Clarkson (2013)

International Journal of Coal Geology 109110 (2013) 101146
Contents lists available at SciVerse ScienceDirect
International Journal of Coal Geology

journal homepage: www.elsevier.com/locate/ijcoalgeo
Review article
Production data analysis of unconventional gas wells: Review of

theory and best practices
C.R. Clarkson
Department of Geoscience, University of Calgary, 2500 University Drive NW Calgary, Alberta, Canada T2N 1N4
a r t i c l e
i n f o
Article history:
Received 28 June 2012
Received in revised form 3 January 2013
Accepted 3 January 2013
Available online 11 January 2013
Keywords:
Production data analysis
Rate transient analysis
Coalbed methane
Shale gas
Review
Best practices
a b s t r a c t
Unconventional gas reservoirs, including coalbed methane (CBM), tight gas (TG) and shale gas (SG), have
become a signicant source of hydrocarbon supply in North America, and interest in these resource plays
has been generated globally. Despite a growing exploitation history, there is still much to be learned about
uid storage and transport properties of these reservoirs.
A key task of petroleum engineers and geoscientists is to use historical production (reservoir uid production
rate histories, and cumulative production) for the purposes of 1) reservoir and well stimulation characterization and 2) production forecasting for reserve estimation and development planning. Both of these subtasks
fall within the domain of quantitative production data analysis (PDA). PDA can be performed analytically,
where physical models are applied to historical production and owing pressure data to rst extract information about the reservoir (i.e. hydrocarbon-in-place, permeability-thickness product) and stimulation (i.e. skin
or hydraulic fracture properties) and then generate a forecast using a model that has been calibrated to the
dynamic data (i.e. rates and pressures). Analytical production data analysis methods, often referred to as
rate-transient analysis (RTA), utilize concepts analogous to pressure-transient analysis (PTA) for their implementation, and hence have a rm grounding in the physics of uid storage and ow. Empirical methods, such
as decline curve analysis, rely on empirical curve ts to historical production data, and projections to the future. These methods do not rigorously account for dynamic changes in well operating conditions (i.e. owing
pressures), or reservoir or uid property changes.
Quantitative PDA is now routinely applied for conventional reservoirs, where the physics of uid storage and
ow are relatively well-understood. RTA has evolved extensively over the past four decades, and empirical
methods are now applied with constraints and rules of thumb developed by researchers with some condence. For unconventional reservoirs, these techniques continue to evolve according to our improved understanding of the physics of uid storage and ow.
In this article, the latest techniques for quantitative PDA including type-curve analysis, straight-line
(ow-regime) analysis, analytical and numerical simulation and empirical methods are briey reviewed, specically addressing their adaptation for CBM and SG reservoirs. Simulated and eld examples are provided to
demonstrate application. It is hoped that this article will serve as practical guide to production analysis for
unconventional reservoirs as well as reveal the latest advances in these techniques.
2013 Elsevier B.V. All rights reserved.
Contents
1.
2.
3.
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Concept of rate transient analysis: ow-regimes
. . . . . . . . . . . . . . . . . . . . . . . . .
Production analysis methods
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.1.
Straight-line (or ow-regime) analysis methods . . . . . . . . . . . . . . . . . . . . . .
3.1.1.
Additional considerations for CBM and shale reservoirs
. . . . . . . . . . . . . .
3.1.2.
Simulated examples of CBM and shale straight-line analysisCases 1 and 2dry shale
3.1.3.
Simulated examples of CBM and shale straight-line analysisCases 3 and 4wet coal
3.2.
Type-curve methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.2.1.
Fetkovich type-curves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.2.2.
Blasingame type-curves . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Tel.: +1 403 220 6445.

E-mail address: clarksoc@ucalgary.ca.
0166-5162/$ see front matter 2013 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.coal.2013.01.002
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
102
103
106
106
110
113
116
116
118
120
102
C.R. Clarkson / International Journal of Coal Geology 109110 (2013) 101146
3.2.3.
Wattenbarger type-curves . . . . . . . . . . . . . . . . .
3.2.4.
Multi-fractured horizontal well type-curves . . . . . . . . .
3.2.5.
Other useful type-curves . . . . . . . . . . . . . . . . . .
3.2.6.
Alternative type-curve matching approaches . . . . . . . . .
3.2.7.
Additional considerations for CBM and shale reservoirs . . . .
3.3.
Analytical and numerical simulation . . . . . . . . . . . . . . . . .
3.4.
Empirical methods . . . . . . . . . . . . . . . . . . . . . . . . .
3.4.1.
Application of Arps decline-curve methodology for CBM wells
3.4.2.
Application of Arps decline-curve methodology for tight gas and
3.4.3.
New empirical approaches for tight gas and shale . . . . . .
3.5.
Hybrid methods . . . . . . . . . . . . . . . . . . . . . . . . . .
4.
Field examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.1.
Field example 1: 2-phase CBM well (vertical) . . . . . . . . . . . . .
4.2.
Field example 2: 1-phase shale gas well (MFHW) . . . . . . . . . . .
5.
Discussion: future development of production data analysis techniques . . . .
5.1.
Analytical (type-curves, straight-line, analytical simulation) . . . . . .
5.2.
Empirical
. . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.3.
Hybrid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6.
Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Nomenclature. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Nomenclature
Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1. Introduction
Advances in reservoir evaluation and drilling/completion technology have enabled commercial production of hydrocarbons from unconventional reservoirs over the past few decades, including natural
gas from coal (coalbed methane or CBM), low-permeability (tight)
gas (TG) and shale gas (SG) reservoirs. In North America, these resource types now contribute signicantly to hydrocarbon supply;
with recent suppression of gas prices, the focus of producers has
been on liquids-rich (i.e. oil, gas condensate) unconventional plays.
The success of unconventional plays in North America has triggered
global activity in the area of unconventional reservoir exploration
and exploitation.
Economics of unconventional reservoirs is tied closely to wellperformance (Haskett and Brown, 2005). Operators have therefore focused extensively on optimizing well performance and hydrocarbon
recovery in the past few decades. Advances in wellbore architecture
design (i.e. horizontal wells), and stimulation technology (i.e. hydraulic fracturing) have enabled commercial production from reservoirs
previously thought of as only source rocks (i.e. coals and organicrich shales). Additional improvements in the areas of core analysis,
well-log analysis, well-test analysis, remote sensing, reservoir surveillance and production technology have allowed operators to better
locate wellbores in the sub-surface to exploit these reservoirs, more
effectively design stimulation programs, and more efciently operate
the wells to achieve maximum performance.
An important reservoir engineering technique that is used for reservoir characterization and evaluation, and for production forecasting, is quantitative production data analysis (PDA). Rate-transient
analysis (RTA) is a relatively new form of PDA, where uid production
data, along with owing pressure information, is used to extract reservoir (i.e. hydrocarbons-in-place and permeability-thickness product) and stimulation information (i.e. skin and hydraulic fracture
properties). The theory behind RTA is exactly analogous to pressure
transient analysis (PTA), and hence the techniques used for RTA are
very similar to PTA. These techniques include production type-curve
analysis, straight-line (ow-regime) analysis and analytical and
numerical simulation. RTA is essentially an inversion problem
where reservoir attributes are obtained from the reservoir signal,
i.e. rate and/or owing pressure data. Once this inversion is solved
for key attributes, these attributes along with other reservoir/uid/
. . .
. . .
. . .
. . .
. . .
. . .
. . .
. . .
shale
. . .
. . .
. . .
. . .
. . .
. . .
. . .
. . .
. . .
. . .
. . .
. . .
. . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
gas wells
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
121
123
125
126
126
127
130
131
132
133
135
137
138
139
141
141
141
142
142
142
143
143
143
well/operating condition data can be used in a model that accurately

captures the physics of the reservoir/well system to generate a forecast of future hydrocarbon production. The RTA approach, combined
with analytical and numerical simulation for forecasting, requires
models for uid storage and transport. These techniques are referred
to as advanced production data analysis methods and require both
production and owing pressure data, along with uid property and
reservoir (volumetric) data to arrive at a quantitative result. Signicant development of the RTA method has been made for conventional
reservoirs in the past four decades.
Empirical methods, such as decline curve analysis, can also be
used for forecasting. For conventional reservoirs, the Arps (1945)
hyperbolic/exponential decline forecasting method remains the
most popular, but because it is an empirical technique, its use can
lead to signicant errors, even for conventional reservoirs. As will
be discussed in detail below, the assumptions behind the Arps decline curve method (i.e. constant operating conditions, static reservoir behavior) are very often violated, causing the method to yield
erroneous results. Ideally, advanced PDA methods should be used
in parallel with empirical methods to constrain the decline-curve
forecast. Workows to ensure this consistency have been provided
in the literature for conventional reservoirs (i.e. Fetkovich et al.,
1996; Mattar and Anderson, 2003) and were provided in an accompanying article by the author (Clarkson, 2013-this volume).
Advanced and empirical production data analysis methods for
unconventional reservoirs continue to evolve. The primary complications associated with the adaptation of these methods for unconventional include complex reservoir behavior, as well as ow behavior
associated with complex wellbore/hydraulic fracture geometry. Analytical solutions to ow equations, on which the advanced PDA techniques are based, typically make many simplifying assumptions
about the reservoir/uids/and hydraulic fracture, including (among
others) slightly-compressible uid behavior, homogeneous and static
reservoir properties.
For shale gas and coal reservoirs, uid storage and ow properties
that complicate analysis include (Clarkson et al., 2012a):
1. Desorption of gas from organic matter and clays (CBM and
organic-rich shale)
2. Ultra-low matrix permeability, which causes transient ow periods
to be extensive (SG)
3. Dual porosity or dual permeability behavior caused by natural fractures or induced hydraulic fractures (or both) (CBM and SG)
4. Additional reservoir heterogeneities, such as multi-layers (ex.
contracting permeability and gas contents in coal seams) and
lateral heterogeneity (ex. coal seam pinchouts)
5. Stress-dependent porosity and permeability caused by a highlycompressible fracture pore volume (CBM and SG)
6. Desorption-dependent fracture porosity and permeability due to
matrix shrinkage effects (CBM)
7. Multi-phase ow of gas and water (2-phase CBM and SG), or gas
and condensate (SG)
8. Non-Darcy ow, including slip-ow and diffusion (CBM and SG)
reservoir. This is a signicant challenge in RTA of SG wells: the separation of reservoir from hydraulic fracture effects.
In CBM, the stimulation treatment can also signicantly affect production characteristics. The two most common methods of stimulating
vertical wells are through the use of hydraulic fracture stimulation and
cavity completions (Palmer, 2010). Horizontal well completions are
also used to exploit lower permeability coals, with or without stimulation of the wellbore.
Production data analysis involves the analysis of production rate
signatures. These signatures are a consequence of the uid ow dynamics in the reservoir, which are often referred to as ow-regimes.
As we will see, the unique reservoir properties, along with completion and stimulation style of coal and shale have a profound impact
on the type and sequence of ow-regimes that are seen, and hence
the methods used for analysis. We begin with an introduction to the
types of ow-regimes encountered for CBM and SG, followed by a discussion of the different types of production data analysis methods
that are currently being used to analyze them. Lastly, we will demonstrate application of the PDA techniques to CBM well and shale gas
wells.
As we will see, correction for some of these affects (1., 5., 6., 7., 8.) requires alteration of primary variables used for RTA, primarily pressure
and time, to pseudovariables that incorporate changes in uid properties, desorption and reservoir properties. If this is done carefully, data
may be converted using pseudovariables allowing the use of RTA
methods designed to analyze single-phase production of slightlycompressible uids, which is the assumption made in existing analytical
solutions to the ow equations. Item 2 suggests that in many shale gas
wells, boundary-dominated ow will not be reached in many wells in a
reasonable time frame, meaning that estimates of hydrocarbon-inplace ultimately contacted by the well will not be accurate. Reservoir
heterogeneities can be accommodated with the choice of solutions
and methods that incorporate these effects; we note however, that
the solutions often greatly oversimplify the true geologic characteristics
of the reservoir.
Not only do reservoir characteristics of SG and CBM impact RTA,
but also the hydraulic fracture network that is created during stimulation. Short and long-term production characteristics of unconventional gas reservoirs can be impacted by the fracture geometry
created through the stimulation process. Because multi-fractured
horizontal wells are now primarily used for exploitation of SG reservoirs, quantitative hydraulic fracture characterization has proven
difcultcreated fracture geometries often do not conform to the
conventional bi-wing planar fracture geometry assumed in conventional reservoirs (Fig. 1). More complex geometries can be created,
often by design, to maximize reservoir contacted in ultra-low permeability reservoirs. When complex geometries are created, resulting in
a stimulated reservoir volume (SRV) (Mayerhofer et al., 2010), hydraulic fracture and reservoir characterization are inseparable, as
the induced hydraulic fracture network often serves to dene the
Simple Fracture
2. Concept of rate transient analysis: ow-regimes

As mentioned in the Introduction, advanced analytical methods
used for quantitative production analysis are classied as ratetransient analysis methods, which are analogous to pressure-transient
analysis methods used to analyze well-test data. RTA essentially involves the analysis of a particular form of well-test, referred to as a
drawdown test, where the well is produced against known wellbore
constraints over a long period of time. In conventional well-test analysis, a short drawdown period typically precedes a lengthier shut-in/
pressure buildup periodone or both of these test periods may be analyzed for reservoir and/or hydraulic fracture properties for stimulated
wells. These tests are typically conducted for a period of days, or perhaps weeks, under highly controlled conditions, with data collected frequently (often on the order of seconds) and accurately (with test
separators for ow periods, and quartz gauges for pressure measurements). For RTA, the drawdown period is essentially the producing
life of the well, possibly interrupted by shut-in periods, with less frequently acquired datarates and owing pressures (if measured) are
often only recorded on a daily basis (at best), but often less frequently.
The data quality for analysis is therefore rarely as good as for PTA, but it
does represent a longer period of time.
Complex Fracture
Fracture Half-Length
Complex Fracture
With Fissure Opening
103
Complex Fracture
Network
Fracture width
Fig. 1. Spectrum of fracture geometries expected for conventional and unconventional reservoirs. Complex fracture networks have nearly equal width and length.
Modied from Warpinski et al. (2008).
104
Both RTA and PTA start with the identication of ow-regimes,

which are characteristic ow patterns or geometries in the reservoir,
which may be analyzed for reservoir and/or hydraulic fracture properties for stimulated wells. These ow patterns in the short-term
are affected by ow to the wellbore or hydraulic fracture network
for stimulated wells, and in the long term by heterogeneities in the
reservoir and ow boundaries. One of the simplest sequences that
can be observed is for a non- or slightly-stimulated vertical well completed in a homogeneous, isotropic reservoir bounded on all sides
(Fig. 2, left side). In the example of Fig. 2, the ow rate at the well is
assumed to be constant (drawdown test), resulting in an early transient radial ow period as the pressure drops in the well-bore (see
Fig. 2a cross-section view, green dashed lines), and a pressure transient propagates out radially away from the well (see Fig. 2b plan
view). This early transient radial period can be interpreted for permeability and skin, if it can be observed in the RTA signature. Slightly
stimulated, or cavity completed CBM wells may exhibit this signature,
for example. Once the pressure transient contacts all boundaries, the
boundary-dominated ow period is entered, and the reservoir depletes like a tank. During boundary-dominated ow, the pressure
drops uniformly throughout the reservoir and can be interpreted to
derive reservoir volume, and most importantly, hydrocarbons-inplace. Observation of this ow-regime is critical to obtain quantitative
estimates of reserves.
The most common way to identify ow-regimes for pressuretransient analysis is to use a pressure-derivative, or rate-normalized
pressure-derivative, versus time on a loglog plot (Fig. 2, right side).
The example in Fig. 2 is for a numerically-simulated vertical well
completed in a dry (gas-production only), homogeneous, isotropic
coal reservoir with instantaneous desorption producing against a
constant owing pressure. The derivative was calculated using the
method of Bourdet (Bourdet et al., 1983, 1989). The time function in
this example has been modied to allow the constant owing
pressure scenario to be converted to the equivalent constant rate condition, and to account for desorption and gas property changes with
pressure (discussed below). The transient radial ow period is identied as a zero slope, and the boundary-dominated ow period, as a
unit slope. For noisy data, as is often seen with production data,
other techniques for ow-regime identication must be considered
(discussed below). Once the ow-regimes have been identied, the
various production analysis methods can be applied.
For hydraulically-fractured wells, the sequence of ow-regimes
may be more complex (Fig. 3, left side) (Cinco-Ley and Samaniego-V,
1981). The example in Fig. 3 is for a vertical well with an inniteconductivity planar hydraulic fracture, completed in a homogeneous
tight gas reservoir with closed boundaries. The sequence of owregimes includes: early (formation) linear ow to the hydraulic fracture; elliptical ow as the pressure transient moves past the ends of
the fracture; pseudoradial ow (analogous to transient radial ow in
the previous example); and boundary-dominated ow. If the fractures
are of nite conductivity (discussed below), at early time, there may
be an additional pressure drop/linear ow along the fractures occurring simultaneously with formation linear owthis ow period is
referred to as bilinear ow. At very early times in the nite conductivity fracture case, there may be a fracture linear ow period, but
this ow period is generally too short to be observed with production
data.
As with the slightly-stimulated well case described above, derivative techniques can be used to identify ow-regimes (Fig. 3, right
side). The example in Fig. 3 is a numerically-simulated vertical well
with a single planar, innite conductivity fracture, completed in a homogeneous, isotropic tight gas reservoir producing against a constant
owing pressure. The time function in this example has been modied to convert the constant owing pressure scenario to the equivalent constant rate condition, and to account for gas property changes
with pressure (discussed below). The formation linear ow period is
3600
3400
3200
a)
3000
2800
2600
Transient Flow
derive k, skin
Boundary Dominated Flow

derive volume
2400
2200
2000
Radial Derivative Plot
b)
d(m(p)/q)/dlnt*ca
1.0E+6
1.0E+5
Radial Flow
(zero-slope)
1.0E+4
1.0E+3
1.0E+0
Boundary-Dominated flow
(unit-slope)
1.0E+1
1.0E+2
1.0E+3
1.0E+4
t*ca, days
Fig. 2. Left side: ow-regimes associated with a slightly-stimulated vertical well subject to constant ow-rate constraint. Pressure changes at the well and the reservoir are shown in
a) cross-section and b) plan views. Lines correspond to isopotential lines; arrows are streamlines. During transient ow (green dashed lines), the pressure transient propagates
radially away from the well in plan view (b) and the well-bore pressure drops (cross-section view). During this ow period, the owing pressure signature over time can be
interpreted for reservoir permeability and near wellbore skin. During boundary-dominated ow (red solid lines), the pressure drops at the same rate everywhere in the reservoir,
and material-balance-like calculations can be performed to interpret the wellbore pressure drop for hydrocarbon-in-place. Right side: identication of ow-regimes for a simulated
(non-stimulated) vertical dry coal well.
Left side: from Clarkson (2011), modied from Fekete RTA short course notes. Right side: modied from Clarkson (2009).
105
4.
3.
1.0E+07
2.
1.
d(m(p)/q)/dlntca
Pseudo -Radial
flow (zero-slope)
1.0E+06
Linear flow
(1/2-slope)
Boundary -Dominated
flow (unit-slope)
1.0E+05
Elliptical flow
1.0E+04
1.0E+00
1.0E+01
1.0E+02
1.0E+03
1.0E+04
tca, days
Fig. 3. Left side: sequence of ow-regimes for a hydraulically-fractured vertical tight-gas well with innite-conductivity fracture. Lines correspond to isopotential lines; arrows are
streamlines. Flow regime 1. (inner black arrows) corresponds to linear ow; ow-regime 2. (elliptical green dashed lines) corresponds to elliptical ow; ow-regime 3. (circular
green dashed lines) corresponds to pseudoradial ow; and ow-regime 4. (circular red line) corresponds to boundary-dominated ow. Right side: identication of ow-regimes for
a simulated hydraulically-fractured vertical tight gas well.
Right side: modied from Clarkson and Beierle (2011).
identied as a 1/2 slope, the radial ow period as a zero slope, and the
boundary-dominated ow period, as a unit slope. A transitional elliptical ow period corresponds to the non-linear portion of the derivative between linear and radial ow.
The sequence of ow-regimes for hydraulically-fractured wells
described above may be visualized by observing gridblock pressure
changes over time in a numerical simulator (Fig. 4). Flow-regimes
13 (transient ow regimes) in Fig. 3 are identied by pressure gradient geometries in Fig. 4.
For multi-fractured horizontal wells (MFHW), the sequence of
ow-regimes may be even more complex (Fig. 5, left side) (Chen and
Raghavan, 1997; Raghavan et al., 1997). The example in Fig. 5 is for a
multi-fractured horizontal well with innite-conductivity planar hydraulic fractures, that are widely spaced, completed in an inniteacting (no outer boundaries) homogeneous tight gas reservoir. The
early ow-regimes are similar to the vertical, hydraulically-fractured
well described above, where ow is restricted to between fractures. Because the latest trend in MFHWs is close hydraulic fracture spacing and/
or multiple perforation clusters per fracturing stage, the transitional
ow-regimes including early elliptical and radial ow may not appear.
Note that it is possible to see 2 linear ow periodsan early linear
ow period where ow is perpendicular to the hydraulic fractures,
and a late or compound linear ow, where ow is perpendicular to
the effective wellbore length (Chen and Raghavan, 1997; Raghavan et
al., 1997; van Kruysdijk and Dullaert, 1989). At some point after early
linear ow, the fractures will interfere, resulting in a ow period that
may appear like boundary-dominated ow, followed by late or compound linear ow.
Derivative techniques can be used to identify ow-regimes (Fig. 5,
right side), but with a great deal of potential variation in the sequence
of ow-regimes seen, as well as their length, depending on both reservoir properties and completion design. As mentioned, some transitional ow-regimes may be absent, depending on hydraulic fracture
spacing and fracture length. Further, early linear ow may be too
short to observe in higher permeability reservoirs and late/compound
linear ow and the second formation radial ow period may be
absent, depending on formation permeability and well-spacing. An
early, sub-linear (appearing as bilinear ow, or skin effects) may
occur as a result of ow-convergence to the horizontal well, cleanup

effects after stimulation, fracture face skin and nite-conductivity
fractures (Nobakht and Mattar, 2012). Finally, the position of
the late linear ow period relative to early linear ow is affected by
fracture spacing (Nobakht et al., 2012a). As with hydraulicallyfractured vertical wells, simulation gridblock pressure gradients are
a useful way to visualize the sequence of ow-regimes over time
(not shown).
The MFHW ow-regime sequence for shale gas wells has recently
been discussed by several authors (i.e. Cheng, 2011; Song and
Ehlig-Economides, 2011). A conceptual diagram showing this sequence is given in Fig. 6, and the corresponding derivative signature
is given in Fig. 7.
The primary difference between the sequence of ow-regimes observed in shales and for conventional reservoirs is the absence of
transitional (elliptical and radial) after the rst linear ow period
(Fig. 6). Fracture storage is usually too short to observe with typical
production data, or masked by cleanup effects. Song et al. (2011)
coined the term pseudo pseudosteady-state for the post-fracture
interferencethey noted that during this timeframe, each fracture
will produce from its own drainage volume, but that the slope on
the derivative is not quite unity because ow is unbounded beyond
the fracture tips. Note that the reservoir volume dened by the hydraulic fracture stimulation is often referred to as a stimulated reservoir volume, or SRV, which many authors believe to be the practical
extent of drainage in shale reservoirs (i.e. Ozkan et al., 2011).
Bello (2009) noted that multi-fractured horizontal wells in shale
can exhibit transient dual porosity characteristics. Moghadam et al.
(2010) similarly documented dual porosity characteristics, and illustrated several sequence of ow-regimes that may occur in such systems, of which two possibilities are shown in Fig. 8. Note that
bilinear ow is possible with this model (scenario b of Fig. 8) due to
simultaneous fracture-linear and matrix-linear ow.
As noted above, a combination of reservoir properties and
wellbore/fracture geometry has a signicant impact on the sequence
of ow-regimes encountered in unconventional reservoirs. For horizontal wells, there are many combinations that may be encountered,
several of which are shown in Fig. 9. These 8 scenarios form the basis
106
a)
Hydraulic Fracture
embedded within an SRV that has also been created as a result of

stimulation; in these cases the conductivity of the primary hydraulic
fracture is greater than the conductivity of induced/natural fractures
within the SRV. Scenario 7 is the basis of the popular trilinear ow analytical model (Brown et al., 2011; Ozkan et al., 2011) discussed
below.
Understanding the ow regime sequence caused by hydraulic
fracture geometry and reservoir properties is critically important
when interpreting rate-transient/decline characteristics of unconventional gas wells. Once the ow-regimes are condently identied,
there are several production analysis methods that may be applied
to obtain hydraulic fracture/reservoir information from them. These
techniques will be discussed next.
3. Production analysis methods
b)
Approximate extent of
elliptical flow pattern
Production analysis can be used to derive the following information about the reservoir and the wellbore or fracture geometry:
1. Estimated ultimate recovery (EUR) and original gas-in-place (OGIP)
from boundary-dominated ow data
2. Fracture or matrix permeability, hydraulic fracture half-length and
fracture conductivity or contacted matrix surface area (for complex fracturing cases), effective wellbore lengthfrom transient
ow data.
c)
Approximate extent of
radial flow pattern
Advanced production data analysis (or rate-transient analysis)

methods use both production rates and owing pressures in the analysis to account for variable operating conditions of the wells. Empirical
methods use only production data. As discussed by Clarkson and
Beierle (2011) there are several production analysis methods that
have been commonly applied for unconventional gas reservoirs
including:
1.
2.
3.
4.
5.
Fig. 4. Simulation case illustrating the transient ow-regimes associated with a vertical
well with an innite conductivity hydraulic fracture. Pressures range from 3300 psi
(red) to 1000 psi (blue). a) Linear ow; b) elliptical ow; c) pseudo-radial ow.
Modied from Clarkson and Beierle (2011).
of several models that have been used to describe production performance in unconventional reservoirs. Scenarios 1 and 2 illustrate an
openhole-completed single horizontal lateral completed in a singleporosity reservoir, and dual-porosity reservoir, respectively. This
completion style continues to be used in low-permeability CBM
plays, such as the Mannville coals of Western Canada (Gentzis and
Bolen, 2008; Gentzis et al., 2009) and the fringe-Fairway Fruitland
coals of the San Juan Basin (Clarkson et al., 2009), but multi-lateral
wells are becoming more common. Scenario 2 has been used to
model multi-fractured horizontal wells where a complex fracture geometry is created, and the SRV exhibits dual porosity characteristics.
Scenarios 3 and 4 are cases where the SRV occupies a region immediately surrounding the horizontal well, but in Scenario 4, the SRV and
background (naturally-fractured) reservoir have differing dual porosity characteristics. Scenarios 5 and 6 correspond to cases where a
low-complexity, primary hydraulic fracture has been created. Scenarios 7 and 8 also contain primary hydraulic fractures, but these are
Straight-line (or ow-regime) analysis methods

Type-curve methods
Analytical and numerical simulation
Empirical methods
Hybrid (analytical and empirical) methods.
In the following, we detail these methods and provide simulated

examples for CBM, TG and SG. The focus will be on straight-line,
type-curve (advanced production analysis methods) and empirical
methods. Analytical simulation will be briey discussed, as will
hybrid methods. Numerical simulation is beyond the scope of the
current discussion.
3.1. Straight-line (or ow-regime) analysis methods
These techniques for production data are analogous to those used
in pressure transient analysis (Lee et al., 2003). Once the owregimes have been identied using derivative analysis or other log
log diagnostic techniques (Figs. 2, 3, 5, and 7) the data corresponding
to specic ow-regimes is analyzed on specialty plots designed to
linearize the dataset for that ow geometry. Depending on the
ow-regime identied, hydraulic-fracture properties or reservoir
properties may be obtained.
To illustrate the application of straight-line analysis for a tight gas
reservoir (CBM and shale gas will be discussed below), we have provided a summary of the expected ow-regimes, diagnostic signature
of the ow-regime on a derivative, the plotting variables for the
ow regime, and the extracted properties for a vertical, hydraulicallyfractured well (Table 1). Note that the ow-regimes in Table 1 are
the same as those identied in Fig. 3, except for bilinear ow
(simultaneous linear ow within the fracture and perpendicular to
the fracture, which may be observed in nite conductivity fractures
discussed below). The specialty plots for each ow-regime (columns 3
and 4, Table 1) are derived from constant rate solutions to the ow
107
4.
1.0E+07
2.
d(m(p)/q)/dlnt
Late Linear
flow
3.
1.
Late Radial
flow
1.0E+06
Fracture
Interference
1.0E+05
B-D flow
Early Linear
flow
1.0E+04
Early Radial
flow
1.0E+03
1.0E-02 1.0E-01 1.0E+00 1.0E+01 1.0E+02 1.0E+03 1.0E+04
t, days
Fig. 5. Left side: sequence of ow-regimes for a multi-fractured horizontal well with planar innite conductivity fractures with wide fracture spacing, completed in a tight gas
reservoir. Lines correspond to isopotential lines; arrows are streamlines. Flow regime 1. (inner black arrows) corresponds to linear ow; ow-regime 2. (elliptical green dashed
lines) corresponds to elliptical ow; ow-regime 3. (elliptical red solid lines) corresponds to fracture interference; and ow-regime 4. (outer black arrows) corresponds to late
(compound) linear. Right side: identication of ow-regimes for a simulated multi-fractured horizontal well completed in a tight gas reservoir.
Right side: modied from Clarkson and Beierle (2011).
equations. Most of these solutions were derived for drawdown

well-tests, as discussed in Lee et al. (2003), except the owing material
balance (boundary-dominated ow) equation which is a modication
of the equation given by Agarwal et al. (1999). The elliptical
ow-regime solution was given in Cheng et al. (2009). In Table 1, t is
real time (hours), ta is pseudotime (hours), m(p) is pseudopressure
(psi2/cp) and mFL and bFL are the slope and intercept, respectively, of
the corresponding specialty plotall other parameters are dened
in the Nomenclature section. As will be discussed below, ratenormalized m(p) function and bilinear, linear and radial superposition
time functions are used to account for variable rates and owing
pressures, and pseudovariables (pressure and time) are used to

account for gas property changes with pressure. These data transformations allow solutions derived for constant rate production of
slightly-compressible uids (the basis for most well test solutions) to
be used.
From the bilinear ow-regime (row 1, Table 1), hydraulic fracture
conductivity (width of fracture, wf, times fracture permeability, kf)
can be estimated from the slope of the bilinear specialty plot (column
3 of Table 1), if permeability is known, along with standard volumetric inputs for the reservoir (thickness, porosity, uid properties).
From the formation linear ow (row 2, Table 1), hydraulic fracture
Boundary
Behavior
Transient Flow to SRV
Pressure Depletion
in SRV
Transient Flow
to Each Fracture
Pseudolinear Flow
Derivative
Slope = 1
Derivative
Slope = 1/2
Linear flow normal to

transverse fractures
Inter-fracture pressure Interference
Fracture Storage
Pseudo
Pseudosteady
State Flow
Derivative
Slope ~
= -1
1
Pseudo pseudosteady
state flow
Compound
Linear Flow
Derivative
Slope = 1/2
Compound
linear flow
Pseudoradial
Flow
Pseudosteady
State
(Interwell
Interference)
Derivative Slope = 1
Derivative
Slope = 0
Pseudoradial
flow
Steady State
(External
Pressure
Support)
P = Constant
Fig. 6. Sequence of ow-regimes that are possible for MFHW completed in shale reservoirs. Note the absence of transitional (elliptical or radial) ow after the rst pseudolinear ow
period.
Modied from Song and Ehlig-Economides (2011).
108
100
Boundary behavior:
Unit slope,
Pseudosteady state
10
RNP or RNP
1
slope,
Compound linear flow
0 slope,
Pseudo radial flow
10-1
Nearly unit slope,

Pseudo pseudosteady state
10-2
10-3
10-4
10-1
slope,
Pseudolinear flow
10
102
103
104
105
106
107
108
te, hours
Fig. 7. Rate-normalized pressure (RNP) and rate-normalized pressure derivative (RNP) signature for MFHW in shale. Note te is material balance time in the terminology of Song and
Ehlig-Economides (2011).
Modied from Song and Ehlig-Economides (2011).
Fracture Boundary-Dominated Flow

(FBD)
Horizontal Well
Fracture Linear Flow (FL)
Boundary-Dominated
Boundary-Dominated
Boundary-Dominated
Matrix Linear + FBD Flow
Boundary-Dominated
b)
Horizontal Well
Boundary-Dominated
Fracture Linear Flow (FL)
Horizontal Well
Boundary-Dominated
Horizontal Well
Boundary-Dominated
Boundary-Dominated
a)
procedures were developed for wells with innite-conductivity and

nite-conductivity fractures. In the nite-conductivity case, A and B are
modied, and bilinear ow regime analysis is also required. The unique
aspect of elliptical ow is that it does not appear as a straight-line on a
semi-log (radial) derivative, and that permeability and xf can be obtained
Boundary-Dominated
half-length (xf) can be estimated from the slope of the linear specialty
plot (column 3 of Table 1), if permeability is known, along with standard volumetric inputs.
The elliptical ow regime (row 3, Table 1) analysis procedure was developed recently by Cheng et al. (2009), and is iterative; separate iterative
Horizontal Well
Horizontal Well
FL + ML => Bilinear (BL) Flow
ML + FBD Flow
2
Fig. 8. 2 possible scenarios (a) and (b) for ow-regime sequences in a transient dual porosity system. For a), > 3 / (yeD) and for b), b 3 /(yeD)2, where and are the
interporosity ow coefcient and storativity ratio, respectively, and yeD is a dimensionless reservoir length.
Modied from Moghadam et al. (2010).
Single Porosity Reservoir

Ex. Horizontal well type-curves of
109
Dual Porosity Reservoir
Horizontal Well
Scenario 1
Scenario 2
Stimulated Reservoir Volume
Scenario 3
Scenario 4
Ex. Nobakht and Clarkson typecurves (Nobakht et al. (2011)
Discrete Hydraulic Fractures
Scenario 5
Scenario 6
Trilinear flow model (Brown et al. 2011)
Scenario 8
Scenario 7
Fig. 9. Possible combinations of reservoir/hydraulic fracture encountered for shale/tight gas reservoirs.
Modied from Clarkson and Pedersen (2010).
Table 1
Summary of ow-regime analysis for hydraulically-fractured gas wells. Note that the unit of time is in hours in this table.
Flow regime
Bilinear
Linear
Loglog diagnostic
Radial derivative: 1/4 slope
Bilinear derivative: zero slope
Elliptical
Radial derivative: 1/2 slope

Linear derivative: zero slope
Radial derivative: variable slope
Pseudoradial
Radial derivative: zero slope
Specialty plot
n
qj qj1
X
1=4
t a;n t a;j1
qn
j1

n
qj qj1
X
1=2
t a;n t a;j1
qn
j1
mpi mpwf
q
vs. (ta)1/4 or t a;BLS
mpi mpwf
q
vs. (ta)1/2 or t a;LS
mpi mpwf
q
vs. ln(A + B)
mpi mpwf
q
vs. log(ta) or t a;RS

n
qj qj1
X
j1
Boundary-dominated
Radial derivative: unit slope
q
mpi mpwf
i mpR
vs. Gi mmppm
pwf
i
qn

log t a;n t a;j1
Extracted properties

1=2 1=4
443:2T
wf kf
k
1=4
mBL h gi cti
p
kxf
40:93T
1=2
mL h gi cti
k 1422T
mE h
xf exp bE
1637T

k mR h
s 1:1513 mbRR log

Gi from x-intercept

k
2
gi cti r w
3:23
110
from the analysis. Prior to the development of the straight-line technique

introduced by Cheng et al., permeability could only be uniquely determined from radial ow, which may take a long time to develop for
tight reservoirs.
From the radial ow-regime (row 4, Table 1), permeability can be
estimated from the slope of the radial specialty plot (column 3 of
Table 1), if standard volumetric inputs for the reservoir are known,
and skin from the y-intercept.
From the boundary-dominated ow period, OGIP can be obtained
from the x-intercept of the specialty plot. Derivation of OGIP from
owing (or dynamic) material balance is analogous to obtaining
OGIP from conventional (static) material balance, except owing
pressures are used instead of static (shut-in) pressures. The owing
material balance (FMB) procedure is also iterative, as discussed in
Mattar and Anderson (2003); kh may also be extracted from the
y-intercept if skin is known.
Because most of the analytical solutions derived for uid ow to a
single well, which are used in straight-line analysis, assume: 1) constant
rate or constant pressure inner (well) boundary-conditions, 2) single
phase ow of a slightly-compressible or non-compressible uid,
3) Darcy (laminar) ow, 4) no chemical reactions between uid and
solid (i.e. adsorption); and 5) a porous medium with negligible porosity
or permeability changes, efforts to linearize the data for each ow
regime must focus on addressing the assumptions of the analytical
solutions. To make the solutions applicable to gas, pseudopressure
and pseudotime are used to account for gas property variations with
pressure:
pi

p
dp
mpi m pwf
gz
p

t dt
t a g ct
:
i
g ct
0
wf
Note that the traditional denition of pseudotime (ta) for production analysis involves the evaluation of gas properties (viscosity and
compressibility) at pore-volume average (reservoir) pressure; as
demonstrated by Nobakht and Clarkson (2012a,b), this may lead to
inaccuracy in reservoir/hydraulic fracture property determination
for ultra-low permeability reservoirs, such as shales; for these cases,
Nobakht and Clarkson (2012a,b) recommend use of a corrected
pseudotime, where the gas properties are evaluated at the average
pressure in the region of inuence during transient ow. Corrections
for non-Darcy ow, adsorption and non-static permeability are
discussed in a later section.
To account for variable production rates and owing pressures at the
well due to operational changes, superposition time should be used for
the x-axis variable and rate-normalized pseudopressure for the y-axis
variable for the specialty plots in Table 1. Use of superposition time,
the form of which is ow-regime specic, allows variable-rate data to
be analyzed using constant-rate solutions to the ow-equations. The
dynamic owing material balance method (Agarwal et al., 1999;
Mattar and Anderson, 2005) was derived using a superposition time
function that is applicable to boundary-dominated ow, which is commonly referred to as material balance time (material balance
pseudotime for gas):

t ca
g ct
qg

t
i
qg dt
:
g ct
Note that for slightly-compressible uids, assuming pore volume

compressibility is constant, the viscosity-total compressibility product is approximately constant. Further, the integral of q over time is
simply the cumulative production. The combination of these allows
a simpler version of material balance time to be used: tc = Q / q.

Some authors have used this simpler form of material balance time
as a rst approximation for gas.
An example application of ow-regime (straight-line) analysis to
the simulated hydraulically-fractured vertical well (tight gas, matrix
permeability = 0.01 mD) case given in Fig. 3 is provided in Fig. 10.
The four ow-regimes identied in Fig. 3 are analyzed on their corresponding specialty plots using the equations given in Table 1; because
the hydraulic fracture was assumed to be innite conductivity, an
early bilinear ow period is not seen or analyzed. Because pseudoradial (Fig. 10c) and boundary-dominated ow periods (Fig. 10d)
are observed in this example, they would rst be analyzed for permeability and OGIP, respectively. The permeability estimate obtained
from radial ow analysis would then be used to calculate fracture
half-length
linear ow analysis (Fig. 10a), which yields an estipfrom
mate of xf k. The elliptical ow period (Fig. 10b) can also be analyzed

to yield an estimate of permeability and fracture half-length to conrm
the estimates from radial and linear ow, respectively. In the absence of
radial or elliptical ow, as is often the case in shale reservoirs, permeability would have to be estimated independently from core data,
pre-fracture stimulation well-testing, or from fracture interference
time associated with multi-fractured horizontal wells, as discussed
below. As noted by Clarkson and Beierle (2011), the parameters derived
from straight-line analysis of this example are in reasonable agreement
with simulation input values, although fracture half-length from linear
ow analysis is somewhat over-estimated due to the use of conventional pseudotime (Eq. (2)), instead of the corrected pseudotime, suggested
by Nobakht and Clarkson (2012a,b). As will be seen, the corrected
pseudotime results in more accurate estimates of xf, particularly for
ultra-low permeability reservoirs.
Multi-fractured horizontal wells can be examined in an analogous
fashion to what is presented above for a vertical well with a single hydraulic fracture, but we note that it is possible to have multiple linear
ow periods (ex. early formation linear ow and/or late compound
linear ow). If the fractures are spaced far enough apart, transitional
(early elliptical and early radial) ow-regimes may be observed,
and if the wells are spaced far enough apart, late elliptical and late radial ow may appear. The equations provided in Table 1 can still be
applied, but if multiple linear ow periods occur, for example, the
early linear ow period can be used to extract a total fracture
half-length (nxf, where n is the number of fractures) for the well (if
permeability is known), and the late linear ow period may be used
to estimate effective well-length (if permeability is known), or permeability, if effective well-length is assumed. Similarly, for radial
ow, early radial ow will yield an estimate of nkh, and late radial
ow will yield an estimate of the true formation kh.
An example is provided below from Clarkson and Beierle (2011);
the example is for a simulated MFHW completed in a tight gas reservoir
(Fig. 11). The derivative signature for this simulated example is shown
in Fig. 5two radial and two linear ow periods are obtained. Analysis
of the late-radial ow period yields a permeability value within reasonable agreement (within 10%) with that input into the numerical simulator (0.01 mD); the total fracture half-length is also within 10%. As with
the previous simulated example, some error in the derived parameters
is due to the use of conventional pseudotimethe estimates with
corrected pseudotime are within 5%.
The simulated examples provided above are for a relatively
straight-forward tight gas (single porosity, permeability = 0.01 mD)
case, with no issues related to desorption, multi-phase ow, nonDarcy ow, or non-static permeability. We now discuss modications
of the straight-line analysis method for CBM and shale wells with
complex reservoir behavior.
3.1.1. Additional considerations for CBM and shale reservoirs
The most common way to incorporate complex reservoir behavior
into straight-line analysis for CBM and shale reservoirs is through
a)
b)
Linear Flow Plot
Elliptical Flow Plot
5.0E+06
Use to calculate xf,

when k is known [from
plot b) or c)].
1.5E+06
4.5E+06
[m(pi)-m(pwf)]/qg
[m(pi)-m(pwf)]/qg
2.0E+06
111
1.0E+06
5.0E+05
Use to calculate k and xf

and compare with
results of plot c) and a),
respectively.
4.0E+06
3.5E+06
3.0E+06
2.5E+06
2.0E+06
1.5E+06
1.0E+06
5.0E+05
0.0E+00
5.0
0.0E+00
0
10
20
30
40
50
6.0
7.0
Superposition Time
c)
d)
Normalized Rate, scf/D/psi2/cp
Radial Flow Plot

5.0E+06
[m(pi)-m(pwf)]/qg
4.5E+06
Use to calculate k and

skin. Skin can be used
to estimate infiniteconductivity-equivalent
xf. k can be used with
plot a) to obtain xf.
4.0E+06
3.5E+06
3.0E+06
2.5E+06
2.0E+06
8.0
9.0
ln(A+B)
1.5E+06
1.0E+06
5.0E+05
Flowing Material Balance

1.0E-03
9.0E-04
Use to calculate OGIP and

drainage area (A).
Pressures obtained from
FMB can be used in
conventional pseudotime
calculations.
8.0E-04
7.0E-04
6.0E-04
5.0E-04
4.0E-04
3.0E-04
2.0E-04
1.0E-04
0.0E+00
0.0E+00
0
1000
Superposition Time
2000
3000
4000
5000
6000
Normalized Cumulative Production, MMscf
Fig. 10. Use of specialty plots (linear, a, elliptical, b, radial, c, and boundary-dominated, d) for analyzing ow-regimes associated with simulated hydraulically-fractured vertical well
in a tight gas reservoir (see Fig. 4). Ideally, if either or both elliptical and radial ows are observed, then permeability is used in linear ow analysis to derive an independent estimate of xf. Analysis tips are also provided (orange boxes). Note that transitional ow-regimes (elliptical and radial) are often not observed in unconventional gas reservoir such as
shale.
alteration of the pseudovariables, pseudotime and pseudopressure, in

Table 1. For example, for a dry gas reservoir, alterations of pseudotime
for a) adsorption, b) non-static permeability, and c) non-Darcy ow,
are given with Eqs. (4)(6), respectively:

t dt

t a g ct
i c
0 g t

t a;NSP
t a;NDF
g ct

t
ki

g ct
k pdt
g ct
0
ka pdt
g ct
0
where:
ct
k p
ki
ka p
k
total compressibility including desorption compressibility

absolute permeability at the pore volume average pressure
of the reservoir
absolute permeability at initial pressure
apparent permeability at the pore volume average pressure
of the reservoir
liquid-equivalent permeability.
To correct pseudotime for adsorption, the total compressibility term

is altered (Bumb and McKee, 1988; Clarkson et al., 2007; Gerami et al.,
2007). To account for non-static permeability, a relationship between

permeability and pore pressure is rst established, and is used to calculate permeability at the pore volume average pressure of the reservoir.
In coal reservoirs, permeability may both decrease due to increase in
stress during depletion, and increase due to matrix-shrinkage effects
during desorptionanalytical models, for which a comprehensive summary was recently provided by Pan and Connell (2012), may be used to
model permeability changes in these cases. In shale reservoirs, it is generally assumed that permeability decreases with an increase in stress
(decrease in pore pressure)Thompson et al. (2010) modeled permeability changes with pore pressure. Clarkson et al. (2012b), Nobakht
et al. (2012b) and Ozkan et al. (2010) recently demonstrated that
non-Darcy ow effects, such as slip ow and diffusion, can be incorporated into rate-transient analysis by including an apparent permeability
change with pressure into the analysis.
Pseudopressure may be corrected for a) non-static permeability
and b) non-Darcy ow, using Eqs. (7) and (8), respectively:

mpi NSP m pwf
NSP
2 i kp
pdp
ki p g z
wf

mpi NDF m pwf
NDF
2 i ka p
pdp:
k p g z
wf
112
As noted by Nobakht and Clarkson (2012a,b), the estimation of

permeability and uid properties in pseudotime at pore-volume average pressure for the reservoir can lead to errors in hydraulic fracture/
reservoir properties derived from rate-transient analysis for ultra-low
permeability reservoirs. They proposed that average pressure in
the region of inuence rst be evaluated, and then the pressuredependent properties (uid properties, adsorption, and permeability)
evaluated at that pressureevaluation procedures for constant pressure (Nobakht and Clarkson, 2012b) and constant rate (Nobakht
and Clarkson, 2012a) well operating constraints were provided.
Corrections for multi-phase ow are more complex. In 2-phase
CBM and shale reservoirs, effective permeability to gas and water
changes as a function of uid saturation, which requires relative
permeability for each phase to be known. Further, alteration of
pseudopressure requires that a relationship between relative permeability and pressure be knownstrictly-speaking, relative permeability is a function of saturation, not pressure. Mavor and Saulsberry
(1996) reviewed the approaches that have been used to perform
two-phase CBM pressure-transient analysis, including: 1) an adaptation of Perrine's (1956) approach which utilizes a multi-phase total
mobility function; 2) Kamal and Six's approach (1993) which utilizes
a multi-phase ow potential with an assumed relationship between
saturation and pressure and 3) Mavor's approach, which also utilizes
a multi-phase ow potential, with a different (from Kamal and Six)
assumption for the relationship between saturation and pressure.
The multi-phase ow potential, as dened by Mavor and Saulsberry
(1996) is:
a)
b)
Early Linear Flow Plot

1.0E+05
Use to calculate total xf,

when k is known [from
plot c)]
[m(pi)-m(pwf)]/qg
8.0E+04
Mp
pb
6.0E+04
4.0E+04
2.0E+04
0.0E+00
0
10
15
20
Superposition Time
c)
Late Radial Flow Plot

4.0E+06
[m(pi)-m(pwf)]/qg
!
krg
k
rw dp
g Bg w Bw
Use to calculate k
and skin. k can be
used with plot a) to
obtain total xf.
3.0E+06
mpMP 2
pb
2.0E+06
1.0E+06
0.0E+00
3
where pb is a base pressure (psia). Mavor and Saulsberry assumed a linear relationship between owing pressure and saturation to allow M(p)
to be calculated. To extend this approach to production analysis, the
pseudotime function would also have to be adjusted, in which case a relationship between pore-volume pressure and saturation would have to
be known. Mohaghegh and Ertekin (1991), who developed type-curves
for two-phase CBM reservoirs, evaluated saturation-dependent variables at the initial gas saturation, which considerably simplies the
analysis. Clarkson et al. (2008) utilized a form of Eq. (9) for production
analysis (owing material balance), but assumed that gas mobility
dominated the two-phase pseudopressure calculationthe equation
used in that work is given below:
Superposition Time
Fig. 11. (a) Simulation grid for numerically simulated MFHW example and analysis of
early linear (b) and late radial (b) ow-regimes using specialty plots.
Procedurally, Eqs. (1)(8) can be used in the ow-regime analysis

equations of Table 1, and analysis can be performed analogously to
the tight gas examples provided above.
krg
pdp
gz
10
where MP refers to multi-phase in the modied pseudopressure.

Clarkson et al. (2008) did not utilize a modied pseudotime in their
work.
To linearize the partial-differential equation governing ow of gas
through a gas-condensate reservoir, Sureshjani and Gerami (2011)
utilized a two-phase pseudopressure and pseudotime. Those authors
also developed a modied material balance pseudotime and owing
material balance plot to analyze boundary-dominated ow data.
Use of modied pseudotime and pseudopressure to perform ratetransient analysis on wells exhibiting multi-phase ow characteristics
represents the most advanced approach; similar modications to the
pseudovariables can be performed for multi-phase oil and CBM wells.
While the approach developed by Sureshjani and Gerami (2011) for
gas-condensate reservoirs is the most rigorous for rate-transient analysis of multi-phase cases, it is not terribly practical as it is applied to
2-phase CBM or shale wells. Use of modied pseudovariables requires
1) that relative permeability be known and 2) relative permeability be
related to pressure, to facilitate pseudopressure calculations. In recent
113
work by Clarkson et al. (2012c), an empirical approach was suggested

for 2-phase CBM reservoirs. They rst represented the y-axis variable
(of the specialty plot) for rate-transient analysis of gas data for CBM
wells producing both gas and water as:
h
i
krg Sg mpi m pwf
qg
11
xf
Note that the inverse of this variable is used for owing material
balance analysis. Moving krg outside of the pseudopressure calculation (see Eq. (11)) removes the need to relate relative permeability
to pressure, but additionally assumes that saturation gradients are
small because a pore-volume average saturation is used in krg calculations. Therefore, Eq. (11) is strictly applicable to high permeability
cases, with the approximation expected to worsen as permeability
decreases. It is further assumed that the production data is dominated
by gas. The need to dene relative permeability for gas as a function
of saturation remains.
The x-axis variable (of the specialty plot) includes a modied
pseudotime dened as follows:
t a;MF

g ct t krg Sg dt
i
g ct
krg
0
12
where MF refers to multi-phase. As above, the starred variables are

corrected for desorption. For variable rate data, Eq. (12) would be
included in the superposition time calculation (see Table 1).
In order to improve consistency in the analysis of 2-phase CBM
wells, which requires knowledge of relative permeability, Clarkson
et al. (2012c) suggested that pressure- and saturation-dependent variables in straight-line (e.g. Eqs. (11) and (12)) and type-curve analysis
(see below) be derived from outputs of analytical or numerical simulation. The rate-transient and simulation results are therefore linked, providing improved consistency, but not necessarily uniqueness.
In the following section, we demonstrate the application of the
straight-line modications for simulated CBM and shale gas examples.
3.1.2. Simulated examples of CBM and shale straight-line
analysisCases 1 and 2dry shale
The rst example is designed to illustrate the application of corrections to straight-line analysis so that CBM and shale wells may
be analyzed. In particular, the corrections for 1) adsorption and 2)
non-static permeability are illustrated. The simulated example is for
a vertical well producing through a single planar hydraulic fracture
of innite conductivity from a dry (single-phase) shale gas reservoir
with a permeability of 0.1 mD (Fig. 12). Other model inputs are
given in Table 2. In this example, the permeability is assumed to be
1) static (Case 1), or 2) exponentially decreasing (Case 2). For Case
2, the permeability change was modeled using the approach of
Ozkan et al. (2010) and Raghavan and Chin (2004).
ppi
k ki e
13
Where is the permeability modulus, assumed = 4 10 4 psi 1

for this example.
Simulated production rates for Case 1 (static permeability) and Case
2 (changing permeability) are given in Fig. 13. Linear ow is followed
directly by boundary-dominated ow in both cases. We see that the primary differences between the two cases are 1) the initial rates and
2) the character of the boundary-dominated (B-D) ow data.
To illustrate the impact of adsorption on the analysis, Case 1 is analyzed by 1) ignoring adsorption and 2) applying the corrections
discussed in Section 3.1.1. The ow-regimes are rst identied with
derivatives (Fig. 14a, b) to determine which straight-line plots to
ye
xe
Fig. 12. Vertical hydraulically-fractured well centered in a rectangular reservoir.
apply to the data. In Case 1 and Case 2, only linear and boundarydominated ow appear. Another tool for ow-regime identication
is the transient productivity index (Araya and Ozkan, 2002;
Medeiros et al., 2008, 2010). Historical workers have used an average
reservoir pressure in the transient PI calculation, but recently
Williams-Kovacs et al. (2012) derived a transient PI that utilizes the
average pressure in the region of inuence for transient linear ow,
which is useful for ultra-low permeability reservoirsthis new PI is
given in Fig. 14c for Case 1.
The primary impact of ignoring adsorption is the OGIP estimate
from owing material balance (Fig. 15a). The owing material balance analysis requires a material balance equation (MBE) to allow
pore-volume average pressure to be calculatedas discussed by
Williams-Kovacs et al. (2012), there are several MBEs available for
analysis of shale gas wells. For the No Corrections Made FMB plot
in Fig. 15a, a conventional gas (p/z-cum) MBE was used that does
not include corrections for adsorption; the Corrections Made FMB
plot uses King's (1993) MBE equation for dry CBM/shale reservoirs,
and includes corrections for adsorption in a modied z-factor calculation (z*). Further sensitivities to MBE selection were discussed by
Williams-Kovacs et al. (2012). We see that when the corrections are
made for adsorption (using King MBE equation) in the FMB plot, the
correct (simulator input) OGIP and drainage area are obtained
(~ 6.7 bscf and 57 ac, respectively). When the corrections for
Table 2
Input parameters for simulated 1-phase (shale gas) vertical well case.
Input parameter
Value
Thickness (ft)
Bulk density (g/cm3)
Porosity (%)
Gas gravity
Initial absolute permeability (mD)
Initial reservoir pressure (psia)
Initial water saturation (%)
Reservoir temperature (F)
Langmuir Volume (scf/ton, in-situ)
Langmuir Pressure (psia)
Drainage Area (ac)
Fracture half-length (ft)
Wellbore Diameter (in.)
Skin factor
Flowing Bottomhole Pressure (psia)
100
2.47
10
0.69
0.1
3500
0
200
89
535.6
57
250
8.4
0
250
114
a)
Gas Production Rates

100000
1.0E+07
B-D flow
d(m(p)/q)/dlnt
Gas Rate, Mscf/D
Linear flow
10000
1000
1.0E+06
Linear flow
Boundarydominated
flow
1.0E+05
100
1.0E+04
1.0E+00
1.0E+01
1.0E+02
1.0E+03
1.0E+04
1.0E+05
Time, days
10
1
10
100
1000
10000
100000
b)
Time, days
Linear Derivative Plot

1.0E+06
Changing k Case (Case 2)
d(m(p)/q)/dt^0.5
Static k Case (Case 1)
Fig. 13. Gas production rates for Case 1 and Case 2.
adsorption are not made (conventional MBE), the FMB underpredicts

the OGIP (~ 5.9 bscf) and overpredicts the drainage area (~ 64 ac).
For linear ow analysis, we use the linear ow plot (see Table 1),
with and without corrections for adsorption (Fig. 15b). For the
Corrections Made linear ow analysis, corrections for adsorption
are made by including adsorption in the pseudotime calculations
(Eq. (4)), whereas for the No Corrections Made linear ow analysis,
Eq. (2) was used. Further, the average pressure used in Eqs. (2) and
(4) was obtained using the procedures outlined by Nobakht and
Clarkson (2012b) and Nobakht et al. (2012b), respectivelyi.e. the
average pressure in the region of inuence for constant owing pressure production. From Fig. 15b, we see that the corrections for
adsorption do not make a signicant difference in the analysis; the fracture half-length derived from the slope of the plot is 254 ft using the
corrected plot and 260 ft using the uncorrected plot, both of which
are in reasonable agreement with the simulator input value of 250 ft.
We note that correction will depend on the adsorption isotherm characteristics as well as drawdownNobakht and Clarkson (2012c) and
Nobakht et al. (2012b) analyze additional cases. We note that the impact of adsorption is to act as a negative skin (Clarkson et al., 2007),
which explains why the uncorrected case yields a longer half-length.
An alternative method for analyzing the linear ow data in this
constant owing pressure case is to use the square-root of time plot
(Wattenbarger et al., 1998) with real time, but using a drawdown correction that accounts for uid property and desorption with pressure:

p
1
315:4T
xf k f cp q
g ct mCP
14
where mCP is the slope of the square-root of time plot for constant
mpi mpwf
vs.
owing-pressure conditions, obtained by plotting
q
time (days) and:
f cp
v

u
u c
u g t i
t
g ct
15
is an analytical drawdown correction derived by Nobakht and

where fcp
Clarkson (2012c) to correct for uid property variation with pressure
and for desorption. The drawdown correction for the constant owing
pressure conventional gas case was derived by Nobakht and Clarkson
(2012b) (Eq. (16)); a drawdown correction for non-Darcy ow was
also derived in Nobakht et al. (2012b). For the complex fracture
1.0E+05
Linear flow
Boundarydominated
flow
1.0E+04
1.0E+00
1.0E+01
1.0E+02
1.0E+03
1.0E+04
1.0E+05
Time, days
c)
Transient Productivity Index Plot

1.0E-01
Linear flow
PI
1.0E-02
1.0E-03
Boundarydominated
flow
1.0E-04
1.0E+00
1.0E+01
1.0E+02
1.0E+03
1.0E+04
1.0E+05
Time, days
Fig. 14. Identication of ow-regimes for Case 1. On semi-log (radial) derivative, linear
ow appears as 1/2 slope, on the linear derivative (b), linear ow appears as a zero
slope. On the modied transient-PI plot (c), linear ow appears as a 1/2 slope,
while boundary-dominated ow is a zero slope. Although real time was used in this example, for noisy data, use of material balance pseudotime or linear superposition
pseudotime would yield a smoother derivative.
geometry case xf should be replaced by contacted matrix surface area

(Eq. (17)), as will be discussed for one of the eld cases below.
f cp
v

u
u g ct
t
i
g ct

p
1
1262T
Acm k f cp q
g ct mCP
16
17
a)
b)

1.0E+06
[m(pi)-m(pwf)]/qg
5.0E-03
qg/[m(pi)-m(pwf)]
115
4.0E-03
3.0E-03
2.0E-03
8.0E+05
6.0E+05
4.0E+05
2.0E+05
1.0E-03
0.0E+00
0.0E+00
1000
2000
3000
4000
5000
6000
7000
8000
50
Corrections Made
100
150
Superposition Time
No Corrections Made
No Corrections Made
Corrections Made
Fig. 15. Flowing material balance (a) and linear ow analysis (b) of Case 1, with and without corrections for desorption. The vertical lines on the linear ow plot indicate the portion
of the plot through which a straight line was tted.
Note that Ibrahim and Wattenbarger (2006) provided an empirical calculation for fcp, but this was found to be less accurate than the
analytical method provided by Nobakht and Clarkson (2012b), and
no corrections for adsorption were provided.
The square-root of time plot for Case 1 is given in Fig. 16note
that the
pplot is not affected by the corrections as real-time is used.
The xf k calculation is however affected by the choice of drawdown
correction (Eq. (15) for adsorption, Eq. (16) for no adsorption)use
of Eq. (15) combined with Eq. (14) yields a fracture half-length of
255 ft (given permeability) using the corrected plot and 263 ft
using the uncorrected plot, both of which are in reasonable agreement with the simulator input value of 250 ft, and with the linear
ow plot analysis above.
Use of the square-root time plot also allows an independent (to
FMB) estimate of OGIP. The OGIP calculation, correcting for adsorption, is:
p
t ehs
200:8Sgi

:
OGIP f cp
m
CP
ct Bg
18
and fcp are dened in Eqs. (15) and (16) above. Eqs. (18) and
where fcp
(19) (combined with Eqs. (15) and (16), respectively), yield OGIP estimates of ~ 6.8 bscf and ~ 5.6 bscf; Eq. (18) (which corrects for adsorption) is in very good agreement with the corrected FMB
analysis, but Eq. (19) yields a substantially smaller OGIP as expected.
Finally, with respect to Case 1, we note that the square-root of
time plot, in combination with the distance of investigation calculation provided below, can be used to estimate permeability of the reservoir:
v
u kt
u
y 0:159t ehs
g ct
20
which, corrected for adsorbed gas is:

v
u kt
u
y 0:159t ehs :
g ct
21
For conventional gas reservoirs with no adsorbed gas, the following equation would be used (Ibrahim and Wattenbarger, 2006):
200:8Sgi

OGIP f cp
ct Bg
p
t ehs
mCP
19
Square-Root Time Plot

2.0E+06
tehs
[m(pi)-m(pwf)]/qg
1.8E+06
1.6E+06
1.4E+06
1.2E+06
1.0E+06
8.0E+05
6.0E+05
4.0E+05
2.0E+05
0.0E+00
10
15
20
25
30
35
40
45
50
Sqrt Time (Days)^0.5

Fig. 16. Square-root of time plot for Case 1. A straight-line is t to the linear portion of this
plotthe green dashed vertical line indicates the end of the linear ow period (tehs).
For Case 1, Eq. (21), when solved for permeability, and knowing
y = ye / 2 (Fig. 12) yields a permeability estimate of 0.10 mD while
Eq. (20) yields a permeability of 0.096 mDthe non-corrected (for
adsorption) equation yields a smaller value of permeability because
ct > ct. As will be illustrated with a eld example later, this approach
for permeability estimation can be used for multi-fractured horizontal wells where the fracture spacing can be ascertained and fracture
interference is evident from the square-root of time plot (Ambrose
et al., 2011).
Turning our attention to Case 2, we again apply the FMB, linear
ow and square-root time analysis with and without correction for
permeability changes, as illustrated in Fig. 17. Unlike the previous
case, the slope of the square-root of time plot changes due to inclusion of permeability change in the pseudopressure calculation. We
see that the impact of not including corrections for permeability in
the FMB plot is underestimation of OGIP (~ 6 bscf), whereas the corrections yield OGIP = 6.7 bscf, which is close to the simulator input.
Both the corrected linear ow plot (with corrected linear superposition pseudotime and corrected pseudotime) and the corrected
square-root of time plot (with corrected pseudopressure and application of fcp with pressure-dependent properties evaluated at average
pressure in region of inuence) yield fracture half-lengths of 271 ft
and 270 ft, respectively, which are larger, but in reasonable agreement with actual value of 250 ft. The uncorrected cases, however,
yield much smaller half-lengths of 182 ft and 183 ft, respectively.
116
a)
qg/[m(pi)-m(pwf)]
5.0E-03
4.0E-03
3.0E-03
2.0E-03
1.0E-03
0.0E+00
0
1000
2000
3000
4000
5000
6000
7000
8000

Corrections Made
b)
No Corrections Made
[m(pi)-m(pwf)]/qg
1.0E+06
8.0E+05
6.0E+05
4.0E+05
2.0E+05
0.0E+00
0
50
100
105
Superposition Time
No Corrections Made
c)
Corrections Made
3.2. Type-curve methods
2.0E+06
[m(pi)-m(pwf)]/qg
1.8E+06
1.6E+06
1.4E+06
1.2E+06
1.0E+06
8.0E+05
6.0E+05
4.0E+05
2.0E+05
0.0E+00
0
10
15
20
25
30
35
40
45
50
Sqrt Time (Days)^0.5

Corrections Made
are obtained from output generated from a model used to historymatch the gas production and water production data. In this case
(Fig. 18), an analytical simulator was used to match the gas production and water production rates of another (commercial) analytical
simulator (Fig. 18a), and output from the match used to create the
FMB plot (Fig. 18b). We see that the 2-phase FMB plot is a straightline pointing towards an OGIP of ~ 540 MMscf, which is consistent
with model input.
Case 4 was also shown by Clarkson et al. (2012c), and illustrates the
impact of relative permeability on transient radial ow analysis
(Fig. 19). The gas production and water production curves were generated using a numerical simulator, and were matched using an analytical
simulator (Fig. 19a). We note that the history-match is not quite as good
as in Fig. 18 because an analytical model was used to match
numerically-generated data (as discussed in Clarkson et al., 2012c).
The radial ow plot, which is corrected for relative permeability
changes, using outputs from the analytical simulator, is shown in
Fig. 19b. The derived absolute permeability (~112 mD) is somewhat
higher than the numerical model input (100 mD), but in reasonable
agreementthe derived skin is slightly negative (1.2), despite the
model input being zero-skin. The errors are related to the slight
mismatch between the analytical model and the numerical model, and
the fact that saturation gradients are ignoredstill, a reasonable rst approximation to permeability, using a method that does not require the
relationship between relative permeability and pressure to be known.
The impact of not correcting for relative permeability changes in
Cases 3 and 4 is shown in Fig. 20. If the relative permeability corrections are not included in FMB analysis, the FMB plot is non-linear
until late in time (cumulative production), where relative permeability effects are not as strong (Fig. 20a). If the radial ow plot is not
corrected for relative permeability to gas (Fig. 20b), a straight-line
t to the data will yield permeability values that are closer to effective
permeability, not absolute permeability.
No Corrections Made
Fig. 17. Flowing material balance (a), linear ow (b) and square-root plot (c) for Case
2, with and without corrections for permeability change. The vertical lines on the linear
ow plot indicate the portion of the plot through which a straight line was tted.
3.1.3. Simulated examples of CBM and shale straight-line

analysisCases 3 and 4wet coal
In Clarkson et al. (2012c), 2-phase CBM cases were analyzed using
the straight-line methodology discussed above. Case 3 corresponds to
an analytically-simulated example, where both gas production and
water production were simulated and analyzed using a 2-phase version of the owing material balance equation (Clarkson et al., 2008).
FMB alone was used for analysis because the simulated data are in
boundary-dominated ow. As mentioned previously, the pressureand saturation-dependent variables used in the straight-line analysis
Type-curve methods involve matching of production data to dimensionless solutions to ow equations, which correspond to different well/
fracture geometries, reservoir types and boundary conditions. Unlike
straight-line methods, type-curves are usually designed to capture multiple ow-regimes associated with a particular well geometry/fracture
geometry/reservoir type, and are cast in dimensionless, not dimensional form. Note that the use of the term type-curve in this work refers
strictly to the use of dimensionless type-curves, analogous to typecurve matching performed for pressure-transient analysis (ex. see textbook by Lee et al., 2003). In petroleum literature, the term type-curve
has also been used to denote ratetime or ratecumulative production
plots for wells representing a certain eld or portion of a eldwe use
the PTA-analog in this work.
The development of type-curve solutions for production analysis
has been a popular activity among researchers and practitioners in
the past several decades, and this review cannot hope to capture all
of that activity. The focus is instead on type-curves that the author
has found particularly useful for the analysis of unconventional gas
wells. Researchers and their afliated students, co-workers and colleagues, that have been particularly fruitful in the development of
production type-curve techniques that have proven useful for unconventional gas analysis include: Fetkovich (i.e. Fetkovich, 1980),
Blasingame (i.e. Amini et al., 2007; Palacio and Blasingame, 1993;
Pratikno et al., 2003), Wattenbarger (i.e. Abdulal et al., 2011;
Wattenbarger et al., 1998), Agarwal (i.e. Agarwal et al., 1999), and
Ozkan (i.e. Araya and Ozkan, 2002). Some of these researchers and afliates have developed the analytical solutions to the ow equations,
in addition to providing the type-curves and type-curve matching
procedures. In this work, we show a few examples of type-curve
b)
2-Phase History-Match
800
400
600
300
400
200
200
100
0
200
400
600
800
1000
8.0E-02
6.0E-02
4.0E-02
2.0E-02
0.0E+00
2-Phase Flowing Material Balance

1.0E-01
Water Rate, STB/D
500
Gas Rate, Mscf/D
1000
qg/[m(pi)-m(pwf)]
a)
1200
200
Time, days
Modelqg
Actualqg
117
400
600
Normalized Cumulative Production , MMscf
Actualqw
Modelqw
Fig. 18. Analytical model history-match (a) and owing material balance analysis (b) of simulated 2-phase (gas + water) CBM case.
Modied from Clarkson et al. (2012c).
usage for unconventional gas reservoirs, most of them generated

from this list of researchers.
As with the straight-line methods described above, the analytical
solutions to ow equations to support type-curve generation are usually derived using liquid solutions, static reservoir properties, and simple boundary-conditions (i.e. either constant ow rate or pressure).
Pseudovariables (dened above) are therefore often employed to account for uid and reservoir property variation with pressure, and
normalized production and superposition time (see Table 1) to account for variable ow-rates and pressures. Solution techniques for
the ow equations will not be described in this workthe reader is referred to comprehensive discussions in texts on pressure-transient
analysis (ex. Lee et al., 2003) and, for production type-curves, the individual papers by researchers cited above.
A critical step in dimensionless type-curve generation is the denition of dimensionless variables. Well-dened dimensionless variables
serve the purpose of condensing many solutions to the ow equations
into a nite set and allow for unique matching of the data. The denition
of the dimensionless variables themselves determines which reservoir,
and well/fracture properties can be obtained from the type-curve
match. It should be noted that dimensionless ratetime, cumulative
production-time, and ratecumulative production type-curves have
been generated in the literature. Auxiliary curves involving derivatives
and integrals have also been developed. The focus here is on rate
time type-curves, but it is noted that the auxiliary curves serve to improve the uniqueness of type-curve matching.
a)
100
10
0
1000
2000
3000
10
4000
0:00634 kg
t
t DA
g ct A
24
where:
Lc
kg
A
Actualqg
Actualqw
is some characteristic dimension of the system used in tD

denitionfor non- or slightly-stimulated wells exhibiting
radial ow, Lc = rwa (apparent wellbore radius) and for
hydraulically-fractured wells Lc = xf (fracture half-length).
is the effective permeability to gas, kg = krg k, where krg is
assumed constant for the single-phase gas case.
is area used in tDA denition.
Note that the unit of time in Eqs. (23) and (24) is days.
Radial Flow Plot

2.0E+03
1.5E+03
1.0E+03
5.0E+02
0.0E+00
0
Superposition Time
Time, days
Modelqg
22
23
krg*[m(pi)-m(pwf)]/qg
100
T
h
i qg
0:000703 kg h mpi m pwf
0:00634 kg
t
tD
g ct L2c
1000
Water Rate, STB/D
Gas Rate, Mscf/D
qD
b)
1000
A common well-test style denition for dimensionless-rate and

time is given below, assuming a gas-dominated (single-phase) reservoir (Clarkson et al., 2012c):
Modelqw
Fig. 19. Analytical model history-match (a) and radial ow analysis (b) of numerically-simulated 2-phase (gas + water) CBM case.
qg/[m(pi)-m(pwf)]
1.0E-01
8.0E-02
6.0E-02
4.0E-02
2.0E-02
0.0E+00
0
200
400
600

Corrections Made
No Corrections Made
krg*[m(pi)-m(pwf)]/qg or [m(pi)-m(pwf)]/qg
118
Radial Flow Plot

2.5E+04
2.0E+04
1.5E+04
1.0E+04
5.0E+03
0.0E+00
1
Superposition Time
No Corrections Made
Corrections Made
Fig. 20. Flowing material balance analysis of Case 3 and radial ow analysis of Case 4, with and without corrections for relative permeability to gas (krg) changes.
As will be discussed further below for type-curve application to

CBM and shale, and was discussed previously for straight-line
methods, the time and pseudo-pressure functions may require alteration, depending on the combination of uid/reservoir properties,
and boundary conditions. For production analysis, many researchers
have altered these dimensionless variable denitions to improve the
uniqueness and usefulness of the type-curves. We demonstrate the
application of several type-curves for vertical and horizontal wells
below.
3.2.1. Fetkovich type-curves
Fetkovich (1980) generated the rst generation of production
type-curves that continue to be useful today. Fetkovich combined analytical solutions for constant owing-pressure radial ow of liquids
with the empirical decline-curve results of Arps (1945) for boundarydominated ow. He altered the dimensionless rate and time denitions
as follows:
qDd qD lnre =r wa 1=2
25
2

tD :
t Dd
r e =r wa 2 1 lnr e =r wa 1=2
26
Note that Fetkovich originally used the liquid-version of

dimensionless-rate and time variables instead of Eqs. 2223. The
resulting type-curve set is given in Fig. 21. To extend the application
of the type-curves to gas, Eqs. 2223 are used, which include gas
properties and pseudopressure.
To use the type-curves, dimensional ratetime data from a well are
plotted on the same loglog scale as the dimensionless type-curves,
and then adjusted to overlay the dimensionless type-curves. Details
of the type-curve matching procedure are provided in Lee and
Wattenbarger (1996) and Poston and Poe (2008). Assuming that both
transient and depletion data are evident in the well production data,
and further that the well exhibits transient radial ow, the match
point allows permeability and skin (from transient data) and EUR
(from depletion data) to be obtained for the well. For the transient
stem (reD = re / rwa), the smaller the reD value, the more stimulated
the well is (more negative skin). Note that in Fig. 21, the transient
stems are plotted as re / rw as opposed to re / rwa to be consistent with
the original work of Fetkovich (1980); however, the derived wellbore
radius is really an apparent wellbore radius, which could be larger or
smaller than the real wellbore radius, depending on whether skin is
negative or positive, respectively. Equations used to derive these parameters (k, s) for liquids (i.e. undersaturated oil reservoirs) are given
in the original work by Fetkovich (1980), and for gas in Fetkovich et
al. (1987). In subsequent work, Fetkovich et al. (1996) addressed such
issues as: the impact of uid-type, reservoir-type, drive mechanism,

and drawdown on b-stem values (depletion data); rate-normalization
for variable operating conditions (owing pressures); reinitialization
when drive or production mechanism has changed; and effects of stimulation and re-stimulation. It should be noted that Fetkovich used
real-time in his analysisthe late time match (b-stem) therefore
depended in part on uid property changes during depletion. For constant owing pressure production of liquids from a single-layer homogeneous reservoir, depletion data should fall down the exponential
depletion stem (b = 0), whereas low pressure gas or saturated oil reservoirs would tend to follow hyperbolic stems (b > 0). A summary of
expected b-stem values, depending on various mechanisms, is provided
in Fig. 22.
As was later demonstrated, the use of pseudo-time (Fraim and
Wattenbarger, 1987, Eq. (2)) allows gas data to be analyzed on the
Fetkovich type-curves, with depletion data following the exponential
decline (b = 0) stem of the type-curve (for single-layer systems). Demonstrated later still (Palacio and Blasingame, 1993), the use of material
balance pseudo-time (Eq. (3)), which is a superposition time function
strictly applicable to boundary-dominated ow, causes depletion data
to fall down the harmonic stem for conventional gas reservoirs. Use of
the material balance part of MB pseudotime converts the variable rate
data to the equivalent constant rate conditionthe constant-rate solution to the ow equations during boundary-dominated ow was
shown to follow a harmonic decline (Palacio and Blasingame, 1993).
Many of the type-curve solutions developed for pressure transient analysis, and recent solutions in support of production analysis, were developed with the constant-rate boundary condition. As was discussed in
the straight-line analysis section, and expanded upon below, further
modications to pseudotime and material balance pseudotime allow
additional pressure- and saturation-dependent to be accounted for unconventional gas reservoirs.
Application of the Fetkovich type-curves is demonstrated for the
simulated dry coal example of Fig. 2, which is reproduced from
Clarkson (2009). In the author's experience, CBM (producing from vertical wells) is among the only unconventional gas types that have transient radial ow signaturesthis is because CBM wells that are not
hydraulically-fracture stimulated, but have dense enough natural
fracturing, may exhibit transient radial ow signatures. Examples
from the Horseshoe Canyon coals, which are often stimulated using
proppantless, high-rate nitrogen injections, were provided in Clarkson
(2009), Clarkson and Bustin (2011) and Clarkson et al. (in press). Examples from the San Juan Basin (Fruitland) Coal and Uinta Basin (Ferron)
Coal were provided in Clarkson et al. (2007, 2012c), respectively.
The Fetkovich type-curve match of the simulated dry coal data is
given in Fig. 23 using real-time (Fig. 23a) and material balance
pseudotime (Fig. 23b). The dimensionless time functions used in
a)
re/rw=10
Fetkovich Type-Curves
re/rw=20
10
119
b)
re/rw=50
Transient
re/rw=100
Depletion
re/rw=200
re/rw=1000
qDd
re/rw=10000
b=0
0.1
rwa
b=0.2
b=0.4
b=0.6
0.01
b=0.8
b=1.0
0.001
0.0001
0.001
0.01
0.1
10
100
re
1000
tDd
Fig. 21. (a) Illustration of the Fetkovich type-curves. (b) Reservoir geometry associated with the Fetkovich type-curves. Note that in (a) the transient stems are plotted as re / rw as
opposed to re / rwa to be consistent with the original work of Fetkovich (1980); however, the derived wellbore radius is really an apparent wellbore radius (b), which could be larger
or smaller than the real wellbore radius, depending on whether skin is negative or positive, respectively.
Modied from Fetkovich (1980).
Eq. (26) for the case of real-time and material balance pseudotime are
given below:
0:00634 kg

t
tD
g ct i r 2wa
27
0:00634 kg

t ca
tD
g ct r 2wa
28
where ct is total compressibility modied for adsorption, rwa is the

apparent wellbore radius (rwa = rwe s), assuming a circular drainage
area, and s = skin. Note that the unit of time in Eqs. (27) and (28) is
was approximated as:
days. For Fig. 23b, tca

g ct Z

i Gi
t ca
mpi mpr :
29
qg
2pi
For the single-phase case given in Fig. 23, kg = k (effective permeability to gas = formation permeability).
Fig. 22. Impact of reservoir-type, uid properties, drive mechanism and operating conditions on b-stem value. Slide courtesy of Dave Reese.
120
a)
b)
10
10
Transient
Depletion
Transient
qDd
qDd
Depletion
0.1
0.1
0.01
0.0001
0.001
0.01
0.1
10
0.01
0.0001
0.001
0.01
tDd
0.1
10
tDd
Fig. 23. Fetkovich type-curve match of simulated dry coal well using (a) real time and (b) material balance pseudotime.
We see that the impact of the use of real-time in the Fetkovich

type-curve is that the depletion data falls down the b= 0.8 stemthis
value is greater than that normally found for low-pressure gas producing from a single-porosity reservoir (0b b 0.5, see Fig. 22) because of
the additional pressure support provided by desorption. Selection of
material balance pseudotime, whose calculation requires use of a material balance equation, results in the data falling down the harmonic
stemmaterial balance time, which is a superposition time function,
converts the variable rate to the equivalent constant rate condition.
calNote that incorrect gas-in-place estimates, which are required for tca
culation, will result in b-stem values different than 1. Further, the selection of either time function does not impact the transient stem match.
Clarkson (2009) noted that the permeability and skin values using the
Fetkovich type-curve match shown above are in reasonable agreement
with simulator input (k = 1, s = 0) and with transient radial ow analysis (straight-line). Further, linkage of the OGIP estimate from owing
calculation for type-curve calculations is validatmaterial balance to tca
ed with the b = 1 match on the type-curve.
3.2.2. Blasingame type-curves
Blasingame and his students and co-workers extended the
Fetkovich type-curves by making them applicable to gas reservoirs
with variable-rate production (through the use of material balance
pseudotime and pseudopressure-normalized rate); they also developed several additional ratetime type-curve sets that are applicable
to hydraulically-fractured vertical wells (i.e. Amini et al., 2007;
a)
Pratikno et al., 2003). Most unconventional gas wells are

hydraulically-fractured, which often results in early ow-regimes
that are non-radial (see Fig. 3). An example type-curve set that is applicable to vertical wells, with planar hydraulic fractures, producing
from a circular reservoir are given below (Fig. 24).
Because the type-curve set was generated using the constant-rate
boundary condition, there is only one depletion stem (b = 1)material
balance pseudotime must therefore be used in the calculation of dimensionless variables for the real data. The type-curve set in Fig. 24 utilizes
the same dimensionless variable format as the Fetkovich type-curves
(Fig. 21), except that qDd and tDd are dened as below:
qDd qD bDpss
30
2
t
bDpss DA
31
t Dd
where bDpss is a pseudo steady-state parameter that depends on FcD and

reD (correlation derived in Pratikno et al., 2003), which in turn are dened as:
F cD
kf wf
kxf
32
r eD
re
xf
33
reD=2
Pratikno-Blasingame Type-Curves
reD=3
100
b)
reD=4
reD=5
10
reD=10
reD=20
reD=30
qDd
reD=40
reD=50
rw
reD=100
0.1
xf
reD=200
reD=300
0.01
reD=400
reD=500
0.001
0.0001
re
reD=1000
0.001
0.01
0.1
10
100
1000
tDd
Fig. 24. (a) PratiknoBlasingame (ratetime) type-curves (Pratikno et al., 2003). Type-curve set shown is for dimensionless conductivity (FcD) = 1000. (b) Illustrates the reservoir/
hydraulic fracture geometry used in the type-curve development.
(re/xf) stem match, but note that the initial slope of the simulated
data is shallower in the nite-conductivity fracture case. Further, a
perfect match to the type-curve was not achieved because the transient data fall between type-curve stemswhen this occurs, then a
range of xf values can be obtained, based on the two closest matching
stems, instead of an exact value.
Pratikno et al. (2003) also utilized the rate-integral and rate-integral
derivative, originally dened by McCray (1990), to guide type-curve
matching particularly in the face of noisy data. The concept of the
rate-integral is given in Fig. 27, and the rate-integral and
rate-integral-derivative functions used by Pratikno et al. (2003) are dened below. At each point of the well producing time, the cumulative
production is divided by time to dene an average rate that yields the
same cumulative production. This calculation is performed at each
time, yielding a smoother rate-prolethe rate-integral derivative is
simply the derivative of this smoothed rate with respect to time. As
noted by Mattar and Anderson (2003), the rate-integral is sensitive to
early time (data) errors, and the rate-integral derivative does not
yield information about ow-regimes (unlike the derivatives used
above), but is useful for pattern recognition.
h
Fracture
Impermeable
Boundaries
xf
Wellbore
Fracture
Impermeable
Boundary
Fracture Flow Geometry
qDdi
Fig. 25. Conceptual model for nite-conductivity vertical, planar fracture. FcD is a measure of the conductivity contrast between hydraulic fracture and reservoir. For nite
conductivity fractures (FcD b 150), pressure drop along the fracture is signicant (leading to bilinear ow period, see Table 1).
Modied from Mavor and Saulsberry (1996).
where kf is the hydraulic fracture permeability, wf is the hydraulic fracture width (see conceptual model in Fig. 25). Because there are different
type-curve sets for each value of FcD, FcD must be dened (in Fig. 24,
FcD = 1000). This fact illustrates an issue with the uniqueness of
type-curve matching for unconventional reservoirsthe more complex
the ow-regime sequence, and consequently the number of reservoir
and fracture parameters that affect them, the more dimensionless parameters required to generate the type-curves. This in turn affects the
uniqueness of the matching process, because more parameters are required to be known to obtain a match. The problem becomes worse
with multi-fractured horizontal wells (see below).
In Fig. 26, we show the match of a simulated vertical well completed in a tight gas reservoir, with a) an innite-conductivity fracture
and b) nite-conductivity fracture. The fracture half-length (xf) and
drainage radius (re) are the same in both cases, giving a similar reD
a)
1
t Dd
qDdid
t Dd
qDd d
dqDdi
dq
t Dd Ddi qDdi qDd
d lnt Dd
dt Dd
35
Fig. 28 shows the analysis of the simulated vertical well with an

innite-conductivity fracture (Fig. 26a), using the PratiknoBlasingame
type-curves with the auxiliary (rate-integral and rate-integral derivative) type-curves in addition to the ratetime type-curves. We see
that including these auxiliary curves does improve match uniqueness,
although ow-regimes are not identied as effectively as shown in
Fig. 3b using the conventional rate-normalized pseudopressure semilog derivative.
Blasingame and his research group have developed several analytical solutions and accompanying type-curve sets that have proven
useful for unconventional gas analysis, including those developed
for elliptical ow (Amini et al., 2007). Further, as with Pratikno et
al. (2003), additional type-curves using a rate versus cumulative production format were generated.
3.2.3. Wattenbarger type-curves
In 1998, Wattenbarger et al. published an important work in
which constant pressure/rate analytical solutions for the case of linear
100.00
Type-Curve set for FcD = 1000
Type-Curve set for FcD = 5
10.00
10.00
qDd
qDd
34
b)
100.00
1.00
0.10
0.01
0.001
121
1.00
0.10
0.01
0.1
tDd
10
100
0.01
0.001
0.01
0.1
10
100
tDd
Fig. 26. PratiknoBlasingame type-curve match of simulated hydraulically-fractured tight gas well with (a) innite conductivity fracture and (b) nite-conductivity fracture. A perfect match was not achieved because data fall between type-curve stems.
122
rate integral
= Q/t
actual rate
ow. The following dimensionless variables (for gas) are used for the
type-curves in Fig. 29 (unit of time is days):
qD
1424T
h
i qg
kg h mpi m pwf
36
0:00634 kg
t
t Dye
g ct y2e
37
0:00634 kg
t:
t Dxf
g ct x2f
38
actual
time
actual
time
Noting:
Fig. 27. Concept of a rate-integral. Figure courtesy of Fekete Associates Inc.
t Dye
ow into hydraulically-fractured tight gas wells were presented. In
this same publication, application of the square-root of time plot
was demonstrated. They noted that many tight gas wells do not exhibit pseudo-radial ow, but instead exhibit long periods of transient
linear ow, followed in many cases by boundary-dominated ow.
This corresponds to the ow-geometry of Fig. 12. The analytical solutions used in type-curve development were developed by El-Banbi
(1998). The straight-line and type-curve techniques developed in
this seminal work have proven extremely useful for both tight gas
and shale gas analysisfor the latter, Bello (2009) later adopted the
methods for analysis of long-term transient linear ow associated
with shale gas wells, resulting in numerous publications (ex. Bello
and Wattenbarger, 2008, 2009, 2010). There are two forms of the linear ow type-curves that this author has utilized extensively for tight
gas and shale gas analysis, which are given in Fig. 29.
The type-curves in Fig. 29 were developed using the constant-rate
solution, and hence the unit-slope (b=1) during boundary-dominated

2
xf
t Dxf :
ye
39
As with the Fetkovich and Blasingame type-curves, material balance pseudotime must be used in real data transformation to force
the data down the harmonic stem (b = 1, constant-rate solution).
Also note that the ye denition used by Wattenbarger et al. (1998)
in the above dimensionless time denitions differs from Fig. 12, and
corresponds to Fig. 30.
Application of these type-curves to Case 1 (simulated shale-gas
well with transient-linear, followed by boundary-dominated ow)
is given in Fig. 31. The material balance pseudotime used in this
case is Eq. (29), which accounts for adsorptionfurther, total compressibility has been altered to account for adsorption. Finally,
owing material balance (Fig. 15, corrected FMB) was used to generate reservoir pressures used in turn for material balance pseudotime
calculationsi.e. the depletion data in Fig. 31 are anchored to FMB
calculations. The OGIP estimate using the type-curve is therefore
tied to the FMB-derived value for consistency. The xf derived from
the type-curve match is consistent with the straight-line analysis
Rate-integral type-curve match
Rate-time type-curve match
Rate-integral-derivative
type-curve match
Fig. 28. PratiknoBlasingame type-curve match of simulated hydraulically-fractured tight gas well with innite conductivity fracture (corresponding to Fig. 26a). Commercial software (Fekete F.A.S.T RTA) was used for analysis. A perfect match was not achieved because data fall between type-curve stems.
a)
b)
Wattenbarger Linear Flow Type-Curves

100
123

10
ye/xf = 1
Linear Flow
LinearFlow
Boundary-Dominated Flow
ye/xf = 2
ye/xf = 3
ye/xf = 4
qD
(ye/xe) qD
10
ye/xf = 6
0.1
ye/xf = 10
0.1
0.01
0.0001
0.01
0.001
0.01
0.1
10
100
0.1
tDye
10
100
1000
tDxf
Fig. 29. Wattenbarger type-curves using two dimensionless variable formats.
performed previously, and from the type-curve match in Fig. 31b,

ye / xf = 10, which is what was input into the simulator.
As will be seen, the linear-to-boundary ow type-curves of Fig. 31
have proven to be very useful for multi-fractured horizontal wells
where the stimulated reservoir volume denes the extent of the reservoir. In such cases, early linear ow may be followed directly by true
boundary-dominated ow, or possibly pseudo-pseudosteady-state
(Figs. 6 and 7) ow. In other cases of higher permeability, the dominant
transient ow regime may be late-linear ow, which in turn may be
followed by well interference. In either of these scenarios, the linear
ow type-curves of Fig. 29 may be useful.
In later work, Wattenbarger and students (i.e. Bello and
Wattenbarger, 2010) developed type-curves that account for transient (linear) dual porosity behavior associated with multi-fractured
horizontal wells in shale. Two ow-regime sequence scenarios associated with transient dual porosity behavior are given in Fig. 8these
scenarios and more have been captured in the new type-curves. The
existence, and timing, of ow-regimes are dependent on the magnitude of the dual porosity parameters and reservoir geometry. In
many shale gas wells, early transient linear (1/2 slope) behavior is
preceded by 1/4 slope or at least less than 1/2 slope behavior.
Abdulal et al. (2011) have argued that this behavior is associated
xf
ye
xe
Fig. 30. Geometry utilized in Wattenbarger type-curve development (Fig. 29).
with bilinear ow, which in turn is simultaneous linear ow in fractures and in the matrix (Fig. 8b). Still others (Nobakht and Mattar,
2012) have argued that the sub-linear slopes are associated with
skin effects, which can be corrected for in type-curve analysis (see
below). Until the early-time skin versus reservoir behavior issue for
shale gas wells is resolved, the usefulness of the more complex (transient dual porosity) type-curves remains in question.
3.2.4. Multi-fractured horizontal well type-curves
Several researchers have developed type-curves for multi-fractured
horizontal wells in both conventional and unconventional (tight gas
and CBM) reservoirs. For conventional reservoirs, the efforts to derive
analytical solutions for multi-fractured wells to support pressure- and
rate-transient analysis by Chen and Raghavan (1997), Larsen and
Hegre (1991), Raghavan et al., and van Kruysdijk and Dullaert (1989)
are particularly noteworthy. Referring to the work by Chen and
Raghavan (1997) and Raghavan et al. (1997), the analytical solutions
developed capture the ow-regimes illustrated in Fig. 5. As noted by
Ozkan et al. (2011), ow convergence beyond the fracture tips (i.e.
compound or late linear ow, or contribution from outside the stimulated reservoir volume) may not occur for ultra-low permeability tight gas
and shale gas reservoirs, and further noted that ow convergence is
mainly linear ow perpendicular to hydraulic fracture surfaces. The
long-term performance of ultra-low permeability reservoirs is therefore
generally dominated by linear ow to fractures (early linear ow of
Fig. 5), which can be represented by an equivalent fracture, whose
length is equal to the aggregate length of hydraulic fractures and the
conductivity of the equivalent fracture is equal to the average conductivities of the individual fractures (Ozkan et al., 2011). Bello (2009)
generalized this concept for shale wells that exhibit complex fracture
behaviorthe dominant linear ow is from the matrix to induced or
natural fractures, and the fractures are best represented in terms of
contacted surface area (or Acm). In such cases, the induced hydraulic
fracture network (see Figs. 1 and 9) represents a transient dual porosity
system. Recent efforts to develop type-curves for shale wells
have therefore focused on developing type-curves for transient dual
porosity systems. Examples include Abdulal et al. (2011), Bello and
Wattenbarger (2010) and Moghadam et al. (2010) to name just a few.
Some researchers (Al-Ahmadi and Wattenbarger, 2011) have even
developed type-curves for triple porosity systems, matrix and two
microfractures (less permeable and more permeable), with applications
to shale gas. Brown et al. (2011) recently presented dimensionless
type-curves based upon the trilinear ow solutionthe conceptual
model for the trilinear ow solution is provided in Fig. 32. Flow in this
system, which corresponds to Scenario 7 of Fig. 9, was approximated
by assuming that the ow-regimes correspond to transient linear ow
124
a)
b)
100
10
ye/xf = 1
ye/xf = 2
ye/xf = 3
ye/xf = 4
1/PwD
(ye/xe) qD
10
ye/xf = 6
ye/xf = 10
0.1
Data
0.1
0.01
0.0001
0.001
0.01
0.1
10
100
0.01
0.1
10
tDye
100
1000
tDxf
Fig. 31. Wattenbarger type-curve match of Case 1 using material balance pseudotime and dimensionless variables adjusted for adsorption.
within the hydraulic fractures, linear ow from the inner reservoir

(stimulated reservoir volume) to the hydraulic fractures and linear
ow from the outer reservoir to stimulated reservoir volume. The
inner matrix was represented as a transient dual porosity (naturally
fractured) system, while the outer reservoir was represented as a single
porosity system. The authors note that contribution from outside the
SRV may be negligible except for very late in the production life of the
well.
As noted in a previous section, early cleanup (multi-phase ow,
convergence ow etc.) has rendered quantitative analysis of early
ow-regimes (i.e. prior to development of matrix linear ow) using
type-curves (and even straight-line methods) difcult. The cleanup
period of multi-fractured horizontal wells in shales may be extensive
due to the large volumes of hydraulic fracturing uid used in stimulation. Nobakht and Mattar (2012) have argued, as have others, that the
dominant transient ow period in shales and many tight formations
is transient linear ow from inner reservoir (SRV) to the induced hydraulic fracture network and that this ow-regime may be masked at
early time due to cleanup effects. Assuming that the cleanup behavior
can be represented by a constant skin, which appears as a positive intercept on the square-root of time plot, the authors proposed that in
such cases, a modied pseudopressure-normalized rate be calculated

as follows:
qm
mpi mpwf
qg
1
p
m t
where:
3

2
mpi m pwf
mpi m pwf
4
5
b
qg
qg
b and m represent the intercept and slope of the square-root of time

plot, respectively. The modied pseudopressure-normalized rate (qm)
can then be used in type-curve matching using the procedures
discussed above. As Nobakht and Mattar (2012) noted, failure to make
this correction may cause the early ow-regimes to be misidentied.
Using Scenario 5 of Fig. 9, Nobakht et al. (2012a) developed a new
type-curve set. Details of the type-curve generation were provided in
that work, but effort was made to dene dimensionless variables that
OUTER RESERVOIR
ko, o, cto
xf
INNER RESERVOIR
NATURALLY
FRACTURED
41
NO-FLOW BOUNDARY
xe
40
HYDRAULIC FRACTURE
kF, wF. F, ctF
kf, f, ctf
km, m, ctm
Horizontal Well
Fig. 32. Conceptual model for trilinear ow solution, used in turn for dimensionless type-curve development by Brown et al. (2011).
125
The following variables may be obtained from the type-curve

match:
44
0:00633 1:417 106 T

A xe ye
X match Y match :
h
g ct
45
g t
ye
2xf
v
p
1:417 106 T u
u0:00633
X match Y match
kxf
t
h
c
i
xf k :
A = xe ye
xe
Ymatch
X match
X match Ymatch
Fig. 33. Element of symmetry used in the development of the Nobakht and Clarkson
MFHW type-curves (Nobakht et al., 2012a) and variables that can be derived from
/tDM or tca / tDM and Ymatch
x-axis
and y-axis

match on the type-curve. Note Xmatch is tca
qg
is mp m
for
gas.
=q
DM
pwf
i
resulted in a unique type-curve set. An element of symmetry was assumed as follows in the development of the Nobakht and Clarkson
type-curves (Nobakht et al., 2012a) (Fig. 33).
The resulting dimensionless type-curve set for two combinations
of ye/xe ratios are provided in Fig. 34.
Note that for ye = 2xf, the type-curves reduce to the ow-regimes
captured in the Wattenbarger type-curve (linear ow followed by
boundary-dominated ow, Fig. 29a), which in turn assumes that the
SRV denes the reservoir. The usefulness of the Nobakht and Clarkson
type-curves is in capturing ow-regimes associated with MFHW for
many combinations of ye (well spacing), xe (fracture spacing) and xf.
As demonstrated by those authors, the type-curves can be used to
assist ow-regime identication and reservoir/hydraulic fracture
characterization. As the authors noted, the ow regime sequence, relationship and length depend strongly on ratios of these parameters.
The dimensionless rate and time variables used in type-curve development are provided below:
qDM qD
!2

xe
ye
ye
xf
t DM t Dxf
42

2
4
xf
ye
:
xe
ye
a)
43
Application of the ratetime type-curves to the simulated MFHW

completed in a tight gas reservoir, previously analyzed with straightline analysis (Fig. 11), is given in Fig. 35. Note that the total gas rate
needs to be divided by the number of fractures (5 in this case) to use
the type-curves.
In this example, the ye/xe ratio is ~ 6, so the type-curves corresponding to this ratio were selected. Only the rst ~ 200 days of production data are selected for analysis. We see that some transitional
ow from early linear (1/2 slope) ow occurs, with the data falling
input)
ye/xf
close to the ye/xf = 10 and 15 stems. The actual (model p
value is ~ 11. The type-curve-derived values for A and xf k serve to

validate the straight-line analysis performed above. In fact, the
author's preferred approach to type-curve
analysis is to use the
p
straight-line-derived estimates for xf k etc. as inputs into the dimensionless variable denitions for the real production data, and observe
the resulting type-curve match, as was done with the linear ow
example in Section 3.2.3the type-curve match and straight-line results should be consistent.
3.2.5. Other useful type-curves
In this review, we cannot possibly cover all of the type-curve sets
that have, and continue to be, used for unconventional gas applications. In this work we have emphasized the use of ratetime
type-curves, but it should be noted that type-curves have also been
generated in cumulative production versus time or rate versus cumulative production formats.
An additional type-curve set that was not illustrated above, but
which the author has used historically for analyzing hydraulicallyfractured vertical wells, are the AgarwalGardner type-curves
(Agarwal et al., 1999). These type-curves have proven very useful
for tight gas applications and dry CBM wells, as illustrated by several
authors, including Cox et al. (2002) and Clarkson (2009), respectively.
Agarwal et al. (1999) also generated other useful formats, such as
b)
Nobakht-Clarkson Type-Curves: ye=xe
Nobakht-Clarkson Type-Curves: ye=6xe

1.E+04
1.E+04
ye/xf=20
ye/xf=20
1.E+03
1.E+03
ye/xf=15
ye/xf=15
1.E+02
1.E+02
ye/xf=10
1.E+00
ye/xf=6
1.E-01
ye/xf=4
ye/xf=10
1.E+01
qDM
qDM
1.E+01
1.E+00
ye/xf=6
1.E-01
ye/xf=4
1.E-02
1.E-02
ye/xf=3
ye/xf=3
1.E-03
1.E-03
ye/xf=2
ye/xf=2
1.E-04
1.E-08
1.E-06
1.E-04
1.E-02
1.E+00
1.E+02
1.E-04
1.E-08
1.E-06
1.E-04
tDM
Fig. 34. Nobakht and Clarkson ratetime type-curves for two combinations of ye/xe.
1.E-02
tDM
1.E+00
1.E+02
126
a)
b)

ye/xf=20
1.E+04
ye/xf=15
ye/xf=15
ye/xf=10
ye/xf=10
1.E+03
1.E+03
ye/xf=20
ye/xf=6
ye/xf=6
1.E+02
ye/xf=4
1.E+01
ye/xf=3
ye/xf=2
1.E+00
1.E+02
ye/xf=4
ye/xf=3
qDM
qDM
1.E+05
ye/xf=2
Data
Data
1.E-01
1.E+01
1.E-02
1.E-03
1.E-04
1.E-08
1.E-06
1.E-04
1.E-02
1.E+00
1.E+02
1.E+00
1.E-06
1.E-05
1.E-04
tDM
1.E-03
1.E-02
1.E-01
1.E+00
tDM
Fig. 35. (a) Nobakht and Clarkson ratetime type-curves for two combinations of ye/xe. (b) is a zoom in of the type-curve match in (a). The type-curve match is for a simulated
MFHW.
3.2.6. Alternative type-curve matching approaches

Conventional type-curve matching procedures involve matching
production data to dimensionless type-curve sets that were previously generated using analytical or numerical models. The type-curve solutions for various combinations of well, reservoir and hydraulic
fracture geometries appear as stems on the type-curve. As illustrated above (Figs. 26, 28, and 35), actual data may fall between stems,
depending on the type-curve resolution, leading to a range of parameter estimates. An alternative approach, as has been demonstrated by
Cox et al. (2002) using the AgarwalGardner type-curve format, is to
history-match actual production data with an analytical or numerical
simulator, and then convert the output of the simulator to dimensionless variables. The actual production data is then compared to the
generated type-curves for the matching processa new type-curve
set is re-generated every time simulation input are adjusted in the
matching process. This process circumvents the resolution limitations
of many existing type-curve sets. The physical model used by analytical or numerical simulator must of course be consistent with the real
well/reservoir combination.
A eld example of a multi-fractured horizontal well in a tight gas
reservoir is given in Fig. 36. In this example, the type-curves were
generated using output from a numerical model used to historymatch the eld data. As discussed by Clarkson and Beierle (2011),
the input parameters used to achieve the type-curve match are consistent with those derived from straight-line analysis.
3.2.7. Additional considerations for CBM and shale reservoirs
As with straight-line analysis, an approach for correcting type-curve
analysis for complex reservoir behavior (adsorption, non-static permeability, non-Darcy ow etc.) is modication of pseudovariables that are
used in dimensionless variable calculations. Corrections for non-Darcy
ow (gas-slippage) in pseudopressure and pseudotime calculations
were discussed in the context of multi-fractured horizontal well typecurves by Nobakht et al. (2012a). Figs. 23 and 31 demonstrated that
the Fetkovich and Wattenbarger type-curves can be used for dry CBM
and shale gas reservoirs in the presence of desorption, respectively,

using the modied material balance pseudotime (Eq. (29)).
Analysis of multiphase production data is the most complex application of type-curves. As noted above, 2-phase CBM reservoirs have been
analyzed using type-curve methods that account for multi-phase ow
(Aminian and Ameri, 2009; Mohaghegh and Ertekin, 1991).
The approach of Clarkson et al. (2012c) is illustrated below for
Cases 3 and 4as with straight-line analysis, output (pressure and
saturations) from the analytical simulator used to match the data is
used in pseudo/dimensionless variable denition (Eqs. (22)(24)).
In Eqs. (22)(24), kg is now a variable that depends on relative permeability. For Case 3 (Fig. 37) all the data fall down the b = 1 depletion stem because of the use of material balance time corrected for
desorption (Eq. (28))the match to b = 1 conrms that OGIP is correct, material balance (reservoir pressures) are calculated correctly
and relative permeability has been correctly accounted for. If relative
permeability to gas is assumed to be constant, the data do not follow
the type-curve, except at late time (Fig. 38).
Case 4, which has transient data in addition to boundarydominated (depletion) data is shown plotted on both the Fetkovich
and Pratikno-Blasingame type-curves (Fig. 39). A match to high re/rw
and re/xf values conrm the non-stimulated nature of the simulated
well (zero skin); the depletion data also fall down the b = 1 stem,
conrming OGIP/material balance.
Horizontal Well with Transverse Fractures

100
qD or d(1/qD)/dln(tD)
ratecumulative type-curves that have been used extensively in the

literature (i.e. Cox et al., 2002).
Lewis and Hughes (2008) developed vertical well type-curves for
dual porosity shale reservoirs, and accounted for the effects of
desorption.
Dimensionless type-curves have also been generated using a
dimensionless productivity index versus time format. This approach
is illustrated in Araya and Ozkan (2002) and Brown et al. (2011).
q D type-curve
10
d(1/qD)/ dln(tD)type-curve
0.1
0.01
0.0001
0.001
0.01
0.1
tD
Fig. 36. Type-curve match of numerical (solid green and blue lines) to eld data for a
multi-fractured horizontal well completed in a tight gas reservoir.
a)
b)
500
800
400
600
300
400
200
100
200
0
200
400
600
800
1000
0
1200
Time, days
Modelqg
Actualqg
Transient
Depletion
qDd
1000
10
Water Rate, STB/D
Gas Rate, Mscf/D
127
0.1
0.01
0.0001
0.001
0.01
0.1
10
tDd
Actualqw
Modelqw
Fig. 37. Analytical model history-match (a) and type-curve analysis (b) of simulated 2-phase (gas + water) CBM case (Case 3).
3.3. Analytical and numerical simulation

Numerical and analytical simulation of unconventional gas reservoirs is only briey touched on in this summary because the subject
warrants a separate review. In the context of rate-transient analysis, analytical and numerical simulation is used to 1) generate type-curves for
type-curve matching (see above); 2) validate parameters (i.e. reservoir
and hydraulic fracture parameters derived from rate-transient analysis)
through history-matching; and 3) to generate production forecasts.
Analytical simulation models are models that use solutions to mathematical models derived analytically, such as from Laplace-Transform
techniques (Lee et al., 2003). Analytical models generally correspond
to relatively simple reservoir behavior and boundary-conditions. Numerical simulation models use solutions to (potentially) more complex
mathematical models that are derived from numerical methods, such as
nite-difference or nite element.
Numerical simulation of coalbed methane wells has received extensive attention and has been comprehensively summarized by
Paul (1996). Numerical simulation of shale gas reservoirs has recently
been summarized in the works of Cipolla et al. (2010), including how
to model complex hydraulic fracture networks (Figs. 1 and 9). Important areas of research are inclusion of non-Darcy effects in numerical
simulators, such as gas slippage and diffusion (i.e. Andrade et al.,
2011; Swami and Settari, 2012). Some authors (Clarkson et al.,
10.00
Transient
Depletion
qDd
1.00
0.10
0.01
0.0001
0.001
0.01
0.1
10
tDd
Fig. 38. Fetkovich type-curve analysis of Case 3, with (green data) and without (red
data) corrections for relative permeability to gas (krg) changes.
2012b) have chosen to model these effects by incorporating apparent

permeability (or transmissibility) changes as a function of pressure in
conventional simulators; Javadpour (2009) noted that non-Darcy effects manifest themselves as apparent permeability changes that are
a function of gas composition, temperature and pressure. Additional
areas of research include coupled geomechanical and ow models,
to enable modeling of hydraulic fracture network creation during
stimulation, dynamic owback of hydraulic fracture and reservoir
uids and long-term production. Further areas of research that will
affect simulation approaches to unconventional reservoirs are nonequilibrium capillarity (i.e. Andrade et al., 2011), multi-component
adsorption and diffusion of gases, and phase behavior of liquids in
nanoporous materials.
Analytical models that have been derived in the literature for CBM
and shale are too numerous to review in this work. Clarkson and
McGovern (2005) summarized single- and 2-phase CBM analytical
models that have been used by Clarkson et al. (2012b) for ratetransient analysis (coupled with straight-line and type-curve analysis), as described above. Seidle (2011) also summarizes analytical
(tank) models that have been used for CBM. For shale gas, many analytical models have been derived historically in parallel with CBM
models (ex. Carlson and Mercer, 1991). In more recent work, the trilinear ow model of Brown et al. (2011) has proven useful for both
rate- and pressure-transient analysisthose authors also provided
asymptotic approximations useful for understanding the origin of different ow regimes. As with numerical simulation, a research focus in
the future will be inclusion of more complex reservoir behavior into
analytical models. For example, Ozkan et al. (2010) recently incorporated gas-slippage effects and non-static fracture permeability effects
into the trilinear ow solutions. Apaydin et al. (2012) recently
discussed the inclusion of discontinuous matrix microfractures in analytical modeling of shale gas reservoirs.
The focus of the following is on analytical history-matching/
forecasting methods that are directly linked with rate-transient analysis. For example, Clarkson and Beierle (2011) derived a multisegmented forecasting method where each segment corresponds to
a ow-equation, from which input is derived from the corresponding
ow-regime analysis. This is illustrated for a simulated multi-fractured
horizontal well with innite (Fig. 40) and nite (Fig. 41) conductivity
planar fractures. In both of these cases, the slope and intercept of the
square-root of time plot for early linear ow analysis were used to forecast early transient linear ow; for the innite conductivity fracture
case, the early linear ow forecast in loglog coordinates follows a 1/2
slope (Fig. 40), whereas it is b 1/2 slope for the nite-conductivity case
128
a)
b)
10.00
Transient
100.00
Depletion
10.00
qDd
qDd
1.00
1.00
0.10
0.10
0.01
0.0001
0.01
0.001
0.01
0.1
10
0.001
0.01
0.1
tDd
tDd
10
100
Fig. 39. Match of Case 4 (2-phase CBM) gas data to (a) Fetkovich and (b) PratiknoBlasingame type-curves.
due to inclusion of skin (Fig. 41). Transitional ow from early-linear to

late-linear is approximated with the radial ow equation, with inputs
from early radial ow analysis. Finally, late linear ow to boundarydominated ow is modeled using the (constant owing pressure)
bounded linear ow equation from Wattenbarger et al. (1998), with inputs derived from late-linear ow analysis and using a pseudotime
function anchored to pressures obtained from owing material balance.
This forecasting approach directly honors parameters derived from
rate-transient (ow-regime) analysis, so that any change in RTAderived parameters is directly reected in the model forecast. Use of
pseudotime and pseudopressure in the analysis ensures that the unique
properties of unconventional gas reservoirs can be incorporated. Additional details of this forecasting methodology are provided in Clarkson
and Beierle (2011) for tight gas reservoirs and Clarkson and Pedersen
(2010) for tight oil reservoirs.
For shale gas reservoirs, transitional ow-regimes (elliptical and
radial ow) are typically not observed, whereas transient linear and
boundary-dominated ow-regimes often are. Bilinear or apparent bilinear ow caused by skin may be seen at early times. We illustrate
the multi-segmented approach using the simulated shale gas example of Case 1 (Fig. 42). In this case, the early linear ow is again
a)
calculated using the slope and intercept of the square-root of time

plot, but after boundaries are reached, the gas rate is calculated
using the following equation (Clarkson et al., 2012d):
h
qgBD
i
mpr m pwf
i :
qgBDi h
mpri m pwf
As with the previous case, reservoir pressures are derived from

owing material balance (which includes corrections for adsorption),
to model boundary-dominated ow. The start of boundarydominated ow is estimated using the distance of investigation equation, knowing distance to the boundary. The approach illustrated in
Fig. 42 is similar to that proposed by Nobakht et al. (2012c) for forecasting shale gas reservoirs, which is referred to as the Simple Yet
Rigorous Method. In the Nobakht et al. approach, later modied by
Nobakht and Mattar (2012) to include skin, transient linear ow is
modeled as in Fig. 42, but boundary-dominated ow is modeled
using the empirical Arps decline curves (discussed below), and is
therefore a hybrid approach because it combines an analytical
method for modeling transient linear ow with an empirical method
b)
Production Rates: Multi-Segment Model
46
100000
10000
10000
Transitional Radial Flow Forecast

Using Radial Flow Equation with
Inputs from Early Radial Flow Analysis
1000
Gas Rate, Mscf/D
Gas Rate, Mscf/D
Early Linear Flow Forecast

Using Sqrt-time Slope and
Intercept
1000
LateLinear Flow to Boundary -Dominated

Flow Forecast Using Bounded Linear Flow Solution
with Inputs from Late Linear Flow + FMB Analysis
100
0.1
100
1
10
100
1000
10000
1000 2000 3000 4000 5000 6000 7000 8000 9000 10000
Time , days
Time, days
Data
Data
Model
Model
Fig. 40. Multi-segment analytical model match of numerically-simulated MFHW completed in tight gas reservoir with innite conductivity fractures: (a) loglog plot and
(b) semi-log plot.
a)
b)

Early Linear Flow Forecast
Using Sqrt-time Slope and
Intercept
Transitional Radial Flow Forecast
Using Radial Flow Equation with
Inputs from Early Radial Flow Analysis
Gas Rate, Mscf/D
10000
1000
Gas Rate, Mscf/D
10000
129
1000
Late Linear Flow to Boundary -Dominated

Flow Forecast Using Bounded Linear Flow Solution
with Inputs from Late Linear Flow + FMB Analysis
100
10
100
1000
100
10000
2000
4000
6000
Time, days
Time, days
Data
Data
Model
8000
10000
Model
Fig. 41. Multi-segment analytical model match of numerically-simulated MFHW completed in tight gas reservoir with nite conductivity fractures: (a) loglog plot and (b) semi-log
plot.
for forecasting boundary-dominated ow. In Fig. 42, boundarydominated ow was modeled analytically using Eq. (46).
The approach illustrated in Fig. 42 and that by Nobakht et al. (2012c)
can be used to model multi-fractured horizontal wells completed in
shale reservoirs (as shown in Clarkson, 2011), but makes several simplifying assumptions about the hydraulic fracture geometry and the reservoir. If a planar fracture geometry is assumed (Fig. 12), then all the
fractures must extend to the boundaries of the reservoir and be evenly
spaced, as shown in Fig. 43, which would cause linear ow to immediately be followed by boundary-dominated ow. Forecasts that use this
assumption (linear to boundary-dominated ow) may be conservative;
in reality, individual hydraulic fractures may be of variable length and
spacing, meaning there could be a long transition between the time of
the rst fracture interference and when all boundaries are reached as
discussed in Ambrose et al. (2011) and further quantied in Nobakht
et al. (2011) using a hybrid technique (described later).
One of the primary advantages of analytical simulation over empirical forecasting (discussed next) is the ability to generate a distribution of forecasts based upon uncertainty of key reservoir and
hydraulic fracture properties. An example of the use of forecasts
a)
generated using stochastic methods was discussed in Clarkson and

McGovern (2005). Application of these techniques for shale gas reservoirs has recently been discussed by Williams-Kovacs and Clarkson
(2011). Very recently, Anderson et al. (2012) described a methodology for quantifying uncertainty associated with parameters obtained
from RTA for shale gas reservoirs.
From a reserves booking perspective, the drainage area is a key parameter controlling individual well reserves, and can be very uncertain.
The method described above, where all boundaries are assumed to be
reached at rst sign of fracture interference will generally lead to a
lower limit estimate of well EUR. Sensitivities to larger drainage areas
can also be performedfor example, using an SRV estimate from microseismic data will generally provide a greater drainage area estimate and
hence a more aggressive forecast. A more aggressive forecast still would
be generated assuming that the drainage area is equal to the well spacing. In each case, the drainage area estimates are defendable and
based upon other physical data derived from drilling and completion
operations. Reserves estimation is beyond the scope of the current discussion, but this author believes that analytical methods, not empirical,
should be the basis of forecasts leading to reserves bookings.
b)
10000
Gas Rate, Mscf/D
Gas Rate, Mscf/D
10000
100000
1000
1000
100
100
10
1
10
100
1000
10000
100000
2000
4000
Time, days
Data
6000
8000
Time, days
Model
Data
Model
Fig. 42. Multi-segment analytical model match of numerically-simulated shale gas example (Case 1): (a) loglog plot and (b) semi-log plot.
10000
130
Planar Transverse Hydraulic Fractures
a)
Horizontal Well
Primary Hydraulic Fractures
Secondary Fractures in SRV
b)
Horizontal Well
Fig. 43. Fracture geometry corresponding to the model assumption of Fig. 42 for multi-fractured horizontal well: (a) planar hydraulic fractures extending to the boundary of the
single porosity reservoir and (b) primary hydraulic fractures and induced hydraulic fracture network (secondary fractures) comprising the SRV. In the latter case, transient linear
ow corresponds to matrix linear ow from the matrix blocks to the primary and secondary fractures (see also Fig. 8), whereas in the former case, transient linear ow is to the
primary hydraulic fractures only. Transient linear ow along primary fractures to the well in both cases is assumed to be too short to be observed.
Lastly, a few additional analytical forecasting methods for shale gas

reservoirs have been developed lately that can also be directly anchored
to RTA estimates. For example, Samandarli et al. (2011) adopted the linear dual porosity approach of Bello and Wattenbarger (2008), based on
the solutions derived by El-Banbi (1998), to forecast shale gas wells. The
solutions correspond to 4 regions (ow-regimes) that would be
expected with linear dual porosity behavior. Application of the equations for forecasting a few real shale gas wells was discussed in
Samandarli et al. (2011). Song and Ehlig-Economides (2011) also describe analytical forecasting methods for shale gas wells that account
for adsorption.
3.4. Empirical methods
Empirical methods use mathematical models for curve-tting production data that are not necessarily derived or are derivable from a
physical model for uid ow through porous media, as with the analytical methods described above. As such, empirical techniques do not
have a rigorous theoretical development, except in some limiting
cases. Once a curve-t is achieved, the calibrated mathematical
model may be extrapolated into the future to generate a forecast. Within this class of techniques is the popular decline-curve method derived
by Arps (1945), which has been used for forecasting both conventional
and unconventional gas wells. Arps' decline-curve forecasting methodology has been applied to single- and 2-phase CBM wells, tight gas and
shale gas wells, with some modication. We begin with a discussion of
the Arps decline curve methodology, followed by a discussion of adaptation of the method for unconventional gas wells. Lastly, recent extensions of the Arps decline-curve methodology such as the Power-Law
Exponential (Ilk et al., 2008a,b) and Stretched-Exponential (Valko,
2009; Valko and Lee, 2010) that attempt to account for the long transient and transitional ow periods associated with tight gas and shale
gas reservoirs, will be described.
The Arps decline method includes the application of the simple
ratetime/ratecum. relations provided in Table 3. As with the discussions above, the focus of the current discussion will be on the ratetime
relationships (Eqs. (47)(49)). An excellent recent summary of the Arps
decline methodology was provided by Poston and Poe (2008).
The Arps ratetime equations (Eqs. (47)(49)) are plotted in
Fig. 44 to illustrate the impact of decline exponent (b) for xed qgi
and Di. Note that the loglog plot of these relationships (Fig. 44a)
has the same shape as the depletion curves on the Fetkovich typecurve (Fig. 21), which uses the Arps equations for boundary ow in
dimensionless form. From the semi-log plot (Fig. 44b), we see that
an increase in b causes attening of the curve; the exponential decline
curve is a straight-line on the semi-log plot.
For application to well data, the model parameters that are adjusted
to match production data include the initial gas production rate (qgi),
Table 3
Summary of Arps' ratetime and cumulativetime relationships (after Blasingame and
Rushing, 2005; Ilk et al., 2008a). See Nomenclature section for denition of variables.
Decline type
Ratetime relationship
Ratecumulative relationship
Exponential (b= 0)
qg = qgi exp[Dit]
(Eq. (47))
qgi
qg
1=b
Gp
qgi
Di
Gp
qgi
1bDi
Gp
qgi
Di
Hyperbolic (0 b b b 1)
Harmonic (b = 1)
1bDi
(Eq. (48))
qgi
qg 1D
i
(Eq. (49))
1 expDi t
h
11=b
11 bDi t
ln1 Di t
a)
b)
Arps Decline Curves
Arps Decline Curves
10000
Gas Rate, Mscf/D
Gas Rate, Mscf/D
10000
131
1000
b=1
100
1000
b=1
100
b=0
b=0
10
10
1
10
100
1000
10000
Time, days
200
400
600
800 1000 1200 1400 1600 1800 2000
Time, days
Fig. 44. Arps ratetime equations plotted on a (a) loglog plot and (b) semi-log plot.
the decline constant (Di) and the decline exponent (b), the latter referred to as b-stem value above in association with the discussion of
Fetkovich type-curves. Often Arps decline analysis is performed in parallel with Fetkovich type-curve analysis, the latter allowing identication of the start of boundary ow (after which the Arps curves are
applied), as well as the parameters (b etc.) to be used in the Arps forecast. Although the Arps decline curves were originally derived from empirical observation, it was later demonstrated by Fetkovich (1971) that
the solution to the ow equation for a well producing a slightly
compressible uid at constant owing pressure during boundarydominated ow has the same form as the Arps exponential (b = 0)
ratetime equation (Eq. (47)). Later still, Palacio and Blasingame
(1993) derived an analytical solution that exhibited the harmonic
form (b= 1, Eq. (49)) for the case of constant or variable owing pressures, if material balance time is used for slightly compressible uids
and material balance pseudotime for gas. As noted above, b = 1 corresponds to the constant-rate solution, so material balance time has the
effect of converting constant or variable owing pressure cases to the
equivalent constant-rate solution. An analytical solution matching the
hyperbolic form has not been derived.
Strictly speaking, the Arps curves were developed to analyze
boundary-dominated ow and are subject to the following conditions
(Lee and Wattenbarger, 1996), specically with respect to application
of Eq. (48) (hyperbolic decline equation):
1. The subject well for analysis is producing at a constant bottomhole
pressure. Changes in bottomhole pressure will affect the decline
characteristics of the well. In fact, the inability to isolate the inuence of the operating constraints of the well from the production
forecast is a major drawback of the Arps technique (Mattar and
Anderson, 2003), and empirical techniques in general, for both
conventional and unconventional reservoirs.
2. The subject well for analysis is producing from a xed drainage
area, with no-ow boundaries. During transient ow, the drainage
area of a well is constantly changing. Further, even during
boundary-dominated ow, the drainage area can change due to
relative changes in production rates in the eld, affecting the decline characteristics.
3. The reservoir has a constant permeability and skin. As discussed
above, effective permeability to gas can change due to multiple
factors in unconventional reservoirs, such as relative permeability
changes, stress-dependent permeability, desorption-dependent
permeability and non-Darcy ow (ex. slip-ow and diffusion). Further, skin changes may be observed during well cleanup
(Clarkson et al., in press).
4. Boundary-dominated ow of the well is being analyzed. As noted

by several authors, and discussed further below, t of the Arps hyperbolic equation to transient data results in b-stem values > 1,
which in turn will result in over-estimation of reserves if the
b-stem value is not adjusted to values b 1 later in the life of the
well, corresponding to boundary-dominated ow. As noted by Lee
and Wattenbarger (1996), if the data t with a decline curve
are transient, there is simply no basis for long-term performance.
Wells producing from unconventional gas reservoirs typically violate most of these assumptions. We will discuss application of the
Arps curves to CBM and tight/shale gas reservoirs below.
3.4.1. Application of Arps decline-curve methodology for CBM wells
The Arps decline curves have been applied to single-phase (gas
production only) and 2-phase (gas production and water production)
CBM wells historically, as summarized by Zuber (1996), and more recently by Seidle (2011). Seidle (2002) analyzed decline characteristics of simulated 2-phase CBM wells and derived decline coefcients
for gas and water that could be related to reservoir behavior; Mavor
et al. (2003) later extended the work to the analysis of Powder
River Basin Coal wells. Both authors noted that high-permeability
CBM wells often exhibit gas production rates that decline exponentially late in time.
CBM wells violate many of the conditions for Arps decline curve
analysis described above. Recently, Rushing et al. (2008) systematically investigated the application of Arps' decline curve methodology
to CBM wells, using simulated data. The impacts of both reservoir
(multi-layer effects, relative permeability, permeability anisotropy,
non-static permeability, isotherm parameters, gas content and sorption time) and operating conditions (different owing pressures) on
Arps decline exponent (b) were investigated. It should be noted
that the guidance for b-stem selection provided in Fig. 22 is for conventional oil and gas reservoirs and does not include some of the
unique attributes of CBM reservoirs, hence the importance of Rushing
et al.'s systematic investigations. The primary conclusions from that
work were:
1. Arps b-exponents derived from the t of long-term production data
(during boundary-dominated ow) ranged between 0.2 b b 0.8 and
fall within the limits dened by Arps (0b bb 1), suggesting that the
Arps approach works well for assessing CBM reserves.
2. The Arps b-exponents are not constant during production decline,
but appear close to 0 (exponential) for early time, and become
more hyperbolic (b > 0) later in the well life. Further, Rushing
To illustrate application of the Arps hyperbolic decline equation to

2-phase CBM data, Eq. (48) is t to the gas production data generated
with a simulator for Case 3 (Fig. 45). A b exponent > 0 and between
0.4 and 0.6 yields a reasonable match to the data. Following the guidance of Rushing et al. (2008) and Zuber (1996), the data were t
sometime after the peak gas production as the early time (after
peak) data look exponential (straight line on semi-log plot), which
could have resulted in too conservative a forecast. As noted above, a
Fetkovich type-curve match to late-time data could be used to estimate the Arps parameters for decline-curve forecasting.
3.4.2. Application of Arps decline-curve methodology for tight gas and
shale gas wells
A primary problem with the application of Arps decline-curves to
tight gas and shale gas wells is that long periods of transient ow and
transitional ow to true boundary-dominated ow may occur, during
which b > 1further, the b exponent often will not be constant during
transient ow (as assumed by Arps). Application of b > 1 to the entire
life of a well often will often lead to aggressive forecasts and overestimated reserves. Several authors have investigated how b changes
during transient ow by noting:
1 dq
q dt
50
d 1
dt D
51
where 1/D is the so-called loss ratio (Arps, 1945; Ilk et al., 2008b).
Note that D = constant and b= 0 for exponential decline (Eq. (47)).
An example of the use of these equations is provided below from
Kupchenko et al. (2008) for a simulated vertical hydraulicallyfractured well (innite conductivity planar fracture) completed in a
tight gas well within a homogeneous isotropic reservoir of nite and innite extent (Fig. 46). In the bounded reservoir, the b exponent changes
b)
Arps fit to 2-Phase CBM Data
1000
500
800
400
600
300
400
200
200
100
0
0
200
400
600
800
1000
0
1200
Arps fit to 2-Phase CBM Data
1000
Water Rate, STB/D
Gas Rate, Mscf/D
a)
from b = 2 during early linear ow to the fracture (period A in Fig. 46),

increases during the transition from linear to pseudo-radial ow
(period B), decreases during the transition to boundary-dominated
ow (period C), and then stays constant during boundary-dominated
ow (period D). The reciprocal decline rate (loss-ratio) was similarly
plotted versus time. The authors also investigated these changes for the
nite-conductivity fracture case (exhibiting bilinear ow), reservoirs
with permeability anisotropy etc. Importantly, during linear ow, bilinear ow, and boundary-dominated ow, the b exponent appears to be
constant and equal to 2, 4 and 0.5, respectively. During transitional
ow, the b exponent is not constant. Kupchenko et al. went on to propose a practical approach forecasting vertical wells in tight gas reservoirs: the b exponent is obtained by tting transient data (they
propose b = 2 as default) followed by estimation of the time at which
boundary-dominated ow is reached (using distance of investigation
calculation), which in turn depends on the drainage area and permeability. Once boundaries are reached, the decline-curve forecast is continued using b 0.5, the exact value of which depends on drawdown
(Fig. 22) and completion efciency. A two-segment decline curve approach was therefore suggested for tight gas wells, where the rst segment is used to approximate transient ow data, and the second is used
to model boundary-dominated ow.
Application of this two-segment approach is illustrated (Fig. 47)
using Case 2 (simulated shale gas well), which exhibits the two ow
regimes: linear followed by boundary-dominated ow. Comparing
Fig. 47 with Fig. 42, we see that the two-segment Arps decline-curve
approach provides a reasonable match to the simulated production
data with b = 2 up until the end of linear ow (time estimated from
linear distance of investigation calculation), followed b = 0.6 for
boundary dominated owb is greater than the expected maximum
of 0.5 for low-pressure gas produced from a single-layer reservoir
due to the effects of adsorption. The advantage of the analytical
model forecasting methodology described above however is that the
forecast can be tied directly to reservoir/hydraulic fracture properties
through rate-transient analysis and nal decline rates during boundary dominated ow are modeled correctly by incorporating uid
property variation and desorption. Further, this simple simulated
case assumes constant owing pressureas noted above, the decline
characteristics are affected by changes in operating conditions
(owing pressures), which can be honored explicitly in the analytical
model approach, but not in the empirical method, which can signicantly affect reserves estimates, as illustrated by Mattar and
Anderson (2003).
As with Arps analysis of CBM data, Rushing et al. (2007) systematically evaluated the application of Arps' hyperbolic decline equation
(Eq. (48)) to tight gas sand reservoirs at high pressure/temperature.
600
100
400
10
200
1
0
200
400
Time, days
Arpsqg
Actualqg
600
800
1000
Water Rate, STB/D
et al. (2008) note that application of exponential decline forecast

to early data may cause an underestimation in reserves.
3. None of the simulated cases exhibited long-term exponential
decline behavior, and CBM reservoir properties such as the adsorption isotherm, fracture permeability anisotropy, gaswater relative
permeability curves and stress-dependent permeability and porosity all affected production decline behavior.
4. Flowing pressure conditions had a strong impact on decline characteristics: higher bottomhole pressures caused longer-term hyperbolic behavior.
Gas Rate, Mscf/D
132
0
1200
Time, days
Actualqw
Arpsqg
Actualqg
Actualqw
Fig. 45. Application of Arps hyperbolic decline ratetime equation (Eq. (48)) to simulated 2-phase CBM example (Case 3) (a) Cartesian plot and (b) semi-log plot. Start of the Arps
model t to the simulated gas data is shown with an arrow.
133
Instantaneous Decline Exponent (b)
8
7
C)
B)
D)
Infinite Acting Reservoir

5
Infinite Acting Reservoir

A: Hydraulic Fracture Linear Flow
B, C: Transition Linear to Pseudo-Radial
D: Infinite Acting, Pseudo-Radial
4
3
2
Bounded Reservoir
A: Hydraulic Fracture Linear Flow
B: Transition Linear to Pseudo-Radial
C: Transition Infinite Acting to BDF
D: BDF (b < 0.5)
b = 0.5
Bounded Reservoir
A)
10
20
30
40
50
60
70
80
90
100
Time, years
Fig. 46. Arps b exponent changes with time for innite-acting and bounded tight gas reservoir.
Modied from Kupchenko et al. (2008).
Simulated cases representing a variety of reservoir and hydraulic fracture properties, as well as reservoir pressure and temperature conditions were generated and production proles matched with Eq. (48).
The primary conclusions from that work are re-stated below:
1. If the traditional (single-segment) Arps decline-curve methodology is applied to either transient or transitional ow, before the
onset of boundary-dominated ow, signicant errors in reserve estimations could occur.
2. Errors in reserve estimation were often b 10% using this methodology, if the well reached late transitional ow.
3. b exponents derived from long-term production data (i.e. 50 years),
ranged from 0.5 to 1.0 for the range of reservoir and hydraulic fracture properties investigated in the Rushing et al. (2007) study.
4. Hyperbolic decline behavior observed in the study is caused by a
variety of reservoir and hydraulic fracture heterogeneities.
3.4.3. New empirical approaches for tight gas and shale

Recognizing the issues with the traditional Arps decline-curve
methodology for tight gas/shale reservoirs, in particular the fact that
the b exponent is not constant but could change throughout the transient/transitional until boundaries are reached, various authors have
modied empirical techniques to account for transient ow. The
two new methods that are used most frequently today are the
Power-Law Exponential (PLE) and Stretched-Exponential (SE) approaches of Ilk et al. (2008b) and Valko (2009), respectively. Of these
two methods, the PLE approach will be discussed in more detail
below.
Using Eqs. (50) and (51) above, Ilk et al. (2008b) systematically
investigated the change in D and b for simulated and eld cases for
tight gas reservoirs and noted that D follows a decaying power-law
function as provided below:
D D D1 t
The Rushing et al. (2007) study underpins the need for a multisegmented Arps decline approach, where b exponents are switched
from > 1, during transient and transitional ow, to b1 during
boundary-dominated ow, to avoid signicant over-estimation of reserves. Although reserve estimation is beyond the scope of the
current review, the results of production analysis/rate-transient analysis are often used to support reserves estimation.
a)
Gas Rate, Mscf/D
Gas Rate, Mscf/D
1
Di
1
:
bt
1000
100
53
2-Segment Arps Decline-Curve
10000
100000
10000
52
which differs in functional form from D derived from the Arps hyperbolic decline equation (Eq. (48)), the latter shown by Blasingame and
Rushing (2005) to be as follows:
b)
2-Segment Arps Decline-Curve
1n
1000
100
10
1
10
100
1000
10000
Time, days
2-Segment Arps
100000
2000
4000
6000
8000
10000
Time, days
Data
2-Segment Arps
Data
Fig. 47. Application of two-segment Arps hyperbolic decline method to match numerically-simulated shale gas example (Case 1): (a) loglog plot and (b) semi-log plot.
which is equivalent in form to the traditional exponential decline relationship (Eq. (47)) of Arps. The model is therefore forced at late
time to follow the conservative (for gas) exponential relationship.
The time-dependence of the b exponent, using the PLE model is
(Currie et al., 2010):
Using Eq. (50), substituting in the denition of D in Eq. (52) and

integrating, the following ratetime relationship was derived:

q
D n
exp D t 1 t
n
q^ i
54
^ n1nt n
D
bt i
:
^ nt n 2
D t D
i
which was further reduced to:

h
i
^ tn :
q q^ i exp D tD
i
55
q q^ i expD t
b)
1.E+04
1.00
1.E+03
500
c)
1,000
1,500
Time, Days
2,000
0.00
2,500
1.E+04
1.E+03
1.E+02
1.E+01
1.E+02
d)
1.E+04
Power-Law Exponential
D - and b - parameters
1.E+01
1/Loss Ratio (D), 1/days
qg_MSCF/day_(PLE)
qg_MSCF/day_all
qg_MSCF/day_edited
9,000
1.E+03
Time, Days
Log-log Plot - q vs. Gp or Np
10,000
8,000
qg, MSCF/day
or qo, STB/day
qg_MSCF/day_(PLE)
qg_MSCF/day_all
qg_MSCF/day_edited
7,000
6,000
5,000
4,000
3,000
D_wo_Dinf
D_w_Dinf
D=1/q*dq/dt
D=dq/dGp
b_wo_Dinf
b_w_Dinf
b=d/dt*1/D
b=d/dt*1/D from Gp
1.E+00
1.E-01
1.E+03
1.E+02
1.E+01
1.E-02
1.E+00
1.E-03
1.E-01
1.E-04
1.E-02
2,000
Loss Ratio Derivative (b)
1.E+05
Gp, BSCF or Np, MMSTB
2.00
1.E+02
Log-log Plot - q vs. time
3.00
qg_MSCF/day_(PLE)
qg_MSCF/day_all
qg_MSCF/day_edited
Gp_BSCF_(PLE)
Gp_BSCF_edited
Gp_BSCF_all
1.E+05
58
To illustrate the behavior of the PLE model, the simulated

multi-fractured horizontal well tight gas example (Figs. 5b, 11, and
35) was analyzed with the PLE approach using software provided
by Texas A&M (Currie et al., 2010). Fig. 48d shows the b exponent
and D calculations from the data along with the model t (note that
the model-derived b and D were calculated with and without D
Cartesian Plot - q vs. time
1.E+06
qg, MSCF/day
or qo, STB/day

n
q q^ i exp t= :
56
a)
57
Ilk et al. (2008b) noted that the parameters ( q^ i and n) in the SE

equation of Valko (2009), given below, are interpreted in the same
way as those parameters in Eq. (55):
Eq. (55) has come to be known as the Power-Law Exponential decline model. q^ i is the rate-intercept obtained from Eq. (55) at t = 0,
^ i is the decline constant in
which differs from qi in Eq. (48), and D
Eq. (55) (D1 / n) and differs from Di in Eq. (48). Further D is the decline
constant at innite time.
From Eq. (55) we see the essential elements of the model. The
early-time transient data is dominated by the second term in the exponential, whereas at late times, the rst term dominates. At late
times, the equation collapses to:
qg, MSCF/day
or qo, STB/day
134
1,000
0
0.0
1.0
2.0
3.0
Gp, BSCF or Np, MMSTB
4.0
5.0
1.E-05
1.E+00
1.E+01
1.E+02
1.E+03
1.E+04
1.E-03
1.E+05
Time, Days
Fig. 48. Power-Law Exponential match to simulated MFHW data. (a) Cartesian plot of q versus t; (b) loglog plot of q versus t; (c) Cartesian plot of q versus cum. and (d) D and b
parameter plot.
a)
b)
Production Rates: Radial Model Forecast
Production Rates: Radial Model Forecast
1000
Gas Rate, Mscf/D
1000
Gas Rate, Mscf/D
135
100
100
10
10
1
10
100
1000
10000
100000
500
1000
Time, days
Data
1500
2000
Time, days
Data
Model
Model
Fig. 49. Radial ow analytical model match of numerically-simulated dry coal gas example: (a) loglog plot and (b) semi-log plot.
Modied from Clarkson (2009).
included). We see that the calculated D generally follows the

power-law decay (Fig. 48d), but is affected somewhat by changes in
the ow-regimes. The model-derived values of D are mostly consistent with the data. Including a non-zero D in the model causes the
predicted D to atten at late time, approaching a constant value as
the well transitions into boundary-dominated ow. The b-exponent
from the data (calculated using two different approaches as described
in Currie et al., 2010), generally decreases over time, but is at at a
constant value of 2 during the linear ow periods (Fig. 48d), as
expected. The model-calculated values generally decrease over time,
but do not show the constant values during linear ow. Again, inclusion of a non-zero D forces the b value to decrease more strongly at
late time as boundary-dominated ow is approached.
The other plots show the ratetime and ratecum. matches of the
PLE model, which are generally good.
A criticism that has been raised regarding the PLE method in its original form is that it contains 4 adjustable parameters, creating nonuniqueness in the matching process. Mattar and Moghadam (2009)
provided guidance on appropriate values of n to use, depending on
the ow-regimes encountered, which results in constraints that can
be applied to the model.
3.5. Hybrid methods
Hybrid methods refer to those that combine analytical methods for
analyzing and forecasting transient and transitional ow, with empirical
methods for analyzing and forecasting boundary-dominated ow. The

original hybrid technique for production analysis was provided by
Fetkovich (1980) in the form of the Fetkovich type-curves (Fig. 21),
where analytical solutions for (constant owing pressure) transient radial ow of liquids were combined with Arps' empirical decline-curves
for boundary-dominated ow. Permeability and skin values obtained
from the transient stems on the type-curve can be used to generate a
forecast for the transient radial portion of the data, while b, Di and qi information obtained from the depletion stem match can be used to generate a forecast for the boundary-dominated ow portion of the data
set. The time to reach boundary-dominated ow (tpss) can also be determined from the type-curve, which determines the point at which the
Arps decline-curve forecast begins. An example for the simulated dry
coal example (Fig. 23) is given below.
The model inputs for this example (Fig. 49) were derived from
type-curve analysis (of transient ow) and straight-line analysis of
transient radial ow and boundary-dominated ow (Clarkson,
2009). Although the boundary-dominated ow portion was modeled
analytically, using a material balance simulator that accounts for adsorption, a similar forecast using the Arps values derived from the
type-curve match is obtained (not shown).
An analogous hybrid approach was developed by Nobakht et al.
(2012c), referred to as the simple yet rigorous method, for analyzing the case where linear ow is followed directly by boundarydominated ow (Fig. 12). In this approach, assuming constant
owing pressure, the slope and intercept of the square-root of time
Planar Transverse Hydraulic Fractures
Horizontal Well
Fig. 50. Fracture geometry for a heterogeneous completion.
136
a)
plot are rst determined. A drainage area is then specied, allowing

calculation of the time at the end of linear ow (telf) as follows:
t elf

n
o32
2
Ah g ct mCP mpi m pwf
i
5
4
200:6T
End of Linear Flow Period 1

59
where mCP is the slope of the square-root of time plot for constant
owing-pressure conditions. Once telf is known, the gas production
rate at the end of linear ow is calculated from:
qelf
mCP
1
p
t elf b
60
where b is the intercept of the square-root time plot. The decline

constant at the end of linear ow can be calculated as follows:
Delf
1
m
:
pCP
p
mCP t elf b 2 t elf
b)
61
With these calculations completed, the forecast for linear ow

(t b telf) is generated using:
qelf
1
p
:
mCP t b
End of Linear Flow Period 2
62
And for t > telf:

qelf
qh

i1=b
1 bDelf tt elf
63
where b is the Arps decline exponent, which must be assumed, or

derived from the Fetkovich type-curve match. The appeal of this
method is that neither xf nor permeability needs to be assumed in
the procedurethe transient linear forecast is dependent on the
slope and intercept of the square-root of time plot. However, a drainage area must be assumed. As with the analytical forecasting method
above (see Fig. 42), forecast sensitivities can be performed using a
range in drainage areas. The difference with the Nobakht et al. (2012c)
approach and the analytical approach described above however, is that
the analytical method does not require an assumed b value as it relies
on material balance calculations during boundary-dominated ow. For
the Case 1 (simulated shale gas example), the two approaches yield
very similar forecasts.
The use of the Nobakht et al. (2012c) approach for multi-fractured
wells relies on the assumption of a homogeneous completion where
the hydraulic fractures are evenly spaced, extend to the reservoir
boundaries, and are all of the same length. This conceptual model is
shown in Fig. 43 and is referred to here as a homogeneous completion. It is possible (and more likely) that the hydraulic fractures
would be uneven in lengththis is depicted in Fig. 50 (for single porosity reservoir), and referred to as a heterogeneous completion.
A conceptual model for the heterogeneous completion case, originally introduced by Ambrose et al. (2011), is that an early linear ow
period between fractures occurs, but only portions of those fractures
interfere, followed by linear ow from the unbounded portions of the
fractures. This is illustrated with a simple three fracture case in
Fig. 51. The colored portions of the gure correspond to that region
of the stimulated reservoir that enters boundary-dominated ow
upon fracture interference. For example, after linear ow period 1
(Fig. 51a), the red portion of the SRV enters boundary-dominated.
The unbounded portions of the fractures continue in linear ow
during linear ow period 2, until additional fracture segments interfere (Fig. 51b), after which unbounded portions of the fractures continue in linear ow during linear ow period 3. In this simple
example, true boundary-dominated ow within the SRV does not
c)
End Linear Flow Period 3
Fig. 51. Illustration of multiple linear ow periods between fractures for simple
3-fracture heterogeneous completion case. (a), (b) and (c) represent the ends of linear
ow periods 1, 2 and 3, respectively.
occur until linear ow period 3 ends, unlike the homogeneous completion case, which enters boundary-dominated ow immediately
after the rst linear ow period. The homogeneous completion
model would therefore be expected to yield a more conservative forecast for the case in Fig. 51.
Nobakht et al. (2011) developed a hybrid model for the heterogeneous completion case as follows:
2
r
X
j1
qj
r
X
j1
h

i

1
1
6 1
7
p 1U tt elfj
p h
j 4

i1=b U tt elfj 5
mCP t elfj 1 bD
mCP t
tt
elfj
elfj
64
137
10,000
Gas Rate, Mscf/D
1,000
100
10
1
1
10
Data
100
1,000
Time,Forecast
days with b=0.8
Homogeneous
Heterogeneous Forecast
10,000
Homogeneous Forecast with b=0.5
Fig. 52. Comparison between homogeneous completion forecast (assuming b = 0.5) forecast, and heterogeneous completion forecast, which is similar to a homogeneous completion forecast with b= 0.8.
where U is the unit step function, and j is a measure of the surface

area to ow in linear ow period j to total surface area to ow. Note
that this again qualies as a hybrid method as the slope of the
square-root of time plot is used to model linear ow periods and
the Arps hyperbolic decline equation is used to model boundarydominated ow.
A eld case (shale well) was modeled in Nobakht et al. (2011) and
reproduced in Fig. 52. This gure compares the forecasts of the homogeneous completion model, which assumes that boundary-dominated ow occurs immediately after the rst linear ow period (i.e.
after initial fracture interference is seen), and the heterogeneous
completion model with seven linear ow periods. The heterogeneous
model forecast, which is similar to an Arps hyperbolic decline forecast
with b = 0.8, is more aggressive than the homogeneous completion
case, which assumes b = 0.5. The heterogeneous completion model
of Nobakht et al. (2011) is currently being updated to account for
complex fracture geometries.
400
80
300
60
200
40
100
20
0
100
200
300
400
0
500
Time, days
Modelqg
Actualqg
Actualqw
Tight gas vertical and multi-fractured horizontal well (MFHW)

examples in Clarkson and Beierle (2011), including individual fracture stage analysis for the MFHW
Dry CBM vertical well examples in Clarkson (2009)
Wet CBM horizontal well example in Clarkson et al. (2012c)
Wet CBM vertical well example with absolute permeability growth
in Clarkson et al. (2009)
d(krg*m(p)/q)/dlnt*ca
100
Until now, the production analysis methods described above have

been illustrated with mostly simulated examples. The following section is designed to illustrate the practical application of the methods
illustrated above. A two-phase CBM vertical well example is shown,
along with a multi-fractured horizontal (dry) shale well example.
Additional eld examples analyzed by the author are included in
the following publications:
b)
500
Water Rate, STB/D
Gas Rate, Mscf/D
a)
4. Field examples

1.0E+05
1.0E+04
Radial Flow
(zero - slope)
1.0E+03
1.0E+02
1.0E+00
1.0E+01
1.0E+02
t*ca, days
Modelqw
Fig. 53. Analytical model history-match (a) and semi-log derivative (b) of 2-phase (gas + water) CBM vertical well. Gas production data is used in the derivative.
Wet shale vertical well example in Mattar et al. (2008)

Dry shale (MFHW) in Clarkson (2011)
Dry shale (MFHW) example with stress-dependent matrix and fracture permeability in Clarkson et al. (2012e)
Flow-back (gas + water) analysis of MFHW shale wells in Clarkson
(2012).
In an accompanying teaching aid article (Clarkson, 2013-this
volume), the workow illustrated below for the (dry) shale case
will be discussed in detail.
a)
Fetkovich Type -Curves
10.00
1.00
qDd
138
0.10
4.1. Field example 1: 2-phase CBM well (vertical)

0.01
0.0001
2-Phase Radial Flow Plot
0.01
0.1
10
tDd
b)
Pratikno-Blasingame Type -Curves
10.00
1.00
0.10
0.01
0.0001
a)
0.001
qDd
This example is reproduced from Clarkson et al. (2012c) and illustrates the use of straight-line, type-curve and analytical simulation
methods that incorporate the effects of 2-phase ow (gas+ water)
and adsorption. For 2-phase CBM, it is recommended that saturationand pressure-dependent variables used in straight-line and type-curve
methods be linked to output from analytical (or numerical) simulation
to improve consistency in the analysis, unless relative permeability information is derived independently with condence. This workow
was demonstrated with simulated Cases 3 and 4.
The example (Figs. 5355) is a hydraulically-fractured vertical
well from the Uinta Basin, U.S. Reservoir, operations and wellbore
input are provided in Clarkson et al. (2012c). A reinitialization of
the data set occurred after an operational upset, possibly due to a
pump changeout. The rst step in the analysis is to input any
known (hard) data (coal thickness and bulk density, initial pressure
0.001
0.01
0.1
10
tDd
1.0E+05
Fig. 55. Match of 2-phase CBM gas data to (a) Fetkovich and (b) PratiknoBlasingame
type-curves.
8.0E+04
6.0E+04
4.0E+04
2.0E+04
0.0E+00
0
Superposition Time
b)
2-Phase Flowing Material Balance Plot
5.0E-02
4.0E-02
3.0E-02
2.0E-02
1.0E-02
0.0E+00
0
200
400
600
800
1000
1200

Fig. 54. Radial ow analysis (a) and owing material balance analysis (b) of 2-phase
(gas +water) CBM vertical well.
and other PVT information, wellhead owing pressures and wellbore

information, adsorption isotherm) into the analytical simulator, then
adjust unknowns (soft data) to achieve a history-match (Fig. 53a).
The primary parameters requiring adjustment were absolute permeability, skin, drainage area and relative permeability. Preliminary estimates of k and s were based on independent information provided by
the operator from well tests, and little adjustment was required.
Drainage area limits were imposed based on well spacing. Relative
permeability curves are the most uncertain matching parameters; initial estimates were based on the curves presented by Mavor (1996)
and required only minor adjustment to achieve the match. The nal
match of the gas production and water production rates is shown in
Fig. 53a, but cumulative gas production and water production were
also matched (Clarkson et al., 2012c). The values obtained from the
history-match were: k = 10.0 md, s = 1.0, OGIP = 1140 MMscf,
and A = 180 ac.
Flow-regimes were identied using a semi-log derivative (Fig. 53b),
utilizing saturation- and pressure-dependent output from the simulator for relative permeability and pseudovariable calculations. The
dominant early time ow regime appears to be radial owlate
time data is assumed to approach boundary-dominated ow (peak
gas production in Fig. 53a), but the derivative at late time is very
noisy.
With the ow-regimes established, straight-line analysis was
performed (Fig. 54), including radial ow analysis of early data to obtain
permeability and skin, and owing material balance analysis of
late-time data to obtain OGIP and A. As with the derivative in Fig. 53b,
pressure- and saturation-dependent variables in the straight-line
analysis utilized pressures and saturations obtained from the analytical

model output after history-match. Radial ow analysis (Fig. 54a)
yielded k = 10.1 and s =2.1, which are in good agreement with
simulation-derived values, although skin is somewhat higher due to uncertainty in drawing the straight line through noisy data. The OGIP from
FMB (~1150 MMscf) is also in good agreement (Fig. 54b).
The Fetkovich (1980) type-curves were selected for analysis
(Fig. 55a) because of the small skin (slightly stimulated well) and
exhibited radial ow period; the PratiknoBlasingame type-curves
for hydraulically-fractured vertical wells were also selected
(Fig. 55b). The type-curve matches yield permeability and skin values
consistent with the simulator history-match derived values. We note
again that pressures and saturation output from the simulator were
used for dimensionless variable calculations.
This example clearly illustrates the need for integrated ratetransient analysis to arrive at a consistent solution. Empirical forecasting was not attempted for this example because no depletion
data were available to match.
4.2. Field example 2: 1-phase shale gas well (MFHW)

The dry shale example is a multi-fractured horizontal well producing from the Barnett Shale.
Details of the well completions and reservoir properties are withheld to preserve operator condentiality. The well was completed in
6 stages, with 5 perforation clusters per stage (total 30 clusters),
using the slickwater hydraulic fracturing technology. The heel to
toe perforated length of the well is 2334 ft.
Gas production and water production data (Fig. 56a) for this well
illustrate that the post-fracture stimulation cleanup period is signicant, with water production rates declining over a multi-week period.
Flowing pressures are also rapidly declining during this period.
Although single-phase analysis of gas production data is performed
in this study, a method for analyzing post-frac gas and water
owback data is discussed in Clarkson (2012).
A loglog diagnostic plot of gas-normalized pseudo-pressure difference [m(pi) m(pwf)] / qg versus time (Fig. 56b), which is useful
for ow-regime identication with noisy production data, suggests
that an early-linear ow period is followed by a departure from linear
ow, possibly due to fracture interference. This interpretation is
supported with the use of derivative analysis (Fig. 57), although the
derivatives tend to amplify the noise in the data.
The data were analyzed by assuming a bounded reservoir, corresponding to the stimulated reservoir volume, where transient linear
ow to the hydraulic fractures is followed immediately by fracture interference, followed by between-fracture depletion (Fig. 30). The
dashboard of analyses for this interpretation is given in Fig. 58,
which illustrates straight-line and type-curve analysis. Because
there is no radial or elliptical ow period to analyze to derive a permeability estimate, the time at which fracture interference occurs,
combined with the fracture spacing (actually half-spacing), is used
in the linear distance of investigation calculation (Eq. (21)) to solve
for permeability. The fracture half-spacing is assumed to correspond
to the 1/2 spacing between perforation clusters (~ 40 ft), and the
time at which fracture interference occurs is 256 days (see
Fig. 58b). The calculated permeability is ~ 100 nD, which is consistent
a)
1800
1000
1200
100
600
10
1
10
0
1000
100
Time, days
Gas Rate
Water Rate
Pwf
Frac
Interference?
d(m(p)/q)/dlnt
Production Rates and Flowing Pressures

10000
Flowing Bottomhole Prressure, psia
1.0E+07
a)
Gas (Mscf/D) or Water (STB/D)
139
1.0E+06
Linear flow
1.0E+05
1.0E+04
1.0E+00
1.0E+01
1.0E+02
1.0E+03
Time, days
b)
Linear Derivative Plot

1.0E+06
b)
Log-Log Diagnostic Plot
1.0E+06
d(m(p)/q)/dt^0.5
m(p)/qg
1.0E+07
Linear flow
1.0E+05
1.0E+04
1.0E+00
Linear
flow
1.0E+05
Frac
Interference?
1.0E+04
1.0E+00
1.0E+01
1.0E+02
1.0E+03
1.0E+04
1.0E+01
1.0E+02
1.0E+03
Time, days
Time, days
Fig. 56. Production rates and owing pressures (a) and loglog diagnostic plot (using
gas data) (b) for MFHW completed in dry shale.
Fig. 57. Identication of ow-regimes for MFHW dry shale case. On semi-log (radial)
derivative, linear ow appears as 1/2 slope and on the linear derivative (b), linear
ow appears as a zero slope. Real time was used in this example.
140
b)
1.0E+06
1.2E+06
8.0E+05
1.0E+06
[m(pi)-m(pwf)]/qg
[m(pi)-m(pwf)]/qg
a)
6.0E+05
4.0E+05
2.0E+05
8.0E+05
6.0E+05
4.0E+05
2.0E+05
0.0E+00
0.0E+00
0
50
100
150
Superposition Time
d)
15
20
25
30
Linear Flow Type-Curves

100
1.0E-02
8.0E-03
10
(ye/xe) qD
Normalized Rate, scf/D/psi2/cp
c)
10
Sqrt Time, (days)^0.5
6.0E-03
4.0E-03
0.1
2.0E-03
0.0E+00
0
500
1000
1500
2000
0.01
0.0001
0.001
0.01
0.1
10
100
tDye
Fig. 58. Linear ow analysis using (a) early linear ow plot and (b) square-root time plot and (apparent), boundary-dominated ow analysis using owing material balance plot
(c) and type-curve match using Wattenbarger type-curves (d). End of linear ow is indicated on the square-root time plot as a vertical green dashed line.
with matrix permeability estimates cited for the Barnett Shale (ex.
Samandarli et al., 2012). Permeability is assumed to be static in this
example; the impact of non-Darcy ow on the analysis, as well as
geomechanical effects, is discussed in Clarkson et al. (2012b).
Using the estimate of permeability, the linear ow plot (Fig. 58a)
and square-root of time plot (Fig. 58b) can be used to derive total
fracture half-length (xf), or contacted matrix surface area (Acm).
Note that pseudotime in this example for linear ow analysis uses
the pressure in the region of inuence (Nobakht and Clarkson,
2012b) and is corrected for desorption (Eq. (4), an adsorption isotherm was available for this well). Note further that the superposition
time function used is linear superposition pseudotime. For the
square-root time analysis, the fcp correction is that of Eq. (16),
which includes the effects of adsorption. There is less scatter in the
linear ow plot due to the use of superpositionnote that Liang
et al. (2011) recommend the use of material balance time with the
square-root of time plot to reduce scatter.
The calculated total xf from the linear ow plot and square-root
time plot are 2415 ft, and 2459 ft, respectively, indicating good agreement between the two analysis approaches. If it is assumed that one
hydraulic fracture was created per perforation cluster, then the average fracture half-length per cluster (30 perf clusters) is 8182 ft. The
calculated Acm values are 2.705 10 6 ft 2 and 2.755 10 6 ft 2it is uncertain whether planar or complex fractures were created in this case
because microseismic and other surveillance data were unavailable.
Flowing material balance analysis was used to provide a minimum
OGIP for the well assuming that fracture interference observed in
Figs. 5657 will be followed immediately by depletion of the SRV
and that there will be no contribution of gas from outside the SRV.
These are conservative assumptions, as will be discussed further
below. The calculated OGIP is ~ 1.5 bscf, corresponding to a drainage
area (given volumetric inputs) of 8 ac. To check the reasonableness
of this estimate, an area corresponding to the SRV can be estimated
by using a) the toe to heel perforated length of the well (SRV length)
and b) the individual fracture half-length from linear ow analysis
(~ 81 ft) (see Fig. 43 for assumed geometry). This results in an area
of ~ 8.5 ac, which is slightly larger than FMB analysisalthough this
is believed to be reasonable agreement, it should be noted that FMB
analysis will generally lead to conservative estimates of contacted
gas-in-place, as discussed in Nobakht and Clarkson (2011).
Type-curve analysis, using the bounded linear type-curves of
Wattenbarger et al. (1998) veries the interpretation of total hydraulic fracture half-length from straight-line linear ow analysis and
OGIP from FMB analysis (Fig. 58d). Corrections to the data for skin
effects were not required in this case.
Using the analytical forecasting approach described above for a
homogeneous completion (transient linear ow followed directly by
boundary-dominated ow), with material balance calculations for
boundary-dominated ow adjusted for desorption, and the SRV
drainage area estimate from FMB analysis, a conservative forecast is
generated for the well (Fig. 59).
There is uncertainty associated with the forecast because it is
uncertain whether true boundaries are reached. If fracture interference is being seen, it is unclear whether this corresponds to interference of all perf clusters simultaneously, as in the homogeneous
completion model (Fig. 43), or just the end of linear ow period 1
a)
Production Forecast
Production Forecast
10000
Gas Rate, Mscf/D
Gas Rate, Mscf/D
a) 10000
1000
100
141
1000
100
10
100
1000
10000
10
10
Time, days
Actual Gas Rate
100
1000
10000
1000
1000
100
0
200
400
600
800
Model Gas Rate
Production Forecast
10000
10000
1000
1000
100
1000
Time, days
Actual Gas Rate
Actual Gas Cum.
b)
Gas Rate, Mscf/D
Gas Rate, Mscf/D
Production Forecast
Cumulative Gas Production, MMscf
Actual Gas Rate
b)10000
10000
Time, days
Model Gas Rate
100
Model Gas Rate
Actual Gas Cum.
Fig. 59. Gas production rate match with analytical model on loglog plot (a) and gas
production rate and cumulative gas production match with analytical model on
semi-log plot. A conservative estimate of the drainage area based upon the owing material balance plot (Fig. 58c) was used to generate the forecast.
in the heterogeneous completion model (Fig. 50). Further, it is

unclear whether true fracture interference is seen, or just operational
upsets, perhaps due to liquid-loading effects. A more aggressive forecast can be obtained by assuming that a) one fracture per stage was
created and that b) true depletion of between-fracture stimulated
reservoir volume is not yet seen. Using linear ow analysis, the
interpreted individual stage xf would be ~ 403 ft, resulting in an estimated SRV area of ~ 43 ac. The new, more aggressive but still reasonable forecast is shown in Fig. 60. If additional independent
information about the SRV can obtained, such as from microseismic
data, additional forecast sensitivities can be performed. Finally, the
actual well-spacing can be used as a high-side estimate.
5. Discussion: future development of production data analysis
techniques
Although great strides have been made in the development of production data analysis techniques for unconventional reservoirs, there
is still much room for improvement. The following is a summary of
the author's view of required future development areas for production analysis techniques.
5.1. Analytical (type-curves, straight-line, analytical simulation)
Further modications to include complex reservoir behaviour, such
as non-Darcy ow (slip-ow and diffusion) and non-static permeability (i.e. Ozkan et al., 2010)
200
400
600
800
100
1000
Cumulative Gas Production, MMscf
10
Time, days
Actual Gas Rate
Actual Gas Cum.
Model Gas Rate

Actual Gas Cum.
Fig. 60. Gas production rate match with analytical model on loglog plot (a) and gas
production rate and cumulative gas production match with analytical model on
semi-log plot. A more aggressive estimate of the drainage area was used to generate
the forecast.
Further modications to account for complex fracture geometries

(i.e. Apaydin et al., 2012)
Incorporation of uid properties contained within nano-porosity,
such as the alteration of critical properties (i.e. Michel et al.,
2012), and incorporation of complex adsorption behavior of heavier
hydrocarbon components of wet gas and gas condensate
Continued development of techniques to account for multi-phase
ow (CBM, gas condensate), such as alteration of pseudovariables
or inclusion of dynamic skin (Clarkson et al., in press)
Integration of rate-transient techniques with surveillance data, such as
microseismic data, chemical and radioactive tracers, distributed temperature surveys etc. to improve the extraction of stage-specic (in
multi-frac'd wells) or layer-specic properties (for wells completed
in multiple zones)
Adaptation of methods to analyze multi-phase, post-stimulation
ow-back data (i.e. Clarkson, 2012), to enable early forecasts
Improved ow-regime identication for wells with noisy data.
5.2. Empirical
Investigation of further constraints to matching parameters by relating them to actual reservoir and hydraulic fracture propertiesex.
Mattar and Moghadam (2009) work with Power-Law Exponential
Continued development of diagnostic methods using empirical approaches to diagnose reservoir behavior and operational problems.
142
5.3. Hybrid
Advancement of these techniques will rely on advancements in
both analytical and empirical techniques, but an immediate development need includes adaptation of these techniques for complex
fracturingcurrent techniques currently applicable mainly to planar fracture case
6. Conclusions
Quantitative production data analysis for unconventional gas reservoirs is a rapidly evolving eld, and new techniques are continuously being introduced. There are no-agreed upon standardized
procedures for deriving reservoir/hydraulic-fracture properties from
production data, or for forecasting wells producing from unconventional reservoirs.
Critical advancements, however, have been made for analytical
methods, such as straight-line (ow-regime) analysis, type-curve analysis, and analytical simulation, in particular the incorporation of reservoir properties unique to unconventional gas reservoirs, such as
adsorption, non-Darcy ow (gas slippage, diffusion), non-static absolute permeability, and multi-phase ow. Adaptations have also been
made to analyze wells exhibiting complex hydraulic fracture geometries such as multi-fractured horizontal wells completed in shale reservoirs. Empirical methods have been altered to account for long transient
and transitional ow periods encountered in some unconventional gas
reservoirshistorical methods were applicable to boundary-dominated
ow only. A new class of techniques, referred to as hybrid techniques
that combine analytical methods for analyzing/forecasting transient
ow, with empirical methods for analyzing/forecasting boundarydominated ow have also been modied recently to account for
common transient and transitional ow-regimes such as long-term
transient linear ow and mixed linear-boundary ow (heterogeneous
completions).
Continuous developments are being made in this eld, which remains one of the most active areas of research in the eld of unconventional gas reservoir engineering. Those practicing reservoir
engineers and geoscientists working in this eld are advised to consult the literature for further advances and best practicesthe value
and future of many oil & gas companies will depend directly on
how well these techniques are used for forecasting production and
optimizing eld development.
Nomenclature
Abbreviations
B-D
boundary-dominated
CBM
coalbed methane
EUR
estimated ultimate recovery
MBE
material balance equation
MFHW multi-fracture horizontal well
OGIP
original gas in place
PDA
production data analysis
PLE
PTA
pressure-transient analysis
RNP
rate-normalized pressure
RNP
rate-normalized pressure derivative
RTA
rate-transient analysis
SE
Stretched-Exponential
SG
shale gas
SRV
stimulated reservoir volume
Field variables
A
drainage area, ac, or major semi-axis of ellipse (Table 1), ft 2
Acm
matrix cross sectional area to ow, ft 2
B
minor semi-axis of ellipse, ft
Bg
Bw
gas formation volume factor, reservoir volume to surface volume

water formation volume factor, reservoir volume to surface
volume
b
Arps' decline exponent, dimensionless
b
y-axis intercept of straight line on specialty plot (Table 1)
formation volume factor, rcf/Mscf
Bg
total compressibility, psi 1
ct

ct
total compressibility, evaluated at average reservoir pressure,
psi1
ct
total compressibility including desorption effects, psi 1
D
Arps' loss ratio, D 1
Delf
decline rate at the end of linear ow, D 1
Di
Arps' initial decline rate (hyperbolic model), D 1
D1
decline constant intercept 1 time unit dened by Eq. (52),
D 1
D
decline constant innite time dened by Eq. (52), D 1
^i
decline constant dened by Eq. (55), D 1
D
fcp
drawdown correction for constant owing pressure case,
dimensionless
drawdown correction for constant owing pressure case,

fcp
accounting for desorption, dimensionless
fracture conductivity, md-ft
Fc
original gas-in-place, MMscf or Mscf
Gi
cumulative gas production, MMscf or Mscf
Gp
h
formation thickness, ft
k
absolute (formation) permeability, md
effective permeability to gas, md
kg
matrix permeability, md
km
relative permeability to gas, dimensionless
krg
relative permeability to water, dimensionless
krw
k p
absolute permeability at the pore volume average pressure
of the reservoir, md
absolute permeability at initial pressure, md
ki
ka p
apparent permeability at the pore volume average pressure
of the reservoir, md
liquid-equivalent permeability, md
k
characteristic dimension, ft
Lc
fracture spacing, ft
LF
m
slope of a straight line on specialty plot (Table 1)
m(p)
real gas pseudopressure, psi 2/cp
m(p)MP real gas pseudopressure accounting for multi-phase ow,
psi 2/cp
m(p)NSP real gas pseudopressure accounting for non-static permeability, psi 2/cp
m(p)NDF real gas pseudopressure accounting for non-Darcy ow,
psi 2/cp
M(p)
multi-phase ow potential, psi/cp
n
time exponent given by Eq. (52), dimensionless
p
pressure, psia
initial reservoir pressure, psia
pi
owing bottomhole pressure, psia
pwf
gas production (surface) ow rate at the end of linear ow,
qelf
Mscf/D
qg (or q) gas production (surface) ow rate, Mscf/D
gas initial production rate, Mscf/D
qgi
gas production rate during boundary-dominated ow,
qgBD
Mscf/D
gas production rate at start of boundary-dominated ow,
qgBDi
Mscf/D
rate intercept dened by Eq. (55), Mscf/D
q^ i
gas rate of division j for heterogeneous completion, Mscf/D
qj
modied pseudopressure-normalized rate
qm
drainage radius, ft
re
wellbore radius, ft
rw
apparent wellbore radius (rwa = rw e s), ft
rwa
s
skin factor, dimensionless
Sg
Sw
t
T
ta,BLS
ta,LS
ta,RS
ta,MP
ta,NSP
ta,NDF
tc
tca
tehs
telf
telfj
ta
tca
U
wf
xe
xf
y
ye
yeD
z
z*
gas saturation, dimensionless

water saturation, dimensionless
time, days or hours
reservoir temperature, F
bilinear superposition pseudotime, h
linear superposition pseudotime, h
radial superposition pseudotime, h
pseudotime accounting for multi-phase ow, h
pseudotime accounting for non-static permeability, h
pseudotime accounting for non-Darcy ow, h
material balance time, days
material balance pseudotime, days
time at the end of the half-slope for linear ow, days
time at the end of ow, days
duration of linear ow for division j in a heterogeneous
completion, days
pseudotime, accounting for desorption, days
material balance pseudotime, accounting for desorption, days
unit step function
fracture width, ft
reservoir half-width (Fig. 37), or full-width (Fig. 14),
rectangular reservoir geometry, ft
fracture half length, ft
distance of investigation for linear ow, ft
half of reservoir length (Fig. 37), or full-width (Fig. 14),
rectangular reservoir geometry, ft
dimensionless reservoir length, dimensionless
gas deviation factor, dimensionless
gas deviation factor, adjusted to account for desorption
effects, dimensionless
Dimensionless variables
dimensionless pseudosteady-state constant
bDpss
dimensionless fracture conductivity
FcD
dimensionless pressure
pwD
dimensionless rate
qD
dimensionless decline rate
qDd
dimensionless decline rate integral
qDdi
dimensionless decline rate integral-derivative
qDdid
dimensionless rate dened in Eq. (42)
qDM
dimensionless wellbore radius; (reD = re / rw) for radial ow
reD
type-curves, and (reD = re / xf) for hydraulically-fractured
well type-curves
dimensionless time
tD
dimensionless decline time
tDd
dimensionless time based on A
tDA
dimensionless time dened in Eq. (43)
qDM
dimensionless time based on xf
tDxf
tDye
Dimensionless time based on ye
Greek variables
ratio of drainage area of division j in a heterogeneous comj
pletion to the total drainage area, fraction
parameter for heterogeneous completion
j
gas viscosity, cp
g
g
gas viscosity, evaluated at average reservoir pressure, cp
water viscosity, cp
w
porosity, fraction
permeability modulus, psi 1

j
product of j and j, which is a measure of surface area to
ow in division j in a heterogeneous completion to total
surface area to ow, fraction
interporosity ow coefcient, dimensionless
storativity ratio, dimensionless
characteristic time parameter for stretched-exponential

decline model (Eq. (58))
143
Subscripts
a
pseudo
c
material balance
ca
material balance pseudo
cp
constant owing pressure case
D
dimensionless variable
Dd
dimensionless decline variable
g
gas
i
initial
BL
bilinear
E
elliptical
L
linear
m
matrix
NDP
non-static permeability
NDF
non-Darcy ow
R
radial
sc
standard conditions
w
water
wa
apparent wellbore
wf
sandface
Superscripts
_
average property
*
altered variable
Units conversion
1 psi
6.895 kPa
1/psi
0.145/kPa
Acknowledgments
Chris Clarkson would like to acknowledge Encana and AITF for
support of his Chair position in Unconventional Gas and Light Oil at
the University of Calgary, Department of Geoscience. He would like
to thank his students, Morteza Nobakht, Farhad Qanbari, Hamid
Behmanesh and Jesse Williams-Kovacs for their great contributions
to research in this difcult eld. Clarkson would also like to thank
his mentors in the eld of reservoir engineering: Tom Blasingame,
Louis Mattar, Mike McGovern and Dave Reeseyou continue to provide inspiration for me. He would like to acknowledge the great
contributions of these people, and their students and colleagues, to
the eld of production analysis: Mike Fetkovich, John Lee, Tom
Blasingame, Bob Wattenbarger, Ram Agarwal, Raj Raghavan, Bobby
Poe, Peter Valko, Michael Economides, Christine Ehlig-Economides,
Erdal Ozkan, Louis Mattar, and Dave Anderson. Finally, Clarkson
would like to acknowledge both Fekete and KAPPA for donation of
their software to his research program.
References
Abdulal, H.J., Samandarli, O., Wattenbarger, R.A., 2011. New type-curves for shale gas
wells with dual porosity model. Society of Petroleum Engineers Paper 149367
Presented at the Canadian Unconventional Resources Conference, Calgary, Alberta.
Agarwal, R.G., Gardner, D.C., Kleinsteiber, S.W., Fussell, D.D., 1999. Analyzing well
production data using combined type curve and decline curve concepts. Society
of Petroleum Engineers Reservoir Evaluation & Engineering 2 (5), 478486.
Al-Ahmadi, H.A., Wattenbarger, R.A., 2011. Triple-porosity models: one further step
towards capturing fractured reservoirs heterogeneity. Society of Petroleum Engineers Paper 149054 Presented at the Saudi Arabia Section Technical Symposium
and Exhibition, Al-Khobar, Saudi Arabia.
Ambrose, R.J., Clarkson, C.R., Youngblood, J., Adams, R., Nguyen, P., Nobakht, M., Biseda,
B., 2011. Life-cycle decline curve estimation for tight/shale gas reservoirs. Society
of Petroleum Engineers paper 140519 Presented at the SPE Hydraulic Fracturing
Technology Conference and Exhibition, The Woodlands, Texas.
Amini, S., Ilk, D., Blasingame, T.A., 2007. Evaluation of the elliptical ow period for
hydraulically-fractured wells in tight gas sandstheoretical aspects and practical
considerations. Society of Petroleum Engineers paper 106308 Presented at the
2007 SPE Hydraulic Fracturing Technology Conference, College, Station, Texas.
144
Aminian, K., Ameri, S., 2009. Predicting production performance of CBM reservoirs.
Journal of Natural Gas Science and Engineering 1, 2530.
Anderson, D.M., Liang, P., Okouma, V., 2012. Probabilistic forecasting of unconventional
resources using rate-transient analysis: Case studies. Society of Petroleum Engineers Paper 155737 Presented at the SPE Americas Unconventional Resource
Conference, Pittsburgh, Pennsylvania.
Andrade, J., Civan, F., Devegowda, D., Sigal, R.F., 2011. Design and examination of requirements for a rigorous shale-gas reservoir simulator compared to current
shale-gas simulator. Society of Petroleum Engineers Paper 144401 Presented at
the North American Unconventional Gas Conference, The Woodlands, Texas.
Apaydin, O.G., Ozkan, E., Raghavan, R., 2012. Effect of discontinuous microfractures on
ultratight matrix permeability of a dual-porosity medium. Society of Petroleum
Engineers Reservoir Evaluation & Engineering 15 (4), 473485.
Araya, A., Ozkan, E., 2002. An account of decline-type-curve of vertical, fractured, and
horizontal well production. Society of Petroleum Engineers Paper 77690 Presented
at the SPE Annual Technical Conference and Exhibition, San Antonio, Texas.
Arps, J.J., 1945. Analysis of decline curves. Transactions of AIME 168, 228247.
Bello, R.O., 2009. Rate transient analysis in shale gas reservoirs with transient linear
behavior. PhD Thesis, Texas A&M University.
Bello, R.O., Wattenbarger, R.A., 2008. Rate transient analysis in naturally fractured shale
gas reservoirs. Society of Petroleum Engineers Paper 114591 Presented at the
CIPC/SPE Gas Technology Symposium, Calgary, Alberta.
Bello, R.O., Wattenbarger, R.A., 2009. Modelling and analysis of shale gas production with
a skin effect. Paper 2009-82 Presented at the Canadian International Petroleum
Conference, Calgary, Alberta.
Bello, R.O., Wattenbarger, R.A., 2010. Multi-stage hydraulically fractured shale gas ratetransient analysis. Society of Petroleum Engineers Paper 126754 Presented at the
SPE North Africa Technical Conference, Cairo, Egypt.
Blasingame, T.A., Rushing, J.A., 2005. A production-based method for direct estimation
of gas-in-place and reserves. Society of Petroleum Engineers Paper 98042,
Presented at the 2005 SPE Eastern Regional Meeting, Morgantown, West Virginia.
Bourdet, D., Whittle, T.M., Douglas, A.A., Pirard, Y.M., 1983. A new set of type curves
simplies well test analysis. World Oil 95106 (May 1983).
Bourdet, D., Ayoub, J.A., Pirard, Y.M., 1989. Use of pressure derivative in well-test interpretation. Society of Petroleum Engineers Formation Evaluation 4 (2), 293302.
Brown, M., Ozkan, E., Raghavan, R., Kazemi, H., 2011. Practical solutions for pressuretransient responses of fractured horizontal wells in unconventional shale reservoirs.
Society of Petroleum Engineers Reservoir Evaluation & Engineering 14 (6), 663676.
Bumb, A.C., McKee, C.R., 1988. Gas-well testing in the presence of desorption for
coalbed methane and Devonian shale. Society of Petroleum Engineers Formation
Evaluation 3 (1), 179185.
Carlson, E.S., Mercer, J.C., 1991. Devonian shale gas production: mechanisms and simple
models. Journal of Petroleum Technology 4 (43), 476482.
Chen, C.C., Raghavan, R., 1997. A multiply-fractured horizontal well in a rectangular
drainage region. Society of Petroleum Engineers Journal 2 (4), 455465.
Cheng, Y., 2011. Pressure transient characteristics of hydraulically fractured horizontal
shale gas wells. Society of Petroleum Engineers Paper 149311 Presented at the SPE
Eastern Regional Conference, Columbus, Ohio.
Cheng, Y., Lee, W.J., McVay, D.A., 2009. A new approach for reliable estimation of
hydraulic fracture properties using elliptical ow data in tight gas wells.
Society of Petroleum Engineers Reservoir Evaluation & Engineering 12 (2),
254262.
Cinco-Ley, H., Samaniego-V, F., 1981. Transient pressure analysis for fractured wells.
Journal of Petroleum Technology 9 (33), 17491766.
Cipolla, C.L., Lolon, E.P., Erdle, J.C., Rubin, B., 2010. Reservoir modeling in shale-gas reservoirs. Society of Petroleum Engineers Reservoir Evaluation & Engineering 13 (4),
638653.
Clarkson, C.R., 2009. Case study: production data and pressure transient analysis of
Horseshoe Canyon CBM wells. Journal of Canadian Petroleum Technology 48 (10),
2738.
Clarkson, C.R., 2011. Integration of Rate-transient and Microseismic Analysis for
Unconventional Gas Reservoirs: Where Reservoir Engineering Meets Geophysics.
Canadian Society of Exploration Geophysicists (CSEG) Recorder, pp. 4461
(December 2011).
Clarkson, C.R., 2012. Modeling 2-phase owback of multi-fractured horizontal wells
completed in shale. Society of Petroleum Engineers Paper 162593 Presented at
the SPE Canadian Unconventional Resources Conference, Calgary, Alberta.
Clarkson, C.R., 2013. Production data analysis of unconventional gas wells: Workow.
International Journal of Coal Geology 109110, 147157 (this volume).
Clarkson, C.R., Beierle, J.J., 2011. Integration of microseismic and other post-fracture
surveillance with production analysis: a tight gas study. Journal of Natural Gas
Science and Engineering 3, 382401.
Clarkson, C.R., Bustin, R.M., 2011. Coalbed methane: current eld-based evaluation
methods. Society of Petroleum Engineers Reservoir Evaluation & Engineering
14 (1), 6075.
Clarkson, C.R., McGovern, J.M., 2005. Optimization of coalbed-methane-reservoir exploration and development strategies through integration of simulation and economics. Society of Petroleum Engineers Reservoir Evaluation & Engineering 8 (6),
502519.
Clarkson, C.R., Pedersen, P.K., 2010. Tight oil production analysis: adaptation of existing
rate-transient analysis techniques. Society of Petroleum Engineers Paper 137352
Presented at the Canadian Unconventional Resources and International Petroleum
Clarkson, C.R., Bustin, R.M., Seidle, J.P., 2007. Production-data analysis of single-phase
(gas) coalbed-methane wells. Society of Petroleum Engineers Reservoir Evaluation
& Engineering 10 (3), 312331.
Clarkson, C.R., Jordan, C.L., Gierhart, R.R., Seidle, J.P., 2008. Production data analysis of
CBM wells. Society of Petroleum Engineers Reservoir Evaluation & Engineering
11 (2), 311325.
Clarkson, C.R., Jordan, C.L., Ilk, D., Blasingame, T.A., 2009. Production data analysis of
fractured and horizontal CBM Wells. Society of Petroleum Engineers Paper
125929 Presented at the SPE Eastern Regional Meeting held in Charleston, West
Virginia.
Clarkson, C.R., Jensen, J.L., Blasingame, T.A., 2012a. Unconventional gas reservoir evaluation:
what do we have to consider? Journal of Natural Gas Science and Engineering 8, 933.
Clarkson, C.R., Nobakht, M., Kaviani, D., Ertekin, T., 2012b. Production analysis of tight
gas and shale gas reservoirs using the dynamic-slippage concept. Society of Petroleum Engineers Journal 1 (17), 230242.
Clarkson, C.R., Jordan, C.L., Ilk, D., Blasingame, T.A., 2012c. Rate-transient analysis of
2-phase (gas + water) CBM wells. Journal of Natural Gas Science and Engineering
8, 106120.
Clarkson, C.R., Nobakht, M., Kaviani, D., Kantzas, A., 2012d. Use of pressure- and ratetransient analysis techniques for analyzing core tests for unconventional reservoirs. Society of Petroleum Engineers Paper 154815 Presented at the SPE Americas
Unconventional Resource Conference, Pittsburgh, Pennsylvania.
Clarkson, C.R., Qanbari, F., Nobakht, M., Heffner, L., 2012e. Incorporating geomechanical
and dynamic hydraulic fracture property changes into rate transient analysis: example from the Haynesville Shale. Society of Petroleum Engineers Paper 162526
Presented at the SPE Canadian Unconventional Resources Conference, Calgary,
Alberta.
Clarkson, C.R., Behmanesh, H, Chorney, L., in press. Production data and pressure transient analysis of Horseshoe Canyon CBM wells, part II: accounting for dynamic
skin. Journal of Canadian Petroleum Technology.
Cox, S.A., Gilbert, J.V., Sutton, R.P., Stoltz, R.P., 2002. Reserve analysis for tight gas. Society
of Petroleum Engineers Paper 78695 Presented at the SPE Eastern Regional Conference, Lexington, Kentucky.
Currie, S.M., Ilk, D., Blasingame, T.A., 2010. Continuous estimation of ultimate recovery.
Society of Petroleum Engineers Paper 132352 Presented at the SPE Unconventional
Gas Conference, Pittsburgh, Pennsylvania.
El-Banbi, A.H., 1998. Analysis of tight gas well performance. PhD Thesis, Texas A&M
University.
Fetkovich, M.J., 1971. A simplied approach to water inux calculationsnite aquifer
systems. Journal of Petroleum Technology 23 (7), 814828.
Fetkovich, M.J., 1980. Decline curve analysis using type curves. Journal of Petroleum
Technology 32 (6), 10651077.
Fetkovich, M.J., Vienot, M.E., Bradley, M.D., Kiesow, U.G., 1987. Decline-curve analysis
using type-curves: case histories. Society of Petroleum Engineers Formation Evaluation 2 (4), 637656.
Fetkovich, M.J., Fetkovich, E.J., Fetkovich, M.D., 1996. Useful concepts for decline curve
forecasting, reserve estimation, and analysis. Society of Petroleum Engineers
Reservoir Engineering 11 (1), 1322.
Fraim, M.L., Wattenbarger, R.A., 1987. Gas reservoir decline curve analysis using type
curves with real gas pseudopressure and normalized time. Society of Petroleum
Engineers Formation Evaluation 2 (4), 671682.
Gentzis, T., Bolen, D., 2008. The use of numerical simulation in predicting coalbed
methane producibility from the Gates coals, Alberta Inner Foothills, Canada: comparison with Mannville CBM production in Alberta Syncline. International Journal
of Coal Geology 74, 215236.
Gentzis, T., Deisman, N., Chalaturnyk, R.J., 2009. A method to predict geomechanical
properties and model stability in horizontal boreholes. International Journal of
Coal Geology 78, 149160.
Gerami, S., Pooladi-Darvish, M., Morad, K., Mattar, L., 2007. Type curves for dry CBM
reservoirs with equilibrium desorption. Journal of Canadian Petroleum Technology
47 (7), 4856.
Haskett, W.J., Brown, P.J., 2005. Evaluation of unconventional resource plays. Society of
Petroleum Engineers Paper 96879, Presented at the SPE Annual Technical Conference and Exhibition, Dallas, Texas.
Ibrahim, M., Wattenbarger, R.A., 2006. Analysis of rate dependence in transient linear
ow in tight gas wells. Society of Petroleum Engineers Paper 100836 Presented
at the Abu Dhabi International Petroleum Exhibition and Conference, Abu Dhabi,
UAE.
Ilk, D., Perego, A.D., Rushing, J.A., Blasingame, T.A., 2008a. Integrating multiple production analysis techniques to assess tight gas sand reserves: dening a new paradigm
for industry best practices. Society of Petroleum Engineers Paper 114947 Presented
at the CIPC/SPE Gas Technology Symposium Joint Conference, Calgary, Alberta.
Ilk, D., Rushing, J.A., Perego, A.D., Blasingame, T.A., 2008b. Exponential vs. hyperbolic decline in tight gas sands understanding the origin and implications for reserve estimates using Arps' decline curves. Society of Petroleum Engineers Paper 116731,
Presented at the Annual Technical Conference and Exhibition, Denver, Colorado.
Javadpour, F., 2009. Nanopores and apparent permeability of gas ow in mudrocks
(shales and siltstones). Journal of Canadian Petroleum Technology 48 (8), 1621.
Kamal, M.M., Six, J.L., 1993. Pressure transient analysis of methane producing coalbeds.
Society of Petroleum Engineers Advanced Technology Series 1 (1), 195203.
King, G.R., 1993. Material-balance techniques for coal-seam and Devonian shale gas
reservoirs with limited water inux. Society of Petroleum Engineers Reservoir
Engineering 8 (1), 6772.
Kupchenko, C.L., Gault, B.W., Mattar, L., 2008. Tight gas production performance using
decline curves. Society of Petroleum Engineers Paper 114991 Presented at the
CIPC/SPE Gas Technology Symposium, Calgary, Alberta.
Larsen, L., Hegre, T.M., 1991. Pressure-transient behavior of horizontal wells with
nite-conductivity vertical fractures. Society of Petroleum Engineers paper 22076
presented at the International Arctic Technology Conference, Anchorage, Alaska.

Lee, J., Wattenbarger, R.A., 1996. Gas Reservoir Engineering. : Gas Reservoir Engineering.
Textbook Series, 5. Society of Petroleum Engineers, Richardson, Texas.
Lee, J., Rollins, J.B., Spivey, J.P., 2003. Pressure Transient Testing. : Textbook Series, 9.
Society of Petroleum Engineers, Richardson, Texas.
Lewis, A.M., Hughes, R.G., 2008. Production data analysis of shale gas reservoirs. Society
of Petroleum Engineers Paper 116688 Presented at the SPE Annual Technical
Conference and Exhibition, Denver, Colorado.
Liang, P., Mattar, L., Moghadam, S., 2011. Analyzing variable rate/pressure data in transient linear ow in unconventional gas reservoirs. Society of Petroleum Engineers
Paper 149472 Presented at the Canadian Unconventional Resources Conference,
Calgary, Alberta.
Mattar, L., Anderson, D.M., 2003. A systematic and comprehensive methodology for advanced analysis of production data. Society of Petroleum Engineers Paper 84472
Presented at the SPE Annual Technical Conference and Exhibition, Denver, Colorado.
Mattar, L., Anderson, D.M., 2005. Dynamic material balance (oil or gas-in-place without
shut-ins). Paper 2005-113 Presented at the Canadian International Petroleum Conference, Calgary, Alberta.
Mattar, L., Moghadam, S., 2009. Modied power law exponential for tight gas. Paper
2009-198 Presented at the Canadian International Petroleum Conference, Calgary,
Alberta.
Mattar, L., Gault, B., Morad, K., Clarkson, C.R., Freeman, C.M., Ilk, D., Blasingame, T.A., 2008.
Production analysis and forecasting of shale gas reservoirs: Case history-based approach. Society of Petroleum Engineers paper 119897 presented at the SPE Shale
Gas Production Conference, Fort Worth, Texas.
Mavor, M.J., 1996. A guide to coalbed methane reservoir engineering, chapter 4. Report
No. GRI-94/0397. Gas Research Institute, Chicago, Illinois.
Mavor, M.J., Saulsberry, J.L., 1996. A guide to coalbed methane reservoir engineering,
chapter 5. Report No. GRI-94/0397. Gas Research Institute, Chicago, Illinois.
Mavor, M.J., Russell, B., Pratt, T.J., 2003. Powder River Basin Ft. Union Coal reservoir
properties and production decline analysis. Society of Petroleum Engineers paper
84427 presented at the SPE Annual Technical Conference and Exhibition, Denver,
Colorado.
Mayerhofer, M.J., Lolon, E.P., Warpinski, N.R., Cipolla, C.L., Walser, D., Rightmire, C.M.,
2010. What is stimulated reservoir volume? Society of Petroleum Engineering Production & Operations 1 (25), 8998.
McCray, T.L., 1990. Reservoir analysis using production decline data. MSc Thesis, Texas
A&M University.
Medeiros, F., Ozkan, E., Kazemi, H., 2008. Productivity and drainage area of fractured
horizontal wells in tight gas reservoirs. Society of Petroleum Engineers Reservoir
Evaluation & Engineering 11 (5), 902911.
Medeiros, F., Kurtoglu, B., Ozkan, E., Kazemi, H., 2010. Analysis of production data from
hydraulically-fractured horizontal wells in shale reservoirs. Society of Petroleum
Engineers Reservoir Evaluation & Engineering 13 (3), 559568.
Michel, G., Sigal, R., Civan, F., Devegowda, D., 2012. Effect of capillary relaxation on
water entrapment after hydraulic fracturing stimulation. Society of Petroleum
Engineers Paper 155787 Presented at the Americas Unconventional Resources
Conference, Pittsburgh, Pennsylvania.
Moghadam, S., Mattar, L., Pooladi-Darvish, M., 2010. Dual porosity typecurves for shale
gas reservoirs. Society of Petroleum Engineers Paper 137535 Presented at the
Canadian Unconventional Resources and International Petroleum Conference,
Calgary, Alberta.
Mohaghegh, S.D., Ertekin, T., 1991. Type-curve solution for coal Seam degasication
wells producing under two-phase conditions. Society of Petroleum Engineers
Paper 22673 Presented at the SPE Annual Technical Conference and Exhibition,
Dallas, Texas.
Nobakht, M., Clarkson, C.R., 2011. Estimation of contacted and original gas-in-place for
low permeability reservoirs exhibiting linear ow. Society of Petroleum Engineers
Paper 149398 Presented at the Canadian Unconventional Resources Conference,
Calgary, Alberta.
Nobakht, M., Clarkson, C.R., 2012a. A new analytical method for analyzing production
data from tight/shale gas reservoirs exhibiting linear ow: constant rate boundary
condition. Society of Petroleum Engineers Reservoir Evaluation & Engineering 15
(1), 5159.
Nobakht, M., Clarkson, C.R., 2012b. A new analytical method for analyzing production
data from shale gas reservoirs exhibiting linear ow: constant pressure
boundary-condition. Society of Petroleum Engineers Reservoir Evaluation & Engineering 15 (3), 370384.
Nobakht, M., Clarkson, C.R., 2012c. Analysis of production data in shale gas reservoirs:
rigorous corrections for uid and ow properties. Journal of Natural Gas Science
and Engineering 8, 8598.
Nobakht, M., Mattar, L., 2012. Analyzing production data from unconventional gas reservoirs with linear ow and apparent skin. Journal of Canadian Petroleum Society
51 (1), 5259.
Nobakht, M., Ambrose, R., Clarkson, C.R., 2011. Effect of heterogeneity in a horizontal
well with multiple fractures on long-term forecast in shale gas reservoirs. Society
of Petroleum Engineers Paper 149400 Presented at the Canadian Unconventional
Resources Conference, Calgary, Alberta.
Nobakht, M., Clarkson, C.R., Kaviani, 2012a. New type-curves for analyzing horizontal
well with multiple fractures in shale gas reservoirs. Journal of Natural Gas Science
and Engineering 10, 99112.
Nobakht, M., Clarkson, C.R., Kaviani, 2012b. New and improved methods for performing
rate-transient analysis of shale gas reservoirs. Society of Petroleum Engineers Journal
15 (3), 335350.
Nobakht, M., Mattar, L., Moghadam, S., Anderson, D.M., 2012c. Simplied forecasting of
tight/shale-gas production in linear ow. Journal of Canadian Petroleum Technology
51 (6), 476486.
145
Ozkan, E., Raghavan, R., Apaydin, O.G., 2010. Modeling of uid transfer from shale matrix to fracture network. Society of Petroleum Engineers Paper 134830 Presented at
the SPE Annual Technical Conference and Exhibition, Florence, Italy.
Ozkan, E., Brown, M., Raghavan, R., Kazemi, H., 2011. Comparison of fracturedhorizontal-well performance in tight sand and shale reservoirs. Society of Petroleum Engineers Reservoir Evaluation & Engineering 14 (2), 248259.
Palacio, J.C., Blasingame, T.A., 1993. Decline-curve analysis using type curvesanalysis
of gas well production data. Society of Petroleum Engineers Paper 25909 Presented
at the SPE Joint Rocky Mountain Regional and Low Permeability Reservoirs Symposium, Denver, Colorado.
Palmer, I., 2010. Coalbed methane: a world view. International Journal of Coal Geology
82, 184195.
Pan, Z., Connell, L.D., 2012. Modelling permeability for coal reservoirs: a review of analytical models and testing data. International Journal of Coal Geology 92 (1), 144.
Paul, G.W., 1996. A guide to coalbed methane reservoir engineering, chapter 6. Report
Perrine, R.L., 1956. Analysis of Pressure Buildup Curves. Drilling and Production Practice. American Petroleum Institute, pp. 482509.
Poston, S.W., Poe, B.D., 2008. Analysis of Production Decline Curves. Society of Petroleum Engineers Book, Richardson, Texas.
Pratikno, H., Rushing, J.A., Blasingame, T.A., 2003. Decline curve analysis using typecurves fractured wells. Society of Petroleum Engineers paper 84287 presented
at the SPE Annual Technical Conference and Exhibition, Denver, Colorado.
Raghavan, R., Chin, L.Y., 2004. Permeability changes in reservoirs with stressdependent permeability. Society of Petroleum Engineers Reservoir Evaluation &
Engineering 7 (4), 308315.
Raghavan, R., Chen, C.C., Agarwal, B., 1997. An analysis of horizontal wells intercepted
by multiple fractures. Society of Petroleum Engineers Journal 2 (3), 235245.
Rushing, J.A., Perego, A.D., Sullivan, R.B., Blasingame, T.A., 2007. Estimating reserves in
tight gas sands at HP/HT reservoir conditions: use and misuse of an Arps decline
methodology. Society of Petroleum Engineers Paper 109625, Presented at the
2007 SPE Annual Technical Conference and Exhibition, Anaheim, California.
Rushing, J.A., Perego, A.D., Blasingame, T.A., 2008. Applicability of the Arps ratetime
relationships for evaluating decline behavior of and ultimate gas recovery of
coalbed methane wells. Society of Petroleum Engineers Paper 114514, Presented
at the CIPC/SPE Gas Technology Symposium 2008 Joint Conference, Calgary,
Alberta.
Samandarli, O., Al-Ahmadi, H., Wattenbarger, R.A., 2011. A new method for historymatching and forecasting shale gas reservoir production performance with a
dual-porosity model. Society of Petroleum Engineers Paper 144335 Presented at
the North American Unconventional Gas Conference, The Woodlands, Texas.
Samandarli, O., Valbuena, E., Ehlig-Economides, C., 2012. Production data analysis in
unconventional reservoirs with rate-normalized pressure (RNP): theory, methodology, and applications. Society of Petroleum Engineers Paper 155614 Presented at
the Americas Unconventional Resources Conference, Pittsburgh, Pennsylvania.
Seidle, J.P., 2002. Coal well decline behavior and drainage areas: theory and practice.
Society of Petroleum Engineers Paper 75519 Presented at the SPE Gas Technology
Symposium, Calgary, Alberta, Canada.
Seidle, J.P., 2011. Fundamentals of Coalbed Methane Reservoir Engineering. PennWell
Corporation, Tulsa, Oklahoma.
Song, B., Ehlig-Economides, C., 2011. Rate-normalized pressure analysis for determination of shale gas performance. Society of Petroleum Engineers Paper 144031
Presented at the North American Unconventional Gas Conference, The Woodlands,
Texas.
Song, B., Economides, M.J., Ehlig-Economides, C., 2011. Design of multiple transverse
fracture horizontal wells in shale gas. Society of Petroleum Engineers Paper
140555 Presented at the SPE Hydraulic Fracture Technology Conference, The
Woodlands, Texas.
Sureshjani, M.H., Gerami, S., 2011. A new model for modern production-decline analysis
of gas/condensate reservoirs. Journal of Canadian Petroleum Technology 50 (7/8),
1423.
Swami, V., Settari, A., 2012. A pore scale gas ow model for shale gas reservoir. Society
of Petroleum Engineers Paper 155756 Presented at the SPE Americas Unconventional Resource Conference, Pittsburgh, Pennsylvania.
Thompson, J.M., Nobakht, M., Anderson, D.M., 2010. Modeling well performance data
from overpressured shale gas reservoirs. Society of Petroleum Engineers Paper
137755 Presented at the Canadian Unconventional Resources and International
Petroleum Conference, Calgary, Alberta.
Valko, P., 2009. Assigning value to stimulation in the Barnett Shale: a simultaneous
analysis of 7000 plus production histories and well completion records. Society
of Petroleum Engineers Paper 119369 Presented at 2009 SPE Hydraulic Fracturing
Technology Conference, The Woodlands, Texas.
Valko, P., Lee, W.J., 2010. A better way to forecast production from unconventional gas
wells. Society of Petroleum Engineers Paper 134231 Presented at SPE Annual Technical Conference and Exhibition, Florence, Italy.
Van Kruysdijk, C.J.P.W., Dullaert, G.M., 1989. A boundary element solution of the transient pressure response of multiple fractured horizontal wells. Paper Presented at
the 2nd European Conference on the Mathematics of Oil Recovery, Cambridge,
England.
Warpinski, N.R., Mayerhofer, M.J., Vincent, M.C., N.R., Cipolla, C.L., Lolon, E.P., 2008.
Simulating unconventional reservoirs: maximizing network growth while optimizing fracture conductivity? Society of Petroleum Engineers Paper 114173
Presented at the SPE Unconventional Reservoirs Conference, 1012 February
2008, Keystone, Colorado, USA.
Wattenbarger, R.A., El-Banbi, A.H., Villegas, M.E., Maggard, J.B., 1998. Production analysis of linear ow into fractured tight gas wells. Society of Petroleum Engineers
146
Paper 39931 Presented at SPE Rocky Mountain Regional/Low Permeability Reservoirs Symposium and Exhibition, Denver, Colorado.
Williams-Kovacs, J.D., Clarkson, C.R., 2011. Using stochastic simulation to quantify risk
and uncertainty in shale gas prospecting and development. Society of Petroleum
Engineers Paper 148867 Presented at the Canadian Unconventional Resources
Williams-Kovacs, J.D., Clarkson, C.R., Nobakht, M., Ambrose, R., 2012. Impact of material
balance equation selection on rate-transient analysis of shale gas. Society of Petroleum Engineers Paper XXXXXX Presented at the SPE Annual Technical Conference
and Exhibition, San Antonio, Texas.
Zuber, M.D., 1996. A guide to coalbed methane reservoir engineering, chapter 3. Report

Clarkson (2013)

Uploaded by

Copyright:

Available Formats

Clarkson (2013)

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Clarkson (2013)

Uploaded by

Copyright:

Available Formats

International Journal of Coal Geology 109110 (2013) 101146

Contents lists available at SciVerse ScienceDirect

International Journal of Coal Geology

Production data analysis of unconventional gas wells: Review of

Tel.: +1 403 220 6445.

C.R. Clarkson / International Journal of Coal Geology 109110 (2013) 101146

well/operating condition data can be used in a model that accurately

C.R. Clarkson / International Journal of Coal Geology 109110 (2013) 101146

2. Concept of rate transient analysis: ow-regimes

C.R. Clarkson / International Journal of Coal Geology 109110 (2013) 101146

Both RTA and PTA start with the identication of ow-regimes,

Boundary Dominated Flow

Radial Derivative Plot

C.R. Clarkson / International Journal of Coal Geology 109110 (2013) 101146

occur as a result of ow-convergence to the horizontal well, cleanup

C.R. Clarkson / International Journal of Coal Geology 109110 (2013) 101146

embedded within an SRV that has also been created as a result of

Advanced production data analysis (or rate-transient analysis)

Straight-line (or ow-regime) analysis methods

In the following, we detail these methods and provide simulated

C.R. Clarkson / International Journal of Coal Geology 109110 (2013) 101146

equations. Most of these solutions were derived for drawdown

pressures, and pseudovariables (pressure and time) are used to

Linear flow normal to

Inter-fracture pressure Interference

C.R. Clarkson / International Journal of Coal Geology 109110 (2013) 101146

Nearly unit slope,

Fracture Boundary-Dominated Flow

Matrix Linear + FBD Flow

Fracture Linear Flow (FL)

procedures were developed for wells with innite-conductivity and

C.R. Clarkson / International Journal of Coal Geology 109110 (2013) 101146

Single Porosity Reservoir

Dual Porosity Reservoir

Stimulated Reservoir Volume

Ex. Nobakht and Clarkson typecurves (Nobakht et al. (2011)

Discrete Hydraulic Fractures

Trilinear flow model (Brown et al. 2011)

Radial derivative: 1/2 slope

Radial derivative: zero slope

vs. (ta)1/4 or t a;BLS

vs. (ta)1/2 or t a;LS

vs. log(ta) or t a;RS

Radial derivative: unit slope

s 1:1513 mbRR log

C.R. Clarkson / International Journal of Coal Geology 109110 (2013) 101146

from the analysis. Prior to the development of the straight-line technique

Note that for slightly-compressible uids, assuming pore volume

a simpler version of material balance time to be used: tc = Q / q.

mate of xf k. The elliptical ow period (Fig. 10b) can also be analyzed

C.R. Clarkson / International Journal of Coal Geology 109110 (2013) 101146

Linear Flow Plot

Elliptical Flow Plot

Use to calculate xf,

Use to calculate k and xf

Radial Flow Plot

Use to calculate k and

Flowing Material Balance

Use to calculate OGIP and