You are on page 1of 17

GEOPHYSICS, VOL. 75, NO. 5 共SEPTEMBER-OCTOBER 2010兲; P. 75A31–75A47, 20 FIGS.

10.1190/1.3483770

Rock-physics diagnostics of depositional texture, diagenetic alterations,


and reservoir heterogeneity in high-porosity siliciclastic sediments
and rocks — A review of selected models and suggested work flows

Per Avseth1, Tapan Mukerji2, Gary Mavko3, and Jack Dvorkin3

INTRODUCTION
ABSTRACT
Rock physics provides a link between geologic reservoir parame-
Rock physics has evolved to become a key tool of reservoir ters 共e.g., porosity, clay content, sorting, lithology, saturation兲 and
geophysics and an integral part of quantitative seismic inter- seismic properties 共e.g., acoustic impedance, P-wave/S-wave veloc-
pretation. Rock-physics models adapted to site-specific dep- ity ratio VP / VS, bulk density, and elastic moduli兲. Rock-physics
osition and compaction help extrapolate rock properties models can be used to interpret observed sonic and seismic veloci-
away from existing wells and, by so doing, facilitate early ex- ties in terms of reservoir parameters or to extrapolate beyond the
ploration and appraisal. Many rock-physics models are avail- available data range to examine certain what-if scenarios, such as
able, each having benefits and limitations. During early ex- plausible fluid or lithology variations. Along this line, rock physics
ploration or in frontier areas, direct use of empirical site-spe- can be used to forecast seismic response to assumed reservoir and
cific models may not help because such models have been overburden properties and conditions.
created for areas with possibly different geologic settings. At Rock-physics models also help infer 共diagnose兲 rock texture of
the same time, more advanced physics-based models can be sandstones or shales if we know porosity and elastic-wave velocity.
too uncertain because of poor constraints on the input param- Such diagnostics assume that, e.g., if velocity-porosity data fall on a
eters without well or laboratory data to adjust these parame- theoretical cemented-rock trend, the rock is cemented. This seem-
ters. A hybrid modeling approach has been applied to silici- ingly circular logic helps us better understand rock properties be-
clastic unconsolidated to moderately consolidated sedi- yond elasticity. For example, if rock is cemented, one may expect
ments. Specifically in sandstones, a physical-contact theory higher strength than in uncemented rock of the same porosity and
共such as the Hertz-Mindlin model兲 combined with theoretical mineralogy. It is also likely that the permeability 共at the same porosi-
elastic bounds 共such as the Hashin-Shtrikman bounds兲 mim- ty兲 of cemented rock is higher than that of uncemented rock. This ef-
ics the elastic signatures of porosity reduction associated fect has a simple physical explanation: Loose pore-filling material
with depositional sorting and diagenesis, including mechani- 共or noncontact cement兲 increases the specific surface area and thus
cal and chemical compaction. For soft shales, the seismic decreases permeability, as opposed to pore-filling material occurring
properties are quantified as a function of pore shape and oc-
as contact cement 共Bosl et al., 1998; Dvorkin and Brevik, 1999兲.
currence of cracklike porosity with low aspect ratios. A work
Local geologic trends can help constrain rock-physics models.
flow for upscaling interbedded sands and shales using
Such trends can be split into two types: compactional and deposi-
Backus averaging follows the hybrid modeling of individual
tional. If we can predict the expected change in seismic response as a
homogenous sand and shale layers. Different models can be
function of depositional environment or burial depth, we will in-
included in site-specific rock-physics templates and used for
crease our ability to locate hydrocarbons, especially where little or
quantitative interpretation of lithology, porosity, and pore
no well-log information is available. Understanding the geologic
fluids from well-log and seismic data.
constraints in an area of exploration reduces the range of expected
variability in rock properties and hence reduces the uncertainties in
seismic reservoir prediction.

Manuscript received by the Editor 11 January 2010; revised manuscript received 11 May 2010; published online 14 September 2010.
1
Odin Petroleum, Bergen, Norway. E-mail: per.avseth@odin-petroleum.no.
2
Stanford University, Energy Resource Engineering Department, Stanford, California, U.S.A. E-mail: mukerji@stanford.edu.
3
Stanford University, Department of Geophysics, Stanford Rock Physics Laboratory, Stanford, California, U.S.A. E-mail: mavko@stanford.edu; dvorkin@
stanford.edu.
© 2010 Society of Exploration Geophysicists. All rights reserved.

75A31
75A32 Avseth et al.

Several workers 共e.g., Vernik and Nur, 1992; Dvorkin and Nur, to describe a mixture of the newly deposited sediment at critical po-
1996; Anselmetti and Eberli, 1997; Florez, 2005兲 recognize that the rosity with additional mineral instead of describing a mixture of
slope of velocity-porosity 共or impedance-porosity兲 trends in sand- mineral and pore fluid. A slight improvement over the modified up-
stones is highly variable and depends largely on the geologic process per HS bound as a diagenetic trend for sands can be obtained by
that controls porosity 共Figure 1兲. Relatively steep velocity-porosity steepening the high-porosity end. An effective way to do this is to
trends for sandstones are representative of porosity variations con- use Dvorkin’s model 共Dvorkin and Nur, 1996兲 for cementing grain
trolled by diagenesis, i.e., porosity reduction from pressure solution, contacts. The contact-cement model captures the rapid increase in
compaction, and cementation. Hence, we often see steep velocity- elastic stiffness of a sand with little change in porosity as the first bits
porosity trends when examining data spanning a great range of of cement are added. The depositional or sorting trends can be de-
depths or ages. The classical empirical trends of Wyllie et al. 共1956兲, scribed with a series of modified HS lower bounds. By combining a
Raymer et al. 共1980兲, Han 共1986兲, and Raiga-Clemenceau et al. contact-cement model with such sorting trends, we can create lines
共1988兲 show versions of the steep, diagenetically controlled veloci- of constant depth but variable texture, sorting, and/or clay content
ty-porosity trend. On the other hand, porosity change resulting from 共Avseth et al. 关2000兴 also refer to these as constant-cement lines兲. Fi-
variations in sorting and clay content tend to yield much flatter ve- nally, we demonstrate how we can use these models to quantify the
locity-porosity trends, meaning that porosity controlled by sedimen- rock texture of sandstones based on North Sea well data.
tation is generally expected to yield flatter trends, which we some- To interpret the observed seismic contrast, we also need to know
times refer to as depositional trends. Data sets from narrow depth the rock properties of shales. These require more complex rock-
ranges or individual reservoirs often 共though not always兲 show this physics models to capture cracklike pore shapes, intrinsic mi-
behavior 共Avseth et al., 2005兲. croporosity, and diagenetic mineral transitions. An upscaling ap-
A range of different models can be used in rock-physics analysis proach is discussed to model the effective rock-physics properties of
共Dræge et al., 2006a; Mavko et al., 2009兲. Every model has certain sands interbedded with shales.
advantages and limitations. We follow Box and Draper 共1987兲 in be-
lieving that “All models are wrong, but some are useful.” Most rock-
physics models relevant to the scope of this paper are aimed at de- ROCK-PHYSICS MODELS — AN OVERVIEW
scribing relations between measurable seismic parameters and rock/
Theoretical models
fluid properties. Although our intent is not to review all models ex-
haustively, many fall within three general classes: theoretical, em- Theoretical models are primarily continuum-mechanics approxi-
pirical, and heuristic. We provide a condensed discussion of mations of the elastic, viscoelastic, or poroelastic properties of
different modeling approaches as well as a more detailed analysis of rocks. Among the most famous are the poroelastic models of Biot
our hybrid approach, where we combine contact theory 共granular 共1956兲, who was among the first to formulate the coupled mechani-
media兲 or pore-shape-constrained models with heuristic bounds to cal behavior of a porous rock embedded with a linear viscous fluid.
predict sedimentary microstructure and geologic trends from elastic The Biot equations reduce to the famous Gassmann 共1951兲 relations
properties. at zero frequency; hence, we often refer to Biot-Gassmann fluid sub-
In particular, diagenetic trends, which connect newly deposited stitution. Elastic models tend to be bounds, inclusion models, dis-
sediment on the Reuss elastic bound with the mineral point at zero placement-discontinuty models, contact models, computational
porosity, often can be described accurately using the upper Hashin- models, and transformations.
Shtrikman 共HS兲 bound 共Hashin and Shtrikman, 1963兲. In fact, we Bounds such as the Voigt-Reuss 共VR兲 or HS are the silent heroes
sometimes refer to it as a modified upper HS bound because we use it of rock models. They are extremely robust and relatively free of ide-
alizations and approximations, other than representing the rock as an
6000 elastic composite. Originally, bounds were treated only as describ-
ing the limits of elastic behavior; some geophysicists even consid-
ered them of limited usefulness. However, they have turned out to be
5000
valuable mixing laws that allow accurate interpolation of sorting and
Processes that
cementing trends and are the rigorously correct equations to describe
P-wave velocity (m/s)

give sediment strength:


4000 compaction, stress, suspensions and fluid mixtures.
diagenesis Inclusion models usually approximate the rock as an elastic solid
Voigt avg. containing cavities 共inclusions兲, representing the pore space. Be-
3000 cause the inclusion cavities are more compliant than solid mineral,
Newly deposited they have the effect of reducing the overall elastic stiffness of the
clean sand rock in an isotropic or anisotropic way. Most inclusion models as-
2000
sume that the pore cavities are ellipsoidal or penny shaped 共Eshelby,
Reuss bound
Suspensions
1957; Walsh, 1965; Kuster and Toksöz, 1974; O’Connell and Budi-
1000 ansky, 1974, 1977; Cheng, 1978, 1993; Hudson, 1980, 1981, 1990;
0 20 40 60 80 100
Crampin, 1984; Johansen et al., 2002兲. Berryman 共1980兲 generalizes
Suspensions Porosity (%)
Sand-clay mixt. the self-consistent description so that the pores and grains are con-
Sand
Clay-free sandstone
Clay-bearing sandstone
sidered to be ellipsoidal inclusions in the composite. Inclusion mod-
els have contributed tremendous insights as elastic analogs of rock
Figure 1. P-wave velocity versus porosity for a variety of water-satu- behavior. However, their limitation to idealized 共and unrealistic兲
rated sediments, compared with the Voigt-Reuss bounds. Data from pore geometries makes comparing the models to actual pore micro-
Yin 共1992兲, Han 共1986兲, and Hamilton 共1956兲. geometry difficult.
Rock physics analysis of siliciclastics 75A33

Displacement discontinuity or excess compliance models have clay behavior in sandstones, the Greenberg-Castagna 共1992兲 VP / VS
been used to model the long-wavelength elastic behavior of rocks relations, and the Gardner et al. 共1974兲 VP-density relations.
with aligned fractures 共Schoenberg and Douma, 1988; Sayers and Empirical relations are sometimes disguised as theoretical. For
Kachanov, 1995; Schoenberg and Sayers, 1995兲. These do not as- example, the popular model of Xu and White 共1995兲 for VS predic-
sume elliptical inclusions but approximate the fractures as infinitesi- tion in shaly sands is based on the Kuster-Toksöz ellipsoidal inclu-
mally thin planes of displacement discontinuity, parameterized by sion formulation. One unknown aspect ratio is assigned to represent
their normal and shear compliances, representing the extra compli- the compliant clay pore space, and a second unknown aspect ratio is
ance of the fractures relative to the background rock. Use of these assigned to represent the stiffer sand pore space. These aspect ratios
models requires estimating the fracture compliances 共or stiffnesses兲 are determined empirically by calibrating to training data. In other
empirically or with model-based methods such as penny-shaped words, this is an empirical model in which the function form of the
cracks. regression is taken from an elastic model. It is useful to remember
Contact models approximate the rock as a collection of separate that all empirical relations involve this two-step process of a model-
grains whose elastic properties are determined by the deformability ing step to determine the functional form, followed by a calibration
and stiffness of their grain-to-grain contacts. Most of these 共Digby, step to determine the empirical coefficients.
1981; Walton, 1987; Norris and Johnson, 1997; Makse et al., 1999;
Bachrach and Avseth, 2008兲 are based on the Hertz-Mindlin 共HM;
Mindlin, 1949兲 solution for the elastic behavior of two elastic Heuristic models
spheres in contact. The key parameters determining the stiffness of Heuristic models are what we might call pseudotheoretical.Aheu-
the rock are the elastic moduli of the spherical grains and the area of ristic model uses intuitive, though nonrigorous, means to argue why
grain contact, which results from the deformability of the grains un- various parameters should be related in a certain way. The best-
der pressure. Dvorkin and Nur 共1996兲 describe the effect of adding known heuristic rock-physics model is the Wyllie time average, re-
small amounts of mineral cement at the contacts of spherical grains. lating velocity to porosity: 1 / V ⳱ ␾ / Vfluid Ⳮ 共1 ⳮ ␾ 兲 / Vmineral. At
As with inclusion models, spherical contact models are useful elastic face value, it looks as if there might be some physics involved. How-
analogs of soft sediments, but they suffer from their extremely ideal- ever, the time-average equation is equivalent to a straight-ray, zero-
ized geometries. They are not easy to extend to realistic grain shapes wavelength approximation, which makes no sense when modeling
or distributions of grain size. Furthermore, the most rigorous part of wavelengths that are very long relative to grains and pores. The Wyl-
the contact models is the formal description of a single grain-to- lie equation is sometimes a useful heuristic description of clean, con-
grain contact. To extrapolate this to a random packing of spheres re- solidated, water-saturated rocks, but it is certainly not a theoretically
quires sweeping assumptions about the number of contacts per grain justifiable one.
and the distribution of contact forces throughout the composite. Other very useful heuristic models are the use of the HS upper and
Computational models are a fairly recent phenomenon 共e.g., lower bounds to describe cementing and sorting trends. Certainly the
Bakke and Øren, 1997; Keehm, 2003; Knackstedt et al., 2009兲. In HS curves are rigorous bounds for mixtures of different phases.
these, the actual grain-pore microgeometry is determined by careful However, we use the bounds as interpolators to connect the mineral
thin section or computed-tomography 共CT兲 scan imaging. This ge- moduli at zero porosity, with moduli of well-sorted end members at
ometry is represented by a grid, and the elastic, poroelastic, or vis- critical porosity. We give heuristic arguments justifying why an up-
coelastic behavior is computed using finite-element, finite-differ- per-bound equation might be expected to describe cementing, which
ence, or discrete-element modeling. Clear advantages of these mod- is the stiffest way to add mineral to a sand, and why a lower-bound
els are freedom from geometric idealizations and the ability to elasti- equation might be expected to describe sorting. However, we are un-
cally quantify features observed in thin sections or 3D real rock im- able to derive these bounds from first principles.
ages.
Transformations include models such as the Gassmann 共1951兲 re-
lations for fluid substitution, which are relatively free of geometric Bound-filling models
assumptions. The Gassmann relations take the measured VP and VS
The bound-filling models that follow provide simple equations
at one fluid state and predict the VP and VS at another fluid state. Ber-
for families of curves spanning the range between upper and lower
ryman and Milton 共1991兲 present a geometry-independent scheme
bounds on elastic moduli and can be used to describe the stiffness of
to predict fluid substitution in a composite of two porous media hav-
rocks that fall in the range between upper and lower bounds. The
ing separate mineral and dry-frame moduli. Mavko et al. 共1995兲 de-
modified HS and modified Voigt averages are useful depth-trend
rive a geometry-independent transformation to take hydrostatic ve-
lines for sand and chalk sediments. The bounding-average method
locity versus pressure data and predict stress-induced anisotropy.
共BAM兲 provides a heuristic fluid-substitution strategy that seems to
Mavko and Jizba 共1991兲 present a transformation of measured dry
work best at high frequency 共Marion, 1990兲. The isoframe model al-
velocity versus pressure to predict velocity versus frequency in flu-
lows one to estimate the moduli of rocks composed of a consolidated
id-saturated rocks.
grain framework with inclusions of nonload-bearing grain suspen-
sions. These models are based on isotropic linear elasticity. All of the
bound-filling models discussed here contain at least some heuristic
Empirical models
elements, such as the interpretation of the modified upper bounds.
Empirical models do not require much explanation. Generally, the The VR and HS bounds yield precise limits on the maximum and
approach is to assume some function form and then to determine co- minimum possible values for the effective bulk and shear moduli of
efficients by calibrating a regression to data. Some of the best- an isotropic, linear elastic composite. Specifically, for a mixture of
known models are Han’s 共1986兲 regressions for velocity-porosity- two components the HS bounds can be written as
75A34 Avseth et al.

f2 The Voigt-Reuss-Hill average is an estimate of elastic modulus

冉 冊
KHSⳲ ⳱ K1 Ⳮ ⳮ1 , 共1兲 defined to lie exactly halfway between the Voigt upper and Reuss
4
共K2 ⳮ K1兲ⳮ1 Ⳮ f 1 K1 Ⳮ ␮m lower bounds 共Figure 3兲. A similar estimate can be constructed to lie
3 halfway between the upper and lower HS bounds 共Figure 3兲. These
two estimates have little practical value except when the constituent
␮HSⳲ ⳱ ␮1 end members are elastically similar, as with a mixture of minerals
without pore space. In this case, an average of upper and lower
f2 bounds yields a useful estimate of the average mineral moduli.

冉 冉 冊冊
Ⳮ ⳮ1 ,
␮m 9Km Ⳮ 8␮m BAM uses the position of porosity/modulus data between the
共␮2 ⳮ ␮1兲ⳮ1 Ⳮ f 1 ␮1 Ⳮ bounds as an indication of rock stiffness. In Figure 4, the HS upper
6 Km Ⳮ 2␮m and lower bounds are displayed for mixtures of mineral and water.
共2兲 The data point A lies a distance d above the lower bound; D is the
spacing between the bounds at the same porosity. In BAM, it is as-
where subscripts 1 and 2 refer to properties of the two components sumed that the ratio d / D remains constant if the pore fluid in the rock
having bulk moduli K1 and K2, shear moduli ␮1 and ␮2, and volume is changed, without changing the pore geometry or the stiffness of
fractions f 1 and f 2. Most commonly, these bounds are applied to de- the dry frame. Though not theoretically justified, BAM sometimes
scribe mixtures of mineral and pore fluid, as illustrated in Figure 2. gives a reasonable estimate of high-frequency fluid-substitution be-
Equations 1 and 2 yield the upper bound when Km and ␮m are the havior.
maximum bulk and shear moduli of the individual constituents and
the lower bound when Km and ␮m are the minimum bulk and shear
moduli of the constituents. The maximum 共minimum兲 shear modu- 40
lus might come from a different constituent than the maximum 共min-
imum兲 bulk modulus. For example, this would be the case for a mix- 35 Hashin-Shtrikman
upper bound
ture of calcite 共K ⳱ 71; ␮ ⳱ 30 GPa兲 and quartz 共K ⳱ 37; ␮ 30
⳱ 45 GPa兲. Bulk modulus (GPa)
Hashin-Shtrikman average

The modified HS bounds also use equations 1 and 2. However, 25 Voigt


with modified bounds, the constituent end members are selected dif-
20 Voigt-Reuss-Hill
ferently, such as a mineral mixed with a fluid-solid suspension 共Fig-
ure 2兲 or a stiffly packed sediment mixed with a fluid-solid suspen- 15
sion. The critical-porosity model 共Nur et al., 1991, 1995兲 identifies a
critical porosity ␾ c that separates load-bearing sediments at porosi- 10
ties ␾ ⬍ ␾ c and suspensions at porosities ␾ ⬎ ␾ c. Modified HS 共or 5 Hashin-Shtrikman
VR兲 equations can be constructed to describe mixtures of mineral lower bound
and the unconsolidated fluid-solid suspension at critical porosity. 0
0 0.2 0.4 0.6 0.8 1
Different values of ␾ c produce a family of curves between the upper Porosity
and lower bounds. The modified upper HS curve has been observed
empirically to be a useful trend line describing, for example, how the Figure 3. Hashin-Shtrikman and Voigt-Reuss bounds for bulk modu-
elastic moduli of clean sandstones evolve from deposition through lus in a quartz-water system. The Voigt-Reuss-Hill curve is an aver-
compaction and cementation 共Gal et al., 1998兲. The modified upper age of the Voigt upper and Reuss lower bounds. The Hashin-Shtrik-
HS curve, constructed as such, is not a rigorous bound on the elastic man average curve is an average of the Hashin-Shtrikman upper and
lower bounds.
properties of clean sand, although sandstone moduli are almost al-
ways observed to lie on or below it.
40
40
35 Hashin-Shtrikman
35 upper bound
30
Hashin-Shtrikman
Bulk modulus (GPa)

30 upper bound
Bulk modulus (GPa)

25
25
20
20 A
15 D
15 d
10
Modified upper
10 Hashin-Shtrikman
bound 5 Hashin-Shtrikman
5 Hashin-Shtrikman lower bound
lower bound Critical porosity 0
0 0.2 0.4 0.6 0.8 1
0
0 0.2 0.4 0.6 0.8 1 Porosity
Porosity
Figure 4. The bounding average method. The position of a data point
Figure 2. Hashin-Shtrikman and modified Hashin-Shtrikman A, described as d / D relative to bounds, is assumed to be a measure of
bounds for bulk modulus in a quartz-water system. the pore stiffness.
Rock physics analysis of siliciclastics 75A35

Fabricius 共2003兲 has introduced the isoframe model to describe All of the bound-filling models provided here contain at least
behavior between the modified upper and lower HS bounds. In this some heuristic elements, such as the interpretation of the modified
model, the modified upper HS curve is assumed to describe the trend upper bounds. The different families of curves are parameterized by
of sediments that become progressively more compacted and ce- different quantities with different physical interpretation, e.g., criti-
mented as they trend away from the lower Reuss average 共Figure 5兲 cal porosity ␾ c, the fraction of load-bearing frame IF, or the pore-
— an empirical result. It is assumed that these rocks contain only space stiffness K␾ . Though somewhat heuristic, these models pro-
grains that are load bearing. Sediments that fall below the modified vide a simple and practical way to model a wide range of velocity-
upper bound are assumed to contain inclusions of grain-fluid sus- porosity trends.
pensions, in which the grains are not load bearing. A family of curves
can be generated 共Figure 5兲, computed as an upper HS mix of frame, Our hybrid approach
taken from the modified upper bound, and suspension, taken from
the lower Reuss bound. IF is the volume fraction of load-bearing We have found a hybrid combination of theoretical, empirical,
frame, and 1 — IF is the fraction of suspension. The isoframe modu- and heuristic models very useful to describe the rock-physics prop-
lus at each porosity is computed from the frame and suspension erties of high-porosity clastic sediments. In this sense, we find our-
moduli at the same porosity. The calculation is done separately for selves thinking more as engineers than physicists — the models that
bulk and shear moduli, from which the P-wave modulus can be cal- work best in practice, for prediction and interpretation purposes,
culated. may not be the models founded on the most advanced physical theo-
The bulk modulus K of an elastic porous medium can be ex- ries.
pressed as It started with Han’s 共1986兲 empirical discovery that the relation-
ship between velocity and porosity in shaly sands could be well de-
1 1 ␾ scribed by a set of parallel contours of constant clay. Amos Nur
⳱ Ⳮ , 共3兲 共1992兲 notes that each of these contours have high- and low-porosity
K Kmin K̃
␾ intercepts with a clear physical interpretation: in the limit of zero po-
where Kmin is the mineral bulk modulus and K̃␾ is the saturated pore rosity, any model should rigorously take on the properties of pure
space stiffness, given by mineral; and in the limit of high porosity 共the critical porosity兲, we
have a suspension that is rigorously modeled with a lower-bound
KminKfluid equation. Eventually, Han’s contours have been replaced by modi-
K̃␾ ⳱ K␾ Ⳮ 共4兲 fied upper bounds, partly because they fit the data better over a large
Kmin Ⳮ Kfluid
range of porosities and partly because we could defend them heuris-
共Mavko et al., 2009兲, where Kfluid is the bulk modulus of the pore flu- tically. We have come to understand that these modified upper
id and K␾ is the dry rock pore-space stiffness defined in terms of the bounds describe the diagenetic or cementation trend for sedimentary
pore volume v and confining stress ␴ c: rocks.
A modified lower bound is an excellent description of the veloci-
1 1 ⳵v
⳱ . 共5兲 ty-porosity sorting trend. Again, this is more of an empirical obser-
K␾ v ⳵ ␴ c vation, aesthetically symmetric to the modified upper bound but not
rigorously defendable. Dvorkin and Nur 共1996兲 introduces the fria-
Figure 6 shows a plot of bulk modulus versus porosity with con-
ble-sand model, which uses a theoretical elastic contact model to de-
tours of constant K␾ . A large K␾ indicates a stiff pore space, and a
scribe clean, well-sorted sands combined with a modified lower
small K␾ indicates a soft pore space. The value K␾ ⳱ 0 corresponds
bound to interpolate these to lower-porosity, poorly sorted sands.
to a suspension.
In summary, we have found the rock-physics modeling approach
presented in this paper to be useful in high-porosity sand-shale envi-
100
ronments. We avoid overmodeling with too much theory that de-
Modified upper
Hashin-Shtrikman 1.0
P-wave modulus (GPa)

80
0.9 Hashin-Shtrikman
upper bound
0.8
60 IF = 1
0.7
0.8
0.6 0.4
40
K/K min

0.6
0.4 Critical 0.5
0.2 porosity 0.3
20 0.4

0.3 0.2
Reuss
0 0.2
0 0.1 0.2 0.3 0.4 0.5 0.6 K phi/K min = 0.1
Porosity 0.1 Reuss
0.0
Figure 5. The isoframe model. The modified upper Hashin-Shtrik- 0 0.2 0.4 0.6 0.8 1
man curve is assumed to describe a strong frame of grains in good Porosity (volume fraction)
contact. The Reuss average curve describes a suspension of grains in
fluid. Each isoframe curve is a Hashin-Shtrikman mix of a fraction Figure 6. Normalized bulk modulus versus pressure, showing con-
IF of frame with 共1 ⳮ IF兲 of suspension. tours of constant K␾ .
75A36 Avseth et al.

pends on model parameters which follow from mathematical conge- The rock-physics diagnostics technique was introduced by Dvor-
niality rather than from geologic processes. We have also found it kin and Nur 共1996兲 as a means to infer rock microstructure from ve-
disadvantageous to become attached to meticulously derived theo- locity-porosity relations. This diagnostic is conducted by adjusting
retical models that can never approach the complexity of nature. At an effective-medium theoretical model curve to a trend in the data,
the same time, we like to honor physical principles because they assuming that the microstructure of the sediment matches that used
make the models universal. As time goes on, it almost seems that we in the model. We review three heuristic hybrid models that have been
throw away more equations and replace them with clever uses of var- used to describe the velocity-porosity-pressure behavior of medium
ious bounds. Another important driver in our approach is the desire — to high-porosity sediments and rocks: 共1兲 the friable-sand model,
to discover, understand, and quantify elastic properties as a function 共2兲 the contact-cement model, and 共3兲 the constant-cement model.
of geologic processes. Although not proving this heuristic approach A very effective approach is to begin by defining the elastic prop-
is correct, the supporting argument for the approach is that the effec- erties of the end members. At zero porosity, the rock must have the
tive medium path chosen from geologic concepts does not violate properties of mineral. At the high-porosity limit, the elastic proper-
the physics of the mixture. An important principle states that if a ties are determined by elastic-contact theory. Then, we interpolate
model falls within elastic bounds, it is realizable 共Norris, 1985兲. between these two end members using upper or lower HS bounds.
The upper bound explains the theoretical stiffest way to mix load-
BRIEF “LIFE STORY” OF bearing grains and pore-filling material, and the lower bound ex-
plains the theoretical softest way to mix these. Hence, we have found
A CLASTIC SEDIMENT
that the upper bound is a good representation of contact cement,
Elastic bounds provide a framework for understanding the acous- whereas the lower bound accurately describes the effect of sorting.
tic properties of sediments. Figure 1 shows P-wave velocity versus Rocks with very little contact cement 共a few percent兲 are not well de-
porosity for a variety of water-saturated sediments, ranging from scribed by the HS upper bound because there is a large stiffening ef-
ocean-bottom suspensions to consolidated sandstones. The Voigt fect during the very initial porosity reduction as cement fills in the
and Reuss bounds, computed for mixtures of quartz and water, are microcracks between the contacts. Then it is dangerous to interpo-
shown for comparison. 共Strictly speaking, the bounds describe the late between the high-porosity and zero-porosity end members. We
allowable range for elastic moduli. When the corresponding P- and therefore include a high-porosity contact-cement model that takes
S-wave velocities are derived from these moduli, it is common to re- into account the initial cementation effect.
fer to them as the upper and lower bounds on velocity.兲
Before deposition, sediments exist as particles suspended in water
The friable-sand model
共or air兲. As such, their acoustic properties must fall on the Reuss av-
erage of mineral mixed with fluids. When the sediments are first de- Dvorkin and Nur 共1996兲 introduce two theoretical models for
posited on the water bottom, we expect their properties to still lie on high-porosity sands. The friable-sand model, or the unconsolidated
共or near兲 the Reuss average as long as they are weak and unconsoli- line, describes how the velocity-porosity relation changes as the
dated. Their porosity position along the Reuss average is determined sorting deteriorates. The well-sorted end member is represented as a
by the geometry of the particle packing. Clean, well-sorted sands well-sorted packing of similar grains whose elastic properties are de-
will be deposited with porosities near 40%. Poorly sorted sands will termined by the elasticity at the grain contacts. The well-sorted end
be deposited along the Reuss average at lower porosities. Chalks member typically has a critical porosity around 40%. The friable-
will be deposited at high initial porosities, 55%–65%. We some- sand model represents poorly sorted sands as the well-sorted end
times call this porosity of the newly deposited sediment the critical member modified with additional smaller grains deposited in the
porosity 共Nur, 1992兲. Upon burial, the various processes that give the pore space. These additional grains deteriorate sorting, decrease po-
sediment stiffness and strength — effective stress, compaction, and rosity, and only slightly increase rock stiffness 共Figure 7兲.
cementing — must move the sediments off of the Reuss bound. With The elastic moduli of the dry, well-sorted end member at critical
increasing diagenesis, the rock properties fall along steep trajecto- porosity are modeled as an elastic sphere pack subject to confining
ries that extend upward from the Reuss bound at critical porosity and
toward the mineral end point at zero porosity. We see below that
these diagenetic trends can be described once again using the
bounds.
Elastic modulus

ROCK-PHYSICS MODELING OF SEDIMENTARY


MICROSTRUCTURE IN HIGH-POROSITY
SANDSTONES
If we want to predict the seismic velocities of a rock, knowing
only the porosity, mineralogic composition, and the elastic moduli of
the mineral constituents, we can at best predict the upper and lower
0.30 0.35 0.40
bounds of seismic velocities. However, if we know the geometric de-
Porosity
tails of how the mineral grains and pores are arranged relative to
each other, we can predict more exact seismic properties. Several
Figure 7. Schematic of the friable-sand model and corresponding
models account for the microstructure and texture of rocks, and sedimentologic variation. Elastic modulus increases slightly with
these in principle allow us to go the other way: to predict the type of deteriorating sorting and associated increasing amount of pore-fill-
rock and microstructure from seismic velocities. ing material.
Rock physics analysis of siliciclastics 75A37

冤 冥
ⳮ1
pressure. These moduli are given by the HM theory as ␾ ␾
1ⳮ

KHM ⳱ 冋 n2共1 ⳮ ␾ c兲2␮2


18␲ 2共1 ⳮ ␯ 兲2
P 册 1/3
, 共6兲
␮dry ⳱
␾c
␮HM Ⳮ z

␾c
␮Ⳮz
ⳮ z, 共12兲

where

␮HM ⳱ 冋
5 ⳮ 4␯ 3n 共1 ⳮ ␾ c兲 ␮
2

5共2 ⳮ ␯ 兲 2␲ 2共1 ⳮ ␯ 兲2
P
2 2
册 1/3
, 共7兲 z⳱ 冉
␮HM 9KHM Ⳮ 8␮HM
6 KHM Ⳮ 2␮HM
. 冊 共13兲

where KHM and GHM are the dry-rock bulk and shear moduli, respec- The saturated elastic moduli Ksat and ␮sat can be calculated from Gas-
tively, at ␾ c 共i.e., depositional porosity兲; P is the effective pressure smann’s equations. Density is given by
共i.e., the difference between the confining pressure and the pore pres-
sure兲; ␮ and ␯ are the shear modulus and Poisson’s ratio of the solid ␳ b ⳱ ␾ ␳ fluid Ⳮ 共1 ⳮ ␾ 兲␳ min, 共14兲
phase; and n is the coordination number 共the average number of con- where ␳ min is the mineral density, which equals 2.65 g / cm3 for
tacts per grain兲. quartz, and ␳ fluid is the fluid density, normally varying from
Poisson’s ratio can be expressed in terms of the bulk K and shear ␮ 1.0 to 1.15 g / cm3 for saline water. For dry rocks, the fluid density is
moduli as follows: zero.
Some of the largest uncertainties with the friable-sand model are
3K ⳮ 2␮ associated with heterogeneous grain contacts, tangential slip, and
␯⳱ . 共8兲
2共3K Ⳮ ␮兲 highly variable coordination number. Several workers demonstrate
how the high-porosity end member can be calibrated to a given data
Assuming hydrostatic pressure, effective pressure versus depth is set by adjusting a slip factor 共e.g., Bachrach and Avseth, 2008; Faust
obtained with the following formula: Andersen and Johansen, 2010兲 or by changing n 共e.g., Florez, 2005;
Dutta et al., 2010兲. Florez 共2005兲 finds that because of the uncertain-


Z
ty in n and the limitations of using an idealized packing model, the
P⳱g 共␳ b ⳮ ␳ fl兲dz, 共9兲 modified HS lower bound 共i.e., the friable-sand model兲 may over-
predict the velocity increase because of deteriorating sorting. Grain
0 packing, which texturally can look very similar to the effect of sort-
ing but is caused by postdepositional mechanical compaction, often
where g is the gravity constant and where ␳ b and ␳ fluid are the bulk
follows a depositional sorting trend in a velocity-porosity crossplot.
density and fluid density, respectively, at a given depth z.
Florez 共2005兲 finds, however, that packing often yields a slightly
The coordination number n depends on porosity, as shown by
steeper slope than the one predicted from the friable-sand model. He
Murphy 共1982兲. The relationship between coordination number and
argues that the combination of packing and sorting could explain the
porosity can be approximated by the following empirical equation:
good match between the friable-sand model and data of sands with
increasing pore-filling material often observed in real in situ well-
n ⳱ 20ⳮ 34␾ Ⳮ 14␾ 2 . 共10兲 log measurements.
Hence, for a porosity of 0.4, n ⳱ 8.6.
The other end point in the friable-sand model is at zero porosity The contact-cement model
and has the bulk K and shear ␮ moduli of the mineral. Moduli of the
poorly sorted sands with porosities between zero and ␾ c are interpo- During burial, sands are likely to become cemented sandstones.
lated between the mineral point and the well-sorted end member us- This cement may be diagenetic quartz, calcite, albite, or other miner-
ing the lower HS bound. One heuristic argument for this is that add- als. Cementation has a more rigid stiffening effect because grain
ing small grains passively in the pore space is the softest way to add contacts are “glued” together. The contact-cement model assumes
mineral to the well-sorted sands; the lower-bound equation is always that porosity reduces from the initial porosity of a sand pack as a re-
the softest way to mix two phases. Another argument follows from sult of the uniform deposition of cement layers on the surface of the
Figure 7. Here, we envision the poorly sorted sand as a few large grains 共Figure 8兲. The contact cement dramatically increases the
grains enveloped by soft “shells” of fine-grained sand. This is the re- stiffness of the sand by reinforcing the grain contacts. In particular,
alization of a lower HS bound. the initial cementation effect will cause a large velocity increase with
At porosity ␾ , the concentration of the pure solid phase 共added to only a small decrease of porosity.
the sphere pack to decrease porosity兲 in the rock is 1 ⳮ ␾ / ␾ c, and The mathematical model is based on a rigorous contact-problem
that of the original sphere-pack phase is ␾ / ␾ c. Then the bulk Kdry solution by Dvorkin et al. 共1994兲. In this model, Kdry and Gdry of dry
and shear Gdry moduli of the dry friable-sand mixture are rock are

n共1 ⳮ ␾ c兲M cSn

冤 冥
ⳮ1
␾ ␾ Kdry ⳱ , 共15兲
1ⳮ 6
␾c ␾c 4
Kdry ⳱ Ⳮ ⳮ ␮HM, and
4 4 3
KHM Ⳮ ␮HM K Ⳮ ␮HM
3 3 3Kdry 3n共1 ⳮ ␾ c兲␮cS␶
␮dry ⳱ Ⳮ , 共16兲
共11兲 5 20
75A38 Avseth et al.

冤 冥
where ␾ c is critical porosity; Ks and ␮s are the bulk and shear moduli 2 0.5
of the grain material, respectively; Kc and ␮c are the bulk and shear 共␾ c ⳮ ␾ 兲
3
moduli of the cement material, respectively; M c ⳱ Kc Ⳮ 4␮c / 3 is the ␣⳱ , 共27兲
compressional modulus of the cement; and n is the coordination 1 ⳮ ␾c

冉 冊
number, defined as average number of contacts per grain. Parame-
ters Sn and S␶ are porportional to the normal and tangential stiff- Kc 2
nesses, respectively. Statistical approximations of the rigorous ce- ⳮ
␮c 3

冉 冊
mentation theory solutions 共Dvorkin et al., 1994兲 are given by the ␯ c ⳱ 0.5 , 共28兲
following equations 共Dvorkin and Nur, 1996兲: Kc 1

␮c 3
Sn ⳱ An共⌳n兲␣ 2 Ⳮ Bn共⌳n兲␣ Ⳮ Cn共⌳n兲, 共17兲

An共⌳n兲 ⳱ ⳮ 0.024153⌳nⳮ1.3646, 共18兲 冉


Ks 2

␮s 3

冉 冊
␯ s ⳱ 0.5 . 共29兲
Ks 1
Bn共⌳n兲 ⳱ 0.20405⌳n ⳮ0.89008
, 共19兲 Ⳮ
␮s 3

Cn共⌳n兲 ⳱ 0.00024649⌳nⳮ1.9864 . 共20兲 A detailed explanation of equations 17–29 and their derivation is
given in Dvorkin et al. 共1994兲. Saturated elastic moduli are calculat-
ed using Gassmann’s equations. Dry and saturated bulk densities are
S␶ ⳱ A␶ 共⌳␶ ,␯ s兲␣ 2 Ⳮ B␶ 共⌳␶ ,␯ s兲␣ Ⳮ C␶ 共⌳␶ ,␯ s兲, calculated using equation 14.
共21兲 The main shortcoming with the Dvorkin-Nur contact-cement
model is that is does not include pressure sensitivity. It is assumed
that the cemented grain contacts immediately lose pressure sensitiv-
A␶ 共⌳␶ ,␯ s兲 ⳱ ⳮ 10ⳮ2共2.26␯ s2 Ⳮ 2.07␯ s ity as the cementation process begins. From in situ observations, we
2 know that cemented reservoirs can have significant pressure sensi-
Ⳮ 2.3兲⌳␶ 0.079␯ s Ⳮ0.1754␯ sⳮ1.342
, 共22兲 tivity 共e.g., Duffaut and Landrø, 2007; Avseth et al., 2009a; Vernik
and Hamman, 2009兲. This could be related to fractures not captured
B␶ 共⌳␶ ,␯ s兲 ⳱ 共0.0573␯ s2 Ⳮ 0.0937␯ s by the microstructural-scale model or by a patchy cementation
where some grain contacts are cemented and others are loose.
2
Ⳮ 0.202兲⌳␶ 0.0274␯ s Ⳮ0.0529␯ sⳮ0.8765
, 共23兲 Hence, the loose contacts should still be pressure sensitive. As with
the HM contact theory for loose granular media, the Dvorkin-Nur
contact-cement model often overpredicts shear stiffnesses in ce-
C␶ 共⌳␶ ,␯ s兲 ⳱ 10ⳮ4共9.654␯ s2 Ⳮ 4.945␯ s mented sandstones. This could be related to a heterogeneous mixture
2 of grain contacts and tangential slip at loose contacts 共Avseth et al.,
Ⳮ 3.1兲⌳␶ 0.01867␯ s Ⳮ0.4011␯ sⳮ1.8186
, 共24兲 2009a兲 or to relative roll and torsion not taken into account in the
contact theory 共e.g., Elata and Berrymann, 1996兲. Furthermore, a
2␮c共1 ⳮ ␯ s兲共1 ⳮ ␯ c兲 steep slope in velocity-porosity crossplots may indicate elastic de-
⌳n ⳱ , 共25兲 formation, pressure solution at grain contacts, or grain interpenetra-
␲ ␮s共1 ⳮ 2␯ c兲
tion without associated cementation 共Florez and Mavko, 2004兲.

␮c The constant-cement model


⌳␶ ⳱ , 共26兲
␲ ␮s The constant-cement model, introduced by Avseth et al. 共2000兲,
assumes that sands of varied sorting 共and therefore varied porosity兲
all have the same amount of contact cement. Porosity reduction is
solely from noncontact pore-filling material 共e.g., deteriorating sort-
ing兲. Mathematically, this model is a combination of the contact-ce-
Elastic modulus

ment model, where porosity reduces from the initial sand-pack po-
rosity to porosity ␾ b as a result of contact-cement deposition and
from the friable-sand model where porosity reduces from ␾ b as a re-
sult of the deposition of the solid phase away from the grain contacts
共Figure 9兲. Considering a given reservoir, this is the most likely sce-
nario because the amount of cement is often related to depth, where-
as sorting is related to lateral variations in flow energy during sedi-
ment deposition. In these cases, we can refer to this as a constant-
0.30 0.35 0.40
depth model for clean sands. However, it is possible that cement can
Porosity (volume fraction)
have a local source and therefore cause a considerable lateral varia-
Figure 8. Schematic of the contact-cement model and the corre- tion in velocity.
sponding diagenetic transformation. Elastic modulus increases To use the constant-cement model, we must first adjust the well-
markedly with increasing amount of contact cement. sorted end-member porosity ␾ b that corresponds to the point shown
Rock physics analysis of siliciclastics 75A39

as an open circle in Figure 9. The dry-rock bulk and shear moduli at Sea. In this case, the rock-physics diagnostics are performed in the
this porosity 共Kb and ␮b, respectively兲 are calculated from the con- VS-versus-porosity domain to avoid significant pore-fluid effects.
tact-cement model. Equations for the dry-rock bulk Kdry and shear The cement volume is estimated by interpolating between the con-
␮dry moduli at a smaller porosity ␾ are then interpolated with a lower stant-cement volume trends, whereas the sorting is defined by the
bound: observed porosity normalized by the high-end-member porosity

冤 冥
ⳮ1 along the given constant-cement trend — the porosity connecting
␾ ␾ the constant-cement model with the contact-cement model.
1ⳮ
␾b ␾b 4 Having estimated cement volume and sorting, we can plot the data
Kdry ⳱ Ⳮ ⳮ ␮b 共30兲 as logs and compare them with other petrophysical logs. Figure 11
4 4 3
Kb Ⳮ ␮b K Ⳮ ␮b shows the resulting estimate of cement volume and sorting. Figure
3 3 11b shows cement volume as magnitude and sorting as superim-
and posed color. For the studied North Sea sandstone interval starting at
around 2 km depth, we observe a clear depth trend in the cement vol-

冤 冥
ⳮ1
␾ ␾ ume. The sorting, however, shows a more erratic pattern without a
1ⳮ
␾b ␾b depth trend, which is expected because sorting is associated with
␮dry ⳱ Ⳮ ⳮ z, 共31兲 depositional trends.
␮b Ⳮ z ␮Ⳮz It is essential to verify the presence of initial cementation predict-
where ed from the rock-physics relations with thin-section observations.

冉 冊
Figure 12 shows a thin section from the relatively clean sands in this
␮b 9Kb Ⳮ 8␮b study; at first glance, the sandstone looks unconsolidated, with
z⳱ . 共32兲
6 Kb Ⳮ 2␮b
a)
The effect of pore fluid can be accounted for by using Gassmann’s 4500 10
共1951兲 equations. Dry and saturated bulk densities are calculated us- 4000
Quartz 9
ing equation 14. Regarding input mineral properties, tables of stan- 8
dard properties of minerals are available in many handbooks and 3500
Constant-cement
7

Cement volume (%)


compilations, such as Ellis et al. 共1988兲, Carmichael 共1989兲, and 3000
trends
6
Mavko et al. 共2009兲. Note that it is possible to arrive at the constant- Increasing cement volume
VS (m/s)

2500 5
cement line by first moving along the friable-sand line and then add-
ing contact cement to the rock 共dashed line in Figure 9兲, which is 2000
Dvorkin-Nur
4
contact cement
consistent with diagenesis following deposition. The pitfalls and 3
1500
limitations mentioned for the friable-sand and the contact-cement
2
models also apply for the constant-cement model. 1000
1
500 Shale
0
ROCK-PHYSICS ESTIMATION OF CEMENT
0 –1
VOLUME AND SORTING — A NORTH 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45
SEA DEMONSTRATION Porosity (volume fraction)

Using the diagnostic models, we can infer the microstructure from b)


4500 1.0
velocity-porosity data. With good local validation of the models, we
can even quantify the degree of sorting and cement volume from 4000 Quartz 0.9

these diagnostic crossplots 共Avseth et al., 2009b兲. Figure 10 shows 3500


0.8
how this procedure is done for some well-log data from the North Sorting = φ /φ c 0.7
3000
0.6
VS (m/s)

Sorting
Constant Contact 2500
cement cement 0.5
2000
0.4
Elastic modulus

1500
0.3
1000 0.2
500 Shale 0.1
Initial 0 0.0
Friable sand 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45
pack
Porosity (volume fraction)
0.30 0.35 0.40
Porosity (volume fraction) Figure 10. Shear-wave velocity-log data versus total porosity and
superimposed diagnostic rock-physics models. Using the models,
Figure 9. Schematic of three effective-medium models for high-po- we can quantify 共a兲 the cement volume and 共b兲 the degree of sorting.
rosity sands in the elastic-modulus/porosity plane and correspond- Green data points are shale data with high GR values and for practi-
ing microstructure characteristics. The elastic modulus may be com- cal reasons are given the value ⳮ1 in cement volume and zero in
pressional, bulk, or shear. sorting.
75A40 Avseth et al.

a) b) c) grains loosely arranged and moderately well sort-


1200 1200 1200 10 ed. A closer investigation, however, reveals the
1.1
9 presence of initial quartz overgrowth covering
1400 1400 1.0 1400 original grain surfaces indicated by dust rims 共ar-
8
0.9 rows, Figure 12兲. This observation confirms what
1600 1600 1600 7 we see in the rock-physics crossplots of the well-
0.8
6 log data. It is interesting that the well-log data
1800 1800 1800
0.7 with tens-of-centimeter resolution reflects what
Depth (m)

5
2000 2000 0.6 2000 we observe at the microscale. Furthermore,
0.5
4 Avseth et al. 共2009b兲 find a very good match be-
2200 2200 2200 3 tween the rock-physics estimated cement volume
0.4
and the total cement estimated from thin sections.
2400 2400 2
0.3 2400
0.2 1
2600 2600 2600 ROCK PHYSICS OF SHALES
0.1 0
2800 2800 2800 –1 Until recently, shales have often been regarded
0 0.5 1 0 5 10 0 0.5 1
(%) as a unique type of lithology among geophysi-
Shale volume Cement volume (%) Sorting
cists, and minor attention has been given to the
great variance in mineralogy, texture, and porosi-
Figure 11. Estimated cement volume and sorting versus depth for a North Sea well. Note
the onset of cement starting at around 2000 m depth 共corresponding to 70° C兲 and 共b兲 the ty of shales during seismic data analysis. This is
increasing cement volume with depth. In spite of this depth trend, there is variability in partly because the rock properties of clay miner-
cement volume that is likely associated with varying shaliness in the sandstones. Sorting, als are difficult to measure in the laboratory but
however, shows a more erratic pattern with no depth trend 关color in 共b兲兴. 共a兲 Shale volume also because the oil industry has given little prior-
versus depth; 共b兲 cement volume versus depth 共dots兲 and sorting versus depth 共color兲; 共c兲 ity to acquiring detailed log data and core samples
sorting versus depth 共dots兲 and cement volume versus depth 共color兲.
in shale sequences. Geologists, however, have
documented the complexity of shales, and there
exists a vast amount of published literature on their geochemistry
a)
and sedimentology 共e.g., Bjørlykke, 1998; MacQuaker et al., 2007;
Peltonen et al., 2008兲. With increased focus on cross-disciplinary in-
tegration, geophysicists are starting to incorporate this geologic
knowledge into modeling and analyzing geophysical data 共e.g.,
Dræge et al., 2006b; Brevik et al., 2007; Mondol et al., 2007; Mar-
cussen et al., 2008兲.
As with sands, we can distinguish between depositional and di-
agenetic trends in shales. Depositional trends will affect clay miner-
alogy; but in particular, the silt content will have great impact on the
seismic properties. Avseth et al. 共2005兲 use a simple isotropic lower
Reuss bound to model vertical velocities of silty shales. However,
shales have a different composition and texture than sandstones.
Therefore, the rock-physics models applicable for sandstones do not
necessarily apply for shales. More rigorous anisotropic modeling of
b) shales has been performed by Hornby et al. 共1994兲, Johansen et al.
共2002兲, Ruud et al. 共2003兲, and Dræge et al. 共2006b兲, among others.
In this paper, we use the shale compaction model 共Dræge et al.,
2006b; Ruud et al., 2003兲 to estimate the anisotropic effective prop-
erties in mechanically compacted shales. More details about this
model are included in the next section and in Avseth et al. 共2008兲.

Quartz cement

ROCK-PHYSICS DEPTH TRENDS


Rock-physics depth trends for sands and shales can be used to
study the expected seismic signatures of sand-shale interfaces as a
function of depth. We use existing empirical porosity-depth trends
Figure 12. Thin sections from Heimdal Formation sands. 共a兲Aloose- for sands and shales as input to rock-physics models of VP, VS, and
ly packed, poorly consolidated sand. 共b兲Analysis of a zoomed-in im- density. We can, for example, use HM theory to calculate the veloci-
age 共partially from the red boxed area in 关a兴兲 confirms the presence of ty-depth trends for unconsolidated sands and shales, whereas Dvor-
quartz overgrowths and contact cement. On detrital quartz grains,
we observe dust rims representing the original grain surfaces that kin and Nur’s 共1996兲 contact-cement model can be used for cement-
have been covered by quartz cement 共arrows兲. Feldspar overgrowth ed sands. The depth trends allow us to study the ability to discrimi-
and calcite cement also occur, yet quartz cement is dominating. nate between pore fluids and lithologies at different depths.
Rock physics analysis of siliciclastics 75A41

Figure 13 is a schematic representation of shale and sand compac- tion 共SCA兲 are used to account for the elongated pores and grains in
tion curves and a sequence of interbedded turbidite sands and marine shales 共Hornby et al., 1994兲. In this study, pores in chemically com-
shales, typical for the North Sea deep-marine environment of Tertia- pacted shales are considered to be isolated, but the pores in the me-
ry age. The depositional porosity in shales is normally much higher chanical compaction regime are connected 共c.f., Dræge et al.,
共60%–80%兲 than in sands 共⬃40%兲, but we expect a shallow cross- 2006b兲. We define a transition zone, where the properties change
over with depth as a result of the mechanical collapse of the shales. from the mechanical to the chemical regime.
The platy clay fabric in the shales is more prone to compaction than In Figure 15, the initial shale 共 ⬍ 1500 m兲 is considered to be
the assemblage of spherically shaped grains in sands and hence the smectite rich, and the deeper 共⬎2200 m兲 illite-rich shale is some-
more rapid mechanical porosity reduction in shales than sands. Dur- what stiffer. The vertical mineral bulk and shear moduli of smectite
ing burial to approximately 2 km depth, sands and shales are ex- are found by calibration to well-log data to be 12.5 and 7.5 GPa, re-
posed mainly to mechanical compaction. The marine shales in the spectively. For illite, these elastic moduli are 21 and 7 GPa, respec-
Tertiary North Sea are very smectite rich, and they have very low tively. The assumption of isotropic mineral moduli for the clay min-
permeability. In thick, smectite-rich shale masses, it is therefore nor- erals is probably incorrect, but the assumption may be realistic for
mal to observe undercompaction and associated overpressure even assemblages of clay crystals that are randomly scattered and mixed
at several hundred meters’ burial depth. At about 70° C, however, with larger silt particles of quartz. In the modeling, we assume rela-
chemical alteration of smectite will begin, and we expect a mineral tively pure shales with 7% quartz. For quartz, the mineral moduli are
transformation to illite. This is a typical mineral transformation seen 37 and 44 GPa, respectively. We assume clay density to be the same
in marine shales all over the world 共Bjørlykke, 1998兲. Bound water as for quartz, 2.65 g / cm3.
in the smectite layers is released when the temperature reaches this There are two counteracting effects on anisotropy. The initial
critical temperature, resulting in a porosity decrease. Moreover, the
presence of potassium cations 共e.g., in feldspar or mica兲 causes Deposition
Porosity
quartz to be produced as a by-product. This quartz can precipitate as
microcrystalline quartz within the shale matrix 共Thyberg et al., Pure

compaction
Mechanical
North Sea Paleocene
shale Clean
2009兲; if connectivity allows, the quartz may precipitate as cement sand
in adjacent sandstones 共Peltonen et al., 2008兲.
Volcanic tuff
Figure 14 shows well-log data from a North Sea well penetrating Smectite-rich shale
siliciclastic sediments and rocks of Tertiary age 共same data investi- ~ 70°C
Unc. sand
~ 2 km

gated in Figures 10 and 11兲, juxtaposed with rock-physics depth


compaction

Cem. SS Illite-rich shale


Chemical

Qz
Qz
trends for shales and sandstones. We observe a good match between
the calculated velocity-depth trends for different lithologies and the Smectite + K+ → Illite + SiO2 + H 2O
well-log data. The sandstone rock-physics models as a function of
depth are modeled by combining HM contact theory for unconsoli- Depth of burial
dated sands with the Dvorkin-Nur contact cement model for cement-
ed sandstones. The input porosity-depth trends are calibrated with Figure 13. Schematic of sand and shale compaction. At 70° C, it is
common to observe a change from mechanical compaction to pre-
local compaction trends according to empirical relations 共e.g., dominantly chemical compaction in siliciclastic systems. In deep
Ramm and Bjørlykke, 1994; Mondol et al., 2007兲. The light-blue marine depositional systems, smectite-rich shales experience illiti-
model curves show how the velocities increase drastically for sands zation and release of bound water, causing a porosity reduction and
as we go from the unconsolidated regime with only mechanical com- mineralogy change with depth. For quartz-rich sands, initial cemen-
paction to the cemented regime with predominantly chemical com- tation tends to start at the same depth level. One possible external
source of cement is in fact derived from the smectite-to-illite transi-
paction. The onset of quartz cement happens at tion in embedding shales.
about 70° C, corresponding to about 2 km burial
0
depth.
Seafloor
We apply the shale compaction model 共Ruud et
500
al., 2003; Dræge et al., 2006b兲 to estimate the an-
isotropic effective properties in mechanically Mechanical
1000 compaction
compacted shales. The first seismically important
Depth (m)

mineral reaction in shales is commonly the smec-


1500
tite-to-illite reaction. The reaction has several im-
plications for the shale; the soft smectite is re-
2000 Gas
placed by stiffer illite that might be distributed Chemical
Oil compaction
differently in the rock. The reaction produces wa-
ter, the amount of solids is decreased 共i.e., illite 2500
has a denser mineral structure than smectite兲,
quartz is generated as a by-product, and porosity 3000
0 0.5 1 2000 3000 4000 1000 2000 3000 2 2.5
is reduced by chemical compaction. When the Vsh Vp (m/s) Vs (m/s) Density (kg/m3)
shales are moving into the chemical compaction
regime, a new set of rock-physics models is ap- Figure 14. Rock-physics depth trends for shales 共blue兲 and sandstones 共cyan兲, juxtaposed
plied to estimate the seismic properties. The an- on North Sea well-log data penetrating a Tertiary sequence of siliciclastic sediments and
isotropic version of a differential effective medi- rocks. A gas zone is indicated in yellow, and an oil zone is in red. The remaining interval
of the Heimdal Formation is brine filled. The Heimdal is embedded in the Lista Forma-
um 共DEM兲 model and self-consistent approxima- tion shale. The leftmost panel shows the shale volume V . sh
75A42 Avseth et al.

alignment of grains leads to more aligned pores and increasing an- VP, VS, and density, we estimate the Thomsen 共1986兲 parameters of
isotropy. However, decreasing porosity leads to decreasing anisotro- anisotropy, ␦ and ␧, which can be significant for interpreting angle-
py, culminating in the anisotropic properties of the solid material dependent seismic reflectivity 共AVO analysis兲. The estimated Th-
共i.e., mixture of quartz and clay兲 at zero porosity. The pores intro- omsen parameters are within the range of experimental values of
duce higher anisotropy than the solid because pores commonly are mixed clay 共kaolinite and illite兲 and quartz derived by Voltolini et al.
weaker orthogonal to the longest axis, whereas the solids in this case 共2009兲 under a uniaxial effective stress 共no lateral strain allowed兲 of
are less dependent on direction of wave propagation. In addition to 5 and 50 MPa. They show that pure clay can have
P-wave anisotropy exceeding 40%, but the pres-
Delta Epsilon ence of silty quartz particles within a shale will
1200
drastically reduce the anisotropy.
1400 Wang 共2002兲 finds that ␧ decreases exponen-
Smectite
rich 1600 tially with increasing porosity, whereas ␦ has a
weak relationship with porosity. The decreasing ␧
1800 ~ 70°C
with burial depth and decreasing porosity that we
Depth (m)

2000 estimate from our modeling does not fit with


Illite Wang’s observations. This is probably a result of
rich 2200 our assumption of isotropic mineral moduli for
2400 the clays 共i.e., individual clay crystals have strong
anisotropy, yet assemblages of clay crystals can
2600
effectively show lower anisotropy, especially
2800 when mixed with quartz兲 together with the micro-
2000 2500 3000 500 1000 1500 2.2 2.4 2.6 0 0.1 0.2
VP (m/s) VS (m/s) Density (kg/m3) Anisotropy structural representation of chemical compaction
Smectite + K → Illite + Si + H 2 O that we include in modeling the illite-rich shales.
Chemical compaction from precipitation of mi-
crocrystalline quartz in illite-rich shales is report-
Figure 15. Modeled rock-physics depth trends of shales, showing the effect of illitization
of marine smectite-rich shales. ed by Thyberg et al. 共2009兲, and this diagenetic
process may cause stronger vertical bindings be-
tween clay particles and hence reduced P-wave
a) anisotropy with decreasing porosity. It is beyond the scope of this pa-
per to verify our anisotropic modeling results with real anisotropy
Shale
measurements in these type of shales, but future research should in-
deed investigate how different geologic processes in shales affect
VPV S

Decreasing Brine sand seismic anisotropy.


Sw
Quartz mineral
Oil sand ROCK-PHYSICS TEMPLATES (RPTS)
Acoustic impedance We can combine the depositional and diagenetic trend models
with Gassmann fluid substitution and make charts or templates of
b) rock-physics models for predicting lithology and hydrocarbons. We
refer to these locally constrained charts as rock-physics templates
共RPTs兲, a technology first presented by Ødegaard andAvseth 共2003兲.
Shale Furthermore, we expand on the rock-physics diagnostics as we cre-
VPV S

Initial ate RPTs of seismic parameters — in our case, acoustic impedance


Oil sand
cementation
versus VP / VS ratios 共Figure 16兲. This will allow us to perform rock-
Brine SS
physics analysis of well-log data and of seismic data 共e.g., elastic in-
Decreasing net to gross version results兲.
Acoustic impedance The RPTs are site 共basin兲 specific and honor local geologic fac-
tors. Geologic constraints on rock-physics models include lithology,
Figure 16. Rock-physics templates 共RPTs兲 can be made from the mineralogy, burial depth, diagenesis, pressure, and temperature. All
rock physics presented in this paper combined with Gassmann’s the- of these factors must be considered when generating RPTs for a giv-
ory to define regions where expected facies and fluids will plot. In en basin. In particular, it is essential to include only the expected
particular, the VP / VS ratio is a great fluid discriminator in siliciclastic lithologies for the area under investigation when generating the
environments. 共a兲 Homogeneous, unconsolidated sands filled with
oil are normally well separated from brine-filled sands in an RPT of RPTs. Water depth and burial depth determine the effective pressure,
VP / VS versus acoustic impedance 共AI兲. However, the effect of initial pore pressure, and lithostatic pressure. The pore pressure is impor-
cement will reduce the fluid sensitivity of sandstones, and the VP / VS tant when calculating fluid properties and determining the effective
ratio of cemented brine sandstones can be similar to the VP / VS ratio stress on the grain contacts of the rock frame carrying the overbur-
of unconsolidated sands filled with oil. The effect of N/G 共i.e., heter- den.
ogeneity兲 normally will move data in the opposite direction in a
VP / VS-AI crossplot. Hence, 共b兲 oil sands with relatively low net-to- Pore-fluid sensitivity in reservoir sandstones is highly affected by
gross can have higher VP / VS than homogeneous brine sands with ini- reservoir heterogeneity and sandstone microstructure, and it is there-
tial cement. fore important to include these geologic factors in the rock-physics
Rock physics analysis of siliciclastics 75A43

analysis. As indicated, initial cement reduces the pressure and fluid useful parameter for quantifying the heterogeneity of sands is N/G,
sensitivity of sandstones. Figure 16 shows schematically the outline the fraction of clean, permeable sand to the complete reservoir in-
of a rock-physics template, where calibrated rock-physics models cluding reservoir sands and intercalating impermeable shales. We
have been selected that fit local data observations 共well-log data or find N/G to be a useful parameter when we upscale from alternating
seismic inversion data兲 of various lithologies and pore fluids. In a thin beds of different lithologies and/or fluid saturations to an effec-
crossplot of acoustic impedance 共AI兲 versus VP / VS, the presence of tive medium during rock-physics analysis of well-log and seismic
diagenetic quartz cement will move brine-saturated sandstone data data. It is also a parameter with which geologists are very familiar.
to plot in an area of very low VP / VS where we may expect hydrocar- Takahashi 共2000兲 formulates one possible methodology to predict
bon-saturated sandstones to plot. On the contrary, reservoir hetero- sand/shale ratio based on statistical rock-physics simulations of var-
geneity and decreasing net to gross 共N/G兲 associated with interbed- ious bedding scenarios. Vernik et al. 共2002兲 predict N/G from P- and
ded sands and shales tend to move data points in the direction of the S-wave impedance-inversion results. Stovas et al. 共2006兲 use effec-
shale cluster. The cement effect is a microstructural effect, whereas tive medium theory combined with Gassmann theory to predict N/G
N/G is a scale effect. When the interbedded shale is relatively soft
and saturation from AVO attributes. Finally, Connolly and Kemper
compared to the sand, the N/G effect will counteract the effect of ce-
共2007兲 use an integrated and data-driven approach to predict N/G
ment on effective rock stiffness — hence, the opposite directions in
from turbiditic reservoirs in offshore Angola.
the AI-VP / VS crossplot. Figure 16 demonstrates why it is important
We suggest a five-step methodology 共Figure 18兲 to model the
to include and understand these geologic factors when analyzing
rock-physics properties of interbedded sands and shales with differ-
rock-physics properties and seismic-fluid sensitivity in reservoir
ent pore-fluid saturation scenarios, where the first four steps follow
sandstones.
Figure 17 shows an RPT including data from two neighboring the generation of the homogeneous rock-physics models outlined
wells penetrating Paleocene sands in the North Sea. One well pene- above. Step 1 is to estimate dry bulk and shear moduli Kdry and ␮dry at
trates a thick, turbiditic gas sand with a thin oil leg, whereas the adja- critical porosity using HM contact theory. In step 2, we interpolate
cent well penetrates a turbidite sand filled with oil. It turns out that between the high-porosity end member and the zero-porosity miner-
the sandstone quality changes from one well to the other, and this al point, choosing the modified lower or upper HS bound. In step 3,
drastically distorts the fluid sensitivity to hydrocarbons. The gas-sat- we apply Gassmann theory, perform fluid substitution, and estimate
urated top Heimdal sands in well 1 show a small increase in acoustic elastic properties of clean sands with uniform, but we vary saturation
impedance, but the oil-saturated sands in well 2 show a significant of brine Sw and hydrocarbons 共1 ⳮ Sw兲, for all porosities.
drop in acoustic impedance. This drastic change in sandstone quality In step 4, we select a characteristic shale to be interbedded with
over a short distance yields a corresponding change in seismic signa- the clean sand units. We can derive typical shale properties from
tures 共see Avseth et al., 2009b兲. well-log data in an area of study, or we can use rock-physics model-
ing. We assume that the shale will be completely impermeable to hy-
drocarbons and that the shale layers will only be saturated with brine
ROCK-PHYSICS MODELING AND UPSCALING
trapped during deposition. It is reasonable to assume that the porosi-
OF SAND-SHALE SEQUENCES
ty of thin-bedded shales at a given depth will be fairly constant, in
As we move from developing thick, hydrocarbon-filled layers to
Step 1: Hertz-Mindlin dry sandstone Step 3: Gassmann fluid sub.
thinner interbedded and laterally discontinuous sequences, we need
to focus on hydrocarbon prediction in heterogeneous reservoirs. One K dry K sat

3.0
K dry = f (Peff)
Contact-cement model

φc φc
2.5 I
Step 2: Hashin-Shtrikman interpolation Steps 4 and 5: Pick a shale
VPV S

K dry K sat + Backus upscaling


Sha
2.0 V le m
ode
l
II
Brine SS
III model (2%
cement) φsh φc
IV φc
1.5 Gas sand model
4 6 8 10 12
Acoustic impedance Figure 18. Work flow for hybrid rock-physics modeling and Backus-
(km/s∗g/cm3) average upscaling of sandstone interbedded with shale. The dry
sandstone is modeled by combining Hertz-Mindlin contact theory
Figure 17. Rock-physics template of VP / VS versus acoustic imped- and Hashin-Shtrikman interpolation 共steps 1 and 2兲. Gassmann fluid
ance 共AI兲 for target zone 共Paleocene兲 of two North Sea wells. Cluster substitution is done to estimate the effect of varying gas versus water
I is the cap-rock shale in both wells; II comprises the brine sand- saturation in the sand layers 共step 3兲. We assume a characteristic
stones in both wells; III and IV are reservoir sandstones in well 1 shale with constant porosity 共step 4兲 to be interbedded with the sands
filled with oil and gas, respectively; and V is the upper oil zone in and use the Backus average 共step 5兲 to estimate effective properties
well 2, with VP / VS higher than the brine sandstones and AI lower for varying N/G ratios. The dashed red arrows in the last subplot in-
than the gas sandstones in well 1. This is counterintuitive and must dicate that the characteristic shale with constant porosity is interbed-
be explained by difference in sandstone quality 共see Figure 16兲. Blue ded with sandstone layers with varying porosity and varying satura-
is shale, cyan is brine sand, yellow is gas sand, and red is oil sand. tion.
75A44 Avseth et al.

contrast to the porosity of the thin-bedded sands that are prone to The models for varying N/G are valid for scales such that the layer
vary with sorting. Hence, we assume a constant characteristic poros- thicknesses are much below the resolution of the elastic wave, and
ity for shale ␾ sh. typically we apply them when the thicknesses are less than about
In step 5, we apply Backus average effective medium theory 共e.g., 1 / 10 of a wavelength 共Backus, 1962兲. A reservoir sand can plot with
Gelinsky and Shapiro, 1997; Mavko et al., 2009兲 to estimate the ef- N / G ⳱ 1 in well-log data; the same sand can plot with N / G ⬍ 1 in
fective, upscaled anisotropic properties of the interbedded shale- seismic data. Hence, these effective models can be used to determine
sand sequences. This we do for various N/G values, ranging from what happens if we go from well-log scale to seismic scale.
zero to one. The Backus average approximates a stack of alternating The models can also be used to interpret scale effects and N/G val-
thin layers of two isotropic media as one effective anisotropic medi- ues in well-log data if these values happen to be less than one, which
um. This medium is characterized by five independent elastic modu- is often the case in turbidite sequences. The reservoir-sand data in
li according to the transverse isotropic elasticity matrix 共Mavko et the studied well match nicely with the models for high N/G values
al., 2009兲. The various elastic moduli that we need further are found 共0.8–1.0兲, and at first glance it appears that the gas-sand data points
from the velocities, densities, and fractions of the alternating sand fall close to the line for homogeneous 共N / G ⳱ 1兲 sandstone with
and shale layers. Finally, we can derive diagnostic rock-physics 100% gas. In Figure 20, however, we remove the brine-filled and oil-
models for varying N/G. filled sand data, leaving only the gas-saturated sands with the cap-
We apply this five-step rock-physics methodology to interpreting rock shales. Now we clearly observe the gas-saturated sands span a
well-log data from a North Sea reservoir. We create a crossplot of wide range of VP / VS. Some of the gas sands seem to have N/G values
acoustic impedance and VP / VS data from the target zone based on the between 0.8 and 0.6, causing VP / VS close to two — a value more typ-
well-log data in Figure 14. This crossplot is compared to rock-phys- ical of brine-saturated sands. This observation is consistent with a
ics models for various N/G made according to the five-step method- patchy saturation behavior. The heterogeneities of the intercalating
ology. The characteristic shale is picked from the cap-rock shale shales are causing a geologic control on the saturation pattern. Even
above the reservoir. This may be somewhat erroneous because the though the gas saturation is uniform at the scale of the sandstone po-
cap-rock shale is not neccesarily equivalent to the interbedded shale rosity, the propagating sound waves will effectively experience
within the resevoir. In Figure 19, we can see the various models for patchy saturation when the interbedded sand-shale thicknesses are
N/G of 1.0, 0.9, 0.8, and 0.6. For each of these N/G, we include vary- larger than the critical diffusion length yet beneath the resolution of
ing gas saturation within the sand layers. It is interesting to note how the sonic or seismic waves.
the decrease in N/G will cause a drastic increase of VP / VS, regardless
of porosity, even for high gas saturation in the sands. The acoustic LIMITATIONS AND CAVEATS
impedances drop drastically with decreasing N/G when the sands
have low porosities but increase slightly when the sands have high The models for the high-porosity sandstones are based on isotro-
porosity. This is of course because of the relative contrast to the inter- pic linear elasticity. All of the bound-filling models provided here
calating shale. contain some heuristic elements, such as the interpretation of the
modified bounds. The grain-contact models are limited by the fol-
3.2
lowing assumptions: the strains are small; grains are identical, ho-
3.0 mogeneous, isotropic, elastic spheres; packings are assumed to be
2.8 random and statistically isotropic; and the effective elastic constants
2.6 Shale point are relevant for long-wavelength propagation, where wavelengths
are much longer 共more than 10 times兲 compared to the grain radius.
VPV S

2.4
Furthermore, effective medium elastic constants based on grain-
2.2 S w = 100% N/G = 0.6
contact models are derived under the mean-field approximation,
2.0 N/G = 0.8

S w = 10% 3.2
1.8 N/G = 0.9
1.6 3.0
S w = 0% N/G = 1.0
2.8
1.4
2 4 6 8 10 12 14 Shale point
2.6
Acoustic impedance
(km/s∗g/cm3)
VPV S

2.4
2.2 S w = 100% N/G = 0.6
Figure 19. Rock-physics template of acoustic impedance versus
VP / VS, including models for varying net-to-gross 共N/G兲 and gas sat- 2.0 N/G = 0.8

uration 共1 ⳮ Sw兲 created by the five-step procedure. The colored 1.8 S w = 10%
N/G = 0.9
model lines represent the clean-sand models without shale interbed-
1.6
ding 共i.e., N / G ⳱ 1.0兲. The cyan line represents clean sands with S w = 0% N/G = 1.0
100% water saturation Sw. Porosity is 40% at the left end point and 1.4
decreases to the right toward the mineral point 共which is outside the 2 4 6 8 10 12 14
range of the plot兲. The dashed orange line is the corresponding 10% Acoustic impedance
3
(km/s∗g/cm )
water saturation and 90% gas saturation, whereas the solid orange
line represents 100% gas. These three sandstone lines “rotate” up-
ward toward the selected shale point when N/G decreases, indicated Figure 20. Studying the gas sands 共yellow兲 in detail, we clearly see
by black lines. For well-log data, the brine sands 共cyan兲 and oil sands they span a VP / VS of 1.5–2. Most of the gas sands have VP / VS of 1.6–
共red兲 fall between N / G ⳱ 1 and 0.8. Most of the gas sands 共yellow兲 1.7, representative for N/G values of 0.9. Hence, very little shale in-
seem to fall on similar porosities, with N/G varying between 1 and tercalation will cause a significant increase in VP / VS compared to ho-
0.6. mogeneous, clean sands 共N / G ⳱ 1兲.
Rock physics analysis of siliciclastics 75A45

which assumes that all grains experience the same mean strain field. suffer from the fact that many input parameters need to be known in
This ignores the actual grain-scale heterogeneous strains and stress- advance, many of which can become “fudge factors” in the models.
es. Mavko et al. 共2009兲 give three caveats about the use of effective- Our hybrid modeling approach for sands and sandstones combines
medium models for granular media. First, granular media have prop- physical-contact theory with heuristic elastic bounds to predict the
erties lying somewhat between solids and liquids and are sometimes microstructure of these rocks from elastic properties. For shales, we
considered a distinct form of matter. Second, complex behavior aris- recommend using inclusion-based models to capture the effect of
es from the ability of grains to move relative to each other, modify varying pore shapes and anisotropy. We have demonstrated using
their packing and coordination numbers, and rotate. Third, observed these rock-physics models on well-log data from a sequence of Pale-
behavior, depending on the stress and strain conditions, is some- ocene sands and shales in the North Sea.
times approximately nonlinearly elastic, sometimes viscoelastic, It is essential to quantify the lithology and rock texture before we
and sometimes somewhat fluidic. can reliably estimate pore-fluid and pressure effects. Therefore, the
Many workers 共Goddard, 1990; Makse et al., 1999, 2004; Duffaut link between geologic processes and rock properties should always
et al., 2010兲 show that this complex behavior causes effective-medi- be integrated in the work flow of any seismic-reservoir characteriza-
um theory to fail in cohesionless granular assemblies. Closed-form tion study. Finally, one should note the gap between the different
effective-medium theories tend to predict the incorrect 共relative to scales, from the pore-scale microstructure to seismic wiggles. We
laboratory observations兲 dependence of effective moduli on pres- use the Backus average as one technique to upscale and estimate ef-
sure and poor estimates of bulk-to-shear moduli ratio. Shear moduli fective seismic properties of interbedded sequences. However, clos-
tend to be overpredicted and often require heuristic corrections 共e.g., ing the gap between different scales remains a challenging problem
Bachrach and Avseth, 2008.兲 Numerical methods, referred to as mo- in rock-physics modeling applied to seismic exploration of hetero-
lecular dynamics or discrete element modeling, which simulate the geneous reservoirs.
motions and interactions of thousands of grains, appear to come
closer to predicting observed behavior 共e.g., Makse et al., 2004; Gar-
ACKNOWLEDGMENTS
cia and Medina, 2006兲. Closed-form effective-medium models can
be useful because it is not always practical to run a numerical simula- Thanks to Anders Dræge at Statoil for contributions on building
tion; however, model predictions of the types presented in this sec- the rock-physics depth trend for shale used in this paper. We also ac-
tion must be used with care. knowledge the sponsors of the Stanford Rock Physics and Borehole
In modeling unconsolidated sands, we have assumed normally Geophysics Project 共SRB兲 and the Stanford Center for Reservoir
compacted 共Middleton and Wilcock, 1994兲 sediments and have esti- Forecasting 共SCRF兲.
mated the effective pressure used in the HM contact theory from
Terzaghi’s principle 共equation 9兲. This principle implies that com-
pression is one dimensional 共i.e., considering only the vertical axis兲 REFERENCES
and pore pressures are hydrostatic. This assumption will be violated
and predictions from HM contact theory will be erroneous in areas Anselmetti, F. S., and G. P. Eberli, 1997, Sonic velocity in carbonate sedi-
ments and rocks, in I. Palaz, and K. J. Marfurt, eds., Carbonate seismolo-
with significant stress anisotropy and/or overpressure conditions gy: SEG, 53–75.
共e.g., Xu, 2002; Vega, 2003兲. Avseth, P., R. Bachrach, T. Bersås, and A. Norenes Haaland, 2009a, Fluid and
For the sand and shale models presented above, a limitation arises stress sensitivity in cemented sandstones: 79th Annual International Meet-
ing, SEG, Expanded Abstracts, 2015–2019.
for mixed mineralogies. For mixed mineralogies, often an effective- Avseth, P., A. Dræge, A.-J. van Wijngaarden, T. A. Johansen, and A. Jørstad,
average mineral modulus must be estimated. As one of the inputs, 2008, Shale rock physics and implications for AVO analysis: A North Sea
demonstration: The Leading Edge, 27, 788–797, doi: 10.1190/1.2944166.
the mineral moduli of clays are variable and not very well known, Avseth, P., J. Dvorkin, G. Mavko, and J. Rykkje, 2000, Rock physics diag-
leading to uncertainties in the model predictions. Nevertheless, we nostic of North Sea sands: Link between microstructure and seismic prop-
have found these models useful for high-porosity siliciclastic rocks erties: Geophysical Research Letters, 27, 2761–2764, doi: 10.1029/
1999GL008468.
in many situations. We are confident that future research will lead to Avseth, P., A. Jørstad, A.-J. van Wijngaarden, and G. Mavko, 2009b, Rock
better models. physics estimation of cementation, sorting and net-to-gross in North Sea
sandstones: The Leading Edge, 28, 98–108, doi: 10.1190/1.3064154.
Avseth, P., T. Mukerji, and G. Mavko, 2005, Quantitative seismic interpreta-
tion: Applying rock physics tools to reduce interpretation risk: Cambridge
CONCLUSIONS University Press.
Bachrach, R., and P. Avseth, 2008, Rock physics modeling of unconsolidated
There are many different rock-physics models and modeling ap- sands: Accounting for nonuniform contacts and heterogeneous stress
proaches for understanding and interpreting the seismic responses of fields in the effective media approximation with applications to hydrocar-
bon exploration: Geophysics, 73, no. 6, E197–E209, doi: 10.1190/
subsurface rocks and their pore fluids. We have listed different types 1.2985821.
of models, ranging from simple empirical models to more complex Backus, G. E., 1962, Long-wave elastic anisotropy produced by horizontal
layering: Journal of Geophysical Research, 67, 4427–4440, doi: 10.1029/
physical and computational models. There are pros and cons associ- JZ067i011p04427.
ated with each model, yet in the end all models are, strictly speaking, Bakke, S., and P. E. Øren, 1997, 3-D pore-scale modelling of sandstones and
wrong. Still, we have no choice but to find a model, or a combination flow simulations in the pore networks: SPE Journal, 2, no. 2, 136–149, doi:
10.2118/35479-PA.
of models, useful 共i.e., yield reasonable predictions or lay the Berryman, J. G., 1980, Long-wavelength propagation in composite elastic
groundwork for reliable interpretation兲 for a given case or scenario. media: The Journal of the Acoustical Society of America, 68, 1809–1831,
Empirical models are useful because of their simplicity and the lim- doi: 10.1121/1.385171.
Berryman, J. G., and G. W. Milton, 1991, Exact results for generalized Gas-
ited number of input parameters. However, they can be erroneous smann’s equation in composite porous media with two constituents: Geo-
when applied to areas with a different geologic setting or depth than physics, 56, 1950–1960, doi: 10.1190/1.1443006.
Biot, M. A., 1956, Theory of propagation of elastic waves in a fluid saturated
the data on which the models were generated. More complex phys- porous solid. I. Low-frequency range: Journal of the Acoustical Society of
ics-based models are normally better for extrapolation exercises but America, 28, no. 2, 168–178, doi: 10.1121/1.1908239.
75A46 Avseth et al.

Bjørlykke, K., 1998, Clay mineral diagenesis in sedimentary basins — A key tropic layered fluid- and gas-saturated sediments: Geophysics, 62,
to the prediction of rock properties: Examples from the North Sea basin: 1867–1878.
Clay Minerals, 33, no. 1, 15–34, doi: 10.1180/claymin.1998.033.1.03. Goddard, J. D., 1990, Nonlinear elasticity and pressure-dependent wave
Bosl, W., J. Dvorkin, and A. Nur, 1998, A study of porosity and permeability speeds in granular media: Proceedings of the Royal Society of London, Se-
using a lattice Boltzmann simulation: Geophysical Research Letters, 25, ries A, 430, no. 1878, 105–131, doi: 10.1098/rspa.1990.0083.
1475–1478, doi: 10.1029/98GL00859. Greenberg, M. L., and J. P. Castagna, 1992, Shear-wave velocity estimation
Box, G. E. P., and N. R. Draper, 1987, Empirical model building and response in porous rocks: Theoretical formulation, preliminary verification and ap-
surfaces: John Wiley & Sons, Inc. plications: Geophysical Prospecting, 40, no. 2, 195–209, doi: 10.1111/
Brevik, I., G. R. Ahmadi, T. Hatteland, and M. A. Rojas, 2007, Documenta- j.1365-2478.1992.tb00371.x.
tion and quantification of velocity anisotropy in shales using wireline log Hamilton, E. L., 1956, Low sound velocities in high porosity sediments:
measurements: The Leading Edge, 26, 272–277, doi: 10.1190/1.2715048. Journal of the Acoustical Society of America, 28, no. 1, 16–19, doi:
Carmichael, R. S., 1989, Practical handbook of physical properties of rocks 10.1121/1.1908208.
and minerals: CRC Press. Han, D., 1986, Effects of porosity and clay content on acoustic properties of
Cheng, C. H., 1978, Seismic velocities in porous rocks: Direct and inverse sandstones and unconsolidated sediments: Ph.D. dissertation, Stanford
problems: Ph.D. dissertation, Massachusetts Institute of Technology. University.
——–, 1993, Crack models for a transversely anisotropic medium: Journal of Hashin, Z., and S. Shtrikman, 1963, A variational approach to the elastic be-
Geophysical Research, 98, no. B1, 675–684, doi: 10.1029/92JB02118. havior of multiphase materials: Journal of the Mechanics and Physics of
Connolly, P., and M. Kemper, 2007, Statistical uncertainty of seismic net pay Solids, 11, no. 2, 127–140, doi: 10.1016/0022-5096共63兲90060-7.
estimations: The Leading Edge, 26, 1284–1289, doi: 10.1190/1.2794387. Hornby, B. E., L. M. Schwartz, and J. A. Hudson, 1994, Anisotropic effective
Crampin, S., 1984, Effective anisotropic elastic constants for wave propaga- medium modeling of the elastic properties of shales: Geophysics, 59,
tion through cracked solids: Journal of the Royal Astronomical Society, 1570–1583, doi: 10.1190/1.1443546.
76, 135–145. Hudson, J. A., 1980, Overall properties of a cracked solid: Mathematical Pro-
Digby, P. J., 1981, The effective elastic moduli of porous granular rocks: ceedings of the Cambridge Philosophical Society, 88, no. 2, 371–384, doi:
Journal of Applied Mechanics, 48, 803–808, doi: 10.1115/1.3157738. 10.1017/S0305004100057674.
Dræge, A., T. A. Johansen, I. Brevik, and C. Thorsen Dræge, 2006a, A strate- ——–, 1981, Wave speeds and attenuation of elastic waves in material con-
gy for modeling the diagenetic evolution of seismic properties in sand- taining cracks: Geophysical Journal of the Royal Astronomical Society,
stones: Petroleum Geoscience, 12, no. 4, 309–323, doi: 10.1144/1354- 64, 133–150.
079305-691. ——–, 1990, Overall elastic properties of isotropic materials with arbitrary
Dræge, A., M. Jakobsen, and T. A. Johansen, 2006b, Rock physics modeling distribution of circular cracks: Geophysical Journal International, 102,
of shale diagenesis: Petroleum Geoscience, 12, no. 1, 49–57, doi: 10.1144/ 465–469, doi: 10.1111/j.1365-246X.1990.tb04478.x.
1354-079305-665. Johansen, T. A., M. Jakobsen, and B. O. Ruud, 2002, Estimation of the inter-
Duffaut, K., and M. Landrø, 2007, VP / VS ratio versus differential stress and nal structure and anisotropy of shales from borehole data: Journal of Seis-
rock consolidation — A comparison between rock models and time-lapse mic Exploration, 11, 363–381.
AVO data: Geophysics, 72, no. 5, C81–C94, doi: 10.1190/1.2752175. Keehm, Y., 2003, Computational rock physics: Transport properties in po-
Duffaut, K., M. Landrø, and R. Sollie, 2010, Using Mindlin theory to model rous media and applications: Ph.D. dissertation, Stanford University.
friction dependent shear modulus in granular media: Geophysics, 75, no. Knackstedt, M. A., S. Latham, M. Madadi, A. Sheppard, T. Varslot, and C.
3, E143–E152. Arns, 2009, Digital rock physics: 3D imaging of core material and correla-
Dutta, T., G. Mavko, and T. Mukerji, 2010, Improved granular medium mod- tions to acoustic and flow properties: The Leading Edge, 28, 28–33, doi:
el for unconsolidated sands using coordination number, porosity, and pres- 10.1190/1.3064143.
sure relations: Geophysics, 76, no. 2, E91–E99, doi: 10.1190/1.3333539. Kuster, G. T., and M. N. Toksöz, 1974, Velocity and attenuation of seismic
Dvorkin, J., and I. Brevik, 1999, Diagnosing high-porosity sandstones: waves in two-phase media, Part I: Theoretical formulations: Geophysics,
Strength and permeability from porosity and velocity: Geophysics, 64, 39, 587–606, doi: 10.1190/1.1440450.
795–799, doi: 10.1190/1.1444589. MacQuaker, J. H. S., K. G. Taylor, and R. L. Gawthorpe, 2007, High-resolu-
Dvorkin, J., and A. Nur, 1996, Elasticity of high-porosity sandstones: Theory tion facies analyses of mudstones: Implications for paleoenvironmental
for two North Sea data sets: Geophysics, 61, 1363–1370, doi: 10.1190/ and sequence stratigraphic interpretations of offshore ancient mud-domi-
1.1444059. nated successions: Journal of Sedimentary Research, 77, no. 4, 324–339,
Dvorkin, J., A. Nur, and H. Yin, 1994, Effective properties of cemented gran- doi: 10.2110/jsr.2007.029.
ular material: Mechanics of Materials, 18, 351–366, doi: 10.1016/0167- Makse, H. A., N. Gland, D. L. Johnson, and L. M. Schwartz, 1999, Why ef-
6636共94兲90044-2. fective medium theory fails in granular materials: Physical Review Let-
Elata, D., and J. Berrymann, 1996, Contact force-displacement laws and the ters, 83, 5070–5073, doi: 10.1103/PhysRevLett.83.5070.
mechanical behavior of random packs of identical spheres: Mechanics of ——–, 2004, Granular packings: Nonlinear elasticity, sound propagation,
Materials, 24, 229–240, doi: 10.1016/S0167-6636共96兲00034-8. and collective relaxation dynamics: Physical Review E, L70,
Ellis, D., J. Howard, C. Flaum, D. McKeon, H. Scott, O. Serra, and G. Sim- 061302.1–061302.19.
mons, 1988, Mineral logging parameters: Nuclear and acoustic: The Tech- Marcussen, Ø., B. Thyberg, C. Peltonen, J. Jahren, K. Bjørlykke, and J. I. Fa-
nical Review, 36, 38–55. leide, 2008, Physical properties of Cenozoic mudstones from the northern
Eshelby, J. D., 1957, The determination of the elastic field of an ellipsoidal North Sea: Impact of clay mineralogy on compaction trends: AAPG Bulle-
inclusion, and related problems: Proceedings of the Royal Society of Lon- tin, 93, no. 1, 127–150, doi: 10.1306/08220808044.
don, Series A: Mathematical and Physical Sciences, 241, no. 1226, Marion, D., 1990, Acoustical, mechanical and transport properties of sedi-
376–396, doi: 10.1098/rspa.1957.0133. ments and granular materials: Ph.D. dissertation, Stanford University.
Fabricius, I. L., 2003, How burial diagenesis of chalk sediments controls son- Mavko, G., and D. Jizba, 1991, Estimating grain-scale fluid effects on veloci-
ic velocity and porosity: AAPG Bulletin, 87, no. 11, 1–24, doi: 10.1306/ ty dispersion in rocks: Geophysics, 56, 1940–1949, doi: 10.1190/
06230301113. 1.1443005.
FaustAndersen, C., and T. A. Johansen, 2010, Test of rock physics models for Mavko, G., T. Mukerji, and J. Dvorkin, 2009, The rock physics handbook,
prediction of seismic velocities in shallow unconsolidated sands: A well 2nd ed.: Cambridge University Press.
log data case: Geophysical Prospecting, doi: 10.1111/j1365– Mavko, G., T. Mukerji, and N. Godfrey, 1995, Predicting stress-induced ve-
2478.2010.00870.x. locity anisotropy in rocks: Geophysics, 60, 1081–1087, doi: 10.1190/
Florez, J.-M., 2005, Integrating geology, rock physics, and seismology for 1.1443836.
reservoir quality prediction: Ph.D. dissertation, Stanford University. Middleton, G. V., and P. R. Wilcock, 1994, Mechanics in the earth and envi-
Florez, J.-M., and G. Mavko, 2004, Pressure-solution and the rock physics ronmental sciences: Cambridge University Press.
diagenetic trend in quartzose sandstones: 74th Annual International Meet- Mindlin, R. D., 1949, Compliance of elastic bodies in contact: Journal of Ap-
ing, SEG, Expanded Abstracts, 1702–1705. plied Mechanics, 16, 259–268.
Gal, D., J. Dvorkin, and A. Nur, 1998, A physical model for porosity reduc- Mondol, N., K. Bjørlykke, J. Jahren, and K. Høeg, 2007, Experimental me-
tion in sandstones: Geophysics, 63, 454–459, doi: 10.1190/1.1444346. chanical compaction of clay mineral aggregates — Changes in physical
Garcia, X., and A. Medina, 2006, Hysteresis effects studied by numerical properties of mudstones during burial: Marine and Petroleum Geology,
simulations: Cyclic loading-unloading of a realistic sand model: Geophys- 24, no. 5, 289–311, doi: 10.1016/j.marpetgeo.2007.03.006.
ics, 71, no. 2, F13–F20, doi: 10.1190/1.2181309. Murphy, W. F. III, 1982, Effects of microstructure and pore fluids on the
Gardner, G. H. F., L. W. Gardner, and A. R. Gregory, 1974, Formation veloc- acoustic properties of granular sedimentary materials: Ph.D. dissertation,
ity and density — The diagnostic basics for stratigraphic traps: Geophys- Stanford University.
ics, 39, 770–780, doi: 10.1190/1.1440465. Norris, A. N., 1985, A differential scheme for the effective moduli of com-
Gassmann, F., 1951, Über die elastizität poröser medien: Vierteljahrsschrift posites: Mechanics of Materials, 4, no. 1, 1–16, doi: 10.1016/0167-
der Naturforschenden Gesellschaft in Zürich, 96, 1–23. 6636共85兲90002-X.
Gelinsky, S., and S. Shapiro, 1997, Poroelastic Backus averaging for aniso- Norris, A. N., and D. L. Johnson, 1997, Nonlinear elasticity of granular me-
Rock physics analysis of siliciclastics 75A47

dia: ASME Journal of Applied Mechanics, 64, 39–49. ly layered reservoirs: Geophysics, 71, no. 3, C25–C36, doi: 10.1190/
Nur, A., 1992, Critical porosity and the seismic velocities in rocks: Eos, 1.2197487.
Transactions, American Geophysical Union, 73, 43–66. Takahashi, I., 2000, Quantifying information and uncertainty of, rock, prop-
Nur, A., D. Marion, and H. Yin, 1991, Wave velocities in sediments, in J. M. erty estimation from seismic data: Ph.D. dissertation, Stanford University.
Hovem, M. D. Richardson and R. D. Stoll, eds., Shear waves in marine Thomsen, L., 1986, Weak elastic anisotropy: Geophysics, 51, 1954–1966,
sediments: Kluwer Academic Publishers, 131–140. doi: 10.1190/1.1442051.
Nur, A., G. Mavko, J. Dvorkin, and D. Gal, 1995, Critical porosity: The key Thyberg, B., J. Jahren, T. Winje, K. Bjørlykke, and J. I. Faleide, 2009, From
to relating physical properties to porosity in rocks: 65th Annual Interna- mud to shale: Rock stiffening by micro-quartz cementation: First Break,
tional Meeting, SEG, Expanded Abstracts, 878–881. 27, 27–33.
O’Connell, R. J., and B. Budiansky, 1974, Seismic velocities in dry and satu- Vega, S., 2003, Intrinsic and stress-induced velocity anisotropy in unconsoli-
rated cracked solids: Journal of Geophysical Research, 79, 4626–4627, dated sands: Ph.D. dissertation, Stanford University.
doi: 10.1029/JB079i035p05412. Vernik, L., and D. Fisher, 2002, Estimation of net-to-gross from P and S im-
——–, 1977, Viscoelastic properties of fluid-saturated cracked solids: Jour- pedance in deepwater turbidites: The Leading Edge, 21, 380–387, doi:
nal of Geophysical Research, 82, 5719–5735, doi: 10.1029/ 10.1190/1.1471602.
JB082i036p05719. Vernik, L., and J. Hamman, 2009, Stress sensitivity of sandstones and 4D ap-
Ødegaard, E., and P. Avseth, 2004, Well log and seismic data analysis using plications: The Leading Edge, 28, 90–93, doi: 10.1190/1.3064152.
rock physics templates: First Break, 22, 37–43. Vernik, L., and A. Nur, 1992, Petrophysical classification of siliciclastics for
Peltonen, C., Ø. Marcussen, K. Bjørlykke, and J. Jahren, 2008, Clay mineral lithology and porosity prediction from seismic velocities: AAPG Bulletin,
diagenesis and quartz cementation in mudstones: The effects of smectite to 76, 1295–1309.
illite transformation of rock properties: Marine and Petroleum Geology, Voltolini, M., H.-R. Wenk, N. H. Mondol, K. Bjørlykke, and J. Jahren, 2009,
26, no. 6, 887–898, doi: 10.1016/j.marpetgeo.2008.01.021. Anisotropy of experimentally compressed kaolinite-illite-quartz mix-
Raiga-Clemenceau, J., J. P. Martin, and S. Nicoletis, 1988, The concept of
acoustic formation factor for more accurate porosity determination from tures: Geophysics, 74, no. 1, D13–D23, doi: 10.1190/1.3002557.
sonic transit time data: The Log Analyst, 219, 54–60. Walsh, J. B., 1965, The effect of cracks on the compressibility of rock: Jour-
Ramm, M., and K. Bjørlykke, 1994, Porosity/depth trends in reservoir sand- nal of Geophysical Research, 70, 381–389, doi: 10.1029/
stones: Assessing the quantitative effects of varying pore-pressure, tem- JZ070i002p00381.
perature history and mineralogy, Norwegian Shelf data: Clay Minerals, Walton, K., 1987, The effective elastic moduli of a random packing of
29, no. 4, 475–490, doi: 10.1180/claymin.1994.029.4.07. spheres: Journal of the Mechanics and Physics of Solids, 35, no. 2,
Raymer, L. L., E. R. Hunt, and J. S. Gardner, 1980, An improved sonic transit 213–226, doi: 10.1016/0022-5096共87兲90036-6.
time-to-porosity transform: 21st Annual Logging Symposium, Society of Wang, Z., 2002, Seismic anisotropy in sedimentary rocks, Part 2: Laboratory
Professional Well Log Analysts, Paper P. data: Geophysics, 67, 1423–1440, doi: 10.1190/1.1512743.
Ruud, B., M. Jakobsen, and T. A. Johansen, 2003, Seismic properties of Wyllie, M. R. J., A. R. Gregory, and G. H. F. Gardner, 1958, An experimental
shales during compaction: 73rd Annual International Meeting, SEG, Ex- investigation of factors affecting elastic wave velocities in porous media:
panded Abstracts, 1294–1297, doi: 10.1190/1.1817522. Geophysics, 23, 459–493, doi: 10.1190/1.1438493.
Sayers, C. M., and M. Kachanov, 1995, Microcrack-induced elastic wave an- Xu, S., 2002, Stress-induced anisotropy in unconsolidated sands and its ef-
isotropy of brittle rocks: Journal of Geophysical Research, 100, no. B3, fect on AVO analysis: 72nd Annual International Meeting, SEG, Expand-
4149–4156, doi: 10.1029/94JB03134. ed Abstracts, 105–108.
Schoenberg, M., and J. Douma, 1988, Elastic wave propagation in media Xu, S., and R. E. White, 1995, A new velocity model for clay-sand mixtures:
with parallel fractures and aligned cracks: Geophysical Prospecting, 36, Geophysical Prospecting, 43, 91–118, doi: 10.1111/j.1365-
571–590, doi: 10.1111/j.1365-2478.1988.tb02181.x. 2478.1995.tb00126.x.
Schoenberg, M., and C. Sayers, 1995, Seismic anisotropy of fractured rock: Yin, H., 1992, Acoustic velocity and attenuation of rocks, isotropy, intrinsic
Geophysics, 60, 204–211, doi: 10.1190/1.1443748. anisotropy, and stress induced anisotropy: Ph.D. dissertation, Stanford
Stovas, A., M. Landrø, and P. Avseth, 2006, AVO attribute inversion for fine- University.

You might also like