(Blasingame) SPE 109625

SPE 109625
Estimating Reserves in Tight Gas Sands at HP/HT Reservoir Conditions: Use and
Misuse of an Arps Decline Curve Methodology
J.A. Rushing, SPE, Anadarko Petroleum Corp., A.D. Perego, SPE, Anadarko Petroleum Corp., R.B. Sullivan, SPE,
Anadarko Petroleum Corp., and T.A. Blasingame, SPE, Texas A&M University
Copyright 2007, Society of Petroleum Engineers

This paper was prepared for presentation at the 2007 SPE Annual Technical Conference and
Exhibition held in Anaheim, California, U.S.A., 1114 November 2007.
This paper was selected for presentation by an SPE Program Committee following review of
information contained in an abstract submitted by the author(s). Contents of the paper, as
presented, have not been reviewed by the Society of Petroleum Engineers and are subject to
correction by the author(s). The material, as presented, does not necessarily reflect any
position of the Society of Petroleum Engineers, its officers, or members. Papers presented at
SPE meetings are subject to publication review by Editorial Committees of the Society of
Petroleum Engineers. Electronic reproduction, distribution, or storage of any part of this paper
for commercial purposes without the written consent of the Society of Petroleum Engineers is
prohibited. Permission to reproduce in print is restricted to an abstract of not more than 300
words; illustrations may not be copied. The abstract must contain conspicuous
acknowledgment of where and by whom the paper was presented. Write Librarian, SPE, P.O.
Box 833836, Richardson, Texas 75083-3836 U.S.A., fax 01-972-952-9435.
Abstract
This paper presents the results of a simulation study designed
to evaluate the applicability of an Arps1 decline curve methodology for assessing reserves in hydraulically-fractured wells
completed in tight gas sands at high-pressure/high-temperature
(HP/HT) reservoir conditions. We simulated various reservoir
and hydraulic-fracture properties to determine their impact on
the production decline behavior as quantified by the Arps
decline curve exponent, b. We then evaluated the simulated
production with Arps' rate-time equations at specific time
periods during the well's productive life and compared
estimated reserves to the true value. To satisfy requirements
for using Arps' models, all simulations were conducted using a
specified constant bottomhole flowing pressure condition in
the wellbore.
Our study indicates that the largest error source is incorrect
application of Arps' decline curves during either transient flow
or the transitional period between the end of transient and
onset of boundary-dominated flow. During both of these
periods (principally the transient period), we observed bexponents greater than one and corresponding reserve estimate
errors exceeding 100 percent. The b-exponents generally
approached values between 0.5 and 1.0 as flow conditions
approached true boundary-dominated flow. Agreement between Arps' suggested b-exponent range and our results using
simulated performance data also indicates that, if applied
under the correct conditions, the Arps rate-time models are
appropriate for assessing reserves in tight gas sands at HP/HT
reservoir conditions.
Introduction
Tight gas sands constitute a significant percentage of the
domestic natural gas resource base and offer tremendous
potential for future reserve and production growth. According
to a recent study by the Gas Technology Institute (GTI),2 tight

gas sands in the US comprise 69 percent of gas production
from all unconventional natural gas resources and account for
19 percent of total gas production from both conventional and
unconventional sources. The same study2 estimates total
domestic producible tight gas sand resources exceed 600 Tcf,
while economically recoverable gas reserves are 185 Tcf.
Most of the resources assessed in the 2001 GTI study were at
depths less than 15,000 ft, yet the natural gas industry
continues to extend exploration and development activities to
much greater depths. In some geologic basins, those depths
are approaching 20,000 to 25,000 ft. Many of these deep
natural gas resources are not only characterized by lowpermeability, low-porosity reservoir properties, but these
reservoirs also exhibit abnormally high initial pore pressure
and temperature gradients i.e. high-pressure/hightemperature (HP/HT) reservoir conditions.
Similar to conventional natural gas resources, tight gas sand
reserves are routinely assessed with Arps decline curve
techniques. The original Arps1 paper suggested the decline
curve exponent, b, should fall between 0 and 1.0 on a semilog
plot. However, we often observe values much greater than
1.0, particularly in tight gas sands at HP/HT reservoir
conditions. Deviations in observed b-exponents from the
expected range suggest Arps' rate-time relationships may not
be valid for modeling the decline behavior of tight gas sands
at HP/HT conditions. More importantly, inappropriate use of
the Arps models may cause significant reserve estimate errors
in these unconventional natural gas resources.
Since these depths and extreme reservoir conditions require
wells that are very expensive to drill, complete and operate; it
is imperative that we understand both the well productivity
and production decline behavior. We also need to determine
the applicability of the Arps rate-time equations for assessing
reserves. To address these concerns, we have conducted a
series of single-well simulation studies to develop a better
understanding of both the short- and long-term production
decline behavior and to identify those parameters affecting the
production decline. In this study we simulated a range of
reservoir and hydraulic fracture properties, including:
z Vertical heterogeneity from layering, permeability contrast
among layers, horizontal permeability anisotropy, and stressdependent reservoir properties;
J.A. Rushing, A.D. Perego, R.B. Sullivan, and T.A. Blasingame
z Variable effective fracture conductivities and lengths, unequal

fracture wing lengths, two-phase and non-Darcy flow, and
stress-dependent fracture properties; and
Table 1 Summary of the Arps' rate-time relations (Ref. 1)

Case
Rate Relation
Hyperbolic:
(0<b<1)
q (t ) =
Exponential:
(b=0)
q(t ) = qi exp( Dit ) ................................ (2)
Harmonic:
(b=1)
q (t ) =
qi
[1 + bDit ]1/b
qi
[1 + Dit ]
............................... (1)
................................... (3)
Description of Numerical Model

We developed a three-dimensional, two-phase (gas-water)
finite-difference model.3-5 We employed a Cartesian rather
than a radial grid geometry so that we could more accurately
model the linear flow patterns characteristic of hydraulicallyfractured wells producing from tight gas sands. We should
note that we developed a reservoir model that addressed
reservoir inflow performance, and other than using a
bottomhole flowing pressure constraint, we did not attempt to
model well outflow performance.
We initially modeled a well drilled on a spacing of 80 acres
per well, but we also evaluated well spacings of 40 and 160
acres per well.
The 80-acre grid system, illustrated
schematically in Fig. 1, was constructed with 1521 grids (39
grids in both the x- and y-directions). We used finer grids
adjacent to the wellbore and at the fracture tips because of the
large pressure gradients observed perpendicular to the fracture
faces and at the fracture tips. During the simulation process,
we also monitored reservoir pressures at two points (Fig. 1) on
the edge of the drainage area so that we could assess flow
conditions (i.e., transient, transition, or boundary-dominated
flow conditions).
Reservoir Model and Properties. We modeled a hydraulically-fractured well completed in multiple layers at an average
depth of 18,000 ft in a HP/HT, low-permeability, low-porosity
gas sand. We assumed that the reservoir fluid was a dry gas
with connate water saturations that are greater than the
irreducible values, but with only a small fraction that is mobile
(this is the analog for a basin-centered gas system.6,7).
Pressure
Monitoring
Point No. 2
z Reservoir temperatures ranging from 300 to 400oF and initial

pore pressure gradients ranging from 0.60 to 0.90 psi/ft.
We evaluated the simulated production with the Arps1 ratetime equations. Reserve estimates were obtained at various
time periods during the wells productive life by extrapolating
the best-fit Arps model through the simulated production. Our
assumed economic conditions for estimating reserves were
either a rate of 50 Mscf/d or a producing time period of 50
years, whichever came first. Reserve estimate errors were
computed by comparing those estimated reserves to the true
value. For this paper, we define the true estimated ultimate
recovery (EUR) to be the 50-year cumulative production
volume. For reference, we also summarize the Arps rate-time
equations in Table 1, given below:
SPE 109625
Wellbore
Pressure
Monitoring
Point No. 1
y
X
Hydraulic
Fracture
Fig. 1 Schematic diagram illustrating x-y Cartesian grid

system for numerical model of hydraulicallyfractured well.
Reservoir Pressures and Temperatures. For the base case,

we assumed an initial pore pressure gradient of 0.90 psi/ft
(pi=16,200 psia at a depth of 18,000 ft) and an initial
bottomhole temperature of 400oF. We also evaluated initial
pore pressure gradients of 0.60, 0.70 and 0.80 psi/ft and a
bottomhole temperature of 300oF. Because of the extreme
reservoir pressure and temperature conditions, the reservoir
model was formulated using a pseudopressure approach3-5,8 so
that we could incorporate pressure-dependent gas properties
and their effect on gas storage and flow capacity.
Reservoir Fluids. We modeled a gas-water system in which
the hydrocarbon fluid is a dry gas (composition shown in
Table 2), i.e. a gas that forms no liquid hydrocarbons over the
pressure depletion path from the reservoir to surface facilities.
Gas deviation or z-factors were computed using the DranchukAbou-Kassem9 correlations with the Wichert-Aziz10
corrections for carbon dioxide and the Casey11 correction for
nitrogen.
Gas viscosities were computed using the
correlations developed by Lee, et al.12 We also assumed the
in-situ connate water is a brine with a total dissolved solids
content of 30,000 ppm.
Table 2 Simulated natural gas composition for a hypothetical HP/HT tight gas sand reservoir.
Gas
Component
Methane
Ethane
Propane
Nitrogen
Carbon Dioxide
Mole
Percent
90.0
3.0
1.5
0.5
5.0
Total mole percent

HC mole percent
100.0
94.5
Total gas specific gravity, g

HC gas specific gravity, HC
0.633
0.585
Reservoir Permeability and Porosity. As a basis for our

simulated cases, we used rock permeability and porosity data
from a comprehensive core description and evaluation program (400 ft of whole core) in a Lower Cotton Valley
Sandstone reservoir at HP/HT reservoir conditions and at
depths of 13,000 to 15,000 ft. Although taken from shallower
zones than evaluated in our study, we have assumed the rock
SPE 109625
Estimating Reserves in Tight Gas Sands at HP/HT Reservoir Conditions: Use and Misuse of an Arps Decline Curve Methodology
1.00E+00
0.30 microns
1.00E-01
HRT 1
0.15 microns
HRT 2
0.075 microns
1.00E-02
HRT 3
0.035 microns
1.00E-03
Pore Size, microns
Absolute Klinkenberg-Corrected Permeability, md
properties and their distribution would be similar for a deeper

reservoir. A significant part of the core evaluation program
included classification of hydraulic rock types.13 When
described on the basis of the dominant pore throat diameter
determined from high-pressure, mercury capillary pressure
data, we observed distinct groupings of rocks having similar
flow and storage properties, i.e., hydraulic rock types (HRT).
Figure 2 is a semilog plot showing the general region of each
hydraulic rock type in porosity-permeability space and
grouped according to empirical relationships developed by
Pittman.14
HRT 4
0.015 microns
1.00E-04
2.0
3.0
4.0
5.0
6.0
7.0
8.0
9.0
10.0
11.0
12.0
Effective Porosity, %
14
Fig. 2 Pittman plot showing hydraulic rock types (HRT)

as a function of permeability, porosity and
dominant pore throat size.
To capture the range of permeability and porosity exhibited by

the four hydraulic rock types, we developed a four-layer case
where each layer represents a different hydraulic rock type
with average properties from Fig. 2. The layers can be
considered to be hydraulic flow units (HFU) which are defined
simply as "groupings of rocks with similar properties."13-15
For modeling purposes, we have assumed the relative
thickness of each HFU is directly proportional to the
percentage of rock types shown in Fig. 2. For example, HRT
1 has the fewest measured points from the coring program and
will represent the thinnest layer. Conversely, HRT 3 has the
most measured data points and will be the thickest layer. In
Table 3 we summarize the properties for each HFU (or layer)
in a reservoir with a gross sand thickness of 200 ft. We should
note that the average properties are "thickness-averaged"
values.
Table 3 Summary of hydraulic flow units (HFU) and
average rock properties, four-layer case.
HFU
1
2
3
4
Avg./Total
HRT
1
2
3
4
h
(ft)
10
30
120
40
200
kg
(md)
0.07291
0.01752
0.00428
0.00086
0.0090
(%)
8.43
7.45
6.57
5.65
6.61
Sw
(%)
10.0
20.0
40.0
50.0
36.4
We also evaluated reservoirs with one, eight and sixteen

layers. For the single-layer case, we assumed a homogeneous,
isotropic system with average effective porosity and absolute
horizontal permeability equal to the thickness-averaged values
shown in Table 3. For the eight- and sixteen-layer cases, we

incorporated more contrast among layer properties by utilizing
a greater range of permeabilities and porosities as shown in
Fig. 2. For example, the permeabilities and porosities given in
Table 4 for the eight-layer case represent the average + one
standard deviation for each HRT.
average rock properties, eight-layer case.
HFU
1
2
3
4
5
6
7
8
Avg./Total
HRT
1
2
3
4
1
2
3
4
h
(ft)
5
15
60
20
5
15
60
20
200
kg
(md)
0.05209
0.01191
0.00292
0.00053
0.09372
0.02314
0.00564
0.00119
0.0090
(%)
7.77
6.73
5.87
4.94
9.10
8.16
7.26
6.35
6.61
Sw
(%)
10.0
20.0
40.0
50.0
10.0
20.0
40.0
50.0
36.4
Similarly, the permeabilities and porosities listed in Table 5

for the sixteen-layer case range from the average + one
standard deviation to the average + two standard deviations.
Note also that thickness-averaged properties shown in Tables
4 and 5 are the same as those for the four-layer case
summarized in Table 3. Maintaining these equalities allows
us to make direct comparisons among the various cases.
average rock properties, sixteen-layer case.
HFU
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
Avg./Total
HRT
1
2
3
4
1
2
3
4
1
2
3
4
1
2
3
4
h
(ft)
2.5
7.5
30
10
2.5
7.5
30
10
2.5
7.5
30
10
2.5
7.5
30
10
200
kg
(md)
0.05209
0.01191
0.00292
0.00053
0.09372
0.02314
0.00564
0.00119
0.03128
0.00629
0.00156
0.00021
0.11454
0.02876
0.00700
0.00152
0.0090
(%)
7.77
6.73
5.87
4.94
9.10
8.16
7.26
6.35
7.10
6.02
5.18
4.24
9.76
8.88
7.95
7.06
6.61
Sw
(%)
10.0
20.0
40.0
50.0
10.0
20.0
40.0
50.0
10.0
20.0
40.0
50.0
10.0
20.0
40.0
50.0
36.4
To illustrate the comparative heterogeneity for each layered

case, we constructed a Modified Stratigraphic Lorenz Plot
(Fig. 3)15-17 which is simply a plot of the normalized
cumulative flow capacity against normalized cumulative
storage capacity from the top of the shallowest layer to the
bottom of the deepest layer. Note that the single-layer case,
which forms a 45-degree line, represents a homogeneous, isotropic system. Deviations from this line indicate heterogeneity. Surprisingly, the four-layer case displays much more
deviation than the other two multi-layer cases. In contrast, the
sixteen-layer case shows the least deviation of all multi-layer
cases. As we will demonstrate later, more heterogeneity is
manifested by hyperbolic decline behavior with larger Arps bexponents than cases with less heterogeneity. In fact, the
heterogeneous cases will display more of the hyperbolic
decline shape characteristic of wells completed in tight gas
sands.
0.90
0.80
0.70
0.90
0.80
0.70
0.60
0.50
0.40
0.30
HRT 1
HRT 2
0.20
0.60
0.10
0.50
0.00
HRT 3
HRT 4
0
2,000
4,000
0.40
6,000
8,000
10,000
12,000
14,000
16,000
Reservoir Pore Pressure, psia
0.30
1-Layer Case
Fig. 5 Stress-dependent, absolute Klinkenberg-corrected

permeability relationships for various hydraulic
rock types (HRT).
4-Layer Case
0.20
8-Layer Case
0.10
16-Layer Case
0.00
0.00
0.10
0.20
0.30
0.40
0.50
0.60
0.70
0.80
0.90
1.00
Normalized Cumulative Storage Capacity, frac.

15-17
Fig. 3 Modified Stratigraphic Lorenz

plot showing
relative heterogeneity for hydraulic flow units (HFU)
for all layered cases.
Stress-Dependent Reservoir Properties. Although all rocks

have some stress-dependent properties, tight gas sands (in
particular), display both stress-dependent porosity and permeability characteristics.18 Observed changes in porosity,
however, are usually much less significant than changes in
permeability. We also note that the magnitudes of the stress
dependencies are functions of hydraulic rock type. Generally,
the lower-permeability, lower-porosity rock types will exhibit
much larger relative changes in stress-dependent properties.
For our simulation study, we have re-scaled laboratory
measurements from the shallower Lower Cotton Valley sands
to reflect the higher pressures in our hypothetical deep gas
reservoirs. Stress-dependent effective porosity and absolute
permeability for each of the four hydraulic rock types used in
our simulation study are illustrated in Figs. 4 and 5,
respectively. Note that the stress measurements have been
converted to multipliers representing a fraction of the original
property value as a function of reservoir pore pressure.
Gas-Water Capillary Pressure Curves. To model the vertical

distribution of fluids in our hypothetical HP/HT tight gas
sands, we also employed gas-water capillary pressure curves
which were measured using core samples from shallower
Cotton Valley Sands. Since no single conventional measurement technique can achieve the low water saturations (and
associated high capillary pressures) in low-permeability rocks,
we used data generated using a hybrid vapor desorption/highpressure porous plate method.19,20 We also have capillary
pressure curves for each HRT. Moreover, we converted the
data to a Leverett J-function21 (Fig. 6) for use in the simulator.
100
HRT 1
Leverett J-Function, dimensionless
Normalized Cumulative Flow Capacity, frac.
1.00
SPE 109625
1.00
Fraction Original Permeability, frac.
HRT 2
HRT 3
10
HRT 4
0.1
0.01
0.00
0.20
0.40
0.60
0.80
1.00
Water Saturation, fraction

1.00
21
Fig. 6 Gas-water Leverett J-Functions

draulic rock types (HRT).
Fraction Original Porosity, frac.
0.98
for various hy-
0.96
0.94
0.92
0.90
0.88
0.86
HRT 1
0.84
HRT 2
HRT 3
0.82
HRT 4
0.80
0
2,000
4,000
6,000
8,000
10,000
12,000
14,000
16,000
Reservoir Pore Pressure, psia
Fig. 4 Stress-dependent effective porosity relationships

for various hydraulic rock types (HRT).
Gas-Water Relative Permeability Curves. Although we are

modeling a basin-centered tight gas sand6,7 in which the
connate water saturations are relatively immobile, we still
incorporated gas-water relative permeability data in our
simulation study. The gas relative permeability curves, shown
in Fig. 7 for each HRT from the Lower Cotton Valley Sands,
were measured using an incremental phase trapping
technique.22 Further, the data have been corrected for gas
slippage or Klinkenberg effects.23 Because of the low permeability, we were able to measure the effective permeability
to water for HRT 1 and 2 only. The data shown for HRT 3
and 4 were derived from single, end-point measurements with
synthetic curves drawn to mimic the general curve shape for
HRT 1 and 2.
SPE 109625
of 3,000 psia after the initial clean-up period and for the
remaining productive life of the well.
1.0
HRT 1 Gas
0.9
HRT 3 Gas
0.7
HRT 4 Gas
1.00
HRT 3 Water
0.5
HRT 4 Water
0.4
0.3
0.2
0.1
0.0
0.0
0.1
0.2
0.3
0.4
0.5
0.6
Gas
0.90
HRT 1 & 2 Water
0.6
Relative Permeability, frac.
Relative Permeability, frac.
HRT 2 Gas
0.8
0.7
0.8
0.9
Water
0.80
0.70
0.60
0.50
0.40
0.30
0.20
1.0
0.10
Water Saturation, %
0.00
0.00
Fig. 7 Gas-water relative permeability curves for various

hydraulic rock types (HRT).
We also modeled several types of stimulation treatments and

properties. Specifically, we investigated fractures with three
different types of generic proppants, each having different
crushing and embedment characteristics. We evaluated the
following range of fracture heterogeneities and the impact of
these heterogeneities on well productivity and performance:
z Effective fracture half-length (equal wing lengths) of 50, 100,
200, 300, 400, and 500 ft;
z Unequal effective fracture half-lengths, including wing length
ratios of 6:1, 3:1 and 1.5:1;
0.20
0.30
0.40
0.50
0.60
0.70
0.80
0.90
1.00
Water Saturation, %
Fig. 8 Gas-water fracture relative permeability profiles

used in this study (ref. 24).
Fraction Original Fracture Conductivity, frac.
Hydraulic Fracture Model. We modeled a vertical fracture

extending over the entire sand thickness. Although some of
our simulation cases evaluated unequal fracture lengths, all
cases assumed each fracture half-length (or fracture "wing")
extended equally in the vertical direction in each HFU or
layer. As mentioned previously, the grid system was
constructed using a Cartesian rather than radial grid geometry
so that we could model the linear flow patterns characteristic
of hydraulically-fractured wells producing from tight gas
sands.
0.10
1.00
0.90
0.80
0.70
0.60
0.50
0.40
0.30
0.20
0.10
0.00
2,000
Proppant No. 1
Proppant No. 2
Proppant No. 3
4,000
6,000
8,000
10,000
12,000
14,000
16,000
Fracture Closure Pressure, psia
Fig. 9 Stress-dependent fracture conductivities for various

types of generic proppants modeled in study.
z Stress-dependent fracture conductivity for three generic types

of proppants (Fig. 9).
Decline Curve Analysis of Simulated Production

Traditional decline curve analysis consists of plotting
production rate against time, history matching the production
data using one of several industry-standard models i.e., the
Arps1 exponential, hyperbolic or harmonic decline models
and extrapolating the established trend into the future. Decline
curve analysis is quite simply a curve fitting process which
does not necessarily have a theoretical basis. An exception to
this statement is the exponential decline case which can be
derived from a single-well model producing at a constant
bottomhole flowing pressure during boundary-dominated flow
conditions.
Simulated Operational Conditions. To better represent

operational conditions, we implemented several flowing
bottomhole pressure constraints in the simulation process to
ensure that we modeled field operations as realistically as
possible. For example, we modeled a well that is choked back
during the initial flow back for the first few weeks during the
clean up period, but is allowed to reach line pressure at the end
of that time period. Specifically, we evaluated the effect of
the maximum pressure drawdown on long-term production
decline behavior, and we observed little differences for
pressure drawdown constraints from 2000 psia to 8000 psia.
For all cases, we used a constant bottomhole flowing pressure
We evaluated the simulated production using common industry practices i.e., assuming the Arps rate-time equations
are applicable. Our objectives were to not only quantify the
Arps decline curve parameters (i.e., initial decline rate, Di,
decline exponent, b, etc.), but to also assess reserves at various
times during the wells productive life. Reserve estimates
were obtained by extrapolating the best-fit Arps model
through the simulated production. Our assumed economic
conditions for estimating reserves were either a rate of 50
Mscf/d or a time period of 50 years, whichever came first.
Reserve estimate errors were computed by comparing those
estimated reserves to the true value. For this paper, we
z A range of fracture conductivities (0.1 FCD < 200) including

absolute fracture permeabilities of 1, 10, 100, and 1000 md
(constant throughout entire fracture);
zVariable fracture permeabilities within the fracture, including
the choked fracture case (lower fracture permeability near the
wellbore);
z Two-phase (gas-water) flow in the fracture using non-linear
fracture relative permeabilities (Fig. 8);24
define the true estimated ultimate recovery (EUR) or

reserves volumes to be the 50-year cumulative production
volume.
Even though the exponential and hyperbolic relations (as
provided by Arps) are empirically derived, the Arps decline
curves are excellent tools for estimating reserves when applied
under the correct conditions. Unfortunately, many of us in the
industry either have forgotten or have chosen to ignore the
conditions under which use of the Arps decline curves are
appropriate. Application of a decline curve methodology
using the Arps models implicitly assumes the following:25
z Extrapolation of the best-fit curve through the current or
historical production data is an accurate model for future
production trends;
z There will be no significant changes in current operating
conditions or field development that might affect the curve fit
and the subsequent extrapolation into the future;
z The well is producing against a constant bottomhole flowing
pressure; and
z Well production is from an unchanging drainage area with noflow boundaries (i.e., the well is in boundary-dominated
(stabilized) flow).
Traditional decline curve analysis is based on the general form

of the Arps1 rate-time (hyperbolic) decline equation:
q (t ) =
qi
[1 + bDit ]1/b
.........................................................(1)
where Di is the initial decline rate, qi is the gas flow rate, and b
is Arps decline curve constant (or decline exponent). Note
that the units of these three variables must be consistent.
Equation 1 has three different forms exponential, harmonic,
and hyperbolic depending on the value of the b-exponent.
Each of these equations has a different shape on Cartesian and
semilog graphs of gas production rate versus time and
cumulative gas production.
The exponential or constant-percentage decline case is a
special case of Eq. 1 where the b-exponent is zero, and is
characterized by a decrease in production rate per unit of time
that is directly proportional to the production rate. The exponential decline equation (b-exponent of zero) is written as:
q(t ) = qi exp( Dit ) ...........................................................(2)
Similarly, the harmonic decline is also a special case of Eq. 1

(b-exponent equals one), and is written as:
q (t ) =
qi
[1 + Dit ]
.............................................................(3)
The Arps hyperbolic decline is given by the general form (Eq.

1) and is "valid" for any condition where the b-exponent varies
between 0 and 1.0. We should note that the value of the bexponent determines the degree of curvature of the semilog
decline plot (log(q) versus t), ranging from a straight line with
b=0 to increasing curvature as b increases. Although there is
no theoretical basis, Arps1 indicated the b-exponent should lie
between 0 and 1.0 but he offered no justification of the
possibility that b might be greater than one. Therefore,
variations in the computed b-exponents outside of the
expected range suggests the Arps' rate-time relationships may
SPE 109625
not be valid for modeling the decline behavior of tight gas

sands at HP/HT conditions.
Effect of Reservoir Properties on Production Decline
We first evaluated the effects of various reservoir properties
and heterogeneities including vertical heterogeneity from
layering, permeability contrast among layers, horizontal
permeability anisotropy, and stress-dependent reservoir
properties on the production decline behavior. Decline
characteristics were quantified and compared using primarily
the Arps decline exponent, b.
Layering and Permeability Contrast Among Layers. The
simulated short- and long-term production profiles for the
single- and multi-layer cases are shown in Figs. 10 and 11,
respectively. Recall that the single-layer case assumes a
homogeneous, isotropic reservoir with absolute permeability
and effective porosity equal to the "thickness-averaged" values
of the multi-layer cases (Tables 3-5). Other reservoir and
hydraulic-fracture input properties used to simulate the
production profiles are summarized in Table 6.
Table 6 Summary of reservoir and hydraulic-fracture input
properties used to simulate production profiles.
Reservoir Property
Well spacing
Initial pore pressure gradient
Initial bottomhole pressure
Bottomhole temperature
Thickness-averaged absolute permeability
Thickness-averaged effective porosity
Thickness-averaged water saturation
Vertical to horizontal permeability ratio
Gross sand thickness
Stress-dependent properties
Hydraulic Fracture Property
Effective half-length
Absolute permeability
Numerical dimensionless fracture conductivity
Vertical to horizontal permeability ratio
Equal wing half-length
Stress-dependent properties
Value
80 acres/well
0.90 psi/ft
16,200 psia
o
400 F
0.0090 md
6.61%
36.4%
0.001
200 ft
No
Value
300 ft
100 md
18.5
1.0
Yes
No
Inspection of the simulated production profiles in Figs. 10 and

11 shows that the single-layer case has a much sharper initial
decline (Dei = 71 days-1) than the 4-, 8- and 16-layer cases (Dei
= 69, 67 and 66 days-1, respectively). Figure 10 also shows
that, after the initial steep decline, the single-layer case has the
flattest production profile during the first 5 years of
production. However, the multilayer cases begin to flatten
substantially after that period. In fact, the multilayer cases
display more of the long-term, hyperbolic decline shape
characteristic of tight gas sands (Fig. 11).
All of these observations concerning the decline behavior are
confirmed by the b-exponents (Table 7) which were computed
from the best-fit Arps1 model through the simulated
production for producing times of 1, 5, 10, 20 and 50 years.
Surprisingly, all b-exponents lie approximately in the range
suggested by Arps1 (i.e., 0 < b < 1) after 50 years of
production.
Based on the results shown in Table 7, it also appears as if the
degree of heterogeneity affects the long-term decline behavior
more than just low permeability i.e., low permeability alone
SPE 109625
is not necessarily sufficient to generate large b-exponents

frequently observed in tight gas sands. Recall the Modified
Stratigraphic Lorenz plot shown in Fig. 3 which indicates that
the 4- and 8-layer cases are the most heterogeneous based on
their deviations from the 45-degree line (representing a homogeneous system). We also see that these two cases have the
largest b-exponents during the last 30 years of production.
Conversely, both the computed b-exponents and the Modified
Stratigraphic Lorenz plot suggest that, even though it has more
layers, the 16-layer case "behaves" more like the single-layer
case than either of the other multi-layer cases.
Gas Production Rate, Mscf/day
1-Layer Case
4-Layer Case
8-Layer Case
16-Layer Case
1,000
100
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
Producing Time, years
Fig. 10 Simulated short-term production profiles for 1-, 4-,

8-, and 16-layer cases.
1-Layer Case
Table 8 Computed reserve estimate errors for 1-, 4-, 8-,

and 16-layer case.
Producing
Time Period
(years)
1
5
10
20
1-Layer
Case
(%)
131.6
58.0
19.2
0.1
4-Layer
Case
(%)
109.6
21.7
7.8
7.3
8-Layer
Case
(%)
117.2
11.1
8.6
7.2
16-Layer
Case
(%)
127.8
20.4
10.9
2.7
As discussed previously, use of the Arps rate-time equations is

strictly valid only when the well has reached true boundarydominated flow conditions. Consequently, we attribute the
large b-exponents shown in Table 8 as well as the large
reserve estimate errors during the early producing times to
incorrect use of the Arps equations during transient and/or
transitional flow periods.
To validate our hypothesis, we monitored and recorded layer
pressures at several positions in the reservoir during the
simulation process. In Fig. 12, we show the layer pressures
for the 4-layer case. Pressures were monitored in each layer at
a point located on the edge of the 80-acre drainage area
directly opposite the wellbore and aligned with the axis of the
hydraulic fracture (i.e., monitoring point no. 1, Fig. 1). Note
that HFU 1 and 2 (the two most permeable layers) show the
earliest and largest pressure reductions, while HFU 3 exhibits
less of a pressure decline. We also observe that the least
permeable layer (HFU 4) does not experience an appreciable
pressure reduction for the first ten years of production.
10,000
4-Layer Case
8-Layer Case
16-Layer Case
1,000
(Table 8) between the estimates and "true value" generally

decline as more production is available for the history match.
Additionally, we generally found a strong correlation between

the computed b-exponents and reserve estimate errors i.e.,
larger reserve estimate errors corresponding to larger bexponents. This correlation suggests that computed bexponents which lie significantly outside the range suggested
by Arps1 (i.e., 0 < b < 1) are not representative of future
decline behavior. Smaller reserve estimate errors later in the
well's productive life also indicate that we may be approaching
the "true" b-exponent for each particular case.
10,000
0.0
100
10
0
10
15
20
25
30
35
40
45
50
20,000
Table 7 Computed b-exponents for 1-, 4-, 8-, and 16-layer

cases.
Producing
Time Period
(years)
1
5
10
20
50
1-Layer
Case
3.62
2.95
1.48
0.58
0.44
4-Layer
Case
2.78
1.35
1.04
1.01
1.00
8-Layer
Case
2.89
1.30
1.18
1.04
1.02
16-Layer
Case
2.97
1.39
1.26
0.96
0.82
We also estimated reserves from an extrapolation of the bestfit Arps1 model through the simulated production and at each
of the time periods shown in Table 7. We then compared
those estimates to the "true" reserve value for each respective
multi-layer case. Not unexpectedly, the differences or errors
Bottomhole Pressure, psia
18,000
Fig. 11 Simulated long-term production profiles for 1-, 4-, 8,

and 16-layer cases
16,000
14,000
12,000
10,000
8,000
6,000
4,000
2,000
0
0
10
15
20
25
30
35
40
45
50

HFU 1
HFU 2
HFU 3
HFU 4
Fig. 12 Layer pressures at monitoring point no. 1 for the 4layer case.
We also monitored the change in reservoir pressures at a point

located on the edge of the 80-acre drainage area, directly
opposite the wellbore but in a direction perpendicular to the
hydraulic fracture axis (i.e., monitoring point no. 2, Fig. 1).

Figure 13 shows that HFU 1 and 2 again exhibit the largest
and earliest pressure reduction. However, unlike pressure
monitoring point no. 1 at which we saw measurable pressure
reductions in the first six months, we observed no significant
pressure changes in HFU 3 for the first two years. Most
significantly, we see very little change in reservoir pressure in
HFU 4 for the first 25 years of production.
SPE 109625
HFU 1 and 2 provides pressure support and helps maintain

higher rates as well as the shallower decline profile during this
early time period. However; after 15 to 20 years of
production, the two cases with lower kv/kh values begin to
flatten more than the kv/kh = 0.1 case. Moreover, the lower
kv/kh cases display more of the long-term hyperbolic shape
characteristic of tight gas sands (Fig. 15).
10,000
kv/kh=0.001
20,000
18,000
16,000
14,000
12,000
10,000
8,000
6,000
kv/kh=0.01
kv/kh=0.1
1,000
4,000
2,000
100
0.0
0
10
15
20
25
30
35
40
45
0.5
1.0
1.5
50
2.0
2.5
3.0
3.5
4.0
4.5
5.0

HFU 1
HFU 2
HFU 3
HFU 4
Fig. 13 Layer pressures at monitoring point no. 2 for the 4layer case.
Fig. 14 Simulated short-term production profiles for kv/kh =

0.001, 0.01 and 0.10.
10,000
Since HFU 4 does not represent a significant percentage

(approximately 13 percent) of the total hydrocarbon pore
volume, it apparently does not affect the overall decline
behavior as significantly as HFUs 1-3. Furthermore, the lack
of boundary-dominated flow in HFU 4 does not prevent the
overall well production decline from "behaving" as if it were
in true boundary-dominated flow conditions after 20 years of
production. We observed similar pressure distributions for the
8- and 16-layer cases.
Vertical-to-Horizontal Permeability Ratio. All of the
simulated production profiles shown in Figs. 10 and 11 were
generated with a vertical-to-horizontal permeability ratio, kv/kh
= 0.001. In this section, we investigate the effects of other
values of kv/kh. Figures 14 and 15 compare the short- and
long-term production profiles, respectively, for the 4-layer
case but with kv/kh values of 0.001, 0.01, and 0.1. Other than
variations in kv/kh, reservoir and hydraulic fracture properties
shown in Table 6 were used as additional input for all cases.
All of the short-term production profiles (Fig. 14) have very
similar initial decline rates (Di =68.8, 66.6, and 60.8 day-1 for
kv/kh = 0.001, 0.01 and 0.1, respectively). However, the
production profile for kv/kh = 0.1 is much flatter than the other
two cases after the initial decline and during the first five years
of production. Vertical flow from less permeable layers into
kv/kh=0.001
Comparison of the layer pressure responses in Figs. 12 and 13

demonstrates several points. First, even in reservoirs with no
permeability anisotropy, we observe an elliptical flow
geometry (with the major ellipse axis centered along the
fracture axis). Secondly, variations in layer flow capacities
cause unequal growth of the elliptical drainage patterns.
Finally, the pressure responses in HFU 4 suggest that the well
did not reach true boundary-dominated flow but was in either
transient or transitional flow for more than 20 years.
kv/kh=0.01
kv/kh=0.1
1,000
100
10
0
10
15
20
25
30
35
40
45
50
Fig. 15 Simulated long-term production profiles for kv/kh =

0.001, 0.01 and 0.10.
The observed decline characteristics are reflected by the

computed b-exponents (Table 9). The b-exponents for the
kv/kh = 0.001 and 0.01 cases are very similar for the entire time
period evaluated. However, b-exponents for the kv/kh = 0.1
cases are generally larger for the first ten years, but begin to
decrease more rapidly than the other cases after that time
period. Again, we observe that the long-term b-exponents are
approaching values between approximately 0.5 and 1.0.
Table 9 Computed b-exponents for kv/kh = 0.001, 0.01 and
0.10.
Producing
Time Period
(years)
1
5
10
20
50
kv/kh
= 0.001
2.78
1.35
1.04
1.01
1.00
kv/kh
= 0.01
2.95
1.37
1.19
1.06
0.95
kv/kh
= 0.1
3.06
1.43
0.76
0.66
0.51
SPE 109625
The differences (or errors) between the 50-year cumulative

production volume and the reserves estimated from
extrapolation of the best-fit Arps decline curve through the
simulated production data are summarized in Table 10.
Similar to the results shown in Table 8, we note a general
correlation between b-exponents and reserve estimate errors
i.e., the largest errors are associated with the largest bexponents. We also observe that the largest errors occur for
cases where there is less than 10 years of production.
The pressure behavior from both Figs. 16 and 17 suggests the

kv/kh = 0.1 case may be in or is approaching boundarydominated flow after less than five years of production, while
the lower value of kv/kh delays the onset of true boundarydominated flow conditions for more than 20 years.
Differences between the duration of transient and transitional
flow periods helps to explain both the smaller reserve estimate
errors as well as the smaller values of the computed bexponent for the kv/kh = 0.1 case.
Table 10 Computed reserve estimate errors for kv/kh =

0.001, 0.01 and 0.10.
Horizontal Permeability Anisotropy. All of the previous

simulated performance cases addressed the impact of vertical
heterogeneities (i.e., layering, permeability contrast among
layers, and kv/kh) on the production decline in these cases
no horizontal permeability anisotropy was considered (i.e., kx
= ky). In this section we evaluate the effects of horizontal
permeability anisotropy as quantified by various values of
ky/kx. Specifically, we evaluated ky/kx = 0.1, 1.0 and 10.
Except for variations in ky/kx, all reservoir and hydraulic
fracture properties shown in Table 6 were used as additional
input for the simulated cases.
Producing
Time Period
(years)
1
5
10
20
kv/kh
= 0.01
(%)
128.5
22.8
8.4
7.9
kv/kh
= 0.001
(%)
109.6
21.7
7.8
7.3
kv/kh
= 0.1
(%)
130.6
30.8
6.8
3.8
The differences in decline behavior for the various values of

kv/kh can again be explained by the simulated layer pressures.
Figure 16 shows the individual layer pressures at monitoring
point no. 1 for kv/kh = 0.001 (color-filled triangles) and kv/kh =
0.1 (color-filled circles). Larger values of kv/kh allow more
effective drainage of all layers resulting in larger pressure
reductions throughout the entire drainage area, including HFU
4 (the least permeable layer). Figure 17 shows similar results
at monitoring point no. 2.
20,000
18,000
16,000
14,000
12,000
10,000
8,000
6,000
4,000
2,000
0
0
10
15
20
25
30
35
40
45
In previous cases which considered vertical heterogeneity, we

observed much less pressure reduction in the layers with the
lowest permeability in a direction perpendicular (y-direction)
to the hydraulic fracture axis, so we would expect variations in
ky/kx to impact the production decline and pressure depletion
even more significantly. Figures 18 and 19 compare the
simulated short- and long-term production profiles,
respectively, for the 4-layer case and with ky/kx values of 0.1,
1.0, and 10. Smaller values of ky/kx result in higher initial
decline rates (i.e., Dei = 76, 69, and 63 days-1 for ky/kx of 0.1,
1.0, and 10, respectively). Conversely, larger values of ky/kx
cause much sharper declines throughout most of the
productive life. This decline behavior is confirmed by the
computed b-exponents (Table 11) which are consistently
larger for the cases simulated with smaller ky/kx values. In
fact, we see b-exponents greater than two for the first ten years
of production for ky/kx = 0.1.
50
100,000

HFU 2, kv/kh=0.1
HFU 3, kv/kh=0.1
HFU 4, kv/kh=0.1
HFU 1, kv/kh=0.001
HFU 2, kv/kh=0.001
HFU 3, kv/kh=0.001
HFU 4, kv/kh=0.001
ky/kx=0.1
Fig. 16 Comparison of layer pressures for 4-layer case at

monitoring point no. 1 for kv/kh = 0.001 and 0.10.
20,000
18,000
16,000
14,000
HFU 1, kv/kh=0.1
ky/kx=1.0
10,000
ky/kx=10
1,000
100
12,000
10,000
10
0.0
8,000
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
6,000
4,000
Fig. 18 Simulated short-term production profiles for 4-layer

case and for ky/kx = 0.10, 1.0 and 10.
2,000
0
0
10
15
20
25
30
35
40
45
50

HFU 1, kv/kh=0.1
HFU 2, kv/kh=0.1
HFU 3, kv/kh=0.1
HFU 4, kv/kh=0.1
HFU 1, kv/kh=0.001
HFU 2, kv/kh=0.001
HFU 3, kv/kh=0.001
HFU 4, kv/kh=0.001
Fig. 17 Comparison of layer pressures for 4-layer case at

monitoring point no. 2 for kv/kh = 0.001 and 0.10.
Differences between the 50-year cumulative production

volume and the reserves estimated from the best-fit Arps
decline curve through the simulated data for ky/kx values of
0.1, 1.0 and 10 are summarized in Table 12. Again, we note
that there is a direct association between the largest reserve
10
estimate errors and the largest computed b-exponents.

Further, the largest reserve estimate errors generally occur
when there is less than 10 years of production and before
reaching boundary-dominated flow conditions.
SPE 109625
and hydraulic fracture properties shown in Table 6 as input for

the simulation.
20,000
18,000
10,000
ky/kx=0.1
ky/kx=1.0
ky/kx=10
1,000
16,000
14,000
12,000
10,000
8,000
6,000
4,000
2,000
100
0
0
10
15
20
25
30
35
40
45
50
10
0
10
15
20
25
30
35
40
45
50
HFU 1, ky/kx=0.1
HFU 2, ky/kx=0.1
HFU 3, ky/kx=0.1
HFU 4, ky/kx=0.1
HFU 1, ky/kx=1
HFU 2, ky/kx=1
HFU 3, ky/kx=1
HFU 4, ky/kx=1
Fig. 20 Comparison of layer pressures at monitoring point

no. 1 for ky/kx = 0.10 and 1.
Fig. 19 Simulated long-term production profiles for 4-layer

case and for ky/kx = 0.10, 1.0 and 10.
20,000
Producing
Time Period
(years)
1
5
10
20
50
ky/kx
= 0.1
3.00
2.75
2.26
1.33
1.13
ky/kx
= 1.0
2.78
1.35
1.04
1.01
1.00
ky/kx
= 10
2.17
1.06
0.96
0.92
0.76
18,000
Table 11 Computed b-exponents for ky/kx = 0.10, 1.0

and 10.
16,000
14,000
12,000
10,000
8,000
6,000
4,000
2,000
0
0
Table 12 Computed reserve estimate errors for ky/kx

= 0.10, 1.0 and 10.
Producing
Time Period
(years)
1
5
10
20
ky/kx
= 0.1
(%)
134.6
53.1
34.9
18.9
ky/kx
= 1.0
(%)
109.6
21.7
7.8
7.3
ky/kx
= 10
(%)
102.0
11.1
7.5
7.1
These observations are confirmed by the simulated layer

pressure responses shown in Figs. 20 and 21 at monitoring
points 1 and 2, respectively. The curves compare layer
pressures for ky/kx = 0.1 (color-filled triangles) and ky/kx = 1
(color-filled circles). The pressure responses demonstrate that
lower values of ky/kx delay the onset of boundary-dominated
flow even more than in the isotropic case. In fact, neither
HFU 3 nor 4 exhibit measurable pressure reductions at
pressure monitoring point no. 2 for at least the first 20 years of
production for the ky/kx = 0.1 case. We should note that the
absence of boundary-dominated flow conditions during the
first 30 years of production for ky/kx = 0.1 also corresponds to
the larger errors shown in Table 12.
Stress-Dependent Reservoir Properties. All tight gas sands
exhibit some stress-dependent characteristics. Typically, reductions in permeability are much greater than that for
porosity. For this paper, we evaluated the impact of stressdependent reservoir properties on the production decline using
the functions shown in Figs. 4 and 5. Except for the inclusion
of stress-dependent reservoir properties, we used the reservoir
10
15
20
25
30
35
40
45
50

HFU 1, ky/kx=0.1
HFU 2, ky/kx=0.1
HFU 3, ky/kx=0.1
HFU 4, ky/kx=0.1
HFU 1, ky/kx=1
HFU 2, ky/kx=1
HFU 3, ky/kx=1
HFU 4, ky/kx=1
Fig. 21 Comparison of layer pressures at monitoring point

no. 2 for ky/kx = 0.10 and 1.
We compared simulated production profiles for all multilayered cases with and without stress-dependent permeability and porosity and we saw little or no differences in the
decline behavior. Apparently, the vertical heterogeneity
caused by reservoir layering and permeability contrast among
layers has a much larger impact on the decline behavior.
These observations are based on results generated using the
functions shown in Figs. 4 and 5, so we should caution that
results may be different with other stress-dependent functions.
The greatest impact occurred in the single-layer, homogeneous, and isotropic case. Following very similar initial
declines, the inclusion of stress-dependent properties caused
the production profile (Fig. 22) to be much flatter during the
first ten years of production, but to decline much faster after
that time period. Although not shown, the computed bexponents match our observations. We also computed larger
reserve estimate errors than those summarized in Table 7
during the first ten years of production when stress-dependent
properties were included.
Lateral Rock Heterogeneity. We frequently encounter
variations in rock properties in low-permeability reservoirs in
the lateral direction (i.e., x- and y-directions) caused by
differential diagenetic events following deposition. Dia-
SPE 109625
genesis defined as any physical or chemical process

causing changes in initial rock properties is the principal
source of reductions in both permeability and porosity in tight
gas sands. The primary diagenetic processes typically seen in
tight gas sands are mechanical and chemical compaction,
quartz and other mineral cementation, mineral dissolution, and
clay genesis. Diagenesis may affect the rock properties
sufficiently enough to cause the rock to act as flow baffles,
and in extreme cases, may create flow barriers.
summarized in Table 6 were used as input for the simulation

cases.
Effective Fracture Half-Length. We first examined the
influence of effective or propped fracture half-length on production decline behavior. All of the previous simulation
studies were generated with an effective hydraulic fracture
half-length of 300 ft, but we also evaluated effective halflengths, Lf = 50, 100, and 500 ft in our study.
The simulated short- and long-term production declines for the
4-layer case with Lf = 50, 100, 300 and 500 ft are shown in
Figs. 23 and 24, respectively. As expected, the cases with
shorter effective fracture half-lengths exhibit steeper initial
declines but are followed by flatter profiles during the early
years of production. Conversely, cases with longer effective
fracture half-lengths display a more gradual initial decline that
is followed by sharper production decline profile.
10,000
with stress-dependent properties
w/o stress-dependent properties
1,000
100
10,000
10
0
10
15
20
25
30
35
40
45
50
Fig. 22 Effects of stress-dependent reservoir properties on

long-term production decline behavior for the
single-layer case.
To evaluate the effects of diagenesis, we randomly populated

the model grid cells in each layer with variable effective
porosity and absolute horizontal permeability values. Each
random distribution was constrained by the maximum and
minimum range of values observed for each HRT shown in
Fig. 2. Moreover, the random property distributions for each
layer or HFU were generated so that the averages equaled the
individual layer values shown in Tables 3-5 for the 4-, 8, and
16-layer cases, respectively. Maintaining these average
properties throughout this work allowed us to make relevant
comparisons among the various cases.
Except for the inclusion of the heterogeneous distribution of
properties, we used the reservoir and hydraulic fracture
properties shown in Table 6 as input for the simulation. We
compared simulated production profiles for all layered cases
with and without heterogeneity and we observed no
significant differences. Similar to the evaluation of stressdependent reservoir properties, it appears as if the vertical
heterogeneity caused by reservoir layering and permeability
contrast among layers has a much bigger effect on the decline
behavior than does the lateral heterogeneity.
Effect of Fracture Properties on Production Decline
In this section, we evaluate the effects of various hydraulic
fracture properties and heterogeneities including variable
effective fracture lengths and conductivities, unequal fracture
lengths ("wings"), stress-dependent fracture properties, and
two-phase and non-Darcy flow on the production decline
behavior. Decline characteristics were again quantified and
compared using the Arps decline exponent, b. Except for
variations in specific hydraulic fracture properties being
evaluated, all reservoir and hydraulic fracture properties
Lf=50 ft
Lf=100 ft
Lf=300 ft
Lf=500 ft
1,000
100
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
Fig. 23 Simulated short-term production profiles for Lf = 50,

100, 300 and 500 ft.
10,000
Lf=50 ft
11
Lf=100 ft
Lf=300 ft
Lf=500 ft
1,000
100
10
0
10
15
20
25
30
35
40
45
50
Fig. 24 Simulated long-term production profiles for Lf = 50,

100, 300 and 500 ft.
Although the initial declines are much steeper, the long-term

declines (as illustrated by Fig. 24 for the shorter effective
hydraulic half-lengths) seem to be slightly flatter for most of
the productive period. Differences between decline profiles
after about ten years of production are, however, very small.
Our observations for the decline behavior are validated by
examination of the computed b-exponents (see Table 13). Although the b-exponents are consistently larger for cases with
shorter fracture half-lengths, all b-exponents approach a value
Table 13 Computed b-exponents for Lf = 50, 100, 300 and

500 ft.
Producing
Time Period
(years)
1
5
10
20
50
Lf =
50 ft
4.01
1.53
1.10
1.07
1.02
Lf =
100 ft
3.60
1.44
1.08
1.06
1.01
Lf =
300 ft
2.78
1.35
1.04
1.01
1.00
Lf =
500 ft
1.91
1.20
1.03
0.96
0.97
Reserve estimate errors for various time periods and computed

from the best-fit Arps1 decline curve through the simulated
data for Lf values of 50, 100, 300 and 500 ft are summarized in
Table 14. We again see that there is a correlation between the
largest reserve estimate errors and the largest computed bexponents shown in Table 13. Interestingly, the reserve
estimate errors are less than 10 percent after ten years of
production. These small errors suggest that each case may be
approaching the "true" b-exponent for all fracture half-lengths
investigated in this study.
Table 14 Computed reserve estimate errors for Lf = 50,
100, 300 and 500 ft.
Producing
Time Period
(years)
1
5
10
20
Lf =
50 ft
(%)
144.6
33.8
14.2
11.3
Lf =
100 ft
(%)
139.3
24.5
9.7
9.3
Lf =
300 ft
(%)
109.6
21.7
7.8
7.3
Lf =
500 ft
(%)
78.0
12.0
6.5
6.2
Figures 25 and 26 show the simulated layer pressures at

monitoring points no. 1 and 2, respectively, for Lf = 50 ft
(color-filled circles) and Lf = 500 ft (color-filled triangles). At
pressure monitoring point no. 1 (Fig. 25), we see that (as
expected) the wells with longer effective fracture half-lengths
will recover gas more effectively in all layers. We even see
significant pressure reductions in the least permeable layer
(HFU 4) after less than 2 years of production for the Lf = 500
ft case. Even though it is delayed for almost 20 years, we also
see small but measurable pressure reductions in HFU 4 for the
Lf = 50 ft case.
Similar to the results from our investigation of reservoir properties on production decline behavior, we also see evidence
of an elliptical flow geometry from the pressure responses
shown at pressure monitoring point no. 2 in Fig. 26. Although
the layer pressure behavior in HFU 1 and 2 are comparable,
we see no measurable pressure reduction in HFU 4 for either
the Lf = 50 or 500 ft case. More significantly, we see
essentially no pressure reduction for either fracture half-length
case after almost 30 years of production.
Based on the pressure responses shown in Figs. 25 and 26, we
offer several conclusions. First, we observe an elliptical flow
geometry for a wide range of effective fracture half-lengths.
The flow duration and geometrical parameters may be
different, but all cases exhibit elliptical flow. Secondly, the
pressure responses in HFU 4 for both the Lf = 50 or 500 ft
SPE 109625
cases suggest that both cases were either in transient or

transitional flow during the first 30 years of production. It
does appear, however, that both cases had either reached or
were approaching boundary-dominated flow. Again, since
HFU 4 does not represent a significant percentage of the total
hydrocarbon pore volume, it apparently does not affect the
overall decline behavior as significantly as HFUs 1-3.
Consequently, the lack of boundary-dominated flow in HFU 4
does not prevent the well from "behaving" as if it were in true
boundary-dominated flow conditions after 30 years.
20,000
18,000
of one after 50 years of production. Unlike previous cases, we

also observe b-exponents much greater than 3.0 for fracture
half-lengths of 50 and 100 ft during the first year of
production.
16,000
14,000
12,000
10,000
8,000
6,000
4,000
2,000
0
0
10
15
20
25
30
35
40
45
50

HFU 1, Lf = 50 ft
HFU 1, Lf = 500 ft
HFU 2, Lf = 50 ft
HFU 2, Lf = 500 ft
HFU 3, Lf = 50 ft
HFU 3, Lf = 500 ft
HFU 4, Lf = 50 ft
HFU 4, Lf = 500 ft
Fig. 25 Comparison of layer pressures at pressure monitoring point no. 1 for Lf = 50 and 500 ft.
20,000
18,000
12
16,000
14,000
12,000
10,000
8,000
6,000
4,000
2,000
0
0
10
15
20
25
30
35
40
45
50

HFU 1, Lf = 50 ft
HFU 2, Lf = 50 ft
HFU 3, Lf = 50 ft
HFU 4, Lf = 50 ft
HFU 1, Lf = 500 ft
HFU 2, Lf = 500 ft
HFU 3, Lf = 500 ft
HFU 4, Lf = 500 ft
Fig. 26 Comparison of layer pressures at pressure monitoring point no. 2 for Lf = 50 and 500 ft.
Absolute Fracture Conductivity. We next evaluated the

effects of absolute fracture conductivity, wfkf, on the
production decline behavior. All of the previous simulated
performance cases were generated using an absolute fracture
conductivity, wfkf = 50 md-ft (i.e., a numerical FCD = 18.5).
Figures 27 and 28 compare the short- and long-term
production decline profiles, respectively, for wfkf = 0.5, 5.0, 50
and 500 md-ft (i.e., numerical FCD = 0.185, 1.85, 18.5, and
185, respectively). Cases with lower effective fracture conductivities exhibit much steeper initial declines followed by
flatter profiles during the early years of production, while
cases with higher effective fracture conductivities display a
gradual initial decline followed by a sharper long-term
production decline profile.
We also note that cases with numerical conductivities of 50
and 500 md-ft are almost identical. These similarities suggest
SPE 109625
the decline behavior for all values of infinitely conductive

fractures will be similar, while the production profiles for
fractures of (low) finite conductivity will vary depending on
the effective conductivity.
10,000
13
largest reserve estimate errors and the largest computed bexponents shown in Table 15. We also note the reserve
estimate errors are less than 10 percent after ten years of
production. These small errors also suggest that each case
may be approaching the "correct" b-exponent for all fracture
conductivities investigated in this study.
Table 16 Computed reserve estimate errors for wfkf = 0.5,
5, 50 and 500 md-ft.
wfkf=5 md-ft
wfkf=50 md-ft
wfkf=500 md-ft
Producing
Time Period
(years)
1
5
10
20
1,000
100
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
Fig. 27 Simulated short-term production profiles for wfkf =

0.5, 5, 50 and 500 md-ft.
10,000
wfkf=0.5 md-ft
wfkf=5 md-ft
wfkf=50 md-ft
100
10
5
10
15
20
25
30
35
40
45
50
Fig. 28 Simulated long-term production profiles for wfkf =

0.5, 5, 50 and 500 md-ft.
The computed b-exponents (Table 15) reflect the decline

behavior shown in Figs. 25 and 26. Similar to those cases
with short effective fracture half-lengths shown previously in
Table 13, cases with low fracture conductivities have bexponents greater than 3 for the first year of production. We
also note that all b-exponents equal or are approaching one
after 50 years of production. In fact, the computed bexponents change little after about ten years of production,
which suggests those cases are approaching boundarydominated flow conditions.
Table 15 Computed b-exponents for wfkf = 0.5, 5, 50 and
500 md-ft.
Producing
Time Period
(years)
1
5
10
20
50
wfkf =
0.5 md-ft
4.34
1.87
1.29
1.14
1.05
wfkf =
5 md-ft
3.71
1.58
1.07
1.04
1.02
wfkf =
50 md-ft
2.78
1.35
1.04
1.01
1.00
wfkf =
5 md-ft
(%)
135.2
38.1
8.2
7.7
wfkf =
50 md-ft
(%)
109.6
21.7
7.8
7.3
wfkf =
500 md-ft
(%)
107.8
10.6
7.6
7.2
Unequal Fracture Wing Lengths. In this section, we

evaluated the effects of unequal hydraulic fracture half-lengths
on the production decline behavior. All previous simulations
were conducted assuming that the fracture "wings" extended
equal distances on either side of the wellbore; however,
heterogeneities in rock properties may cause fractures to grow
unequally during the stimulation treatment.
wfkf=500 md-ft
1,000
wfkf =
0.5 md-ft
(%)
138.6
41.8
10.5
8.2
Although not shown, we also monitored the simulated layer

pressures for each of the fracture conductivities that were
modeled. We observed similar pressure responses as those
shown in Figs. 25 and 26 for various effective fracture halflengths. In general, the pressure responses for fracture cases
with higher conductivities behaved similarly to those for
longer, effective fracture half-lengths, while fractures with
lower fracture conductivities behaved like wells with short,
effective fracture half-lengths.
wfkf =
500 md-ft
2.34
1.14
1.02
1.01
1.00
Reserve estimate errors for various time periods are summarized in Table 16. We again see a correlation between the
For this study, we modeled wing length ratios ranging from

1.0 (i.e., equal lengths), 1.5, 3 and 6.0. Results from our study
indicate that variations in fracture wing ratios do not significantly affect long-term production profiles but this variable
will cause variations in the short-term decline behavior.
Figure 29 illustrates the decline behavior for the first five
years. In general, larger wing length ratios exhibit steeper
initial declines and flatter profiles during this time period.
10,000
Lf1/Lf2=6
wfkf=0.5 md-ft
Lf1/Lf2=3
Lf1/Lf2=1.5
Lf1/Lf2=1
1,000
100
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
Fig. 29 Simulated short-term production profiles for Lf1/Lf2

= 1.0, 1.5, 3.0 and 6.0.
Computed b-exponents are summarized in Table 17 for the

same wing length ratios shown in Fig. 29. We see larger wing
length ratios result in larger b-exponents for the first five years
of production. However, we also notice that the b-exponents

for all wing length ratios start to approach a value of 1.0 after
ten years of production. In fact, the values change little for the
last 30 years of production.
Table 17 Computed b-exponents for hydraulic fracture
wing length ratios of 1, 1.5, 3 and 6.
Producing
Time Period
(years)
1
5
10
20
50
Lf1/Lf2
= 1.0
2.78
1.35
1.04
1.01
1.00
Lf1/Lf2
= 1.5
2.91
1.37
1.11
1.01
1.01
Lf1/Lf2
= 3.0
3.05
1.38
1.12
1.03
1.03
Lf1/Lf2
= 6.0
3.23
1.40
1.12
1.03
1.03
The computed reserve estimate errors are compiled in Table

18. We observe the same relationships that we have seen in
all of our previous results between the computed b-exponents
and reserve estimate errors. The largest reserve estimate
errors occur during the first five years of production when the
wells are experiencing either transient or transitional flow
conditions.
SPE 109625
choked fracture cases with kf1/kf2 < 1 have b-exponents

significantly greater than 3.0 during the first year. As we
would expect, the corresponding reserve estimate errors
(Table 20) exceed 100 percent. We also note that the bexponents for the kf1/kf2 = 1 case tend to "stabilize" after about
10 years of production, while the choked fracture cases
continue to change after that time period. All cases, however,
equal or approach 1.0 after 50 years of production.
10,000
kf1/kf2=0.01
14
kf1/kf2=0.10
kf1/kf2=1.0
1,000
100
0.0
0.5
1.0
1.5
Table 18 Computed reserve estimate errors for Lf1/Lf2 = 1,

1.5, 3 and 6.
Lf1/Lf2
= 1.0
(%)
109.6
21.7
7.8
7.3
Lf1/Lf2
= 1.5
(%)
133.9
22.9
10.0
7.3
Lf1/Lf2
= 3.0
(%)
129.4
23.4
10.1
7.6
Lf1/Lf2
= 6.0
(%)
135.6
23.6
10.1
7.7
Although not shown, the layer responses at pressure monitoring points 1 and 2 confirm that HFUs 1-3 have experienced
significant pressure depletion during the first 20 years of
production. Similar to previous results, HFU 4 does not
experience much pressure reduction for the first 20 years
(especially at monitoring point no. 2). But, since HFU 4
represents a much smaller percentage of the total connected
hydrocarbon pore volume than HFUs 1-3, the overall well
decline stabilizes and behaves as if it were in true boundarydominated flow. This explains why the computed b-exponents
shown in Table 17 appear to be stabilizing after 20 years of
production.
Variable Fracture Conductivity (Choked Fracture Case).
Next, we considered the effects of variable fracture
conductivities, or more specifically, the "choked" fracture case
in which the fracture conductivity near the wellbore is lower
than in the fracture towards the tip. The variable fracture conductivity is quantified by the ratio of fracture conductivities,
kf1/kf2, where kf1 and kf2 are the near (wellbore) and far (field)
fracture conductivities, respectively. Figures 30 and 31 present the short- and long-term production profiles, respectively,
for kf1/kf2 = 0.01, 0.10 and 1.0. As illustrated in Fig. 30, the
effect of lower fracture conductivity in the near-wellbore area
causes steep initial declines in the production rates followed
by relatively flat profiles for the first five years. Long-term
production profiles also tend to remain flat for the entire
producing period shown in Fig. 31.
Computed b-exponents for the production profiles shown in
Figs. 30 and 31 are summarized in Table 19. Both of the
2.5
3.0
3.5
4.0
4.5
5.0
Fig. 30 Simulated short-term production profiles for kf1/kf2 =

0.01, 0.10 and 1.0.
10,000
kf1/kf2=0.01
Producing
Time Period
(years)
1
5
10
20
2.0
kf1/kf2=0.10
kf1/kf2=1.0
1,000
100
10
0
10
15
20
25
30
35
40
45
50
Fig. 31 Simulated long-term production profiles for kf1/kf2 =

0.01, 0.10 and 1.0.
Table 19 Computed b-exponents for kf1/kf2 = 0.01,
0.10 and 1.0.
Producing
Time Period
(years)
1
5
10
20
50
kf1/kf2
= 0.01
4.73
1.69
1.22
1.15
1.01
kf1/kf2
= 0.10
4.20
1.54
1.14
1.06
1.00
kf1/kf2
= 1.0
2.78
1.35
1.04
1.01
1.00
Table 20 Computed reserve estimate errors for kf1/kf2

= 0.01, 0.10 and 1.0.
Producing
Time Period
(years)
1
5
10
20
kf1/kf2
= 0.01
(%)
115.2
26.1
10.1
7.9
kf1/kf2
= 0.10
(%)
114.9
23.0
8.7
7.6
kf1/kf2
= 1.0
(%)
109.6
21.7
7.8
7.3
Stress-Dependent Fracture Conductivity. Most hydraulic

fractures, particularly those in reservoirs with high initial
stress conditions (similar to HP/HT gas reservoirs), experience
some reduction in conductivity created during the initial
stimulation treatment. The most common causes of conductivity reduction are proppant crushing and/or embedment
into the reservoir rock. Some proppants (i.e., resin-coated
ceramic, bauxite, and resin-coated sands) are designed to
minimize conductivity reductions in high-stress reservoirs.
For this paper, we evaluated the impact of stress-dependent
reservoir properties on the production decline using the
functions shown in Fig. 9. These curves represent the fraction
of original fracture conductivity as a function of the net
fracture pressure for various types of generic proppants.
Except for the inclusion of stress-dependent reservoir
properties, we used the reservoir and hydraulic fracture
properties shown in Table 6 as input for the simulation.
We compared simulated production profiles for all multilayered cases with and without stress-dependent permeability and porosity and we saw little or no differences in the
decline behavior. Again, the vertical heterogeneity caused by
reservoir layering and permeability contrast among layers has
a much more significant effect on the decline behavior than
the stress-dependent fracture conductivity.
Two-Phase and Non-Darcy Fracture Flow. The final
fracture characteristics evaluated in our study were two-phase
and non-Darcy flow phenomena. Two-phase flow conditions
typically occur primarily early in the production history
immediately following the stimulation treatment and during
the fracture clean-up period. Non-Darcy flow occurs during
any phase of the production history and is dependent on the
flowrate and pressure drawdown.
We modeled two-phase flow during the fracture cleanup
period using a method similar to that described in Reference
24. Two-phase (gas-water) flow in the fracture was accounted
for using the relative permeability curves shown in Fig. 8. We
also varied fracture conductivity from 50 to 500 md-ft.
Simulation results show that two-phase flow primarily affects
very early production declines, but has little to no impact on
the long-term decline behavior.
We also evaluated non-Darcy flow in the fracture. For this
study, we modeled non-Darcy flow using a rate-dependent
skin, D.3-5 Similar to the two-phase flow simulations, we also
varied fracture conductivity from 50 to 500 md-ft. Again, the
simulation results suggest that non-Darcy flow affects primarily the early decline behavior but has no substantial impact
on the long-term production profiles. For both the two-phase
and non-Darcy flow simulations, vertical heterogeneity from
layering and permeability contrast among layers had a much
larger effect on the long-term production decline.
Effect of Well Spacing on Production Decline
All of the previous results evaluating the effects of reservoir
and hydraulic fracture properties were generated for a well
drilled on an 80-acre spacing. In this section, we evaluate the
impact of other well spacings on the production decline
behavior. Except for variations in well spacing, all reservoir
15
and hydraulic fracture properties shown in Table 6 were used

as input for the simulation cases.
Figure 32 compares the simulated long-term production
profiles for well spacings of 40, 80 and 160 acres per well. As
expected, we observe much sharper initial decline rates as well
as much steeper long-term profiles for wells drilled on closer
spacing. The computed b-exponents for each well spacing
also reflect our observations concerning both short- and longterm decline characteristics. As shown in Table 21, computed
b-exponents are consistently lower for smaller well spacings
(for the time periods shown). Surprisingly, we also note that
all b-exponents, regardless of the spacing, lie approximately in
the range 0.50 < b < 1.0 after 50 years of production.
10,000
40-Ac Spacing
SPE 109625
80-Ac Spacing
160-Ac Spacing
1,000
100
10
1
0
10
15
20
25
30
35
40
45
50
Fig. 32 Simulated long-term production profiles for well

spacings of 40, 80 and 160 acres per well.
Table 21 Computed b-exponents for well spacings of 40,
80 and 160 acres per well.
Producing
Time Period
(years)
1
5
10
20
50
40 Acres
per
Well
2.11
1.15
1.02
0.97
0.84
80 Acres
per
Well
2.78
1.35
1.04
1.01
1.00
160 Acres
per
Well
2.98
1.96
1.09
1.06
1.05
The reserve estimate errors for the three well spacings considered in our study are shown in Table 22. Similar to the
previous results for various reservoir and hydraulic fracture
properties, we see a direct correlation between the computed
b-exponent and reserve estimate error. The results also show
that the largest errors for the 40-acre and 80-acre well spacing
cases occur with less than five years of production, while the
errors for the 160-acre spacing case are still quite significant
(greater than 10%) for the first 20 years of production.
Table 22 Computed reserve estimate errors for well
spacings of 40, 80 and 160 acres per well.
Producing
Time Period
(years)
1
5
10
20
40 Acres
per Well
(%)
67.0
10.5
7.2
5.5
80 Acres
per Well
(%)
109.6
21.7
7.8
7.3
160 Acres
per Well
(%)
122.2
25.9
17.4
14.7
Figures 33 and 34 present the simulated layer pressures at

monitoring points no. 1 and 2, respectively, for well spacings
16
of 40-acres per well (color-filled circles) and 80-acres per

wells (color-filled triangles). At pressure monitoring point 1
(Fig. 33), we see that all HFUs for both well spacings
experience some pressure depletion within the first year of
production. As expected, the largest pressure reductions occur
for the 40-acre spacing case.
Figure 34 shows the pressure response at pressure monitoring
point no. 2. Unlike most of the previous cases in which we
saw little or no pressure change for the entire 50-year
producing period, we see significant pressure depletion in
HFU 4 for the 40-acre spacing case after 20 years of
production. Based on the pressure responses shown in Figs.
33 and 34, we conclude that the well is in boundary-dominated
flow after 20 years of production. And, we would expect the
reserve estimate errors to be significantly lower after that time
period.
SPE 109625
To assess the impact of pressure, we also simulated several

cases with initial reservoir pressures of 10,800, 12,600, and
14,400 psia corresponding to pore pressure gradients of 0.60,
0.70 and 0.80 psi/ft, respectively. All simulated cases used to
evaluate the effects of reservoir pressure were generated with
a bottomhole temperature of 400oF. Although not shown, the
simulated production profiles generally exhibited lower initial
production rates and sharper initial declines for lower initial
reservoir pressures. However, both the intermediate- and
long-term decline behavior were very similar for all pressures.
Moreover, differences between the computed b-exponents and
reservoir estimate errors were also very minor.
We also compared the production decline characteristics for
reservoir temperatures of 300oF and 400oF for the same range
of initial reservoir pressures. Again, we observed very little
differences in the results generated using these two temperatures for the complete range of pressures considered.
20,000
Conclusions
Based on the results of our simulation study, we offer the
following conclusions about use of an Arps decline curve
methodology for evaluating reserves in tight gas sands at
HP/HT reservoir conditions:
18,000
16,000
14,000
12,000
10,000
8,000
6,000
4,000
2,000
0
0
10
15
20
25
30
35
40
45
50

HFU 1, 40-Ac
HFU 1, 80-Ac
HFU 2, 40-Ac
HFU 2, 80-Ac
HFU 3, 40-Ac
HFU 3, 80-Ac
HFU 4, 40-Ac
HFU 4, 80-Ac
Fig. 33 Comparison of layer pressures at pressure monitoring point no. 1 for well spacings of 40- and 80-acres
per well.
20,000
18,000
16,000
14,000
12,000
10,000
8,000
6,000
4,000
2,000
0
0
10
15
20
25
30
35
40
45
50

HFU 1, 40-Ac
HFU 2, 40-Ac
HFU 3, 40-Ac
HFU 4, 40-Ac
HFU 1, 80-Ac
HFU 2, 80-Ac
HFU 3, 80-Ac
HFU 4, 80-Ac
Fig. 34 Comparison of layer pressures at pressure monitoring point no. 2 for well spacings of 40- and 80-acres
per well.
Effect of Reservoir Pressure and Temperature on

Production Decline
The final phase of our simulation studies addressed the effects
of both reservoir pressure and temperature conditions. Recall
that all of the previous results were generated with an initial
reservoir pressure of 16,200 psia (i.e., initial pore pressure
gradient of 0.90 psi/ft at an average well depth of 18,000 ft)
and a bottomhole reservoir temperature of 400oF.
1. The most significant error source for reserve evaluations

using a traditional Arps decline curve methodology is
incorrect application of Arps' decline curves during either
transient flow or the transitional period between the end of
transient and onset of boundary-dominated flow. During
both of these periods (principally the transient period), we
observed b-exponents greater than one and corresponding
reserve estimate errors exceeding 100 percent.
2. Although only a few of the simulated cases evaluated in our
study reached true boundary-dominated flow during the first
50 years of production, we found that the reserve estimate
errors were quite often less than 10 percent when the well
was in the "late" transitional flow period. It appears that
reserve estimate errors using the Arps models will be
negligible if the least permeable layers contributing to flow
represent a small percentage of the overall contacted pore
volume.
3. b-exponents computed from the long-term (i.e, 50 years)
decline characteristics generally approached values between
0.5 and 1.0 for the range of reservoir and hydraulic fracture
properties and heterogeneities investigated in our study.
Agreement between Arps' suggested b-exponent range and
our simulated evaluations also indicates that, if applied
under the appropriate conditions, the Arps rate-time models
are appropriate for assessing reserves in tight gas sands at
HP/HT reservoir conditions.
4. Hyperbolic decline behavior commonly observed in wells
producing from tight gas sands is caused by various types of
both reservoir and hydraulic fracture heterogeneities. Our
simulated results demonstrate that low permeability alone in
a homogeneous system is not sufficient to generate large bexponents and flat declines.
5. Although not shown, terminal decline rates for the simulated
production ranged from about 1.5 to 5.0 percent for the
ranges of reservoir and hydraulic fracture properties and
heterogeneities evaluated in our study. These rates are based
on our reservoir inflow model and do not account for
wellbore problems that might affect well outflow
performance. These results also suggest that much larger
SPE 109625
terminal decline rates may not be attributable to reservoir

phenomena, but may be caused by operational problems
(e.g., liquid loading in the wellbore, loss of fracture
conductivity, plugging or closure of perforations, etc.).
Acknowledgments
We would like to express our thanks to Anadarko Petroleum
Corp. for permission to publish this paper.
Nomenclature
b
D
Di
qi
h
HFU
HRT
FCD
kg
kv
kh
kx
ky
Lf
wfkf
kf
pi
g
HC
Sw
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
Arps' decline exponent, dimensionless

rate-dependent skin factor, (MMscf/d)-1
initial decline rate, (days)-1
initial gas production rate, Mscf/day
gross sand thickness, ft
hydraulic flow unit
hydraulic rock type
dimensionless fracture conductivity = wfkf/kgLf
absolute Klinkenberg-corrected permeability, md
absolute vertical permeability, md
absolute horizontal permeability, md
absolute horizontal permeability in the x-direction, md
absolute horizontal permeability in the y-direction, md
effective fracture half-length, ft
fracture conductivity, md-ft
absolute fracture permeability, md
initial bottomhole reservoir pressure, psia
effective porosity, frac.
gas specific gravity, dimensionless
gas hydrocarbon component specific gravity,
dimensionless
= water saturation, frac.
References
1.
Arps, J.J.: "Analysis of Decline Curves,: Trans., AIME, vol.

160, 228-247.
2. Prouty, J.L.: "Tight Gas in the Spotlight," in GasTIPS, Gavin,
J.J. and Lang, K.R. (Editors), Gas Technology Institute, vol. 7,
no. 2, Chicago, IL (2001), 4-10.
3. VIP-Core Reference Manual, Landmark Graphics Corp. (2002),
Houston, TX.
4. VIP-Executive Reference Manual, Landmark Graphics Corp.
(2002) Houston, TX.
5. VIP-Executive Technical Reference Manual, Landmark
Graphics Corp. (2002) Houston, TX.
6. Gwilliam, B. and Nuccio, V.F.: "Basin-Centered Gas Systems
of the U.S.," U.S. Department of Energy, National Energy
Technology Laboratory, Report DE-AT26-98FT40031, April
1998-November 2000.
7. Law, B.E.: "What Is A Basin-Centered Gas System?," paper
presented at the Rocky Mountain Association of Geologists
2000 Basin-Centered Gas Symposium, Denver, CO, October 6.
8. Al-Hussainy, R, Ramey, H.J. and Crawford, P.B.: "The Flow of
Real Gases Through Porous Media," J. Pet. Tech. (May 1966)
624-636; Trans., AIME, 237.
9. Dranchuk, P.M. and Abou-Kassem, J.H.: "Calculations of zFactors for Natural Gases Using Equations of State," J. Can.
Pet. Tech. (July-Sept. 1975) 34-36.
10. Wichert, E. and Aziz, K.: "Calculate z's for Sour Gases,"
Hydrocarbon Processing (May 1972) 119-122.
11. Lee, W.J. and Wattenbarger, R.A.: "Properties of Natural
Gases," Chap. 1, in Gas Reservoir Engineering, vol. 5, SPE
Textbook Series, Society of Petroleum Engineering, Dallas, TX
(1996), 1-36.
17
12. Lee, A.L., Gonzalez, M.H., and Eakin, B.E.: "The Viscosity of
Natural Gases," J. Pet. Tech. (Aug. 1966) 997; Trans., AIME,
237.
13. Newsham, K.E. and Rushing, J.A.: "An Integrated Work-Flow
Model to Characterize Unconventional Gas Resources: Part
I-Geological Assessment and Petrophysical Evaluation," paper
SPE 71351 presented at the 2001 SPE Annual Technical
Conference and Exhibition, New Orleans, LA, Sept. 30-Oct. 3.
14. Pittman, E.D.: "Relationship of Porosity and Permeability to
Various Parameters Derived From Mercury Injection-Capillary
Pressure Curves for Sandstone", AAPG, v. 76, No. 2 (Feb. 1992)
191-198.
15. Gunter, G.W., et al. "Early Determination of Reservoir Flow
Units Using an Integrated Petrophysical Model," paper SPE
38679 presented at the 1997 SPE Annual Technical Conference
and Exhibition, San Antonio, TX, October 5-8.
16. Gunter, G.W., et al. "Overview of an Integrated Process Model
to Develop Petrophysical Based Reservoir Descriptions," paper
SPE 38748 presented at the 1997 SPE Annual Technical
Conference and Exhibition, San Antonio, TX, October 5-8.
17. Rushing, J.A. and Newsham, K.E.: "An Integrated Work-Flow
Model to Characterize Unconventional Gas Resources: Part
II-Formation Evaluation and Reservoir Modeling," paper SPE
71352 presented at the 2001 SPE Annual Technical Conference
and Exhibition, New Orleans, LA, Sept. 30-Oct. 3.
18. Thomas, R.D. and Ward, D.C.: "Effects of Overburden Pressure
and Water Saturation on Gas Permeability of Tight Gas
Sandstone Cores," J. Pet. Tech. (Feb. 1972) 120-124.
19. Newsham, K.E., et al.: "Use of Vapor Desorption Data to
Characterize High Capillary Pressures in a Basin-Centered Gas
Accumulation with Ultra-Low Connate Water Saturations,"
paper SPE 84596 presented at the 2003 SPE Annual Technical
Conference and Exhibition, Denver CO, Oct. 5-8.
20. Newsham, K.E., et al.: "A Comparative Study of Laboratory
Techniques for Measuring Capillary Pressures in Tight Gas
Sands," paper SPE 89866 presented at the 2004 SPE Annual
Technical Conference and Exhibition, Houston, TX, Sept. 2629.
21. Leverett, M.C.: "Capillary Behavior in Porous Sands", Trans.
AIME (1941).
22. Bennion, D.B., et al.: "Determination of True Effective In-Situ
Gas Permeability in Subnormally Water Saturated Tight Gas
Reservoirs," J. Can. Pet. Tech., vol. 43, no.10 (Oct. 2004) 2732.
23. Rushing, J.A., et al.: "Measurement of the Two-Phase Gas
Slippage Phenomenon and Its Effect on Gas Relative
Permeability in Tight Gas Sands, paper SPE 84297 presented at
the 2003 SPE Annual Technical Conference and Exhibition,
Denver CO, Oct. 5-8.
24. Sullivan, R.B., et al.: "Evaluation of Nonlinear Fracture
Relative Permeabilities and Their Impact on Water-Frac
Performance in Tight Gas Sands," paper SPE 98329 presented at
the 2006 SPE International Symposium and Exhibition on
Formation Damage Control, Lafayette, LA, Feb.15-17.
25. Lee, W.J. and Wattenbarger, R.A.: "Decline-Curve Analysis for
Gas Wells," Chap. 9, in Gas Reservoir Engineering, vol. 5, SPE
Textbook Series, Society of Petroleum Engineering, Dallas, TX
(1996), 214-225.

(Blasingame) SPE 109625

Enviado por

Dados do documento

Título original

Direitos autorais

Formatos disponíveis

Compartilhar este documento

Compartilhar ou incorporar documento

Opções de compartilhamento

Você considera este documento útil?

Este conteúdo é inapropriado?

Direitos autorais:

Formatos disponíveis

(Blasingame) SPE 109625

Enviado por

Direitos autorais:

Formatos disponíveis

SPE 109625

Copyright 2007, Society of Petroleum Engineers

to a recent study by the Gas Technology Institute (GTI),2 tight

J.A. Rushing, A.D. Perego, R.B. Sullivan, and T.A. Blasingame

z Variable effective fracture conductivities and lengths, unequal

Table 1 Summary of the Arps' rate-time relations (Ref. 1)

q(t ) = qi exp( Dit ) ................................ (2)

Description of Numerical Model

z Reservoir temperatures ranging from 300 to 400oF and initial

Fig. 1 Schematic diagram illustrating x-y Cartesian grid

Reservoir Pressures and Temperatures. For the base case,

Total mole percent

Total gas specific gravity, g

Reservoir Permeability and Porosity. As a basis for our

Pore Size, microns

Absolute Klinkenberg-Corrected Permeability, md

properties and their distribution would be similar for a deeper

Fig. 2 Pittman plot showing hydraulic rock types (HRT)

To capture the range of permeability and porosity exhibited by

We also evaluated reservoirs with one, eight and sixteen

shown in Table 3. For the eight- and sixteen-layer cases, we

Similarly, the permeabilities and porosities listed in Table 5

To illustrate the comparative heterogeneity for each layered

J.A. Rushing, A.D. Perego, R.B. Sullivan, and T.A. Blasingame

Reservoir Pore Pressure, psia

Fig. 5 Stress-dependent, absolute Klinkenberg-corrected

Normalized Cumulative Storage Capacity, frac.

Fig. 3 Modified Stratigraphic Lorenz

Stress-Dependent Reservoir Properties. Although all rocks

Gas-Water Capillary Pressure Curves. To model the vertical

Leverett J-Function, dimensionless

Normalized Cumulative Flow Capacity, frac.

Fraction Original Permeability, frac.

Water Saturation, fraction

Fig. 6 Gas-water Leverett J-Functions

Fraction Original Porosity, frac.

for various hy-

Reservoir Pore Pressure, psia

Fig. 4 Stress-dependent effective porosity relationships

Gas-Water Relative Permeability Curves. Although we are

HRT 1 & 2 Water

Relative Permeability, frac.

Relative Permeability, frac.

Fig. 7 Gas-water relative permeability curves for various

We also modeled several types of stimulation treatments and

Fig. 8 Gas-water fracture relative permeability profiles

Fraction Original Fracture Conductivity, frac.

Hydraulic Fracture Model. We modeled a vertical fracture

Fracture Closure Pressure, psia

Fig. 9 Stress-dependent fracture conductivities for various

z Stress-dependent fracture conductivity for three generic types

Decline Curve Analysis of Simulated Production

Simulated Operational Conditions. To better represent

z A range of fracture conductivities (0.1 FCD < 200) including

J.A. Rushing, A.D. Perego, R.B. Sullivan, and T.A. Blasingame

define the true estimated ultimate recovery (EUR) or

Traditional decline curve analysis is based on the general form

Similarly, the harmonic decline is also a special case of Eq. 1

The Arps hyperbolic decline is given by the general form (Eq.

not be valid for modeling the decline behavior of tight gas

Inspection of the simulated production profiles in Figs. 10 and

is not necessarily sufficient to generate large b-exponents

Gas Production Rate, Mscf/day