GEOPHYSICS, VOL. 68, NO. 2 (MARCH-APRIL 2003); P. 430–440, 8 FIGS., 2 TABLES.

10.1190/1.1567211

Tutorial

Gassmann fluid substitutions: A tutorial

Tad M. Smith∗ , Carl H. Sondergeld‡ , and Chandra S. Rai‡

ABSTRACT pore fluid, and the new effective bulk modulus and den-
sity are calculated.
Fluid substitution is an important part of seismic A direct result of Gassmann’s equations is that the
attribute work, because it provides the interpreter shear modulus for an isotropic material is independent
with a tool for modeling and quantifying the vari- of pore fluid, and therefore remains constant during the
ous fluid scenarios which might give rise to an ob- fluid substitution process. In the case of disconnected or
served amplitude variation with offset (AVO) or 4D cracklike pores, however, this assumption may be vio-
response. The most commonly used technique for lated. Once the values for the new effective bulk modulus
doing this involves the application of Gassmann’s and bulk density are calculated, it is possible to calculate
equations. the compressional and shear velocities for the new fluid
Modeling the changes from one fluid type to another conditions.
requires that the effects of the starting fluid first be re- There are other approaches to fluid substitution (em-
moved prior to modeling the new fluid. In practice, the pirical and heuristic) which avoid the porous frame cal-
rock is drained of its initial pore fluid, and the moduli culations but, as described in this tutorial, often do not
(bulk and shear) and bulk density of the porous frame yield reliable results. This tutorial provides the reader
are calculated. Once the porous frame properties are with a recipe for performing fluid substitutions, as well
properly determined, the rock is saturated with the new as insight into why and when the approach may fail.

INTRODUCTION to highlight some of the commonly encountered caveats and

pitfalls inherent to the modeling process. By understanding the
Fluid substitutions are an important part of any seismic at-
equations and processes discussed in this tutorial, the reader
tribute study, as they provide the interpreter with a valuable
should be able to program a spreadsheet or build a user pro-
tool for modeling various fluid scenarios which might explain
gram to perform fluid substitutions on a routine basis. We also
an observed amplitude variation with offset (AVO) anomaly. A
include a simple case study from the deepwater Gulf of Mexico,
common empirical approach for modeling velocity-porosity re-
using wireline log data.
lationships was proposed by Wyllie et al. (1956, 1958) and later
As with any technology, recognition of pitfalls and the abil-
modified by Raymer et al. (1980). The most commonly used
ity to diagnose and solve difficult problems comes with time
(and theoretically sound) approach at seismic frequencies is
and experience. Consequently, care should always be exercised
based on the work of Gassmann (1951), which relates the bulk
when performing fluid substitutions, and experienced users
modulus of a rock to its pore, frame, and fluid properties.
should be consulted.
Unfortunately, many users of Gassmann’s theory never de-
velop an understanding of the strengths and weaknesses of the BACKGROUND
model, simply because of the widespread reliance on commer-
cially available software. The purpose of this tutorial is to guide Modeling the effects of mineral composition, porosity, and
the reader through practical Gassmann fluid substitution, and fluid on seismic velocities is generally based on a combination

Manuscript received by the Editor October 18, 2001; revised manuscript received October 1, 2002.
∗
Formerly Veritas DGC, Houston, Texas; presently Newfield Exploration, 363 North Sam Houston Parkway E., Houston, Texas 77060. E-mail:
[email protected].
‡University of Oklahoma, 4502 E. 41st Street, Tulsa, Oklahoma 74135. E-mail: [email protected].
°c 2003 Society of Exploration Geophysicists. All rights reserved.

430
 Gassmann Fluid Substitutions 431

of empirical relationships and theoretical formulations. Em- Second, Gassmann’s equation is valid only at low enough
pirical approaches attempt to establish relationships between frequencies such that pore pressures are equalized over a
porosity, fluid velocities, and the velocity of the rock matrix, length scale much greater than a pore dimension and much
and generally involve regression analyses of core or log mea- less than the wavelength of the passing seismic wave. For high
surements (e.g., Wyllie et al., 1956, Raymer et al., 1980, Han frequencies, original formulations presented by Biot (1956,
et al., 1986; Eberhart-Phillips, et al., 1989). However, these re- 1962) should be used. At sonic logging frequencies, Gassmann
lationships are not based on physical principles and generally may or may not be applicable (Berryman, 1999), although re-
do not work well for fluid substitution modeling. liable results are often obtained for clean sands with high ef-
The most commonly used theoretical approach for fluid fective porosities, such as those encountered in the deepwater
substitutions employs the low-frequency Gassmann theory Gulf of Mexico. Unreliable results often occur when apply-
(Gassmann, 1951). Gassmann’s equation relates the saturated ing Gassmann’s equation to low-porosity or shaley sands or to
bulk modulus of the rock to its porosity, the bulk modulus of carbonate rocks. This is because basic assumptions regarding
the porous rock frame, the bulk modulus of the mineral matrix, frequency or pore connectivity are violated.
and the bulk modulus of the pore-filling fluids: In this tutorial, we focus on simple application of equation (1)
µ ¶ and assume that all the model assumptions are met. Additional
K∗ 2 pitfalls and sources of error are discussed in a later section.
1−
Ko
K sat = K∗ + , (1) DEFINITIONS
φ (1 − φ) K∗
+ − 2
K fl Ko Ko Prior to delving into Gassmann, we must first define and
briefly discuss the bulk and shear moduli of the rock, as well
where K sat = the saturated bulk modulus (undrained of pore as the bulk modulus of the pore-filling fluid. In addition, we
fluids), K o = the bulk modulus of the mineral matrix, K fl = the must also discuss techniques for mixing complex mineralogies
bulk modulus of the pore fluid, K ∗ = the bulk modulus of and multiple fluids. For excellent and comprehensive reviews
the porous rock frame, (drained of any pore-filling fluid), and of rock and fluid properties, the reader is referred to Batzle
ø = porosity. and Wang (1992), Castagna et al. (1993), Mavko et al. (1998),
Application of this equation is a two-part process, whereby and Wang (2001).
we first determine the bulk modulus of the porous rock frame
(the bulk modulus of the rock drained of its initial pore-filling
Rock properties
fluid, also referred to as the “dry frame” bulk modulus), after
which we calculate the bulk modulus of the rock saturated with Gassmann’s equation relates the saturated bulk modulus of
any desired fluid. the rock (K sat ) to its porous frame properties and the proper-
When applying equation (1) to core measurements, it is im- ties of the pore-filling fluid [equation (1)]. The bulk modulus,
portant to recognize that the “dry frame” bulk modulus, K ∗ , or incompressibility, of an isotropic rock is defined as the ratio
represents the “dry frame” property of the rock with a small of hydrostatic stress to volumetric strain. Values for bulk mod-
amount of moisture present (see Clark, et al., 1980). In prac- ulus can be obtained either by dynamic laboratory measure-
tice, these values would come from controlled humidity-dried ments of velocities or from analysis of wireline log data (typical
core samples. Predicted velocities will be too fast if the porous static measurements for the moduli violate the assumption of
rock frame is assumed to be absolutely dry. In this tutorial, infinitesimal strain, and should be avoided for fluid substitu-
rather than confuse the reader by referring to the “dry frame” tion purposes). Whether from dynamic velocity measurements
properties, we instead refer to the “porous rock frame”. or from wireline log data, we can relate the bulk modulus of
a rock, K sat , to its compressional velocity, shear velocity, and
bulk density through the following relationship:
Model assumptions µ ¶
4
Application of Gassmann’s equation is based on several as- K = ρ B V p2 − Vs2 , (2)
3
sumptions. First, the model assumes that the rock is homo-
geneous and isotropic, and that the pore space is completely where ρ B is the bulk density of the rock, V p is its compressional
connected. This assumption is violated if the rock framework velocity, and Vs is its shear velocity. Note that this equation is
is composed of multiple minerals with large contrasts in elas- easily rewritten and solved for V p . If velocity and density mea-
tic stiffness (Berge, 1998), or if it composed of preferentially- surements are used to calculate K sat from wireline log data, the
oriented anisotropic minerals (Brown and Korringa, 1975). resulting value will be the bulk modulus of the rock saturated
Fortunately, Gassmann’s equation is free of assumptions about with the in-situ pore fluid. If velocity measurements are made
pore geometry, although the pore system must be connected on controlled humidity-dried core samples, the calculated bulk
and fluids must be moveable. However, when multiple pore modulus will represent the bulk modulus of the porous rock
types are present in a rock, it is often necessary to use more framework, K ∗ .
complex models (Berryman and Milton, 1991; O’Connell, 1984; The shear modulus (G), or shear stiffness, of a rock is defined
Berryman and Wang, 2000). Note that many of the assump- as the ratio of shear stress to shear strain. As with the bulk
tions regarding pore connectivity and pore type are probably modulus, it can be determined by laboratory tests or analysis
violated in low-porosity rocks. Carbonate rocks, in particular, of wireline log data, and is given by the following equation:
may be problematic due to the diverse pore types often ob-
served in thin section and the low connectivity of these pores. G = ρ B Vs2 . (3)
432 Smith et al.

The shear modulus, G, is also frequently represented by the Because there typically are two or more fluid phases occupy-
symbol µ. ing the pore space of a reservoir rock, we must calculate a bulk
If velocities are in kilometers per second and densities are modulus and density of the individual fluid end members, and
in grams per cubic centimeter, the resultant moduli (K and G) then mix the fluids according to the following physical rules.
will be in gigapascals (GPa). It is important to recognize that Gassmann’s equation assumes all the pore space is connected
the saturated bulk modulus of a rock may be sensitive to the and pore pressure is equilibrated throughout the rock. Thus,
composition of the pore-filling fluid, whereas the shear modu- the assumptions of a homogeneous fluid, uniformly distributed
lus is insensitive to pore fluid [that is, G dry = G wet (Biot, 1956; throughout the pore space, allows the bulk modulus of the fluid
Berryman and Milton, 1991; Berryman, 1999)] . Thus, the shear mixture to be calculated via the isostress, or Reuss, average:
modulus does not vary during the course of a fluid substi- ·X
n ¸−1
Si
tution. This understanding is fundamental to application of K fl = , (5)
Biot-Gassmann theory in general, and Gassmann’s equation i=1
Ki
specifically.
where K fl is the bulk modulus of the fluid mixture, K i is the bulk
A fourth equation is also necessary for performing fluid sub-
modulus of the individual phases, and Si is their saturation.
stitutions. This equation describes the relationship between the
For a simple two-component hydrocarbon-water system, this
fluid density (ρfl ), porosity (φ), grain density of the rock matrix
equation becomes
(ρg ), and the rock bulk density (ρ B ): · ¸−1
Sw (1 − Sw )
ρ B = ρg (1 − φ) + ρfl φ. (4) K fl = + , (6)
Kw K hc
This equation is also easily rewritten and solved for porosity. where Sw is the water saturation, K w is the bulk modulus of
the water, and K hc is the bulk modulus of the hydrocarbon.
USING GASSMANN’S EQUATION
This equation is easily expanded to account for additional fluid
Before we begin fluid substitution using equation (1), we phases.
must first determine (1) the porosity of the rock, φ, (2) the For the fluid density, a simple volumetric mix of the end-
properties of the fluids (K fl , ρfl ) that occupy the pore space, member components is used to calculate the density of the
(3) the bulk modulus of the mineral matrix (K 0 ), and (4) the fluid mixture:
bulk modulus of the porous rock frame (K ∗ ). All four compo- X
n

nents may be defined or inferred through laboratory measure- ρfl = Si ρi , (7)

ment or analysis of wireline log data. i=1

where Si is the saturation of the individual components, and ρi

Porosity is the density of the individual components. For a simple two-
component hydrocarbon-water system, this equation becomes:
Porosity is routinely calculated from core data or from anal-
ysis of wireline log data [equation (4)] is rewritten and solved ρfl = Swρw + (1 − Sw)ρhc , (8)
for porosity. Because logging tools do not directly measure
porosity or bulk density, calibration of the log-derived poros- where ρw is the density of water, and ρhc is the density of the
ity to measured core porosity is highly desirable, and in some hydrocarbon. As with equation (6), this equation is easily ex-
instances (e.g., when dealing with complex lithologies or low- panded to include additional fluid phases.
porosity rocks) may significantly alter the results of a fluid sub-
stitution. Core calibration may also be particularly important Matrix properties
if the formation is invaded by drilling fluids.
To calculate the bulk modulus of the mineral matrix, K o , in-
Fluid properties formation on the composition of the rock must be available.
If core samples are available, mineral abundance may be de-
Prior to performing a fluid substitution, we must also know termined using conventional laboratory techniques [e.g., point
something about the bulk modulus and density of the in-situ counting of thin sections, X-ray diffraction, or Fourier trans-
pore-filling fluid, as well as those of the new fluid we wish form infrared analysis (Sondergeld and Rai, 1993)]. In the ab-
to model. Three approaches are commonly used for deter- sence of core data, lithology can be approximated from wire-
mining these values: (1) the properties are measured directly line logs by simple clay volume (Vclay ) analysis and assuming
(at reservoir temperatures and pressures) from pore fluids re- the two mineral end members are quartz and clay. For more
covered from the reservoir, (2) the properties are calculated complex lithologies, however, other techniques must be ap-
from equations of state (see McCain, 1990; Danesh, 1998), or plied which allow the volumetric abundance of the mineral
(3) the properties are calculated from an empirical calcula- constituents to be estimated (e.g., deterministic techniques,
tor (e.g., Batzle and Wang, 1992). Note that typical laboratory multilinear regression, neural networks, etc.).
analyses of reservoir fluids yield the isothermal properties of Once the mineral abundances are determined, K 0 is cal-
the fluid bulk modulus, whereas wave propagation involves culated via application of Voigt-Reuss-Hill (VRH) averaging
the adiabatic properties. Generally, the difference between the of the mineral constituents [alternatively, Hashin-Shtrikman
two is small unless the fluid has a relatively high gas-oil-ratio (HS) averaging may be used (Hashin and Shtrikman, 1962)].
(GOR). For additional detailed information on the behavior A VRH average is simply the average of the harmonic (Reuss
of fluids, the reader is referred to McCain (1990), Batzle and average) and arithmetic means (Voigt average) for the min-
Wang (1992), and Danesh (1998). eral constituents (Figure 1). If compliant minerals are located
 Gassmann Fluid Substitutions 433

at grain contacts, it may be more appropriate to use a simple computed from completely dry samples will be too high. This is
Reuss average when calculating K 0 . because the first few monolayers of water chemically weaken
For a simple rock consisting of two minerals, this can be the rock (Clark et al., 1980). Thus, it is this slightly wet, or
expressed as drained, measurement of K ∗ which should be used when per-
· ¸
F1 F2 −l forming Gassmann fluid substitutions on core data.
K Reuss = + , (9) When working with wireline log data, one common approach
K1 K2 for determining K ∗ is to rewrite equation (1) such that (Zhu
K Voigt = [F1 K 1 + F2 K 2 ], (10) and McMechan, 1990)
1 µ ¶
K vrh = [K Voigt + K Reuss ], (11) φ Ko
2 K sat + 1 − φ − Ko
K fl
where F1 and F2 are the volumetric fractions of the two com- K∗ = . (12)
φ Ko K sat
ponents, and K 1 and K 2 are the bulk moduli of the two com- + −1−φ
ponents. Note that these equations may be readily expanded
K fl Ko
to include additional mineral components. Values for the bulk
Thus, the saturated bulk modulus of the rock (K sat ) for the
moduli of the most common sedimentary minerals are included
in-situ conditions is first calculated using equation (2), and the
in Table 1. See Carmichael (1989) for a more comprehensive
other terms (K fl , K o , and φ) are calculated using the processes
list of seismic rock properties.
described above. For purposes of calculating K ∗ , K fl is the bulk
modulus of the pore-filling fluids under in-situ conditions. Note
Frame properties that uncertainties in porosity, water saturation, fluid properties,
or lithology will lead to errors in the calculated value of K ∗ .
Prior to applying Gassmann’s equation, it is necessary to Since porosity appears explicitly in equations (1) and (12), it
determine the bulk modulus of the porous rock frame, K ∗ . This is tempting to use these equations to model porosity-velocity
is the low-frequency, drained bulk modulus of the rock. Once relationships. Note, however, that the value of K ∗ depends, in
determined, K ∗ is held constant during the course of a fluid part, upon the porosity of the rock. Thus, porosity modeling
substitution. Note that the shear modulus, G [(equation (3)], can only be accomplished if a relationship can be established
is also a frame property of the rock and is therefore also held between the porous frame properties of the rock and poros-
constant during the typical fluid substitution process. ity (c.f., Budiansky and O’Connell, 1976; Murphy et al, 1993).
K ∗ may be derived either from (1) velocity measurements Thus, it is incorrect to use equations (1) and (12) for porosity
on controlled humidity-dried core, (2) application of empiri- modeling without properly adjusting K ∗ .
cal relationships or effective medium theory (Budiansky and
O’Connell, 1976; Gregory, 1976; Murphy et al., 1993; Spencer
CALCULATING VELOCITIES
et al., 1994; Vernik, 1998; Wang, 2000, 2001), or (3) direct calcu-
lation from log data (Zhu and McMechan, 1990). When work- Assume we have calculated porosity (φ) along with the ma-
ing with core data, it is important to recognize that velocities trix and frame properties of the rock (K 0 , K ∗ , and G). This
allows us to use equation (1) to calculate a new saturated bulk
modulus for any desired fluid [the bulk modulus of a fluid can
be calculated using equation (5)].
Next, we must use equations (4) and (7) to calculate a new
bulk density for the rock. Once the new bulk density of the rock

Table 1. Bulk modulus, shear modulus, and bulk density of

common rock forming minerals.∗ Values are from Carmichael
(1989).

Bulk Modulus Shear Modulus Density

Mineral (GPa) (GPa) (g/cm3 )
Quartz 37 44 2.65
“Average” 37.5 15 2.62
feldspar
Plagioclase 75.6 25.6 2.63
feldspar
clay variable variable variable
FIG. 1. Matrix properties are often calculated using a Pyrite 147.4 132.5 4.93
VRH-average, which is simply the average of the Reuss [equa- Hematite 100.2 95.2 5.24
tion (9) and Voigt equation (10)] averages of individual min- Calcite 76.8 32 2.71
eral end-member components. In this diagram, we mix pure Dolomite 94.9 45 2.87
quartz with wet clay. The Voight mix is the “stiff” upper bound, Siderite 123.7 51 3.96
whereas the Reuss mix is the “weak” lower bound. The VRH Anydrite 44.8 29.1 2.98
average is simply the average of the Voigt and Reuss bounds.
∗
The VRH mixing scheme is used to determine the matrix bulk Note that the bulk modulus of clay minerals are highly vari-
modulus of the rock (i.e., the bulk modulus of the minerals with able. For more information on clay properties, see Wang et al.,
no effective porosity present). 2001.
434 Smith et al.

is calculated, we can rewrite equation (2) and calculate V p : rect the log data. If homogeneous saturations are assumed, the
v changes to the sonic data will be small (unless Sxo > ∼90%).
u
u 4 Sonic corrections may be larger if patchy saturation is assumed
u K sat + G
t 3 . (Endres and Knight, 1989; Packwood and Mavko, 1995), or if
Vp = (13)
ρB gas is unevenly distributed among pores of different shapes
(Knight and Nolen-Hoeksema, 1990; Castagna and Hooper,
We also rewrite equation (3) and use the new bulk density to 2000).
calculate the new shear velocity:
s
G Dry frame bulk modulus, K ∗
Vs = . (14)
ρB Application of Gassmann’s equation is dependent upon ac-
curate determination of the porous frame properties of the
Since the shear modulus, G, is held constant during the course
rock (K ∗ and G). In order to calculate the shear modulus, G,
of a fluid substitution, we see from equation (14) that the shear
it is only necessary to know the bulk density and shear veloc-
velocity of a gas sand will be faster than the shear velocity of
ity of the formation. However, in order to calculate the bulk
a brine sand. A summary workflow for Gassmann fluid substi-
modulus of the porous rock frame, certain assumptions must
tutions is included in Table 2.
be made about the matrix and fluid properties (see previous
discussion). Thus, K ∗ is interpretive, in part, and errors in either
CAVEATS AND PITFALLS matrix or fluid properties will lead to errors in the calculated
As with all models, certain assumptions are made during the value for K ∗ . Indeed, in many cases with complex lithology,
fluid substitution process (discussed earlier). Failure to under- K ∗ may be negative (a physical impossibility) or unrealisti-
stand these assumptions, or simple application of the wrong cally large. Sources of anomalous K ∗ values may be when (1)
assumptions, can lead to unrealistic and potentially costly an- porosities or matrix properties are incorrect, (2) fluid proper-
swers. In this section, we review some of the additional pitfalls ties are incorrect (McLean and Alberty, 2001), or (3) initial
and that may lead to incorrect models. assessments of water saturation are incorrect. A technique we
recommend for evaluating K ∗ is to compute the ratio of K ∗ to
G. For clean sands, this ratio is usually close to 1 (Spencer et al.,
Invasion effects
1994; Wang, 2001), whereas for shaley sands it may approach
When processing log data for seismic modeling, it is impor- 2–3. Theoretical models, as well as empirical observations, sup-
tant to evaluate the log response for the effects of invasion port this behavior (Budiansky and O’Connell, 1976; Nur et al.,
(Alberty, 1994, Walls and Carr, 2001). Because water satu- 1995; Wang, 2001) (Figure 2). Note that this approach may not
rations are typically computed from a deep resistivity device work for sands with complex mineralogy or in carbonate rocks.
(which looks several feet into the formation), calculated SW val- Indeed, Ramamoorthy and Murphy (1998) show that the K ∗ /G
ues most likely do not reflect the saturation seen by the sonic ratio can vary substantially in carbonates.
and density logs, which have much shallower depths of investi-
gation. Thus, it may be necessary to apply invasion corrections Patchy saturation
to the sonic and density response, if possible.
Invasion corrections require accurate estimates of water sat- A typical application of Gassmann’s equation assumes that
uration in the invaded zone (Sxo ), as well as accurate estimates all fluid phases are immiscible and homogeneously distributed
of porosity. In addition, knowledge of the mud and forma- throughout the pore space (homogeneous saturation). These
tion fluid properties (K and ρ) is necessary. If these data are assumptions are thought (expected) to be satisfied in systems
available or can be reasonably estimated, it is possible to use which have come to equilibrium over geologic time. However,
equation (4) and Gassmann’s equation [Equation (1)] to cor- this equilibrium distribution of phases may be disturbed dur-
ing drilling, production, and water flooding, and the return to
Table 2. Workflow for application of Gassmann’s equation. equilibrium may require time frames longer than those en-
countered during logging or between seismic surveys used in
1. Log edits and interpretation. 4D monitoring. Thus, it is reasonable to expect that fluids might
2. Shear velocity estimation (if necessary). not be uniformly or homogeneously distributed throughout the
3. Calculate K and G for the in-situ conditions pore space in a reservoir or borehole (Brie et al., 1995). Further-
[equations (2) and (3)].
more, in water-wet rock, water is preferentially drawn into the
4. Calculate K 0 based on lithology estimates [VRH or HS
mixing; equations (9)–(11)]. smaller size pores and cracks, leaving the larger voids or pores
5. Calculate fluid properties (K and ρ). preferentially occupied by the hydrocarbons. This may lead to
6. Mix fluids for the in-situ case according to Sw a segregation of phases and, perhaps more importantly, the in-
[equations (5) and (7)]. ability of pore pressures to equilibrate in the time scale of wave
7. Calculate K ∗ [equation (12)]. propagation. Thus, conditions can and probably do exist where
8. Calculate new fluid properties (K and ρ) at the desired Sw
water and hydrocarbon distribution are not uniform and the
[equations (5) and (7)].
9. Calculate the new saturated bulk modulus of the rock application of the Reuss average for fluid mixture properties
using Gassmann [equation (1)]. is inadequate. This class of nonuniform phase distributions is
10. Calculate the new bulk density [equation (4)]. sometimes referred to as “patchy saturation,” and can be mod-
11. Calculate the new compressional velocity [equation (13)]. eled if the following assumptions are made: (1) the shear mod-
12. Calculate the new shear velocity [equation (14)]. ulus does not vary as a function of pore fluid; (2) the fluids are
 Gassmann Fluid Substitutions 435

spatially distributed in aggregates whose dimensions are much form saturation into the Gassmann modeling allows us to con-
smaller than the seismic wavelength; and (3) the different pore tinue to ignore the details of the pore geometry. However, mi-
fluid “patches” have dimensions which do not allow fluid pres- crostructural details can exist which lead to preferential fluid
sure equalization during the time scale of wave propagation. segregation that can not be adequately modeled using either
Hill (1963) and Berryman and Milton (1991) have developed simple Gassmann or patchy saturation (e.g. Endres and Knight,
the formalism for estimating the effective bulk modulus under 1989; Castagna and Hooper, 2000).
these conditions: Figure 3 shows the effect of patchy (nonhomogeneous) satu-
 −1 ration on a high-porosity (30%) sand and a lower-porosity sand
Xn  (19%); both examples are from wireline log data. The percent
xi 4
K eff = 
 µ ¶
 − G, (15) difference between patchy and homogeneous saturation for
4 3 both scenarios is shown in Figure 3c. For the lower-porosity
i=1 Ki + G
3 sand case, the difference between patchy and homogeneous
saturation is less than 5%, whereas the difference approaches
where n = the number of patches with different fluid content, 12% for the higher-porosity sand case (when Sw approaches
xi = the volume fraction of the ith patch, G = the shear modu- ∼80%). Clearly, it becomes important to consider the potential
lus of the rock, K i = the bulk modulus of the rock completely effects of patchy saturation for high-porosity sands (Figure 3b),
saturated with the ith fluid, and K eff = the effective bulk mod- although it will be most problematic for the case of low gas sat-
ulus of the rock. uration (Sw ≈ 80%). Thus, consideration of nonuniform versus
Importantly, Hill (1963) demonstrated that equation (15) homogeneous saturation may become particularly important
holds for general aggregates, where the shear modulus is the when applying invasion corrections to wireline log data and in
same for all constituents. In addition, the Backus (1962) av- modeling time-lapse seismic responses.
erage for isotropic layers having a constant shear modulus re-
duces to equation (15). Note that nonuniform saturation will Internal consistency and saturation modeling
have no effect on the calculated shear velocity, because the
shear modulus of a rock is independent of the pore-filling fluid Fluid substitution models should always be carefully eval-
under the conditions discussed above. Incorporating nonuni- uated for quality and internal consistency. For instance, when

FIG. 2. Model-based and empirical relationships between the shear modulus, G, and the porous frame bulk
modulus, K ∗ . The self-consistent models of Budiansky and O’Connell (1976) and the critical porosity models
of Nur et al. (1995) yield similar results and suggest that K ∗ /G should be greater than 1 for shaley sands. The
empirical result of Wang (2001) is also included for comparison. His data show that the K ∗ /G ratio should be
approximately equal to 1. Note that for shaley sands and low porosity sands, many of the assumptions discussed
in this tutorial are violated.
436 Smith et al.

oil or gas is substituted into a brine sand, bulk densities should To aid as a quality-control check on fluid substitutions and
always decrease and shear velocities should always increase in order to best understand how velocities and densities should
(assuming no adjustments were made to K ∗ , G, or porosity). vary for any given sand, it is often instructive to generate
If the expected changes in Vs or density are not observed to saturation models, where Gassmann’s relationships are used
occur or if the magnitude of change is not what is expected, to calculate V p , Vs , and ρ B as a function of water saturation
errors must exist in one or more of the input parameters. For (Figure 4). These types of models are particularly important
density and shear velocity, these errors will most likely be in for understanding the rare case when compressional velocities
grain density or porosity. for a gas sand are faster than those for oil in the same sand
(Figure 4b). This situation can occur when the bulk density de-
creases at a faster rate than the saturated bulk modulus as gas
saturation is increased.

CONVENTIONAL APPLICATION

Conventional application of Gassmann’s equation provides

the interpreter with a powerful tool for evaluating the various
fluid scenarios which might give rise to an AVO anomaly (e.g.,
Sbar, 2000). In this paper, we use velocities and densities from
a gas sand in a deepwater Gulf of Mexico well to illustrate
conventional application of Gassmann’s equation (Figure 5).
Using the processes described in this paper, we replace the

FIG. 3. Homogeneous versus nonhomogeneous saturation for

a gas-brine system. (a) High-porosity sand, (b) low-porosity
sand, and (c) percent difference plot. Both models are de- FIG. 4. Saturation models generated using Gassmann’s equa-
rived from wireline log data. Note from (c) that patchy sat- tion. (a) High-porosity sand from the deepwater Gulf of
uration may be particularly problematic at low gas saturations Mexico, (b) low-porosity sand from South America. Note the
(Sw ≈ 0.80). Patchy saturation [or other dispersion mecha- small differences between the gas and oil case at low water
nisms (Brie et al., 1995; Castagna and Hooper, 2000)] may also saturations (Sw < 0.30). Also note from (b) that the gas sand
be important to consider when correcting wireline sonic logs velocity at low values of Sw is actually faster than the oil sand
for invasion. velocity.
 Gassmann Fluid Substitutions 437

in-situ gas with brine (Figure 5). Inferred fluid properties are distinct from the gas response. In the analysis of actual seis-
included in Figure 5. mic data, it is this deviation from the brine and shale “back-
Large differences are observed in the density and compres- ground” trend which will allow us to identify gas or oil anoma-
sional velocity response between gas and brine, whereas much lies (see Castagna and Swan, 1997). Construction of this type
smaller differences are observed in shear velocity. Note, how- of crossplot from petrophysical data also allows us to quantify
ever, that the shear velocity for gas is faster than the shear ve- the difference between different fluid scenarios. It is also com-
locity for brine. Examination of equation (14) shows that this mon to replace Poisson’s ratio in these crossplots with elas-
behavior is to be expected. Also note from this figure that the tic or shear impedance (Connolly, 1999; Vernik and Fisher,
calculated porous frame bulk modulus [K ∗ , calculated using 2001).
Equation (12)] is nearly equal in value to the shear modulus. Simple interface models are also constructed for this sand
This is to be expected for clean quartz sands. (Figure 7) using Shuey’s three-term approximation (Shuey,
In Figure 6, we cross-plot acoustic impedance and Poisson’s 1985). Notice that the in-situ gas case generates a positive
ratio. We note from this that the brine sands and shale are class III AVO response (negative gradient), whereas the brine

FIG. 5. Gas sand from a deepwater Gulf of Mexico well (water depth is approximately 1000 m). In relatively
high-porosity rocks such as these, there is a large contrast between the gas and brine response. Also included in
this plot are curves showing the shear modulus, G, and the dry frame bulk modulus, K ∗ . Note that K ∗ is nearly
equal to G for sands with low clay contents, whereas it may be much higher than G for shaley sands.
438 Smith et al.

case generates a negative AVO response (positive gradient). which might give rise to an observed seismic anomaly. We
In both cases, large differences in intercept are observed. have attempted in this tutorial to provide the reader with a
The interface models shown in Figure 7 are derived from in- practical understanding of the “nuts-and-bolts” behind con-
terval averages calculated from the log data, and therefore rep- ventional application of Gassmann. However, this technology
resent a single point reflection with infinite half-spaces above is often perceived as being “black box,” and is frequently ap-
and below the reflector. This allows us to develop an under- plied via commercial software packages without much knowl-
standing of the AVO response these rocks are capable of gen- edge of how the calculations are being made or where the
erating in the absence of complications resulting from sand potential pitfalls may occur. In particular, invasion correc-
thickness and geometry, seismic frequency, and the shape of tions and proper determination of the porous frame properties
the seismic wavelet. Unfortunately, the response calculated (in particular, K ∗ ) are critical for generating internally con-
from these simple interface models is not always representa- sistent and meaningful results. Failure to be aware of these
tive of the observed response. Thus, we must scale up our data problems can, and frequently does, lead to models which are
to the seismic bandwidth for a direct comparison to seismic wrong. Once reliable models are obtained, the user can quan-
data. tify the differences between the hydrocarbon and brine case
Figure 8 is a comparison of the offset synthetic response for by crossplot analysis, interface modeling, and offset synthetic
both gas and brine. For these models, a simple bandpass filter analysis.
was applied with a frequency spectrum of 5–10–40–50 HZ. Note
that the AVO response is still observed at the seismic band-
width. Clearly, AVO (and other AVO-based attributes) could ACKNOWLEDGMENTS
be used to help differentiate gas and brine for this particular
The authors thank Bruce Wagner (BP Americas), Yafei Wu
example. Caution should always be used, however, when com-
(VeritasDGC), and Dave Went (VeritasDGC) for early and
paring these types of models to the actual seismic gathers, as
critical reviews of the original manuscript. In addition, the qual-
improper processing of the seismic data can significantly alter
ity of the manuscript was greatly improved by the comments
the amplitude response of the data.
from John Castagna, Yue Feng Sun, Enders A. Robinson,
and two anonymous reviewers. The authors are particularly
CONCLUSIONS indebted to Keith Katahara (BP Americas) for several re-
views of this manuscript and for numerous stimulating and
Gassmann’s equations provide the seismic interpreter with enlightening discussions on fluid substitutions and seismic rock
a powerful framework for evaluating various fluid scenarios properties. Editorial support was provided by Harold Cansler
(Petrophysical Solution) and Mandy Jones and Ken Williams
(VeritasDGC).

FIG. 6. Acoustic impedance versus Poisson ratio crossplot of

the well data shown in Figure 5. Brine sands and shale can
be used to define the “background” trend against which gas FIG. 7. Interface models [Shuey’s three-term approximation
anomalies may be quantified (Castagna and Swan, 1997). The (Shuey, 1985)] for the sand shown in Figure 5. Note that large
magnitude of deviation from the shale and brine sand “back- differences in intercept and gradient are observed between the
ground” will be variable, and is largely dependent on rock type gas and brine cases. The in-situ response is for gas, and is calcu-
and quality. For low-porosity sands, small differences are typ- lated using the measured velocities and density from wireline
ically observed between the “background” trend and the gas log data. The brine response is calculated using Gassmann’s
response. equation. RHOB = bulk density.
 Gassmann Fluid Substitutions 439

FIG. 8. Offset synthetics for gas and brine over the interval shown in Figure 5. Both models use the same color
scale. These models show that even at the seismic bandwidth (5–10–40–50 HZ), a strong AVO response is
generated for gas.
REFERENCES Louvain, 351–356.
Biot, M. A., 1956, Theory of propagation of elastic waves in a fluid satu-
Alberty, M., 1994, The influence of the borehole environment upon rated porous solid. I. Low frequency range and II. Higher-frequency
compressional sonic logs. Trans. Soc. Prof. Well Log Analysts 35th range: Acoust. Soc. Am., 28, 168–191.
Ann. Logging Symp., paper 5. ———1962, Mechanisms of deformation and acoustic propagation in
Backus, G. E., 1962, Long-wave elastic anisotropy produced by hori- porous media: J. Appl. Phys., 33, 1482–1498.
zontal layering: J. Geophys. Res., 67, 4427–4440. Brie, A., Pampuri, F., Marssala A. F., and Meazza, O., 1995, Shear sonic
Batzle, M., and Wang, Z., 1992, Seismic properties of pore fluids: Geo- interpretation in gas-bearing sands; SPE Ann. Tech. Conf., 701–
physics, 57, 1396–1408. 710.
Berryman, J. G., 1999, Origin of Gassmann’s equations: Geophysics, Brown, R., and Korringa, J., 1975, On the dependence of the elastic
64, 1627–1629. properties of a porous rock on the compressibility of the pore fluid:
Berryman, J. G., and Milton, G. W., 1991, Exact results for general- Geophysics, 40, 608–616.
ized Gassmann’s equation in composite porous media with two con- Budiansky, B., and O’Connell, R. J., 1976, Elastic moduli of a cracked
stituents: Geophysics, 56, 1950–1960. solid: Internat. J. Solids and Structures, 12, 81–97.
Berryman, J. G., and Wang, H. F., 2000, Elastic wave propagation and Carmichael, R. S., 1989, Practical handbook of physical properties of
attenuation in a double-porosity dual-permeability medium: Inter- rocks and minerals: CRC Press.
nat. J. Rock Mech. and Min. Sci., 37, 63–78. Castagna, J. P., Batzle, M. L., and Kan, T. K., 1993, Rock physics-the
Berge, P. A., 1998, Pore compressibility in rocks, in Thimus, J.-F., link between rock properties and AVO response, in Castagna, J. P.
Abousleiman, Y., Cheng, A.H.-D., Coussy, O., and Detournay, E., and Backus, M. M., Eds., Offset-dependent reflectivity—theory and
Eds., Biot Conference on Poromechanics: Université Catholique de practice of AVO analysis,: Soc. Expl. Geophys., 131–171.
440 Smith et al.

Castagna, J., and Hooper, J., 2000, A simple method for fitting P- saturated porous rocks, in Johnson, D. L., and Sen, P. N., Eds., Physics
wave velocity versus saturation curves: 70th Ann. Internat. Mtg., and chemistry of porous media: Amer. Inst. Physics, 166–175.
Soc. Expl. Geophys., Expanded Abstracts, 1887–1890. Packwood, J. L., and Mavko, G., 1995, Seismic signatures of multi-
Castagna, J. P., and Swan, H. W., 1997, Principles of AVO crossplotting: phase reservoir fluid distribution: Application to reservoir moni-
The Leading Edge, 4, 337–342. toring: 65th Ann. Internat. Mtg., Soc. Expl. Geophys., Expanded
Clark, V. A., Tittman, B. R., and Spencer, T. W., 1980, Effects of volatiles Abstracts, 910–913.
on attenuation and velocity in sedimentary rocks: J. Geophys. Res., Ramamoorthy, R., and Murphy, W. F., 1998, Fluid identification
85, 5190–5198. through dynamic modulus decomposition in carbonate reservoirs:
Connolly, P., 1999, Elastic impedance: The Leading Edge, 18, 438–452. Trans. Soc. Prof. Well Log Analysts 39th Ann. Logging Symp.,
Danesh, A., 1998, PVT and phase behaviour of petroleum reservoir paper Q.
fluids: Elsevier. Raymer, L. L, Hunt, E. R., and Gardner, J. S., 1980, An improved
Eberhart-Phillips, D., Han, D-H., and Zoback, M. D., 1989, Empiri- sonic transit time-to-porosity transform: Trans. Soc. Prof. Well Log
cal relationships among seismic velocity, effective pressure, porosity, Analysts 21st Ann. Logging Symp.
and clay content in sandstone: Geophysics, 54, 82–89. Sbar, M. L., 2000, Exploration risk reduction: An AVO analysis in
Endres, A. L., and Knight, R., 1989, The effect of microscopic fluid the offshore middle Miocene, central Gulf of Mexico: The Leading
distribution on elastic wave velocities: The Log Analyst, Nov-Dec, Edge, 19, 21–27.
437–444. Shuey, R. T., 1985, A simplification of the Zoeppritz equations: Geo-
Gassmann, F., 1951, Uber die Elastizität Poröser Medien: Vier. der physics, 50, 609–614.
Natur. Gesellschaft in Zürich, 96, 1–23. Sondergeld, C. H., and Rai, C. S., 1993, A new exploration tool: Quan-
Gregory, A. R., 1976, Fluid saturation effects on dynamic elastic prop- titative core characterization: Pageoph, 141, 249–268.
erties of sedimentary rocks: Geophysics, 41, 895–921. Spencer, J. W., Cates M. E., and Thompson, D. D., 1994, Frame moduli
Han, D-H., Nur, A., and Morgan, D., 1986, Effects of porosity and of unconsolidated sands and sandstones: Geophysics, 59, 1352–1361.
clay content on wave velocities in sandstones: Geophysics, 51, 2093– Vernik, L., 1998, Acoustic velocity and porosity systematics in silici-
2107. clastics: The Log Analyst, July-Aug., 27–35.
Hashin, Z., and Shtrikman, S., 1962, A variational approach to the Vernik, L., and Fisher, D., 2001, Estimation of net-to-gross from P and
elastic behavior of multiphase materials: J. Mech. Phys. Solids, 11, S impedance: Part I—Petrophysics: 71st Ann. Internat. Mtg: Soc.
127–140. Expl. Geophys., Expanded Abstracts, 207–210.
Hill, R., 1963, Elastic properties of reinforced solids: Some theoretical Walls, J. D., and Carr, M. S., 2001, The use of fluid substitution modeling
principles: J. Mech. Phys. Solids, 11, 357–372. for correction of mud filtrate invasion in sandstone reservoirs: 71st
Knight, R., and Nolen-Hoeksema, R., 1990, A laboratory study of the Ann. Internat. Mtg., Soc. Expl. Geophys., Expanded Abstracts, 385–
dependence of elastic wave velocities on pore scale fluid distribu- 387.
tions: Geophys. Res. Lett., 17, 1529–1532. Wang, A., 2000, Velocity-density relationships in sedimentary rocks,
Mavko, G., Mukerji, T., and Dvorkin, J., 1998, The Rock physics hand- in Wang, Z., and Nur, A., Eds., Seismic and acoustic velocities in
book: Tools for seismic analysis in porous media: Cambridge Univ. reservoir rocks, 3: Recent developments: Soc. Expl. Geophys, 8–23.
Press. Wang, Z., 2001, Fundamentals of seismic rock physics: Geophysics, 66,
McCain, W. D., 1990, The Properties of petroleum fluids: PennWell 398–412.
Books. Wang, Z., Wang, H., and Cates, M. E., 2001, Effective elastic properties
McLean, M., and Alberty, M, 2001, Dissolved gas in brine: Its impact of clays: Geophysics, 66, 428–440.
on fluid substitution: Trans. Soc. Prof. Well Log Analysts 42nd Ann. Wyllie, M. R. J., Gergory, A. R., and Gardner, L. W., 1956, Elastic wave
Logging Symp., paper QQ. velocities in heterogeneous and porous media: Geophysics, 23, 41–
Murphy, W., Reischer A., and Hsu, K., 1993, Modulus decomposition 70.
of compressional and shear velocities in sand bodies: Geophysics, Wyllie, M. R. J., Gergory, A. R., and Gardner, L. W., 1958, An exper-
58, 227–239. imental investigation of factors affecting elastic wave velocities in
Nur, A., Mavko, G., Dvorkin, J., and Gal, D., 1995, Critical poros- porous media: Geophysics, 23, 459–493.
ity: The key to relating physical properties to porosity in rocks: Zhu, X., and McMechan, G. A., 1990, Direct estimation of the bulk
65th Ann. Internat. Mtg., Soc. Expl. Geophys., Expanded Abstracts, modulus of the frame in fluid saturated elastic medium by Biot
878. theory: 60th Ann. Internat. Mtg., Soc. Expl. Geophys., Expanded
O’Connell, R. J., 1984, A viscoelastic model of anelasticity of fluid Abstracts., 787–790.