IPTC 18194

Fractured Reservoir Modeling and Interpretation

Fikri Kuchuk, SPE, Denis Biryukov, and Tony Fitzpatrick, Schlumberger
Copyright 2014, International Petroleum Technology Conference
This paper was prepared for presentation at the International Petroleum Technology Conference held in Kuala Lumpur, Malaysia, 10-12 December 2014.
This paper was selected for presentation by an IPTC Programme Committee following review of information contained in an abstract submitted by the author(s). Contents of the paper, as presented, have not been reviewed by
the International Petroleum Technology Conference and are subject to correction by the author(s). The material, as presented, does not necessarily reflect any position of the International Petroleum Technology Conference, its
officers, or members. Papers presented at IPTC are subject to publication review by Sponsor Society Committees of IPTC. Electronic reproduction, distribution, or storage of any part of this paper for commercial purposes without
the written consent of the International Petroleum Technology Conference 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 where and by whom the paper was presented. Write Librarian, IPTC, P.O. Box 833836, Richardson, TX 75083-3836, U.S.A., fax +1-972-952-9435

Abstract. Fractures are common features of many well-known reservoirs. Naturally Fractured Reservoirs (NFR)
consist of fractures in igneous, metamorphic, and sedimentary rocks (matrix). Faults in many naturally fractured
carbonate reservoirs often have high-permeability zones and are connected to numerous fractures with varying con-
ductivities. In many NFRs, faults and fractures frequently have discrete distributions rather than connected fracture
networks. Because fractures are often created by faulting, faults and fractures should be modeled together. Ac-
curately modeling naturally fractured reservoir pressure transient behavior is important in hydrogeology, the earth
sciences, and petroleum engineering, including ground-water contamination to shale gas and oil reservoirs. For more
than 50 years, conventional dual-porosity type models, which do not include any fractures, have been used for model-
ing fluid flow in naturally fractured reservoirs and aquifers. They have been continuously modified to add unphysical
matrix block properties such as the matrix skin factor.
In general, fractured reservoirs are heterogeneous at different length scales. It is clear that even with millions
of grid blocks, numerical models may not be capable of accurately simulating the pressure transient behavior of
continuously and discretely NFRs containing variable conductivity fractures. The conventional dual-porosity type
models are obviously an oversimplification; their serious limitations and consequent implications for interpreting well
test data from NFRs are discussed in detail. These models do not include wellbore-intersecting fractures, even though
they dominate the pressure behavior of NFRs for a considerable length of testing time. Fracture conductivities of
one to infinity dominate transient behavior of both continuously and discretely fractured reservoirs, but again dual-
porosity models do not contain a single fracture. Our fractured reservoir model is capable of treating thousands of
fractures that are periodically or arbitrarily distributed with finite- and/or infinite-conductivities, different lengths,
densities, and orientations.
Appropriate inner boundary conditions are used to account for wellbore-intersecting fractures and direct wellbore
contributions to production. Wellbore storage and skin effects in bounded and unbounded systems are included in
the model. Three types of damaged skin factors that may exist in wellbore-intersecting fracture(s) are specified.
With this highly accurate model, the pressure transient behavior of conventional dual-porosity type models are
investigated, and their limitations and range of applicabilities are identified. The behavior of the triple-porosity
models are also investigated. It is very unlikely that triple-porosity behavior is due to the local variability of matrix
properties at the microscopic level. Rather, it is due to the spatial variability of conductivity, length, density, and
orientation of the fracture distributions.
Finally, we have presented an interpretation of a field buildup test example from an NFR using both conventional
dual-porosity models and our fractured reservoir model.

Introduction
Fractures occur in all subsurface and surface rock formations and are common features in many aquifers and well-
known hydrocarbon reservoirs. Naturally fractured reservoirs contain a large percentage of the hydrocarbon reserves
of the world and are very prolific producers. All geological formations are fractured to some extent as a result of stress
triggered by the Earth’s crust weight, high fluid pressure, tectonic forces, and/or thermal loading. Fractures occur at
a variety of scales from microscopic to continental. NFRs include fractured igneous, metamorphic, and sedimentary
rocks (matrix). Faults in many naturally fractured carbonate reservoirs often have high-permeability zones and are
connected to numerous fractures that have varying conductivities. Furthermore, in many NFRs, faults and fractures
can be discrete rather than the connected network dual-porosity systems. The Committee on Fracture Characteri-
zation and Fluid Flow (1996) stated that fractures create traps, act as conduits to oil and gas migration, can present
barriers or baffles to fluid flow, and have a significant influence on the behavior and performance of reservoirs. Here
the word “fractures” includes both fractures and faults. The Committee on Fracture Characterization and Fluid Flow
(1996) presented a comprehensive overview that covers most geological, hydrogeological, and fluid flow aspects of
2 IPTC 18194

fractures. Berkowitz (2002) presented an extensive review of flow and transport properties in fractured porous media
in the geology, geothermal systems, hydrogeology, and petroleum engineering literature.
Accurately modeling naturally fractured reservoir pressure transient behavior is important in hydrogeology, the
earth sciences, and petroleum engineering: in terms of ground-water contamination, waste-water, nuclear waste
disposal, hot-dry rock geothermal systems, and shale gas and oil reservoirs. These include understanding different
spatial scales, from pores to fracture corridors. In fact, modeling fractured formations has been going on for almost a
century, and appropriate algorithms have been developed and continuously improved to include more heterogeneity of
NFRs. Characterizing NFRs is difficult and cannot be achieved by any single discipline or measurement. It is shown
that their equivalent-medium representation has serious shortcomings. One of the real challenges is determining
spatial distributions of fractures using petrophysical and geophysics data with dynamic reservoir data, such as
pressure transient test and production data.
Naturally fractured (also called two-porosity, double porosity, dual-porosity, or fissured ) reservoirs can be de-
scribed as sedimentary formations fractured by tectonic, thermal, or chemical processes. In these reservoirs, Pressure
Transient (diffusion) Test (PTT) behavior is predominantly controlled by fracture properties: distribution, connec-
tivity, conductivity, and density. Matrix block properties to some extent also contribute to the overall behavior of
the system, provided that they are reasonably permeable. Nelson (1985) classified NFRs into four categories from
the reservoir point of view. With regard to reservoir simulations and performance as well as pressure transient test
interpretation, Kuchuk and Biryukov (2014) divided fractured reservoirs into the following four categories:

1. Fractures create a network, communicate hydraulically with each other globally, and provide overall conduc-
tivity (permeability) of the reservoir. The matrix provides overall storage capacity (porosity), but it should
be permeable enough to provide conductivity for the flow from the matrix into the fractures. We call these
continuously fractured reservoirs. They can also be called dual-porosity-permeability fractured reservoirs. In a
sense, both fracture and matrix have distinct porosities and permeabilities. Joints often create a set of orthog-
onal continuous fracture networks (Pollard and Aydin, 1988). NFRs frequently have nonorthogonal continuous
fracture sets (Pollard and Aydin, 1988), as well as secondary fractures that are nonorthogonal to each other
and to primary fracture sets.
2. Fractures do not form a continuous conductive network; only a limited number of fractures communicate
hydraulically with each other. Both fractures and matrix provide conductivity, but the overall storage capacity
(porosity) is in the matrix. These are usually called discretely fractured reservoirs. The combination of open
and partially sealing faults often creates discrete conductive systems. Stratified reservoirs that consist of both
fractured and nonfractured layers stacked vertically should also be called discretely fractured reservoirs. In
reservoirs with a continuous network of faults and/or fractures, assuming some of them are mineralized and/or
closed, this process also creates discretely fractured reservoirs. If matrix permeabilities are high or similar to
fracture permeabilities, or ultra low, the pressure transient behavior of fractured reservoirs will be similar to
the behavior of homogeneous reservoirs. Many discretely NFRrs have nonorthogonal discontinuous fracture
sets (Pollard and Aydin, 1988), while only a few have orthogonal ones.
3. Fractures are nonconductive (sealed due to mineralization and/or shale smearing due to faulting). In these
reservoirs, both conductivity (permeability) and storage capacity are provided by the matrix. These are called
compartmentalized reservoirs.

4. Fractures alone are conductive; the matrix has no permeability and/or porosity. Reservoirs with these fractures
are called unconventional fractured basement reservoirs.
The above definitions are applied to both conventional and unconventional NFRs, such as shale and tight gas and oil
formations. Although there are plenty of naturally fractured sandstone reservoirs, in general, fractures are abundant
in carbonate formations. Nelson (1985) gave an extensive list of fractured reservoirs, such as Spraberry Trend,
West Texas; Miocene Monterey Formation, California; Asmari Formation, Iran; Ain Zalah and Kirkuk Fields, Iraq;
Dukhan Field, Qatar; Cantarell, Mexico; and Fahud and Yibal, Oman. Many of these reservoirs contain continuously
and discretely fractured regions depending on the location, whether at the crest or flank. Daniel (1954) reported that
the fractures appear to be interconnected over the productive area in the Ain Zalah Field in Iraq, and the drainage
of the oil from the first and second pays can probably be achieved through two or three wells since all of the oil
enters wells through fractures. He also reported that Kirkuk Field in Iraq, with wells suitably located on the flanks
of the highest dome (Baba), could eventually drain the entire field. The network of fractures are widespread and
all-pervading, and the fractures are closely spaced. It is clear that these are continuously fractured reservoirs. He
also reported that the wells of Dukhan Field in Qatar will not give such large yields as the other two Iraqi Fields,
and that more wells are required to obtain the volume of production. It seems likely that its fractures are not closely
spaced, are rare (not high density), and appear to be more effective vertically than laterally. This behavior indicates
a discretely fractured reservoir.
IPTC 18194 3

Naturally fractured reservoir modeling is very challenging due to the complicated geology. The spatial resolution
of petrophysical log, core, and seismic, which we call geoscience data, is still very limited, although it has been
improving steadily—in particular, seismic (Freudenreich et al., 2006). Petrophysical logs and cores provide vertically
high-resolution near-wellbore static data. Acoustic logs and seismic provide the geological structure at different scales,
and high-vertical resolution wellbore seismic profiles (VSP) enhance the surface seismic geological interpretation.
Acoustic and seismic data can be further enhanced by using nearby outcrop analogues. Unfortunately, these data
are still not acquired systematically.
The second challenge to NFR modeling is the lack of 4D dynamic data, which makes it nearly impossible to
estimate medium (macroscopic)-scale reservoir flow properties. Currently, 4D numerical simulation models are de-
veloped using geoscience data, and then reservoir parameters are adjusted by history-matching production, pressure,
and water-cut data that are normally 2D (1D spatially and 1D temporally). This dimensional incompatibility causes
a long and arduous history-matching process, and often produces a nonunique reservoir model with unrealistic pa-
rameter distributions, i.e., several plausible models, all of which could satisfy the historical performance of the
field.
It should be stated from the outset that reservoir multiphase transport properties, e.g. relative permeabilities and
capillary pressures, are very important for reservoir management and EOR, etc. However, before obtaining multi-
phase transport properties, we must have a reasonably accurate reservoir model, along with single-phase transport
properties such as permeabilities and fracture conductivities. Casciano et al. (2004), Rogers et al. (2007), Casabi-
anca et al. (2007), and Morton et al. (2012) presented integrated interpretation methodologies for reservoirs with
natural fractures, incorporating openhole log and seismic data, and the preliminary geological reservoir model. This
is an important step towards reconciling static and dynamic reservoir data, and updating the geological reservoir
model with meaningful parameters. New measurements, provided that they are acquired systematically, and inver-
sion techniques are now almost sufficient for obtaining an accurate reservoir model and its single-phase transport
properties with a reasonable degree of certainty, principally due to these new techniques: 1) Accurate semi-analytical
pressure transient reservoir models for NFRs with arbitrarily distributed finite and/or infinite conductivity vertical
fractures and faults (Biryukov and Kuchuk, 2012), 2) Grid-based inversion methods for spatial feature identification
and parameter estimation (Booth et al., 2010, 2012), and 3) Global sensitivity analysis (Morton et al., 2013).
There is still, however, a great deal of uncertainty in the modeling of NFRs, particularly in terms of reservoir
properties and insufficient dynamic data. More importantly, almost 50 years of the incorrect treatment of fluid flow
in NFRs has been based on the antiquated Barenblatt et al. (1960) and Warren and Root (1963) sugar-cube (spheres,
slabs, cylinders, rectangular parallelepipeds, etc., elements are also used) dual-porosity models. Using the Warren
and Root (1963) technique, as shown in Fig. 12, a naturally fractured reservoir is transformed into an equivalent
(fictitious) homogenous fractured medium and an equivalent homogenous matrix medium using the s− > sf (s)
transformation, where f (s) = ω(1−ω)s+λ
(1−ω)s+λ , s is the Laplace transform variable, λ is the interporosity flow coefficient,
and ω is the storativity ratio. As Braester (2009) stated, naturally fractured reservoirs are heterogeneous, and most
fracture parameters normally exhibit power-law distributions. It is, therefore, unrealistic to model naturally fractured
reservoirs with only two ω and λ parameters, although there may be a few reservoirs that can be modeled using the
s− > sf (s) transformation. Over the years, many different forms of f (s) have been introduced, depending on the
matrix shape and/or the interporosity flow conditions. One of the objectives of this paper is to examine the validity
of the f (s) transformation.
For the sake of completeness, let us Matrix source element
introduce the Warren and Root (1963)
Matrix
and Kazemi (1969) models in Fig. 1. In
this figure, (b) shows an orthogonal set Matrix
Matrix
of continuous fracture systems with rect-
angular parallelepiped matrix blocks, which Matrix

is supposed to be a representation of the

real fractured system shown in (a). How-
Fractures
ever, Warren and Root (1963) did not Vugs
Fractures
use the model given in (b) for NFRs, but Wellbore Homogeneous reservoir
Actual reservoir Sugar cube representation Slab representation
instead used the 1D homogeneous model
(a) (b) Warren & Root model (c) Kazemi model (d) 1D Fictitious reservoir model
shown in Fig. 1 (d) with rectangular par-
allelepiped matrix blocks as source terms.
Kazemi (1969) solved the homogenous Figure 1: Idealization of a naturally fractured heterogeneous porous medium: (a)
multilayer model shown in (c), in which Actual reservoir, (b) the Warren and Root (1963) model, (c) the Kazemi (1969) model,
all high permeability layers are called “frac- and (d) the fictitious “equivalent” Warren and Root (1963) homogenous dual-porosity
reservoir model.
tures” intersect the wellbore horizontally;
but these fractures do not inconceivably exhibit any linear formation flow regimes.
The Warren and Root (1963) dual-porosity model is an equivalent (fictitious) homogenous medium, and does not
4 IPTC 18194

contain any fractures, as shown in Fig. 12. In any reservoir, if a pressure transient buildup test exhibits two parallel
straight lines (two identical infinite-acting flow regimes) on the Horner plot shown in Fig. 2 (Bourdet et al., 1983)
and/or its derivative first goes downward, second goes through a minimum, third goes upward, and finally flattens to
the second infinite-acting flow regime on a log-log plot (Fig. 3), then the reservoir is usually assumed to be naturally
fractured and the Warren and Root (1963) model is used. Fig. 3 also presents dimensionless pressure derivatives for
the pseudosteady-state and transient flow conditions in spherical matrix blocks, where λ = 10−7 and ω = 0.06, with
and without wellbore storage and skin (Kuchuk and Biryukov, 2014). For comparison, the same figure also presents
the dimensionless derivative for a homogenous system without wellbore storage and skin effects.

10 CD=0, S=0 & pseudosteady-state spherical blocks

CD=1000, S=10 & pseudosteady-state spherical blocks
8 CD=0, S=0 & transient spherical blocks
7 CD=1000, S=10 & transient spherical block
3905 6 CD=0, S=0 & homogenous model
5
slope = 14 psi/cycle
4

3
3900 unit slope
Shut-in pressure, psi

dpD/dlntD
first radial third radial
3895 1 second radial
slope =8.6 psi/cycle 8
7
6
5
3890 4

slope = 14 psi/cycle 3

2
3885 m = 1/2

0.1
3880 10
2
10
3
10
4
10
5
10
6
10
7
10
8
2 3 4 5 6 7 8 2 3 4 5 6 7 8 2
1 10 100 tD
(tp +Δt)/Δt
Figure 3: Dimensionless derivatives for a dual-porosity model with
Figure 2: Horner plot of the BU data given by Bourdet et al. (1983), and without wellbore storage and skin effects for various matrix block
after Kuchuk and Biryukov (2014). flow conditions.

The first pressure derivative plot for a field buildup (BU) test from a fractured reservoir was presented by Bourdet
et al. (1983), as shown in Fig. 4. The same plot also shows the result of the history match between the field BU data
and the dual-porosity (Warren and Root, 1963) model with a pseudosteady-state flow condition for the matrix blocks.
Fig. 4 shows that the match is quite good. As can be seen in Fig. 2, it is remarkable that three well-defined semilog
straight lines are identified on the Horner plot, although none of them is observed on the derivative plot shown in
Fig. 4. It can also be observed from this figure that the derivative (Fig. 3 of SPE 158096) exhibits a 1/2 slope linear
flow regime at very early times, and a second linear flow regime at late times. The Warren and Root (1963) model
does not exhibit any linear flow regime (see Fig. 3).

3
m = 1/2 2

2 m = - 1/2 m=1
m = 1/2
100
9
Derivative, psi

Derivative, psi

109 8
7
8 6
7
6 5
5 4
4
3
m = 1/3
3
2
measured
2 computed

10
1
0.1 1 10 100 1000
0.001 0.01 0.1 1 10 100
Shut-in time, hr Shut-in time, Δt, hr

Figure 4: Comparison of measured and computed pressure deriva- Figure 5: Pressure derivative of the BU data given in Casabianca
tive of the BU data given by Bourdet et al. (1983). et al. (2007) from a fractured chalk reservoir (Machar oilfield).

Casabianca et al. (2007) presented the derivative of a buildup test from the Machar oilfield, which is a fractured
chalk reservoir. An integrated interpretation methodology uses a Discrete Fracture Network (DFN) model to un-
derstand the derivative behavior of the BU test shown in Fig. 5. As can be noticed from this figure, the derivative
behavior of the buildup test of the Machar oilfield is similar to the derivative shown in Fig. 4, except that they have
1/2 and 1/3 slopes at the middle times, respectively.
Casciano et al. (2004) first stated that the uniform distribution of the fluid flow in fully connected fracture
networks is very seldom observed in a reservoir. They used DFN models for NFRs as alternatives to continuous
fracture network systems to interpret the well test data from a reservoir for which the derivative of a buildup test
is shown in Fig. 6. As can be seen from the derivative curve, there is a short positive unit-slope (m = 1) wellbore-
storage dominated period, lasting about 0.6 minute, then a 1/2 linear flow regime until 0.25 hr. A short duration
IPTC 18194 5

wellbore-storage dominated period is normally expected when a high conductivity fracture(s) intersects the wellbore
before the linear flow, regardless of fracture azimuth and dip angle. After the linear flow, the derivative exhibits
bilinear flow regimes for 5 hrs, then the derivative flattens before going down slightly (which could be real or simply
due to the producing time and/or boundary effect). We probably cannot conclude that the short flattening of the
derivative after 10 hrs is due to a radial flow regime.

4
100 m = 1/2
2

10
8
m = 1/4
Derivative, psi

6
10

(tD/CD)p'D
4

2
m=0
m = 1/2
1
8
1
6 m=1 m=1
4

0.1 0.1
1 2 3 4 5
0.001 0.01 0.1 1 10 100 10 10 10 10 10
Shut-in time, Δt, hr tD/CD

Figure 6: Pressure derivative of a buildup test from a fractured Figure 7: Pressure derivative of a buildup test from a fracture
carbonate reservoir, modified from Casciano et al. (2004). reservoir, modified from Ayestaran et al. (1989).

Another buildup test derivative was presented by Ayestaran et al. (1989), as shown in Fig. 7. The reservoir in this
example is a Tertiary limestone about 350 ft thick, which has a low matrix porosity but is extensively fractured (Zaman
et al., 1989). The buildup test derivative exhibits a very brief wellbore-storage dominated period without unit slope,
and then flattens (m = 0 slope) for a short time, perhaps because of a radial flow regime in the finite-conductivity
fractures that intersects the wellbore. After these two brief flow periods, the derivative exhibits a pseudosteady-state
flow regime (m = 1 slope) in the fracture system. This is simply due to the very high conductivities compared to
the low matrix permeabilities (kf  km ): as the fluid in the fractures is depleted, it is not replenished readily by the
matrix blocks. As can be observed in Fig. 7, the formation linear flow (the matrix contributions) starts while the
pseudosteady-state flow is progressing. This also implies that many fractures are communicating with each other.
Finally, a long linear flow regime is observed on the derivative plot.
When we compare the three derivatives of the well test data from different NFRs shown in Figs. 5, 6, and
7, these derivatives are entirely dissimilar. The first two reveals a DFN model and the third one reveals a fully
connected fracture network, at least within the volume that is investigated by the buildup test. These derivatives
have no similarity to the derivatives of the conventional dual-porosity type models shown in Fig. 3. In many well
test interpretations, therefore, most half-slope-linear flow regimes are attributed to faulted channel reservoirs, or to
reservoirs with two or three intersecting impermeable boundaries (faults).

Geological Setting
Most naturally fractured reservoirs, particularly carbonate formations, contain primary (macro), secondary, and
tertiary (minor) fractures, as shown in Fig. 8. (It should be noticed that this figure is just an illustration to show
the complexity of fractured formations; the fracture classification may not be geologically accurate.) The figure
also shows that fractured reservoirs are heterogeneous, and that fracture properties (length, aperture, orientation,
etc.) can significantly vary. As we stated above, the conventional dual-porosity type models for this type of NFR are
clearly an oversimplification and cannot be representative. Naturally fractured formations also contain microfractures
and vugs (small cavities in rock); the dimensions of both are normally less than 100 mm. NFRs are sometimes
hydraulically or acid fractured, or simply acidized. Another important fluid flow aspect of fractured reservoirs is
the presence of fracture corridors. These are clusters of parallel or near-parallel (globally in the same direction)
fractures. Fracture corridor size may vary from less than 1 m to 10 m wide, a few meters to 100 m height, and 100 m
to 1000 m lateral length. They usually have very high permeabilities, which can be above 10 Darcy. They affect the
PTT behavior of wells significantly, particularly if they are within the drainage area of the well.
In general, open fractures are very conductive whether they are in a discrete or continuous fracture network.
Furthermore, many mineralized fractures can also be conductive. After the mineralization of fractures, the formations
have been under continuous tectonic forces over millions of years, and stress magnitudes and directions may change
with time. Moreover, compaction as formations are buried deeper and chemical reactions, particularly by active
6 IPTC 18194

chemical agents in migrated gas, oil, and/or water, between fracture and mineralized material surfaces may alter
fracture conductivities. It is conceivable that under these forces, the fracture plane and mineralized material surface
may detach from each other or do not strongly adhere to each other by cement materials. A thin crack between the
fracture plane and mineralized surface does not have to be thick to be conductive; a 0.001-mm tiny crack yields about
a permeability of 800 mD. It should be pointed out that not all fractures have these tiny cracks between fracture
and mineralized material surfaces in a given formation.
Consequently, mineralization of some fractures, and less
conductive secondary and tertiary fractures increase hetero-
geneity in NFRs. Most fracture properties exhibit power-law
distributions. For instance, a small number of fractures may
have large spacing, while hundreds or thousands of fractures secondary fractures

have small spacing (Belfield and Sovich, 1994); see Fig. 9.

However, Odling (1997) cautions us that the power-law dis-
tribution in nature has upper and lower limits, at least for
Tertiary fractures
the fracture length distribution, although it could be valid
up to a few kilometers.
Aarseth et al. (1997) also presented the average joint Macrofractures

spacing data from both limestone and sandstone reservoirs

with different lithologies; see Fig. 10. They pointed out that
the spacing data given in this figure is consistent with a power
law relationship.

Figure 8: An outcrop of a naturally fractured formation,

where macrofractures, secondary, and tertiary (minor) frac-
tures are visible, from Kuchuk et al. (2014).

Figure 10: Logarithmic plot of average joint spacing against

Figure 9: Power-law distributions of fracture spacing, modified mechanical unit thickness for 12 different localities, from Aarseth
by Nurmi et al. (1995) after Belfield and Sovich (1994). et al. (1997).

Dual-Porosity Models for NFRs

A lumped parameter model for naturally fractured reservoirs was first formulated by Barenblatt et al. (1960) and
extended by (Warren and Root, 1963) for more general cases.

Representative Elementary Volume (REV) for Dual-Porosity Models. In this section we will discuss
constructing an REV for a fractured medium because it is relevant to fracture reservoir modeling. Consider a porous
medium consisting of fractures and matrix blocks, as shown in Fig. 11. We choose a representative elementary volume
(VREV ) that is sufficiently large so that vmm  vm  VREV , where vmm is a representative macroscopic Darcy’s
porous medium volume, vm is a representative matrix element volume, and VREV is a representative elementary
volume. vm should contain a large number of vmm to be statistically meaningful and representative. VREV should
contain a large number of matrix elements (vm ) and fracture segments (vf is a representative fracture segment
volume) to be statistically meaningful and representative as well. All these quantities are shown in Fig. 11. For
IPTC 18194 7

simplicity, let us assume that vm and VREV have cubical shapes with dimensions lf and lV , respectively, where lf is
the length of a fracture segment. For practical purposes, lf ≈ lm , where lm is the dimension of the matrix element
cube vm . Furthermore, it is assumed that vf is a rectangular parallelepiped with dimensions lf , lf , and b, where b is
the fracture aperture, and the pore volume is a spherical shape with a radius rp . These all imply that rp  lf  lV .
The characteristic length at the pressure transient well test scale for building an REV will be the reservoir
(formation) thickness h, or the external drainage radius re of a well. Normally h  re ; if so, the characteristic length
must be the reservoir thickness h because natural fractures are three-dimensional objects. Even for 2D problems, we
must have a large number of VREV elements in the vertical direction so that we have a statistically meaningful and
representative medium. Thus, VREV should satisfy rp  lf  lV  h.

Matrix Fractures Solid grains

vmm
Pore space

Fracture segment homogenized equivalent fractured media

vf
lV
vf vf homogenized equivalent

VREV
matrix source element
flow
matrix
element lf
vm
vf
Fracture segment

Figure 11: Warren and Root (1963) REV, a single matrix element Figure 12: Naturally fractured porous (dual-porosity) medium and
surrounded by intersecting fracture segments, and the macroscopic a representative elementary volume (REV) V for matrix elements
control volume vmm of a matrix element. and fractures.

Macroscopic Darcy’s representative elementary volume vmm :

Most carbonate formations are heterogeneous, contain micro fractures and vugs, and have diverse pore types. These
formations are interbedded with grainy and muddy stones (wackstone, packstone, grainstone, mudstone, etc.). In
general, low porosity and permeability dolomite usually causes strong vertical stratification within the formation. The
presence of small amounts of dolomite can deteriorate the reservoir quality (permeability). Microscopic heterogeneity
is controlled by both the depositional environment and diagenetic history. Carbonate porosity is dominated by
micro (intragranular) porosity, and large intergranular pores usually enhance permeability. Petrophysical and flow
characteristics of carbonate formations are governed to a large extent by their pore level heterogeneity. Ramakrishnan
et al. (2001) presented a method that parametrizes the pore structure in terms of a multiporosity system of fractures,
vugs, and intragranular and large intergranular porosities. Their method is based on physics and core measurements,
and relies minimally on empirical correlations.
Excluding fractures, Ramakrishnan et al. (2001) provide three qualitatively different contributors to porosity:
vuggy, intergranular, and intragranular. The matrix, excluding fractures, is thus an assembly of grains with a
fraction fg , separated by intergranular fraction fm . Any amount of grains removed from the lattice arrangement
contributes to a vug fraction fv . Vugs range in size from hundreds of microns to a few centimeters. The grains may
be microporous, with a porosity equal to φµ . This classification is illustrated in Fig. 13. Quantitatively,

fv + fm + fg = 1 ; fv + fm + fg φµ = φ ; fµ = fg φµ . (1)

In this scheme, when φµ = 0, we automatically obtain the classical grainstone with no microporosity. Similarly, when
fm = 0, a mudstone is obtained. An intercrystalline rock with interdispersed vugs will have φµ = 0, but fv 6= 0. If
microfractures are present in the matrix and not highly connected to main fractures and/or their conductivities are
low, the total porosity of vmm becomes φmm = fmf + fv + fm + fg φµ , where fmf is the microfracture fraction.
Figure 14, given by Akbar et al. (2001), presents wellbore resistivity images in Track 2 and Track 3 from a West
Texas carbonate formation. Track 1 from ECS (Elemental Capture Spectroscopy) indicates volumes of carbonate
in blue, clay in brown, and quartz in yellow. It should be noted from this figure that vugs are imbedded in the
matrix and contribute to the matrix permeability and porosity. If vugs are directly connected to the fracture system,
then they must be a part of that system. In most carbonate formations, fracture and vug apertures have the same
order of magnitude: the aperture may vary from a few µm to 10 mm. On the other hand, fracture lengths may
8 IPTC 18194

0 100

10,399

10,400

Intergranular, fm 10,401

Absence of grains
Vug, fv 10,402

Intragranular, fμ
10,403

10,404

10,405

Figure 13: Schematic of pore partitioning, with microporous Figure 14: Fractured, vugular carbonate rock with shale filling
grains, modified from Ramakrishnan et al. (2001). pores, from Akbar et al. (2001).

vary from a few millimeters to several thousand meters, while vug lengths may vary from a few µm to 100 mm.
Thus, fracture-intersecting vugs will not make any observable change in the fracture conductivity, provided that the
fracture conductivity is not very low.
Let us assume that the matrix is an isotropic porous medium and has a normal or log-normal permeability dis-
tribution. If both microfractures and vugs are embedded in the matrix, an effective permeability for vmm , which
includes matrix, microfractures, and vugs, can be obtained by solving the Stokes equation at the microscopic scale for
Newtonian fluid at a low Reynolds number, or by using a homogenization technique. Let the effective permeability
of vmm be kmm .

Macroscopic representative elementary volume VREV for a fractured medium:

As stated above, most naturally fractured reservoirs, particularly carbonate formations, contain primary (macro),
secondary , and tertiary (minor) fractures. Let us assume that these fractures communicate hydraulically with each
other in the domain of interest, i.e., they create a hydraulically connected network. Thus, the fracture network is a
continuous porous medium. Matrix blocks are isolated from each other and are dispersed in the continuous medium.
Let each fracture segment (vf ) be a rectangular parallelepiped with a distinct porosity and permeability. Here we
assume that Darcy’s Law is valid for the flow through the fractures.
Let us take a VREV volume from the fractured medium that is large enough to include a statistically significant
number of fracture segments and matrix blocks, as shown in Fig. 11, while satisfying rp  lf  lV  h. If the
fracture segments and matrix blocks have a similar order of magnitude of permeabilities, VREV can be treated as
a single effective continuum porous medium. The effective permeability and porosity of a homogenous VREV can
therefore be obtained by using an upscaling or homogenization technique.
Some of the requirements for VREV given above were included in the following three papers. For a common
REV, Warren and Root (1963) stated, “If useful solutions are to be obtained utilizing the smoothed pressures in the
primary [matrix] and secondary [fractured] regions, p1 and p2 [shown in Fig. 2 of the Warren and Root (1963) paper],
it is obvious that the volume considered [VREV ] must be small in comparison with the volume of the reservoir and
must be large in comparison with the size of the matrix elements.” Gringarten (1984) stated, “A basic assumption
in the Barenblatt et al. (1960) model is that any infinitesimal volume [VREV ] contains a large proportion of each
of the two constitutive [matrix system and the fracture network] media. As a consequence, each point in space is
associated with two pressures, namely: (1) the average fluid pressure, pf (p2 ), in the most permeable medium in
the vicinity of the point, and (2) the average fluid pressure, pm (p1 ), in the least permeable medium in the vicinity
of that same point.” Raghavan (1993) stated, “In this model [Warren and Root (1963)], the matrix system and the
fracture network are visualized as two overlapping continua that are uniformly distributed—an infinitesimal volume
[VREV ] contains both the fracture and matrix systems.”
Concerning the mathematical models for pressure transient tests in NFRs, we can make the following three
IPTC 18194 9

remarks:

1. If the permeabilities of the fracture segments and matrix blocks have a similar order of magnitude, the system
can be treated as a single effective continuum porous medium.
2. If the permeabilities of the fracture segments are a few orders of magnitude larger than those of the matrix
blocks, but not km  kf , then, as suggested by Bear (1993), a macroscopic representative elementary volume
VREV can be constructed for two overlapping continua: the equivalent homogenous medium of the network of
fractures, and the equivalent homogenous medium of the matrix elements; Fig. 12.
3. If km  kf , then a macroscopic representative elementary volume VREV cannot be constructed. Analytical
(Biryukov and Kuchuk, 2012) or numerical techniques should be used to model the transient pressure behavior
of fractured reservoirs without fracture segments homogenization.

Given the above three conditions, in the next section we will describe mathematical models for pressure transient
tests in NFRs. Models for single effective continuum porous media are simple, and many solutions to them are
available for various reservoir and wellbore geometries. First we will discuss dual-porosity models, and second, more
general models without fracture segments homogenization, i.e., without an equivalent homogenous porous media for
the fractured systems.

Mathematical Dual-Porosity Reservoir Models

Given the macroscopic representative elementary volume VREV for a fractured medium, let us construct a Warren and
Root (1963) type dual-porosity model. Let Vf be a volume consisting of the total bulk volume of fracture segments,
and Vm be a volume consisting of the total bulk volume of matrix elements (Fig. 11), where VREV = Vm + Vf . If
each of the connected fracture segments has a normal porosity distribution and a normal or log-normal permeability
distribution, an effective porosity and permeability can be obtained for Vf using an upscaling or homogenization
technique. Vf is now an equivalent (fictitious) homogenous medium that does not contain any fractures; see Fig. 12.
The same procedure is also applied to obtain an equivalent (fictitious) homogenous medium Vm for matrix elements
with effective properties. It should be stated that Vm is not a matrix element of the actual fractured medium. In
the mass balance continuity equation for Vf , Vm becomes a source term. If matrix block sizes are considerably
different within the macroscopic representative elementary volume VREV , Vm can be partitioned into a few smaller
equivalent homogenous matrix elements (Belani and Jalali, 1988; Cinco-Ley et al., 1985; Johns and Jalali, 1991; Liu,
1981; Spivey and Lee, 2000). These papers all included the matrix-block-size distribution in f (s) (see above), i.e.,
in VREV , and assumed the distribution is the same spatially throughout the fractured reservoir. This contradicts
many studies (Aarseth et al., 1997; Belfield and Sovich, 1994) on fracture spacing, which is directly related to the
matrix-block-size distribution. As can be observed from Figs. 9 and 10, the fracture spacing is quite variable and
often exhibits a power-law distribution.
2 km
In the dual-porosity models, the λ (αrw kf ) parameter is related to the matrix block size via α. Let us assume
that the ratio of two α values of two matrix blocks is the order 10, which is a moderately high value. If km and
kf are locally (in VREV ) the same in each of the blocks, the α ratio of 10 gives, for instance, a volume ratio of two
blocks as 103/2 ≈ 31, which is a large difference in terms of matrix block size. On the other hand, an α ratio of
10 slightly affects the pressure behavior of the system. Therefore, the use of a single equivalent homogenous matrix
element Vm would be sufficient for the macroscopic well test scale. As we said, except for the size, it is unlikely that
matrix elements that are in close proximity to each other will have very different properties.
It is possible for calcite or other minerals to deposit (called mineralization) on the actual matrix surfaces and in
the fractures. Fracture-filled mineralization generally takes place in the fracture void space (aperture). Therefore,
if there is a reduced permeability region due to calcite deposits or diagenesis on the surface of each physical matrix
block, then the overall permeability of the equivalent homogenous matrix element Vm will be reduced. In other words,
because the matrix element control volume Vm contains a large number of physical matrix blocks, the core matrix
permeability will differ from the reduced upscaled permeability of the volume Vm . There will be no damaged skin
on the surface of the equivalent homogenous medium, Vm , although mineralization will prolong the matrix element
transient period in the actual fractured reservoirs due to the permeability heterogeneity created by the non-uniformly
distributed calcite deposited regions in Vm .
Warren and Root (1963) undoubtedly assigned the same VREV (representative elementary volume) for fracture
and matrix media (two overlapping continua), where each medium has distinct properties. The pressure diffusion
equation for the equivalent homogenous medium of the network of fractures [Warren and Root (1963) called it
secondary porosity medium, denoted by the subscript f ] can be easily written in a 1D cylindrical radial coordinate
system as  
kf 1 ∂ ∂pf ∂pf
r − qm−f = φf (ct )f , (2)
µ r ∂r ∂r ∂t
10 IPTC 18194

where qm−f is the source term per unit volume that is the flow from Vm of the equivalent homogenous medium of the
matrix elements [Warren and Root (1963) called it primary porosity medium, denoted by the subscript m] into Vf of
the equivalent homogenous medium; see Fig. 12. This equation is given by Warren and Root (1963) with some minor
differences. The qm−f source term is normally obtained by solving the one-dimensional pressure diffusion equation
in spherical coordinates that is written as

 
km 1 ∂ 2 ∂pm ∂pm (t)
r = φm cm . (3)
µ r2 ∂r ∂r ∂t

Two-Horizontal Slab Models. Let us assume that we partition an NFR, as shown in Fig. 15, into two horizontal
slabs: one for the equivalent homogenous medium of the network of fractures, and the other for the equivalent
homogenous medium of matrix elements. This is a two-layer reservoir with crossflow, where the system is layered in
the z direction and is infinite in the x and y directions. A vertical well is completed only in the bottom layer (see
Fig. 15). If the permeability slightly varies in the horizontal direction, then for transversely isotropic media, we can
assume (kx )i = (ky )i = (kh )i and (kz )i = (kv )i . The flow of a slightly compressible fluid of constant compressibility
and viscosity is assumed for each layer. It is assumed that only the bottom layer is open to production. The pressure
diffusion equation for this two-horizontal-slab model in the i-th layer can be written as

∂ 2 pi ∂ 2 pi ∂ 2 pi ∂pi
(kh )i 2
+ (k h ) i 2
+ (kz )i 2
= (φµct )i , (4)
∂x ∂y ∂z ∂t

where i = f, m (fracture and matrix). As can be observed from Eq. 2, there is no source term qm−f in Eq. 4. The
fluid flows directly from the matrix layer into the equivalent fractured layer.
If the contact between the slab surfaces is perfect (Carslaw
and Jaeger, 1959; Hahn and Ozisik, 2012), i.e., there is no pres-
sure loss at the interface, then the continuity of pressure and Well
flux at the interface can be written as

pf (x, t) = pm (x, t) at the interface (5)

Equivalent homogenous medium for matrix

interface
and

kf ∂pf km ∂pm
. .
(x, t) = (x, t) at the interface. (6) k f ∂ pf k m ∂ pm
µ ∂n µ ∂n pf (x, t ) = pm (x, t )
µ ∂n
(x, t ) =
µ ∂n
(x, t )

Equivalent homogenous medium for fractures

Kazemi (1969) was the first to solve Eq. 4 for an NFR, but he
did not state the interface conditions he used. The solution of
Eq. 4 with the interface conditions given by Eqs. 5 and 6 is well
known and has been published in the literature; see Kuchuk Figure 15: Schematic of a two-layer slab model for an
(1994). NFR.
If the contact between the slab surfaces is imperfect (Carslaw and Jaeger, 1959; Hahn and Ozisik, 2012), i.e., a
sudden pressure drop occurs across the interface, the boundary condition for this case at the interface can be written
as
kf ∂pf  km ∂pm
(x, t) = α0 pf (t) − pm (t) =

(x, t) at the interface, (7)
µ ∂n µ ∂n

where α0 is the contact resistance [we use α0 to differentiate it that is given by Barenblatt et al. (1960) in Eq. 9], which
can be considered an infinitesimally thin layer (Carslaw and Jaeger, 1959), like the skin factor over the wellbore, and
pf and pm denote the pressure of the fluid in the equivalent fractured and matrix media, respectively. Raghavan
(1993) also called the interface condition given by Eq. 7 as the resistance interface condition.
Now let us look at the two-horizontal-slab model, for which Barenblatt et al. (1960) were the first to use the
resistance interface condition. Barenblatt et al. (1960) stated that on the basis of dimensional analysis (Sedov,
IPTC 18194 11

1954)∗ , we obtain for v [velocity] an expression of the type

α
v= (pf − pm ), (8)
µ

where α is some new dimensionless characteristic of the fissured rock. Barenblatt et al. (1960) further stated that
for the mass q of the liquid that flows from the pores into the fissures per unit of time, per unit of rock volume, the
following equation is valid
ρα
q= (pf − pm ), (9)
µ

where ρ is the density of the liquid. The resistance interface condition (Eq. 9) has been known in the heat conduction
literature since the early 1900s (Carslaw and Jaeger, 1959; Jaeger, 1955), and simply creates an additional pressure
drop between the equivalent matrix and fractured media.
Barenblatt et al. (1960) presented a solution for Eq. 4 without the horizontal flow in the equivalent matrix medium
2 2 2
[(kh )i ∂∂xp2i + (kh )i ∂∂yp2i = 0, i = m and (kz )i ∂∂zp2i = 0, i = f ] with the interface condition given by Eq. 9 (Eq. 7). Serra
et al. (1983) also presented a solution for Eq. 4 without the horizontal flow for i = m using the interface condition
given by Eq. 5. The de Swaan–O (1976) solution for the same problem is slightly different from the Barenblatt et al.
(1960) solution because he used a different interface condition. The solution of Eq. 4 with the interface condition
given by Eq. 7 is given in detail by Bourdet (1985). In the Bourdet (1985) model, there is no flow in the vertical
2
direction. Thus, (kz )i ∂∂zp2i = 0, where i = m, f .
It should be again stated that the slabs are equivalent homogenous media: one is for the the network of fractures,
and the other is for matrix elements. Perhaps it is slightly easier to solve Eq. 4 with the interface condition given by
Eq. 7 than with the interface conditions given by Eqs. 5 and 6, and without the horizontal or vertical flow. These
solutions exhibit look-alike NFR pressure behaviors. It is difficult to justify the resistance interface condition given
by Eq. 7 for most geological formations. It is clearly a non-verifiable claim for most formations, and neither cores nor
petrophysical measurements support the resistance interface condition given by Eq. 7. Although the two-horizontal-
slab models discussed above produce look-alike NFR pressure behaviors, they cannot be physically justified or verified.

The Pseudosteady-State Matrix Element Warren and Root (1963) Models. Although Warren and Root
(1963) did not reference it, the resistance interface condition given by Eq. 9 is the same as Eq. 1-5 of Warren and
Root (1963). Equation 2 given above for the equivalent homogenous fractured medium in a dimensionless form can
be easily written as  
1 ∂ ∂pf D ∂pmD ∂pf D
rD − (1 − ω) =ω (10)
rD ∂rD ∂rD ∂tD ∂tD

and, using the resistance interface condition given by Eq. 7, for the equivalent homogenous matrix element as

∂pmD
(1 − ω) = λ(pf D − pmD ), (11)
∂tD

where
kf h
pf D = [pi − pf (t)], (12)
141.2qµ

0.0002637kf t
tD = 2
, (13)
µ(φm (ct )m + φf (ct )f )rw

the storativity ratio, ω, is defined as

φf (ct )f
ω= , (14)
φm (ct )m + φf (ct )f

the interporosity flow coefficient, λ, is defined as

2 km
λ = αrw , (15)
kf
∗ This book is about fluid dynamics and mechanics. It does not refer to heat conduction, pressure diffusion, porous media, and Eq. 8.
12 IPTC 18194

where α is called a geometric parameter for the “heterogeneous region;” it is also called a matrix block shape factor.
Warren and Root (1963) used rectangular parallelepipeds for the equivalent homogenous matrix elements. In general,
different matrix block (Vm , which is not an actual matrix block of an NRF) shapes have been used: 1) Spherical
(cubes), 2) Horizontal slabs (layers), 3) Cylindrical (rectangular parallelepipeds), and 4) Multiple block sizes. It
should be noted that the Laplacian of the left-hand side of Eq. 3 is replaced by the resistance interface condition
given by Eq. 7, resulting in Eq. 11.
Now let us investigate the pseudosteady-state flow condition from the matrix elements to the equivalent fractured
medium. The solution for an internally and externally bounded spherical matrix element producing at a constant-rate
at the internal surface of the spherical matrix element is given by Chatas (1966) as

√ √  √ 
sreD cosh s(reD − 1) − sinh s(reD − 1)
p̄mD (s) = n√ √  √ o . (16)
s s(reD − 1) cosh s(reD − 1) + (sreD − 1) sinh s(re − 1)

For long times, as s → 0 (tmD → ∞), Eq. 16 becomes

2 (reD − 1)reD
p̄mD (s) = 2 + 1)s2 + (r 2 + 1)s (17)
(reD − 1)(reD eD

and in the time domain

2tmD (reD − 1)reD
pmD (tmD ) = 2 + 2 + 1) . (18)
(reD − 1)(reD + 1) (reD

This equation, as expected, clearly shows that the spherical matrix element reaches the pseudosteady-state flow
condition at large times.
The linearity of the diffusivity equation allows us to express the solution for a constant-pressure or constant-rate
inner boundary condition (van Everdingen and Hurst, 1949) as

1
q̄D = . (19)
s2 p̄D

Using Eqs. 18 and 19, and taking the inverse Laplace Transform of the resultant equation, we obtain the flow rate
at inner boundary at large times as

2
 
(reD + 1) −2tmD
qmD (tmD ) = exp . (20)
(reD − 1)reD (reD − 1)2 reD

Like any closed system, the flow rate exhibits an exponential decline.
The constant-rate solution at the surface of the spherical matrix element from Carslaw and Jaeger (1959) can be
written as  !
∞
q  3t µrs 2µrs X 1 ηm αn2 t 
∆pm (t) = + − exp − , (21)
4πrs2 φ(ct )m rs 5km km n=1 αn2 rs2

km
where ∆pm = po − pm , po and pm denote the initial reservoir and matrix pressures, respectively, ηm = φm µ(c t )m
is
the diffusivity of the matrix, rs is the radius of the spherical matrix element, and αn = 1, 2, . . ., are the positive roots
of tan(α) = α.
The dimensionless pressure at the surface of the spherical matrix element can be written from Eq. 21 as

∞
1 X 1 
2

pmD (tmD ) = tmD + +2 exp −αn t mD . (22)
5 α2
n=1 n
IPTC 18194 13


For long times, as tmD → ∞, exp −αn2 tmD → 0, Eq. 21 becomes

1
pmD (tmD ) = tmD + . (23)
5

Using Eqs. 23 and 19, and taking the inverse Laplace Transform of the resulting equation, we obtain the flow rate
at the outer boundary boundary (surface of the sphere) at large times as

qmD (tmD ) = 5 exp (−5tmD ) . (24)

Like in any closed system, the flow rate again exhibits an exponential decline for a spherical matrix element producing
from its outer surface. The dimension of the outer radius is not in Eq. 24 because tmD in this case is based on rs ,
the radius of the spherical matrix element, thus reD = 1.
Therefore, it obvious from Eqs. 20 and 24 that the transient solutions for any type of finite volume matrix elements
should include the pseudosteady-state flow condition at large times. i.e., the transient flow condition will turn into
pseudosteady-state at large times. This large time should not be confused with s → 0 (tD → ∞), where the time
is based on the wellbore radius for the equivalent fractured medium. We know that the Warren and Root (1963)
dimensionless wellbore pressure solution at large times can be written as s → 0+ (t → ∞), then f (s) = 1 as

√
K0 ( s)
p̄D (s) = √ √ . (25)
s s K1 ( s)

Thus, the slope of the semilog straight line at large times, mf , for the equivalent fracture medium from Eq. 25 can
be written as
qµ
mf = , (26)
2πkf h

from which the equivalent fracture medium permeability kf is obtained. This is not an average system (NFR)
permeability. On the other hand, many field well test examples [many of them were presented in Kuchuk and
Biryukov (2014)] indicate that the long term transient behavior does not yield the equivalent fracture medium
permeability kf as the Warren and Root (1963) model suggests. The long term transient, provided that the matrix is
not ultra tight, usually yields the equivalent matrix medium permeability because Vf t  Vmt in any given reservoir,
which always has a finite volume, where Vf t and Vmt are the equivalent fracture and matrix medium total volumes,
respectively. The equivalent matrix medium permeability is the effective permeability due to matrix, vugs, and minor
and micro fracture permeabilities. This effective permeability is always larger than the matrix core permeabilities.
Therefore, Warren and Root (1963) model type models usually underestimate both fracture network and matrix
permeabilities considerably.
Transient Matrix Element Models.
Dual-porosity models with the transient flow condition in matrix elements (Fig. 3) were first introduced numerically
by Kazemi (1969), and subsequently analytically by de Swaan–O (1976) without using the resistance interface
condition given by Eq. 7. In principle, it is permissible to use the dual-porosity models given by Eq. 2, provided
that the conditions stated above are met. However, there are serious problems with Eq. 2 when it is used for
pressure transient test interpretation in fractured reservoirs: 1) The inner (wellbore) boundary condition and 2) The
interporosity skin factor for the matrix. Both of these two issues will be discussed in the next section.
In summary, the Stokes equation, which describes the microscopic velocities associated with fluid paths in porous
media, is an approximation (simplification) of the Navier-Stokes equation. With further simplifications we can derive
Darcy’s law, which provides an accurate description of the macroscopic relationship between superficial velocity and
pressure drop. In this sense, the dual-porosity models with pseudosteady-state or transient flow conditions in matrix
elements are not an approximation or simplification for the transient behavior of the natural fractured reservoirs,
although they can be used for a few specific cases.
Figure 16 presents derivatives for a dual-porosity model with both pseudosteady-state (PSS) and transient flow
(TR) conditions in matrix elements for λ values of 10−7 and 10−5 , and ω values of 0.01 and 0.1. As can be seen
from this figure, the derivatives of the pseudosteady-state case are significantly different from the derivatives of the
transient case. The derivatives of the transient case are flat, the dip never goes below 0.20, and exhibit more or less
a radially composite look-alike behavior. None of the 33 field buildup test derivatives from 32 different well-known
naturally fractured reservoirs reviewed by Kuchuk and Biryukov (2012) exhibits a similar behavior to the transient
14 IPTC 18194

case. This is also true for many other derivatives published in the petroleum literature. However, Kikani and Walkup
(1991) presented a field example, for which a radially-composite NFR model was used for interpretation. It should
be noted that their composite model is different from the transient flow matrix models shown in Fig. 16.

-5 -4 -3 -2 -1 0 1
10 10 10 10 10 10 10
1 radial flow -5
radial flow 1
λ = 10 -7
FD = 1 λ = 10 ω = 0.01 PSS

Derivative, dpD/dlntD
Derivative, dpD/dlntD

-7 radial flow (slope m) radial flow (m)

λ = 10
0.1
-7 ω = 0.01 PSS
0.1 -5 λ = 10
λ = 10 ω = 0.1 PSS
ω = 0.01 PSS -6
ω = 0.1 PSS FD = 10 λ =1.5*10
ω = 0.01 TR ω = 0.007 S = 2 TR
ω = 0.1 TR
ω = 0.01TR 0.01
3 4 5 6 7 8
ω = 0.1 TR 10 10 10 10 10 10
tD
0.01
1 2 3 4 5 6 7 8
10 10 10 10 10 10 10 10
tD Figure 17: Derivatives of a dual-porosity model with the
pseudosteady-state flow condition (PSS) and transient (TR) with
a matrix skin flow condition in spherical blocks, and the deriva-
Figure 16: Derivatives of dual-porosity models with the tives, for which tD is shown by the top axis, for FD = 1 and 10
pseudosteady-state (PSS) and transient (TR) flow conditions in that is obtained from the Biryukov and Kuchuk (2012) solution
spherical matrix elements. using the Warren and Root (1963) geological model; see Fig. 18.

Moench (1984) introduced an interporosity skin factor for the matrix elements. This concept of interporosity skin
is also called restricted interporosity flow condition. Stewart and Asharsobbi (1988) have further elaborated on the
interporosity skin in the dual-porosity models with transient flow conditions in matrix elements. As we stated above,
mineral depositions, particularly calcite and anhydrite, may occur in fractures and matrix elements, including the
matrix face (Davidson and Snowdon, 1978; van Golf-Racht, 1982). However, it is not possible for calcite deposits
in fractures and over the matrix blocks uniformly distributed throughout the formation that can be characterized
by a single interporosity skin (Sma ). In fact, high, low, and nonconductive fractures are usually observed in the
same wellbore, but this pattern does not repeat itself uniformly throughout the reservoir. Furthermore, cores and
high-resolution image logs indicate that calcite deposits are not usually observed on matrix block faces. Fracture-
filled mineralization generally takes place in the fracture void space (aperture). As shown in Fig. 19, high and low
conductive and closed fractures with many different sizes, dip angles, and azimuths are distributed throughout the
system (Souche and Rotsch, 2007).

Orientation
Perpendicular to bedding Conjugated

Matrix

Well Main
fault
Fracture

2lw
2lw Matrix 1400 m 2100 m

Cemented (mineralized)

Open

Sub-seismic fault zone

Figure 18: Warren and Root (1963) uniformly distributed

fracture and matrix block geological model, after Kuchuk and Figure 19: Open and closed fractures, and two faults intersecting
Biryukov (2014). a horizontal well, from Souche and Rotsch (2007).

Including a single interporosity skin (Sma ) in the dual-porosity models with the transient flow condition in matrix
elements does not make sense anyway because both fractures and matrix elements are equivalent homogenous media,
and are not physical matrix blocks and fracture segments of NFRs. Nevertheless, Fig. 17 presents derivatives of a
dual-porosity model with PSS and TR-Sma (pseudosteady-state flow conditions in spherical elements and transient
flow conditions with an single interporosity skin). We have used the transient spherical blocks with interporosity
IPTC 18194 15

skin given by Gringarten (1984) as

q q 
λ 15(1−ω)s 15(1−ω)s
5s λ coth λ −1
f (s) = ω + q q . (27)
15(1−ω)s 15(1−ω)s
1 + 2Sma λ coth λ −1

As can be seen in this figure, the derivative of the pseudosteady-state (PSS) spherical blocks without skin, with λ
value of 10−7 and ω values of 0.01, behaves almost identically to the derivative of the transient spherical blocks with
interporosity skin, with λ value of 1.5 × 10−6 , ω values of 0.007, and Sma = 2. As stated above, the Barenblatt
et al. (1960) and Warren and Root (1963) resistance interface condition dominates the transient behavior of their
dual-porosity models, but not the assumption of the pseudosteady-state flow condition in matrix blocks. The similar
behavior of the pseudosteady-state and the transient with the interporosity skin introduces serious nonuniqueness
problems. In the history matching of the field data, depending on which model we use, the λ and ω values will result
in a 15 times difference in kf /km and a 1.4 times difference in the matrix to fracture volumes.
Figure 17 also presents derivatives of a continuous fractured reservoir model, which is a 2D representation of
the Warren and Root (1963) geological model with uniformly distributed fractures and matrix blocks, as shown in
Fig. 18. In this model, the matrix blocks consist of 65.6 ft × 65.6 ft × h rectangular parallelepipeds with a square
base, and fractures with an aperture b, a length of 65.6 ft (2lw ), and a height h. For this case, the well is in a matrix
block. For this model, a meaningful REV for an equivalent fractured and matrix media cannot be built because the
matrix size (65.6 ft) does not satisfy rp  lf  lV  h, i.e., a reasonable formation thickness (h) usually is not
much greater than 65.6 ft (lV  h).
k b
In Fig. 17, only two derivatives for FD = 1 and 10 are shown for the model given in Fig. 18, where FD = kmflw ,
kf , b, and lw denote the fracture permeability, aperture, and the half length or reference length, respectively, and
km denotes the matrix permeability. These derivatives exhibit behavior similar to the derivatives of dual-porosity
models with PSS and TR-Sma . As shown in Fig. 17, all derivatives first exhibit a well-stabilized radial flow regime
(the m value gives semilog straight line slope); second, they go down rapidly and then up, finally, exhibiting another
radial flow regime. According to the dual-porosity models with PSS and TR-Sma , both radial flow regimes yield
the equivalent fracture medium permeability kf ; see Eq. 26. On the other hand, in accordance with the (Biryukov
and Kuchuk, 2012) model, the radial flow regimes for both FD = 1 and 10 yield the equivalent matrix medium
permeability km . It should be stated that the equivalent matrix medium is the same for both the dual-porosity
and (Biryukov and Kuchuk, 2012) type models. As explained, the equivalent matrix medium permeability includes
matrix, vugs, and minor and micro fracture permeabilities. This effective permeability is always larger than the
matrix core permeabilities. Although the unphysical dual-porosity models with PSS and TR-Sma have led to obtain
better history matches with the field data, they have caused serious inaccuracies in estimating NFR parameters from
transient well test data, i.e., substantially under estimating both fracture network and matrix permeabilities.

Problems with the Inner (Wellbore) Boundary Condition

The inner boundary conditions used in the dual-porosity models are unsuitable for fractured reservoirs because none
of the fractures in the reservoir intersects the well. In reality when we drill a well there are two possibilities:
1. The well does not intersect any fractures, i.e., the well is in the matrix.

2. The well intersects one or multiple fractures and one or multiple matrix elements, because well diameters
are much larger than fracture apertures. There is no possibility, therefore, that the well will only intersect
fracture(s).
In the near-wellbore formation, fractures may intersect the wellbore, as shown in Figs. 20(a) and 21(a). Most
wellbore formation images indicate that fractures intersect as swarms or clusters with certain dips and azimuths. In
general, dip angles are usually higher than 45◦ but rarely 90◦ (vertical fractures). Therefore, there is a high degree
of certainty that the well may intersect a few fractures at least. If fractures are open with a reasonable aperture,
they dominate the wellbore behavior. Flow regimes for this case changes from a 1/2 slope fracture linear flow to a
fracture radial flow, depending on the fracture conductivities. If fractures are not intersecting the wellbore as shown
in Figs. 20(b) and 21(b), the initial flow regime will be radial due to matrix blocks until the effects of the fractures
in the vicinity of the wellbore begin. This radial flow regime will yield the matrix (the formation) kh of the producing
interval of the wellbore. After these initial flow regimes, the pressure derivative starts going downward due to the
effects of the fracture.
16 IPTC 18194

Well Well
Well Well Porous Porous
medium medium
Porous Porous (matrix) (matrix)
medium medium
(matrix) (matrix)
Fractures Fractures
Fractures
Fractures

(a) (b)
(a) (b)
Figure 21: Schematic of near-wellbore formation in a discretely
fractured formation, where fractures may intersect each other locally
Figure 20: Schematic of near-wellbore formation in a continuously and are in pressure communication with the wellbore through the
fractured formation, modified from Bear (1993): (a) fractures inter- matrix (formation), modified from Bear (1993): (a) fractures inter-
sect the wellbore and (b) fractures do not intersect the wellbore. sect the wellbore and (b) fractures do not intersect the wellbore.

As shown in Figs. 20 and 21, it is unlikely for a vertical well to intersect any fractures in fractured reservoirs if
matrix sizes are large. The dual-porosity models are usually applied to continuously fractured reservoirs regardless
of matrix size, whether the well is connected to fractures or matrix blocks.

The Well Intersects a Matrix Block only. It is highly unlikely for a vertical well to intersect any fractures
in sparsely fractured reservoirs. As with fracture spacing, matrix element size must have power-law distribution; see
Fig. 9. Therefore, matrix element size may vary from a few inches to a few hundred feet. If a fully penetrated vertical
well is produced at a constant rate q, then the inner boundary condition for the well in a matrix can be written as


2πkm h ∂p(r, t) 
− = q, t > 0. (28)
µ ∂r r=rw

For the Warren and Root (1963) dual-porosity model, it is


2πkf h ∂p(r, t) 
− = q, t > 0. (29)
µ ∂r r=rw

Normally km  kf ; this is why the Warren and Root (1963) model cannot even be applied to continuously fractured
reservoirs when the well is in the matrix. If the well is in the matrix, the derivatives (Fig. 16) of the Warren and
Root (1963) geological model (Fig. 18) for FD = 1 and 10 are very similar to the derivatives of the dual-porosity
models with the pseudosteady-state and transient with the interporosity skin. This is the unintended consequence
of using dual-porosity models.
The following example, which was presented by Kuchuk et al. (2014), will demonstrate how important it is to
know whether the well is in the matrix or intersecting fractures. Let us simulate the pressure and rate transient
behaviors of a reservoir configuration containing 40 fractures (Fig. 22) that are vertical and disjointed, predominately
in the northwest to southeast (NW to SE) direction in a homogeneous porous medium (matrix). Figure 23 shows
the pressure derivatives for the model shown in Fig. 22 for various FD values. As can be seen from this figure,
the pressure derivatives exhibit two plateaus; they also show a three-zone radially composite reservoir look-alike
behavior.
Let us assume that the well in the fractured reservoir shown shown in Fig. 22 is hydraulically fractured in the
NW to SE direction, where FD = 1000 and lwD = 5 (the dimensionless hydraulic fracture length). For this model,
Fig. 24 presents the derivatives, which are dominated by the hydraulic fracture. The formation linear (m = 1/2)
flow regimes are clearly observed for all FD values.
Wellbore-intersecting vertical fracture solutions with uniform flux or infinite conductivity in the dual-porosity type
models with both pseudosteady-state and transient-interporosity-flow models were presented by Houze et al. (1988).
In actuality there are no fractures in the system, and the hydraulic fracture is in the 1D homogenous porous medium
to which the matrix blocks contribute. Of course, it is well known that hydraulic fractures in fractured reservoirs
usually intersect other fractures. Figure 25 presents derivatives for an infinite-conductivity vertical fracture in a
IPTC 18194 17

5 N

4 1
9
8
3 7
6
2

Derivative, dpD/dlntD
5
y coordinate

1
4

0
3
Well
-1
FD = 0.01
-2 2 FD = 0.1
FD = 1
FD = 10
-3 FD = 100

-4
0.1
-5 0.001 0.01 0.1 1 10 100
-5 -4 -3 -2 -1 0 1 2 3 4 5
Dimensionless time, tD
x coordinate E

Figure 22: Distributions of fractures in a discretely fractured Figure 23: Dimensionless pressure derivatives of the discretely
reservoir, where the well is located at (0,0), from Kuchuk and fractured reservoir shown in Fig. 22, where the well does not inter-
Biryukov (2012). sect any fractures, for various fracture conductivities, from Kuchuk
et al. (2014).

1 −4
λ=10 ω = 0.1 TR λ=10−4 ω = 0.1 PSS
Derivative, dpD/dlntD

Derivative, dpD/dlntD
0.1

−4
0.1 λ=10 ω = 0.01 TR

0.01 m = 1/2 −4
FD = 0.01 λ=10 ω = 0.01 PSS
FD = 0.1
m = 1/2 FD = 1
FD = 10
FD = 100

0.01
0.001 -3 -2 -1 0 1 2 3 4 5
-5 -4 -3 -2 -1 0 1 2 10 10 10 10 10 10 10 10 10
10 10 10 10 10 10 10 10
Dimensionless time, tD tD

Figure 24: Dimensionless pressure derivatives for the discretely Figure 25: Dimensionless derivatives for an infinite-conductivity
fractured reservoir shown in Fig. 22 with a FD = 1000 hydraulic vertical fracture in a dual-porosity reservoir with pseudosteady-
fracture in the NW to SE direction, from Kuchuk et al. (2014). state and transient spherical matrix blocks.

dual-porosity fractured reservoir for pseudosteady-state and transient flow conditions in spherical matrix blocks for
λ = 10−4 , ω = 0.1, and 0.01. When we compare the Figure 25 derivatives with those in Fig. 24, there is no
similarity except for the early the formation linear (m = 1/2) flow regime.
Figure 26 shows the dimensionless flow rates for a di-
mensionless constant-pressure inner boundary condition for
F = 0.01
the hydraulically fractured NFR shown in Fig. 22; it also F = 0.1
D
D
F =1 D
shows the dimensionless rate for FD = 1000 without the hy- 10 F = 10
Dimensionless rate, qD

D
F = 100 D
without hydraulic fracture (F = 100)
draulic fracture. Although the fracture conductivity some- D

what affects the rate behavior, the hydraulic fracture in-

creases the production (Kuchuk et al., 2014) significantly.
Including hydraulic or wellbore-intersecting fractures in the 1

NFR for pressure transient solutions is important for reser-

voir management and for determining stimulation strate-
gies. Figure 26 shows that if a well in a discretely fractured
reservoir (in which the well is only in the matrix) is hy- 0.1
0.01 0.1 1 10 100 1000
draulically fractured, the production increases 100 times at Dimensionless time, tD
the beginning. For the long times, the production increase
reduces to 10 times. Figure 26: Dimensionless rate for the discretely fractured reser-
voir shown in Fig. 22 with a FD = 1000 hydraulic fracture in the
NW to SE direction, from Kuchuk et al. (2014).

Triple-Porosity Reservoir Models

The concept for the PTT behavior of triple-porosity reservoirs began with Liu (1981), and Abdassah and Ershaghi
(1986) developed further. Since the 1990s, triple-porosity models have been applied to many different well and NFR
types, and are still in use. The triple-porosity concept is not based on solid evidence, or detailed analytical or
numerical work. This is yet another extension of dual-porosity models, where an NFR has been transformed into an
 18 IPTC 18194

equivalent (fictitious) homogenous fractured medium and two equivalent homogenous matrix media using s− > sf1 (s)
and s− > sf2 (s) (see above for the f (s) definition) for the matrix media. As in dual-porosity models, sugar-cube,
spheres, slabs, cylinders, rectangular parallelepipeds, etc., elements are also used for the equivalent matrix medium
representation.
Abdassah and Ershaghi (1986) and Liu (1981) assumed that fractured reservoirs have two different matrix block
types, with two ω and λ values for each, where the REV (Representative Elementary Volume) is uniformly distributed
throughout the reservoir. It is highly unlikely that this level of partitioning at the macroscopic level in the REV can
be done at the scale of the macroscopic Darcy’s law. In other words, pressure diffusion is not sensitive to this level of
partitioning of porous media, including small vugs, unconnected micro fractures, intergranular macroporosity, and
intragranular microporosity, etc. For instance, if vugs are directly connected to the fracture system, then they are
part of the fracture system. If they are isolated, then they are a part of the matrix. Therefore, having multiple
matrix block sizes with different properties in a single REV is not useful for either pressure or rate transient tests in
fractured reservoirs or in any reservoir for that matter. Of course, in fractured reservoirs, the matrix block size and
its permeability may vary significantly spatially.
The model shown in Fig. 27 was presented by Morton et al. (2013) for a discretely fractured reservoir. The
fractured reservoir solution of Biryukov and Kuchuk (2012) is used to generate the wellbore pressure data with the
input model parameters that are given in Table 2 of the Morton et al. (2013) paper. In addition to the model
parameters, fracture density, length, and spacing have truncated lognormal distributions. Fracture conductivity
has a lognormal distribution. The model consists of about 300 fractures, as shown in Fig. 27, and all fractures
are predominately in the east to west direction, extending about 1640 ft (500 m) from the well. Figure 28 presents
derivatives for a number of realizations of the model. As can be seen from this figure, the derivatives for different cases
exhibit many different look-alike (not identical) behaviors: dual-porosity pseudo-steadystate and transient matrix
flow, radially composite, triple-porosity, etc. Of course, the Morton et al. (2013) model has nothing in common
with the the dual- or triple-porosity models. Therefore, the triple-porosity behavior observed from the well test in
naturally fractured reservoirs is most likely due to the variability of fracture properties: their density, distributions,
conductivities, and the large spatial scale variability of matrix properties. In other words, NFRs may exhibit a
triple-porosity behavior because the secondary and tertiary conductivities can be significantly lower than those of
major fractures. In addition, low conductivity fractures are easily filled with sediments, or mineralized, because of
their small apertures (see Fig. 8). It is highly unlikely that the triple-porosity behavior is due to the microscopic level
local variability of matrix properties because the matrix, as in dual-porosity models, is an equivalent homogenous
matrix medium.
600 N

400 1

200
Derivative, dpD/dlntD

m=0 m=0
y axis, m

0 well

-200

-400

-600
-600 -400 -200 0 200 400 600 0.1
x axis, m E 0.0001 0.001 0.01 0.1 1 10 100 1000
Dimensionless time, tD
Figure 27: Directionally but arbitrarily distributed fractures
with different conductivities in a discretely fractured reservoir,
where the well is located in the matrix at the center of the co- Figure 28: Derivatives for a number of realizations of the dis-
ordinate system x = 0 and y = 0, after Morton et al. (2013). cretely fractured reservoir model shown in Fig. 27.

Dual- and triple-porosity, resistance interface condition, matrix pseudosteady-state and transient flow conditions,
interporosity skin factor, multiple matrix elements, etc., with other slight variations of the Barenblatt et al. (1960)
and Warren and Root (1963) type models, have increased rapidly since the 1980s, but there are still significant
discrepancies that have persisted between the field data and the published models. Recently, more realistic analytical
and numerical solutions for NFRs have been published by Biryukov and Kuchuk (2012); Bogdanov et al. (2003);
Doonechaly et al. (2013); Izadi and Yildiz (2009); Wei et al. (1998). The study of Wei et al. (1998) has proved that
the dual-porosity Barenblatt et al. (1960) and Warren and Root (1963) type models are not realistic for NFRs. Next
we will give the details of Biryukov and Kuchuk (2012).
IPTC 18194 19

Mathematical Model for Continuously and Discretely Fractured Reservoirs

Most of the published solutions on the pressure transient behavior of a hydraulically fractured well or a wellbore-
intersecting natural fracture(s) included the flow entering from the formation into the wellbore, i.e., the solutions do
not include the wellbore. Most of these solutions require an a priori known value of the equivalent-pressure point.
The equivalent-pressure point solutions slightly diverge from the true solutions as the effects of top and bottom
boundaries become dominant, particularly for partially penetrated fractures. They further diverge when the effects
of the lateral boundaries, such as no-flow, faults, and fractures, are felt in the wellbore (Kuchuk et al., 1991). The
pressure-averaging solutions of Kuchuk and Wilkinson (1991); Ozkan and Raghavan (1991); Riley et al. (2007) are
less sensitive to boundaries and discontinuities, none of these solutions included the wellbore.
The Biryukov and Kuchuk (2012) solutions, which we have used in this paper, have two major contributions
regarding hydraulically fractured wells in naturally fractured and non-fractured reservoirs, and regarding wellbore-
intersecting natural fractures. First, the pressure (potential) at the open surface ∂Ω1 , which includes both wellbore
and fractures(s), must be uniform, i.e.,

∆p(x, t) = ∆pwu (t) on ∂Ω1 , t > 0, (30)

where x is the spatial position vector, ∆p(x, t) = p0 − p(x, t), and p0 is the initial pressure imposed at t ≤ 0. Second,
the inner boundary conditions for the pressure transient problems in fractured reservoirs are of the mixed type. In a
mixed boundary value problem, we have both the Neumann and Dirichlet boundary conditions imposed on the inner
boundary surface. The boundary condition at the open (producing) section of the inner boundary ∂Ω1 , for which
the total volumetric flow rate q is specified, can be written as

ZZ n
 o
− n · T grad ∆p(x, t) dS = q(t) on ∂Ω1 , t > 0, (31)
∂Ω1

where n is the outer normal vector to the surface ∂Ω1 , T is the mobility tensor given as

 
λx 0 0
T = 0 λy 0 , (32)
0 0 λz

k
λx = kµx , λy = µy , λz = kµz are mobilities, µ is the fluid viscosity, q is the total flux outwards from the surface ∂Ω1 ,
and for the no-flow section of the inner boundary

n  o
n · T grad ∆p(x, t) = 0 on ∂Ω2 , t > 0. (33)

If the fracture is fully penetrating the entire formation (the thickness of the formation is the same as the fracture
height), then Eq. 31 is sufficient for defining the inner boundary condition. If the fracture is partially penetrating,
both Eqs. 31 and 33 have to be satisfied. The inner boundary condition for both fully penetrating fractures and the
wellbore for an a priori unknown uniform pressure pwu over the wellbore and/or the wellbore-intersecting conductivity
fracture given is as (Biryukov and Kuchuk, 2012)

∆p = ∆pf 0 = ∆pwu on Γi , i ∈ I0 , (34)

k 2π ∂∆p k X li ∂∆p
Z Z  
rw (rw cos ϕ, rw sin ϕ, t)dϕ + dxi = −q(t), (35)
µ 0 ∂r µ −li ∂yi
i∈I0
∆p = ∆pf j on Γi , i ∈ Ij , j > 0, (36)
k X li ∂∆p
Z  
dxi = 0, j > 0, (37)
µ −li ∂yi
i∈Ij

where {r, ϕ} denote polar coordinates with the origin associated with the wellbore position, Γi is the i-th fracture
20 IPTC 18194

with the half length li , Ij denotes a set of indices i belonging to a j-th group of connected fractures (including a
wellbore), j = 0 corresponds to a group containing the well, j = 0 corresponds to fractures intersecting the wellbore,
and lj denotes indices for the intersecting fractures.
Biryukov and Kuchuk (2012) give solutions for the con-
rw damaged zone
stant sandface flow rate (qsf ). If the flow rate varies, then xf = lw = fracture half-length ks1 = damaged zone permeability
wellbore pressure at any time in a reservoir can be written
from the convolution integral (Muskat, 1937) as fracture b well
w e ll ks1 kf
kf = fracture permeability
b = fracture open width
xs1 = damaged zone half-length
Zt
second damaged zone (a)
pw (t) = po − qsf (τ ) gw (t − τ ) dτ (38)
0

ks2
and in the Laplace domain ks2 = second damaged zone permeability
xs1 = second damaged zone half-length
(b)
po
p̄w (s) = − q̄sf (s)ḡw (s), (39)
s

where t is time and gw is the impulse response at the well-

bore. If the flow rate is constant at the surface, the wellbore
pressure, including skin and storage effects in the Laplace
domain, can be written from Eq. 39 (see pages 60 to 66 of
Kuchuk et al. (2010) for the details) as
(c)
Figure 29: Schematic of fracture damage mechanisms: (a)
choked fracture, (b) damaged fracture face, and combination
of (a) and (b), modified from Soliman (2009).

po qgw (s)
p̄w (s) = −  . (40)
s s 1 + Csḡw (s)

Eq. 40 in terms of the dimensionless wellbore impulse re-

sponse can be written as
ḡwD (s)
p̄wD (s) = , (41)
1 + CD sḡwD (s)

and in terms of the dimensionless pressure and skin (S) as

sp̄D (s) + S
p̄wD (sD ) = n  o , (42)
s 1 + CD s sp̄D (s) + S

where
4πkh lw ∆pw kh t C
pwD = , tD = 2
, CD = 2
, (43)
qµ φµct rw 2πφct hrw

C is the wellbore storage coefficient, S is the skin factor, and s is the Laplace transform variable that is related
to the dimensionless time tD . The above equations in this section provide a general framework for interpreting
simultaneously measured pressure data with an arbitrarily varying flow rate as a function of time for pressure
transient well testing. The dimensionless pressures (pD ) for the formation pressure are given in the Biryukov and
Kuchuk (2012) paper.
The inner boundary conditions given above become much more important for low conductivity fractures, partic-
ularly in deviated wells. They also become important if the vicinity of the wellbore and/or perforations are damaged
regardless of the fracture conductivities. None of the previous published solutions on the pressure transient behavior
of a hydraulically fractured well or a wellbore-intersecting natural fracture satisfies these conditions. The published
solutions are, therefore, applicable only to a few simplified special cases. Cinco-Ley and Samaniego-V (1981) and
Soliman (2009) have specified three types of damaged skin factors that may exist in fractured wells, as shown in
Fig. 29:
1. Choked fracture (a), where the fracture opening area into the wellbore is damaged,
2. Damaged fracture face (b) due to excessive fluid leakoff, and
IPTC 18194 21

3. Choked fracture with a damaged fracture face (c).

With downhole shut-in, it is possible to determine the skin factor for a choked fracture (a) or a damaged fracture
face (b) from pressure transient well test data. If the fracture is choked for its entire length, then it is not possible
to determine the skin factor for either case. For this case and the case shown in Fig. 29 (c), only one total damage
skin factor can be determined.
Cinco-Ley and Samaniego-V (1981) included the skin factor in the fracture linear flow dimensionless wellbore
pressure as √
pD = πtD + Sf s , (44)

where Sf s is the skin factor for a choked fracture without a wellbore. Wong et al. (1986) included skin and wellbore
storage effects only in the bilinear flow dimensionless wellbore pressure (Soliman, 2009) as

hp i
(1/4)
pD = πtD / 2FD Γ (5/4) + Sf s (45)

Of course, these solutions have two fundamental problems: 1) Although very short in time, when we have a choked
fracture skin factor, it must be included in the fracture linear flow, while prolonging the fracture linear flow period
particularly for moderate and low fracture conductivities, and 2) With the wellbore storage and fracture skin included
only in the bilinear flow, the true fracture behavior will be considerably distorted. As can be seen from Fig. 5 of
Wong et al. (1986), all derivatives finally reach the bilinear flow regime after the skin and wellbore storage effects.
For most cases, however, the bilinear flow regime will be dominated by the wellbore storage and fracture skin effects.
As can be observed in Fig. 30 (Kuchuk and Biryukov, 2013), none of the derivatives exhibits a 1/4-slope bilinear
flow regime, and the derivatives for FD = 10 and 100 a 1/3-slope trilinear flow regime. Thus, the fracture skin and
wellbore storage significantly affect the derivative behavior of the fracture-intersecting wells in NFRs.
As stated by (Kuchuk and Biryukov, 2013) with the exception of openhole wells, all wells are perforated to create
hydraulic fractures or to establish communication with natural fractures. In many cases, there is a damaged zone in
the vicinity of the wellbore. The hydraulic fracture path is sometimes misaligned with the wellbore axis because the
fracture may align itself with either the maximum or intermediate principal in-situ stress. Furthermore, the wells
are not always vertical; instead they are often slightly or considerably deviated. Therefore, the well intersection with
a hydraulic or natural fracture(s) may not be along the wellbore, but will be slightly nonaligned with the wellbore
axis. As we know from the wellbore image logs, most wellbore-intersecting natural fractures are not aligned with
the wellbore axis. If the perforations were damaged and/or not deep enough, then we would have another type
of choked fracture by the wellbore itself. These three types of skin zones significantly affect the pressure transient
behavior of hydraulically or naturally fractured wells. Including the wellbore in the solutions of hydraulically or
naturally fractured wells is therefore essential. For a given wellbore and fracture, these three types of skin zones
may occur simultaneously: wellbore choked, fracture choked, and damaged fracture face. Any combination of two of
them may also occur. For instance, excessive fluid leakoff may take place if the wellbore is choked due to ineffective
and/or damaged perforations. Identifying and removing these three types of damaged skin zones need very different
techniques and procedures.

10
6 FD = 0.01
5
4
FD = 0.1
FD = 1
Derivative, dpD/dlntD

3
FD = 10 1
2 FD = 100
6
FD = 10
1
dpD/dlntD

6 m=0
5
4 CD = 0
0.1 CD = 0.001
3

m = 1/3 m=1 CD = 0.01

2
CD = 0.1
m=1
0.1 m = 1/2
0.0001 0.001 0.01 0.1 1 10 100 1000
tD 0.01
-5 -4 -3 -2 -1 0 1 2
10 10 10 10 10 10 10 10
tD
Figure 30: Derivatives of a discretely fractured reservoir shown in
Fig. 20 of (Kuchuk and Biryukov, 2013) for various finite conduc-
tivities with CD = 0.001 and S = 5 in an infinite reservoir, where Figure 31: Derivative of a single wellbore-intersecting fully-
one of the fractures intersects with the wellbore, from (Kuchuk penetrated infinite-conductivity fracture with skin and storage in
and Biryukov, 2013). an infinite reservoir, from (Kuchuk and Biryukov, 2013).
22 IPTC 18194

Figure. 31 presents the derivatives for a single vertical wellbore-intersecting fully-penetrated-infinite-conductivity

fracture with various dimensionless wellbore-storage values and an S value of 1 (Kuchuk and Biryukov, 2013). For
all the storage cases, the derivatives clearly exhibit a positive unit-slope (m = 1) wellbore-storage dominated period.
The derivative without the wellbore storage effect (CD = 0) exhibits a 1/2-slope formation linear flow regime. As can
be observed from Fig. 31, even if it is distorted, the derivative for CD = 0.001 still reveals the fractured nature of the
well, because after a minimum (dip) we can observe a short 1/2-slope period. For the CD = 0.01 case, it is difficult
to determine if the well intersects an infinite-conductivity fracture. The derivative behavior for the CD = 0.1 case
is strange; after the unit-slope period, it immediately becomes flat (its value approaches 0.5—a radial flow regime),
almost without a transition period after a maximum arch interval due to the wellbore storage [see Fig. 2.13 of Kuchuk
et al. (2010)].

A Field example
This field test example is given by Lee et al. (2003). It is a buildup test from a well in a naturally fractured gas
reservoir. The variable gas flow rate data and pressure-buildup-test data are given in Tables 7.1 and 7.2 of Lee
et al. (2003), respectively. The authors used both semilog [Horner] and type-curve analysis techniques with a dual-
porosity matrix-pseudosteady-state model to estimate the fractured reservoir permeability, skin factor, interporosity
flow coefficient, λ, and storativity ratio, ω. The input model parameters are given in Lee et al. (2003).
The objective of showing this example is not to criticize their interpretation, or to state that their interpretation
is incorrect. Any judgment on the interpretation of these examples would be unfair because the interpretation tools
have been improved significantly in recent years. In fact, the first author of this paper in Barua et al. (1985) also
used a dual-porosity model with pseudosteady-state matrix blocks for the interpretation of the data given by Bourdet
et al. (1983), as shown in Figs. 2 and 4 in this paper.
As shown in Fig. 33, the buildup test derivative exhibits a positive unit-slope wellbore-storage dominated period,
lasting about 6 hours, then a 1/2 slope linear flow regime continues until 100 hr. This is a very long 507-hr test and
the derivative exhibits a linear flow regime towards the end of the test. That means that the wellbore-intersecting
fracture(s) dominates the entire test, except a 6 hr wellbore-storage dominated period. The derivative of the Lee
et al. (2003) buildup data most likely indicates a discretely fractured reservoir because of the lack of unit slope; see
7.
Figure 33 shows the history matches of the pseudopressure change and its derivative of the Lee et al. (2003)
buildup data (denoted as measured) with the pseudopressure change and its derivative computed from the Biryukov
and Kuchuk (2012) NFR model with the fracture and reservoir configurations shown in Fig. 32. As can be seen from
Fig. 33, the model matches the measured data very well and the estimated parameters are given in the Model 2
column of Table 1.
The Lee et al. (2003) buildup data are matched to the three different Warren and Root (1963) dual-porosity
pseudosteady-state matrix block models. As can be observed from Fig. 34, both the pressure change and its derivative
(denoted as measured) are matched very well, except for Model 4, with the three different dual-porosity models.
The estimated values of model parameters for each model are shown in Table 1 as Model 3, Model 4, and Model
5. The data are also matched with a bounded single-porosity model as Model 6. A reasonably good match is
obtained, as shown in Fig. 34.
IPTC 18194 23

Δpseudopressure & derivative, psi /cp

7
10
pseudopressure

2
6
10

secondary fracture FD = 10 main fracture FD = 100 derivative

xfs = 16 ft xf = 196 ft
5
10 m= 1/2
measured
measured
computed
computed

well
4
10
0.01 0.1 1 10 100 1000
Time, hr

Figure 33: Comparison of the pseudopressure change and its

derivative (denoted as computed) from the history match of the
reservoir model given by Fig. 32 to the measured pseudopressure
Figure 32: Fracture configurations in the reservoir model used change and its derivative (denoted as measured) of the Lee et al.
for the history matching of the Lee et al. (2003) buildup data. (2003) buildup data.

Table 1: Estimated parameters from the Lee et al. (2003) buildup data by using various models.

Parameters Model 1 Model 2 Model 3 Model 4 Model 5 Model 6

kf , mD 0.154 3.3E5§ 0.72 1 0.2 NA∗
km , mD 5 0.115 0.00015 0.000216 0.000055 2
C, RB/psi 0.00145 0.4 0.4 0.265 0.4 0.3
S - 3.505 0.58 0 3.53 -3.72 8.5
†
matrix PSS NA PSS PSS PSS SP‡
ω 0.02371 NA 0.01 0.01 0.016 NA
λ 1.78E-8 NA 1.25E-4 1.3E-4 1.65E-4 NA
Lwn , ft ∞ ∞ 5000 5000 ∞ 106
Lws , ft ∞ ∞ 70 65 ∞ 45
Lwe , ft ∞ ∞ 70 65 ∞ 57
Lww , ft ∞ ∞ 5000 5000 5000 5000
§
the fracture aperture b is assumed 0.0066 ft (2 mm).
∗
NA denotes Not Applicable.
†
pseudo-steadystate matrix flow and ‡ single-porosity.
Lwn denotes the distance from the wellbore to the North boundary.
Lws denotes the distance from the wellbore to the south boundary.
Lwe denotes the distance from the wellbore to the east boundary.
Lww denotes the distance from the wellbore to the west boundary.

The Lee et al. (2003) semilog analysis results in Table 1 are given in the Model 1 column. They reported results
from what they called “qualitative type-curve analysis” as ω = 0.00398 and λ =4.19E-5. Lee et al. (2003) stated that
the estimates of ω and λ are in poor agreement with the values from semilog analysis. The lack of closer agreement
is related to the fact that the early-time test data do not overlay the 0.5 line of the derivative type curve. This
suggests that the initial straight line identified in the semilog analysis is not correct. Hence the estimates of ω and
λ are not correct.
The parameter α (a parameter characteristic of the geometry) in the Warren and Root (1963) model is give by
Gringarten (1984) as
4n(n + 2)
α= , n = 1, 2, 3, (46)
`2
where ` is the equivalent matrix medium characteristic length, which we call the characteristic length lV of the
macroscopic representative elementary volume VREV , and n = 3 for the equivalent cubic matrix block. ` is not the
characteristic length (lf ≈ lm , see Fig. 11) of the matrix elements of the actual fractured reservoir. For given λ,
24 IPTC 18194

rw , and kf values, and using n = 3 and ` from Eq. 46 for a cubic matrix bloc, the matrix permeability km can be
calculated as
λkf λ`2 kf
km = 2
= 2
. (47)
αrw 60rw

The matrix permeability for each model given in Table 1 is obtained using Eq. 47 with rw = 0.3 ft and ` = 3 ft. As can
be observed from Table 1, the matrix permeabilities obtained from the dual-porosity models, including the Lee et al.
(2003) estimate, are about one-tenth of a microdarcy. Even for ` = 6 ft, they will still be less than one microdarcy.
As stated above, we cannot make ` too large because it has to satisfy lf  ` ≈ lV  h. Even if we quadruple `,
km values will still be very low. Usually dual-porosity models with the pseudosteady-state condition provide better
history matches with the field data (as in this example, all the matches are very good), but both fracture network
and matrix permeabilities are significantly underestimated for NFRs. It is difficult to imagine a well in a fractured
reservoir that can produce about 150 Mscf/D gas from a fractured system permeability of 0.1 mD and the matrix
0.1 microdarcy. Our analysis indicates that the well intersects a reasonable high conductivity fracture, which in turn
intersects a few minor fractures. The matrix permeability we estimated is about 0.1 mD.
Due to the lack of geoscience data, we cannot claim that

Δpseudopressure & derivative, psi /cp

our interpretation is unique. Our fractured reservoir model 10
7
pseudopressure

2
is one of many possible models. On the other hand, un-
likely the dual-porosity model, it is a realistic one for the
6
Lee et al. (2003) data. This is a prime example for the 10

nonuniqueness of the well testing inversion problem, which

can only be addressed by constraining the model with geo- measured
measured
5 model 3
science data during the inversion process. For instance, if 10 m= 1/2 derivative model 3
model 4
model 4
we had wellbore images for the Lee et al. (2003) data, we model 5
model 5
model 6
could have eliminated some other plausible reservoir mod- model 6
4
els. 10
0.01 0.1 1 10 100 1000
Time, hr

Figure 34: Comparison of the pseudopressure change and its

derivative (denoted as computed) from the history match of a
various models to the measured pseudopressure change and its
derivative (denoted as measured) of the Lee et al. (2003) buildup
Conclusions data.
In this paper, modeling continuously and discretely fractured reservoirs is examined from the microscopic and macro-
scopic representative elementary volume (REV) point of view. It is shown that dual-porosity models consist of two
overlapping continua: 1) The equivalent homogenous medium of the network of fractures, and 2) The equivalent
homogenous medium of the matrix elements. Both equivalent fractured and matrix media are homogenized or up-
scaled versions of the actual fractures and matrix blocks. Therefore, using fracture and matrix face skin factors due
to calcite deposits and mineralization (these are not the wellbore damage skin factors), and multiple matrix blocks
in the same REV in dual-porosity models is not consistent with homogenization or upscaling. For the same reason,
the microscopic level local variability of matrix properties do not show a triple-porosity behavior.
It is shown that if the permeabilities of the fracture segments are a few orders of magnitude larger than those of
the matrix blocks, but not km  kf , then a macroscopic representative elementary volume can be constructed, and
therefore dual-porosity models can be used for fractured reservoirs. If km  kf , then a macroscopic REV cannot be
constructed, therefore analytical or numerical techniques should be used to model the transient pressure behavior of
fractured reservoirs without fracture segment homogenization or upscaling.
The resistance interface condition, which specifies fluid transport between the equivalent matrix and fractured
media, dominates the dual-porosity model pressure behavior, but is not the actual pseudosteady-state flow condition
in the matrix medium. The resistance interface condition for pseudosteady-state flow and the interporosity skin with
transient flow in the matrix are unphysical, and introduce serious nonuniqueness problems for dual-porosity models.
It is shown that wellbore-intersecting fractures dominate the pressure transient behavior of both continuously
and discretely NFRs. If the well is in the matrix, we observe a matrix-block radial flow regime until the effects of
the fractures are felt in the wellbore. After the first radial flow, the derivatives go downward rapidly at early times
until they reach a minimum, depending on the conductivities.
It is shown that the inner boundary (wellbore) used in conventional dual-porosity models are incorrect if any
fractures intersect the wellbore, even though the REV conditions are satisfied. The appropriate inner boundary
conditions with wellbore storage and skin effects for NFRs are prescribed, and three types of damaged skin factors
are specified.
Based on these grounds, the dual-porosity models are not approximate, basic, or simple models for the naturally
fractured reservoirs. They can be used for a few specific cases of NFRs only.
IPTC 18194 25

It is shown that the triple-porosity model behavior is not due to variable matrix block size in a REV, but is rather
due to the spatial variation of conductivity, length, density, and orientation of fracture distributions.
Finally, field buildup test data from an NFR are interpreted using both the conventional dual-porosity and our
fractured models. It is concluded that the conventional dual-porosity model with the pseudosteady-state matrix flow
condition is not appropriate for this test and significantly underestimates both fracture and matrix permeabilities.
Nomenclature
C = wellbore storage constant, RB/psi [m3 /Pa]
ct = compressibility
F = fracture conductivity
h = formation thickness
k = permeability
l = fracture half-length
p = pressure
q = flow rate
r = radius or radial coordinate
S = Skin factor
t = time
x = coordinate
y = coordinate
z = vertical coordinate
η = diffusivity for pressure
µ = viscosity
φ = porosity
Subscripts
D = dimensionless
f = fracture
o = initial or original
m = matrix
w = wellbore

Acknowledgments
The authors are grateful to Schlumberger for permission to publish this paper, and would like to thank Kirsty Morton and
Amine Ennaifer for their contributions for understanding pressure transient tests from NFRs, particularly for Amine performing
the history matching for the field example with the dual-porosity models.

References
Aarseth, E. S., Bourgine, B., Castaing, C., Chiles, J. R., Christensen, N. R., Eeles, M., Fillion, E., Genter, A., Gillespie, R. A., Hakansson,
E., Jorgensen, K. Z., Lindgaard, H. F., Madsen, L., Odling, N. E., Olsen, C., Reffstrup, J., Trice, R., Walsh, J. J., and Watterson, J.
1997. Interim guide to fracture interpretation and flow modeling in fractured reservoirs. Technical Report EUR 17116 EN, European
Commission.

Abdassah, D. and Ershaghi, I. 1986. Triple-porosity systems for representing naturally fractured reservoirs. SPE Formation Evaluation,
1(2):113–127.

Akbar, M., Vissapragada, B., Alghamdi, A. H., Allen, D., Herron, M., Carnegie, A., Dutta, D., Olesen, J.-R., Chourasiya, R. D., Logan,
D., Stief, D., Netherwood, R., Russell, S. D., and Saxena, K. 2001. A snapshot of carbonate reservoir evaluation. Oilfield Review,
12(4):20–41.

Ayestaran, L., Nurmi, R., Shehab, G., and El Sisi, W. 1989. Well test design and final interpretation improved by integrated well testing
and geological efforts. Paper SPE 17945, Middle East Oil Show, Bahrain, 11–14 March.

Barenblatt, G. I., Zeltov, Y. P., and Kochina, I. 1960. Basic concepts in the theory of seepage of homogeneous liquids in fissured rocks.
PMM (Journal of Soviet Applied Mathematics and Mechanics), 24(5):1286–1303.

Barua, J., Kucuk, F., and Gomez-Angulo, J. 1985. Application of computers in the analysis of well tests from fractured reservoirs. Paper
SPE 13662, SPE California Regional Meeting, Bakersfield, CA.

Bear, J. 1993. Flow and Contaminant Transport in Fractured Rock, chapter Modeling Flow and Contaminant Transport in Fractured
Rocks, pages 1–37. Bear, J. and Tsang, C.F. and de Marsily, G, (editors). Academic Press, Inc, San Diego, CA., US, 1st edition.

Belani, A. and Jalali, Y. 1988. Estimation of matrix block size distribution in naturally fractured reservoirs. Paper SPE 18171, SPE
Annual Technical Conference and Exhibition, Houston, Texas, 2-5 October.

Belfield, W. and Sovich, J. 1994. Fracture statistics from horizontal wellbores. Paper HWC-94-37, SPE/CIM/CANMET International
Conference on Recent Advances in Horizontal Well Applications, March 20–23, Calgary, Canada.

Berkowitz, B. 2002. Characterizing flow and transport in fractured geological media: A review. Advances in Water Resources, 25(8–
12):861–884.
26 IPTC 18194

Biryukov, D. and Kuchuk, F. 2012. Transient pressure behavior of reservoirs with discrete conductive faults and fractures. Transport in
Porous Media, 95:239–268. 10.1007/s11242-012-0041-x.

Bogdanov, I., Mourzenko, V., Thovert, J.-F., and Adler, P. 2003. Pressure drawdown well tests in fractured porous media. Water Resour.
Res., 39(1):1021. doi:10.1029/2000WR000080.

Booth, R., Morton, K., Onur, M., and Kuchuk, F. 2010. Grid-based inversion of pressure transient test data. Proceedings of European
Conference on the Mathematics of Oil Recovery XII. Oxford, UK.

Booth, R., Morton, K., Onur, M., and Kuchuk, F. 2012. Grid-based inversion of pressure transient test data with stochastic gradient
techniques. Int. J. Uncertainty Quantification, 2(4):395–405.

Bourdet, D. 1985. Pressure behavior of layered reservoirs with crossflow. Paper SPE-13628-MS, SPE California Regional Meeting, 27-29
March, Bakersfield, California.

Bourdet, D., Whittle, T., Douglas, A., and Pirard, Y. 1983. A new set of type curves simplifies well test analysis. World Oil, 6:95–106.

Braester, C. 2009. Groundwater Flow Through Fractured Rocks, pages 22–42. GROUNDWATER –Vol. II, Edited by Luis Silveira and
Eduardo J. Usunoff. EOLSS/ UNESCO, 1st edition.

Carslaw, H. C. and Jaeger, J. C. 1959. Conduction of Heat in Solids. Clarendon Press, Oxford.

Casabianca, D., Jolly, R. J. H., and Pollard, R. 2007. The machar oil field: waterflooding a fractured chalk reservoir. Geological Society,
London, Special Publications, 270(1):171–191.

Casciano, C., Ruvo, L., Volpi, B., and Masserano, F. 2004. Well test simulation through discrete fracture network modelling in a fractured
carbonate reservoir. Petroleum Geoscience, 10(4):331–342.

Chatas, A. 1966. Unsteady spherical flow in petroleum reservoirs. Society of Petroleum Engineers Journal, 6(02):102–114.

Cinco-Ley, H. and Samaniego-V, F. 1981. Transient pressure analysis: Finite conductivity fracture case versus damaged fracture case.
Paper SPE 10179, SPE Annual Technical Conference and Exhibition, San Antonio, Texas, 4–7 October.

Cinco-Ley, H., Samaniego-V, F., and Kuchuk, F. 1985. The pressure transient behavior for naturally fractured reservoirs with multiple
block size. Paper SPE 14168, SPE Annual Technical Conference and Exhibition, Las Vegas, NV., September 22–25.

Committee on Fracture Characterization and Fluid Flow 1996. Rock Fractures and Fluid Flow:Contemporary Understanding and Appli-
cations. The National Academies Press.

Daniel, E. 1954. Fractured reservoirs of middle east. AAPG Bulletin, 38(5):774–815.

Davidson, D. and Snowdon, D. 1978. Beaver river middle devonian carbonate: Performance review of a high-relief, fractured gas reservoir
with water influx. Journal of Petroleum Technology, 30(12):1672–1678.

de Swaan–O, A. 1976. Analytic solutions for determining naturally fractured reservoir properties by well testing. SPE J., Trans. AIME
261:117–122.

Doonechaly, N. G., Rahman, S., and Cinar, Y. 2013. A new finite-element numerical model for analyzing transient pressure response of
naturally-fractured reservoirs. Paper SPE 166477-MS, SPE Annual Technical Conference and Exhibition, New Orleans, Louisiana, 30
September- 2 October 2013.

Freudenreich, Y., Angerer, E., del Monte, A. A., and Reiser, C. 2006. Characterizing fracture networks: an effective approach using
seismic anisotropy attributes. First Break, 24(6):67–72.

Gringarten, A. 1984. Interpretation of tests in fissured and multilayered reservoirs with double-porosity behavior: Theory and practice.
J. of Petroleum Technology, 36(4):549–564.

Hahn, D. W. and Ozisik, M. N. 2012. Heat Conduction. Wiley, New York, 3 edition edition.

Houze, O. P., Horne, R. N., and Ramey, H. J. 1988. Infinite conductivity vertical fracture in a reservoir with double porosity behavior.
SPE Formation Evaluation, 3(3):510–518.

Izadi, M. and Yildiz, T. 2009. Transient flow in discretely fractured porous media. SPE Journal, 14(02):362–373.

Jaeger, J. 1955. Conduction of heat in a solid in contact with a thin layer of a good conductor. Quart. J. Mech. Appl. Math, 8:101–106.

Johns, R. and Jalali, Y. 1991. Comparison of pressure-transient response in intensely and sparsely fractured reservoirs. SPE Formation
Evaluation, 6(4):513–518.

Kazemi, H. 1969. Pressure transient analysis of naturally fractured reservoirs with uniform fracture distribution. SPE J., 9(4):451–462.

Kikani, J. and Walkup, G. W. 1991. Analysis of pressure-transient tests for composite naturally fractured reservoirs. SPE Formation
Evaluation, 6(2):176–182.

Kuchuk, F. 1994. Pressure behavior of MDT packer module and DST in crossflow-multilayer reservoirs. J. of Petroleum Science and
Engineering, 11 (2):123–135.

Kuchuk, F. and Biryukov, D. 2012. Transient pressure test interpretation from continuously and discretely fractured reservoirs. Paper
SPE 158096, SPE Annual Technical Conference and Exhibition, San Antonio, Texas, 8–10 October.
IPTC 18194 27

Kuchuk, F. and Biryukov, D. 2013. Pressure transient tests and flow regimes in fractured reservoirs. Paper SPE 166296-MS, SPE Annual
Technical Conference and Exhibition, New Orleans, Louisiana, 30 September- 2 October 2013.

Kuchuk, F. and Biryukov, D. 2014. Transient pressure test interpretation from continuously and discretely fractured reservoirs. SPE
Reservoir Evaluation & Engineering, 17(01):82–97.

Kuchuk, F., Biryukov, D., and Fitzpatrick, T. 2014. Rate transient and decline curve analyses for continuously (dual-porosity) and
discretely naturally fractured reservoirs. Paper SPE 170698-MS, SPE Annual Technical Conference and Exhibition, Amsterdam, The
Netherlands, 27–29 October.

Kuchuk, F., Goode, P. A., Wilkinson, D. J., and Thambynayagam, R. K. M. 1991. Pressure-transient behavior of horizontal wells with
and without gas cap or aquifer. SPE Formation Evaluation, 6(1):86–94.

Kuchuk, F., Onur, M., and Hollaender, F. 2010. Pressure Transient Formation and Well Testing: Convolution, Deconvolution and
Nonlinear Estimation. Elsevier, New York.

Kuchuk, F. J. and Wilkinson, D. 1991. Pressure behavior of commingled reservoirs. SPE Formation Evaluation, 6 (1):111–120.

Lee, W., Rollins, J., and Spivey, J. 2003. Pressure Transient Testing. SPE Textbook Series Vol. 9. Society of Petroleum Engineers,
Richardson, Texas.

Liu, C. Q. 1981. Exact solution for the compressible flow equations through a medium with triple-porosity. Applied Mathematics and
Mechanics, 2(4):457–462.

Moench, A. 1984. Double–porosity models for a fissured groundwater reservoir with fracture skin. Water Resource Research, 20:831–846.

Morton, K., Booth, R., Chugunov, N., Fitzpatrick, A., and Kuchuk, F. 2013. Global sensitivity analysis for natural fracture geological
modeling parameters from pressure transient tests. Paper SPE 164894, SPE EUROPEC/EAGE Annual Conference and Exhibition,
10–13 June, London, United Kingdom.

Morton, K., de Brito Nogueira, P., Booth, R., and Kuchuk, F. 2012. Integrated interpretation for pressure transient tests in discretely
fractured reservoirs. Paper SPE 154531, SPE EUROPEC/EAGE Annual Conference and Exhibition, 4–7 June, Copenhagen, Denmark.

Muskat, M. 1937. The Flow of Homogeneous Fluids Through Porous Media. J. W. Edwards, Inc., Ann Arbor, Michigan.

Nelson, R. A. 1985. Geologic Analysis of Naturally Fractured Reservoirs. Gulf Publishing, Houston, Houston, Texas, 1st edition.

Nurmi, R., Kuchuk, F., Cassell, B., Chardac, J.-L., and Maguet, L. 1995. Horizontal highlights. Schlumberger Middle East Well
Evaluation Review, 16:6–25.

Odling, N. E. 1997. Scaling and connectivity of joint systems in sandstone from western norway. J. of Structural Geology, 19(10):1257–
1271.

Ozkan, E. and Raghavan, R. 1991. New solutions for well–test–analysis problems: Part 1 – analytical considerations. SPE Formation
Evaluation, 6(3):359–368.

Pollard, D. D. and Aydin, A. 1988. Progress in understanding jointing over the past century. Geological Society of America Bulletin,
100(8):1181–1204.

Raghavan, R. 1993. Well Test Analysis. Prentice Hall, Boston.

Ramakrishnan, T. S., Ramamoorthy, R., Fordham, E., Schwartz, L., Herron, M., Saito, N., and Rabaute, A. 2001. A model-based
interpretation methodology for evaluating carbonate reservoirs. Paper SPE 71704, SPE Annual Technical Conference and Exhibition,
New Orleans, Louisiana, 30 September3 October.

Riley, M., Brigham, W., and Horne, R. 2007. Analytic solutions for elliptical finite-conductivity fractures. Paper SPE 22656, SPE Annual
Technical Conference and Exhibition, 6-9 October, Richardson, Texas.

Rogers, S., Enachescu, C., Trice, R., and Buer, K. 2007. Integrating discrete fracture network models and pressure transient data for
testing conceptual fracture models of the valhall chalk reservoir, norwegian north sea. Geological Society, London, Special Publications,
270(1):193–204.

Sedov, L. I. 1954. Metody podopiya i razmernostei v mechanike (in Russian). GITTL.

Serra, K., Reynolds, A., and Raghavan, R. 1983. New pressure transient analysis methods for naturally fractured reservoirs. Journal of
Petroleum Technology, 35(12):2271–2283.

Soliman, M. 2009. Transient Well Testing, chapter 11–Well-test analysis of hydraulically fractured wells, pages 281–331. Monograph
Series 23, Edited by M. Kamal. Society of Petroleum Engineers, Richardson, 1st edition.

Souche, L. and Rotsch, M. 2007. An end-to-end approach to naturally fractured reservoir modeling: Workflow and implementation.
SEG/EAGE Research Workshop 2007, Perugia, Italy, September 3-6.

Spivey, J. P. and Lee, W. J. 2000. Pressure transient response for a naturally fractured reservoir with a distribution of block sizes. Paper
SPE 60294, SPE Rocky Mountain Regional/Low-Permeability Reservoirs Symposium and Exhibition, Denver, Colorado, 12-15 March.

Stewart, G. and Asharsobbi, F. 1988. Well test intepretation for naturally fractured reservoirs. Paper SPE-18173-MS, SPE Annual
Technical Conference and Exhibition, 2-5 October, Houston, Texas.
28 IPTC 18194

van Everdingen, A. and Hurst, W. 1949. The application of the laplace transformation to flow problems in reservoirs. Trans., AIME,
186:305–324.

van Golf-Racht, T. D. 1982. Fundamentals of fractured reservoir engineering. Elsevier Scientific Publ. Co., Inc., New York.

Warren, J. E. and Root, P. J. 1963. The behavior of naturally fractured reservoirs. SPE J., 3(3):245–255.

Wei, L., Hadwin, J., Chaput, E., Rawnsley, K., and Swaby, P. 1998. Discriminating fracture patterns in fractured reservoirs by pressure
transient tests. Paper SPE 49233, SPE Annual Technical Conference and Exhibition, New Orleans, Louisiana, 27–30 September.

Wong, D., Harrington, A., and Cinco-Ley 1986. Application of the pressure derivative function in the pressure transient testing of
fractured wells. SPE Formation Evaluation, 1(5):470–480.

Zaman, A., Yousaf, M., and Ahmed, S. 1989. Well test design and final interpretation improved by integrated well testing and geological
efforts. Paper SPE 17998, Middle East Oil Show, Bahrain, 11–14 March.