mexico, Villahermosa, Mexico (Feb. 1-3,
2000).
Chang, M-M: Deriving the Shape Factor of a Fractured Rock Matrix, U.S. Department of Energy (National Institute for Petroleum and
Energy Research, Bartlesville, OK, NIPER-696 (DE93000170) (Sept. 1993).
Chen, H. Y., Poston, S. W., and Wu, C. H.: "Characterization of the Austin Chalk Producing Trend," SPE 15533 presented at the 61st SPE
Annual Technical Conference, New Orleans, LA (Oct. 1986).
Chen, H. L., Lucas, L. R., Nogaret, L. A. D., Yang, H. D., and Kenyon, D. E.: Laboratory Monitoring of Surfactant Imbibition With
Computerized Tomography, SPE Res. Engr. and Eval., (February 2001).
Christiansen, R. L., Gilman, J. R., Jargon, J. R., Kazemi, H., and Smith, R. E.: "Enhanced Oil Recovery in Naturally Fractured Reservoirs,"
AIChE 51d presented at the 1989 AIChE Summer National Meeting Philadelphia, PA (August 1989).
Dershowitz, W., LaPointe, P., Eiben, T., and Wei, L.: Integration of Discrete Fracture Network Methods with Conventional Simulator
Approaches, SPE Res. Eval. & Eng., 165-170, (April 2000).
Dershowitz, W., Hermanson, J., Follin, S. and Mauldon, M.: "Fracture intensity measures in 1-D, 2-D, and 3-D and sp, Sweden,Pacific
Rocks 2000, Girard, Liebman, Breeds & Doe (eds) c 2000 Balkema, Rotterdam, ISBN 90 5809 155 4. pg 849-853.
Dykstra, H.: "The Prediction of Oil Recovery by Gravity Drainage", J. Pet. Tech. 818-830, (May 1978).
Gale, Julia F. W.: Specifying Lengths of Horizontal Wells in Fractured Reservoirs, SPE Res. Eval. & Eng., 266-272, (June 2002).
Gauthier, B. D. M., Garcia, M., and Daniel J. -M.: Integrated Fractured Reservoir Characterization: A Case Study in a North Africa
Field, SPE Res. Eval. and Engr., 284-294 (Aug. 2002).
Gilman, J. R. and Kazemi, H.: "Improvements in Simulation of Naturally Fractured Reservoirs," Soc. Pet. Engr. J. 695-707 (Aug. 1983).
Gilman, J. R. and Kazemi, H.: "Improved Calculations for Viscous and Gravity Displacement in Matrix Blocks in Dual-Porosity
Simulators," J. Pet. Tech, 60-70 (Jan. 1988).
Gilman, J. R., Tiss, M. Jargon, J. R., and Hinchman, S. B.: Dual-Porosity Simulation of the Zebbag Reservoir, Ezzaouia Field, Presented
at the Second ECLIPSE International Forum, Houston, TX, (April 15-19, 1996).
Gilman, J. R.: "Practical Aspects of Simulation of Fractured Reservoirs ", presented at the Seventh International Forum on Reservoir
Simulation, Schlosshotel Bhlerhhe, Bhl/Baden-Baden, Germany (June 23rd - 27th, 2003).
14 CSUG/SPE 146580
Hallenbeck, L.D., Sylte, J.E., Ebbs, D.J., Thomas, L.K.: Implementation of the Ekofisk Field Waterflood SPE Form. Eval., , p. 284,
(1991).
Haws, G. W. and Hurley, N. F.: "Applications of Pressure-Interference Data in Reservoir Characterization Studies, Big Horn Basin,
Wyoming", SPE 24668 presented at the 67th SPE Annual Technical Conference and Exhibition, Washington, DC, (Oct. 4-7, 1992).
Heinemann, Z.E. and Mittermeir, G.M. Rigorous Derivation of the Kazemi-Gilman-Elsharkawy Generalized Dual Porosity Shape Factor
Paper B044 presented at the 10th European Conference on the Mathematics of Oil Recovery - Amsterdam, The Netherlands, (Sept. 4-
7, 2006)
Hossain, M. M., Rahman, M. K. and Rahman, S. S.: A Shear Dilation Stimulation Model for Production Enhancement from Naturally
Fractured Reservoirs, SPE Journal, 183-195, (June 2002).
Iwere, F.O., Moreno, J. E., Apaydin, O. G., Len Ventura, R., Garcia, J. L. Vug Characterization and Pore volume Compressibility for
Numerical Simulation of Vuggy and Fractured Carbonate Reservoirs SPE 74341, SPE International Petroleum Conference and
Exhibition, Villahermosa, Mexico, (Feb. 2002).
Kazemi, H., Merrill, L. S., Porterfield, K. L., and Zeman, P. R.: Numerical simulation of Water-Oil flow in Naturally Fractured Reservoirs,
Soc. Pet. Eng. J. 317-26, (Dec. 1976).
Kazemi, H., Gilman, J. R., and El-Sharkawy, A. M.: "Analytical and Numerical Solution of Oil Recovery from Fractured Reservoirs Using
Empirical Transfer Functions," Soc. Pet. Eng. Res. Engr., 219-227, (May 1992).
Kazemi, H. and Gilman, J. R.: "Chapter 6. Multiphase Flow in Fractured Petroleum Reservoirs", in Flow and Contaminant Transport in
Fractured Rock, Edited by J. Bear, C-F. Tsang, and G. de Marsily, Academic Press, San Diego, CA, 267-323 (1993).
Kazemi, H. and Shinta, A. A.: Determining Orientation and Conductivity of High Permeability Channels in Naturally Fractured
Reservoirs, Reservoir Characterization IV, Edited by Bill Linville, Pennwell Books, (1993).
Kyte, J. R.: A Centrifuge Method To Predict Matrix-Block Recovery in Fractured Reservoirs, Soc. Pet. Eng. J., p. 164, (1970)
Lake, Larry W.: Enhanced Oil Recovery, Prentice Hall, p. 18, (1989).
Lim, K. T. and Aziz, K.: Matrix-Fracture Transfer Shape Factor for Dual-Porosity Simulation, J. of Petroleum Science and Engineering,
169-178, (1995).
Lucia, F. Jerry, Carbonate Reservoir Characterization, Springer-Verlag Berlin Heidelberg, (1999)
MacDonald, A. E., Beckner, B. L., Chan, H. M., Jones, T. A. and Wooten, S. A.: Some Important Considerations in the Simulation of
Naturally Fractured Reservoirs, SPE 21814 presented at the Rocky Mountain Regional Meeting and Low-Permeability Reservoir
Symposium, Denver, CO (April 15-17, 1991).
Malik, Q. M., and Islam, M. R.: CO2 Injection in the Weyburn Field of Canada: Optimization of Enhanced Oil Recovery and Greenhouse
Gas Storage With Horizontal Wells SPE 59327, SPE/DOE Improved Oil Recovery Symposium, Tulsa, Oklahoma, (April 3-5, 2000)
Maloney, D. and Doggett, K.: Multiphase Flow in Fractures DOE Contract DE-AC22-94PC91008, Report SCA 9730, (1977).
Mattax, C. C., and Kyte, J. R.: "Imbibition Oil Recovery from Fractured, Water-Drive Reservoirs", Soc. Pet. Eng. J., 177-184, (June 1962).
Nelson, R. A.: Geologic Analysis of Naturally Fractured Reservoirs, Gulf Professional Publishing Co., (2
nd
Ed. 2001).
Oda, M.: Permeability Tensor for Discontinuous Rock Masses, Geotechniques, V. 35, 483-495, (1985).
O'Neill, Niel: Fahud Field Review: A Switch from Water to Gas Injection J. Pet. Tech, p. 609, (1988).
Pereira, C, Kazemi, H., Ozkan, E.: Combined Effect of Non-Darcy Flow and Formation Damage on Gas Well Performance In Dual-
Porosity/Dual Permeability Reservoirs SPE 90623 presented at SPE ATCE, September 28. (2004)
Penuela, G., Civan, F., Hughes, R. G. and Wiggins, M.L.: Time-Dependent Shape Factor for Interporosity Flow in Naturally Fractured
Gas-Condensate Reservoirs, SPE 75524, SPE Gas Technology Symposium, Calgary, (April 30-May 2, 2002).
Phelps, R. E. and Strauss, J. P.: Capturing Reservoir Behavior by Simulating Vertical Fracture and Super-K Zones in the Ghawar Field,
SPE Res. Eval. & Eng, 333-340, (Aug. 2002).
CSUG/SPE 146580 15
Reis, J. C., Miller, M. A.: Oil Recovery from Naturally Fractured Reservoirs by Steam Injection Methods, DOE Contract DOE/BC-
14661-10, DE95000140, (May 1995).
Richardson, J. G., Sangree, J. B., and Sneider, R. M.: "Oil Recovery by Gravity Segregation", J. Pet. Tech., 581-582, (June 1989).
Rothkopf, B., and Wadleigh, E.: Re-tooling Improves Field Efficiency, The American Oil and Gas Reporter, 91-95, (Sept., 1994).
Saidi, Ali M.: Twenty Years of Gas Injection History into Well-Fractured Haft Kel Field (Iran) SPE 35309, Intl. Petroleum Conference
& Exhibition of Mexico, Villahermosa, Mexico, (March 5-7. 1996).
Sarma, P., Aziz, K.: New Transfer Functions for Simulation of Naturally Fractured Reservoirs with Dual-Porosity Models paper SPE
980231 presented at the SPE Annual Technical Conference and Exhibition, Houston, Texas, (26-29 September, 2003).
Satman, A.: "Decline Curve Analysis for Naturally Fractured Reservoirs: A Comparison of Models" unsolicited paper SPE 14473, (April
19, 1985).
Sydansk, R. D. and Moore, P. E.: Gel Conformance Treatments Increase Oil Production in Wyoming, Oil and Gas J. V. 90, N. 3, 40-45,
(Jan. 20, 1992).
Wang, X.: "Stereological Interpretation of Rock Fracture Traces on Borehole Walls and Other Cylindrical Surfaces" PhD Dissertation,
Virginia Polytechnic Institute and State University September 16, 2005, Blacksburg, Virginia
Warren, J. E. and Root, P. J.: "The Behavior of Naturally Fractured Reservoirs," Soc. Pet. Eng. J. 245-255, (Sept. 1963).
Zhang, X., Morrow, N. R., and Ma, S.: Experimental Verification of a Modified Scaling Group for Spontaneous Imbibition, SPE Res.
Eng., V. 11, 280-285, (1996).
Zimmerman, Chen, Hadgu and Bodvarsson: A Numerical Dual-Porosity Model with Semi-Analytical Treatment of Fracture/ Matrix
Flow, Water Resources Research, V. 29, N. 7, 2127-2137, (July 1993).
Table 1. Average properties for all cases
Case kfe |f
km |m o, 1/ft
2
L=
(8/o)
0.5
ft
|f/(|f+|m) kfe/km kmo, md/ft
2
A=
(Akmo)/kfe
+
Case 1 1.97E+01 0.0004 2.53E+00 0.1149 0.0152 40.4 0.0033 29.73 3.52E-02 13653.30
Case 2 6.59E+01 0.0038 3.51E+00 0.2364 0.0284 21.4 0.0177 49.70 8.56E-02 10541.45
Case 3 4.64E+02 0.0054 5.48E+01 0.1250 0.0407 16.2 0.0456 66.60 2.19E+00 33423.98
Case 4 1.99E+00 0.0037 1.16E-02 0.0280 0.0050 92.5 0.0872 411.51 3.00E-05 202.67
Case 5 1.08E-01 0.0033 5.06E-03 0.0909 0.0013 82.4 0.0362 29.26 6.95E-06 413.15
Case 6 4.01E-04 0.0010 3.52E-05 0.0770 0.0032 50.0 0.0131 14.04 1.13E-07 1957.22
Case 7 5.26E+01 0.0058 1.38E+00 0.1449 0.0070 47.5 0.0387 926.13 6.75E-03 1276.83
+A=160acres for all cases
16 CSUG/SPE 146580
Figure 1. Idealization of a fractured system
with fracture spacing Lx and fracture width wf
Figure 3. Gilman Plot for conventional reservoirs, Cases 1 to 4. A) Carbonate, light oil, pressure depletion via horizontal and vertical
wells, moderate aquifer influx, B)Carbonate, light oil, pressure maintenance via gas and water injection in horizontal wells, C)
Sandstone, heavy oil, waterflood via vertical wells; D)Quartzite, low matrix permeability, gas reservoir with vertical and horizontal
wells
L
x
L
x
w
f
Gilman Plot
1.00
10.00
100.00
1000.00
0.0000 0.0200 0.0400 0.0600 0.0800 0.1000
f
f
/(f
f
+f
m
)
k
f
e
/
k
m
>0. to 1478.
>1478. to 4441.
>4441. to 9510.
>9510. to 15581.
>15581. to 22050.
>22050. to 29327.
>29327. to 39029.
>39029. to 51233.
>51233. to 74122.
>74122. to 229912.
lA=(Adkms)/kfe
Intellectual Property of iReservoir.com, Inc
0.10
1.00
10.00
100.00
1000.00
0.0000 0.0100 0.0200 0.0300 0.0400
k
f
e
/
k
m
f
f
/(f
f
+f
m
)
Gilman Plot
>0. to 2163.
>2163. to 5056.
>5056. to 28388.
>28388. to 68730.
>68730. to 135408.
>135408. to 213857.
>213857. to 341834.
>341834. to 622681.
>622681. to 6452860.
>6452860. to 793982524.
lA=(Adkms)/kfe
Intellectual Property of iReservoir.com, Inc
Figure 2. Modified schematic distribution of fracture
reservoir types I-III relative to reservoir porosity and
permeability. Courtesy Ron Nelson, PhD, Broken N
Consulting (after Nelson, 2001).
1.00
10.00
100.00
1000.00
0.0000 0.0200 0.0400 0.0600 0.0800 0.1000
k
f
e
/
k
m
f
f
/(f
f
+f
m
)
Gilman Plot
>0. to 16.
>16. to 19.
>19. to 22.
>22. to 25.
>25. to 30.
>30. to 37.
>37. to 47.
>47. to 58.
>58. to 91.
>91. to 33739.
lA=(Adkms)/kfe
Intellectual Property of iReservoir.com, Inc
0.10
1.00
10.00
100.00
1000.00
0.0000 0.0050 0.0100 0.0150 0.0200
k
f
e
/
k
m
f
f
/(f
f
+f
m
)
Gilman Plot
>0. to 735.
>735. to 1349.
>1349. to 2389.
>2389. to 4118.
>4118. to 6993.
>6993. to 11847.
>11847. to 20576.
>20576. to 39851.
>39851. to 90546.
>90546. to 872404.
l
A
=(A
d
k
m
s)/k
fe
Intellectual Property of iReservoir.com, Inc
I
Type II
III
%of Total Porosity
%
o
f
T
o
t
a
l
P
e
r
m
e
a
b
i
l
i
t
y
100 % k
f
100% k
m
100% f
m
100% f
f
M
All Fractures
Little or No Fractures
A B
C
D
CSUG/SPE 146580 17
Figure 4. Gilman Plot for unconventional oil reservoirs, Cases 5 and 6: A) Unconventional, light oil, pressure depletion via
hydraulically fractured horizontal wells; B) Unconventional, light oil/condensate with free gas phase, pressure depletion via
hydraulically fractured horizontal wells (single-well characterization)
Figure 5. A) Gilman Plot for synthetic example and B) Normalized production for 10-yr forecast with shape factor increased by a
factor of 10 or decreased by a factor of 10 compared to base simulation
Figure 6. Comparison of A) average values for all cases in Nelson Plot and B) plot of inter-porosity flow ratios
Gilman Plot
1.00
10.00
100.00
1000.00
0.0000 0.0200 0.0400 0.0600 0.0800 0.1000
f
f/(f
f+f
m)
k
f
e
/
k
m
>0. to 152.
>152. to 202.
>202. to 248.
>248. to 296.
>296. to 358.
>358. to 449.
>449. to 575.
>575. to 755.
>755. to 1007.
>1007. to 4969.
lA=(Adkms)/kfe
Intellectual Property of iReservoir.com, Inc
1.00
10.00
100.00
1000.00
0.000 0.005 0.010 0.015 0.020
k
f
e
/
k
m
f
f
/(f
f
+f
m
)
Gilman Plot
>0. to 1001.
>1001. to 1112.
>1112. to 1224.
>1224. to 1390.
>1390. to 1613.
>1613. to 1891.
>1891. to 2225.
>2225. to 2781.
>2781. to 3615.
>3615. to 4283.
l
A
=(A
d
k
m
s)/k
fe
Intellectual Property of iReservoir.com, Inc
1.00
10.00
100.00
1000.00
10000.00
0.0000 0.0200 0.0400 0.0600 0.0800 0.1000
k
f
e
/
k
m
f
f
/(f
f
+f
m
)
Gilman Plot
>0. to 14.
>14. to 34.
>34. to 71.
>71. to 139.
>139. to 265.
>265. to 464.
>464. to 855.
>855. to 1521.
>1521. to 3286.
>3286. to 47463.
lA=(Adkms)/kfe
Intellectual Property of iReservoir.com, Inc
Case 1
Case 2
Case 3
Case 4
Case 5
Case 6
Case 7
1.00
10.00
100.00
1000.00
0.0000 0.0200 0.0400 0.0600 0.0800 0.1000
k
f
e
/
k
m
f
f
/(f
f
+f
m
)
Case 1
Case 2
Case 3
Case 4
Case 5
Case 6
Case 7
10.00
100.00
1000.00
10000.00
100000.00
1.00 10.00 100.00 1000.00
l
A
=
(
A
k
m
s
)
/
k
f
e
k
fe
/k
m
A B
A B