Вы находитесь на странице: 1из 6

SPECIAL

S h a l e sSECTION: Sh
haal el se s

5RFNSK\VLFVRIRUJDQLFVKDOHV
LEV VERNIK and JADRANKA MILOVAC, Marathon Oil Corporation

S everal publications over the last 20 years have established


Downloaded 12/10/17 to 167.205.22.105. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

that black organic shales are generally characterized


by strong velocity anisotropy, low velocity in the bedding-
normal direction, and relatively low density and porosity
(e.g., Vernik and Nur, 1992; Vernik and Liu, 1997). These
rock properties are of interest in seismic attribute studies,
but even more so in geomechanical applications related to
reservoir characterization and hydraulic fracture stimulation.
Organic shales typically, but not necessarily, express strong
vertical transverse isotropy (VTI), with axis of symmetry
perpendicular to the bedding-parallel lamination and clay-
particle-preferred orientation. High-resolution SEM imaging
(Figure 1) immediately reveals all these textural features in
addition to silt grains and the lenticular distribution of the
solid organic matter (kerogen), which in turn is characterized
by significant intraparticle porosity in mature shales. Any TI
medium is described by the five independent elastic constants
(C11, C33, C44, C66, and C13) and detailed multidirectional
ultrasonic core measurements can provide the most accurate Figure 1. Dual-beam SEM-FIB image of Woodford Shale containing
means for their adequate evaluation. Single-well sonic- and 12% lenticular kerogen by volume (black), preferred orientation of
clay aggregates, and scattered silt grains of quartz, carbonates, and
density-log data fall short of the full TI tensor description feldspars. (Courtesy of Bureau of Economic Geology.)
resulting in major ambiguities related to empirical correlations.
In this paper we focus on reported core measurements,
but complement them with in-house core and logging re-
sults as well as heuristic models, which can be adopted in
data analysis and further utilized in predictive mode. We also
present a curious example of the geomechanical application
of these data, specifically in constructing log-based minimum (1)
horizontal stress profiles critical in the process of horizontal
well landing and hydraulic fracturing of these extremely low- where E3 and G13 = C44 are the effective Youngs and shear
permeability rocks. moduli, E3m and G13m are the crack-free (matrix) moduli,
which are slightly stress-dependent, and viso, Eiso, and Giso the
Stress sensitivity best isotropic fit of the anisotropic matrix properties (Sevos-
Elastic properties of mature organic shales are characterized tianov and Kachanov, 2008). We assume the latter can be
by enhanced stress-sensitivity due to bedding-subparallel mi- obtained as linear extrapolation from the high-stress moduli
crocracks that originate in the process of hydrocarbon genera- measured in three directions with respect to the symmetry axis
tion and overpressure developement (e.g., Vernik and Landis, into the low-stress domain. The model suggests that elastic
1996). It is interesting that immature and some postmature compliances are proportional to the crack density parameter ,
shales have reduced stress sensitivity, especially in the normal which in turn can be used to constrain the horizontal-perme-
pore-pressure environments. Consistent with geomechanical ability enhancement due to cracks. Progressive crack closure
applications (Sayers, 2010), we will focus on the anisotropic during confining pressure application will gradually change
engineering parameters such as Youngs moduli (E1 and E3) the crack density, which can be predicted to decay exponen-
and Poissons ratios (v13 , v12, and v31) computed from ultra- tially according to = o exp(-dm),where o is the zero-stress
sonic (dynamic) elastic tensor measurements (Cij). crack density and d is a function of crack aspect ratios and
Let us consider the bedding-normal elastic stiffnesses of roughness of crack faces controlling the crack-closure stress.
one of the most stress-sensitive shale samples coming from the The measurement results and the model predictions
deeper part of Bakken Formation, immediately below the oil are shown in Figure 2a. We find an excellent match for the
window (Vernik, 1993). This shale core can be approximated Youngs modulus and a slight underprediction in the shear,
as a medium with TI matrix enhanced by partially oil- and which can be easily accounted for by the likely inaccuracy of
gas-saturated bedding-parallel microcracks, so the following the linear extrapolation to determine the crack-free response.
theoretical model (Sevostianov and Kachanov, 2001) is appro- The vertical effective stress mv for the core location is estimat-
priate for the bedding-normal direction: ed to be around 27 MPa, meaning that the vast majority of

318 The Leading Edge March 2011


Shales
Downloaded 12/10/17 to 167.205.22.105. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

Figure 2. Measurements (blue data points, from Vernik, 1993) and model predictions (red lines) of elastic stiffening with isostatic confining
pressure on partially oil-saturated Bakken Shale from the wet-gas window (a). Relationship between and b for the same core sample (b).

Figure 3. (a) Relationship between compressional and shear velocity for bedding-normal (hot colors) and bedding-parallel (cold colors) ultrasonic
core measurements from Bossier Shale at 15 MPa (squares) and Bakken and Woodford shales at 70 MPa (circles). (b) The same template for
dipole sonic data obtained in wells in Bakken, Woodford, and Bossier shales.

bedding-subparallel maturation cracks may be open in situ Our ability to understand the relationships among the
and detectable by sonic logs. Even more intriguing is the rela- three Thomsen anisotropy parameters is critical in constrain-
tionship between the anisotropic parameters and b, showing ing the TI tensor from sonic logs, VSP, and seismic data. As
a crossover from < b to > b at about the same stress level we will see later, this change in b/ ratio with crack closure has
in this partially gas-saturated oil shale (Figure 2b). Of course, important implications in geomechanics as well.
this Bakken Shale core shows the greatest stress sensitivity in
our data set; nonetheless we can argue that a typical range VP versus VS
of b/ ratio variation for the organic shales in situ lies in the In order to review the VP versus VS relationship in organic
range from 0.4 to 0.8i.e., the anisotropy of these rocks is shales, we have compiled several published data sets (Vernik
positively anelliptical (Vernik and Liu, 1997). and Nur, 1992; Vernik and Liu, 1997) and in-house core and

March 2011 The Leading Edge 319


Shales

log data from Bossier, Woodford, and Bakken shales. The


database covers a wide range of total organic carbon (TOC)
content (315 wt%), XRD-based clay content (2050%),
porosity (512%), saturation (wet, gas, and oil shales), and
effective stress (1330 MPa).
Bedding-normal (0 incidence) and bedding-parallel (90
Downloaded 12/10/17 to 167.205.22.105. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

incidence) ultrasonic core measurements are plotted in Fig-


ure 3a. For reference, both inorganic shale and gas-saturated
arenite sand trends (Vernik and Kachanov, 2010) are added
as well as some constant VP/VS lines. The plot immediately re-
veals the peculiarity of the organic shales characterized by the
relatively narrow range of VP/VS from 1.6 to 1.7, independent
of the wave-propagation direction. It is obvious that none of
the existing models relating shear- to compressional-wave ve-
locity can be applied to shales containing even a relatively
small amount of kerogen.
Figure 3b shows the same VP versus VS template for the
dipole sonic data acquired in three vertical wells. The vast ma-
jority of data points still fall within the similarly narrow VP/VS
range regardless of saturation, porosity, or effective stress. In-
deed, all these parameters seem to be secondary in controlling
the reduced velocity ratio typical of organic shales as com- Figure 4. Bulk density versus total organic carbon (TOC) for several
pared to their inorganic counterparts. Again, because of its cores sampled from organic shale formations worldwide superposed by
thinly-laminated distribution, even a relatively small amount model realization for 5% porosity (green) and 10% porosity (red).
of kerogen typical of Bossier Shale (710% by volume of the Uncertainty range for mineral density on kerogen-free basis, lnk, is
shown for each realization.
solid rock matrix corresponds to TOC content of 34% by
weight) may cause a dramatic shift in VP(0)/VS(0) toward
the gas-saturated sand trend; whereas any deviation of the The solid matrix density lm is given by nonkerogen and kero-
log data acquired in organic shales back toward the inorganic gen components:
shale trend can only be interpreted in terms of the significant
subvertical fracture density. (3)

Effects of kerogen and porosity where kerogen density can vary from 1.21.4 g/cm3 depend-
There are several challenges in modeling composition and ing on maturity; the nonkerogen component usually consists
porosity effects on velocities in organic shales, which are of clay, quartz, carbonates, and pyrite; and the density of this
critical in mapping their organic richness and reservoir prop- component lnk normally varies between 2.722.81 g/cm3. Fi-
erties from seismic amplitudes and/or impedances. First, nally, porosity was computed from bulk density as = (lm
their porosity is not easily and unambiguously determined lb)/(lm lf) with lk =1.3 g/cm3, lf =0.8 g/cm3 (oil shale),
from either core or log data. Second, it is unclear how we and lf = 0.5 g/cm3 (gas shale).
can employ micromechanics modeling for shales in general, It is evident from Figure 4 that the vast majority of our or-
because these lithologies clearly defy the main assumptions ganic shales fall into the 510% porosity range with Bakken,
of Gassmanns theory and their dry-frame compliances are Woodford, and Bazhenov shales on the lower and Bossier
hard to define due to rock-fluid interaction effects. For prac- shales on the high end of this range. Hence, the bulk-density-
tical applications, we choose to employ (1) the mass-balance- log data can be successfully employed in porosity modeling
based porosity computation using a combination of TOC if the kerogen volume and mineral composition are properly
and XRD mineralogy data to derive the solid matrix density constrained.
and (2) the simple empirical model relating porosity to elas- Backus averaging can be used as a first approximation
tic stiffnesses and velocities measured in the bedding-normal to derive the bedding-normal compressional stiffness of the
direction. solid matrix if we partition it into the following three compo-
From the mass-balance equation the relationship between nents: (1) total clay, (2) kerogen, and (3) silt:
volume of kerogen (K) and TOC is given by:
(4)
(2)
where vcl is the total clay content in the solid matrix volume,
where Ck (0.70.85 depending on maturity level) is the con- C33cl = 33.4 GPa accounts for the effective, imperfect clay
stant relating kerogen to TOC, is the porosity, and lb, lf, mineral preferred orientation (Vernik and Kachanov, 2010),
and lk are the bulk, fluid, and kerogen densities, respectively. C33K = Mk = 9.5 GPa (based on VPk = 2.7 km/s), and C33ncl

320 The Leading Edge March 2011


Shales

ness to porosity is maximally simple, does not take into ac-


count the effective stress, and is based on the loosely defined
concept of the critical porosity of mud:

(5)
Downloaded 12/10/17 to 167.205.22.105. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

where p = [log(C33c) log(C33m)]/log(1 c) with


defined as the mud stiffness at the critical porosity of 40%
making sure that the model trend lines converge on the criti-
cal porosity point. This model has been shown to accurately
fit the worldwide inorganic shale trends (Vernik and Kacha-
nov, 2010) when VPc = 1.51.6 km/s (i.e., close to the mud
suspension velocity). The difficulty now is in selecting the ap-
propriate VPc to be used for mature organic shales that have
significant hydrocarbon saturation. We have selected, some-
what arbitrarily, VPc = 1.0 km/s. Finally, the bedding-normal
velocity was computed as VP(0) = {C33/[lm(1 ) + lf ]}1/2.
Figure 6 shows several modeled velocity trends in the
VP(0) versus porosity plane, where each line represents ker-
ogen contents of 0, 10, 20, and 30%. Superposed in these
Figure 5. Solid matrix stiffness C33m model based on Backus theory templates are the sonic velocity and bulk-density-log-based
for two different realizations of clay content. porosity for two Bakken and Woodford wells (Figure 6a) and
a Bossier well (Figure 6b). Notice the clay content effect be-
Mquartz = 97.0 GPa. Incidentally, the shear-wave velocity of tween the two data sets. Core TOC data from these three
kerogen should be of the order of VSk = 1.8 km/s to fit the wells yielded the following statistics for kerogen content: (1)
experimental data reported. Bakken, K = 0.270.03, (2) Woodford, K = 0.130.04, and
Clay content (predominantly illite) in organic shales rang- (3) Bossier, K = 0.080.03. Thus, referring to Figure 7, we
es from 50% in Bossier and Bazhenov to 2025% in more obtain a satisfactory match to our model results, which can
siliceous and carbonate-rich shales of the Bakken, Niobrara, be used for sonic porosity computations as well as forward
and Woodford formations. Figure 5 displays C33m versus kero- modeling of seismic response.
gen models for vcl ranging from 25 to 45% and confirms that
the total clay content has a measurable effect on elasticity, Application: Stress profiling
second only to the kerogen concentration. Geomechanical applications of core/log data in organic shale
Our empirical model relating the bedding-normal stiff- are numerous, but the most critical question that is yet to be

Figure 6. Compressional velocity versus porosity for log data acquired in vertical wells in (a) Bakken and Woodford shales and (b) Bossier Shale.
Mean values of kerogen content computed from core TOC data are 0.27, 0.13, and 0.08 for Bakken, Woodford, and Bossier, respectively. Note a
small and unaccounted for effect of the lower effective stress in Bossier.

March 2011 The Leading Edge 321


Shales

by Katahara (2009), most models based on the Shmin equa-


tion above contain questionable assumptions (e.g., Biots po-
roelastic constant equal to unity) and often are limited by
data availability.
To test the isotropic versus anisotropic stress-coupling-
factor computation, we have selected our Woodford well. The
Downloaded 12/10/17 to 167.205.22.105. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

isotropic ko was computed from VP(0)/VS(0) as logged


by the dipole sonic tool, while the anisotropic ko = C13/C33,
where C13 was constrained by the b/ ratio range from 0.4 to
0.8, as indicated before. To estimate , we used our database
of organic shales to derive an empirical formula = b exp[
VP(0)], where b = 102. Because C13 is a known function of
b, C33, and C44 (Thomsen), we can estimate the likely range
for anisotropic ko with the fair amount of certainty. Inciden-
tally, the isotropic computation of ko is a special case (b =
0) of the more general anisotropic formula, which is reduced
to C13/C33 = 1 2(C44/C33) when b = 0, allowing the coupling
factor to be computed from bedding-normal velocity ratio
squared.
The results of this exercise are presented in Figure 7. It
is clear that the anisotropic solution increases the stress-cou-
pling factor with the b/ ratio; however, it should be em-
phasized that the final Shmin computation is to a larger degree
affected by the pore pressure and the pore-pressure gradient
Figure 7. Log-derived stress profile (isotropic and anisotropic was as high as 0.71 psi/ft in the Woodford Shale, leading
realizations) in Woodford Shale. Three realizations correspond to b = 0 to a significant overall dilution of the anisotropic effect on
(isotropic), b = 0.4 (index 1), and b = 0.8 (index 2), where was
derived empirically from the VP(0) log. Minifrac test result, shown in ko. Even more curious is that the minifrac data in this shale
the last track, matches the isotropic case. (Figure 7) yielded Shmin which is better approximated by the
incorrect isotropic solution reminding us once again that
answered is: Can we use the logs in computing/constrain- sometimes two wrongs can make a right.
ing the magnitude of the minimum horizontal stress Shmin?
There is no doubt that minifrac tests provide the most reli- Conclusions
able information on Shmin, yet they are very expensive and do The main controls on elastic properties of organic shales are
not allow the computation of continuous stress or a fracture kerogen content, porosity, clay content, and effective stress.
gradient profile. It has been noticed in some unconventional A high level of velocity anisotropy is primarily related to the
plays that a simple uniaxial strain solution such as Shmin = lenticular distribution of kerogen and clay-mineral-preferred
ko(Sv Pp) + Pp with the stress coupling factor ko = v/(1 v) orientation parallel to the bedding plane. The TI anisotropy
sometimes approximates the fracture closure stress, which is further enhanced by bedding-subparallel microcracks,
is the most reliable measure of Shmin. Poissons ratio is sim- which can survive in-situ stresses especially in highly over-
ply computed from the VP/VS ratio assuming the formation pressured shales.
is isotropic. More sophisticated methods add lateral strain All organic shales with TOC content in excess of 3%
components to possibly account for tectonic stresses and in- (about 7.5% of kerogen volume in the solid shale matrix) are
troduce anisotropic ko = C13 /C33 (Thomsen, 1986). As argued characterized by a relatively low bedding-normal VP/VS ratio

322 The Leading Edge March 2011


Shales

ranging from 1.6 to 1.7 and approaching gas-saturated sand


trends on the lower velocity end. Theoretical micromechan-
ics-based modeling of shales in general and organic shales in
particular is highly problematic due to the non-Gassmann
nature of these rocks. Alternatively, empirical models ac-
counting for compositional and textural characteristics can
Downloaded 12/10/17 to 167.205.22.105. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

have practical applicability in sonic porosity derivation and


forward modeling of sweet spotshigher kerogen content
and higher permeability due to higher crack density.
Dipole sonic data can be quite useful for continuous
fracture gradient profile computation in boreholes penetrat-
ing organic shale formations. Empirical correlations among
anisotropic parameters and bedding-normal velocities, which
are readily attainable in vertical wells, may help constrain the
popular approach of fracture gradient estimation and shed
more light on the anisotropic effects of stress coupling.

References
Katahara, K., 2009, Lateral earth stress and strain: 79th Annual Inter-
national Meeting, SEG, Expanded Abstracts, 21652169.
Sayers, C., 2010, Geophysics under Stress: Geomechanical Applica-
tions of Seismic and Borehole Acoustic waves: SEG 2010 Distin-
guished Instructor Short Course.
Sevostianov, I. and M. Kachanov, 2001, Authors response: Jour-
nal of Biomechanics, 34, no. 5, 709710, doi:10.1016/S0021-
9290(00)00208-6.
Sevostianov, I. and M. Kachanov, 2008, On approximate symme-
tries of the elastic properties and elliptic orthotropy: International
Journal of Engineering Science, 46, no. 3, 211223, doi:10.1016/j.
ijengsci.2007.11.003.
Thomsen, L., 1986, Weak elastic anisotropy: Geophysics, 51, no. 10,
19541966, doi:10.1190/1.1442051.
Vernik, L. and A. Nur, 1992, Ultrasonic velocity and anisotropy
of hydrocarbon source rocks: Geophysics, 57, no. 5, 727735,
doi:10.1190/1.1443286.
Vernik, L., 1993, Microcrack-induced versus intrinsic elastic anisot-
ropy in mature HC-source shales: Geophysics, 58, no. 11, 1703
1706, doi:10.1190/1.1443385.
Vernik, L. and C. Landis, 1996, Elastic anisotropy of source rocks:
Implications for hydrocarbon generation and primary migration:
AAPG Bulletin, 80, no. 4, 531544.
Vernik, L. and X. Liu, 1997, Velocity anisotropy in shales: A petrophysi-
cal study: Geophysics, 62, no. 2, 521532, doi:10.1190/1.1444162.
Vernik, L. and M. Kachanov, 2010, Modeling elastic proper-
ties of siliciclastic rocks: Geophysics, 75, no. 6, 171182,
doi:10.1190/1.3494031.

Acknowledgments: We thank Marathon Oil Corporation for per-


mission to publish this paper, and Mark Kachanov (Tufts Univer-
sity) and Colin Sayers (Schlumberger) for constructive comments.

Corresponding author: lvernik@marathonoil.com

March 2011 The Leading Edge 323

Вам также может понравиться