You are on page 1of 10

GEOPHYSICS, VOL. 79, NO. 6 (NOVEMBER-DECEMBER 2014); P. D389–D398, 10 FIGS., 1 TABLE.

10.1190/GEO2014-0089.1

Estimating the influence of stress state on compaction behavior

Matt R. Hauser1, Brent A. Couzens-Schultz1, and Alvin W. Chan2

neering and operational decisions that are critical to safe, successful,


ABSTRACT and economic well design and execution as well as providing key
inputs to exploration and development. A variety of different rela-
Relationships between the compaction state and effective tionships have been developed to relate effective stress to porosity
stresses are the basis for most quantitative pore-pressure estimators (e.g., Rubey and Hubbert, 1959; Hottman and Johnson,
and stress estimates. Common practice uses only a single 1965; Eaton, 1975; Gutierrez et al., 2006), and they have been used
element of the stress tensor, the vertical stress, for these cal- successfully for many years. However, as industry moves to increas-
culations; mean stress formulations also exist, although they ingly complex environments, the applicability of these transforms
are less widely applied. Using simple models and field data may be challenged.
from two distinct stress regimes, we examined the validity It is well established that the full stress tensor can be important to
and limitations of the vertical-stress approach as well as a understanding mechanical compaction, and that both normal and
mean-stress approach, showing that in complex stress set- shear stresses may have a significant effect (e.g., Roscoe et al.,
tings, both can perform very poorly. We evaluated a method 1958; Wong et al., 1997). Either explicitly or implicitly, though,
for incorporating shear stresses into compaction relations by standard models tie compaction solely to vertical stress. Most of
using state boundary surface (SBS) formulations from soil the work developing and calibrating compaction-stress transforms
mechanics and demonstrated how the resulting model may has been carried out in areas with comparatively simple stress con-
be calibrated and applied to field data. This approach was ditions such as within extensional basins. In such settings, where
found to perform much better in the complex stress environ-
other stress components are well correlated to the vertical stress,
ment, providing more stable calibration behavior and more
a transform based solely on vertical stress is expected to perform
reliably extrapolating to stress states beyond those present
well. In more complex settings, calibrations based on vertical stress
in the calibration data. Although vertical and mean stress
are, at best, confined to subregions where stress conditions might be
compaction models may work well in simple stress environ-
expected to be somewhat consistent. However, if nonvertical
ments, we discovered that incorporation of shear stress is
stresses become decoupled from the vertical stress, such forms
necessary for models in complex stress settings. Although
are likely to perform less satisfactorily.
the addition of shear stress significantly improves agreement
Extension of existing compaction models to 3D mean stress has
with field data, it also increases the complexity of the model
been proposed by some authors (e.g., Harrold et al., 1999; Goulty,
as well as the requirements for calibration data. We therefore
evaluated the settings in which each of these three ap- 2004; Katahara, 2005), but application of mean stress is not
proaches — vertical stress, mean stress, and SBS — may be common. The importance of shear stress has also been noted
most appropriate. (e.g., Bowers, 2002; Nikolinakou and Chan, 2012), although to
our knowledge, no practical application of these components has
been made in the literature. Although not typically used in pressure
and stress prediction, various models do exist to better incorporate
INTRODUCTION the full stress tensor in compaction calculations, as well as finite
element or finite difference engines with which to apply them. How-
The relationship between the compaction state and effective ever, such computations depend on necessary prerequisites.
stress is central to most quantitative estimates of stress and pore First, it is prudent to validate the models against field data —
fluid pressure. These quantities, in turn, provide a basis for engi- much of which falls in more extreme stress ranges than those in

Manuscript received by the Editor 22 February 2014; revised manuscript received 11 June 2014; published online 16 September 2014.
1
Shell Exploration & Production Company, Houston, Texas, USA. E-mail: matt.hauser@shell.com; brent.couzens@shell.com.
2
Shell International Exploration & Production Inc., New Orleans, Louisiana, USA. E-mail: alvin.w.chan@shell.com.
© 2014 Society of Exploration Geophysicists. All rights reserved.

D389
D390 Hauser et al.

which the models were derived. Second, the necessary material SOIL MECHANICS FRAMEWORK
parameters for the models may be unavailable or may be very dif-
ficult or costly to obtain for many areas of interest; a viable meth- It is commonly observed in soil compaction measurements that
odology would ideally be able to self-calibrate or to extrapolate in a the relationship between sediment compaction (e) and maximum
stable fashion in these situations. Finally, for many applications re- MES p 0 is nearly linear in e- ln p 0 space, as diagramed in Figure 1
lated to well design and execution, it is advantageous to be able to (e.g., Fjaer et al., 2008). The degree of compaction is represented
make very rapid good-quality estimates of compaction behavior; here by the void ratio, e:
although finite-element or finite-difference calculations can provide
ϕ
robust solutions, they may often be ill suited for real-time opera- e¼ ; (1)
tional applications needing calculations that can be completed in 1−ϕ
a few minutes.
where ϕ is the sediment porosity. For sediments undergoing iso-
To address these issues, this paper examines the potential for
tropic compaction where all principal stresses are equal, the void
contributions due to the full stress tensor using a soil mechanics
ratio for a given maximum MES, termed the preconsolidation stress
framework. We will begin by reviewing relevant assumptions and
pc0 , is given by
developing some simple relations with which we can evaluate the
impact of nonvertical stresses. We intentionally choose very basic e ¼ e0 − λ ln pc0 . (2)
formulations to allow analytic solutions that can be used to quickly
illustrate and evaluate the basic hypotheses. More complex soil This relation is represented by the compaction line (solid black)
mechanics and geomechanical models could be incorporated using in Figure 1 with slope λ. Sediments at stresses less than their pre-
a similar workflow, though they would be subject to the same con- consolidation stress rebound slightly, again in a nearly linear way,
ditions mentioned above. along the so-called swelling line (gray) with slope κ as shown in the
Once the basic framework is established, we compare typical figure. The void ratio along this swelling line is
models based on vertical effective stress (VES) or mean effective
stress (MES) with a new model that incorporates shear stress e ¼ e0 − λ ln pc0 þ κðln pc0 − ln p 0 Þ; (3)
through a state boundary surface (SBS) formulation. These models
will be evaluated using two sets of field data. First, we will consider where pc0 is the maximum MES reached by the sediments and p 0 is
a regional data set from the Gulf of Mexico, which is thought to the current stress. These equations display the functional form used
have a simple extensional stress setting. We will then consider a by Bowers for his widely used effective-stress transform (Bowers,
data set from a fold belt near Borneo, in which typical vertical-stress 1995), although that form is typically cast in terms of VES rather
compaction models break down, evaluating both mean-stress and than mean stress.
shear-compaction models. In soil mechanics, a nonisotropic stress state is often described by
the ratio of the generalized shear stress q to the mean stress p 0 de-
fined as

1
p 0 ¼ ðσ 1 þ σ 2 þ σ 3 Þ;
3
pffiffiffiqffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
q ¼ 1∕ 2 ðσ 1 − σ 2 Þ2 þ ðσ 1 − σ 3 Þ2 þ ðσ 2 − σ 3 Þ2 ;
q
K ¼ 0; (4)
p

where σ i are the principal effective stresses. Although defining the


stress state with only two parameters representing generalized nor-
mal and shear stresses is still a substantial simplification of the full
stress tensor, this is found to be an adequate parameterization for
modeling a wide range of laboratory and field observations in soil
and rock mechanics (e.g., Fjaer et al., 2008).
The ratio K ¼ q∕p 0 is zero for the isotropic case discussed
above. Nonisotropic compaction paths that still maintain constant
stress ratios, having a fixed K > 0, will follow lines in the
e- ln p 0 plot that are parallel to the K ¼ 0 line but lie to its left
(the black dotted line in Figure 1). More complex stress paths would
follow compaction curves also to the left of the K ¼ 0 line, but they
would generally be nonparallel and possibly curvilinear.
So long as the stresses in a region maintain a near-constant K, the
Figure 1. Typical soil mechanics compaction-effective stress rela- compaction trend is expected to be parallel to the K ¼ 0 path and
tions. Compaction along stress paths with constant stress ratio (K) is
assumed proportional to ln p 0 , where p 0 is the MES. Loading the equations above should still provide a usable relation between
moves along lines with slope λ, whereas unloading or swelling stress and compaction. This requirement is expected to be reason-
moves along lines of slope κ. ably satisfied in many extensional basin settings and even in more
Stress path and compaction D391

complex environments in which the stress range under considera- Cam-Clay form in which the yield surface is an ellipse defined
tion for both calibration and prediction is sufficiently narrow. In by the expression
these situations it is only necessary to change e0 in equations 2
and 3, an adjustment which happens as a matter of course in stan- M2
dard calibration workflows for empirical or semiempirical trans- p0 ¼ pc0 : (7)
M2 þ K2
forms. A difficulty for this process arises, though, if the stresses
of the region under study do not conform to a constant-K stress This surface is constructed such that the apex of the ellipse in the
path or if the region with calibration data has a different K-value p 0 -q plane lies on the critical state line and its intersection with the
than the region to be analyzed. K ¼ 0 line (the p 0 axis) falls at pc0 ; thus, the slope of the critical
To begin addressing this issue, we observe in Figure 1 that the state line M is the only additional parameter needed to define the
swelling line provides a relationship between the compaction states shape of the yield surface (Roscoe and Burland, 1968). Combining
reached along a line with constant K > 0 and states reached along equations 5 and 7, we can write
the K ¼ 0 line. In the example shown in the figure, sediments
loaded along the K > 0 path to p 0 will have the same void ratio  ðκ∕λ−1Þ
M2 þ K 2
as sediments loaded to pc0 along a K ¼ 0 line and then unloaded p0 ¼ p00 . (8)
to p 0 . The same void ratio will also be obtained by loading along M2
K ¼ 0 to p00 and stopping. Because the paths give the same void
ratio e, the right side of equation 2 (for K ¼ 0 compaction to Equation 8 gives a relationship between an arbitrary stress state
p00 ) must be equal to the right side of equation 3 (for K ¼ 0 com- p 0 ; q (where q ¼ K · p 0 ) and the K ¼ 0 stress state (p00 ; 0) with
paction to pc0 followed by swelling to p 0 ); by setting these two ex- the same compaction state or porosity. The path traced by a constant
pressions equal, we can solve for a relationship between the three porosity with varying values of K is known as the SBS; this surface
stress states: is dependent on the yield surface but will in general fall slightly to
the right of the yield surface in p 0 -q space as illustrated in Figure 2.
0ð1−κ∕λÞ From equation 2, we can find p00 (K ¼ 0) given a porosity esti-
p00 ¼ p 0κ∕λ pc . (5)
mate from data such as density, velocity, or resistivity
We next examine the relationship between these states in the p 0 -q
plane as shown in Figure 2. Stress paths for the two K ¼ q∕p 0 ratios p00 ¼ eðe0 −eÞ∕λ ; (9)
from Figure 1 are again illustrated with K ¼ 0 along the horizontal
axis and the arbitrary K > 0 path; two schematic yield surfaces, and using equation 8, we can then write
YS1 and YS2, are also shown. YS1 corresponds to a K ¼ 0 pre-
consolidation stress of p00 and YS2 corresponds to a K ¼ 0 precon-  ðκ∕λ−1Þ
M2 þ K 2
solidation stress of pc0 . During a monotonic increase in stress p0 ¼ eðe0 −eÞ∕λ ; (10)
(loading), the yield surface will move as the sediment compacts
M2
such that for a given p 0 ; q state, the sediment is always on its yield
surface. In this example, a sediment compacting on the K > 0 line where e0 and λ describe K ¼ 0 compaction and M and κ are fit to
at (p 0 ; q) will have YS2, the same surface obtained by (pc0 ; 0) along the calibration data. With this equation, we can in principle use a
the K ¼ 0 path. K ¼ 0 compaction law to estimate stress or compaction along ar-
For any given stress ratio K, the intersection of the constant-K bitrary stress paths. So long as we know the stress state of the cal-
line with the yield surface will define a unique nonzero stress state ibration data with reasonable accuracy, we do not even need K ¼ 0
(p 0 ; q). If we assume that the locus of the yield surface is completely data to calibrate our compaction trend; in fact, the more spread in K
determined by its K ¼ 0 preconsolidation stress pc0 , then for any pc0
and K the corresponding MES is uniquely determined:

p 0 ¼ gðpc0 ; KÞ. (6)

In principle, equations 5 and 6 provide the relationships neces-


sary to relate any two of these three stresses to one another. This
is precisely what we need to use stress-compaction data from
one stress state to estimate the relationship in another stress
state. For example, if we have a transform such as equation 2
for a K ¼ 0 state, then a given compaction e might imply a
K ¼ 0 effective stress of p00 as shown in Figure 1. If we know
the actual stress has K > 0, then the same e would actually corre-
spond to an MES p 0 as shown in that figure. Equations 5 and 6 give
us the relationship between p00 inferred from our calibrated K ¼ 0 Figure 2. Stress states plotted in mean stress (p 0 ) — shear stress (q)
transform and the actual p 0 we need to estimate in situ stress and space. The K ¼ 0 and K > 0 lines shown here correspond to those
pore fluid pressure. plotted in e- ln p 0 space in Figure 1. Notional yield surfaces cor-
responding to K ¼ 0 preconsolidation stresses p00 and pc0 are also
In general, this relationship may not be solvable analytically. For shown as YS1 and YS2, respectively. A representative constant-
some simple yield surface formulations, however, we can write a porosity line, known as the SBS, is shown for the preconsolidation
closed-form solution. For this paper, we will use the modified stress p00 .
D392 Hauser et al.

we have in the calibration data, the better defined the M and κ To that end, we begin again with the poroelastic equations that
parameters will be. form the basis of the 1D work:
The remaining issue is the determination of K. In the most gen-  
eral case, this can be an extremely complicated undertaking that is 1 1 1 α
εij ¼ S − S δ þ S δ − Pδ ; (12)
quite difficult to calibrate or validate. As in the above analysis, 2μ ij 3 mm ij 9B mm ij 3B ij
though, we will appeal to simple models to allow evaluation of
the hypothesis and leave more complex analyses to future work. with total stress components Sij , pore-fluid pressure P, shear and
bulk moduli μ, B, and Biot coefficient α. The boundary conditions
are similar to those leading to equation 11, but lateral compaction is
ESTIMATION OF THE STRESS STATE incorporated as an extrinsic strain, T, along one horizontal axis.
For the two areas we will examine in this paper, we begin With S33 ¼ Sv , S22 ¼ SH ¼ maximum horizontal stress, S11 ¼
by assuming the principal stresses are horizontal and vertical. Sh ¼ minimum horizontal stress; these conditions are
Although it is straightforward to derive solutions for a more general ε11 ¼ 0 ε22 ¼ T εij ¼ 0; i ≠ j
stress state when needed, the geometry of the sediments in the areas (13)
S33 ¼ Sv Sij ¼ 0; i ≠ j.
of interest here suggests a grossly vertical-horizontal alignment. In
keeping with our goal of seeking the simplest practical solutions For application to field data we will define Sv solely as the in-
with which to evaluate our hypotheses, we therefore use the alge- tegral of the overburden load. In actuality, the lateral compression of
braically simpler horizontal-vertical case. the section would result in arching and stress redistributions much
In many environments, the ratio of horizontal-to-vertical stresses more complex than those captured in these equations. However, we
estimated from field data is found to follow relatively simple trends will assume for this work that such modifications are second order
as shown for instance by Hubbert and Willis (1957), Mathews and in terms of their impact on the compaction modeling. As discussed
Kelly (1967), or Breckels and van Eekelen (1981) (it should be em- in the Introduction, this is done in an attempt to make a clearer il-
phasized that the “k” used by these authors is not the stress ratio lustration of the concepts with the full expectation that more accu-
described above, but rather it is the ratio between the vertical and rate models may be needed for many applications.
horizontal stress). In the case of the extensional Gulf of Mexico data With these boundary conditions and constitutive equations, and
set used in the next section, we specifically find that the model that taking α ¼ 1, straightforward algebraic manipulation yields the fol-
best fits data from the study wells is a 1D poroelastic model of the lowing relations for the effective stresses:
type proposed by Eaton (1969), with the horizontal-to-vertical
stress ratio given by ν∕ð1 − νÞ, and using local log-based estimates ν 2μν
σh ¼ σ þ T;
of Poisson’s ratio (ν). Although there are limitations in the physical 1−ν v 1−ν
assumptions leading to this model, it agrees well with available data ν 2μ
σH ¼ σv þ T. (14)
on stress state, such as casing shoe leak-off pressures, lost-circula- 1−ν 1−ν
tion events, etc. We thus take it as a practical basis for stress esti-
mates in our passive settings, in keeping with the goal of simple In the p 0 -q domain used for the “Soil mechanics framework” sec-
forms with which to test the compaction models. tion, this yields
Given an assumption of vertical and horizontal principal stresses, 1þν
and further assuming horizontal stress isotropy to illustrate our p0 ¼ ðσ þ 2μTÞ;
workflow, we can combine this local stress ratio formalism with
3ð1 − νÞ v
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
the conventional definitions of generalized mean and shear stress 1
q¼ ð1 − 2νÞ2 σ 2v − ð1 þ νÞð1 − 2νÞ2μTσ v þ 4μ2 T 2 .
shown in equation 4 to yield simple relations for the stress param- 1−ν
eters needed for compaction modeling: (15)
1þν
p0 ¼ σ ;
3ð1 − νÞ v
1 − 2ν APPLICATION IN A SIMPLE
q¼ σ ; EXTENSIONAL SETTING
1−ν v
3ð1 − 2νÞ To illustrate the application of this framework, we begin in a sim-
K¼ . (11)
1þν ple extensional setting in which existing methods already provide
good-quality relationships between stress and compaction. The data
The situation is somewhat more complicated in the second study used here are from a set of 51 wells located in the deepwater Gulf of
area, which is comprised of a fold belt with strike-slip and reverse- Mexico, as shown in Figure 3. An analysis of the compaction
fault conditions overlain by drape sediments, which appear to be in behavior in these wells using a standard VES approach is presented
normal-fault stress conditions. The simple 1D poroelastic model by Hauser et al. (2013). Within this set, there were pressure mea-
applied above may still be usable as a first-order estimate in the surements in 405 distinct sand units with associated basic petro-
drape sediments but is clearly inadequate for deeper strike-slip and physical data in the 10 ft of shale immediately overlying each
reverse-faulting intervals. We nevertheless desire a similarly tract- sand as described in that paper.
able stress model to reach our goal of simple solutions to the com- For each pressure point in the data set, the porosity ϕ and the
paction equations that can be used to understand and evaluate the Poisson ratio ν of the associated bounding shale were calcu-
hypothesis. lated from the petrophysical data. Density data were available
Stress path and compaction D393

for approximately 90% of the points to determine porosity and good performance here. The resulting fit gives similar residual error
were used to construct a density-velocity transform, which allowed in effective-stress space to the free fit.
estimation of porosity for the remaining units. Compressional To prioritize a good fit in the e- ln p00 space, we can use equation 2
velocities were available for all bounding shales, and a regional to provide a relationship between λ and e0 . Each stress-void ratio
VP ∕VS trend, validated with available shear velocities from this data point in the data set (ei , p0i0 ) implies a specific λ − e0 relation:
set, was used to provide the remaining data necessary to compute ν.
0 .
e0;i ¼ ei þ λi ln p0;i (16)
In most cases, density logs were not available to the mudline; there-
fore, the vertical stress was calculated from a regional depth-based
density trend validated against the local density data. We may also note that equation 2 implies that for a given e0 , λ
0
The stress states for the data were estimated using equation 11 there will be some limiting effective stress p0;L for which e becomes
and are shown in the p 0 -q plane in Figure 4. For effective stresses zero. Although reaching zero porosity at a finite stress appears to be
beyond approximately 500 psi, it can be seen that the stress path an artifact of the initial functional assumption, we can nevertheless
follows a nearly linear trend. Although this is not a strictly propor- use this as a basis to finish our reparameterization. Having assumed
tional trend as stipulated by a constant-K assumption, following the that the sediments display a consistent compaction behavior, we
discussion in the “Soil mechanics framework” section, it never-
theless provides some understanding of the regional success of
models built solely on VES without consideration of other stress
tensor components.
To fit the data with equation 10, four parameters must be deter-
mined: M, e0 , κ, and λ. The most direct path is to perform a simple
χ 2 minimization on the difference between p 0 predicted by equa-
tion 10 and p 0 from the data set by free adjustment of these param-
eters. The motivation for equations 8 and 10, though, was the
hypothesis that the SBS allows us to relate p 0 for an arbitrary stress
ratio K to an equivalent p00 for K ¼ 0 compaction with the same
void ratio. Stress states with arbitrary and potentially scattered stress
ratios K, which would have similarly scattered compaction lines,
could then be recast into equivalent K ¼ 0 void ratio-stress pairs
e; p00 , which would be described by a single compaction trend of
the form in equation 2. The e0 and λ parameters for our fit should
thus describe this equivalent K ¼ 0 compaction behavior, but the
K ¼ 0 compaction line defined by the e0 , λ found in a free fit
of the four parameters as described above is not in good agreement
with the data in the e- ln p00 space. Because the relationship in this
space (equation 2) was the starting point of our derivation, we rep- Figure 4. Estimated stress states of Gulf of Mexico compaction
data from well locations shown in Figure 3.
arameterize the fitting variables to provide constraints that ensure

Figure 3. Locations of extensional-setting wells.


D394 Hauser et al.

may further assume that they should display a similar compaction more complex function proposed by Hauser et al. (2013), it does fit
0 . Because e becomes zero at
end point at some effective stress p0;L as well or better than VES-based exponential or power law forms
this stress, by definition, typically in use by the pore pressure community.
The fit in e- ln p00 space is shown in Figure 5b. Again, it can be
e0 seen that the data are not fully described by the model developed in
λ¼ 0 . (17)
ln p0;L this paper. There are several possible reasons for this. The equation
for the yield surface is integral to the shape of the compaction re-
Using equation 17, we can remove λ from equation 16 and aver- lation of equation 10. Different yield surface models will result in
age it across the data set to yield a “best” estimate of e0 : different shapes for the compaction trend, and the modified Cam-
Clay yield surface used here may be too simple to capture the actual
P
1∕N e compaction behavior of the shales; certainly, a more complex yield
e0 ¼ Pi 0 . (18) surface can be constructed to provide a much better fit. It may also
1 − N ln1p 0 ln p0;i
0;L be observed that the distribution of the field data in the e- ln p 0
space is roughly bilinear, having different slopes at low and high
0
We now have a parameterization in terms of M, κ∕λ, and p0;L , stresses. It is possible that as stresses increase, new physical com-
which by construction will favor a linear fit to the data in the paction mechanisms come into play; within the approach taken
e- ln p 0 space. Optimizing on these three parameters leads to the here, this would require a yield surface that changed its fundamental
relations shown in Figure 5. Figure 5a shows the data and the fit shape with increasing stress — unlike the simple form used here but
using equation 10 in a standard VES space for comparison with consistent with modern clay models such as MIT-E3 (Whittle and
the VES-based trend presented by Hauser et al. (2013) for these Kawadas, 1994), which might therefore provide a better fit to
data. As the authors note, the actual distribution of the data cannot the data.
be fully described by functional forms such as a single exponential. Although it is clear that better fits are possible, we consider the
Similar departures from log-linear compaction have been observed performance of the SBS model with this data set to be encouraging
in studies by other authors (e.g., Long et al., 2011). Although the fit because a reasonable fitting procedure yields results that are com-
of equation 10 shown here does not match the data as well as the parable to most compaction models typically used for pore pressure
prediction. As stated at the outset, this is not an
a) b) environment in which the SBS model is expected
to improve predictions; instead, adequate model
performance here is a prerequisite to consider us-
ing it in more complex settings.

APPLICATION IN A
HETEROGENEOUS STRESS
SETTING
In complex stress environments, the simple
vertical-effective stress approach becomes prob-
Figure 5. Compaction data for Gulf of Mexico example. (a) Void ratio — VES plot lematic. It is in these settings that an approach
comparing the state-boundary approach (labeled “MCC model”) to the conventional
VES-type model presented in Hauser et al. (2013) (labeled “VES model”) and based on the state-boundary surface may prove
(b) e- ln p00 plot of field data and the state-boundary fit. (The logarithm is taken with quite valuable, even with limitations such as
p00 in pounds per square inch.) those highlighted in the prior section. To evaluate

Figure 6. Schematic cross section of the fold belt region. Rotation of the Borneo block results in deep-seated folds overlain by comparatively
undeformed drape sediments.
Stress path and compaction D395

this model in such a setting, we use data from a deepwater fold belt We first evaluate the fold data in a simple mean-stress sense with-
near Borneo. out incorporating the SBS. Figure 7b shows the data in e- ln p 0
The play examined here is located outboard of the shelf, offshore space. The mean stress of the drape data is estimated using the uni-
northeast Borneo. A schematic cross section of the geologic struc- axial form of equation 11. The results of estimating the mean stress
ture is shown in Figure 6. The underlying fold belt has experienced in the fold interval with the same uniaxial approach is shown in
shortening due to the clockwise rotation of northwest Borneo blue, demonstrating a similar scatter to the e-VES plot.
against the relatively undeformed Sundaland Block to the northwest To incorporate the tectonic shortening, the mean stress of the fold
(Simons et al., 2007), whereas shallow sediments that drape the data is calculated based on Equations 15 and 19 with the parameters
folds appear relatively undeformed. Within the folds, the maximum a, b to be determined. A compaction trend of the form in equation 2
horizontal stress appears to be oriented margin-normal with an over- is defined using the parameterization of equations 17 and 18 such
all strike slip to reverse-faulting stress regime (e.g., Kreemer et al, that the trend is determined by data statistics and a single fitting
2000; King et al., 2010). parameter p0;L0 . This results in three free parameters p 0 , a, and
0;L
Data were collected for 26 wells in this area with 1074 pressure b to be found.
points associated with approximately 470 distinct units. Couzens- We could in principle determine the compaction line, i.e., the
Schultz and Azbel (2013) present a subset of these data in terms of parameter p0;L0 , using only drape data because it is presumed to

VES, highlighting the probable impact of lateral stresses on the data be independent of tectonic effects represented by the parameters
set. Shale members associated with each pressure unit were deter- a and b. These tectonic parameters could then be determined by
mined based on a gamma rays (GR) cutoff, and corresponding minimizing the scatter in the compaction trend of the fold data be-
velocity and density data were tabulated. Shale porosity was calcu- cause we assume only these data experience significant tectonic
lated from density logs, and overburden pressure
was estimated by vertical integration of the den-
sity logs (in keeping with the assumptions used a) b)
in deriving equation 14). An exponential fit of
density with respect to the depth-below-mudline
was used to extrapolate available data back to the
water bottom; density at the water bottom was
taken as 1.8 g∕cc based on shallow data.
Figure 7a shows the compaction trend of these
data using only VES. The data from sediments
that drape the folds (“drape” data) appear to
follow a trend similar to those seen in exten-
sional settings such as the one discussed in the
previous section. This observation supports the
assumption that these sediments are in or near
a passive normal faulting stress environment.
Figure 7. Compaction data for the fold belt example. (a) Void-ratio VES plot comparing
The data from the fold interval (“fold” data), fold data to drape data. Although the drape data (green circles) appear to follow a typical
however, are widely scattered suggesting more extensional trend, the fold data (grey squares) are far more scattered. (b) Void-ratio/
complex compaction mechanics (Couzens- mean-effective-stress plot. The drape data follow a well-defined trend, whereas the fold
Schultz and Azbel, 2013). data using the same uniaxial stress assumptions (open light-blue squares) are very scat-
tered. The solid red squares show the fold data with mean stress including a tectonic term
This additional scatter is expected in areas fit to minimize this scatter.
where the overburden is not the only load applied
to the system. To estimate that effect here, we
begin by assuming that the forces acting on
the sediments are controlled by an applied lateral
strain in addition to the vertical loading, as dis-
cussed in the “Estimation of the stress state” sec-
tion and represented in equation 15. We further
assume that the lateral strain will be solely a
function of distance F from the axial surface
of the fold forelimb or the distance to the thrust
fault cutoff, and that this strain will decrease with
increasing distance (e.g., Hafner, 1951; Ellis
et al., 2004; Buiter et al., 2006). For this analysis,
the tectonic strain T is parameterized as

a
T¼ ; (19)
bþF

with a and b being the fitting parameters. Figure 8. Distribution of fold data relative to distance from the forelimb.
D396 Hauser et al.

stresses. However, the relatively small amount of drape data (∼18


distinct units) makes any determination of p0;L 0 from this subset very

uncertain. Instead, all three parameters are determined simultane-


ously by jointly varying them to minimize the χ 2 difference between
the model compaction line and the estimated p 0 of combined drape
and fold data.
Fold data were incorporated into the fitting process in two differ-
ent ways: (1) selecting fit parameters based on all fold and drape
data or (2) fitting with all drape data but only fold data farther than
2500 ft from the fold forelimb. Figure 8 shows the distribution of
fold data with respect to the distance from the forelimb. Including
all data in the optimization gives the most optimistic view of the
performance of the model, whereas optimization only on distal data
(presumed to have lower tectonic stresses) gives an indication of
how well the model might extrapolate beyond its calibration set.
The residual errors for both cases are shown in Table 1, and the Figure 9. Compaction data for fold belt example using the state-
results (e, ln p 0 ) for fold data with parameters from case 1 are plot- boundary-surface model to calculate p00 . Drape data are plotted
ted as red squares in Figure 7b along with the derived compaction as green circles. Using only a uniaxial stress model to estimate
trend. It can be seen in the figure that the drape data match the (p00 ; q) results in fold data points plotted as light-blue open squares.
Incorporating estimated stress due to compression, as described in
compaction trend reasonably well, lending additional qualitative the text, leads to the red squares.
support to the results of the joint minimization. The scatter in
the fold data is reduced by incorporating lateral stress into a mean
stress model, but even when fitting with all data, there is still con-
siderable spread.
To investigate the SBS model, the fitting parameters M, κ∕λ, and
0
p0;L are defined as in the prior section. The stress state in the drape
is estimated with equation 11 and in the fold with equations 15, 19.
0
We now optimize on five parameters (M, κ∕λ, and p0;L , a, b) using
the same two data partitions described above; the resulting residual
errors are shown in Table 1. Figure 9 displays the results in e- ln p00
space with the same symbols/colors as Figure 7 for comparison.
Comparing the full-stress points in Figures 7b and 9, there ap-
pears to be a slightly better constrained trend for the SBS model,
but the difference between the SBS and mean-stress models is not
striking. Examining the residual errors, though, it becomes clear
where the SBS approach may add value. Figure 10 shows the
residual mean error and the standard deviation of the mean for both
models as a function of distance from the fold forelimb, using
parameters optimized for the drape/distal fold data (case 2 above).
Figure 10. Distribution of residual errors from mean-stress and
For sediments approaching the forelimb, where lateral stresses and SBS models as a function of distance from the fold axis; parameters
resultant shear stresses will be highest, the mean-stress model de- for both models were fit using only data farther than 2500 ft from
teriorates rapidly. The SBS approach, on the other hand, appears the forelimb. Fit quality is comparable within the fitting range, but
insensitive to the position of the sediments. Although the mean- the SBS model extrapolates much better to shorter distances (higher
stress approach can provide reasonably good results within its lateral stresses).

Table 1. Residual errors for the mean stress and SBS models. Errors for two fitting scenarios are shown for each model: fit to
all data or fit only to data farther than 2500 ft from the fold (distal). Residual errors are presented for the full data set as well
as by distance from the fold. The SBS model can be seen to perform better close to the fold.

Mean stress model SBS model


Fit to all fold data Fit to distal fold data Fit to all fold data Fit to distal fold data

Residual error All data −1.8%  20.7% −10%  20.1% −3.2%  18% −4.6%  17.9%
F > 2500 ft 1.7%  20.1% −4.4%  18.5% −3.5%  19% −3.6%  18.9%
F < 2500 ft −6.4%  20.7% −17.3%  19.8% −2.9%  16.7% −5.8%  16.4%
F < 2000 ft −14.4%  19.9% −26.4%  18.4% −3%  18.2% −6.6%  17.5%
Stress path and compaction D397

calibration range if conditions are fairly constant, it does not per- VES = vertical effective stress
form well for shear stress states away from the mean K ¼ q∕p 0 VP ∕VS = compressional/shear velocity
trend of the fitting data set. The SBS model provides consistent re- α = Biot’s constant
sults both within its calibration range and beyond. The statistics in δij = Kronecker’s delta
Table 1 show that these behaviors of the models hold not only for ε = strain tensor
calibration with distal data but also for calibration using all available ϕ = porosity
data. The same characteristics that lead to better extrapolation of the κ = slope of log-linear swelling line
SBS model also lead to a more stable calibration behavior that is λ = slope of log-linear compaction line
relatively insensitive to the choice of calibration data. μ = shear modulus
ν = Poisson’s ratio
σ = effective stress tensor
CONCLUSIONS
We have defined a framework based on soil mechanics and po-
roelasticity formulations with which both shear and normal stresses REFERENCES
can be incorporated into compaction modeling. Although this Bowers, G. L., 1995, Pore pressure estimation from velocity data: Account-
framework provides the scope for including complex soil and tec- ing for overpressure mechanisms besides undercompaction: Presented at
tonic models, the examples in this paper illustrate that even simple IADC/SPE Drilling Conference.
Bowers, G. L., 2002, Vertical vs. mean effective stress — Which one is bet-
models can substantially improve the accuracy of the calculations, ter for pore pressure estimation?: Presented at SEG/EAGE Summer Re-
reducing residual errors in complex stress settings. It is also appar- search Workshop, Geopressure: Conceptual Advances, Applications, and
ent that the model parameters can be derived from field data and Future Challenges.
Breckels, I. M., and H. A. M. van Eekelen, 1981, Relationship between hori-
successfully extrapolated to different stress regimes. Although zontal stress and depth in sedimentary basins: Presented at 56th Annual
not undertaken in this work, inversion of the modeled equivalent Fall Technical Conference and Exhibition of the Society of Petroleum
K ¼ 0 mean stress p00 to obtain corresponding pore pressure should Engineers of AIME.
Buiter, S. J. H., A. U. Babeyko, S. Ellis, T. V. Gerya, B. J. P. Kaus, A. Kell-
be a straightforward process. ner, G. Schreurs, and Y. Yamada, 2006, The numerical sandbox: Com-
It is clear from the data considered in this paper that a full state- parison of model results for a shortening and an extension experiment:
Geological Society London, Special Publications, 253, 29–64, doi: 10
boundary-surface model will not add value in all settings. In exten- .1144/GSL.SP.2006.253.01.02.
sional settings, there appears to be more value in improving the Couzens-Schultz, B. A., and K. Azbel, 2013, Predicting pore pressure in
void-ratio/effective-stress trendline than in improving the full stress active fold-thrust systems: An empirical model for the deepwater Sabah
foldbelt: Journal of Structural Geology, doi: 10.1016/j.jsg.2014.07.013.
tensor definition. Even in some tectonically active systems, a simple Eaton, B. A., 1969, Fracture gradient prediction and its application in oil
mean-stress model may be adequate — such as the distal part of the field operations: Journal of Petroleum Technology, 21, 1353–1360,
fold data set examined here. However, in complex settings with doi: 10.2118/2163-PA.
Eaton, B. A., 1975, The equation for geopressure prediction from well logs:
significant variations in the stress state, the inclusion of the SBS Presented at SPE 50th Annual Fall Meeting.
model may substantially improve calculations relating compaction, Ellis, S., G. Schreurs, and M. Panien, 2004, Comparisons between analogue
and numerical models of thrust wedge development: Journal of Structural
stresses, and pore fluid pressures. Geology, 26, 1659–1675, doi: 10.1016/j.jsg.2004.02.012.
Fjaer, E., R. M. Holt, A. M. Raaen, R. Risnes, and P. Horsrud, 2008, Petro-
leum related rock mechanics, 2nd ed.: Elsevier.
ACKNOWLEDGMENTS Goulty, N. R., 2004, Mechanical compaction behavior of natural clays and
implications for pore pressure estimation: Petroleum Geoscience, 10, 73–
We thank Shell International Exploration and Production Incor- 79, doi: 10.1144/1354-079302-552.
Gutierrez, M. N., N. Braunsdorf, and B. Couzens, 2006, Calibration and
porated and Shell Exploration and Production Company (SEPCo) ranking of pore pressure prediction models: The Leading Edge, 25,
for permission to publish this work. 1516–1523, doi: 10.1190/1.2405337.
Hafner, W., 1951, Stress distribution and faulting: Bulletin of the Geological
Society of America, 62, 373–398, doi: 10.1130/0016-7606(1951)62[373:
SDAF]2.0.CO;2.
NOMENCLATURE Harrold, T. W. D., R. E. Swarbrick, and N. R. Goulty, 1999, Pore pressure
estimation from mudrock porosities in tertiary basins: Southeast Asia:
ab = strain calibration parameters AAPG Bulletin, 83, 1057–1067.
Hauser, M. R., T. Petitclerc, N. R. Braunsdorf, and C. D. Winker, 2013,
e = void ratio Pressure prediction implications of a Miocene pressure regression: The
p0 = mean effective stress Leading Edge, 32, 100–109, doi: 10.1190/tle32010100.1.
Hottman, C. E., and R. K. Johnson, 1965, Estimation of formation pressures
p00 = K ¼ 0 (isotropic) mean effective stress from log-derived shale properties: Journal of Petroleum Technology, 17,
0
p0;L = limiting effective stress (fitting parameter) 717–722, doi: 10.2118/1110-PA.
pc0 = preconsolidation stress Hubbert, M. K., and D. G. Willis, 1957, Mechanics of hydraulic fracturing:
Transactions of the AIME, 210, 153–168.
q = generalized shear stress Katahara, K. W., 2005, Selecting which stress to use in velocity-stress mod-
B = bulk modulus els for pressure estimation: 75th Annual International Meeting, SEG, Ex-
F = distance to fold forelimb panded Abstracts, 1223–1226.
King, R. C., R. R. Hillis, M. R. P. Tingay, and A.-R. Damit, 2010, Present-
K = stress ratio (shear to mean effective) day stresses in Brunei, NW Borneo: Superposition of deltaic and active
M = slope of critical state line margin tectonics: Basin Research, 22, 236–247, doi: 10.1111/j.1365-2117
MES = minimum effective stress .2009.00407.x.
Kreemer, C., W. E. Holt, S. Goes, and R. Govers, 2000, Active deformation
P = pore fluid pressure in Eastern Indonesia and the Philippines from GPS and seismicity data:
S = total stress tensor Journal of Geophysical Research B, 105, 663–680, doi: 10.1029/
1999JB900356.
SBS = state boundary surface Long, H., P. B. Flemings, J. T. Germaine, and D. M. Saffer, 2011, Consoli-
T = extrinsic strain dation and overpressure near the seafloor in the Ursa Basin, Deepwater
D398 Hauser et al.

Gulf of Mexico: Earth and Planetary Science Letters, 305, 11–20, doi: 10 Wyoming in light of fluid pressure hypothesis: Geological Society of
.1016/j.epsl.2011.02.007. America Bulletin, 70, 167–205, doi: 10.1130/0016-7606(1959)70[167:
Mathews, W. R., and J. Kelly, 1967, How to predict formation pressure and ROFPIM]2.0.CO;2.
fracture gradient: Oil & Gas Journal, 65, no. 8, 92–106. Simons, W. J. F., A. Socquet, C. Vigny, B. A. C. Ambrosius, S. Haji Abu, C.
Nikolinakou, M. A., and A. W. Chan, 2012, Soil model for rock properties Promthong, C. Subarya, D. A. Sarsito, S. Matheussen, P. Morgan, and W.
prediction in exploration settings: Presented at 46th US Rock Mechanics/ Spakman, 2007, A decade of GPS in Southeast Asia: Resolving Sunda-
Geomechanics Symposium. land motions and boundaries: Journal of Geophysical Research, 112,
Roscoe, K. H., and J. B. Burland, 1968, On the generalized stress-strain B06420, doi: 10.1029/2005JB003868.
behavior of ‘wet’ clay’, in Engineering plasticity: J. Heyman, and F. Whittle, A., and M. Kawadas, 1994, Formulation of MIT-E3 constitutive
A. Leckie, eds., Cambridge University Press, 535–609. model for overconsolidated clays: Journal of Geotechnical Engineering,
Roscoe, K. H., A. Schofield, and C. P. Worth, 1958, On the yielding of soils: 120, 173–198, doi: 10.1061/(ASCE)0733-9410(1994)120:1(173).
Geotechnique, 8, 22–53, doi: 10.1680/geot.1958.8.1.22. Wong, T. F., C. David, and W. Zhu, 1997, The transition from brittle faulting
Rubey, W., and M. K. Hubbert, 1959, Role of fluid pressure in mechanics of to cataclastic flow in porous sandstones: Mechanical deformation: Journal
overthrust faulting, Part II: Overthrust belt in geosynclinal area of western of Geophysical Research, 102, 3009–3025, doi: 10.1029/96JB03281.

You might also like