ARMA 13-527

Iterative 3D geomechanical modeling to determine distribution and impact

of stresses in deepwater Gulf of Mexico subsalt and near-salt environments
Nagy*, Z.R., Lee, D.W. and Sayers, C.M.
Schlumberger, Houston, Texas, U.S.A.
Zapata, T.R. and Silvestro, J.L.
YPF Services USA Corp., The Woodlands, Texas, U.S.A.

*now with Shell Exploration & Production Company, Houston, Texas, U.S.A.

Copyright 2013 ARMA, American Rock Mechanics Association

th
This paper was prepared for presentation at the 47 US Rock Mechanics / Geomechanics Symposium held in San Francisco, CA, USA, 23-26
June 2013.
This paper was selected for presentation at the symposium by an ARMA Technical Program Committee based on a technical and critical review of
the paper by a minimum of two technical reviewers. The material, as presented, does not necessarily reflect any position of A RMA, its officers, or
members. Electronic reproduction, distribution, or storage of any part of this paper for commercial purposes without the written consent of ARMA
is prohibited. Permission to reproduce in print is restricted to an abstract of not more than 200 words; illustrations may not be copied. The
abstract must contain conspicuous acknowledgement of where and by whom the paper was presented.

ABSTRACT: A 3D finite-element model (FEM), covering an area of 300 km2 in the deepwater Gulf of Mexico, was used to
determine the stress distribution around salt structures, and to study the impact on well design decisions. A 3D model was
necessary to accurately determine the full stress tensor (i.e., changes of stress magnitude and orientation around the salt body). An
iterative workflow to fine-tune properties and stresses has been developed to validate wellbore failure and to explain drilling
events. The 3D FEM was built using a geological model, high-quality seismic velocity data, and results from 1D mechanical earth
models. The model indicates that the stress magnitude and orientation are modified by salt morphology. In particular locations at
the salt/sediment interface, this phenomenon resulted in significant stress rotation and high von Mises stress. In these
circumstances, the safe drilling mud weight window can quickly disappear, forcing the operator to manage both wellbore instability
and mud loss, and with no choice but to sidetrack or select a new well location. The described case study illustrates the application
of the 3D FEM, which was able to explain drilling events (including an abandoned wellbore), provide input for calculating the mud
weight window (range between shear failure and minimum stress), and guide well design. The significance of this study is that, as
more well trajectories are drilled in these complex environments, robust, predictive 3D FEMs will be necessary to assess potential
risk prior to drilling and to improve well planning.
The conventional industry-standard approach to estimate
1. INTRODUCTION
the stresses along a trajectory planned to reach a subsalt
Many major discoveries in the deepwater Gulf of target mostly include 1-dimensional (1D) analytical
Mexico are located beneath thousands of feet of salt. calculation of properties, pressures, and stresses. The
This salt environment (supra- and subsalt) is intrinsic properties, formation pressures, and stress
characterized by complex geological structures and gradients are often migrated, from the existing offset
represents significant drilling challenges for operators. wells to the planned well location, along geological
Many of these challenges have been described [1], and markers. These simple assumptions and results do not
include drag zones composed of deepwater shales, low- explain and cannot predict drilling events around
strength strata and reactive shales, subsalt “rubble-” or complicated salt environments because the assumptions
breccia-like zones, and active or relict translation that the largest compressive principal stress (P1) is
surfaces (faults and “gouge-zones”). These challenges vertical, that rotation of stresses around the salt body can
are mostly associated with the sedimentary strata in the be ignored, or that rock properties are laterally constant
vicinity of the salt and sediment contact (e.g., fractured are generally erroneous.
carapace facies in the suprasalt or at the lower salt
Three-dimensional (3D) numerical simulation
interface), which is characterized by relatively low
techniques have been developed for solving a complex
minimum stress and fracture gradient, perturbed vertical set of equations with appropriate boundary conditions
and horizontal stresses, horizontal stress anisotropy and needed to estimate stresses and fluid pressures in salt
stress-arching [2‒4].
environments [2, 5‒8]. Detailed description and 2. METHODOLOGY: ITERATIVE 3D
comparison of the modeling techniques is included in MODELING
Luo et al. [4]. Numerical simulation to solve complex
geomechanical problems has proven suitable; however, 2.1. Wellbore-Centric Modeling
it has not become a routine practice due to the perceived 1D mechanical earth models (MEMs) were constructed
lack of well information or lack of quality subsalt for three existing offset wells in the area of interest
seismic velocity data prior to drilling, which makes the (Figure 1). The MEM is defined as a mathematical
results uncertain. Furthermore, basin-wide models representation of rock properties, pore pressure, and in
usually have low spatial resolution; therefore these situ stresses as a function of the depth for a particular
models are mostly impractical for wellbore-centric interval. The development of the earth model followed a
studies. Finally, it is difficult to validate the results of a detailed workflow, which comprises ten steps [10]. Input
3D model with wellbore data; therefore, the 3D results data used to construct the models includes formation
are often considered qualitative only. evaluation logs (compressional and shear velocity,
gamma ray, density, neutron porosity, and checkshots)
In this paper, we describe a modeling approach in which from wireline or logging-while-drilling operations;
3D numerical simulation covering 300 km2 (115.8 mi2) formation markers and lithology descriptions, drilling
was used to predict stresses around a salt body in a data with list of drilling events from offset wells, fluid
deepwater field in the Gulf of Mexico (Figure 1). While and mud logging information, and petrophysical
recent publications in which authors describe attempts to interpretations (porosity and volume of clay).
overcome the shortcomings associated with spatial Proprietary correlations for mechanical properties based
resolution or data quality [8, 9] by incorporating local on laboratory measurements on cores were adapted from
gridding techniques or well data and high-quality a nearby field. Pore pressures and stresses were derived
processed seismic interval velocity, respectively; we from available wellbore density and sonic data, and
show examples that validate the 3D model by using calibrated against in situ formation measurements and
wellbore data to explain drilling events. The calibrated well events (kicks, results of leakoff test (LOT), and
iterative model was used to estimate the mud weight mud losses).
window for a planned well in the same prospect. The
combined 3D models generated from well log-derived 2.2. Field-Centric Modeling
and locally calibrated elastic properties, strength, Factors considered while building the 3D geological
pressures, and stresses provide a more realistic earth model for the study area included the structural
model in contrast to the past published stress models, complexity of the salt (overhangs), orientations
which are mostly uncalibrated and are built on spatially (pinchout) of the reservoir sands relative to the base of
constant properties (layer-cake). salt, and the positions of existing offset wells. In
addition to the 3D interpreted surface representing the
water bottom, the top and base of salt and the reservoir
sand surfaces were used to generate the structural
framework. In the framework model, truncation rules
(e.g., erosional, base, discontinuous, and conformable)
were accurately represented, as interpreted from the
available seismic image data. A 3D MEM was generated
on the basis of well log-derived rock properties (1D
MEMs) and the available 3D seismic data. In sections
where well logs were not available due to the absence of
logging operations, such as in the overburden,
correlations were used to predict the required properties
from the seismic velocity. The 3D MEM, which
incorporated reservoir geometry and mechanical
properties was exported from the geological model and
imported into the finite element stress modeling routine
of proprietary geomechanics modeling software [11].
The geological model was extended by adding
sideburdens on the four vertical sides and the
underburden (below the reservoir). This improved the
Figure 1. Location map of the study area (magenta rectangle) model’s aspect ratio and ensured a reasonable transfer of
in the deepwater Gulf of Mexico showing available offset stresses within the reservoir and from the overburden in
wells, A, B, and C, and the reservoir top surface in the
the preproduction stress phase of the modeling. The
background. Scale: 1 U.S. offshore block is 4.8 km  4.8 km
embedded model consisted of 148  72  358 I,J,K cells,
(3  3mi).
for a total of 3.81 million cells. Properties in the good match was achieved between the minimum
sideburden section were assumed constant for each layer principal stress (P3) and field data (mud losses and LOT
and reflected average layer values. Properties in the data) at well locations within the model. The 3D stress
underburden section were extended downward from the model computes the six Cartesian components of the
model bottom. stress tensor (sxx, syy, szz, sxy, syz, sxz) in each cell, from
The initial stress state of the field was determined with which the three orthogonal principal stresses and stress
finite element geomechanics software [11]. Inelastic directions were determined (P1, P2, and P3).
constitutive laws for sediments surrounding the salt were Computed stresses inside the salt converged to an
applied in multiple modeling runs, while the fluid isotropic conditions after relaxation. This was verified
pressures were derived from 3D seismic interval velocity by evaluating the ratio between the deviatoric stress q
data, using a velocity-to-pore-pressure transform. The and the mean stress p inside the salt.
Mohr-Coulomb plastic yield criterion was used to model
2.3. Iterative Workflow
intact rock behavior. Effective stresses were used during
The stresses in the initial 1D MEMs were calculated
simulations, and then were added to pore pressure to
assuming that the orientation of the largest principal
obtain the total stresses according to the effective stress
stress was vertical and orthogonal to the intermediate
principle. The salt formation was modeled as a linear
and minimum principal stresses (i.e., P1 = SV, P2 = SH,
elastic perfectly-plastic material (with high Poisson’s
and P3 = Sh) (Figure 2). Because of the absence of
ratio, low Young’s modulus, very low cohesion, and
strength measurements from cores in the study area and
zero pore pressure) to ensure isotropic principal stresses
within the salt formation. there is no way to directly measure SH, the unconfined
compressive strength (UCS) and the horizontal stress
ratio (Sh/SH) were fine-tuned so that the calculated shear
Initial condition: and tensile failure were consistent with oriented (four-
P1 = SV, P2 = SH, and P3 = Sh arm) caliper, image logs, and reported drilling events
Subsequent condition:
P1 ≠ SV, P2 ≠ SH, and P3 ≠ Sh
(results of shoe tests and mud losses). The values of
UCS and Sh/SH had considerable uncertainty when
1D/3D MEM
imported into the stress model with the other properties,
First run: First run:
which include density (r), pore pressure (Pp), static
sxx, syy, szz, sxy, syz, sxz Pp, r, n, E, UCS, T, , Sh/SH Young’s modulus (E), tensile strength (T), friction angle
Second run: Second run: (), and Poisson’s ratio (n). The main outputs of the 3D
s’xx', s’yy, s’zz, s’xy, s’yz, s’xz UCS (T)
stress model include the stress and strain tensors, from
which principal stresses (effective and total),
3D stress displacements, and deviatoric stresses were calculated
model
(Figure 2). The six-component stress tensor, extracted
Six-component stress tensor (magnitude and
orientation), total principal stresses (P1, P2, P3), from the 3D model along the wellbore trajectory,
total volumetric strain , plastic volumetric strain
allowed the recalculation of the wellbore stresses (and
Figure 2. Schematic workflow showing the iterative approach shear and tensile failure) in a new 1D MEM, using
applied during the execution of the study. Some Nomenclature inclined far-field stresses (P1’, P2’, and P3’). In this new
for explanation of symbols. 1D iteration, the far field stresses were fixed (P3’ = P3)
because it was calibrated by well events, and the UCS
The stress initialization process, similar to that used by was further adjusted until the shear failure matched the
Fredrich and colleagues [2], was performed in several profile of the oriented (four-arm) caliper, image logs,
discrete steps. First, rollers on the sides (allowing the and was consistent with reported drilling events (tight
model to move freely in the vertical plane) were hole, drag, and overpull). The last step of the workflow
introduced, while vertical motion of the base was included the export of this updated UCS’ to the stress
prevented. The model was subjected to gravity loading, model (Figure 2).
with the appropriate pore pressure for sediments
calculated separately. This loading induced a reaction at 3. CASE STUDY BACKGROUND
the base of the model. Second, the model was fixed at The area of interest is located in the deepwater Gulf of
corner points and the stress from the first run was Mexico and covers 300 km2 (115.8 mi2) that includes
applied at the base. While the loads were applied, approximately 13 US offshore blocks (Figure 1).
stresses were calculated on the sides of the model on the
basis of the prescribed boundary conditions (ratio of the Drilling of the recent exploratory well, Well A, resulted
far field stresses Sh/SH, where Sh is minimum horizontal in a serious well-control incident followed by mud
stress, and SH is maximum horizontal stress). Third, the losses, and finally loss of the wellbore before entering
boundary conditions were adjusted iteratively until a the salt body. The existing two other offset wells, Wells
B and C, 17.7 km (11 mi) and 32.2 km (20 mi) away, obtained using the compressional velocity and the shear
respectively, successfully reached subsalt targets after wave velocity according to
considerable non-productive time. The planned well is
𝑉𝑝2 − 2𝑉𝑠2
located approximately 4.8 km (3 mi) to the northwest 𝜈𝑑𝑦𝑛 =
from Well A, and the play is defined by a three-way salt 2(𝑉𝑝2 − 𝑉𝑠2 )
(4)
closure. There is a significant thickness variation of the
suprasalt sedimentary sequence in the study area, The same correlations for Young's modulus and
varying from 370 m (1200 ft) at Well B to 1830 m (6000 Poisson's ratio were used for both sands and shales.
ft) at the planned well and Well A. Well A was drilled Because no laboratory measurements of rock strength
on the top of a local salt high that plunges steeply toward were available in any of the offset wells, a proprietary
a minibasin to the southeast, which is filled with up to correlation, defined for an adjacent field, was used for
4900m (16,000 ft) of thick suprasalt sedimentary rocks sands and shales. For rock strength, Horsrud-type [12]
that onlap the salt body. The thickness of the salt varies equations were defined for sand and shales and
from 500 to 8000 m (1400 to 26,000 ft) and is expressed as
characterized by a high width-to-height aspect ratio 𝑈𝐶𝑆𝑠𝑎𝑛𝑑 /𝑠ℎ𝑎𝑙𝑒 = 𝑎 + 𝑏𝑉𝑝𝐶 (5)
within the study area and has no significant salt
overhangs (therefore simplifying the construction of the where the unit of UCS is expressed in psi units, Vp is in
geological model). ft/s, and a, b, and C are calibration parameters for sand
and shale. Because the original equation was defined in
4. RESULTS units of MPa and Vp in km/s, unit conversion parameters
were used for consistency.
1D properties, pressures, and sand stresses were
calculated from well log data. In sections in which well The friction angle  was according to the method in
logs were not available due to the absence of logging Plumb [13], which is valid for both grain-supported and
operations, such as in the overburden, correlations were clay-supported rocks; whereas the tensile strength was
used to predict the required properties from the seismic estimated as 10% of the UCS for both sand and shale.
velocity data. These properties were calibrated using The results of the mechanical property profiles (elastic
available wellbore data, and validated within the properties and rock strength) along the trajectories of
geological model for consistency, as described below. Well A and Well B are shown in Figure 3 and Figure 4,
respectively.
4.1. Mechanical Properties
As with all mechanical properties, local correlations The salt lithology was assumed to be made up
from core are preferred for calibrating log mechanical exclusively of halite and was modeled with properties
properties. For this study, however, no core data were published previously [8, 14].
available for calibration; therefore a proprietary
The 3D MEM containing key elastic properties and rock
correlation, defined for an adjacent field, was used to
strength was generated on the basis of well log-derived
establish the static Young’s modulus:
rock properties, and the available 3D seismic interval
𝐸𝑠𝑡𝑎 = (𝑎 − 𝑏𝜙)𝐸𝑑𝑦𝑛 velocity volume was used where log data were not
(1)
available. Poisson’s ratio for the suprasalt sands and
where Esta is the static Young's modulus, a and b are shales was correlated with Vp and expressed as
calibration parameters, is total porosity, and Edyn is 𝜈 = −0.0004 × 𝑉𝑝 + 0.62493
the dynamic Young's modulus. The dynamic Young's (6)
modulus is obtained from the density ρ, the where Vp is in ft/s, and n is dimensionless.
compressional wave velocity Vp, and the shear wave
velocity Vs : Static Young’s modulus for the suprasalt sands and
shales was correlated with Vp and expressed as
3 𝑉𝑝2 𝑉𝑠2 − 4
𝐸𝑑𝑦𝑛 = 𝜌𝑉𝑠2 𝐸 = 0.003 × 𝑒 0.00082 ×𝑉𝑝
𝑉𝑝2 𝑉𝑠2 − 1 (7)
(2)
where E in Mpsi, Vp is in ft/s.
where the density and velocities are in consistent units.
4.2. Pore Pressure
To calculate the static Poisson's ratio n from the sonic The pore pressure model was built for the area of interest
data, the following correlation was used: using the seismic interval velocity from the e-Wave
𝜈 = 𝑎 × 𝜈𝑑𝑦𝑛 volume provided by WesternGeco [e.g., 15]. The
(3)
accuracy of the seismic velocity was investigated at the
where a is a locally derived calibration parameter. In the available offset well locations as described in the
above equation, ndyn is the dynamic Poisson's ratio, Appendix.
 where z is depth below the mudline, and a and b are
fitting parameters.

Figure 3. Profiles of elastic properties and rock strength

calculated from compressional and shear velocities and
density logs in Well A. Track 3: input data; Track 4: static Figure 4. Profiles of elastic properties and rock strength
Young’s modulus and Poisson’s ratio; Track 5: UCS; Track 6: calculated from compressional and shear velocities and
tensile strength; Track 7: friction angle. The extractions of the density logs in Well B. Track 3: input data; Track 4: static
same static properties from the 3D MEM (thick blue, red, Young’s modulus; Track 5: UCS and tensile strength; Track 6:
black and orange curves in Tracks 4‒7) are shown for quality Poisson’s ratio; Track 7: friction angle. The extractions of the
check. same static properties from the 3D MEM (thick blue, red and
orange curves in Tracks 4, 5, 7) are shown for quality check.
The velocity-to-pore pressure transform was calibrated
using available pressure measurements ‒ MDT A composite pore pressure transform was used to
pressures, mud weight and equivalent circulating density estimate the pore pressure distribution in the area of
(ECD) data ‒ and drilling events from the existing offset interest, consisted of fitting parameters for the subsalt
wells. reservoir section and for the supra- and subsalt shales.
This calculated pore pressure, as also observed for the
In this study, the method of Eaton [16] was employed, Well C, indicated a significant pressure reversal below
which estimates the vertical component of effective the salt. The phenomenon has been widely described in
stress σ obtained from the P-wave seismic interval the deepwater Gulf of Mexico [e.g., 18].
velocity, v, according to the relation:
𝑛
4.3. Vertical Stress
𝜎 = 𝜎𝑛𝑜𝑟𝑚𝑎𝑙 𝑉 𝑉𝑛𝑜𝑟𝑚𝑎𝑙 (8) The vertical component of the total stress, SV, at any
point was assumed to be given by the combined weight
where σnormal and Vnormal are the vertical effective stress
of the rock matrix and the fluids in the pore space
and seismic velocity that would occur if the sediment
overlying the interval of interest. This may be calculated
was normally pressured, and n is an exponent controlling from an integral of density:
the sensitivity of velocity to effective stress; in the
𝑧
current study n = 3. To use Eaton’s method, the 𝑆𝑉 𝑧 = 𝑔 𝜌𝑏 𝑧 𝑑𝑧 + 𝜌𝑤 𝑔𝑧𝑤
deviation of velocity from Vnormal and the velocity of 0 (10)
normally pressured sediments must be estimated. This
was obtained by fitting the sonic log to an analytical where SV is the overburden stress (psi units), g is the
function of the form proposed by Galperin [17]: acceleration due to gravity,rw is the sea water density
1/𝑎 (assumed to be 1.03 g/cm3), zw is the water depth, z is the
𝑉 𝑧 = 𝑉0 1 + 𝑏𝑧 (9) true vertical depth (TVD, ft) below mud line (z = TVD –
zw – za), and za is the height of kelly bushing above the minimum principal stress is mainly aligned in the
sea surface. In the absence of a density log, rb in the horizontal plane; therefore, P3 = Sh.
suprasalt section was estimated using the Amoco In addition, the validity of the magnitude and location of
equation [19]. This can be expressed as failure calculated in the 1D model provided a
𝜌 𝑧 = 𝜌0 + 𝑎 × 𝑧 𝑏 consistency-check of the six-component stress tensors.
(11)
4.5.1. Minimum stress magnitude
where the unit for rb is g/cm ; r0 = 1.6 g/cm , a = 0.053,
3 3
In the absence of reliable LOTs or minifrac data, a novel
b = 0.28, and z is TVD below the mudline. approach introduced by Edwards et al. [21] was used in
4.4. Horizontal Stresses the study to estimate the magnitude of the Sh. Mud loss
The magnitude of the minimum horizontal stress (Sh) in occurred in the subsalt section of Well B, during which
the 1D MEMs was calculated using the poro-elastic while-drilling service was recording the downhole
equation including the effect of tectonic strains, which annular pressure and flow rate. The wellbore breathing
can be expressed as [20] was characterized by mud losses when the pumps were
𝜈 𝐸 𝐸𝜈
turned on, followed by mud gains when the pumps were
𝑆ℎ − 𝛼𝑃𝑝 = 𝑆𝑉 − 𝛼𝑃𝑝 + 𝜀ℎ + 𝜀 turned off. This effect is caused by opening and closing
1−𝜈 1−𝜈 2 1 − 𝜈2 𝐻 (12) of fractures and simulates an extended-LOT, which
𝜈 𝐸 𝐸𝜈 yielded an accurate value for Sh. Examples of annular
𝑆𝐻 − 𝛼𝑃𝑝 = 𝑆 − 𝛼𝑃𝑝 + 𝜀 + 𝜀
1−𝜈 𝑉 1 − 𝜈2 𝐻 1 − 𝜈2 ℎ (13) pressure signatures before and after mud losses observed
in Well B are shown in the Appendix. Mud losses also
where α is Biot's coefficient, Pp is the pore pressure, and occurred in the suprasalt section of Well A when the
h and H are strain components in the direction of the mud weight in the hole was increased due to well control
minimum and maximum horizontal stress, respectively. issues. Excessive ECD most likely caused the fracturing
The Young's modulus (E) and Poisson's ratio (ν) were and the loss of drilling fluid to the formation. The
calculated from the correlations described in Eqs. 1, 3, 6 location of the loss zone and the downhole conditions,
and 7. A value of α = 1 was used for the Biot's interpreted from the while-drilling pressure log, were
coefficient. used to calibrate Sh (Figure 5).
Given the tectonic setting of the Gulf of Mexico area, we
first considered that the strains h and H should be not
only small, but very close to each other. This essentially
produced stress ranges similar to that used by Fredrich et
al. [2] (varying from Sh=SH=0.7SV to Sh=SH=SV). The
magnitude of the maximum horizontal stress (SH) is
difficult to quantify because of the lack of any technique
to measure it directly. An iterative method was
employed to reduce the error in estimating SH; by
changing h and H in the poro-elastic equation during
the iterations, the correct SH should reproduce, as closely
as possible, the observed response of wellbore failure
indicated by the oriented caliper data and wellbore
events.
4.5. Model Validation
One of the challenges of the 3D approach is the
definition of the model accuracy when trying to validate
it with wellbore data. To this end, we compared output
data calculated in the 1D and 3D models for the
magnitude and the azimuth of the minimum principal Figure 5. 1D profiles of wellbore (Sh and SV – solid black) and
stress (Sh and P3) in particular. We acknowledge that the principal (far-field) stresses (P1-red, P2-orange, P3-blue) in
3D stress model does not calculate Sh, and because P3 = wells A and B. The profiles were calibrated with mud losses
Sh is certainly not always the case; the comparison of (yellow star) in both wells. Note the good match between the
these outputs may not be fully valid. This certainly stresses calculated by the different models. Horizontal lines
would be the case where stresses are perturbed such as in are drawn at every 1000’ for vertical scale.
the zone adjacent to the salt. However, since the location
of comparison is approximately 1140 m (3750 ft) below In the 3D stress model, boundary conditions were
the salt (discussed in Section 4.5.2.), we predict that the adjusted iteratively during the stress initialization
process until a good match was achieved between the
minimum principal stress (P3) and the mud losses event arrow) is NW-SE, which is similar to the elongation
in Well B. interpreted from the wellbore (blue: bit size, red: hole size).

Since the mud loss events are located in different

stratigraphic intervals and are distant from each other, 5. DISCUSSION
they provide an adequate calibration to validate the 5.1. Migration of Abnormally High Pore Pressure at
stress profile calculated in the 1D and 3D stress models, the Salt Flank
as shown in Figure 5. A major well-control event occurred in the suprasalt
4.5.1. Azimuth of minimum stress section of Well A when drilled into an unexpected
Breakout occurs when the stresses around the wellbore overpressured shale. A velocity slow-down anomaly was
exceed those required to cause compressive failure of the observed in the seismic velocity data (Figure 7a), which
borehole wall. Around a vertical or near-vertical is attributed to high pressure that has migrated through
wellbore, stress concentration is greatest in the direction permeable formation from the mini-basin from southeast
of Sh; therefore, breakouts are oriented approximately of Well A. This pressure charged the adjacent shales that
perpendicular to SH [22]. The direction of Sh was onlapped the salt flank or pinched out against the fault
estimated using oriented caliper information (C1-3 and seated at the salt, and caused the unexpected abnormal
C2-4 from oil-based mud imaging tool) acquired in the pressure in Well A (Figure 7b). Located at the structural
subsalt section of Well B to identify the breakouts and apex, this fault acted as a trap, whereas the suprasalt
their direction [22, 23]. The analysis indicated that the shale and silty shales acted as a seal for abnormal
azimuth of the elongations varies slightly between 120 pressure migrating from the minibasin. A pressure
(+180) and 150 (+180) degrees. This direction of buildup can also be seen at the footwall side of this fault.
elongation is considered the azimuth of Sh. The
orientation of SH was further investigated from the
analysis of shear wave anisotropy measured in the
subsalt section of Well B. A shear-wave splits into a fast
and slow shear wave when traveling through a formation
that has intrinsic stress anisotropy.
In a vertical well, the fast shear will have polarization in
the direction of the maximum horizontal stress, while the
slow shear will have polarization parallel to the direction
of minimum horizontal stress [24]. The fast shear
azimuth in Well B varies little with depth and is
approximately oriented at ~65 to 80 degrees.
The azimuth of the minimum principal stress (P3) was
calculated in each cell of the 3D model from the
components of the stress tensor (sxx, syy, szz, sxy, syz,
sxz). Figure 6 shows the azimuth for the layer located at
the approximate depth at which elongation data from the
wellbore was also available. The P3 azimuth from the
3D model agrees with the Sh azimuth interpreted from
the caliper enlargement; therefore, the consistency
between the 1D MEM and the 3D stress model was
verified.

Figure 7. NW-SE cross-section through Well A and the

planned well showing the distribution of seismic interval
velocity in ft/s (a) and pore pressure converted to mud weight
equivalent units of pound per gallon (ppg) (b). Velocity
slowdown (pressure charge) extends down the salt flank but
this charge is not evident at the planned well location.

Figure 6. Map view showing the orientation of the minimum 5.2. Stress Orientation, Rotation and Stress Anisotropy
principal stress (P3) just above the first reservoir sand in Well Adjacent to the Salt Body
B as shown in data derived from the 3D model (left) and from The changes in vertical, horizontal and shear stresses
wellbore data (right). Note that the orientation of P3 (blue cause principal stress directions to be perturbed and
become reoriented adjacent to a salt body [2, 8]. The orthogonal to the base of salt surface (Figure 9, top),
isotropy of the stresses inside the salt should induce whereas the mid (P2) and minimum principal stresses
rotation of the principal stresses at the rock/salt (P3) are rotating in and out of the plane (Figure 9,
interfaces [2, 4, 14]. Similar observations were made in middle and bottom)
the study area with the principal stresses in the sediment
near the salt oriented parallel and perpendicular to the
salt/sediment interface. Visualization of the stress
orientation calculated from the six-component tensor is
presented with arrow(s) pointing in the direction of
principal stresses within in each cell. Note that the
presentation of orientations includes out-of-plane
directions, which may not be truthfully represented on
cross-section.

Figure 9. Orientation of the principal stresses along a NE-SW

cross-section in the vicinity of Well B; maximum (P1) (top),
mid (P2) (middle) and minimum (P3) (bottom). For clarity
purposes, only the stresses in the subsalt sediments are shown
(base of salt is at top). The size of the vectors is proportional
to the stress magnitude.

Fredrich and colleagues [2] found up to 35% horizontal

Figure 8. Orientation of the principal stresses along a NW-SE stress anisotropy adjacent to the salt body despite that Sh
cross-section in the vicinity of Well A; maximum (P1) (top), and SH were defined equal in the far field. Others [3]
mid (P2) (middle) and minimum (P3) (bottom). For clarity postulated that the stress regime is associated with
purposes, only the stresses in the suprasalt sediments are thrusting in this region, where SV = P3, though, fault
shown (top of salt is at bottom). The size of the vectors is elements near the salt body are rare. Small thrust faults
proportional to the stress magnitude.
with NE-SW compressional features have been
interpreted at the top of the salt to the west of Well A
The 3D model shows that the stress perturbation in the
(not presented in this paper). Although, it is not clear in
vicinity of the salt flank at Well A caused the maximum
the seismic image data; nevertheless, it seems to follow
principal stress (P1) to be nearly horizontal (Figure 8,
the salt-ridge with a NE-SW trend and perpendicular to
top), whereas the mid principal stress (P2) is close to
the main extensional faults. This stress reorientation
vertical (Figure 8, middle). The maximum principal
(with a constant wellbore trajectory) may have caused
stress (P1) in the subsalt section at Well B is oriented
significant increase in the shear failure limit, while narrow mud weight window. As pointed out by others
showing a decrease in tensile failure limit in Well A. workers [8], this representation is somewhat incomplete,
Actually, the stress rotation caused the tensile failure in since this excludes trajectory-dependent limits such as
this zone to be lower than Sh, which was calculated in shear and tensile failure. Nevertheless, it illustrates the
the 1D MEM. As a consequence, the change in failure effect of the salt geometry on the mud weight window
limits (i.e., mud weight window) combined with the and the drillability of the formation. In Figure 11, there
applied casing design resulted in simultaneous wellbore is a gradual decrease in width of the window adjacent to
instability and mud losses to drilling-induced fractures, the steeply dipping salt flank, and it practically becomes
and ultimately in loss of the well. zero as the salt/sediment interface is reached. Because
Well A was drilled into the top of this structure, it
The von Mises stresses are good indicators of the
experienced serious drilling problems related to the
locations of the maximum horizontal stress anisotropy,
narrow mud weight window. In contrast, the mud weight
and can be used as a well design tool during the planning
window at the planned well location is larger and allows
phase. Locations of high von Mises stresses coincide
a more achievable casing design and safer drilling
with narrow mud weight window [2, 4]; therefore it is a
operation.
useful tool for evaluating drillability or casing design
schemes. The trajectory of Well A penetrated the zone of
high von Mises stresses (Figure 10), and the well
became undrillable due to wellbore instability and mud
losses.

Figure 10. NW-SE cross-section of von Mises stresses (in psi)

indication of stress perturbation (high von Mises stresses) at Figure 11. Cross section of the study area showing difference
the salt/sediment interface in the vicinity of Well A. between the minimum principal stress (P3) and pore pressure
(in ppg), which identifies regions characterized by a narrow
mud weight window. For clarity purposes, the salt has been
5.3. Application of the Model to the Planned Well removed from the plot. Red color indicates areas with a
Design narrow window (< 0.5 ppg); blue color indicates areas with a
Detailed mapping of the suprasalt section in the study wide window (> 2 ppg) for safe drilling. 2x vertical
area showed numerous normal faults, which are the exaggeration.
consequence of vertical extension in the suprasalt section
during stress relaxation throughout geologic time [4]. In a more regional context, the drillability profile (Figure
The planned well is positioned at a favorable location, 11) demonstrates that a drilling operation is predicted to
since the trajectory will bypass most of these faults. In be more difficult if approaching a convex (i.e., salt
addition, no velocity slowdown is present above the salt bulging outward) salt/sediment interface, whereas a
or at the salt/sediment interface (Figure 7a). concave geometry provides a relatively larger drilling
Furthermore, the extensional faults at the salt flank form window [e.g., 8]. This is a result of horizontal extension
a structural apex that seems to contain the pressure and contraction caused by the salt geometry during stress
preventing from diffusing through this topographic high, relaxation [4]. The caveat of this approach is that high
as evidenced by the pressure change (~0.5 to 1.5 mud pressure that develops locally, as shown in this study,
weight equivalent in ppg) across this zone (Figure 7b). overprints the stress signature (e.g., in the embayment
The results of 3D stress modeling can provide a “quick- adjacent to the steep salt flank).
look” means to predict potential drilling problems ‒ a
volume of the difference between P3 and pore pressure 6. CONCLUSIONS
(in ppg) identifies areas of the model that have a narrow Regardless of obvious drawbacks in terms of the amount
mud weight window and, therefore, will be more of available data and model resolution to calculate
difficult to drill. A cross-section of such a volume is failure along a single trajectory, we successfully used the
shown in Figure 11, where the red color indicates a combination of 1D/3D static geological and
geomechanical models and 3D finite-element model to method showed that those unsafe drilling conditions
characterize stress state adjacent to a salt body located in were not present in the planned well location.
the deepwater Gulf of Mexico. During the modeling
approach, several iterations were performed in the 1D
and 3D domain to fine-tune the values for UCS and the ACKNOWLEDGEMENTS
Sh/SH in order to validate wellbore data and to explain
The authors acknowledge the support from
drilling events. The following main conclusions are
WesternGeco, allowing us to use the seismic volume
drawn from this modeling study:
from the e- Wave dataset. We thank, in particular, Eldin
 Reliable wellbore-centric results generated from 1D Burns for all the discussions and support, and thanks
and 3D mechanical earth models (MEMs) that Denise Comeaux, Scott Cornish, Jennifer Chesterman,
utilize wellbore data, high-quality 3D seismic data, and Cynthia Swenceski (all WesternGeco) for preparing
and a geological framework model prove to be a the data for this study. We thank J M Evans
more realistic representation of the properties of an (Schlumberger) for proof-reading an earlier version of
area than uncalibrated, layer-cake type models the document.
published in the literature.
The data for this publication was released by YPF
 In this study, one of the key findings is that the 3D Services USA Corp, Chevron USA and Schlumberger.
seismic velocity model clearly shows an over-
pressured zone above the salt at the Well A location
that is pressure connected to the adjacent minibasin. REFERENCES
The velocity and pore pressure model also show that 1. Wilson, S.M., and J.T. Fredrich 2005. Geomechanics
the overpressure is trapped by extensional faults and considerations for through- and near-salt well design. In
is pressure-separated from the planned well location, Society of Petroleum Engineers Annual Technical
reducing the drilling risk for the planned well. Conference and Exhibition, Dallas, Texas, 9‒12
October 2005, SPE 95621.`
 Not only does the complexity of the salt geometry
impact the stress magnitudes, but rotation of the 2. Fredrich, J.T., D. Coblenz, A.F. Fossum, and B.J.
principal stresses is shown at the rock/salt interface. Thorne 2003. Stress perturbations adjacent to salt
Visualization of the stress orientation demonstrates bodies in the deepwater Gulf of Mexico. In Society of
Petroleum Engineers Annual Technical Conference and
that reoriented stresses exist at the reservoir level in
Exhibition, Denver, Colorado, 5‒8 October 2003, SPE
Well B, which agrees with the azimuth of borehole 84554.
enlargement.
3. Dusseault, M.B., V. Maury, F. Sanfilippo and J.F.
 Similar to previous works, this study demonstrates Santarelli 2004. Drilling around salt: risk, stress, and
that salt geometry plays a significant role during uncertainties. Gulf Rocks 2004, In 6th North America
stress relaxation. Horizontal extension occurs at a Rock Mechanics Symposium (NARMS), Houston,
convex salt/sediment interface and causes a narrow Texas, 5‒9 June 2004, ARMA/NARMS 04-647.
mud weight window and challenging drilling 4. Luo, G., M.A. Nikolinakou, P.B. Flemings and M.R.
conditions, whereas horizontal contraction at a Hudec 2012. Geomechanical modeling of stresses
concave interface provides a relatively larger drilling adjacent to salt bodies: Part 1 – Uncoupled models.
window and safer drilling conditions. Contrary to AAPG Bulletin, 96: 43‒64.
this, we illustrated that the embayment adjacent to
5. Fredrich, J.T., B.P. Engler, J.A. Smith, E.C. Onyia and
the steep salt flank is an example in which high D.N. Tolman 2007. Predrill estimation of subsalt
pressure had developed locally and overprinted the fracture gradient: analysis of the Spa Prospect to
stress signature. validate nonlinear finite element stress analysis. In 2007
 The 3D stress modeling results offer a quick-look SPE/IADC Drilling Conference, Amsterdam, The
Netherlands, 20‒22 February 2007, SPE/IADC 105763.
tool for predicting potential safe and unsafe drilling
condition, and thereby, the application contributes to 6. Koupriantchik, D., S.P. Hunt, P.J. Boult and A.G.
evaluating well locations and realistic casing design. Meyers 2005. Geomechanical modeling of salt diapirs:
The difference (in ppg) between minimum principal 3D salt structure from the Officer Basin, South
stress (P3) and pore pressure, when applied to the Australia. In Proceedings to the 11th International
Conference on Computer Methods and Advances in
study area predicted an ever-narrowing mud weight
Geomechanics, Torino, Italy, 19 June 2005.
window at Well A upon approach to the
salt/sediment interface on the steeply dipping salt 7. Mackay, F., N. Inoue, S.A.B. da Fontoura and F.
flank, which, in hindsight, explained the serious Botelho 2008. Analyzing geomechanical effects while
drilling problems encountered in that well. The same drilling sub salt wells through numerical modeling. In
2008 Indian Oil and Gas Technical Conference and
 Exhibition, Mumbai, India, 4‒6 March 2008, SPE formations. In Rocky Mountains Meeting and Low-
113216. permeability Reservoirs Symposium, Denver, Colorado,
15‒17 April 1991, SPE 21847.
8. Adachi, J.I., Z.R. Nagy, C.M. Sayers, M.F. Smith and
D.F. Becker 2012. Drilling adjacent to salt bodies: 21. Edwards, S.T., T.R. Bratton and W.B. Standifird 2002.
definition of mud weight window and pore pressure Accidental geomechanics – capturing in-situ stress from
using numerical models and fast well planning. In SPE mud losses encountered while drilling. In 2002
Annual Technical Conference and Exhibition, San SPE/ISRM Rock Mechanics Conference, Irving, Texas,
Antonio, Texas, 8‒10 October 2012, SPE 159739. 20‒23 October 2002, SPE/ISRM 78205.
9. den Boer, L.D., C.M. Sayers, Z.R. Nagy, P.J. Hooyman 22. Plumb, R.A and S.H. Hickman 1985. Stress-induced
and M.J. Woodward 2006. Pore pressure prediction borehole elongation: A comparison between the four-
using well-conditioned seismic velocities. First Break, arm dipmeter and the borehole televiewer in the Auburn
24: 43‒49. geothermal well. Journal of Geophysical Research,
90‒B7: 5513-5521.
10. Plumb, R., S. Edwards, G. Pidcock and D.W. Lee 2000.
The mechanical earth model concept and its application 23. Plumb, R.A. and J.W. Cox 1987. Stress directions in
to high-risk well construction projects. In 2000 eastern North America determined to 4.5 km from
IADC/SPE Drilling Conference, New Orleans, borehole elongation measurements. Journal of
Louisiana, 23–25 February 2000, IADC/SPE 59128. Geophysical Research, 92‒B6: 4805‒4816.
11. VISAGE 2009. Introduction to Geomechanics and 24. Plona, T.J., K.W. Winkler, B.K. Sinha and R.
VISAGE 2009.1 Software. Software Solution Training D’Angelo 1998. Measurement of stress direction and
Manual, Schlumberger Information Solutions, 2009 mechanical damage around stressed boreholes using
dipole and microsonic techniques. In SPE/ISRM Eurock
12. Horsrud, P. 2001. Estimating mechanical properties of
’98, Trondheim, Norway, 8‒10 July 1998, SPE 47234.
shale from empirical correlations. SPE Drilling &
Completion, June 2001, pp 68‒73, SPE 56017.
13. Plumb, R.A. 1994. Influence of composition and APPENDIX
texture on the failure properties of clastic rocks. In
EUROCK, Delft, The Netherlands 29‒31 August 1994, A.1. Quality Check of Seismic Velocities
SPE/ISRM 28022. The seismic interval velocity model used in this study
14. Fossum, A.F. and J.T Fredrich 2002. Salt Mechanics
was obtained using tilted transversely isotropic, full
Primer for Near-Salt and Sub-Salt Deepwater Gulf of wave inversion (TTI FWI). The accuracy of the
Mexico Field Developments. Sandia National velocities was investigated at the offset wells by
Laboratory, Report SAND2002‒2063. comparing with the interval velocities obtained by
upscaling available sonic logs for these wells. Figure A-
15. den Boer, L.D., C.M. Sayers, S. Noeth, A. Hawthorn,
P.J. Hooyman, and M.F. Smith 2011. Using 1 compares the velocity from the sonic log with the
tomographic seismic velocities to understand subsalt velocity obtained from the seismic velocities. Very good
overpressure drilling risk in the Gulf of Mexico. In agreement is observed between the upscaled sonic data
Offshore Technology Conference, Houston, Texas, 2–5 and seismic-derived velocity data.
May 2011, OTC 21546.
16. Eaton, B. A. 1975. The Equation for geopressure
prediction from well logs. In 50th Annual Fall Meeting
of SPE-AIME, Dallas, TX, Sept 28 Oct. 1975, SPE
5544.
17. Galperin, E. 1974. Vertical Seismic Profiling, Society of
Exploration Geophysicists Special Pub. no. 12, Society
of Exploration Geophysicists, Tulsa, Oklahoma.
18. Zhang, J., W. Standifird and C. Lenamond 2008.
Casing ultradeep, ultralong salt section in deep water: a
case study for failure diagnosis and risk mitigation in
record-depth well. SPE Annual Technical Conference
and Exhibition, Denver, Colorado, 21‒24 September
2008, SPE 114273.
19. Traugott, M. 1997. Pore pressure fracture pressure
determinations in deepwater, Deepwater Technology
Supplement to World Oil. August 1997.
20. Thiercelin, M.J., and R.A. Plumb 1991. A core-based
prediction of lithologic contrast in East Texas
 A - Last connection without losses
400 15.8

350
15.7
300
15.6
250

Flow, gpm

ECD, ppg
200 15.5

150
15.4
100
15.3
50

0 15.2
0 20 40 60 80 100 120
Time, min

B - Moment of losses
400 15.8

350
15.7
Figure A-1. Comparison of P-wave velocity from the seismic
300
cube (red curves) with the P-wave velocity from the suprasalt
15.6
well (top) and subsalt well (bottom). The wellbore (sonic) 250

Flow, gpm

ECD, ppg
velocity is represented in a 2D histogram.
200 15.5

A.2. Estimating Mud Losses using Annular Pressure 150

15.4
Data 100
The signature of annular pressure in the recorded-mode 15.3
50
log is characteristic of mud losses [21]. The pressure
signal during the last connection without losses was 0 15.2
square shaped (Figure A-1a) in the subsalt section in 180 200 220 240 260 280 300
Time, min
Well B. After the losses occurred, the pumps were
turned on and off many times while several connections
C - 1st connection
were made until the total depth was reached. The 400 15.8
fracture opened and closed with each cycle of the mud 350
pumps. Careful choice of opening and closing pressures 15.7

was determined on the time dataset to estimate the 300

15.6
closure pressure. Location of the “thief zone” is 250
Flow, gpm

ECD, ppg
important because it is required for calibrating the stress
200 15.5
model. One method compares the oil-based mud
invasion profile using time-lapse resistivity data 150
15.4
acquired across the zone of losses. Since this data was 100
not available, the resistivity data acquired during 15.3
50
logging-while-drilling services and, following that, on
wireline tools were used. This analysis, however, did not 0 15.2
960 980 1000 1020 1040 1060 1080
clearly indicate the zone of losses. Review of the drilling
Time, min
report and the annular pressure time-data indicated that
the point of losses is likely to have occurred on-bottom
while drilling the sand/shale interface at the reservoir
section.
 D - 2nd connection G - 5th connection
400 15.8 400 15.8

350 350
15.7 15.7
300 300
15.6 15.6
250 250
Flow, gpm

Flow, gpm
ECD, ppg

ECD, ppg
200 15.5 200 15.5

150 150
15.4 15.4
100 100
15.3 15.3
50 50

0 15.2 0 15.2
1230 1250 1270 1290 1310 1330 1350 2220 2240 2260 2280 2300
Time, min Time, min

E - 3rd connection Figure A-2. Signature of annular pressure during the mud-loss
400 15.8 event in Well B. Legend: Black curve is downhole pressure
350 (converted to mud weight equivalent in ppg); Red curve is
15.7 flow rate in gallons per minute (gpm). From top to bottom: A
300 shows last connection without losses; B shows the moment of
15.6 losses at 220 min. Note the characteristic drop in ECD from
250
Flow, gpm

15.48 ppg to 15.34 ppg while pump rate remained constant

ECD, ppg

200 15.5 (250 gpm); C shows 1st connection after the losses with
150 sigmoid-shaped curve characteristic of fracture closure and
15.4 reopening; D, E shows 2nd and 3rd connections after the losses
100 with only fracture closure interpreted; F, G shows 4th and 5th
15.3 connections after the losses with fracture closure and
50
reopening.
0 15.2
1590 1610 1630 1650 1670 1690 1710
Time, min

F - 4th connection
400 15.8

350
15.7
300
15.6
250
Flow, gpm

ECD, ppg

200 15.5

150
15.4
100
15.3
50

0 15.2
1860 1880 1900 1920 1940
Time, min