Professional Documents
Culture Documents
This paper was prepared for presentation at the 2009 SPE Hydraulic Fracturing Technology Conference held in The Woodlands, Texas, USA, 19–21 January 2009.
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 have not been
reviewed by the Society of Petroleum Engineers and are subject to correction by the author(s). The material does not necessarily reflect any position of the Society of Petroleum Engineers, its
officers, or members. Electronic reproduction, distribution, or storage of any part of this paper without the written consent of the Society of Petroleum Engineers is prohibited. Permission to
reproduce in print is restricted to an abstract of not more than 300 words; illustrations may not be copied. The abstract must contain conspicuous acknowledgment of SPE copyright.
Abstract
Sonic log data and core measurements are often used to develop models of in-situ stress profiles and rock elastic properties
for use in hydraulic fracture design treatments in unconventional reservoirs. “Unconventional” reservoirs, for the purposes of
this paper, include shale-gas, coalbed methane, and tight-gas projects. In unconventional reservoirs, many of the assumptions
underlying rock property estimation and stress profiling in conventional reservoirs may not apply. In fact, the result of using
conventional approaches may be an incorrect and often misleading stress profile. Fracture geometry predicted using
conventionally derived rock properties and stresses might also be inaccurate.
Methods of deriving rock properties from log and core measurements and the effect of various parameters on resulting
moduli and stress estimates are examined. The paper also discusses the effects of rock anisotropy and inhomogeneity on
static and dynamic properties. The impact of organic materials and trapped gas on sonic logs and conventional mechanical
properties interpretation and the use of synthetic sonic logs are also presented. For static and dynamic measurements on core
samples, the effects of the condition of recovered core and applied laboratory procedures on measurement results are also
considered.
Potential errors resulting from the use of inappropriate mechanical properties for stress profiling and fracture geometry
prediction can be significant. The paper identifies common pitfalls in core and log interpretation. A recommended procedure
to determine useful and accurate rock mechanical properties for stress profile prediction and fracture design is presented.
Introduction
Full-waveform sonic logs and core measurements are commonly used to derive “calibrated” models of in-situ stress profiles.
The results are used as input to hydraulic fracture design simulators, which are used to determine optimum perforation
placement, job size, pump rate, fluid and proppant requirements, and other design variables for optimum reserve recovery.
The methods used to determine rock mechanical properties and stresses assume that the raw input measurements are valid
and that the conditions of measurement are appropriate to conditions during hydraulic fracturing. In unconventional
reservoirs, these assumptions may not be valid.
Because of the potential for large reserves, there is an increasing interest in unconventional reservoirs. For the purpose of
this discussion, unconventional reservoirs are considered to be tight and ultra-tight gas sands (less than 0.01 md effective
permeability), gas-shales, and coalbed methane (CBM) reservoirs. These reservoirs have several characteristics in common:
They all have low, very low, or nearly immeasurable “matrix” permeability. They may be self-sourcing and can contain
organic carbon within the hydrocarbon maturation window, and may be actively generating hydrocarbons at the time of
discovery and development. Many have abnormal pore pressures, relative to a hydrostatic gradient. Many occur in regions
with significant tectonic stress or strain overprints, hence anisotropic stress fields. Production from these reservoirs is
commonly enhanced through the presence of some kind of fracture or micro-fracture network.
These complexities affect the behavior of core and log measurements. Interpretation of core and log data may require
different paradigms and assumptions than those commonly applied in more conventional reservoir systems. Using
conventional assumptions when dealing with measurements in unconventional reservoirs can lead to significant errors in the
derived rock mechanical properties and estimated stress profile. These errors can lead to incorrect predictions of fracture
containment and overall geometry, conductivity, and post-frac performance.
horizontal and to follow some form of the uniaxial strain model. Equation 1 gives a slightly expanded form of the uniaxial
strain estimate of minimum horizontal stress.
ν
Pc = σ h = [σ ]
− α v Pp + α h Pp + ε h E + σ t (1)
(1 − ν )
v
The observed fracture closure pressure (Pc) is assumed to equal the minimum horizontal stress (σh) in this model. The
magnitude of σh is assumed to be controlled by the vertical uniaxial strain with externally applied horizontal tectonic stress
and strain offsets. As discussed by Thiercelin (1994), the model is very simplistic considering the complex deposition,
diagenetic, and deformational history of most reservoir systems. Warpinski (1998) suggests that the two poroelastic constants
(αv and αh) are equal for isotropic materials. This is incorrect and misinterprets the physical meaning of the pore pressure
terms in Equation 1. The first term in brackets, involving αv, is the vertical net effective stress causing compaction of the rock
framework. In this term, the internal pore pressure acts against the externally applied overburden stress to reduce net
intergranular stress. The pore pressure term must be corrected for cementation, consolidation, and other poroelasticity effects.
The second pore pressure term, involving αh, does not involved net intergranular stress but refers to internal fluid pressure
only. The pore pressure acts equally in all directions and is in direct hydraulic communication with the fracturing fluid. In
this case, no poroelastic effect should be applied, and the αh term should be set to unity or removed from the equation.
The observed minimum stress commonly differs from that calculated from the uniaxial strain assumption. The differences
may be caused by errors in estimation of elastic properties (α, ν and E), pore pressure, and overburden stress, or through
externally applied horizontal stresses and strains. In Equation 1, only stresses and strains normal to the plane of minimum
stress are considered, although transverse stresses may be generated through other applied strains. Since stresses and strains
other than the minimum cannot be resolved, they will not be considered in detail here. The maximum stress induced by an
applied lateral strain (ε) is limited by the shear failure envelope of the material, as discussed by Thiercelin (1994). Accurate
determination of the shear limit requires knowledge of the complete in-situ net stress tensor and rock strength (in shear) along
the potential plane of weakness. Typically, none of these can be determined from field data. Instead, a practical maximum
strain limit can be set for use in calibrating stress profiles. The practical limit can be derived from an assumed material
cohesion and friction angle along with an estimate of vertical net effective stress. Using the applied strain boundary condition
has been shown to give more accurate derived stress profiles than a constant stress offset (Blanton, 1999). The stress offset
may be useful to describe residual background stress after inset of shear failure in an active fault environment.
Application of Equation 1 requires representative values of rock mechanical properties. Values of Young’s Modulus (E)
and Poisson’s Ratio (ν) can be determined from core samples (both static and dynamic acoustic tests) and from well log
(dynamic acoustic) measurements. The accuracy and applicability of these derived properties has a direct impact on the
calculated stress profile, even when it is “calibrated” to a single point of measured closure stress. When rock properties and
environmental or test conditions affect the apparent mechanical properties, errors in stress and fracture geometry can result.
Unconventional reservoirs offer many opportunities for these errors to occur. In addition, Equation 1 implies that the stress
profile affecting fracture propagation and geometry is the result of an equilibrium deformation state for an elastic medium.
Unconventional reservoirs, especially coal and shale, may behave as ductile or plastic materials rather than elastic ones.
During hydraulic fracturing, the rock must respond to rapidly induced deformation. In very low permeability systems, a rapid
strain rate can outrun the ability to dissipate internal pore pressure. This can result in deformation behavior controlled by
“undrained” moduli instead of “drained” moduli. Many more complex factors must be considered in deriving a stress profile
and fracture geometry in these complex reservoirs.
ν=
(R − 2 ) (2)
(2 R − 2)
E = 13447 ρ b
(3R − 4)
(DTC 2 R(R − 1)) (3)
In these equations, the observed formation bulk density (in g/cm3) is given by ρb, and R is the square of the travel-time
ratio:
DTS 2
R= (4)
DTC 2
SPE 118703 3
The travel times (DTC and DTS) are the reciprocals of the compressional and shear acoustic wave velocities (Vp and Vs)
respectively. Equations 2-4 are equivalent to the dynamic moduli definitions of Lama and Vutukuri as cited by Warpinski
(1998). The mechanical properties derived from sonic measurements are assumed to correlate to “static” measurements made
on core samples. Unfortunately, things other than variations in elastic rock properties significantly affect the sonic velocities.
These other factors include fractures and laminations, external stress, temperature, borehole conditions (breakouts, mud
weight, borehole size, and tool eccentricity), pore pressure, and pore fluid saturation. In many cases, there is no correction or
adjustment made to the observed sonic transit times for any of these effects.
30.0%
20.0%
15.0%
10.0%
5.0%
0.0%
DTSM DTCO R PR YME PR/(1-PR)
Figure 1: Propagation of errors with a 5% error in sonic velocities
Errors of this magnitude obviously have a large effect on the final stress profile calculated using these mechanical
properties. Calculating the same mechanical properties using a correlation based on compressional velocity alone (no shear
input) leads to a reduction in the potential error. Dynamic Young’s Modulus can easily be obtained using compressional
velocity only. Figure 2 shows a regression through 327 tight-gas and shale-gas core samples that gives an excellent estimate
of the dynamic modulus measured under controlled laboratory conditions. Using a lithology volume fraction weighted
average value of ν (based on the curves presented in Mullen, 2008) can give a good estimate of Poisson's Ratio. For a
4 SPE 118703
sand/shale lithology with an assumed +/- 5% error in the compressional velocity, the Poisson’s Ratio error is 9%, with a total
error of 11% for the Young’s Modulus (if the correct shale fraction is known). The error in the ν/(1-ν) term is 14%, or half
the error resulting from using both the shear and compressional velocities in equation 4.
18
16
-2.1557
y = 54857x
14 2
R = 0.9665
12
10
0
40 60 80 100 120 140 160
Vp Transit Time, micro-sec/ft
Figure 2: Estimation of dynamic Young’s Modulus from Vp alone
Generating a stress profile based on the shear/compressional ratio results in the magnification of any inherent errors in the
measurements of both shear and compressional wave slowness. Both the shear and compressional slowness values from
acoustic borehole logs can be subject to significant error of measurement and interpretation. Using a diagnostic injection test
to measure minimum in-situ stress can help to ensure the validity of the calculated stress profile. Using a correlation not
based on shear velocity, and correcting compressional velocity data for saturation and environmental effects also results in
less error prone mechanical properties.
In Equation 5, DTFL is the volume-average fluid acoustic velocity for a gas-water mixture and DTMA is the matrix
transit time. Figure 3 shows the computed change in apparent Poisson’s Ratio (ν) using an assumed DTC at 100% Sw of 64.5
μs/ft (DTMA=53 μs/ft and DTFL=197 μs/ft) and a constant DTS of 126 μs/ft. The assumption of constant DTS is based on
the (probably incorrect) belief that the acoustic shear wave is transmitted only through the solid matrix and is unaffected by
pore fluid. A change in gas saturation from 0% to 60% causes ν to drop from 0.32 to 0.03 for these assumptions. This
magnitude of change is greater than the typical contrast between clean sand and plastic shale, and it drastically affects the
estimated stress profile. Laboratory measurements of sonic velocity in water at various pressures, and in oil at and below the
saturation pressure (Merkel, 2001), show that velocity changes of this magnitude can occur in gas-water systems and in
water-oil systems. In oil reservoirs, the acoustic velocity changes dramatically at the saturation pressure when free gas is
generated. Similar effects can be expected in organic rich shale in the gas generation (maturation) window.
SPE 118703 5
0.35
0.3
Apparent Poisson's Ratio
0.25
0.2
0.15
0.1
0.05
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Gas Saturation
Figure 3: Effect of gas saturation on Poisson’s Ratio for variable DTC with constant DTS
fracture pressure of bounding rocks or of the shale itself (Rondeel, 2001). Some (unreleased) field data confirms pore
pressures in shale nearly equal to the fracture closure stress in surrounding layers. Net effective stress within the shale can be
very low, even at great depths, making the formation susceptible to shear failure at low differential stress. The presence of
gas in the pore space, along with the elevated internal pressure, results in dramatic slowing of the compressional acoustic
signal and abnormally high observed values of DTC.
As an example, Figure 4 shows the total high-resolution GR (HSGR) curve and a total GR composed of the sum of the
scaled spectral GR input curves (GR_SPEC). The individual spectral GR components are plotted in the far-right track of the
figure, after scaling to API units. Scaling is accomplished by multiplying the thorium signal (ppm) by 4.0, uranium (ppm) by
8.0, and potassium volume fraction (v/v) by 1500. These scaling factors have been found to give comparable total GR curves
in most cases, although other scaling factors have been published (http://server4.oersted.dtu.dk/research/RI/SNG/SNG-
logs.html, 2008). The GR_TH+K curve in Figure 4 is the scaled total GR, without the uranium component. This curve,
sometimes referred to as the Computed GR (CGR) gives a better appreciation of actual mineralogy and clay content for the
organic rich shale.
The log section at far left in Figure 4 shows the measured and derived (synthetic) DTC curves for the well. The DTCO
curve is the measured compressional travel time. It is significantly affected by the organic material and generated gas in the
shale. The result is an abnormally slow DTC compared to the synthetic DTC curves that match the measured DTC above and
below the gas/TOC zone. Using the anomalous measured DTC curve will generate incorrect estimates of dynamic elastic
properties for the rock. Use of a synthetic DTC curve, or DTC corrected for the presence of gas saturation and TOC, may
give a better estimate of rock elastic properties. Improved estimates of rock elastic moduli generate more accurate computed
stress profiles when elastic deformation is expected. Only careful calibration of the computed stress profile to a directly
measured stress allows the best of the many possible DTC interpretations to be selected. Many field cases have been studied
in which the measured DTC clearly leads to incorrect stress profiles while synthetic DTC curves give useful results.
DTC from
Resistivity
DTC from
PHIN and GR
DTC Measured
Measured total GR
and GR_SPEC
Passey (1990) technique, which relates TOC percentage to the sonic and resistivity response. In estimating rock properties, it
is advantageous to use the same response to eliminate the gas and TOC effect.
The synthetic DTC curves in Figure 4 were derived using models similar to those presented by Mullen (2007) and
calibrated to the local log conditions in zones expected to be at 100% Sw. Local calibration or scaling of the derived DTC
curves is recommended to detect small variances in the measured DTC that can show trapped gas in the region investigated
by the sonic tool. In organic-rich shale formations, the effect on the measured sonic is typically large, as in the case shown.
The relative magnitudes of the derived and measured DTC curves are systematic and diagnostic. In formations with high
gas or TOC content, the measured DTC will be slowest. The DTC computed from average total porosity (neutron-density
crossplot porosity) will generally be very close to the measured DTC curve as both neutron and density response will be
affected by gas and TOC. The DTC curves derived only from neutron porosity will give the next lower values of DTC, and
DTC from deep resistivity will give the fastest estimate of transit time slowness. In developing synthetic rock mechanical
properties and stress profiles, use of the DTC from resistivity often gives the best correlation to observed closure stress and
fracture height development.
Elastic moduli (ν and E) can be derived from the synthetic DTC curves using lithology volume fractions, as demonstrated
by Mullen (2007). These moduli can also be derived directly from conventional open-hole log responses using simple
models. For example, useful estimates of Poisson’s Ratio can be derived using power-law functions of resistivity and
computed GR, as shown by Equations 8 and 9:
PR_GR=CPG*GREPG (8)
PR_RESIST=CPR*RESISTEPR (9)
The coefficients in Equations 8 and 9 are obtained by correlation of the derived synthetic curves to values of ν computed
from DTC and lithology, or travel-time ratio (R), in areas of good borehole condition and 100% water saturation (usually
non-reservoir intervals). Similarly, good estimates of E can be derived directly from open-hole log responses using simple
linear scaling, power-law, and exponential equations, as shown in Equations 10-12:
YME_GR=CEG*GR+OEG (10)
YME_RESIST=CER*RESISTEER (11)
Other useful estimates of mechanical properties can be derived from various combinations of log measurements,
including cased-hole pulsed neutron logs. The key is that dynamic properties derived from measured full-waveform acoustic
logs can be used to develop correlations when the data from the sonic logs are valid. This is only the case in 100% water-
saturated rocks with good borehole conditions and intact rock with minimal open fractures. Even then, the pore pressure must
be nearly constant across the zone used for calibration. As shown by Merkel, et al (2001) both shear and compressional
velocity vary with net stress to the 1/3 power. If pore pressure changes locally, especially in a hydrocarbon sourcing shale,
then all observed velocities will be poor representations of rock elastic properties. Elastic properties computed from
uncorrected (measured) sonic-derived properties will be wrong, and the estimated stress profile will be incorrect due to both
erroneous elastic moduli and incorrect pore pressure assumptions.
after recovery. In most cases, the observed strains are believed to be caused by opening of fractures normal to the maximum
in-situ stress (Warpinski, 1989, Yassir, 1998). These observations make it a near certainty that a core sample from an
unconventional reservoir will not represent in-situ stress or mechanical properties without some attempt to restore the
sample’s integrity.
Effects of stress and strain relaxation and cycling on static and dynamic moduli
When core samples are subjected to static or dynamic testing in a laboratory, the confining stress, net effective stress,
stress history, pore pressure, temperature, and saturation state may all affect the results. Traditionally, core samples are tested
in the “as received” condition and the primary compaction cycle is used to derive static moduli. If the core has developed
open microfractures during coring and handling, and is partially gas saturated, these measurements may not reflect the
expected in-situ conditions. In general, the core will deform more easily because of the increased compliance of the fracture
network, giving a Young’s Modulus that is too low to represent deformation during fracturing. Applied axial loads will close
fractures and generate less lateral strain than would be the case for an intact sample, leading to abnormally low Poisson’s
Ratio. The absolute magnitude and differential stress state established at the start of the test also has a significant effect on the
observed material properties. Often the confining stress is poorly defined.
DIfferential Stress
E4
E5
E2
E3
E1
Axial Strain
Figure 5: Effect of stress cycling on apparent Young’s Modulus
One way to partially offset the effects of core relaxation and microfracture evolution is to stress-cycle the sample before
taking measurements of stress and strain to describe moduli for stress profiling or fracture geometry estimates. Figure 5
shows a hypothetical stress-strain profile for a core sample under cyclic loading. Five different Young’s Modulus values are
illustrated corresponding to the initial compaction modulus (E1), low net stress tangent modulus (E2), unloading tangent
modulus (E3), high net stress tangent modulus (E4), and high net stress secant modulus (E5). As described by Briaud (2008),
many other modulus values may be defined, and the proper modulus for use in a particular case must be selected to represent
local deformation and stress conditions. In an unconventional reservoir, core E1 is likely to represent a partially failed state
with open fractures, and is not representative of conditions during fracturing. Assuming the rock has not been subjected to
multiple stress cycles over a short time, the unloading modulus (E3) is not representative. The secant modulus (E5) may be
useful to estimate long-term residual deformation after multiple stress cycles, but is not useful for fracture geometry or stress
prediction. This leaves the high or low stress tangent moduli E2 and E4.
During hydraulic fracturing, the rock is subjected to fairly rapid deformation, starting at a stable in-situ stress state. If the
in-situ stress and strain conditions could be maintained throughout the coring and handling process, then measuring the
tangent modulus at the average net effective stress should be the correct conditions (Warpinski, 1998). Because the core must
be re-compacted, a better estimate can more often be obtained using a higher net effective stress than that existing in the
SPE 118703 9
reservoir. If a prolonged linear stress-strain regime exists through several stress cycles, the selection of proper stress state is
simplified and the overall linear trend can be used. In general, the most useful modulus is that which describes the process
being modeled.
120 0.5
DTS 0.45
110
0.4
DTC A Dry
PR A 0.3
PR B
90
ν 0.25
0.2
80
0.15
Dry
0.1
70
DTC
0.05
60 0
0 500 1000 1500 2000 2500 3000 3500 4000 4500
Average Net Stress, psi
Figure 6: Effect of saturation and stress on sonic velocity and apparent Poisson’s Ratio
The initial low-stress velocity measurements on the samples (A and B) were conducted on dry cores. The measurements
were repeated at 100% brine saturation at the same stress. The measured shear slowness, DTS, increases from approximately
112 to 117 μs/ft because of the saturation change alone. Slowing of the shear wave in a liquid saturated medium is not
generally expected, but it is consistently observed in core tests. At the same conditions, the observed DTC values drop from
73 to 66 μs/ft and 70 to 67 μs/ft for the two cores. The combined effect of the saturation change is to increase the apparent
dynamic Poisson’s Ratio (PR in legend) from 0.13 to 0.27 for Sample A and 0.18 to 0.25 for Sample B. The magnitude of the
shift in ν due to saturation change alone is more than the typical difference between clean sand and shale. The change in
acoustic velocity and apparent ν with stress change of 3000 psi is less than the saturation effect.
In addition to saturation, the frequency of the acoustic wave also affects the apparent velocity and calculated dynamic
mechanical properties. The data in Figure 7 show measurements of wet (100% brine saturation) and dry cores over a range of
nearly six orders of magnitude of frequency. The effects of water saturation are consistent with the data in Figure 6: DTS
increases with increasing liquid saturation while DTC decreases slightly with increasing liquid saturation. The apparent
10 SPE 118703
dynamic value of ν increases by 50% on average from the dry to saturated state. As frequency increases, the apparent value
of ν also increases by roughly 40% over the range tested. The frequency range is representative of the difference between
typical laboratory conditions (1000-1,000,000 Hz) and well logging conditions (1-1000 Hz). Clearly, the effects of saturation
are at least as significant as the effects of frequency. Both affect the relationship between laboratory and field (well log)
measurements.
140.0 0.3
120.0
0.25
100.0
0.2
Poisson's Ratio
DT, microsec/ft
80.0
0.15
60.0
0.0 0
1 10 100 1000 10000 100000 1000000
Frequency, Hertz
Figure 7: Effect of frequency and saturation on acoustic velocity and apparent Poisson’s Ratio
Laboratory Procedures
Vertical core samples were obtained from whole core samples, faced to a right cylinder, and tested at the “as received”
saturation. Each sample was inserted into a rubber jacket and a radial Linear Variable Displacement Transducer (LVDT) was
placed around the lateral surface of the sample. The sample was mounted between pistons with ports on the contacting
surfaces for controlling pore pressure, and acoustic crystals for acquiring the sonic travel time. All samples reported here
were tested at 0 psi imposed pore pressure and room temperature. The entire assembly was mounted in a pressure vessel that
allows application of confining pressure and axial stress. The top piston extends through the top of the pressure vessel
enabling the application of axial load. The pressure vessel was then loaded into a computer-controlled load frame where
another LVDT was attached for axial strain measurements. The net reservoir stress was calculated for each sample using the
initial reservoir pressure, overburden stress, and the estimated mechanical properties. The confining and axial pressures were
increased at the same rate to the computed net confining pressure. The low axial stress sonic measurement was obtained at
that point. Data logging was begun and the axial and radial strain were measured as the axial load was increased while
holding the confining pressure constant. At near maximum axial stress, the high axial stress sonic velocity was again
determined.
0.4
Low Axial Stress PR
Hgh Axial Stress PR y = 0.9575x
0.35 Linear (Low Axial Stress PR)
Linear (Hgh Axial Stress PR)
y = 0.9258x
0.3
Static Poisson's Ratio
0.25
0.2
0.15
0.1
0.1 0.15 0.2 0.25 0.3 0.35 0.4
Dynamic Poisson's Ratio
Figure 8: Comparison of core dynamic and static Poisson’s Ratio measurements
the dynamic modulus for both high and low axial stress states. Various models have been proposed to correlate dynamic
modulus to a more representative static modulus. Each of the proposed models is based on a specific core data set and
implied rock type and may not be generally applicable to all rock types. The data sets used here are presented to develop
useful correlations specifically for tight-gas and shale-gas reservoirs.
14.000
Low Axial Stress Dyn E
High Axial Stress Dyn E
12.000
Static Young's Modulus, MMpsi
10.000
8.000
6.000
4.000
2.000
0.000
0.000 2.000 4.000 6.000 8.000 10.000 12.000 14.000
Dynamic Young's Modulus, MMpsi
Figure 9: Comparison of core dynamic and static Young’s Modulus values
Eissa and Kazi (1988) are credited with the early work in the area of dynamic to static modulus conversions. They
presented two models for well consolidated, low porosity samples. The first model proposed a linear scaling relationship
given by Equation 13, while the second model proposed a log linear relationship using the product of core bulk density (ρ, in
gm/cm3) and dynamic modulus (Equation 14). In both the Eissa and Kazi models, the dynamic modulus is given in GPa
(approximately 6.895 times modulus in million psi).
The proposed models (Equations 13 and 14) were applied to the tight-gas and shale-gas data sets. To obtain the best fit
with the current data set the coefficients of Equation 14 were adjusted to those given by Equation 15. The coefficients for
Equation 13 were found to give a best-fit straight line for the current data. In addition to the modified Eissa-Kazi models, a
power law fit of the form suggested by van Heerden (1987) was also tested. The best-fit power-law model is given by
Equation 16.
Figure 10 shows the static Young’s Modulus values for the entire data set computed from the input dynamic modulus. Of
the three equations tested, the modified E-K Log-Linear model (Equation 15) gives the best results by far over the entire
range of moduli tested. It is possible that some of the data scatter is reduced by using measured core bulk density
measurements made under the same saturation conditions as the corresponding dynamic moduli. While saturation state
strongly affects the dynamic Young’s Modulus, the effects of partial saturation on static modulus are not yet known, but are
expected to be minor.
SPE 118703 13
14.000
Eissa-Kazi Linear
Power-Law Model
12.000
Measured Static Young's Modulus, MMpsi
10.000
8.000
6.000
4.000
2.000
0.000
0 2 4 6 8 10 12 14
Computed Static Young's Modulus, MMpsi
Figure 10: Conversion of core dynamic to static Young’s Modulus
Figure 11 shows a simplified drawing of the three core sample orientations for a core through a horizontally laminated
medium. Sample A in Figure 11 shows a vertical plug with an axial load representing the vertical net overburden stress.
Lateral deformation is driven by the axial compression normal to bedding. Loading in this orientation appears to derive the
appropriate Poisson’s Ratio for use in the modified uniaxial strain equation for horizontal stress estimation (Equation 1). Bi-
lateral anisotropy due to microfractures and grain orientation effects will still cause unequal lateral strains that can be
averaged through radial strain measurement.
σz=σv σr≠σv
σz=σh
σr=σh B σr≠σH
σz≠σv
σr≠σH
σr≠σH
Analysis of axial compression of a horizontal core plug parallel to bedding (B) is more problematic. In this case, the
applied axial load more closely represents a maximum horizontal stress, as in a thrust-fault environment. The geometry is
proposed to represent displacement of the fracture surface by fracturing fluid pressure. Here, the uniform radial confining
stress must represent both the vertical stress and the second horizontal stress, which will normally be far from equal in field
conditions. For a normal stress state, the radial stress normal to bedding in this laboratory geometry will be far less than the
vertical net stress. This can lead to enhanced shear failure along bedding. A potentially more serious concern exists if the
sample is not homogeneous. Any bedding layers of high modulus will concentrate load and can give a higher than normal
modulus, as compared to an integrated average value. If stress concentration reaches the shear failure limit, the local hard
streaks may fail before the softer surrounding rock, leading to odd mechanical behavior. In dynamic measurements, first-
arrival acoustic signals will be channeled through the hard streaks, altering interpretations of dynamic properties.
In the case of the 45-degree core orientation (C), the relative angle between the core plug axis and bedding may control
failure and apparent moduli. In the orientation shown, the tendency for shear failure along the (potentially) weak bedding
planes is maximized. All these core measurements will show differences in acoustic velocity and apparent static or dynamic
moduli depending on orientation for locally heterogeneous samples. The difficulty lies in applying these measurements to
field predictions of moduli and stress. In most instances, the only wellbore acoustic wave velocities that can be measured will
be vertical, and they will correspond most closely to the orientation of sample A in Figure 11. The large-scale measurements
will inherently be affected by heterogeneities not captured in intact core plug samples. Overall, the measurement of core
properties at different orientations may provide additional information about rock heterogeneity, but these measurements
offer no solution to the problem of predicting fracture geometry or closure stress profile.
Acoustic anisotropy
Acoustic anisotropy can be divided into two main categories: vertical velocity anisotropy and vertical to horizontal
anisotropy. The use of vertically polarized dipole sonic logs has become common in recent years. In theory, a fast and slow
shear wave velocity can be determined from an orthogonally polarized transmitter-receiver pair. Variations in shear velocity
SPE 118703 15
are theoretically related to changes in stress, but can also be caused by changes in saturation, bedding features, fractures, and
other environmental factors. Differentiating vertical to horizontal anisotropy seeks to address the tendency of formations to
fail in horizontal or low-angle planes (as in delamination or disking of cores). Currently, however, no acoustic borehole
logging tools can actually measure horizontal compressional and shear velocities. Vertical to horizontal anisotropy appears to
be interpreted based on reflected and refracted tube and surface waves. If properly interpreted, a high degree of vertical to
horizontal anisotropy should correlate to better fracture height containment and possibly higher treating pressures.
POIS and
PRRESIST PRACT
PRDTC
DTC
DTCRESIST
The curves in the second track are the reported fast and slow shear slowness reported from the crossed-dipole shear log. A
maximum shear anisotropy of 3 μs/ft out of 150 μs/ft (2%) is observed in a few spots over the logged interval, while most of
the log shows no discernable shear velocity anisotropy. It is interesting to note that this area is highly faulted and has a
known dominant fracture orientation with significant stress anisotropy. It could be that the rock is hard enough that a change
in horizontal stress of more than 500 psi has no measurable effect on shear velocity. Other local factors (saturation, organics,
and fractures) may have larger impacts. Two stress profiles are presented in Figure 12, computed using the log-derived
mechanical properties and Equation 1. The resulting stress profiles are compared to the radioactive tracer log acquired after
the fracture stimulation treatments. The stress profile derived from the acoustic DTS/DTC ratio fails to predict the contained
fracture observed in the lower perforation set. In contrast, the confined fracture is predicted by the resistivity derived stress
profile. In this case both stress profiles would predict a contained fracture geometry for the upper treatment but the
magnitudes of the predicted stresses differ by nearly 1000 psi. In most cases the properties derived from sonic velocity
measurements fail to give useful predictions of fracture geometry or treating pressure.
Conclusions
Estimation of in-situ stress profiles in unconventional reservoirs is complicated by many factors. Determining pore
pressure when matrix permeability is vanishingly small can be nearly impossible by conventional means such as wellbore
measurements or even pressure transient tests. Pore pressure can vary locally and to a large degree when driven by
hydrocarbon maturation. The local pore pressure is a primary driver of the total closure stress. Pore pressure also affects the
net effective stress in the system and can alter acoustic velocities and apparent dynamic moduli.
Acoustic velocities from sonic logs, and dynamic moduli derived from them, can be affected by gas saturation and TOC
in tight formations and shale where near-wellbore flushing while drilling is ineffective. The rock volume interrogated and
integrated by the sonic log may be at a highly variable saturation state, and the measured sonic velocity may respond more to
changes in saturation than to stress or rock property variations. Fractures, laminations, and averaging of dissimilar lithology
response can also affect velocity and dynamic properties derived from borehole sonic logs.
SPE 118703 17
Saturation, stress history, anelastic strain, and many other factors also affect core samples used for both static and
dynamic measurements of elastic moduli. Core data can be used with care to develop models for deformation and correlation
to dynamic properties. Stress cycling and proper saturation of cores is important in deriving useful data for both static and
dynamic core measurements. Core samples typically represent point values of intact rock and will give indications of higher
strength and isotropy than may actually be present in larger scale samples.
Use of core samples at various orientations can provide useful qualitative information about small-scale anisotropy. This
information should be extended to large-scale estimates of stress and deformation with extreme caution. Stress and strain
boundary conditions in these oriented core laboratory tests may not be appropriate for field application.
Use of synthetic sonic logs and elastic moduli is strongly encouraged in unconventional reservoirs. If the derived data
confirms the direct acoustic measurements, then it is possible to derive some degree of comfort. When the synthetic derived
data and measured sonic data disagree, there is usually a good and definable reason. If the cause of the discrepancy can be
traced to pore fluid, pore pressure, or discontinuity effects, then the synthetic data typically give better and more useful
estimates of the necessary rock properties for both stress profile calculation and fracture geometry evolution.
Nomenclature
References
Al-Shayea, N. A.: “Effects of testing methods and conditions on the elastic properties of limestone rock,” Engineering Geology v. 74, pp.
139–156. 2004.
Blanton, T.L., and Olson, J.E.: “Stress Magnitudes from Logs: Effects of Tectonic Strains and Temperature,” SPE Reservoir Evaluation &
Engineering, Volume 2, Number 1, Pages 62-68, February 1999.
Briaud, Jean-Louis, “Intro to Moduli’”, ceprofs.tamu.edu/briaud/, 2008.
Ciccotti, M. and Mulargia, F.: “Differences between Static and Dynamic Elastic Moduli of a Typical Seismogenic Rock,” Geophys. J. Intl.,
v. 157, pp. 474-477, 2004.
Crain, E. R., CRAIN'S PETROPHYSICAL HANDBOOK, http://www.spec2000.net/chapters/Chapter06.htm, Rocky Mountain House,
Alberta, Canada, 2008.
Eissa, A., Kazi, A.: “Relation between static and dynamic Young’s moduli of rocks,” Int. J. Rock Mech. Min. Sci. Geomech. Abstr.
Volume 25, Number 6, Pages 479– 482, 1988.
Holmes, M., Holmes, A., and Holmes, D.: “Modification of the Wyllie Time Series Acoustic Equation, to Give a Rigorous Solution in the
Presence of Gas,” presented at the SPWLA 45th Annual Logging Symposium, Noordwijk, The Netherlands, June 6-9, 2004.
http://Hyperphysics.phy-astr.gsu.edu/hbase/sound/souspe3.html, 2008.
http://server4.oersted.dtu.dk/research/RI/SNG/SNG-logs.html, 2008.
Knight, R., Dvorkin, J. and Nur, A.: “Acoustic Signatures of Partial Saturation,” GEOPHYSICS, v. 63, n. 1, pp. 132–138, Jan-Feb 1998.
La Gesse, J., and Hurley, N.: “Predicting Source Rock Quality using GR Wireline Response and Umaa vs. ρmaa Crossplots in the Lewis
Shale, Green River Basin, Wyoming,” paper SPE 114963 presented at the 2008 SPE Annual Technical Conference and Exhibition,
Denver, Colorado, U.S.A., 21-24 September 2008.
Li, Yongyi and Schmitt, D. R.: “Well-Bore Bottom Stress Concentration and Induced Core Fractures,” AAPG Bulletin, V. 81, No. 11 P.
1909–1925, November 1997.
Lim, S. S., Martin, C. D., and Christiansson, R.: “Estimating In-Situ Stress Magnitudes from Core Disking,” In-Situ Rock Stress – Lu, Li,
Kjerholt, and Dahle (eds), Taylor and Francis Group, London, ISBN 0-415-40163-1, 2006.
Merkel R. H. and Barree, R. D., and Towle, G.: “Seismic Response of Gulf of Mexico Reservoir Rocks with Variations in Pressure and
Water Saturation,” THE LEADING EDGE, pp. 290-299, March 2001.
Miskimins, J. L. and Barree, R. D.: “Modeling of Fracture Height Containment in Laminated Sand and Shale Sequences,” paper SPE
80935 presented at the Production and Operations Symposium held in Oklahoma City, OK, March 22-25 2003.
Miskimins, J. L., Hurley, N., and Graves, R.: “A Method for Developing Rock Mechanical Property Logs using Electrofacies and Core
Data,” paper SPE 77783 presented at the SPE Annual Technical Conference and Exhibition, San Antonio, TX, Sept 29-Oct 2, 2002.
Morales, R. H. and Marcinew, R. P.: “Fracturing of High Permeability Formation: Mechanical Properties Correlations,” paper SPE 26561
presented at the 68th Annual Technical Conference and Exhibition, Houston, TX, October 3-6, 1993.
Mullen, M., Russel Roundtree, R, Barree, R.: “A Composite Determination of Mechanical Rock Properties for Stimulation Design (What
to do When You Don’t Have a Sonic Log),” paper SPE 108139 presented at the 2007 SPE Rocky Mountain Oil & Gas Technology
Symposium, Denver, Colorado, U.S.A., 16–18 April 2007.
Passey, Q. R., Creaney, S., Kulla, J. B., Moretti, F. J., Stroud, J. D.: "A Practical Model for Organic Richness from Porosity
and Resistivity Logs", AAPG Bull., Dec. 1990.
Perreau, P.J., Heugas, O., and Santarelli, F.J.: “Tests of ASR, DSCA, and Core Disking Analyses to Evaluate In-Situ Stresses,” paper
number 17960 presented at the Middle East Oil Show, March 11-14, 1989, Bahrain.
Rickman, R., Mullen, M., Petre, E,, Grieser, B., Kundert, P.: “A Practical Use of Shale Petrophysical Properties for Stimulation Design
Optimization: All Shale Plays are not Clones of the Barnett Shale,” paper SPE 115258 presented at the 2008 SPE Annual Technical
Conference and Exhibition, Denver, Colorado, U.S.A., 21-24 September 2008.
Rondeel, H.E.: “HYDROCARBONS,” Tekst voor de cursus Grondstoffen en het Systeem Aarde (HD 698), pp. 17-30 December 2001.
Schatz, J. F., Olszewski, A. J., and Schraufnagel, R. A.: “Scale Dependence of Mechanical Properties: Application to the Oil and Gas
Industry,” paper SPE 25904 presented at the SPE Rocky Mountain Region/Low Permeability Reservoirs Symposium, Denver, CO,
April 12-14, 1993.
Stieber, S. J., and Thomas, E. C.: “The Distribution of Shale in Sandstones and its Effect upon Porosity,” SPWLA Sixteenth Annual
Logging Symposium, June 4-7, 1975.
Thiercelin, M. J., and Plumb, R. A.: “Core-Based Prediction of Lithologic Stress Contrasts in East Texas Formations,” SPE Formation
Evaluation, pages 251-258, Dec. 1994.
Van Heerden, W.L.: “General relations between static and dynamic moduli of rocks,” Int. J. Rock Mech. Min. Sci. Geomech. Abstr.
Volume 24, Number 6, Pages 381– 385, 1987.
Warpinski, N.R., Peterson, R.E., Branagan, P.T., Engler, B.P., and Wolhart, S.L.: “In Situ Stress and Moduli: Comparison of Values
Derived from Multiple Techniques,” paper SPE 49190 presented at the 1998 SPE Annual Technical Conference and Exhibition held
in New Orleans, Louisiana, 27-30 September 1998.
Warpinski, N.R., Teufel, L.W.: “A Viscoelastic Constitutive Model for Determining In-Situ Stress Magnitudes From Anelastic Strain
Recovery of Core (includes associated papers 19042 and 19892 ),” Journal SPE Production Engineering, Volume 4, Number 3,
Pages 272-280, August 1989.
Yassir, N., Wang, D. F., Enever, J.R., and Davies, P.J.: “Experimental Analysis of Anelastic Strain Recovery of Synthetic Sandstone
Subjected to Polyaxial Stress,” Paper 47238 presented at the SPE/ISRM Rock Mechanics in Petroleum Engineering, Trondheim,
Norway, July 8-10, 1998.