See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/324165953

Estimating Fracture Closure Stress in Naturally Fractured Reservoirs with

Diagnostic Fracture Injection Tests (ARMA-18-225)

Article · June 2018

CITATIONS READS

2 462

2 authors:

Hanyi Wang Mukul Sharma

University of Texas at Austin University of Texas at Austin
40 PUBLICATIONS   299 CITATIONS    458 PUBLICATIONS   6,367 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Numerical Modeling of Particulate Flow Systems View project

New P3D Modeling of Hydraulic Fracture Propagation and Closure View project

All content following this page was uploaded by Hanyi Wang on 05 October 2018.

The user has requested enhancement of the downloaded file.

Citation Export:
Wang, H and Sharma, M.M. 2018. Estimating Fracture Closure Stress in Naturally Fractured Reservoirs with Diagnostic Fracture Injection Tests. Paper ARMA-2018-225 presented at the 52nd US Rock Mechanics / Geomechanics
Symposium, held in Seattle, Washington, June 17-20.

ARMA 18–225

Estimating Fracture Closure Stress in Naturally Fractured

Reservoirs with Diagnostic Fracture Injection Tests
HanYi Wang
The University of Texas at Austin, Austin, Texas, USA
Mukul M. Sharma
The University of Texas at Austin, Austin, Texas, USA

Copyright 2018 ARMA, American Rock Mechanics Association

This paper was prepared for presentation at the 52nd US Rock Mechanics / Geomechanics Symposium held in Seattle, Washington, USA, 17–20
June 2018. This paper was selected for presentation at the symposium by an ARMA Technical Program Committee based on a technical and
critical review of the paper by a minimum of two technical reviewers. The material, as presented, does not necessarily reflect any position of
ARMA, its officers, or members. Electronic reproduction, distribution, or storage of any part of this paper for commercial purposes without the
written consent of ARMA is prohibited. Permission to reproduce in print is restricted to an abstract of not more than 200 words; illustrations may
not be copied. The abstract must contain conspicuous acknowledgement of where and by whom the paper was presented.

ABSTRACT: Diagnostic Fracture Injection Tests (DFIT), which have also been referred to as Injection-Falloff Tests, Fracture
Calibration Tests or Mini-Frac Tests in the literature, have evolved into a commonly used technique to obtain in-situ stress.
Simplifying assumptions used in traditional G-function model often lead to an underestimation of the closure stress, even for a
planar fracture geometry. When a DFIT is conducted in naturally fractured reservoirs, the stimulated natural fractures can either
alter the effective reservoir permeability within the distance of investigation or interact with the hydraulic fracture to form a
complex fracture geometry. In this study, we present a new pressure transient model for DFIT analysis in naturally fractured
reservoirs, and it is shown that using established methods to pick closure stress often lead to significant over or underestimates,
regardless of whether the reservoir is heavily fractured or sparsely fractured. Our proposed “variable compliance method” gives a
much more accurate and reliable estimation of closure stress in both homogenous and naturally fractured reservoirs.
the pressure decline can be obtained based on a G-
1. INTRODUCTION
function approach. Wang and Sharma (2017b) showed
Diagnostic fracture injection tests (DFIT) involve that that the traditional “tangent line method (Barree et al.
pumping a fluid (typically water), at a constant rate for a 2009) ” underestimates the closure stress while the
short period of time, creating a relatively small hydraulic “compliance method” (McClure et al. 2016) estimates
fracture before the well is shut in. The pressure transient the onset of closure rather than the closure stress. Based
data after shut-in is analyzed to obtain in-situ stresses on their analysis, a new “variable compliance method”
and reservoir properties. In addition, the net pressure was proposed to estimate the closure stress. Wang and
trends in a DFIT can be also used to infer the induced Sharma (2018) further confirmed that fracture
fracture complexity in different geological settings compliance is pressure-dependent during closure, and
(Potocki 2012) and stress-dependent un-propped fracture the mechanically closed fracture is still subject to leak-
conductivity can be estimated through forward modeling off. The stress-dependent un-propped fracture
or inverse calculation (Wang and Sharma 2018). This conductivity can be estimated by analysis of DFIT data.
valuable information obtained from DFITs provides key
Previous papers have been mainly focused on planar
input parameters for modeling hydraulic fracture
fractures in homogeneous reservoirs, and relatively little
propagation (Wang 2015; Wang 2016; Wang et al. 2016),
work has been done to interpret DFIT data in naturally
stimulation design (Ramurthy et al. 2011), development
fractured reservoirs with fracture pressure dependent
of reservoir models (Mirani et al. 2017; Loughry et al.
leak-off (FPDL) and variable fracture compliance. In
2015; Wang 2017) and post-fracture analysis (Fu et al.
this study, we extend our previous work to investigate
2017).
pressure transient behavior in naturally fractured
The advent of fracturing pressure decline analysis was reservoirs.
pioneered by the work of Nolte (1979 and1986). With
the assumptions of power-law fracture growth,
negligible spurt loss, constant fracture compliance (or
2. THE GOVERNING EQUATIONS
stiffness), constant fracture surface area after shut-in and 2.1. 1D Fracture Pressure Dependent Leak-off
Carter’s leak-off (a constant pressure boundary with Variable Compliance
condition), a remarkably simple and useful equation for
The transient pressure response during fracture closure is 𝑘 𝑑𝑃 𝐴𝑓 𝑑𝑃𝑓
𝑞𝑓 = − 2𝐴𝑓 =− at x = 0 (5)
derived using the following assumptions: 𝜇𝑓 𝑑𝑥 𝑆𝑓 𝑑𝑡
1. Reservoir is isotropic and homogeneous and contains Rearranging Eq. (5), we get the boundary condition at
a single slightly compressible fluid.
2. The fluid viscosity, formation porosity and total
the fracture surface:
compressibility are independent of pressure. 𝑘 𝑑𝑃 1 𝑑𝑃𝑓
3. Reservoir permeability is low so that poroelastic = at x = 0 (6)
𝜇𝑓 𝑑𝑥 2𝑆𝑓 𝑑𝑡
effects caused by fluid leak-off are negligible
4. Gravity effects are negligible and pressure gradients With initial condition (disregarding the pressure
are small thorough out the reservoir. disturbance during the short injection period)
5. Mechanically closed fracture has residual fracture
width so it is still subject to leak-off (Wang and 𝑃 = 𝑃0 at t = 0, x > 0 (7)
Sharma 2018) and the fracture leak-off surface area
remains unchanged. 𝑃 = ISIP at t = 0, x = 0 (8)
6. Fracture pressure is uniformly distributed inside the where 𝑃0 is the initial pore pressure and ISIP is the
fracture. instant shut-in pressure. The governing equation Eq.(1),
7. The pore pressure disturbance caused by fracture plus the initial condition Eq. (7)-(8) and boundary
propagation is negligible. This means that fluid leak-
off during pumping is very small and the duration of
condition Eq.(6) uniquely describe the pressure transient
injection is very short (typically 3-5 minutes) while behavior during DFITs. To account for wellbore storage
the total shut-in time can be hours, days or even effects, the fracture stiffness 𝑆𝑓 needs to be replaced by
weeks. the fracture-wellbore system stiffness, which is defined
8. Leak-off is linear and perpendicular to the fracture as:
surface and pressure interference among fractures and
late time radial flow has not been developed yet. 𝑆𝑓 𝐴𝑓
𝑆𝑠 = (9)
𝐴𝑓 + 𝑆𝑓 𝑉𝑤 𝑐𝑤
Assuming 1D linear Darcy flow and a slightly
compressible, single-phase fluid in the reservoir, the where 𝑉𝑤 is the wellbore volume and 𝑐𝑤 is the water
differential form of the mass balance can be written as: (injection fluid in the wellbore) compressibility.
∂P 1 ∂ ∂P 2.2. Fracture Stiffness of Complex Fracture
= (k ) (1)
∂t 𝜇𝑓 𝜙𝑐𝑡 ∂x ∂x Geometry
Because fracture stiffness (or compliance) represents
where P is the pressure, k is formation permeability,
surface normalized compressibility, for a fracture with
which can be a function of P. μf is fluid viscosity, ϕ is complex fracture geometry, the total fracture stiffness is
formation porosity and ct is total formation controlled by the stiffness of multiple individual fracture
compressibility. From a material balance perspective segments, as illustrated in Figure.1.
(fluid compressibility is negligible compared to that of
the fracture), the total rate of fluid leak-off into the
formation, qf , after shut-in equals the rate of shrinkage
of total fracture volume, Vf , as pressure declines:
𝑑𝑉𝑓
𝑞𝑓 = − (2)
𝑑𝑡
Fracture stiffness, which is the reciprocal of fracture
compliance, is defined as (Wang and Sharma 2017a;
Wang et al. 2018): Fig. 1 Illustration of complex fracture geometry.
𝐴𝑓 𝑑𝑃𝑓
𝑆𝑓 = (3) The complex fracture is formed by individual segments
𝑑𝑉𝑓
with different orientations and fracture surface area. Let
where 𝑃𝑓 is pressure inside fracture and 𝐴𝑓 is half of the the ith fracture surface area of one wall denoted as Af,i
total fracture surface area (only account for 1 wall). and its pressure-dependent stiffness denoted as Sf,i , the
Substituting Eq.(3) into Eq.(2), we have total fracture stiffness can be determined from the
𝐴𝑓 𝑑𝑃𝑓 stiffness of each individual fracture segment (analogous
𝑞𝑓 = − (4) to capacitors in parallels):
𝑆𝑓 𝑑𝑡
𝐴𝑓
At the boundary of the fracture surface, both Darcy’s 𝑆𝑓 = (10)
𝐴𝑓,1 𝐴𝑓,2 𝐴𝑓,𝑛
law and material balance have to be honored: +
𝑆𝑓,1 𝑆𝑓,2 + ⋯ + 𝑆𝑓,𝑛
Where,
𝐴𝑓 = 𝐴𝑓,1 + 𝐴𝑓,2 + ⋯ + 𝐴𝑓,𝑛 (11)
2.3. Pressure Dependent Permeability
In previous literature (Barree et al. 2009), the
phenomenon of pressure-dependent permeability is often
called as pressure-dependent leak-off (PDL), however,
this is not an appropriate term, because leak-off is
always pressure-dependent, even if the permeability is
constant (Wang and Sharma 2017b). There are many
equations that can be used to express the formation
effective permeability as a function of pressure, such as
a linear, power law or hyperbolic relationship. To make
a general case and use as few free parameters as
possible, we can use a general formula to relate
formation permeability and pressure:
𝐼𝑆𝐼𝑃−𝑃
𝑐 the location where the DFIT is done, the natural
𝑒 𝑐 − 𝑒 𝐼𝑆𝐼𝑃−𝑃𝑜𝑟𝑖𝑔𝑖𝑛𝑎𝑙
k(P) = (𝑘𝑒 − 𝑘0 ) + 𝑘0 (12)
𝑒𝑐 − 1
where 𝑘𝑒 is the effective reservoir permeability at the
end of pumping and 𝑘0 is the original reservoir
permeability. 𝑃𝑜𝑟𝑖𝑔𝑖𝑛𝑎𝑙 is the pressure where reservoir
permeability drops to its original value and c a free
parameter controls the evolution of permeability. 𝑘𝑒 is
larger than 𝑘0 because the surrounding permeability is
enhanced by stimulated fracture networks during
pumping. After the pressure drops to or below 𝑃𝑜𝑟𝑖𝑔𝑖𝑛𝑎𝑙 ,
the stimulated fracture networks close back to the
original in-situ state and the effective permeability
returns to its original value. Figure.2 demonstrates the
pressure-dependent permeability for different values of c.
Fig. 2 Illustration of pressure-dependent permeability with
different values of c
2.4. Pressure Dependent Fracture Compliance
Fracture closure is a gradual process with increasing
fracture stiffness (or decreasing fracture compliance) as
fracture closes from tip to wellbore or closes from higher
stress region to lower stress region in layered formations.
A method to obtain a pressure-dependent fracture
compliance based on fracture geometry, rock properties
and surface roughness has been well discussed by Wang
and Sharma (2017a) and Wang et al. (2018). The
influence of surface roughness on fracture compliance is
described by a contact law relating the fracture width to
the net closure stress for fractured rocks (Willis-Richards
et al. 1996):
σref w0
σc = ( − 1) for wf ≤ w0 (17)
9 wf
where wf is the fracture aperture and, w0 is the contact
width, which represents the fracture aperture when the
contact normal stress is equal to zero, σc is the contact
normal stress on the fracture, and σref is a contact
reference stress, which denotes the effective normal
stress at which the aperture is reduced by 90%.
3. DFIT IN NATURALLY FRACTURED
RESERVOIRS
All reservoirs are fractured to some degree. Depending
on the density and dimensions of natural fractures and
fractures can impact hydraulic fracture propagation and
the adjacent flow capacity. If the reservoir is heavily
fractured and these natural fractures are well-connected,
as shown in Figure 3, then the formation can be
considered as homogenous within the distance of
investigation and the effective permeability is controlled
by both fracture and matrix. High-pressure fluid
injection during the pumping period enhances the local
effective permeability because it widens the natural
fractures. However, as pressure declines during shut-in,
the effective permeability will gradually decline until it
reaches the original value. The original effective
permeability does not necessarily represent the matrix
permeability, because the existence of natural fractures
may already have enhanced the permeability.
Fig. 3 Heavily fractured reservoir with well-connected natural
fractures
If the reservoir is sparsely or moderately fractured, and
the stimulated natural fractures are discrete and not well-
connected, then these natural fractures or isolated
fracture networks will not change the overall flow
capacity in the reservoir (Wang 2017), but they can
interact with hydraulic fractures, and generate a complex
fracture geometry, as shown in Figure 4. Under this
scenario, the fracture compliance and leak-off surface
area are impacted by the sequential closure of individual
fracture segments or branches, where the branches with
the highest normal stress perpendicular to local fracture
orientation will close first, and the fracture segments that
open against the minimum principal stress will be the
last to close. If the hydraulic fracture does not initiate
along the maximum principal stress direction because of
the local hoop stress around perforations, non-planar
fractures can also be formed. In this section, we are
going to investigate the pressure transient behavior
under these scenarios and discuss the implications of
estimating closure stress using different methods.
shown in Figure 5. As can be seen, when the fluid
pressure inside the fracture is relatively high, the fracture
volume declines linearly with pressure under constant
fracture stiffness (or compliance). However, as the
pressure declines to a certain level the fracture volume
and pressure depart from a linear relationship, and
fracture stiffness starts to increase noticeably. In the end,
when the fluid pressure drops to a sufficient low level,
the residual fracture volume (supported by asperities)
becomes more or less insensitive to pressure drop with
extremely high fracture stiffness (low compliance). This
is because when the pressure inside the fracture is
relatively high, it is still wide open and its stiffness is a
constant. However, as pressure continues to drop, the
fracture will gradually close from its edges to its center,
which increases the fracture stiffness gradually.

Fig. 4 Low to Moderate Fractured Reservoir with Discrete

Natural Fractures

3.1. Base Case

Before we embark on investigating the pressure transient
behavior with closing fracture in fractured reservoirs,
Fig. 5 Illustration of the fracture stiffness and volume
we’ll first examine pressure decline behavior for a planar
evolution from modeling
fracture and discuss the causes for the discrepancy for
estimated closure stress using established methods. Figure.6 shows the normalized tiltmeter response (its
Table 1 shows the input parameters for the Base Case slope is proportional to fracture compliance or the
Scenario. reciprocal of fracture stiffness) plotted against wellbore
Table 1. Input parameters for the Base Case scenario pressure during the shut-in period of 2B well at three
Fracture type PKN
different stations from the GRI/DOE M-site. The
Fracture height 12 m tiltmeter demonstrates that soon after shut-in, the
Fracture half-length 50 m measured displacement declines linearly with pressure
Pumping time 5 min (roughly constant fracture stiffness). After the pressure
ISIP 40 MPa declines to a certain level, the measured displacement vs
Minimum in-situ stress 35 MPa pressure departs from a linear relationship. As pressure
Initial pore pressure 20 MPa continues declining, more and more fracture surface area
Young's modulus 20 GPa comes into contact with the rough fracture surfaces and
-1
Total compressibility 6.3e-4 MPa asperities, and the fracture stiffness increases gradually.
In-Situ fluid viscosity 0.3 cP
This field measurement is consistent with the general
Formation original permeability 0.0005 md
Poisson's Ratio 0.25
trend from out modeling results shown in Figure.5. It is
Initial porosity 0.03 challenging to estimate closure stress based on
Contact width of asperities 2 mm mechanical closure from tiltmeter measurement. If we
Contact reference stress of asperities 5 MPa pick the closure stress using the “tangent line method”
on tiltmeter measurement (Craig et al. 2017), the closure
stress is 15 MPa, and if we pick the closure stress using
With the known fracture geometry, rock modulus, the “compliance method”, the closure stress is 21 MPa.
minimum in-situ stress and surface roughness There is a 6 MPa discrepancy between these two
(represented by two contact parameters 𝑤0 and 𝜎𝑟𝑒𝑓 ), methods. As discussed by Wang and Sharma (2017b),
the fracture stiffness and volume evolution can be the “tangent line method” only marks the end of fracture
uniquely determined at different fluid pressures (Wang storage dominated flow, so it always underestimates
and Sharma 2017a; Wang et al. 2018), the results are closure stress while the “compliance method” can
overestimate closure stress, because it only marks the
beginning fracture closure on edges or tips. The
“variable compliance method” proposed by Wang and
Sharma (2017b) provides an alternative way to estimate
closure stress that not only accounts for the difference
between two endpoints (fracture compliance start to
change and fracture compliance become small enough to
be considered as negligible), but also compensates for
the impact of leak-off, which is reflected on G-function
time or the square root of time.

Fig. 7 Pressure decline and its derivatives of Base Case

Fig. 6 Tiltmeter measurement from 2B well of GRI/DOE M- scenario
site project

To illustrate the benefit of the “variable compliance

method” and compare the estimated closure stress from
different methods, the pressure-dependent fracture
stiffness from Figure. 5 is used as a boundary condition
for Eq.(6) to simulate the pressure decline response, all
other parameters are provided in Table 1. Figure.7 shows
the pressure decline and its derivatives on the G-function
and square root of time plots. Compared to the input
closure stress of 35 MPa, the estimated closure stress by
the “tangent line method” is 32.2 MPa (-2.8 MPa error),
the estimated closure stress by the “compliance method”
is 36.8 MPa (1.8 MPa error), and the estimated closure
stress by the “variable compliance method” is 35.2 MPa
(0.2 MPa error). It is clear that the closure stress
estimated by the “variable compliance method” is more
reliable. Detailed comparisons with different fracture
geometry, surface roughness and reservoir properties
have been presented by Wang and Sharma (2017b), and
hence will not be discussed further.
Figure.8 shows the pressure profiles in the formation
away from the fracture surface at different shut-in times.
Together with Fig.6, we can notice that the closure time
is around 8.85 hr when fluid pressure drops to the close
stress, and the corresponding distance of investigation is
less than 7 m at the time of closure. This is very typical
in low permeability formations.
Fig. 8 Pressure profile and distance of investigation for Base
Case Scenario.
Now the question remains: what unique features will
emerge from DFIT data in naturally fractured reservoirs
and how do we reliably estimate closure stress under
these contexts? Are these established methods are
reliable ways to estimate closure stress? In the
following sections, we’ll examine these cases and
discuss the associated challenges in identifying closure
stress in naturally fractured reservoir.
3.2. Heavily Fractured Reservoir with Pressure
Dependent Permeability
First, let’s investigate the case where the reservoir is
heavily fractured and all the stimulated fractures form
well-connected networks that alter the effective
permeability around the hydraulic fracture. The effective
permeability is highest at the end of pumping and
gradually declines as pressure drops and fluid leak-offs
into formation. Assumes the effective permeability
follows Eq.(12), with c=3 and 𝑃𝑜𝑟𝑖𝑔𝑖𝑛𝑎𝑙 is 4.5 MPa
below the ISIP, all the other input parameters are the
same as the base case. Figure.9 shows the pressure
decline and its derivatives with different values of
enhanced effective permeability 𝑘𝑒 at the end of
pumping. From the results, we can clearly observe that
there exists a “bump” on the pressure derivative during
fracture closure before it reaches its final peak. And in
all cases, the pressure derivatives “bump” end at a
fracturing pressure that is 4.5 MPa below the ISIP,
which is the exactly value of 𝑃𝑜𝑟𝑖𝑔𝑖𝑛𝑎𝑙 . So the values of
𝑃𝑜𝑟𝑖𝑔𝑖𝑛𝑎𝑙 can be readily obtained from field DFIT data if
it exhibits this pressure-dependent permeability
signature. We can also observe that the higher the 𝑘𝑒 , the
large the initial slope of pressure derivatives.
Fig. 9 Pressure decline and its derivatives with pressure-
dependent permeability
Because the effect of pressure-dependent permeability
masks the initial phase of fracture closure, so the
moment when the fracture stiffness starts to change
cannot be identified. In such cases, the “compliance
method” is hard to apply. We can, however, modify our
“variable compliance method” approach as follows: first,
we identify the dimensionless G-time or the square root
of time at the end of the first pressure derivative “bump”
(i.e., end of pressure-dependent permeability) and at the
intersection of pressure derivative and a tangent line
passing through the origin (i.e., the closure point picked
by the “tangent line method”). Then average the values
of these two dimensionless G-time or the square root of
time, finally extrapolate back to the pressure curve that
corresponds to the averaged G-time or the square root of
time. Table 2 shows the estimated closures by “tangent
line method” and “variable compliance method”. As can
be seen, the “tangent line method” consistently
underestimates the closure stress. This stems from the
fact that the “tangent line method” assumes constant
fracture stiffness and Carter’s leak-off during closure,
but these two assumptions are not valid (Wang and
Sharma 2017b). The closure stress estimated by
“variable compliance method” gives significantly lower
errors and is a more reliable approach to estimate closure
stress.
Table 2. Comparison of estimated closure stress from different
methods with pressure-dependent permeability
Tangent Line Method Variable Compliance Method
Scenarios Estimated Absolute Estimated Absolute
𝝈𝒉𝒎𝒊𝒏 Error 𝝈𝒉𝒎𝒊𝒏 Error
𝑘𝑒 = 𝑘0 32.2 MPa -2.8 MPa 35.2 MPa 0.2 MPa
𝑘𝑒 = 5𝑘0 32.3 MPa -2.7 MPa 34.4 MPa 0.4 MPa
𝑘𝑒 = 10𝑘0 32.1 MPa -2.9 MPa 34.5 MPa 0.5 MPa
𝑘𝑒 = 20𝑘0 32.2 MPa -2.8 MPa 34.4 MPa 0.4 MPa
3.3. Discrete Natural Fractures
In the case of sparsely or moderately fractured reservoirs,
even though the stimulated natural fractures are discrete
and not well-connected, they can interact with
propagating hydraulic fracture and generate complex
fracture geometry. To make modeling and analysis
trackable, we’ll first investigate a simplified version of
complex fracture geometry that only consist two sets of
fractures with different orientations, as shown in
Figure.10, then more general scenarios will be discussed
at the end. In Figure.10, we can see that the main
hydraulic fracture (i.e., fracture set 1) is perpendicular to
the minimum principal stress, while the natural fractures
(i.e., fracture set 2) are perpendicular to the maximum
principal stress. As pressure declines inside fracture after
shut-in, the natural fracture that opens against the
maximum principal stress will close first, and the
hydraulic fracture that opens against the minimum
principal stress will be the last to closure. Here, there are
two possible scenarios are likely to happen during this
sequential closure process: 1) the closed natural fracture
remain hydraulic connectivity with the main hydraulic
fracture. 2) the closed natural fracture lost the hydraulic
connectivity with the main hydraulic fracture. We’ll
discuss these two scenarios and their implications in
interpreting DFIT data separately.
3.3.1. Closed Fracture Remain Hydraulic
Connectivity
If the intersection of fractures are well aligned (not
skewed) and supported by asperities and tortuous walls,
then it is likely that the closed natural fractures can
maintain hydraulic connectivity to the open fracture.
 hydraulic fracture. Because the fracture stiffness
evolution is smooth and gradual, so the closure of
natural fracture is undetectable on G-function or the
square root of time plot. But from the pressure
derivatives, we can notice that the larger the surface area
of natural fracture, the sooner the pressure derivatives
curve upward and the faster the pressure decline during
closure.

Fig. 10 Top view of complex fracture geometry with two sets

of fractures.

In this case, the sequential closure natural fractures or

branches of a complex fracture will not alter the total
leak-off surface area, but will change the overall fracture
stiffness evolution. Assume a complex fracture that
consists of two sets of fracture, as shown in Figure.10.
The total surface areas are 𝐴𝑓1 and 𝐴𝑓2 for fracture set
1 and fracture set 2, respectively. The maximum
principal stress is 38 MPa and the minimum principal
stress is 33 MPa. We further assume that all fracture
closes like a PKN fracture. Figure.11 shows the fracture
stiffness evolution with different surface area ratio. As
can be seen, if we only have fracture set 2 (i.e., natural
fracture), the fracture stiffness will increase at 39.5 MPa,
and if we only have fracture set 1 (i.e., hydraulic
fracture), the fracture stiffness will increase around 34.5
MPa. The overall fracture stiffness evolution is
calculated using Eq.(10) with different surface area ratio. Fig. 12 Pressure decline and its derivatives of complex
Because for complex fracture geometry, the overall fracture geometry (remain hydraulic connectivity after
fracture stiffness is like a parallel system of capacitors, closure)
so the changes in fracture stiffness is not directly
proportional to the surface area ratio, but is mostly Table 3 summarizes the estimated closure stress
influenced by the fracture set that has the highest ratio of (minimum principal stress) by different methods. As
surface area to stiffness. always, the “tangent line method” underestimates the
closure stress, while “compliance method” overestimates
closure stress. The “compliance method” gives a much
larger error if the surface area of natural fracture is large.
This is because “compliance method” picks the pressure
when natural fracture edges or tips start to closure,
which can be much higher than the minimum principal
stress. The “variable compliance method” yields the
least error.
Table 3. Comparison of estimated closure stress of complex
fracture geometry (hydraulic connectivity remains after
closure)
Tangent Line Compliance Variable Compliance
𝑨𝒇𝟐 Method Method Method
Fig. 11 Fracture stiffness evolution for a complex fracture
𝑨𝒇𝟏 𝝈𝒉𝒎𝒊𝒏 Error 𝝈𝒉𝒎𝒊𝒏 Error 𝝈𝒉𝒎𝒊𝒏 Error
geometry (hydraulic connectivity remains after closure) (MPa) (MPa) (MPa) (MPa) (MPa) (MPa)
0.5 31 -2 34.8 1.8 33.4 0.4
Figure.12 shows the pressure decline and its derivatives 1 31.1 -1.9 35.4 2.4 33.6 0.6
of complex fracture geometry when closed natural 5 31.6 -1.4 37.8 4.8 33.3 0.3
fractures remain their hydraulic connectivity to the main 10 32 -1 39.2 6.2 33.5 0.5
In fact, the information on fracture complexity can’t be
detected under such cases, and compared to a single
hydraulic fracture scenario (i.e., only fracture set 1 on
Fig.10), the impact of natural fractures on overall
fractures stiffness evolution can be substituted by a
“rougher” fracture walls with larger contact width and
higher contact reference stress (Wang and Sharma
2017b) to obtain earlier and steeper increase in fracture
stiffness as pressure declines. So based on before-closure
DFIT data alone, we may not be able to distinguish the
causes to the increase of fracture stiffness from fracture
complexity or surface roughness or recession in barrier
layers. Nevertheless, as discussed by Wang and Sharma Fig. 13 Pressure decline and its derivatives of complex
(2017b), the “variable compliance method” is insensitive fracture geometry (lose hydraulic connectivity after closure)
to fracture geometry, surface roughness and reservoir
properties, and from the above results, we see that it is
also much less sensitive to the existence of natural
fractures, so the “variable compliance method” is more
reliable way of estimate closure stress, especially when
we are not sure whether the created fracture geometry is
complex or planar.
3.3.2. Closed Fracture Lost Hydraulic
Connectivity
If the intersections of fractures are skewed and not well-
aligned (such as T-shaped fracture), then it is likely that
the closed fracture will lose its hydraulic connectivity to
the open fracture. In such case, the sequential closure
natural fractures or branches of a complex fracture will
alter both the total leak-off surface area and the overall
fracture stiffness evolution. Assume the same complex
fracture as discussed in the previous section. The
maximum principal stress is 37 MPa and the minimum
principal stress is 33 MPa. We further assume the natural
fractures will gradually lose their hydraulic connectivity
to the open hydraulic fracture when fluid pressure drops
around 37 MPa. Figure.13 shows the fracture stiffness
evolution with different surface area ratio when natural
fractures lose their hydraulic connectivity after closure. Fig. 14 Pressure decline and its derivatives of complex
Compared to Figure.11, we can notice that the overall fracture geometry (lose hydraulic connectivity after closure)
fracture stiffness evolution is not monotonically
increasing with declining pressure anymore. The surface Table 4. Comparison of estimated closure stress of complex
area ratio will shape the fracture stiffness evolution fracture geometry (lose hydraulic connectivity after closure)
before the closure of natural fractures, but once the Tangent Line Compliance Variable Compliance
pressure drops to a certain level and disconnect the 𝑨𝒇𝟐 Method Method Method
natural fractures, the overall fracture stiffness becomes 𝑨𝒇𝟏 𝝈𝒉𝒎𝒊𝒏 Error 𝝈𝒉𝒎𝒊𝒏 Error 𝝈𝒉𝒎𝒊𝒏 Error
(MPa) (MPa) (MPa) (MPa) (MPa) (MPa)
the fracture stiffness of the remaining hydraulic fracture.
0.5 30.5 -2.5 35.4 2.4 33.4 0.4
Figure.14 shows the pressure decline and its derivatives 1 30.6 -2.4 35.5 2.5 33.4 0.4
of complex fracture geometry when closed fractures lose 5 30.7 -2.3 35. 2.7 33.6 0.6
hydraulic connectivity. It can be clearly observed that 10 30.9 -2.1 35.8 2.8 33.6 0.6
there are two peaks on pressure derivatives curve, the
first one is associated with the closure process of natural If we only analyzed the portion of data after the first
fractures, and the second one is associated with the peak of pressure derivatives and use it to estimate the
closure process of the hydraulic fracture. closure stress associated with the closure of the
hydraulic fracture, we can compare the results from
different methods, as shown in Table 4. Again, we can
notice that “tangent line method” underestimates closure
stress while “compliance method” overestimates closure
stress. The “variable compliance method” is more
reliable and gives the least errors.
In reality, a complex fracture can consist of many
fracture sets with different orientations and surface area.
The approach and analysis presented for our simplified
cases are still valid for these more general scenarios.
Assuming that a complex fracture has four sets of
fractures with equal surface area, but open against
different in-situ stresses (37 MPa, 35 MPa, 33MPa and
31 MPa, where 31 MPa is the minimum principal stress)
because of their orientations, Figure.15 shows the
evolution of fracture stiffness as a function of fracturing
pressure. As can be seen, if natural fractures gradually
lose their hydraulic connectivity around the normal
stress that is perpendicular to that fracture surface, the
overall fracture stiffness will exhibit three peaks before
it finally reflects the fracture stiffness of the last open
fracture.
Fig. 15 Fracture stiffness evolution of complex fracture
geometry with four sets of fracture
Figure.16 shows the corresponding pressure decline and
its derivatives. As expected, four peaks emerge on the
pressure derivatives and if we only analyze the portion
of data after the third peak to estimate closure stress, the
“tangent line method” underestimates the closure stress
by around 1 MPa. The “variable compliance method”
gives a closer value. In general, estimated closure stress
by the “tangent line method” can be regarded as the
lower bound of closure stress and the closure stress
estimated by the “compliance method” can be viewed as
the upper bound of closure stress.
4. COMPARISON WITH EXPERIMENT
Craig et al. (2017) compared and discussed fracture
closure stress estimated by “Compliance Method” and
“Tangent Line Method” using tiltmeter measurement
and laboratory experiments.
Fig. 16 Pressure decline and its derivatives of complex
fracture geometry with four sets of fracture
Figure.17 shows the G-function plots using data
extracted from laboratory experiments from their study.
As expected the pressure decline is affected by tip
extension and fluid transient behavior along the fracture
soon after shut-in, so a brief pressure derivative spike
often immediately follows shut-in. In many cases, this
pressure derivative spike may mask the “open fracture
period ” (with constant stiffness) and make it difficult to
locate the moment when the fracture stiffness starts to
increase. If the closure time is not long enough, this
early time abnormal pressure (large pressure drop during
a short time interval) can’t be distinguished from the
signature of pressure-dependent permeability, which is
normally the case in laboratory experiments where the
fracture closes in less than 1 minute. In the soft rock case
of Figure.17, the imposed closure stress is 7 MPa, but
the closure stress estimated by the “Tangent Line
Method” is only 3 MPa, which is significantly lower.
Because the early-time pressure derivative spike masks a
large portion of before closure data, the “compliance
method” and “variable compliance method” are not
applicable. In the moderately hard rock case of
Figure.17, the imposed closure stress is 16 MPa, the
closure stress estimated by “Tangent Line Method” is
14.32 MPa and the closure stress estimated by
“compliance method” is 20.16 MPa. The closure stress
estimated by “variable compliance method” is 17.1 MPa
at the averaged G-function time of 1.745. This result is
consistent to our simulated synthetic cases presented in
this study and previous study (Wang and Sharma
2017b). Laboratory experiments also show that the
“variable compliance method” gives the least error and
“Tangent Line Method” normally underestimates closure
stress.
Fig. 17 Laboratory fracture closure experiment for soft (upper
figure) and moderately hard rock (lower figure) (Craig et al.
2017)
5. DISCUSSION
Examining Eq.(1)~Eq.(9) closely, it is evident that the
pressure decline is governed by a linear Partial
Differential Equation (PDE) system in the formation that
is coupled with an Ordinary Differential Equation
(ODE) at the fracture surface. The minimum in-situ
stress is just one of many implicit factors that shape the
evolution of the fracture pressure in the boundary ODE.
The causes for the error associated with “tangent line
method” stems from the traditional G-function plot
analysis, where the assumptions of Carter’s leak-off and
constant fracture stiffness during closure (changes
abruptly to infinite at closure stress) are invalid. The
causes for the error associated with “compliance
method” is due to the fact that the increase of fracture
stiffness can happen even before fracturing pressure
drops to minimum in-situ stress.
For a planar fracture geometry, the fracture closes
progressively from its edges to the center with a gradual
increase in stiffness, if both numerical modeling (as
shown in Fig. 5) and tiltmeter measurements (as shown
in Fig.6) only shows smooth curves without a clear
indication of closure stress, then the pressure decline
signature that reflects the combined effects of pressure-
dependent leak-off and fractures stiffness evolution is
not going to yield additional information on closure
stress. The “variable compliance method”, provides an
alternative way to estimate the minimum in-situ stress.
By history matching the DFIT data, using properties of
fracture surface roughness and formation flow capacity,
the ranges of un-propped fracture conductivity and
formation permeability can be obtained for various
fracture dimensions (Wang and Sharma 2018).
In naturally fractured reservoirs, it is difficult to quantify
such information through a history match. In the case of
pressure-dependent permeability, the pressure response
is affected by the declining effective permeability and
the increase of fracture stiffness. Those two independent
mechanisms are intertwined and cannot be separated
from the before-closure data, especially in the case
where the period of pressure-dependent permeability
spans a large portion of before-closure data. In other
words, the information on the individual evolution of
fracture stiffness and effective permeability is lost and
cannot be recovered quantitatively from DFIT data. In
the case of complex fractures, if the closed fracture
segments retain their hydraulic connectivity to the main
hydraulic fracture, we will get a smooth pressure decline
response. In such cases, we do not know what the
topology of the complex fracture is and we will not be
able to constrain the fracture dimension using a simple
material balance calculation. Without a good estimation
of the initial fracture geometry to start with, a
meaningful history match and uncertainty analysis are
not possible. If the closed fracture segments lose
hydraulic connectivity to the main hydraulic fracture, we
will observe multiple pressure derivative peaks on G-
function or square root of time plots. In such cases, we
can discern, at least, how many sets of fractures (with
different orientations) exist. However, we still have no
knowledge of the fracture topology, because the
information on the individual fracture surface area and
their stiffness are smeared together and can’t be
separated. The situation can get more complicated if
some of the closed fracture segments retain their
hydraulic connectivity to the main fracture while others
do not. Nevertheless, our proposed “variable compliance
method” provides an alternative way to estimate in-situ
stress, which can be more reliable than established
methods, even in naturally fractured reservoirs.
6. CONCLUSIONS
In this study, we present a new pressure transient model
for DFIT analysis in naturally fractured reservoirs,
which not only preserves the physics of unsteady-state
reservoir flow behavior, elastic fracture mechanics,
material balance, variable fracture compliance during
fracture closure, but also incorporates pressure-
dependent effective permeability and sequential closure
of complex fracture branches. Conclusions reached from
the analysis presented in this paper include the following:
1) Fracture closure stress (i.e., minimum in-situ stress)
can’t be pinpointed from DFIT data because of the
nature of progressive fracture closure behavior.
2) Heavily fractured reservoirs that exhibit pressure-
dependent effective permeability can partly mask
some information on fracture stiffness (or
compliance) evolution and the associated mechanical
closure process.
3) In sparsely or more moderately fractured reservoirs
with complex fracture geometry, if the closed fracture
maintains its hydraulic connectivity to the open
fracture, then fracture complexity cannot be detected
from before-closure data and the pressure decline
response resembles a fracture with “rougher” fracture
walls. However, if the closed fracture loses its
hydraulic connectivity to the open fracture, then
fracture complexity can be detected qualitatively by
the existence of multiple peaks on the pressure
derivative plot.
4) The conventional “tangent line method” consistently
underestimates closure stress while “compliance
method” tends to only reflect the onset of fracture
closure. Our proposed “variable compliance method”
gives a much more accurate and reliable estimation of
closure stress in both homogenous and naturally
fractured reservoirs.
REFERENCES
1. Barree, R.D., Barree, V.L. and Craig, D., 2009. Holistic
fracture diagnostics: consistent interpretation of prefrac
injection tests using multiple analysis methods. SPE
Production & Operations, 24(03), pp.396-406.
http://dx.doi.org/ 10.2118/107877-PA
2. Craig, D. P., Barree, R. D., Warpinski, N. R., and
Blasingame, T. A. 2017. Fracture Closure Stress:
Reexamining Field and Laboratory Experiments of
Fracture Closure Using Modern Interpretation
Methodologies. Paper presented at the SPE Annual
Technical Conference and Exhibition, 9-11 October,
San Antonio, Texas, USA.
https://doi.org/10.2118/187038-MS
3. Fu, Y., Dehghanpour, H., Ezulike, D. O., and Jones, R.
S. 2017. Estimating Effective Fracture Pore Volume
From Flowback Data and Evaluating Its Relationship to
Design Parameters of Multistage-Fracture Completion.
SPE Production & Operations, Vol (32)04:423-439.
https://doi.org/10.2118/175892-PA
4. Loughry, D., Epps, D. and Forrest, J., 2015. Using Pad
ISIP, DFIT, and ESP Data to Generate a Pore Pressure
Model for the Midland Basin. Paper presented at the
Unconventional Resources Technology Conference,
San Antonio, Texas, USA, 20-22 July.
https://doi.org/10.15530/URTEC-2015-2162973
5. McClure, M.W., Jung, H., Cramer, D.D. and Sharma,
M.M., 2016. The Fracture Compliance Method for
Picking Closure Pressure from Diagnostic Fracture-
Injection Tests. SPE Journal 21(04):1321-1339.
http://dx.doi.org/10.2118/179725-PA
6. Mirani, A., Marongiu-Porcu, M., Wang, H., and
Philippe, E. Production-Pressure-Drawdown
Management for Fractured Horizontal Wells in Shale-
Gas Formations. 2017. SPE Reservoir Evaluation &
Engineering (In Press).
http://dx.doi.org/10.2118/181365-PA
7. Nolte, K. G. 1979. Determination of Fracture
Parameters from Fracturing Pressure Decline. Paper
8341 presented at the SPE Annual Technical
Conference and Exhibition, Las Vegas, Nevada, 23-26
September. http://dx.doi.org/10.2118/8341-MS
8. Nolte, K. G. 1986. A General Analysis of Fracturing
Pressure Decline with Application to Three Models.
SPE Formation Evaluation. Vol (1) 06:571-583.
http://dx.doi.org/10.2118/12941-PA
9. Potocki, D. J. 2012. Understanding Induced Fracture
Complexity in Different Geological Settings Using
DFIT Net Fracture Pressure. Paper SPE 162814
presented at the SPE Canadian Unconventional
Resources Conference, Calgary, Alberta, Canada, 30
October-1 November.
http://dx.doi.org/10.2118/162814-MS
10. Ramurthy, M., Barree, R.D., Kundert, D.P., Petre, J.E.
and Mullen, M.J., 2011. Surface-area vs. conductivity-
type fracture treatments in shale reservoirs. SPE
Production & Operations, 26(04), pp.357-367.
https://doi.org/10.2118/140169-PA
11. Wang, H and Sharma, M.M. 2017a. A Non-Local
Model for Fracture Closure on Rough Fracture Faces
and Asperities. Journal of Petroleum Science and
Engineering, Vol (154):425-437. http://dx.doi.org/
10.1016/j.petrol.2017.04.024
12. Wang, H and Sharma, M.M. 2017b. New Variable
Compliance Method for Estimating Closure Stress and
Fracture Compliance from DFIT data. Paper SPE
187348 presented at the SPE Annual Technical
Conference and Exhibition held in San Antonio, TX,
 USA, 09 – 11 October. http://dx.doi.org/10.2118/
187348-MS
13. Wang, H. 2015. Numerical Modeling of Non-Planar
Hydraulic Fracture Propagation in Brittle and Ductile
Rocks using XFEM with Cohesive Zone Method.
Journal of Petroleum Science and Engineering, Vol
(135):127-140.
http://dx.doi.org/10.1016/j.petrol.2015.08.010
14. Wang, H. 2016. Numerical Investigation of Fracture
Spacing and Sequencing Effects on Multiple Hydraulic
Fracture Interference and Coalescence in Brittle and
Ductile Reservoir Rocks. Engineering Fracture
Mechanics, Vol (157): 107-124.
http://dx.doi.org/10.1016/j.engfracmech.2016.02.025
15. Wang, H. 2017. What Factors Control Shale Gas
Production and Production Decline Trend in Fractured
Systems: A Comprehensive Analysis and Investigation.
SPE Journal, Vol 22(02): 562-581.
http://dx.doi.org/10.2118/179967-PA
16. Wang, H. and Sharma, M.M. 2018. Estimating
Unpropped Fracture Conductivity and Compliance
from Diagnostic Fracture Injection Tests. Paper SPE
189844 presented at the SPE Hydraulic Fracturing
Technology Conference held in The Woodlands, TX,
USA, 23-25 January. http://dx.doi.org/10.2118/189844-
MS
17. Wang, H., Marongiu-Porcu, M., and Economides, M. J.
2016. Poroelastic and Poroplastic Modeling of
Hydraulic Fracturing in Brittle and Ductile Formations,
Journal of SPE Production & Operations, 31(01): 47–
59. http://dx.doi.org/10.2118/168600-PA
18. Wang, H., Yi,S., and Sharma, M.M. 2018. A
Computationally Efficient Approach to Modeling
Contact Problems and Fracture Closure Using
Superposition Method. Theoretical and Applied
Fracture Mechanics, Vol(93): 276-287. doi:
https://doi.org/10.1016/j.tafmec.2017.09.009
19. Willis-Richards, J., K. Watanabe, H. Takahashi. 1996.
Progress toward a stochastic rock mechanics model of
engineered geothermal systems. Journal of Geophysical
Research 101 (B8): 17481-17496,
http://dx.doi.org/10.1029/96JB00882.

View publication stats