You are on page 1of 11

Estimation of formation fluid pressure using high-resolution

velocity from inversion of seismic data and a rock physics model


based on compaction and burial diagenesis of shales
NADER C. DUTTA, Schlumberger, Houston, USA
JALAL KHAZANEHDARI, Schlumberger, Stavanger, Norway

ccurate assessment and


prediction of formation
fluid pressure and fracture
gradient form an essential
part of the planning process
during various phases of
hydrocarbon exploration
and exploitation; these
include regional basin
analysis and basin highgrading, prospect identification, analysis, and drilling
campaign to lift hydrocarbons. While the literature is
full of activities in this field
that are linked to the need
for drilling safety by providing help with casing and
mud weight programand
quite justifiably sovery
little discussion addresses
another, and very important
need: seal integrity and trap
configuration analysis.
Overpressured formations, in which the pore fluid Figure 1. Integrated pore-pressure prediction workflow.
pressure is higher than the
corresponding hydrostatic pressure, form an excellent trap ical approaches such as the Eaton-type methods based on
for hydrocarbons. However, if the pore fluid pressure normal compaction trend (NCT) analysis and empirically
exceeds a threshold dictated by the strength of the rock, the relating the observed deviations from the trend to mud
seal may have been breached in the geologic past. This will weight data from analog wells. This is unfortunate as this
cause the hydrocarbons to migrate away. This process will approach is not based on the fundamentals of rock physics
be further facilitated by the presence of hydrocarbons in dip- and, therefore, lacks reliability in prediction, unless the well
ping formations due to fluid migration and buoyancy effects. used in the analysis is the exact analog of the well to be
Thus, reliable estimates of formation pressure are critical to drilled.
In this article, we describe various steps in an integrated
understanding the hydrocarbon habitat, from regional to
prospect scale. Although many operators use seismic data pore-pressure prediction workflow that focus on estimatto estimate pore pressure during the basin ranking and ing the pore pressure and its uncertainty due to various
prospect high-grading phase, such practices are regrettably mechanisms such as disequilibrium compaction, diagenesis of smectite to illite, aquifer relief, and buoyancy. This
not routine in the industry.
During the drilling phase, pore-pressure estimation pro- workflow combines seismic velocity, inversion, and well
vides help prior to drilling with designing a safe mud weight data and provides a velocity-to-effective-pressure transform
program that is typically about 0.5 pounds per gallon (PPG) using rock physics-based methodology. A velocity-to-preshigher than the true formation pressure and 0.5 PPG less sure calibration is established at well locations using both
than the true fracture pressure. For deepwater clastics, as wireline and drilling data. A high-resolution velocity cube
the water depth increases, this safe operating window dimin- is generated by combining seismic inversion and velocity
ishes, causing drilling problems and raising costs. During analysis results and then calibrated to pressure. Finally, geodrilling, the predrill estimates of pore pressure are updated logic, seismic, and well data are used to identify different
using several drilling parameters and LWD/MWD data. lithologies, fluids, and possible variation of pore pressure
Estimation of pore pressure using seismic data such as in the sand from the background shale. Uncertainty in prevelocity is well known. However, the transform that relates dicted pressure is estimated through a Monte Carlo simuvelocity to pressure depends on various factors such as lation technique.
Figure 1 shows how the proposed workflow for interock/fluid type, overpressure mechanism, burial rates, and
history of sediments, diagenesis of rocks such as those of grated pore-pressure estimation and analysis brings together
shales, and others. Although we have made some progress geologic, well, and seismic data. The key steps include:
in these areas, much work remains. Due to the complexity velocity-to-pressure calibration using rock physics princiof the process, many practitioners have reverted to empir- ples, seismic velocity analysis, seismic inversion, generation
1528

THE LEADING EDGE

DECEMBER 2006
Downloaded 21 Nov 2009 to 64.91.188.154. Redistribution subject to SEG license or copyright; see Terms of Use at http://segdl.org/

of a high-resolution velocity cube, pressure estimation, sand


pressure correction, and uncertainty in predicted pressure.
In what follows, we discuss these steps briefly. More details
on some of these steps can be found in the articles by Dutta
listed in the suggested reading section.
Velocity to effective pressure calibration using rock physics
principles. Both overburden pressure (the total weight of
rock and fluid at a given depth) and the effective pressure
(the difference between the overburden pressure and the
hydrostatic pressure) are estimated from seismic velocity, if
well data are not available.
Bulk density, b, required for this calculation can be
obtained from various velocity-to-density transforms such
as the Issler (1992) method modified to be applicable in the
shallow part of the stratigraphy and accounting for variable
lithology. The following relation gives the original Issler
transform:
(1)
where is the fractional porosity, x is referred to as acoustic
component, and tm and t are the matrix and rock transient time, respectively.
The overburden pressure, S, is then given by
(2)
where g is the acceleration due to gravity. The effective pressure, Pe, is given by
(3)
where Pf is the pore pressure. The factor known as the
Biot consolidation coefficient is defined as:
(4)
where KF is the bulk modulus of the dry rock frame (drained)
and KS is the bulk modulus of the solid material. For soft
rocks = 1 and for hard rocks = 0. The value of can be
estimated either through laboratory measurements or field
data.
The critical parameter is the effective pressure. This can
also be estimated from velocity through rock physics principles. In the rock physics literature, such relationships are
well known and well established. Unfortunately, many practitioners of pore-pressure analysis did not and still do not
use that approach directly and instead use empirical correlations based on (1) NCT analysis; (2) estimation of deviation, v, of a velocity field from the NCT; and (3) correlating
v directly to pore pressure data from analog wells. These
methods assume disequilibrium compaction as the main
abnormal pressure mechanism. There are, however, several
problems with most trend-based geopressure approaches:
In many geologic settings, sediment loading is so fast
that quite often pressures in these sediments are above
hydrostatic at shallow depths which could prevent development of a normal compaction trend or make its estimation difficult.
Velocity and pressure calibration data at shallow depths
are often unavailable (due to hole size) and the determination of the normal trend and subsequent pressure
estimations are unreliable.
There is no general rule regarding the shape of the NCT.
Such trends cannot be transferred from one basin to

Figure 2. Diagenetic depth profile showing the modeled transition from a


smectite-rich rock to a illite-rich one. The depth and extent of the transition zone depends heavily on the thermal gradient and the burial rates of
sediments. A variation of 10% in the thermal gradient is also shown.
Kinetic parameters for the transition are obtained by data fit to X-ray
diffraction data on mixed-layer illite/smectite clays from fine-grained (less
than 5 micron) sediments from many wells. The pictures at the bottom are
the electron microscope images of a smectite-rich (right) and illite-rich
(left) sandstone.

another or even in the different part of the same basin


due to using depth as a substitute for geologic variables
that control pressure distribution.
To overcome some of the above-mentioned problems
associated with methods that use NCT trend-based velocity-to-pressure transform techniques, Dutta (1983) first developed a unique rock-physics-based geopressure model that
has been modified and refined over decades as more data
are made available. A brief description of an earlier version
of this method to estimate effective pressure from velocity
(or slowness) is provided by Lopez (2004) and termed the
Simms-Dutta approach. We have continued to evolve this
type of method as our understanding of rock physics of various rock types and its seismic response continue to improve,
controlled laboratory measurements become more sophisticated, and access to well-log data widens. For example,
development of the current model and the reliability of its
predictions are greatly facilitated by the access to more than
100 000 wells from the Gulf of Mexico through an external
commercial organization specializing in providing such services.
The current version of Duttas method uses a rock
DECEMBER 2006

THE LEADING EDGE

Downloaded 21 Nov 2009 to 64.91.188.154. Redistribution subject to SEG license or copyright; see Terms of Use at http://segdl.org/

1529

Instead of CEC to define the end-member rock properties, one uses velocities appropriate for smectites and
illites.
4) Pore pressure is then calculated as the difference between
the overburden pressure and the effective stress, using
the Terzaghi principle (Equation 3) with an appropriate
Biot coefficient.

Figure 3. (right) Plot of sonic log and normal compaction curve used to
estimate pore pressure according to Eaton method. The red curve labeled
NCT is the predicted normal compaction trend from the current model
(Dutta method). In the traditional Eaton method, this curve would have
been a straight line. Although such an empirical curve can be constructed
by curve fitting, we note that it would have a wrong intercept and violate
the critical porosity model dictated by the rock physics principle. (left) Plot
of pore-pressure as a function of depth. The predicted pore pressure using
the Eaton (red) and Dutta (blue) methods are displayed.

Figure 4. (left) Slowness versus effective stress and temperature relations


for deepwater sediments using the Dutta method. (right) Interval transit
time or slowness versus depth trends at various pore pressure conditions
from 9 to 15 ppg for a location in the Gulf of Mexico based on the model
on the left.

physics model that relates effective stress to lithology, porosity, seismic velocity, and time-temperature history of sediments. Burial diagenesis of shales (smectite-to-illite
transformation) is accounted for by a kinetic model based
on first-order rate theory and calibration based on X-ray diffraction data from analyses of fine-grained sediments (less
than 5 microns) of many wells. The model does not use a
cation exchange capacity (CEC) parameter in the analysis
as discussed by Lopez in the so-called Simms-Dutta method.
The method has four components:
1) The bulk density is estimated from a rock model that
relates velocity to temperature (and time) and bulk density. Overburden pressure is then calculated by integrating the estimates of bulk density as in Equation 1.
2) The effective pressure is estimated using another rock
model that relates void ratio (and temperature and time)
to effective pressure (compaction model). Void ratio is
calculated from porosity using the velocity-derived bulk
density of the rocks. Void ratio is defined as =/(1).
3) Time-temperature history is computed from seismic data
using a simple burial history model and is related to an
illite-to-smectite transformation using a kinetic theory.
1530

THE LEADING EDGE

The accounting for two major overpressure mechanisms


(smectite-to-illite diagenesis that changes rock properties and
releases extra water in the pore space, and compaction disequilibrium) is an integral part of the current rock model.
In shale-dominated basins, the smectite-to-illite transformation is mainly controlled by seabed temperature, geothermal gradients associated with various geologic units,
the rate of burial of these units, and their rock types (Figure
2). This transformation results in porosity and permeability changes. This crystallographic transformation occurs in
three stages of dewatering and could result in total of about
4% volume increase (Osborne and Swarbrick, 1997).
The range of overpressure and its depth of occurrence
depend heavily on the rate of heating of each sedimentary
unit. Numerous experiences in studying pressure systems
from the Gulf of Mexico, West Africa, and southeast Asia
suggest that diagenesis is an important element in many
pressure systems. A consequence of this phenomenon is
that for each lithology, compaction is a function of effective
stress and temperature and not depth. Therefore, the notion
of virgin compaction trend as discussed by Bowers and
others is redundant in this model. Figure 3 shows estimation of pore pressure using both a trend-based method (least
squares fitting) and Duttas current rock-physics-based
approach. It is nice to see that, without using well data, the
rock physics approach makes a good prediction of pore
pressure and further captures the onset of overpressure
zone better than the trend-based approach. Further, in shallow intervals, the trend-based method predicts unrealistic
pressure values. One way to overcome this problem is to
apply multiple trends at different intervals. This could be
dangerous as there is no rule as how many trend lines
should be used. The whole prediction, in that case, could
easily become a curve-fitting exercise rather than a realistic
compaction trend estimation based on first principles of
rock physics.
Seismic velocity and pore pressure at regional and prospect
scales. The rock velocity, which is needed for pore pressure
estimation and, in fact, for any other subsurface characterization, is a fundamental intrinsic property of rocks that
mainly depends on the grains, pores, and fluid properties
and their interaction, as well as extrinsic properties such as
pressure and temperature (Khazanehdari et al., 1998). Thus,
it is vital to ascertain that true rock velocities are estimated
from analysis of seismic data.
Dutta (2002) discussed the use (and abuse) of seismic
velocities for pore-pressure analysis. It is well documented
that significant differences exist between the velocity field
obtained using different seismic techniques such as a conventional method based on the Dix analysis, and stacking
and reflection tomography (Wilhelm et al., 1988). A standard
NMO stacking velocity analysis can be insufficient in complex media because of its simplified layered velocity model.
Instead, the traveltime tomography algorithms tend to provide more accurate velocity data in structurally complex
areas (for example, Sayers et al., 2002). In general, these
methods provide a velocity that is more appropriate to correct the seismic raypath for certain subsurface structure (for

DECEMBER 2006
Downloaded 21 Nov 2009 to 64.91.188.154. Redistribution subject to SEG license or copyright; see Terms of Use at http://segdl.org/

Figure 5. (top) A stacked seismic section with some horizons from a 3D data volume. (bottom) Interval velocity derived from the prestack data of above
using the Dix model. Vertical scale is km.

example, dipping reflectors) and the acquisition system


(source-receiver offsets). When the geologic environment
suggests that there could be sizable lateral variation in the
velocity field, the seismic data should include a prestack
depth migration (PSDM) process so that the images are in
true depth. However, the resulting velocity field is usually
too high after this process, leading to pressure estimates that
could be too low. In such case, one could transform the data
back into the time domain after PSDM and perform another
velocity analysis as suggested by Al-Chalabi (1994). In that
classic paper, Al-Chalabi clearly distinguished between processing velocity (identical to seismic velocity terminology in
this paper) and propagation velocity. His definition of propagation velocity is identical to that which we call rock velocity in this paper.
In any case, the interval seismic velocity should be calibrated with well data if such data are available. In exploration applications with no or minimal well control, one
should check the derived interval velocity field with
expected ranges of velocities versus depth based on rock
physics principles for that particular geologic environment.
An example of this is shown in Figure 4. On the left, we show
a template of slowness (inverse of velocity) versus effective
stress for various values of temperature for shales based on
Duttas model described earlier. This template is then used
to predict velocities at various states for assumed values of
1532

THE LEADING EDGE

pore pressure (Pf) in PPGall the way from hydrostatic to


fracture pressure (on the right of Figure 4). The curves on
this figure are predicted rock velocities model versus depth
for increasing pore pressure at every 2-PPG increment. We
note that the acceptable values of velocities from seismic
analysis, if performed according to the steps discussed
above, should be within the ranges predicted from rockphysics-based model. Of course, this QC procedure will
identify zones that are outside the predicted range. Such discrepancies must be investigated and rationalized thoroughly.
In summary, we recommend the following steps as part
of seismic velocity analysis and conditioning and extracting velocities that are close to rock velocity needed for pressure estimation:
1) Log editing and calibration and tie with seismic data after
proper upscaling.
2) Estimation of anisotropy parameters using borehole seismic data.
3) Gather conditioning (should include: multiples suppression, residual NMO, and anisotropy correction)
4) Combination of manual and automatic velocity picking.
The automatic velocity picking should be spatially consistent and horizon constrained. The well information
(check-shot-calibrated sonic logs and VSP) should be an
integrated part of this step and used to constrain the

DECEMBER 2006
Downloaded 21 Nov 2009 to 64.91.188.154. Redistribution subject to SEG license or copyright; see Terms of Use at http://segdl.org/

autopicking.
5) QC and comparison with borehole elastic velocity and iterate
from step 3 if there is significant
inconsistency between estimated borehole and seismic
velocities.
6) In regions with complex raypath
trajectories and wave propagation effects, reflection tomography is a tool that should be used
to estimate a geologically meaningful velocity model.
Finally, one should compare
seismic velocities thus derived with
a velocity template based on a rockphysics-based model and valid for
that geologic basin. Any discrepancies must be noted and rationalized
and corrective steps should be
undertaken, if necessary.
Our experience shows that the
above steps are necessary for obtaining good-quality, stable, and geologically consistent interval velocity
for pore-pressure analysis.

Figure 6. Interpreted horizons (autopick) for the data of Figure 5 for automated velocity analysis using
SCVA procedure as discussed in the text. Each dot in this figure represents an event that is continuous
and interpretable for 20 traces around the dot. A blank portion indicates where no continuous horizon
was found. Interval velocity analysis is performed at each location using prestack data and a generalized liner inversion procedure at each time sample. Vertical scale is seconds.

Figure 7. A comparison of SCVA/AVMB (top) interval velocity with those from Dix analysis (bottom). The top figure shows more details, even in the
low-frequency model, than shown in the bottom figure. Vertical scale is km.
DECEMBER 2006

THE LEADING EDGE

Downloaded 21 Nov 2009 to 64.91.188.154. Redistribution subject to SEG license or copyright; see Terms of Use at http://segdl.org/

1533

Figure 8. 3D porepressure image showing high-pressure


zones and pressure
regressions.

Figures 58 show an example of velocity analysis and


pore-pressure imaging over an area of 650 km2 covered with
a 3D seismic survey (deepwater clastics). Figure 5 (top)
shows an example of a portion of the stacked seismic data.
The bottom of Figure 5 is the interval velocity resulting
from Dix analysis (with three geologic horizons) and conditioned to represent rock velocity as discussed above. Figure
6 shows all horizons that are present from an automated
velocity analysis algorithm within an established criterion.
Figure 7 (top) is the resulting interval velocity that uses a
generalized inversion scheme to convert rms velocities into
interval velocities (at every time sample) as opposed to a
Dix-type approach where such analysis would be unstable.
For comparison, the bottom of Figure 7 shows the interval
velocity from Dix-type analysis. The comparison shows that
although Dix analysis is consistent with the top figure in
gross details, the top figure (from spatially consistent velocity analysis, SCVA) contains more details that are consistent
with the geologic model. Figure 8 is a 3D pore-pressure
cube obtained from SCVA analysis over the entire area of
650 km2. The cube shows high pressure as well as pressure
regression zones. These results have been verified by drilling.
3D pressure cubes such as Figure 8 are extremely valuable
for regional basin analysis and prospect high grading.
Seismic inversion and high-resolution pressure analysis
for drilling. Although the detailed velocity analysis techniques discussed above yield detailed pressure variations
within a minibasin, such analyses are not appropriate for
drilling applications. The same comment applies to any
other technique that is inherently low frequency in nature,
such as reflection tomography for velocity analysis. These
methods do not yield velocities at a scale that drillers require.
For drilling applications, we recommend deriving interval
velocities at much finer scale using either poststack seismic
inversion (PSSI) of amplitudes in conjunction with any
1534

THE LEADING EDGE

Figure 9. (top) High-frequency velocity versus two-way time obtained


from PSSI method. (bottom) The low-frequency trend of the velocity (from
SCVA) inversion procedure that was used to create the model at the top.
Vertical scale is m/s.

acceptable low-frequency model, such as SCVA, as discussed above, or reflection tomography or prestack waveform seismic inversion (PWSI). Below, we discuss the use
of these techniques for pressure estimation for drilling applications.
PSSI. The use of surface seismic stacked data for ampli-

DECEMBER 2006
Downloaded 21 Nov 2009 to 64.91.188.154. Redistribution subject to SEG license or copyright; see Terms of Use at http://segdl.org/

Figure 10. Composite display of seismically derived pore pressure (left) and the high-resolution interval velocity (right) after combining seismic velocity
and inversion results using PSSI. The two wells and the horizontal slice displaying the pore pressure are also shown.

tude inversion is well documented; it is frequently used for


impedance analyses for rock/fluid identification. However,
this technique can also be used for pore-pressure analysis
as it can yield the much higher resolution required for
drilling applications. Seismic inversion is excellent at deriving impedance contrasts across layered interfaces within the
reflection data, which reveal short-scale (high-frequency,
tens of meters) variations of impedance; however, it is not
able to determine the large-scale (low-frequency, hundreds
of meters) vertical trends.
The conversion of the band-limited inversion results
(also referred to as relative impedance)
to absolute impedance is accomplished by adding a low-frequency
trend to the inversion result. This can
be done either during or after the
inversion. Typically, a weighted projection of impedance log data (from
the available wells) onto the position
of each trace in the 3D volume is
added to inversion results to obtain
absolute acoustic impedance. Another
approach would be to use the seismic
velocity field (SCVA or Dix) as a lowfrequency trend. This approach also
requires estimation of a low-frequency
density trend to obtain a seismically
constrained low-frequency imped-

ance. There are numbers of geostatistical techniques that can


be deployed for density modeling (both high- and low-frequency trends). The aim is to extend the range of the hard
data (well density) by correlated soft data (velocity).
By removing the full-frequency density trend from the
computed absolute impedance, an estimate of the high-resolution velocity field is obtained that is more suited to
drilling applications than those from the low-frequency
velocity analyses techniques. We follow this approach in both
1D and 3D. Obtaining velocity through seismic inversion
provides the necessary high-frequency components, which

DECEMBER 2006

THE LEADING EDGE

Downloaded 21 Nov 2009 to 64.91.188.154. Redistribution subject to SEG license or copyright; see Terms of Use at http://segdl.org/

1535

Figure 11. 1D prestack full waveform inversion (PWSI) as discussed in the text for a location in the deepwater Gulf of Mexico. The first panel shows
the seismic CMP gather; the second panel shows the synthetic gather generated from the PWSI method that best fit the data on the left. The third,
fourth, and fifth panels show VP, VP/VS, and bulk density derived from the inversion, respectively. The blue curves show the result from inversion. The
pink curve denotes the low-frequency velocity. The green curve is the filtered version of the blue curves, and the yellow curves signify the uncertainty in
the estimation of the attributes. The zone labeled SWF denotes where mild shallow waterflow (SWF) was observed in this location.

are missing in kinematics-based seismic velocity analysis


such as the semblance analysis (SCVA, Dix) or reflection
tomography. An example of this approach is given in Figure
9 in 1D and in Figure 10 in 3D. The details of pressure variations are better suited for well planning and updates in real
time with drilling data. It should be noted that the high frequency velocity data could be also influenced by lithological variation. Seismic inversion for both acoustic and shear
impedances using prestack inversion techniques can help
to discriminate better between lithology and pressure effects.
PWSI. This technique also provides higher frequencies
in the velocity estimation and is an alternative to PSSI
described above for drilling applications and well design.
Details of the inversion technique are published by Mallick
(1995) and used for pressure estimation by Mallick and
Dutta (2002) for analysis of shallow water flow problems
and will not be discussed here. Briefly, PWSI is an algorithm
that extracts detailed P-velocity, S-velocity, density, and
Poissons ratio from prestack CMP gathers that have been
processed for AVO analysis. The PWSI uses a genetic algorithm to optimize an initial elastic earth model (VP, VS, and
density) at discrete drilling locations. An example of the predicted high-frequency information is shown in Figure 11.
The left in the first panel shows the actual CMP gather
obtained from using the Q-Marine technology of
Schlumberger in the deepwater Gulf of Mexico. The second
1536

THE LEADING EDGE

panel shows the match obtained by inversion of the CMP


gather using PWSI approach. The resulting VP, VP/VS, and
bulk density are shown in panels 46, respectively. The predicted bulk densities are used to estimate overburden stress,
and VP is used to obtain effective stress using Duttas rockphysics approach described earlier. Because this technique
uses more parameters than the PSSI approach, this approach
can yield nonunique and ambiguous results. To minimize
problems associated with this, we use rock-physics constraints on the inversion scheme. Figure 12 shows the predicted pore pressure, overburden pressure, and fracture
pressure (third panel) that use the velocity data from Figure
11. It should be noted that conventional velocity analysis
couldnt be performed at shallow depths such as shown in
Figure 12; however, inversion can yield useful velocity information at these depths.
Sand pressure estimation. Besides disequilibrium compaction and shale diagenesis, there are other pressure mechanisms for which we must account. These are lateral transfer
of fluids updip through the inclined aquifers and buoyancy
pressure associated with hydrocarbon replacing brine in a
reservoir. For well planning and drilling applications, these
mechanisms could be important, especially if the structural
relief of a reservoir is significant. England et al. (1987) and
others show that if a permeable sand is loaded asymmetri-

DECEMBER 2006
Downloaded 21 Nov 2009 to 64.91.188.154. Redistribution subject to SEG license or copyright; see Terms of Use at http://segdl.org/

porosity, b is bulk density, f is fluid


density, g is gravitational acceleration,
Z is depth, and L is sand relief.
Through a combination of seismic
inversion, attribute analysis, and 3D
visualization techniques, the sand
bodies and their properties can be
characterized, thus allowing an estimation of possible excess pressure due
to structural relief. Figures 1315 show
a sand body delineated from surrounding shale. A computer program
makes an estimation of the sand
geometry and its total relief and computes the excess pressure due to structural relief using the model of Stump
et al. Once the sands are delineated
and the excess pressure due to structural relief is corrected, the pressure
variation due to fluid buoyancy (oil,
gas) can also be easily modeled by
allowing for density changes. An
example is shown in Figure 15.
Uncertainty in predicted pressure.
There are many sources of uncertainty
in estimated pore pressures such as
velocity and the parameters used to
describe the rock model. The estimation of the error associated with seismic velocity is not a trivial task. If the
seismic data were treated for noise,
multiples, anisotropy, and any other
geometric factors that affect the velocity data quality, the difference in seismic and well velocities could be used
as a measure of velocity uncertainty.
The error in density estimation could
come from the original density logs,
velocity-to-density transform, and the
rock physics parameters. Similar
errors exist in the parameterization of
velocity to effective stress transforms
and quantification of the diagenetic
profile. We have made an attempt to
capture these uncertainties as well.
Figure 16 shows the results from a
Monte Carlo simulation technique
aimed at capturing the cumulative
uncertainty associated with pore-pressure estimation at a proposed well
location that accounts for most of
Figure 13. Two sand bodies mapped using the acoustic impedance and velocity data. The sands on
these uncertainties. It is clear that
the left show the relative pressure difference between the sand and the surrounding shale. The sands
uncertainties are large. We suggest
on the right show the total pressures.
that a unique answer for pore pressure
cally by relatively impermeable shales, the consequent sand at a given depth prior to drilling the well is not possible;
and shale pressures will differ. The sand will maintain a one must manage the risk. The current procedure is intended
hydrostatic pressure gradient while the shale could main- to reduce the risk. Eliminating the risk is not possible. Only
tain a nearly lithostatic pressure gradient. Equation 5 shows by incorporating real-time drilling data and updating the
a model for estimation of reservoir pressure due to struc- model as drilling continues can one reduce the uncertainty.
tural relief in otherwise pressured shales (Stump et al., 1998).
Conclusions. Pore-pressure estimation using velocity is not
(5) an art but a science. It is based on sound principles of geology, and its detection is based on fundamentals of seismic
Here, DP is the amount of overpressure within the sand, wave propagation, seismic data acquisition, processing, and
is matrix compressibility, f is fluid compressibility, is velocity conditioning, and rock physics analysis. Although
Figure 12. (left) Gamma-ray curve at the well corresponding to the seismic data in Figure 11. (middle) High-frequency interval slowness derived from PWSI inversion (blue) with superimposed NCT
obtained from the rock model of Dutta (red). (right) Predicted pore pressure, overburden pressure,
fracture pressure using Eatons model, and an estimate of fracture pressure that is based on the
assumption that the fracture pressure is 90% of the overburden pressure.

DECEMBER 2006

THE LEADING EDGE

Downloaded 21 Nov 2009 to 64.91.188.154. Redistribution subject to SEG license or copyright; see Terms of Use at http://segdl.org/

1537

Figure 14. Seismic attribute analysis and classification used to identify potential sand bodies (second panel from left). A computer program estimates
the geometry of the individual sand bodies and computes the excess pressure due to structural relief (third panel). The pressure cube that is more representative for a background shale lithology (first panel) corrected according to the estimated pressure for each individual sands body (fourth panel).

discrete locations in a 3D volume (say, performing Dix-type


analysis of velocity at a grid of
0.5 km to save money!). This will
capture more details of pressure
variation.
However, for drilling applications and well planning, this
approach is not sufficient due to
lack of higher frequencies. We
recommend using seismic inversion for these types of applications using the low-frequency
trend as a starting point. We suggested two such approaches:
post- and prestack impedance
Figure 15. Relative difference between pressure in an inclined sand and the surrounding shale as function of inversion (PSSI and PWSI). The
fluid type.
latter yields both density and
Poissons ratio that can be used
empiricism is essential due to the fact that the relationship for overburden and fracture gradient calculation, respecbetween velocity and stress contains parameters that are not tively.
dimensionless, one must resort to basic physics for a meaningful prediction. We outlined a workflow that honors this Suggested reading. Seismic velocityA critique by Al-Chalabi
principleboth in data conditioning and rock model build- (First Break, 1994). Accounting for overpressure mechanisms
ing. Essentials of data conditioning are aimed at extracting besides undercompaction SPE Drilling and Completion by
propagation or rock velocity from seismic velocity analysis. Bowers (SPE, 1995). Shale compaction and abnormal pore presWe outlined some steps for doing that. We also discussed sures: A model of geopressures in the Gulf Coast Basin by
that pressure analysis at a regional, basin, or prospect scale Dutta (SEG 1983 Expanded Abstracts). Shale compaction, burshould be an integral part of exploration and high-grading ial diagenesis, and geopressures: A dynamic model, solution and
of portfolio. Low-frequency models of velocity may be ade- some results by Dutta (in Thermal Modeling in Sedimentary
quate for that. However, attempts should be made to use Basins, Editions Technip, 1986). Geopressure by Dutta
all the data (every trace at every time sample) and not at (GEOPHYSICS, 1987). Fluid flow in low permeable porous media
1538

THE LEADING EDGE

DECEMBER 2006
Downloaded 21 Nov 2009 to 64.91.188.154. Redistribution subject to SEG license or copyright; see Terms of Use at http://segdl.org/

Figure 16. (left) 3D velocity cube, error field associated with the velocity cube (STD), mean pore pressure in ppg, and standard deviation of pore pressure. (right) Extracted predicated pore pressure and its uncertainty at a well location.

by Dutta (in Migration of Hydrocarbons in Sedimentary Basins,


Editions Technip, 1987). Deepwater geohazard prediction using
prestack inversion of large offset P-wave data and rock model
by Dutta (TLE, 2002). Geopressure prediction using seismic
data: current status and the road ahead by Dutta (GEOPHYSICS,
2002). Fracture gradient predictions and its application in oil
field operations by Eaton (Journal of Petroleum Technology, 1969).
The movement and entrapment of petroleum fluids in the subsurface by England et al. (Journal of the Geological Society, 1987).
Practical application of the geostatistical scaling laws for data
integration by Frykman and Deutsch (Petrophysics, 2002).
Effects of porosity and clay content on wave velocities in sandstones by Han et al. (GEOPHYSICS, 1986). A new approach to
shale compaction and stratigraphic restoration, BeaufortMackenzie Basin and Mackenzie Corridor, northern Canada
by Issler (AAPG Bulletin, 1992). The effects of pore fluid pressure, confining pressure and pore fluid type on the acoustic properties of a suite of clean sandstones by
Khazanehdari et al. (American Association
of Drilling Engineers, 1998). 3D pore-pressure prediction and uncertainty analysis
by Lopez et al. (TLE, 2004). Model-based
inversion of AVO data using genetic algorithm by Mallick (G EOPHYSICS, 1995).
Shallow water flow prediction using
prestack waveform inversion of conventional 3D seismic data and rock modeling by Mallick and Dutta (TLE, 2002).
The effects of high temperatures and pressures on diagenesis and porosity destruction in reservoir sandstonesexamples
from the North Sea and Gulf of Thailand
by Osborne and Swarbrick (Indonesian
Petroleum Association, 1997). Predrill

pore pressure prediction using seismic data by Sayers et al.


(GEOPHYSICS, 2002). Pressure differences between overpressured sands and bounding shales of the Eugene Island 330 field
(Offshore Louisiana, USA) with implications for fluid flow
induced by sediment loading: Overpressures in Petroleum
Exploration by Stump et al. (Bull. Centre Rech, Elf Explor. Prod.,
Mem, 1998). Theoretical Soil Mechanics by Terzaghi (Wiley, 1943).
Seismic pressure-prediction method solves problem common
in deepwater Gulf of Mexico by Wilhelm et al. (Oil and Gas
Journal, 1998). TLE
Acknowledgments: The authors thank Schlumberger for permission to publish this work. We also thank Ran Bachrach and many other colleagues
for their input.
Corresponding authors: NDutta@houston.westerngeco.slb.com;
jkhazanehdari@slb.com

DECEMBER 2006

THE LEADING EDGE

Downloaded 21 Nov 2009 to 64.91.188.154. Redistribution subject to SEG license or copyright; see Terms of Use at http://segdl.org/

1539

You might also like