1527101

GEOPHYSICS, VOL. 67, NO. 6 (NOVEMBER-DECEMBER 2002); P. 20122041, 34 FIGS., 4 TABLES.
10.1190/1.1527101
In honor of the new millennium (Y2K), the SEG Research Committee is inviting a series of review articles and tutorials that
summarize the state-of-the-art in various areas of exploration geophysics. Further invited contributions will appear during
the next year or twoSven Treitel, Chairman, SEG Research Subcommittee on Y2K Tutorials and Review Articles.
Y2K Review Article
Geopressure prediction using seismic data: Current status
and the road ahead
N. C. Dutta
ABSTRACT
The subject of seismic detection of abnormally high-
pressured formations has received a great deal of atten-
tioninexplorationandproductiongeophysics becauseof
increasing exploration and production activities in fron-
tier areas (such as the deepwater) and a need to lower
cost without compromising safety and environment, and
manage risk and uncertainty associated with very ex-
pensive drilling. The purpose of this review is to capture
the best practice in this highly specialized discipline
and document it. Pressure prediction from seismic data
is based on fundamentals of science, especially those of
rock physics and seismic attribute analysis. Nonetheless,
since the rst seismic application in the 1960s, practition-
ers of the technology have relied increasingly on em-
piricism, and the fundamental limitations of the tools
applied to detect such hazardous formations were lost.
The most successful approach to seismic pressure pre-
diction is one that combines a good understanding of
rock properties of subsurface formations with the best
practicefor seismic velocity analysis appropriatefor rock
physics applications, not for stacking purposes. With the
step change that the industry has seen in the application
of the modern digital computing technology to solving
large-scale exploration and production problems using
seismic data, the detection of pressured formations can
now be made with more condence and better resolu-
tion. The challenge of the future is to break the com-
munication and the language barrier that still exists
between the seismologists, the rock physicists, and the
drilling community.
INTRODUCTION
Since the publication of SEG Reprint Series No. 7 (Dutta,
1987a), the landscape of the drilling and exploration ac-
tivity has changed dramatically. Encouraged by the prolic
reservoirs and the large size of the recent discoveries, most
exploration, production, and drilling activities are now in the
deepwater environment [water depths greater than 1500 ft
(450 m), as dened by the U.S. Minerals Management Service
(MMS) Shallow water ow evaluation: Current technology
and capabilities, in Shallow Water Flow: A Joint Indus-
try Forum, 1998]. In the annual budget of a typical explo-
ration and production (E&P) company, about 40% of the
E&P capital expenditure is spent in drilling. In the deepwa-
ter, drilling a well is very expensive, the average cost being
around $24 million (mostly due to high rig rates, approximately
$200 000 per day). Deepwater drilling is also very risky. In the
1999 Shallow Water Flows drilling conference in League City,
Texas, the industry reported that about 30% of deepwater well
Manuscript received by the Editor August 10, 2001; revised manuscript received April 29, 2002.
WesternGeco, Seismic Reservoir Services, 3600 Briarpark Drive, Houston, Texas 77042. E-mail: ndutta@houston.westerngeco.slb.com.
c 2002 Society of Exploration Geophysicists. All rights reserved.
costs was due to troubles, mainly related to geopressure, where
pore pressure of uids in the formation at a depth exceeded
that of the hydrostatic headof the uidat the same depth. Thus,
the subject of geopressure has received, and will continue to
receive, a great deal of attention, especially in deepwater.
As shown in Table 1, a recent survey conducted by MMS
(reported at the 1998 conference) suggests that by 2007, about
69% of the hydrocarbon production in the Gulf of Mexico
would be coming from offshore elds in the deepwater [water
depths greater than 2600 ft (790 m)]. Further, another study
by the MMS (at the 1999 conference) indicated that before the
end of 2001, offshore production from deepwater reservoirs
indeed exceeded that from shallow water for the rst time in
history. Internationally, the oil and gas industry also has had
a remarkable success in the deepwater clastic provinces off
the coasts of Angola and Nigeria. Thus, E&P activities in the
deepwater worldwide are continuing at a very fast pace. These
activities are expected to continue in the foreseeable future.
In fact, MMS (at the 1999 conference) projected that over the
2012
Downloaded 21 Jul 2009 to 136.250.230.21. Redistribution subject to SEG license or copyright; see Terms of Use at http://segdl.org/
Geopressure Prediction 2013
ve years begining in 2001, the industry will spend about $100
billion in E&P projects worldwide in the deepwater. These
are typically long-term projects (approximately, seven years
from exploration to the production phase), and much of the
monetary commitments has already beenmade by the industry.
The success of the industry in the deepwater exploration for
hydrocarbons is largely a result of our better understanding of
the deepwater geology, aided by the vast improvements in the
acquisition, processing, and interpretation of 3-D seismic data.
Signicant advancements in drilling technology and facilities
engineering, such as subsea completions and construction and
the usage of megarigs in deepwater, have also contributed to
the success of the petroleumindustry. Several rigs are currently
under constructionwhicharecapableof drillinginwater depths
greater than of 8000 ft (2400 m). In the reservoir description
and imaging areas, the availability of high-speed computers
and large-scale algorithms, such as the sophisticated reservoir
simulators and 3-D prestack depth imaging, played a very sig-
nicant role in building industrys condence in multibillion
dollar projects.
Prediction of geopressure before drilling is critical at several
stages in the exploration and development process. In the ex-
ploration phase, it can assist in assessing the seal effectiveness
and in mapping hydrocarbon migration pathways. It can also
assist in the analysis of trap congurations and basin geome-
try, and provide calibration for basin modeling. In the drilling
phase, an accurate pore-pressure prediction and the ability to
update and revise predictions quickly can be vital for safe and
economic drilling. For example, a major cause of cost overruns
in deepwater drilling is due to stuck pipes and lost circulation
of drilling uids, resulting in costly lost rig time. These are usu-
ally related to erroneous pore-pressure estimates, which give,
among other variables, the mud weights to be used in a given
well andcasingdepths towithholdtheformationpressurewhile
drilling. Estimates of proper geopressure and fracture pressure
(dened as the pressure at which tensile fractures are created)
are also essential for an optimized casing program design and
for avoiding well control problems, such as blowouts. In the
deepwater, often too conservative estimates for casing pro-
grams (mainly due to lack of reliable predrill pressure esti-
mates) lead to costly disappointments, such as not being able
to reach the intended target depth or having to redrill the well.
Before a well is drilled, especially infrontier areas suchas the
deepwater provinces, seismic data are the only available data.
Ever since the pioneering work by Pennebaker (1968), seismic
data have been used extensively for pore pressure analysis.
Many authors (e.g., Reynolds, 1970, 1973; Reynolds et al, 1971;
Bilgeri and Ademeno, 1982; Dutta, 1997) described how seis-
mic velocities can be used for geopressure analysis. The seismic
methods detect changes of interval velocities with depth from
a velocity analysis of common midpoint (CMP) seismic data.
Table 1. Percent of estimated production, deepwater U.S.
Gulf of Mexico.
Water depth (ft) 1997 2002 2007
0650 59 39 17
6502600 13 14 14
>2600 28 47 69
Source: MMS, Department of Interior, USA, 1998.
These methods exploit the fact that a geopressured formation
exhibits several of the following properties when compared
with a normally pressured section at the same depth: (1) higher
porosities, (2) lower bulk densities, (3) lower effective
stresses, (4) higher temperatures, (5) lower interval veloci-
ties, and (6) higher Poissons ratios. Each of these indicators
affects seismic interval velocities and reection amplitudes
which are the keys to seismic detection of geopressure.
Unfortunately, the seismic velocities, obtained from a stack-
ing velocity analysis of CMP gathers, have quite often been
misused for pore pressure analysis. For this, one needs rock
velocities. The rock velocity is dened as the velocity of sound
wave through a piece of a rock, a rock composite, or a partic-
ular rock formation containing pore uids, akin to, say, check-
shot, sonic log, or laboratory measurements. This can be, and
usually is, very different from the interval velocity obtained
from the stacking velocity. The stacking velocities were termed
processing velocities by Al-Chalabi (1973; 1994). He is right in
drawing attention to the fact that the purpose of the processing
velocities is to produce a stacked seismic section to highlight
the structural details. However, the interval velocities derived
from conventional stacking velocity analysis, without special
reprocessing, such as 3-D pre-stack depth imaging [including
dipmoveout (DMO)] andanisotropy processing of large-offset
reection data, usually do not resemble the rock velocities.
A routine use of the conventional stacking or processing ve-
locities, without aproper understandingof howtheseismic data
are acquired, processed, and interpreted and of the datas lim-
itations, can result in disastrous consequences in geopressure
analysis.
A simplistic geopressure prediction process using velocities
involves several steps: (1) obtainseismic velocities, (2) recondi-
tion and calibrate the velocities, (3) relate seismic velocities to
rock velocities, (4) construct a rock model that relates velocity
to effective stress and porosity, and (5) obtain effective stress
and pore and overburden pressures using the rock model and
the conditioned and / or calibrated seismic velocity. A major
purpose of this paper is to draw attention to the state of the
art for obtaining and conditioning seismic velocities for pres-
sure analysis. Even if the rock model (the one that relates rock
velocities to pore pressures) were perfect, there would remain
an uncertainty in the predicted pore pressure due to inherent
differences between the stacking or processing velocities and
the rock velocities.
Besides a stacking velocity analysis, the rock velocities can
also be obtained from inversion of traveltime (tomographic
inversion) and amplitudes of seismic data (poststack as well
as prestack), again with proper conditioning. Several papers
have dealt with this subject in the context of the pore-pressure
analysis work (Dutta and Ray, 1997; Mallick, 1999). These ve-
locities contain more delity than those from the conventional
stacking-velocity analysis. However, these are also inherently
nonunique and uncertain. This paper will also discuss this
subject.
There is a considerable source of misunderstanding in the
literature regarding the nomenclature used by seismologists to
describe various velocities and their failure to communicate
these properly to the drilling communitythe clients of the
seismologists in this endeavor. The nomenclature includes in-
terval and apparent velocity; average and root-mean-squared
(rms) velocity; stacking and migration velocity; instantaneous,
2014 Dutta
phase, and group velocity. Not all of these velocities are appro-
priate for pressure prediction.
Although I will describe various pressure mechanisms, a
comprehensive discussion of the subject and how indicators
other than seismic data can be used for pressure prediction are
beyond the scope of this paper. However, I will discuss briey
how seismic data, being a remote sensing tool, can be used to
predict pressure in real time, in conjunction with other indica-
tors, such as measurement-while-drilling/logging-while drilling
(MWD/LWD) logs, vertical seismic proles (VSP), and drill bit
seismic.
BASIC PRESSURE CONCEPTS AND ORIGIN
OF GEOPRESSURE
Denitions and pressure concept
Pore pressure or formation pressure, P, is dened as the
pressure acting on the uids in the pore space of a formation.
Hydrostatic pressure, P
h
, is the pressure caused by the weight
of a column of uid:
P
h
=
f
gz, (1)
where z,
f
and g are the height of the column, the uid den-
sity, and acceleration due to gravity, respectively. The size and
shape of the cross-section of the uid column have no effect on
hydrostatic pressure. The uid density depends on the uid
type, concentration of dissolved solids (i.e., salts and other
minerals) and gasses in the uid column, and the tempera-
ture and pressure. Thus, in any given area, the uid density
is depth dependent. In the SI system, the unit of pressure is
Pascal (abbreviated by Pa), and in the British system, the unit
is pounds per square inch (abbreviated by psi). We note that
1 Pa =1.45 10
4
psi =1 N/m
2
. This is a rather small unit and
for most practical applications, it is customary to use megapas-
cals (MPa), where 1 MPa =10
6
N/m
2
.
The formation pressure gradient, expressed usually in
pounds per square inch per foot (abbreviated by psi/ft) in the
British system of units, is the ratio of the formation pressure, P
(in psi) to the depth, z (in feet). It is not the true instantaneous
gradient, dP/dz. In general, the hydrostatic pressure gradient,
P
g
(in psi/ft), can be dened by
P
g
= 0.433 uid density (in g/cm
3
). (2)
We note that 1 psi/ft =0.0225 MPa/m. The hydrostatic pressure
gradient of 0.465 psi/ft (0.0105 MPa/m), typical for the offshore
Gulf of Mexico, assumes a salt concentration of 80 000 ppm of
NaCl at 77
F.
The overburden pressure, S(z), at any depth is the pressure
which results fromthe combined weight of the rock matrix and
the uids in the pore space overlying the formation of interest.
This is expressed as
S = g
z
0
b
(z) dz, (3)
where
b
is the depth dependent bulk density given by
b
=
f
+(1 )
g
, (4)
where ,
f
, and
g
are the fractional porosity, the pore uid
density, and the density of the matrix (grain density), respec-
tively. The overburden pressure is depth dependent and in-
creases with depth. In the literature, the overburden pressure
has also been referred to as the geostatic or lithostatic pressure.
Table 2 contains a brief summary of how overburden pres-
sures may be computed when the available data are either
inadequate or incomplete, such as a missing density log or un-
reliable or inadequate analog well data for a particular basin.
The effective pressure or differential pressure, , is the pres-
sure, which is acting on the solid rock framework. Accord-
ing to Terzaghis principle (Terzaghi, 1943), it is dened as the
difference between the overburden pressure, S, and the pore
pressure, P:
= S P. (5)
It is that controls the compaction process of sedimentary
rocks; any condition at depth that causes a reduction in
will also reduce the compaction rate and result in geopressure.
Figure 1 shows an example of a typical pressure versus depth
prole in a clastic sequence. This prole is typical for the
Table 2. Computation of overburden pressure.
Sea level to sea bed Depends on salinity of water
column (1.011.05 g/cm
2
)
Sea bed to top of Density log from offset well
logged interval Density-transit time relations plus
velocities seismic velocities
Regional trends
Top of logged interval Density log
to total depth Density log from offset well
Density-transit time relations plus
seismic velocities
Intervals of poor Density log from offset well
quality density log Density-transit time relations plus
seismic velocities
Frontier areas or deeper Density-transit time relations plus
intervals where no well seismic velocities
data are available Regional trends
FIG. 1. A typical pressure prole in a clastic basin. Top of hard
pressure is dened as that pressure where effective stress is
1000 psi. When effective stress reaches that value, the likeli-
hood of seal failure increases considerably. The pore pressure
gradient at seal failure limit is higher than 0.8 psi/ft.
Louisiana and Texas Shelf, offshore Gulf of Mexico. The tran-
sition from the hydrostatic interval to the overpressured inter-
val in these areas is fairly well dened and can range from
a few hundred to several thousand feet in thickness. How-
ever, drilling experiences in the Pliocene-Pleistocene forma-
tions of the deep water Gulf of Mexico [water depth greater
than 2 600 ft (790 m)] have shown that in these areas the tran-
sition zone is usually not well developed, and the pore pres-
sures are usually higher than the hydrostatic pressure at shal-
low depths and continue to build up gradually with depth (see
Figure 2), withoccasional occurrences of pressure reversal. The
variables that are required for prediction and assigning risks
for prospectivity are:
1) Depth of the top of the overpressured zone.
2) Depth of the top of the hard-pressure zone (dened as
a lower limit of 1000 Psi effective stress).
3) Seal failure limit.
4) Shape of the effective stress versus depth prole.
Although there is no universally accepted scale expressing the
degree of geopressuring, the nomenclature introduced for the
Gulf of Mexicos tertiaryclastics byDutta(1987a, his Table1on
page 5) will be used in the present work. The effective pressure,
, plays a key role in the model discussed here. Note that geo-
pressuring implies a low effective stress and a higher porosity,
higher than what one would obtained had the rock compacted
normally (pore pressure remaining hydrostatic while undergo-
ing burial and compaction). A low and high tend to lower
the rock velocity. Most authors use relationships that relate
the velocity to the effective stress, porosity, and lithology, and
hence, pore pressure.
FIG. 2. A typical pressure versus depth prole in the deep-
water basins. Note that at shallow depths, the pore, fracture,
and overburden pressures are close to each other. This poses a
considerable drilling hazard (shallow water ow sands) in the
deepwater.
Table 3 contains denitions of some equivalent terms used
in the literature. Note that many practitioners use the density
of circulating mud (referred to as mud weight in the indus-
try and recorded by the mud logger on the rig) as a substi-
tute for pore-pressure, especially in the calibration phase of a
pore-pressure prediction project. This is a bad practice. Mud
weights are usually higher than the true formation pressure by
unknown amounts and are unsuitable for pressure calibration.
Further, the drilling community uses a terminology related to
mud weights that geophysicists should be aware of. Some of
the more commonly used terms are given in Table 4.
Table 3. Denitions.
Normal pressure (hydrostatic pressure or normal uid pres-
sure) is the pressure exerted by a static column of water of the
same height as the overlying pore uids and the same density
as the pore water.
Normal pressure = pressure gradient of water depth
Pore pressure (uid pressure or formation pressure) is the
pressure exerted by the pore uids. Units: psi/ft 19.268 =
ppg (pounds per gallon); ppg 0.0519 = psi/ft; g/cm
3
0.433 = psi/ft.
Pore pressure = normal pressure + over/underpressure
Overpressure (geopressure) is the excess pressure above nor-
mal pressure.
Overpressure = pore pressure normal pressure
Overburden pressure (lithostatic pressure or geostatic pres-
sure) is the pressure exerted by the overlying pore uid and
rocks.
Overburden pressure = overburden gradient depth
Terzaghis relationship (net differential stress or net overbur-
den stress or net conning stress) states that the total stress is
jointly supported by the pore uid and the rock matrix.
S = overburden pressure = pore pressure + effective
pressure = P +
Effective vertical stress is the stress applied to the rock matrix.
Effective pressure () = overburden pressure pore pres-
sure = S P
Buoyant pressure is the excess pressure created in conned
reservoir by the density difference between hydrocarbons and
water
P = (water gradient hydrocarbongradient) height of
hydrocarbon column
Pore pressure = normal pressure + overpressure (brine-
lled) + buoyancy pressure.
Table 4. Drilling terminology.
Balanced drilling is when the hydrostatic pressure of the mud
column is equal to the formation pressure. It seldom happens.
Overbalanced drilling is when the hydrostatic pressure of the
mud column is greater than the formation pore pressure. The
usual case.
Underbalanced drilling is when the hydrostatic pressure of the
mudcolumnis less thantheformationpressure. Never planned,
but it happensespecially in thick shales!
Differential pressure is the difference between the hydrostatic
pressure exerted by the mud column and the formation pres-
sure. Do not confuse with the effective stress.
Effective circulating density is the apparent increased mud den-
sity of a circulating mud column. When the mud is circulated,
additional pressure (backpressure) is placed against the for-
mation due to frictional effects in the mud column. This is en-
gineering terminology.
2016 Dutta
Origin of geopressure
A comprehensive discussion of various geopressure mecha-
nisms is contained in Fertl (1976) and Dutta (1987a, chapter 2).
Development of geopressure indicates that uid movement is
retarded both vertically and laterally. This can be caused by a
rapid burial of low-permeability sediments (rapid enough to
prevent compaction water to leave the system), or expansion
of pore uid from diagenetic and lithologic changes, or both.
Some of the important mechanisms related to geopressure are:
1) Mechanical compaction disequilibrium (Hubbert and
Rubey, 1959).
2) Clay dehydration and alteration due to burial diagenesis
(e.g., see Dutta, 1987a, chapter 2).
3) Dippingor lenticular permeablebeds embeddedinshales
(e.g., see Fertl, 1976).
4) Buoyancy (e.g., see Fertl, 1976).
5) Tectonism/uplift and erosion (e.g., see Dutta, 1987a,
chapter 2).
6) Aquathermal pressuring (e.g., see Barker, 1972; Dutta,
1987a, chapter 2).
Mechanical compaction disequilibrium is believed to be the
primary cause of geopressure in sedimentary basins (Hubert
and Rubey, 1959). There, low-permeability sediments (such as
high-porosity clay) accumulate at rates faster than they can
dewater and compact due to gravity. Therefore, the pore uid
is forced to support some portion of the combined weight of
the overlying rocks and uids. The magnitude of the overpres-
sure due to mechanical disequilibrium and the depth where
the pressure deviates from hydrostatic condition are dictated
by the burial history of the sediments and the hydraulic com-
munication with neighboring formations and faults.
Smectite or swelling clay is a common component of clay,
especially in the Gulf of Mexico. It contains a considerable
amount of water bound in the clay platelets. At 150250
F,
smectite begins to dehydrate and transformto illite, with potas-
sium feldspar acting as a catalytic agent. This process releases
interlayer bound water from the smectite into free pore water,
andcauses anadditional increase inpore pressure anddecrease
of effective stress. In addition, the process also causes a long-
range ordering of clay platelets and redistribution of effective
stress (Hower et al., 1976). The interlayering of smectite and
illite layers is characterized by an ordering parameter RN. RN
means that at least N illite layers separate every pair of smec-
tite layers. Hower et al. (1976) noted this ordering as being
gradual. Some details of this ordering phenomenon are given
in Dutta (1987a). The entire process is kinetic in nature and
controlled by the temperature and time history of burial of the
sediments.
The third mechanism, due to ow of uids through dipping
or lenticular sands, is a local phenomenon. When these sands
are embedded in geopressured shales, they provide a very ef-
cient mechanism by which high-pressured shales can dewater,
especially if the sand has a considerable structural relief. This is
because the sand has much higher permeability than the encas-
ing shales which, inturn, causes the pore uidpressure gradient
within the sand to be hydrostatic. This causes a transmittal of
high pressure from the downdip positions of the sands to their
crestal positions. This can also happen when sands are juxta-
posed by faulting. This phenomenon, in principle, can cause
lateral changes in the velocity eld around the sand with a
signicant structural relief. In the literature this has also been
termed as the centroid phenomenon (Traugott, 1997).
Oil and gas are lighter than water. Hence, when water is
replaced by hydrocarbons, it results in an increase in the pore
pressure at the reservoir level due to buoyancy effect. The key
factors that dictate the amount of overpressuring caused by
this mechanism are the density of hydrocarbons, the height of
the column of hydrocarbons, and the density of pore water at
that depth.
In provinces such as Orinoco Delta, Venezuela, Trinidad,
Sumatra, and California, tectonic compression and shear (tec-
tonic wrench faulting) can cause signicant geopressure. Rapid
uplifting and erosion of a sealed compartment, while maintain-
ing the pore uid pressure within the compartment, will also
cause pore pressure tobe abnormally highat its depthof burial.
Seismic detection of pore pressure due to these mechanisms is
very difcult.
Aquathermal pressuring has been a much-discussed phe-
nomenon in the literature. It is also controversial. This phe-
nomenon is due to the fact that the coefcients of thermal
expansion of pore uids are greater than the corresponding
coefcients of the rock matrix. Consequently, when a com-
partment is sealed and buried, the pore uids are subjected to
a high pressure due to the pore uid volume expansion caused
by increased temperature. However, there is considerable dis-
agreement as to how perfect the seal must be for this mecha-
nism to signicantly contribute to high pore pressure.
As far as the seismic detection of abnormally high pore pres-
sure is concerned, the rst two mechanisms (mechanical com-
paction disequilibrium and clay dehydration) can be detected
relatively easily compared to the remaining mechanisms. This
is because the compaction properties of sediments dealing with
both mechanical disequilibrium and clay dehydration are rel-
atively well behaved and easier to describe and model using
well-known rock physics and seismic principles (see below).
Further, these mechanisms can be reconciled with observa-
tions easier relative to the other mechanisms. However, there
still remains a certain ambiguity; the contribution from the
other mechanisms can never be ruled out completely.
Subsurface stress
Under hydrostatic stress, water within the pore spaces of the
rocks is connected to water in the sediments and in the sea
above. At low rates of sedimentation, it is possible for water
to be expelled at a rate adequate to maintain the hydrostatic
equilibrium. However, at rapid burial rates with relatively im-
permeable shales, this equilibrium is not maintained; the uid
motion is retarded, and the pore uid begins to support the
overburden, resulting in pressure increase.
Terzaghi (1943) describes the state of subsurface stress fairly
adequately. According to him, the vertical overburden stress,
S, is the sumof two quantities: the formation (pore) uid stress,
P, and the vertical effective stress acting on the rock frame,
v
.
Thus, as stated in equation (5),
S =
v
+ P. (6)
For example, if we assume S/z =1.0 psi/ft, and P/z =0.46,
where z is depth in feet, then
v
/z =0.54 psi/ft. However, if
P/z =0.70 psi/ft and S/z remains 1.0 psi/ft, then
v
/z becomes
0.30 psi/ft. Therefore, geopressuring (caused by compaction
and trapping of uid) causes a reduction in rock frame stress.
Note that for this case, the effective stress increases linearly
with depth at a rate of approximately 0.54 psi/ft, until pore
pressure becomes abnormally high. At this point, the effec-
tive stress decreases, causing a decrease in the velocity and the
density of the rock. If the pore pressure, P, approaches the
overburden pressure, S, then
v
approaches zero, and the seal
failure occurs via creation of open fractures (tensile fractures).
This is known as hydraulic seal failure, and its likelihood of
occurrence provides a risking criterion for a prospect.
The horizontal effective stress,
h
, is related to
v
through
Poissons ratio, . According to Eaton (1968),
h
=

1
v
. (7)
For poorly consolidated silty shales, 0.4, and
h
0.67
v
.
Thus, according to Eaton (1968), tensile fractures are likely to
occur before extensional fractures.
I will drop the sufx v in
v
and use to denote vertical
effective stress unless otherwise noted. Inthe following section,
I present a brief discussion on how the effective stress can be
related to porosity and velocity.
ROCK VELOCITY, POROSITY, AND EFFECTIVE STRESS
Rock velocity
The velocity of sound (compressional, seismic, or P) waves
propagating through a piece of rock is called the rock veloc-
ity. The rock velocity depends on many parameters: porosity,
uid saturation, state of stress, pore and conning stress, pore
structure, temperature, pore uid type and its thermodynamic
state, lithology, clay content, cementation, and frequency of
the propagating waves. Furthermore, these parameters are not
independent of each other. A full description of these depen-
dencies can be obtained from controlled laboratory measure-
ments. In Nur and Wang (1989, 1992), the reader can nd a
collection of some pertinent papers that deal with this com-
plex subject. Below, I point out a few salient features relevant
to my discussions. Obviously, the lithology dependence of rock
velocity is a signicant point because gross lithologies (possi-
bly) exhibit large P-wave velocity distributions. However, the
velocity is not a unique parameter to identify lithology. Two
different lithologic rocks under identical environmental con-
ditions can indeed have the same velocity. Similarly, a given
rock can have different velocities, depending upon its environ-
mental conditions. A few comments are noteworthy for the
deepwater setting:
1) The distributions of sandstone and shale velocities over-
lap considerably. In the deepwater settings, for normally
compacted sediments, often the sand and shale velocities
overlap completely. On the other hand, the shale bulk
densities are typically higher (by about 0.1 g/cm
3
) than
the corresponding sand densities at a given environmen-
tal condition.
2) Rock velocities in overpressured zones are lower than
those found at hydrostatic pressures. Low velocities in
highly pressured zones are often related to thick shale
sequences that have low sand layer occurrence (less than
10% of total thickness). This is most common in uncon-
solidated and relatively young sediments (e.g., Gulf of
Mexicos Tertiary sediments).
3) In shallow water sediments, sands and shales often have
locally identiable velocity-versus-depth trends. Such
trends are very useful for pressure analysis. However,
in deepwater sediments, such trends are usually absent.
Pore pressures higher than normal often happen in these
sediments just below the mudline; for example, about
10003000 ft (300900 m) below the sea bed in the
Mississippi Canyon area. In deepwater settings, these
pressured sands are known to be hazardous, and drilling
through them has cost the industry an enormous sum
of money. This is known as the shallow water sand ow
problem, and the reader is referred to the recent confer-
ence proceedings on this subject (Minerals Management
Service, Shallow Water Flow: A Joint Industry Forum,
1998).
Although laboratory measurements have added much to our
understanding of rock properties, such measurements of ve-
locities on core samples are plagued by one problemthat of
restricted sampling. Rock velocities appropriate for large-scale
geologic formations quite often differ from those measured on
small core samples. This is referredtoas the upscaling problem-
a discussion of which is beyond the scope of the present work.
There is another issue with this type of measurement: fre-
quency. How can one reconcile data from ultrasonic measure-
ments (10
6
Hz) withthose fromseismic frequency bandwidth
(10100 Hz)? Thus, proper scaling and calibration of velocities
are critical to success for geopressure prediction.
A direct source of the velocity measurements appropriate at
seismic frequencies is a checkshot survey (or a zero-offset VSP
survey). In this type of measurement, the sampling interval is
of the order of 1001000 ft (30300 m), and the sources employ
frequency bandwidths of 10100 Hz, which overlap the explo-
ration seismic frequency bandwidth. However, velocity details
of the lithologic variations are grossly smoothed in the data as
is the case with the seismic waves.
Acoustic logs are a direct source of in-situ rock velocity data.
Measurements are obtained at a much higher frequency (about
510 kHz). These measurements are carried out using sources
and receivers in a borehole, and the data are obtained from re-
fraction of sound waves along the interface between the bore-
hole and the drilling uid. There, the sampling intervals are
of the order of several feet. The velocity information obtained
from sonic logs is most frequently used for pressure analysis,
especially for the calibration purpose. Further, these measure-
ments are the only ones that can practically yield rock veloc-
ity data on shales under in-situ conditions. This is particularly
important when we note that every subsurface pressure analy-
sis technique is invariably carried out on shales. Hottman and
Johnson (1965) clearly noted in their paper that the shale prop-
erties (velocity, porosity, etc.) are more consistent indicators of
pressure than those for sandstone, and hence are ideally suited
for pressure analysis.
Relationship of porosity and effective stress
It is well known that a compaction process causes reduction
of porosity. Before relating the velocity to the effective stress,
2018 Dutta
and hence the pore pressure via Terzaghis relationship, we
must understand the compaction process. This means relating
the porosity to the effective stress for a given rock type under
hydrostatic conditions.
Porosity is the most direct measure of compaction. Athy
(1930) put forward an exponential function relating porosity
to depth under normal pressure conditions:
(z) =
0
e
cz
, (8)
where (z) is porosity at depth z,
0
is the porosity at z =0,
and c is a constant. Since its inception, many authors suggested
several variations of the above model. Magara (1978) made a
compilation of these models. Following Rubey and Hubbert
(1959), Smith (1971) and Dutta (1983, 1987b), Athys com-
paction equation can be recast as
=
0
e
K
, (9)
where the coefcient K is related to the bulk density of the
sediments and the density of pore water, and is the effective
stress. Equation (9) is termed the exponential relationship by
Stump et al. (1998). Taking the natural logarithm of both sides,
one nds
ln = ln
0
K. (10)
Equation (10) represents a straight line of ln versus . Dutta
(1987b), and also other published constitutive relationships be-
tween porosity and effective stress are:
Terzaghi (1943): = 1
0
C/4.606
, (11)
Palciauskas and Domenico (1989): = 1
0
e
,
where is constant, (12)
Dutta (1987b): =
0
e
(T)
. (13)
In equation (13), is void ratio given by
= /(1 ), (14)
and T is temperature, (T) is a functionof temperature (Dutta,
1987b) and
0
is a constant. The coefcients C, (T), and
0
depend on lithology.
Once the effective stress, , is known from the porosity, ,
using one of the porosity versus effective stress relationships
described above, the pore pressure is easily obtained from
Terzaghis relationship in equation (6). It is noted that the con-
stitutive relations of the type described above are not unique;
these are usually obtained by curve tting a large set of data
from wireline logs. Nonetheless, the user must try to under-
stand the underlying assumptions behind these empirical com-
paction relationships because these constitute the foundation
of a rock model needed for pore pressure analysis. For exam-
ple, the relationship given in equation (13) employs explicit
thermal dependency of shale compaction. This is based on the
assumption that as shales are buried, compaction process not
only causes porosity reduction as in equations (11)(12), but
also shale mineralogic changes and, hence, changes in the ef-
fective stress due to increased temperature.
Relationship of rock velocity and porosity
The literature is full of relationships between porosity (or
bulk density) and velocity of a given rock type (Nur and Wang,
1989). However, very little is published about such relation-
ships for shales, where geopressuring occurs. For sonic logs,
the Pickett-type equation (Pickett, 1963) is frequently used:
t = A
0
+ B
0
, (15)
where t is the sonic interval transit time (slowness) (usu-
ally expressed in microseconds per foot), and A
0
and B
0
are
lithology-dependent constants.
Another commonlyusedrelationshipis that of Gardner et al.
(1974). This is usually expressed as
b
= c
V
d
, (16)
where V is interval velocity, and c
and d
are lithology-
dependent constants.
Note that the well-known Wyllie time-average equation
(Wyllie et al., 1956; Gardner et al., 1974) relating porosity to
transit time, namely,
t = t
f
+(1 )t
m
, (17)
where t
f
and t
m
are the transit times through the uid and
solid phases, respectively, is a special case of the Pickett-type
equation.
Issler (1992) published a very useful porosity-sonic transit-
time equation for shales and mudstones based on logs and core
measurements. It is an extension of a similar relationship pub-
lishedearlier by Raiga-Clemenceauet al. (1988). Isslers (1992)
equation is
t = t
m
(1 )
x
, (18)
where t
m
is the travel time for the matrix, t is the log-derived
traveltime, and x is an acoustic formation factor dependent on
lithology. This is very similar to Archies resistivity equation.
The resistivity of uid-lled porous rocks, R, is expressed in
terms of , resistivity of water, R
w
, and formation factor, F:
R = FR
w
= a
m
R
w
. (19)
Here, a and m are lithology dependent constants (Archie,
1942).
Vernik (1994) developed velocity-porosity curves for four
general lithology types: (1) cleanarenites (<2%clay), (2) aren-
ites (215%clay), (3) wackes (1535%clay), andshales (>35%
clay). According to Bowers (1995), Gardner et al.s (1974) rela-
tion, equation (16), is an approximate upper bound for poorly
consolidated shale velocity as a function of bulk density. It is
indeed violated by well-lithied shales.
Relationship of velocity or transit time and effective stress
All seismic methods for pressure prediction use either ex-
plicitly or implicitly a relation between the rock velocity and
the effective pressure. A typical example of the effect of an
external pressure and the differential pressure (taken to be the
difference between the external or conning pressure and pore
uid pressure as a rst approximation) on the velocity of Berea
sandstone is shown in Figure 3. The gure shows that:
1) The velocity increases with the effective stress (P in
Figure 3). The increase is at rst rapid, but becomes grad-
ual until an approximately constant terminal velocity is
attained.
2) For the effective stress, the velocity at rst increases by a
small amount rapidly and then becomes almost constant.
3) For small effective stresses, the behavior is very similar
to that at the zero effective stress. At higher differentials,
the initial velocity rise is not observed.
4) Further, different samples of equal porosity from the
same rock, or the same sample usedrepeatedly, may show
a variation in measured velocity for a xed low external
pressure. However, the terminal velocity inall these cases
is the same (not shown in Figure 3).
All of the above observations made by Wyllie (1957) are also
valid in deepwater settings, although velocity values are quite
different [lower than the values given by Wyllie (1957)]. Rela-
tionships such as those in Figure 3 are extremely useful for de-
veloping an expression relating the effective stress to the rock
velocity. For applications to deepwater sediments, note that
the nonlinear portions of these curves at low effective stresses
are the most relevant to pore pressures at deeper depths. Fur-
ther, shallow water ow (SWF) zones occur for very low effec-
tive stressesabout 100300 psi is very typical. This is also in
the nonlinear portion of the curve of Figure 3.
FIG. 3. Velocities through an oil-wet sandstone at varying con-
ning and pore pressure conditions (from Wyllie, 1957) show-
ing clearly that the rock velocities are directly dependent on
the effective pressure (difference between the conning and
the pore pressure).
Because of its importance in deepwater drilling, especially
for very shallow drilling beneath the mudline and for very
deep water hazards associated with geopressure, Prasad (2002)
has recently carried out high-quality acoustic measurements
of compressional and shear-wave velocities of unconsolidated
sediments at low effective stresses. As the effective stress ap-
proaches low values, the P-wave velocity approaches that of
suspensions in a uid (water velocity), while the S-wave ve-
locity approaches zero. Thus, as sediments pass through the
transition zone from load bearing to a suspension [typically,
up to 2000 ft (600 m) below the mudline], the V
p
/V
s
ratio
is expected to rise almost exponentially. Figure 4 shows just
that: the V
p
/V
s
ratios for water-lled sediments (the upper
curve) are found to be as high as 20 (Poissons ratio of 0.499)
as the effective stress approaches values less than 0.2 Mpa.
This result will have a signicant impact not only on the
SWF prediction by seismic techniques, but also for deepwa-
ter hazards, where prestack inversion of P-wave seismic data
for Poissons ratio is used as a pressure indicator (see below).
Recently, Huffman and Castagna (2001) also made very similar
observations.
For geopressure estimation using velocities, we need a rela-
tionship between velocity and effective stress for a given lithol-
ogy. This can be obtained simply by eliminating porosity (or
bulk density) between the relationships described above: ve-
locity (or transit time) versus porosity (or bulk density), and
porosity (or bulk density) versus effective stress. For example,
using Isslers (1992) relation, equation (18), and the exponen-
tial relationship of equation (9), one obtains the following re-
lationship that relates transit time to effective stress directly:
= 1/k ln
1)
, (20)
FIG. 4. Laboratory measurements of the Poissons ratios of soft
clastics as a function of effective pressure (M. Prasad personal
communication, 2000). Note that at very high pore pressures,
Poissons ratios approach high values, close to that appropri-
ate for water (0.5). Poissons ratio is a good indicator of pore
pressure and SWF sands.
2020 Dutta
where =t /t
m
, and =1/x. Thus, velocity data (or )
from seismic data, after processing to recover the rock veloci-
ties, can be used as input to equation (20) to derive the effective
stress. The overburden stress can be calculated, again using ve-
locity in equation (18), to derive the porosity, and hence, the
bulk density of equation (4). The corresponding pore pressure
is givenby subtracting fromthe overburdenpressure as given
in equation (6).
Empirical relations
There are several empirical approaches to pressure predic-
tion using velocity (Hottman and Johnson, 1965; Mathew and
Kelly, 1967; Eaton, 1972). The essential steps are dened as
follows:
1) Perform either a standard or, preferably, a high-
resolution velocity analysis (VA) (see below) and obtain
interval velocity. The interval velocity versus depth plot
is usually expressed as interval transit time versus depth
on a semilog scale.
2) Construct a normal compaction trend line for the area
under investigation and overlay this trend line on top of
the seismically derived acoustic log. Some examples are
shown in Figure of Pennebaker (1968, his Figures 13).
Here, normal compaction trend denotes a depth versus
velocity relationship for a rock which has compacted un-
der hydrostatic pressure conditions.
3) The deviation, (t ), of the seismically derived interval
velocity at a given depth from the normal compaction
trend is used as an indicator of geopressuring. For this
purpose, one uses the available calibration charts which
relates (t ) topore pressure gradients. Dutta (1987a, his
Figure 3.2) shows examples of these calibration charts for
several geologic provinces. Of course, this procedure tac-
itly assumes that the interval velocities indicate porosity
(andcompaction) variationrather thanvariationinlithol-
ogy. These calibration charts are usually obtained by cor-
relating sonic logs to pressures derived from repeat for-
mation tester (RFT) or measurement on drillstem tester
(MDT) logs.
Apopular empirical approach is Eatons method (Eaton, 1968,
1972), which relates pore pressure to interval transit time. The
pertinent equation for seismic or sonic logs is
P = S (S P
h
)(t /t
n
)
3.0
, (21)
where t
n
is the slowness for normally compacted rock. This
is derived from analyses of sonic logs from wells in the strati-
graphic section where pressures are known to be hydrostatic.
Approaches such as these are good as long as a normal com-
paction trend can be constructed. However, if the sedimenta-
tion rates are very high, as in the deepwater Gulf of Mexico,
the normal compaction trends are usually absent. This is be-
cause the sediments under these conditions are overpressured
at very shallow depths below mudline. Any attempt to derive
a normal compaction trend line, by tting a couple of pressure
measurements or even mud weights by trial-and-error, must
be viewed with considerable skepticism. Such trend lines are
likely to vary from well to well, even in the same minibasin
with the same geology and rocks. Further, such normal com-
paction trends often have unphysical limits of the velocities
when porosities of sediments are either close to or higher than
the critical porosities as well as when porosities approach
zero. Here, the critical porosity is dened as being the porosity
at which a uid-lled porous rock behaves more like a suspen-
sion than a compacted sediment or a rock (Vernik, 1994).
SOURCES OF SEISMIC VELOCITY
FOR GEOPRESSURE ANALYSIS
The preceding section dealt with the relationships between
the rock velocity and the effective stress as well as between the
rock velocity and porosity. The second relationship yields bulk
density and, hence, overburdens stress from velocity by as-
suming values for grain and pore-uid densities. The rst rela-
tionship then yields the pore pressure upon subtraction from
the overburden pressure. Thus, velocity provides information
for both pore pressure and effective pressure. Before a well is
drilled, seismic is often the only source of the velocity. In this
section, I review various techniques to obtain seismic veloci-
ties, with emphasis on the ways to condition these velocities for
geopressure analysis. Some of the techniques used for this pur-
pose are Dix-velocity inversion (Dix, 1955), migration velocity
analysis, tomographic velocity inversion, poststack impedance
inversion, and prestack inversion.
Dix velocity inversion
Dating back to the pioneering work by Pennebaker (1968),
it has been known that the conventional stacking or process-
ing velocities obtained from an analysis of CMP recording and
processing of seismic data can provide an abundant source of
velocities for geopressure prediction. These velocities are ob-
tained from an analysis of the traveltime equation valid for the
horizontally stratied earth model (Taner and Koehler, 1969):
t
2
(x) = c
0
+c
1
x
2
+c
2
x
4
+. . . . , (22)
where t (x) denotes traveltime at offset x, and t (0) is the trav-
eltime for x =0. The coefcients in equation (22) are given
by
c
0
= t
2
(0), (23)
and
c
1
= 1/V
2
rms
, (24)
where V
rms
is the root-mean-square velocity, and c
2
, c
3
, etc. are
complicatedfunctions that dependonlayer thickness andinter-
val velocities. Figure 5 describes a simple horizontally stratied
earth and a ray path through this medium emanating from a
source, S , and captured at a receiver, R. The interval velocity
and the travel time for the Nth layer are denoted by V
N
and
t
N
, respectively. V
rms
down to the point D is dened as
V
2
rms
= t (0)
1
N
i =1
V
2
i
t
i
(0), (25)
where t
i
(0) is the vertical two-way time through the i th layer
and
t (0) =
N
i =1
t
i
(0). (26)
A further approximation to the traveltime equation given in
equation (22) can be realized by assuming that the offset is
small compared to the depth (small-spread approximation).
In this case, the series in equation (22) can be truncated as
follows:
t
2
(x) = t
2
(0) + x
2
/V
2
rms
. (27)
This may be compared with the well-known NMO equation:
t
2
(x) = t
2
(0) + x
2
/V
2
NMO
. (28)
Comparing equation (28) with equation (27), we realize that
the velocity required for NMO correction for a horizontally
stratied medium is equal to the rms velocity, provided the
small-spread approximation is made. The hyperbolic moveout
velocity should be distinguished fromthe stacking velocity that
optimally allows stacking of traces in a CMP gather. The hy-
perbolic form is used to dene the best stacking path:
t
2
st
(x) = t
2
st
(0) + x
2
/V
2
st
, (29)
where v
st
is the velocity that allows the best t of the traveltime
curve on a CMPgather to a hyperbola within the spread length.
This hyperbola is not necessarily the small-spread hyperbola
implied by either equation (27) or equation (28). The differ-
ences can be signicant as shown in Figure 6. In this gure,
label (a) denotes the actual traveltime, (b) denotes the best-t
hyberbola over the full offset range, and (c) is the small-spread
hyberbola. The difference between the stacking velocity, V
st
,
and the NMO velocity, V
NMO
, is called the spread-length bias
and should be accounted for in any velocity analysis software.
For the larger acquisition cables employed these days, higher
order terms in equation (22) become important and should be
included as a part of the velocity analysis. It is obvious that
doing so implies that the moveout or the traveltime equation
is no longer hyperbolic.
Equation (29) contains the basis for velocity analysis for a
CMP gather. This describes a straight line on the t
2
versus x
2
plane. The slope of the line is 1/V
2
st
, and the intercept value at
FIG. 5. NMO for a horizontally stratied earth model.
x =0 is t (0). Inpractice, a least-squares methodis usedtodothe
curve tting. Equation (25) yields interval velocities, V
i
, from
the rms velocities. This process is called velocity inversion. The
traditional form of the Dixs interval velocity equation for the
Nth layer in terms of the rms velocities and travel times (valid
for horizontal layers) is
V
N
=
V
2
rms,N
t
N
V
2
rms,N1
t
N1
t
N
t
N1
. (30)
Here, V
N
is the interval velocity of the Nth layer, V
rms,N
is the
rms velocity of the Nth layer and t
N
is the traveltime to the Nth
layer.
Proper stacking of a dipping event requires a velocity that
is greater than the velocity of the medium above the reector.
This suggests that a horizontal layer with a high velocity can
yield the same moveout as a dipping layer with a low veloc-
ity and, hence, can yield the same stacks in small-spread ap-
proximation. This ambiguity can result in a different stacking
velocity function, for seemingly similar looking stacks! Modern
processing techniques involving dip moveout (DMO) usually
account for this effect. Therefore, any pressure analysis work
should be carried out on migrated and DMO-processed data.
A standard display for continuous velocity analysis is shown
in Figure 7. Shown on the left in Figure 7 is the typical sem-
blance analysis window. It shows how various velocity (hor-
izontal axis) functions are used to dene the best-t hyper-
bola at a particular traveltime (vertical axis). Also shown are
the best-t rms and the interval velocity functions dened in
equation (25). The gathers inthe middle are displayedfor qual-
ity control. These gathers indicate seismic events, which are
attened to estimate the interval velocities shown on the left
panel. This is anessential displayduringvelocityanalysis; it sug-
gests how at the events are and how much freedom one has
in changing the velocity eld, within the signal-to-noise (S/N)
FIG. 6. Errors associated with NMO assumptions. The NMO
approximation assumes a small-spread hyperbola. On the
other hand, the stacking velocity is derived from the best-t
hyperbola over the entire offset range. Here, (a) is the actual
traveltime, (b) is the best-t hyperbola over the full offset range
OA, and (c) is the small-spread hyperbola.
2022 Dutta
F
I
G
.
7
.
P
a
n
e
l
s
s
h
o
w
i
n
g
s
t
a
n
d
a
r
d
v
e
l
o
c
i
t
y
a
n
a
l
y
s
i
s
.
T
h
e
p
a
n
e
l
o
n
t
h
e
l
e
f
t
s
h
o
w
s
s
e
m
b
l
a
n
c
e
a
n
a
l
y
s
i
s
f
o
r
r
m
s
a
n
d
i
n
t
e
r
v
a
l
v
e
l
o
c
i
t
y
d
e
t
e
r
m
i
n
a
t
i
o
n
b
a
s
e
d
o
n
t
h
e
i
n
p
u
t
s
e
i
s
m
i
c
g
a
t
h
e
r
s
h
o
w
n
i
n
t
h
e
m
i
d
d
l
e
.
A
r
a
n
g
e
o
f
v
e
l
o
c
i
t
y
f
u
n
c
t
i
o
n
s
i
s
u
s
e
d
t
o
d
e
t
e
r
m
i
n
e
t
h
e
b
e
s
t
s
t
a
c
k
i
n
g
v
e
l
o
c
i
t
y
.
T
h
e
p
a
n
e
l
o
n
t
h
e
r
i
g
h
t
s
h
o
w
s
t
h
e
n
a
t
u
r
e
o
f
t
h
e
s
t
a
c
k
o
f
a
d
j
a
c
e
n
t
g
a
t
h
e
r
s
u
s
i
n
g
t
h
e
v
e
l
o
c
i
t
y
f
u
n
c
t
i
o
n
s
h
o
w
n
o
n
t
h
e
l
e
f
t
p
a
n
e
l
.
ratio in the data, yet keeping the seismic events attened. Pick-
ing velocity functions based on maxima in the semblance anal-
ysis will yield nice looking stacks, but not necessarily the
best velocity for pressure analysis. This criterion will tend to
yield velocities systematically increasing with depth and will
ignore geopressuring which, on the other hand, suggests that
the velocity may not increase with increasing depth. The panels
on the right side of Figure 7 show velocity elds for adjacent
gathers. These displays are also important; they indicate the
spatial consistency of the velocity eld. Any velocity analysis
for geopressure prediction work must include this step indicat-
ing lateral continuity of geology, unless changed by faulting or
other geologic features such as pressure compartments. How-
ever, such geologic events must have a seismic signature in
many gathers, and not just one; otherwise such signatures are
false indicators of geologic features and should be discarded
while interpreting the velocities.
For pore pressure prediction, one often uses horizon veloc-
ity analysis (HVA), as discussed in Yilmaz (1987). This is an
efcient and accurate way to get velocity information at every
CMP location along selected key horizons, as opposed to the
conventional velocity analysis that is usually carriedout tempo-
rally at selected CMP locations. HVAis tedious and expensive,
but highly recommended for pressure prediction work.
No matter which method is used, we must realize that the
quality of the velocity function is dictated by the quality of the
stack: the function that attens a certain event in a gather. This
means using a quantitative measure to obtain a velocity func-
tion based on the amplitude and continuity of stacked event(s).
Note that this velocity function does not yield rock velocity.
This has a signicant consequence in that not all velocity func-
tions that yieldthe best stackcanbe consideredtoproduce true
rock velocities. The velocity estimation from seismic stacking
velocity analysis has limitations because of the S/N ratio of the
gathers, muting, the spread length used for data acquisition,
the stacking fold, the choice of coherency measure, true de-
parture from hyperbolic moveout (e.g., anisotropy), the time
gate length, the bandwidth of the data, and the complex struc-
ture with spatially variable velocities. In general, as events on
a velocity analysis are picked deeper (in time), the quality of
the velocity function degrades. This is because one encounters
a range of velocity functions, which atten an event, creating
ambiguity and lack of precision in the picked velocity function.
For the sake of completeness, I summarize here the limitations
for calculation of interval velocities (with Dixs model) from
stacking velocities (Yilmaz, 1987): layered geometry assump-
tions for Dix-velocity calculation, assumption of homogeneous
and isotropic layer properties, lack of precision in data analysis
(picking of reection times, etc.), improper use of the Dix cal-
culation (interval velocities calculated in a narrow interval),
poor resolution, raypath bending through complex geology,
and spatial sampling.
Velocity smoothing, calibration, resolution,
and interpretation
Velocity smoothing is an essential step in conditioning veloc-
ities for pore-pressure prediction work. Smoothing (and inter-
polation) is done bothintime andspace. Anexample is givenin
Figure 8. The top panel in Figure 8 shows unsmoothed stacking
velocities, whereas the bottom panel shows smoothed veloci-
ties. Obviously, the interval velocities fromthese two models in
this gure would yield two different interval velocities. Which
model is valid? One does not knowthis a priori. For this reason,
velocity calibration is an essential step in conditioning seismic
velocities for pressure prediction. This step is discussed below.
Lateral velocity variations do occur frequently. This involves
encountering low-velocity zones in an otherwise high-velocity
medium or vice versa. This will involve different moveout
curves, and hence velocities, as shown in Figures 9 and 10. Ef-
fects suchas those showninthese gures must be accountedfor
in the velocity analysis work prior to geopressure estimation.
As noted earlier, the Dix-model assumes a layered earth and
ignores raypathbending while attening a seismic event. This is
unphysical. The raypath bending will cause a stack of at layers
to behave as if it were apparently anisotropic, although each
layer may be intrinsically isotropic (Yilmaz, 1987). The resul-
tant effect is to cause nonhyperbolic moveout at larger offset
ranges. In the jargon of the velocity analysis software packages,
this is often referred as the residual moveout correction. This
correction is very important and must be accounted for before
any pressure analysis work progresses. This usually requires
a large offset data set. Errors such as this often degrade the
conventional semblance analysis plots for smaller offsets.
Figure 11 shows the improvement in velocity analysis with
increasingly sophisticated corrections applied to a given set of
data (i.e., a gather). Included are the fourth-order correction of
FIG. 8. Unsmoothed (top) versus smoothed (bottom) stacking
velocities fromseismic data. Smoothing function can affect the
quality of the resulting velocity function in an adverse way for
geopressure applications.
2024 Dutta
equation (22), the raypath bending effect, and the transversely
anisotropy correction. Note that if we restricted ourselves to
smaller source-receiver distances (less than or equal to 30
),
none of these effects would be detectedall would yield ap-
proximately the same velocity function. This is because one is
trying to atten a hyberbola in the small angle range where it
is already close to being at. That would result in erroneous
interval velocity and, hence, erroneous pressure values. Thus, I
recommend that pressure analysis be performed on a data set
containing as large a source-receiver distance as possible. Not
only will this provide a way to account for fourth and higher or-
der corrections to NMO velocities, but it will also help account
for anisotropy corrections.
Velocity calibration is an essential step. It is usually done
using checkshot data from well surveys. Due to the acquisition
geometry, the seismic interval velocities are usually higher than
the vertical velocities measured from the well surveys. There
are two methods to account for the discrepancy between the
seismic and the checkshot velocities. The rst method involves
comparing the time-depth relationships from checkshot sur-
veys to those derived from the seismic data. The second, the
preferred method, involves comparing the interval velocities
fromthe checkshots to those derived fromthe seismic stacking
velocity analysis using Dixs method. The procedure involves
obtaining a correction function versus two-way time at the well
location (or the analog well location), which is then applied to
the entire velocity eld. Typically, several velocity functions
FIG. 9. Effect of lateral velocity variation on NMOfor the model shown in the top panel, where a velocity anomaly is present. The
lower panels depict the shapes of the moveout curves. The vertical axis denotes travel time in milliseconds and the horizontal axis
denotes velocity in meters/second. Note the differences in the moveout behavior where the ray paths are sampling a transition
zone between the low and the high velocities.
around a well are used to construct this calibration function.
This allows an error analysis, which must follow such a calibra-
tion scheme. The entire procedure is shown schematically in
Figure 12. I recommend that the procedure be tested on cross-
ing seismic lines for accuracy and consistency. Pressure predic-
tion should never be done based on a single velocity function at
a CMP location.
Any application of seismic velocity for pressure prediction
must include an understanding of the velocity resolution asso-
ciated with the velocity analysis process. Figure 13 shows the
vertical (temporal) and lateral resolution of the conventional
velocities. Temporally, the interval velocities from the conven-
tional stacking velocity analysis do not have frequencies higher
than 24 Hz. This leads to pressure analysis in layers no thinner
than perhaps 625 ft [190 m] for typical Gulf Coast sediments.
Seismic inversion of either stacked data (poststack acoustic
impedance inversion) or prestack data [using full wave-form
analysis and inversion of gathers at various common depth
points (CDPs)] can add higher resolution.
The standard error,
i nt
, in the interval velocity obtained
from rms velocities with the standard error, , when the inter-
val is of thickness h and average depth D, is given by
i nt
1.4D/h. (31)
Thus, if we want a precision of 5% in the interval velocity of,
say, a layer between 1000 and 1200 m depth, we will need a
precision in the rms velocity at the top and the base of the
layer given by
= h
i nt
/(1.4D) = 200 5%/(1.4 1100) = 0.6%.
(32)
Such precision is very difcult to obtain.
Velocity interpretation is also an essential step in converting
stacking velocities to rock velocities. There are many pitfalls
in the procedure, some of which have already been pointed
out. The major pitfall is related to the fact that multiples have
low rms velocities than primaries arriving at the same time.
Figure 14 shows velocity analysis for a single CDP in which the
stacking quality of various user-selected velocity functions is
shown. The contouring indicates the quality of the stack, and
the optimum stacking velocity function is obtained by con-
necting the contoured highs, as shown by the heavy line in the
middle and connecting the velocity contours on the plot. Note
the secondary, low-velocity function in the lower left-hand side
of the plot. This represents the optimum stacking velocities for
the multiples for this CDP at this location. However, a priori
one does not know whether this lowering of velocity is due to
higher than normal pore pressure in that interval or to multi-
ples. It could also be due to lithology changes (fromcarbonates
to clastics, for example). Ambiguities of these types are very
common in the velocity analysis steps and must be reconciled
by using other data and knowledge to interpret the seismic ve-
locity eld. Even the very simple step of velocity smoothing as
described above can lead to either destroying the local geol-
ogy or creating false geology, either of which could lead to an
FIG. 10. Velocity prole in laterally varying velocity zones. The top panel shows the raypaths through a medium containing a
volocity anomaly, as shown in Figure 9. The bottom panel shows V
rms
.
unphysical pressure model, and must be avoided.
Error analysis must accompany every velocity analysis em-
ployedfor pressure prediction. InFigure 15, we showthe errors
in stacking velocity estimation for a particular case. The gure
depicts a histogram of errors in interval velocity analysis as a
function of two-way interval time thickness. A typical velocity
trend is used for this example. Interval thickness is shown in
milliseconds, anderrors invelocity are showninmeters/second.
Figure 15 clearly shows that the errors in interval velocity are
greatest for the thin layers at large depths. This inherently will
pose a limit on the accuracy of predicted pressures using seis-
mic interval velocity.
Guide to velocity analysis for geopressure work
Quite often velocities are misused for pressure prediction;
stacking velocities need much conditioning before being used
for this purpose, as discussed earlier. Before work begins, a
clear understanding of the purpose of the work must be de-
ned: Are the velocities needed for a regional understanding
of pressure [typically a gridof 100 100 miles (160 160 km)]?
A detailed image of subsurface pressure at the prospect scale
[3 3 miles (5 5 km)]? At a reservoir scale [100200 ft (30
60 m) layer thickness]? At a wellbore scale [several feet layer
thickness at target depth, say, 10 000 ft [3000 m])? The care
and details employed for each of these scales vary and require
integration of a host of data other than seismic, such as well
2026 Dutta
velocities, logs, and geology. Velocity analysis is a tedious pro-
cess and is usually carried out repetitively; at each step, ad-
ditional data and interpretation are incorporated to improve
resolution and accuracy. The steps below are intended to pro-
vide a guide to the process:
1) A general understanding of the geology should precede
any velocity analysis. Using stacked and interpreted seis-
mic sections does this best. All available wells must be
posted at appropriate locations along with the key geo-
logic horizons.
2) Conventionally processed stacking velocities are usually
unsuitable for pressure prediction work because they are
created for imaging and may have very little to do with
the rock velocity.
3) The seismic gathers must be available for quality control
and examined for the S/N analysis.
4) The processing ow must be clearly identied on the
seismic section. The velocities must be processed for
demultiple, dip moveout (DMO), and prestack migra-
tion (DMO is not necessary if Kirchhoff migration is
used). For areas with salt (such as in the deepwater Gulf
of Mexico), one must apply appropriate salt mask and
mute.
5) The stacking velocity, rms velocity, or the interval veloc-
ities given at the top of a stacked seismic section should
FIG. 11. Comparison of moveout methods showing successive improvements in the velocity analysis of a CMP gather. The panel
on the extreme right is the correct velocity model, whereas the others are not. Note that these changes are noticeable only when
large offsets are used for analysis.
not be used for pressure prediction work without looking
at the gathers and semblance or velocity spectrum plots
for quality assurance.
FIG. 12. A processing ow chart for velocity conditioning and
calibration for pore pressure analysis.
6) Detailed velocity analysis is best done on a workstation.
Essential steps include closely spaced velocity analysis,
lateral consistency in the velocity eld, smoothing, cali-
bration, and interpretation.
FIG. 13. Velocity resolution from seismic data.
FIG. 14. Velocity analysis panel for a single CMP gather. Var-
ious user selected velocity functions are also shown. The con-
touring indicates the quality of the stack, and the optimum
stacking velocity function is obtained by connecting the con-
toured highs. Note the secondary, low-velocity function on the
left-hand side of the plot. This represents the optimumstacking
velocities for multiples at this location.
7) Usual criteria for picking the velocity (such as semblance
maximum on the velocity spectra) may not be suitable
for pressure prediction work. Picking that velocity which
attens an event fromnear to far offsets may work better.
If an event cannot be attened across its full offset range,
then every effort must be made to optimize attening it
as far an offset range as possible (say fromnear to middle
range).
8) Velocitycalibrationis anessential stepinconditioningve-
locities for pressure prediction. Comparing seismic rms
and interval velocities with those fromcheckshot surveys
does this best. However, if checkshot data are not avail-
able, a notion of rock velocity must be used to constrain
the velocity eld from analog studies, if possible.
9) Never make a prediction based on a single velocity func-
tion, say at the well location. The velocity eld must
be checked for lateral and temporal consistency, so that
any wild uctuation or spikes are not present. I rec-
ommend that any velocity analysis for pressure predic-
tion work should proceed with examination of at least a
dozen seismic gathers around each site of a calibration or
pseudowell.
10) Velocity picking should be done in several steps of suc-
cessively increasing detail. Most detailed picking should
be event oriented to ensure that no bias is introduced by
Dix calculations for intervals of varying thickness.
11) Quality control displays such as NMO-corrected gathers
should be used to check accuracy of velocity picks
12) As the histogram in Figure 15 indicates, never pick ve-
locities in layers with thickness (in time) less than 50 ms,
especially at relatively large depths (say, at 3-s two-way
time or more).
13) For a regional-scale velocity analysis for pressure predic-
tion work, the velocities should be picked at least in a
1 1 km grid or smaller. Special care must be exercised
FIG. 15. Errors in stacking velocity estimation as a function of
two-way time. The histogramillustrates the accuracy of a veloc-
ity analysis as a function of two-way time thickness for a typical
velocity trend. The errors in interval velocity are greatest for
thin layers at large depths.
2028 Dutta
while interpolating the velocity eld. This is because the
interpolation process can exaggerate any velocity spikes
caused by anomalous picks.
14) Velocities must be smoothed by a simple mathematical
function, such as a low-order polynomial. Complicated
smoothing algorithms, such as splines, may not be worth
the trouble they may pose. These functions tend to follow
the undulations in the velocity eld too faithfully and
sometime create geology when none is present.
15) Velocity structures observed within a spread length must
be investigated carefully. These may not be due to geo-
logic variations.
16) Every effort must be made to relate the seismic interval
velocities to the rock velocities. The velocity eld must be
constrained by the knowledge of the range of known rock
velocities in the area. For example, velocities more than
10 000 ft/s (3000 m/s) in the deepwater Gulf of Mexico
must be carefully examined. Charts such as the one for
deepwater Gulf of MexicogiveninFigure 16 for twocases
(one when the rock is under the hydrostatic pressure con-
dition and the other when the rock is under the fracture
pressure condition) are extremely useful. Any velocity
function that is outside this range, for example, should be
considered suspicious and must be checked and repicked.
Thus, velocity analysis should be guided by a notion
of expected rock velocity, not by semblance analysis
alone.
17) Any comparison with well log sonic velocities must
be done only after the sonic log has been checkshot
corrected and ltered to simulate the low-frequency
velocities.
18) Usually analysts are trained to pick the faster velocities
as the depth increases. This may not be a good practice
when picking velocities for a quantitative pressure anal-
ysis. In fact, as pore pressure increases, the velocities do
not increase with the depth as rapidly with burial as it
would have without any pressure effect. The rms veloci-
FIG. 16. Seismic slowness interpretation using slownesses de-
rived from a rock model. The solid line shows slowness from
a CMP gather. The short dash curve shows a predicted trend
of slowness at that location had the rocks compacted normally
and maintained a hydrostatic pore-uid pressure at all depths.
The heavy dashes show the expected trend if the pore pressure
is close to the hydraulic fracture limit.
ties can, in fact, decrease with increasing depth across a
major interval velocity inversion.
19) Special care must be exercised with the layer where ve-
locity has been picked last (namely, at the end of the data
set where reliable velocity picks can be made). Quite of-
ten, the interval velocity is held constant in that layer and
then extrapolated to depths, beyond where there is no ac-
tual data to pick velocity. This is a bad practice and should
be avoided because this can result in unphysical pressure
values when used by unsuspecting users unfamiliar with
the practice.
20) All velocity analyses must be accompanied by error anal-
yses. Every pick must include a quality assurance proce-
dure (for example, 1 for good, 2 for questionable, and 3
for bad, although this is subjective judgement based on
the inter preters assessment of the S/N of the data).
Tomographic inversion
The previous section dealt with velocity estimation using
stacking velocities following the algorithm proposed by Dix
(1955). Although this is the most commonly used approach,
it does have many limitations, as discussed earlier. There are
many other techniques, all based on some sort of inversion
procedure, which can add value to pressure analysis work be-
yond what is possible by the Dix approach. The tomographic
inversion technique is one of these procedures. It can provide
velocity elds that better relate to the geologic structures than
what is feasible with conventional stacking velocity analysis.
This method and other inversion techniques discussed below
are gaining more popularity as the cost of computing has come
down signicantly. Geophysicists are using more and more ve-
locities fromvarious inversiontechniques for pressure analysis.
Tomographic inversion is one of a class of multidimensional
inversion methods that provides more detail of the subsurface
3-D velocity eld associated with 3-D geologic structures (Aki
and Lee, 1976; Gjoystdal and Ursin, 1981; McMechan, 1983;
Bishop et al., 1985; Chiu and Stewart; 1987; Hubral and Krey,
1980). It is basically a 3-D traveltime inversion technique. The
fundamental concept is to start with an initial guess of the sub-
surface structure and the associated parameters, such as veloc-
ity, which characterize the model. This is usually done in the
depth domain. Rays are then traced through this initial-guess
model using Snells law. The mismatch or a residual travel time
is dened next. It is the difference between the computed trav-
eltimes and those observed from the real data. This residual is
used as a guide to update the initial model by solving a version
of the least-squares equations. The iterative process is carried
until a convergence criterion is met. This is dened as the case
when the residual times become small or the model parameters
change insignicantly. Figure 17 shows a typical owchart for
the tomographic inversion process. The process provides both
the depth and the velocity elds associated with reectors. It is
very effective for identifying andaccounting for lateral velocity
variation and various stacking velocity anomalies, such as the
one illustrated in Figures 9 and 10.
The tomographic method is inherently nonunique. Different
starting models will give different nal solutions. In this regard,
the seismic tomography technique is different and more com-
plicatedthanmedical X-ray tomography. The complicating fac-
tors are unknowndepthtoreectors, ray-bending effects which
are nonlinear, and raypath coverage (which is irregular and
has limitedview-angle coverage). Therefore, the procedure can
yield velocities that may not be appropriate for pressure anal-
ysis in all cases . Velocity constraints, usually provided either
from well logs or VSP data, are essential to limit the range of
uncertainty. When properly conditioned, tomographic velocity
inversion will provide a laterally extensive and reliable image
of the velocity eld, which will approximate to the 3-D rock
velocities more closely.
Areviewof the literature suggests that the velocity inversion
approach is very useful in the cases where velocity contrasts in
subsurface rocks are sizable. Examples are clastics overlaid by
carbonates, shallow gas layers or pockets, and very high over-
pressures. Quite often, we see a remarkable improvement in
the quality of the tomographic image over that obtained using
Dixs interval velocities. However, the technique sometimes
misses subtle changes in the velocity elds due to the limited
bandwidth of reection seismic data. The methods of seismic
reection tomography and migration may be viewed as com-
plementary processes. Migration requires velocities in order
to image reector boundaries, whereas reection tomography
requires reector positions to estimate velocities. However, it
is for this reason that the tomography derived velocities are
more appropriate for pressure analysis than those from the
straightforward depth-imaging process.
Prestack amplitude inversion
While Dixs velocity analysis and the tomography inversion
method use seismic traveltimes, the velocities from prestack
inversion methodologies include both traveltime and seismic
amplitudes. These methods, therefore, involve full waveform
analysis techniques (a wave-equation-based approach). They
use raw seismic gathers as input to produce a general velocity
model. Work by Tarantola (1983) and Mora (1987) are note-
worthy contributions. These full waveform methods represent
a very general type of multidimensional inversion. They em-
ploy inversion schemes based on nonlinear least squares and
proceed by iteratively updating the earth parameters until a
FIG. 17. A workow for velocity analysis using tomography.
good t between the observed data and the modeled data is
achieved.
There are several weaknesses to this inversion procedure.
First, the technique requires fairly accurate a priori knowledge
of the low-wavenumber velocity eld. This may be handled by
using a starting model based on, for example, tomographic in-
version. Second, these procedures are very computer intensive
and require good S/Nratios in the seismic data and careful am-
plitude processing to preserve the true amplitudes. While we
can achieve some gain in the S/N ratios by modern processing
algorithms, many geologic environments will have inherently
poor S/N ratios. Examples are subsalt and subbasalt areas. As
far as computing speeds are concerned, parallel computers are
very powerful tools and are ideally suited for this type of in-
version, but are not routinely available. Third, the results are
nonunique, as is the case with every inversion method. This is
perhaps the most serious weakness. Imposing rock-property
based constraints to the inversion process, we can reduce this
weakness. These constraints are various interrelationships be-
tween P-waves, S-waves, and densities of various rock types.
Unless the inversion process accounts for these relationships
as constraints in the procedure, the resulting velocities may be
useless for pressure work. Unfortunately, most of the available
algorithms do not use any constraints of these types.
Mallick (1995, 1999) developed a prestack inversion method
based on the genetic algorithm (GA) approach. This is a statis-
tical optimization technique that operates much like biological
evolution and yields P-waves, S-wave, and densities for a given
seismic gather, along with the low-frequency velocity trends.
The procedure is outlined in Figure 18. In the GA procedure,
rst a random population of elastic earth models is generated
within a specied search interval of the parameter space. For
each of these random models, synthetic seismograms are cre-
ated with a full wave-equation simulator. Thus, all multiples,
modeconversions, andtransmissionlosses areaccountedfor, as
well as tuning and propagation effects through velocity gradi-
ent zones. The modeled synthetic data are then compared with
the real data to yield a tness value for each model. Genetic
operations of reproduction, crossover, mutation, and update
2030 Dutta
performed on the generated models follow this comparison.
The iterative process continues until the tness values in the
generated models converge.
An example of the application of the GA procedure is given
in Figure 19. The left panels show stacked data and the cor-
responding synthetic stack. The right panels show the acoustic
FIG. 18. A owchart detailing the steps involved in the prestack full waveform inversion of a CMP gather using
the GA procedure.
FIG. 19. An example of results from the GA procedure for a 2-D line. The inversion was applied to every CMP gather for the
stacked data shown in the upper left panel. The lower left panel shows the synthetic stack generated from stacking the velocities
obtained from the GA inversion. The acoustic impedance and the Poissons ratios obtained from the prestack inversion are also
shown.
impedance (the product of velocity anddensity), andestimated
Poissons ratio. Figure 20 shows a comparison between ob-
served (black) and inverted (red) acoustic impedance at wells
Aand Dof Figure 19. The method has the capability of produc-
ing a much higher frequency image, although still much lower
than what can be obtained from well data.
The GA procedure suffers from the same weaknesses as
those discussed above for other prestack inversion techniques:
nonuniqueness is a severe problem here also. However, by
constraining the models with rock properties information, this
method can provide velocities for overpressure analysis with
much higher resolution than is possible with either the Dix or
tomographic inversion approaches.
Poststack amplitude inversion
The poststack amplitude inversion concept is essentially
based on 1-D inversion of seismic data. It visualizes the earth
as a series of layers at each CMP location. Each layer is
characterized by density, velocity, and layer thickness. Each
layer has equal two-way time intervals. The procedure con-
verts estimates of reection coefcients into an image of acous-
tic impedance with depth. Lavergne and Willm (1977) and
Lindseth (1979) discuss the procedure in detail. Except for
some modications to take advantage of todays computing
environment, the basic approach remains unchanged. For nor-
mally incident pressure waves, the reection coefcients for
layer interfaces are related to the acoustic impedance by
C
k
= (
k+1
v
k+1
k
v
k
)/(
k+1
v
k+1
+
k
v
k
), (33)
where C
k
is the reection coefcients of the kth interface and
k
v
k
is the acoustic impedance of the kth layer. Then, the acous-
tic impedance for the next k +1th layer is
k+1
v
k+1
=
k
v
k
(1 +C
k
)/(1 C
k
). (34)
This operation is often called trace integration. It is compli-
cated, however, by the fact that the eld data lacks the low
frequency trend (05 Hz, for example). Lindseth (1979) sug-
gested using a velocity analysis approach to add to the trace
integrated high-frequency impedance. This procedure is still
used widely. Using a relationship between density and veloc-
ity, acoustic impedance traces are converted to velocities. Thus,
each seismic trace in the stacked data domain is converted into
FIG. 20. A comparison of the acoustic impedance obtained
from the GA procedure (labeled inverted) with those from
the well data (labeled log) at locations A and D of Figure 19.
a pseudowell velocity log trace. These velocities can be used for
pressure prediction (Dutta and Ray, 1997). Kolb et al. (1986)
showed some excellent examples of nonuniqueness inherent
in the method and showed very clearly that the low-frequency
trend in 1-D inversion is critically important.
Poststack inversion techniques can be applied to any seis-
mic data volume. Figure 21 presents an example of a 2-D cross
section from a 3-D data volume. Figure 21a shows the interval
velocity as a function of two-way time from the conventional
stacking velocity analysis based on the Dix-model, including
all the corrections discussed earlier. Figure 21b shows the high-
resolution interval velocity derived from the poststack acous-
tic inversion. The details in the velocity eld are obvious in
Figure 21. Figure 22 shows a comparison with a band-limited
version of the velocity log from a well on the seismic data. Ve-
locity traces such as shown in Figure 22 are extremely useful
in providing pressure estimates needed for well planning, es-
pecially when the wells are deviated. This is quite often the
case in the appraisal and development phases of a eld. How-
ever, note that velocity calibration, especially that of the low-
frequency trend, is essential for reliable velocity prediction by
this method. Therefore, this may not be useful for exploration
wells inrankwildcat areas unless areliablelow-frequencytrend
is available.
APPLICATIONS OF SEISMIC VELOCITY
FOR PRESSURE PREDICTION
Dix model
From a review of the existing literature, it is clear that most
people use seismic data for pressure prediction in the predrill
phase of the work. The Dix model is the most commonly em-
ployed technique for the velocity input. However, as noted ear-
lier, the frequency content of the velocity function is no more
than 24 Hz. Thus, the pressure estimates using conventional
FIG. 21. Velocity from poststack inversion of seismic data:
(a) Dix velocity model, (b) poststack inversion. Note that the
reservoir section(bright amplitude at CMPlocation2200 and
at 3230 ms is missing in the Dix model due to lack of velocity
resolution. This is not the case withthe velocities obtainedfrom
the poststack inversion. Figure from Dutta and Ray (1997).
2032 Dutta
velocity analysis is fairly gross and may not provide estimates
within reservoir layers, where repeat formation tester (RFT)
measurements are made. However, at the exploration stage,
pressures in the reservoir scale are usually not an issue.
Figure 23 show a owchart for pressure prediction using any
kind of seismic velocity. For the Dix model, we assume that the
FIG. 22. Velocity from poststack inversion of seismic data and
comparison with the well velocity versus two-way time for the
section shown in Figure 21 at the bright spot location (CMP
2200).
FIG. 23. A schematic diagram showing the pertinent steps in model-based geopressure prediction using seismic
data.
velocities are conditioned following the procedure described
earlier in this paper. We note that seismic velocities are used
twice: once for generating a pseudobulk density, which will
eventually yield overburden stress, and then for deriving an ef-
fective stress at each velocity location (CMP location). Here I
employ a user-specied transform between velocity and effec-
tive stress, as discussed earlier. Alternatively, the same veloc-
ities can also be used in the traditional fashion, such as in the
method due to Eaton (1972) or Hottmann and Johnson (1965).
Those methods rely on normal compaction trend (NCT) analy-
sis of seismic velocity versus depthfunction(s), andthenuse the
velocity deviation fromthe NCTas a measure of pore pressure,
employing empirical calibration functions (Dutta, 1987a).
Once velocity analysis is completed, pressure prediction can
be made in different dimensions, depending upon need: 1-D,
2-D, and 3-D. I now present some examples, which show the
power of seismic velocity as a tool for pressure prediction.
3-D applications
For 3-Dapplications (for regional pressure work), the veloc-
ities come from either a 3-D velocity survey or a grid of closely
spaced 2-D seismic lines. Figures 2425 show an application of
this procedure in the deepwater Gulf of Mexico. Here, a 3-D
model of effective stress has been developed over a prospec-
tive play fairway. The model covers an area of 140 102 km,
with the water depth greater than 330 m. Available checkshot
data from a number of wells provided velocity calibration and
anisotropy correction. The velocity analysis and interpreta-
tion followed the procedure outlined above. Analysis shown in
Figures 2425 used velocity grids no more than 0.5 km apart.
Figure 24 is a map of effective stress projected at a prospec-
tive horizon over many blocks. The color codes in this gure
represent the risk associated with hydraulic seal failure: red de-
notes areas with high likelihood, green denotes areas with least
likelihood, and yellowsignies areas with moderate risk of seal
failure, respectively. Such maps have helped explorationists to
high-grade areas of low top-seal risk, and downgrade risky ar-
eas. Figure 25 shows a map of the top of hard pressure in the
same fairway as a function of two-way time. Here, the top of
hard pressure has been dened as the depth (or time) where
the effective stress reaches a threshold value of 1000 psi. For
most cases, this corresponds to a pore pressure that is about
90% of the overburden pressure.
1-D/2-D applications
As discussed earlier, at the prospect scale, lateral resolu-
tion of the seismic velocity analysis can be greatly enhanced
by picking velocities at every CMP. In this manner, a very de-
tailed subsurface image of pressure and effective stress can be
obtained, both at prospect and at well-bore scales. Figure 26
shows an example of such an analysis from a deepwater Gulf
FIG. 24. Effective stress has been color coded for its proximity to the hydraulic seal failure limit of 1000 psi for a data set in the
deepwater Gulf of Mexico. Green represents a low likelihood of seal failure. Yellow indicates uncertainty based on an analysis of
the estimated errors in the effective stress. Certain basins stand out as either lowrisk (e.g., Auger basin) or high risk (e.g., Amundsen
basin). The updip margins of most basins show high risk of prospectivity. This allowed explorationists a quantitative means of not
only high-grading prospects in this play fairway but also opening new opportunities. The inset shows the extent of the area of study
in the Gulf of Mexico. Figure from Dutta (1997).
of Mexico prospect on a 2-D seismic line. It shows the effec-
tive stress as a function of two-way time and CMP locations.
The gure exhibits the existence of a pressure cell associated
with stratigraphic variations within the prospect. This is due to
uid ow along the dipping sand beds to the left that causes
differential compactions of sediments across the structure. It
also indicates pressure traps in the vertical direction, shown
as reversals of the effective stresses. A comparison of the pre-
dicted pressures with the RFT data from the well is shown
in Figure 27. The comparison is good, and the predictions are
within 400 psi of the formation pressure.
Figure 28 shows another example from the deepwater Gulf
of Mexico. This gure is a plot of effective stress versus two-
way time for every CMP location. The data are from a eld
in the deepwater Gulf of Mexico (Dutta and Ray, 1997).
Figure 28a uses the Dix model, whereas Figure 28b uses the
high-resolution velocity model from poststack impedance in-
version. The improvement in the effective stress in the latter
model is obvious; the seal distribution around the bright spot
is clearly visible in the right side of the gure at about 3400 ms.
2034 Dutta
Further, an additional reservoir at about 3340 ms is also vis-
ible. Figure 29 shows a plot of the high-resolution pore pres-
sure versus two-way time at the well location (full line) and
a comparison with data from six RFT (black dots) within the
top reservoir. The data are extracted from Figure 28. The thin
line in this gure shows pressure estimate from the Dix model.
We note that the high-resolution velocity model predicts the
pressure reversal at the reservoir level. The Dix model does
not yield such detail; it just provides a low-frequency pressure
trend.
Tomographic inversion
Figure 30 shows the predicted pore pressure from a deepwa-
ter Gulf of Mexicoplay. Greenrepresents normal pressures and
yellows a pore pressure of 10 lbs/gallon (ppg), which is dened
to be the overpressure threshold. The lateral pore pressure
varies about 12 ppg betweenthe footwall andthe hanging wall
of the normal faults. These results were obtained using seismic
velocity analysis in conjunction with tomographic inversion of
velocities to add high frequencies to the data. The procedure
is very complex; it uses all the prestack data and requires cal-
FIG. 25. A map of two-way time to the top of hard geopressure in a deepwater, Gulf of Mexico play fairway. The inset shows
the extent of the area of study in the Gulf of Mexico. Figure from Dutta (1997).
ibration. However, once calibration wells are available, it is a
powerful tool for geopressure analysis.
Prestack inversion (GA procedure)
Not much is available in the literature regarding the applica-
tion of this technology for pressure prediction; it is expensive,
and software needs to be run by experts. However, if the S/N
ratio is high and there is a good background low-frequency ve-
locity trend that has been conditioned for rock velocity analy-
sis, the technology can yield a high-resolution image of subsur-
face pore pressure. At present, only 1-D applications on CMP
gathers have been made. An example is given in Figure 31.
The gure shows the input CMP gather on the left, the acous-
tic impedance in the middle, and the resulting pore pressure
prole on the right. The inversion process followed the data
in the 20-ms interval and used a rock model to convert the
resulting P-wave velocity to the effective stress and the pore
pressure. Results such as these are useful in well-planning ap-
plications, provided calibration data are available and a proper
interpretation of the results is made. A limited testing on this
procedure shows its potential power. Some recent applications
of this technology for pore pressure prediction and SWF iden-
tications in the deepwater Gulf of Mexico are given in Dutta
et al. (2001) and Dekok et al. (2001a, b). Nonetheless, it is an
emerging technology, and needs further validation under a va-
riety of geologic conditions so that a condence limit can be
established. The procedure also yields Poissons ratios that can
be used to estimate fracture gradient.
Use of seismic pressure prediction techniques in real time
With proper care, any of the seismic-based techniques dis-
cussed above may be used for pressure estimates in real time,
provided a suitable strategy is implemented. The strategy is
rather simple: obtain a pressure image of a pseudowell pres-
sure prole before a well is drilled and update that image or
the prole based on real-time and other data obtained while
drilling. This must be a t-for-purpose strategy. It will depend
on the expected pressure prole for a given well. For example,
if the predrill model in a given geologic situation suggests that
the onset of the geopressure is gradual, as is the case when
the velocity data have well-dened normal compaction trends,
prediction ahead of the bit in real time may not be critical.
On the other hand, if the predrill model suggests that the
onset of geopressure may be abrupt, detection of the prole
ahead of the bit will be critical. It will also be critical if the
FIG. 26. A cross-section of the effective stress versus two-way time over a prospect in the deepwater Gulf of Mexico. A pressure
cell is clearly visible, and is bounded by salt on the left-hand side. The discovery well location through the bright spots is also
noted on the seismic section.
prole is expected to contain pressure reversals (associated
with thin sands). The latter is one of the major cause of well
problems.
Thus, the strategy for real-time prediction might be as fol-
lows. Start with the best possible velocity function(s). Take this
function to the rig and update it up to the bit level using ei-
ther an intermediate checkshot or seismic-while-drilling data,
if available. Using the newly calibrated velocity function, com-
pute pore pressures, again up to the bit level. Calibrate this
pressure model using other real-time data such as LWD/MWD
logs, D-exponents and connection, and background and trip
gases which are pressure indicators on the rig while a well is
being drilled. Next, apply this calibration function to the seis-
mic data and obtain an update ahead of the bit, both for the
velocity and the pressure. This continuous prediction and re-
peated calibration on the rig can be done at every connection
time during the drilling process (typically every 100 ft, or 30 m).
There are several real-time tools for the velocity update.
They are drill-bit seismic-while-drilling tools, such as Tomex
(Western Altas), Seisbit (Agip), and DbSeis (Schlumberger);
seismic measurement while drilling (SMWDof Schlumberger);
and the LWD acoustic tool congured for a while-drilling en-
vironment. If real-time velocity tools are unavailable, inter-
mediate VSP logs can be substituted for the update of the
seismic model. Of course, the VSP data must be processed
2036 Dutta
for a look-ahead image (Dutta et al., 1998). Figure 32 gives a
owchart of the process.
An example of this type of approach is taken from the work
of Meehan et al. (1998a, b, c). Figure 33 shows a velocity plot
containing data from three sources from an offshore well in
southeast Asia. The seismic velocity at the well location prior
todrilling, the velocitydatafromVSP, andfromdrill-bit seismic
(which employs the drilling bit as a seismic source) are shown.
The shapes of these velocity functions are similar. The trend
FIG. 27. Pressure depth proles for the discovery well shown in
Figure26. Computedpressures frombothseismic (predrill) and
sonic (postdrill) velocities are shown and compared with the
measured pressures from repeat formation testers denoted by
RFT. The overburden pressure was estimated from the seismic
velocity and found to be in good agreement with that obtained
from integrating the bulk density log (postdrill).
FIG. 28. Effective stress from poststack inversion of seismic
data of Figure 21 versus two way time. (a) The result from the
Dix velocity, (b) the result from the high-resolution velocity
model from poststack impedance inversion.
changes in these functions indicate the top of overpressure.
However, the seismic velocity curve shows the top to be at a
deeper time (2500 ms) than the other two velocity functions.
Therefore, a calibration at the rig site using either of the two
data sets would enable the seismic to be much more robust in
the look-ahead sense. The computed pore pressure from these
velocity functions as well as those fromother drilling data (mud
weight, pressure from the rate of penetration data labeled as
DXC pore pressure, and true formation pressure data labeled
as RFT/DST) is shown in Figure 34. The gure also shows pore
pressure estimated from sonic log (post drilling) and that from
seismic while drilling (SWD). This type of technique can be
easily implemented on a visualization system at a rig site.
DISCUSSION
Pressure prediction from seismic data has two major com-
ponents: a rock model that relates effective stress to veloc-
ity, and the velocity eld. The errors in the predicted pres-
sure arise from both sources. It is true that even if the rock
model were perfectan unlikely casethere would still be
errors in the predicted pore pressure due to inherent errors
in the seismic velocity eld. Drilling experience has indicated
that when seismic velocities are processed and conditioned as
outlined in this paper, one can obtain pressures to within 0.50
ppg at target depths in the deepwater, provided that the low-
frequency trends in the seismic interval velocity curves are of
good quality and lie within 510% of well velocities. However,
FIG. 29. A detailed comparison of the computed pore pres-
sure at the well location of the data shown in Figure 28 us-
ing poststack inversion of seismic data and the RFT data. The
low-frequency trend of this gure is based on the Dix velocity
data of Figure 28 at the well location.
seismic velocities still lack resolution, even when including
the high-frequency components provided by various inversion
methodologies. Quite often, pressure regressions within thin
overpressured sands and shales (thickness less than a quar-
ter wavelength of the dominant frequency of seismic waves)
are not detected by seismic techniques prior to drilling. This
can result in expensive downtime. At present, there still is no
consistently reliable methodology to predict the occurrence of
such an event.
Velocity smoothing, calibration, and interpretation are im-
portant andessential steps inthe process leading toseismic pre-
diction of pore pressure. The industry has suffered a great deal
in this area due to two factors: routine use of commonly avail-
able stacking velocities without proper conditioning for pres-
sure prediction, and lack of communication with the drilling
community about the limitations of seismic velocity informa-
tion. For example, picking stacking velocities in the intervals
shorter than, say, 50 ms can result in a reversal of interval ve-
locity. This will lead to a false prediction of a pore pressure
(reversal of pore pressure). Further, such reversals are often
exaggerated by trace interpolation schemes in the seismic dis-
play programs. These are false indicators of geology and have
nothing to do with the pressure compartments or reversals as-
sociated with the real geology.
Before using any type of inversion technique, a great deal
of attention must be paid to the low-frequency trend obtained
either fromband-passed well data or froma variation based on
FIG. 30. Predicted pressure using tomography inversion procedure for a Gulf of Mexico prospect (fromLee et al., 1999). Pressure
is in pounds/gallon (ppg).
FIG. 31. Predicted acoustic impedance and pore pressure from
full waveformandprestackinversionof the CMPgather shown
at the left using the GA algorithm as discussed in the text and
in Figure 18.
2038 Dutta
Dix-type models. Otherwise, the inversion can yield an appar-
ently noisy velocity function rendering any pressure prediction
totally useless. Not all seismic data can be inverted for high-
frequency velocity information, especially if the S/N ratio is
low. Such is the case in the subsalt and subbasalt areas. For
prestack inversion, the data must have sufcient offset range
at larger depths (angles greater than 35
o
) so that the move-
out velocities can be derived with some condence. Otherwise,
the amplitude analysis for high-frequency velocity estimation
would be fraught with uncertainty.
There is a current misunderstanding about the utility of
depth imaging velocities for pore pressure work. Even if one
obtains a focused image by this process, it can not guarantee
that the associated velocities will be useful for lithology and
uid work. Al-Chalabi (1994) clearly articulated this. He also
recommended a process for conditioning migration velocities
so that they become suitable for such analysis. Whether it will
be sufcient for pore pressure prediction remains to be seen.
While I was reviewingtherecent literatureontheapplication
of seismic data for pressure prediction, it became obvious that
most of these applications dealt with Tertiary clastics (i.e., sand
and shale sequences). Further, the authors made either explicit
or implicit assumptions that the pressure mechanism is due to
mechanical compaction. Such is not always the case, as, for
example, in carbonate provinces. Pressure mechanisms in car-
bonates are poorly understood, and their seismic signature is
usually absent in the seismic velocity eld. Additionally, veloc-
ity variations in carbonates are often controlled by secondary
porosity and fractures. Thus pressure variations in carbonate
might not be detected by the methodologies discussed in this
paper.
The mechanical compaction model assumes that there is a
normal compaction trend of velocity versus depth. This trend
suggests an increase of velocity with depth. Geopressuring,
then, would cause a velocity reversal. However, not every ve-
locity reversal obtained from seismic data is indicative of geo-
pressure. Thus, there is an inherent ambiguity in the seismic
velocity eld: is the trend variation due to pressure or lithol-
FIG. 32. A procedure for real-time calibration, update, and look-ahead prediction of high-frequency velocities
using VSP and seismic data. These velocities can be used to update the pressure model in real time (namely, on
the rig while a well is being drilled and when there is a pause in drilling while a casing is being set). RC=reection
coefcient, VA=velocity analysis.
ogy? This ambiguity becomes more acute when using high-
frequency velocity from various seismic inversion methodolo-
gies. One then would ask additionally whether the velocity
variation has something to do with uid substitution in hy-
drocarbon reservoirs, namely, a lighter uid such as oil or gas
replacing a denser uid (brine). This could also cause a lower-
ing of velocity in sandstones, and thus compound the problem.
Such issues can only be dealt with by interpretation and mod-
eling (uid substitution, for example), but one cannot remove
all the ambiguities.
What measure can one use to determine how well a certain
pressure prediction methodology works? Is a comparison of
the predicted pressure at a depth with the measured uid pres-
sures at the same depth such as those obtained from RFTs
or drill-stem tests sufcient? That is not necessarily the case.
FIG. 33. A comparison of velocities from seismic, drill-bit seis-
mic, and VSP. This suggests a way for updating the seismic ve-
locity model in real time and predicting a velocity eld ahead
of the drill bit in real time (taken from Meehan et al., 1998a, b)
While this comparison is conceptually simple, the reality is far
fromthe truth: the formation pressures are always measured in
permeable formations (sandstone), whereas most of the pres-
sure indicators (rock-modelbased transforms) are only valid
for impermeable formations (shales). So how can one obtain a
true measurement of pore pressure in the shales? If the shales
are in hydraulic communication with adjacent sandstones, then
there is no problem. However, this is not always the case. Fur-
ther, before a well is drilled, one does not know the exact lo-
cation of shales and sandstones, let alone whether they are in
hydraulic communication or not. Thus, in the predrill sense, the
seismically predicted trend of pressure with depth is perhaps
the most reliable indicator. The details of pressure regressions
and other variations can be reliably tracked only if a calibra-
tion well is available within the same geologic and formation
environment.
CONCLUSIONS
Velocity conditioning is critical to a successful pressure anal-
ysis. This requires better seismic acquisition, processing, and
interpretation techniques. The conventional stacking velocities
fromseismic data are usually unsuitable for pressure work, un-
less these velocities are processed to yield rock or propagation
velocities (Al-Chalabi, 1994) by integration either with a rock-
physical model or with velocities from surrounding wells. We
must make every effort to include lateral changes in velocities
to obtain consistent pressure images. Velocity analysis must in-
clude steps to correct for anisotropy. Otherwise, the estimated
velocities couldbe muchfaster thanthe true velocities. This will
FIG. 34. Real-time prediction of pore pressure from drill-bit SWD data from a well drilled in Vietnam (taken from Meehan
et al., 1998c)
underestimate pore pressure consistently, even to the extent of
producing pressures below their hydrostatic values.
For pressure work based on seismic velocities, a variety of
tool kits can be recommended. The data quality and the pur-
pose of the application dictate the choice of a particular tool.
In this respect, the scale at which pressure predictions are to
be made should be considered.
Inversion techniques are extremely valuable. They add reso-
lution, but always require a reliable low-frequency model and
calibration. Prestack migration methods are essential for pres-
sure analysis in structurally complicated areas such as subsalt
basins. However, any inversion technique, no matter whether
it be poststack or prestack, must start with a reliable and petro-
physically acceptable rock-velocity model.
Many applications in the deepwater using the methodology
discussed in this paper show that with proper analysis and con-
ditioning of seismic data, one can obtain pore pressures to
within 0.5 ppg at the typical target depths (10 00015 000 ft,
or 30004500 m). However, there is a serious limitation to a
methodology mainly based on seismic information. Methods
based on seismic information are not reliable if beds are thin-
ner than typical wavelengths contained in the velocity curve.
These hidden beds are usually the source of unexpected pres-
sure problems while drilling.
Velocities interpretation is a critical step in this process, one
that should not be done separately. An interdisciplinary team,
including geologists, geophysicists, petrophysicists, and drilling
engineers, will do it best. Further, pressure prediction is es-
sentially an expectation and risk management process. Team-
work and open communication between the team members,
2040 Dutta
especially regarding issues of the limitations of the technology,
are important elements for success of the work.
THE ROAD AHEAD
Although much progress has been made in the use of seismic
velocity and amplitude for pressure prediction, there are still
challenges ahead. Perhaps the most daunting is in the subsalt
and subbasalt areas, where seismic images are unreliable and
velocities are questionable. A key question is: are the prestack
depth imaging velocities in these areas suitable for pressure
prediction? Al-Chalabi (1994) has clearly stated that a well-
focusedimage does not equate withaccurate depths. This state-
ment also applies for pressure prediction; a well-focused image
does not equate with a reliable pressure image. Researchers
need to address this problem since the industry expends con-
siderable resources in predicting pressures below salt.
There are other methods available to us beside seismic. For
example, the basin modeling approach may be quite suited for
subsalt pressure prediction, although I know of no publication
on this subject. However, the parameters in the basin modeling
approach (for example, permeability of various rocks and their
compaction properties) are numerous, and not all of them can
be handled easily prior to drilling. Predicted results are very
sensitive to these parameters. Perhaps an approach that com-
bines the basin modeling approach with seismic techniques is
a step forward in this area.
Ahead-of-the-bit pressure prediction in real time is also a
challenge. This has two components: hardware and a t-for-
purpose methodology. The hardware has to do with the devel-
opment of various SWD tools and getting the downhole data
in real time to the surface for applications (i.e., telemetry). The
limitation of this methodology depends on development of a
procedure that will yield look-ahead velocity from the SWD
data, again in real time, so that a pressure prole ahead of the
bit can be generated while drilling. Data turnaround time and
impacting the drilling process are both critical issues. A key
question in this regard is: what is meant by real time? Do we
want ahead-of-the bit pressure during every turn of the bit?
Or is it sufcient to have pressures updated during a pause
in the drilling, say during casing? Answers to these questions
are critical to developing an appropriate methodology for real-
time pressure prediction.
As stated in this paper, rock models are vital to a successful
application of pressure prediction from seismic data. For ex-
ample, there are relationships among velocity, porosity, effec-
tive stress, temperature and lithology, with built-in parameters
that can be obtained from conventional well logs or laboratory
data. Our current understanding of rock models is rather poor
in several respects; for example, in the case of soft clastics, car-
bonates, and rocks altered by diagenesis. A related issue deals
with rock models upscaled to match seismic data. We need to
develop more predictive rock models with fewer built-in pa-
rameters.
The use of seismic inversion technologies for pressure pre-
diction is a very promising area, especially with continued in-
crease in computing power. The mathematics underlying these
techniques is well understood, and related computer algo-
rithms are plentiful. What is lacking, however, is a step change
in the inversion technology to impose rock-physicsbased con-
straints on the inversion for various seismic attributes, includ-
ing, but not limited to, velocities. Such constraints are essential
inconditioning velocities fromaninversionprocess sothat they
become useful for pressure analysis work.
ACKNOWLEDGMENT
The author is grateful to WesternGeco for allowing him to
work on this paper and to publish it.
REFERENCES
Aki, K., and Lee, W. H. K., 1976, Determination of three-dimensional
velocity anomalies under a seismic array using rst P arrival times
fromlocal earthquakes 1. Ahomogeneous initial model: J. Geophys.
Res., 81, 43814399.
Al-Chalabi, M., 1973, Series approximations in velocity and traveltime
computations: Geophys. Prosp., 21, 783795.
1994, Seismic velocityA critique: First Break, 12, 589596.
Archie, G. E., 1942, Theelectrical resistivitylogas anaidindetermining
some reservoir characteristics Trans. Am. Inst. Mech. Eng., 146, 54
62.
Athy, L. F., 1930, Compaction and oil migration: AAPG Bull., 14, 25
36.
Barker, C., 1972, Aquathermal pressuringRole of temperature in
development of abnormal pressure zones: AAPG Bull., 56, 2068
2071.
Bilgeri, D., andAdemeno, E. B., 1982, Predicting abnormally pressured
sedimentary rocks: Geophys. Prospect., 30, 608621.
Bishop, T. N., Bube, K. P., Cutlers, R. T., Langan, R. T., Love, P. L.,
Resnick, J. R., Shuey, R. T., Spindler, D. A., and Wyld, H. W., 1985,
Tomographic determination of velocity and depth in laterally vary-
ing media: Geophysics, 50, 903923.
Bowers, G. L., 1995, Pore pressure estimation from velocity data: Ac-
counting for overpressure mechanisms besides undercompaction:
SPE Drilling and Completions, June, 119.
Chiu, S. K. L., and Stewart, R. R., 1987, Tomographic determination of
three-dimensional seismic velocity structure using well logs, vertical
seismic proles, andsurface seismic data: Geophysics, 52, 10851098.
DeKok, R., Dutta, N. C., Mallick, S., Utech, R., andVauthirn, R., 2001a,
Shallow waterow detection using prestack inversion, Ann. Mtg.,
Eur. Assn. Geosci. Eng., Extended Abstracts, 841843.
DeKok, R., Dutta, N. C., Khan, M., andGelinsky, S. G., 2001b, Deepwa-
ter geohazarddetectionusing prestackinversion: 71st Ann. Internat.
Mtg., Soc. Expl. Geophys., Extended Abstracts, 613616.
Dix, C. H., 1955, Seismic velocities from surface measurements:
Geophysics, 20, 6866.
Dutta, N. C., 1983, Shale compaction and abnormal pore pressuures: A
model of geopressures in the Gulf Coast Basin: 53rd Ann. Internat.
Mtg., Soc. Expl. Geophys., Expanded Abstracts, 542544.
Ed., 1987a, Geopressure: Soc. Expl. Geophys.
1987b, Fluid owin lowpermeable porous media; in B. Doligez,
Ed., Migration of hydrocarbons in sedimentary basins: Editions
Technip, 567595.
1997, Pressure prediction from seismic data: Implications for
seal distribution and hydrocarbon exploration and exploitation in
the deepwater Gulf of Mexico; in Moller-Pederson, P., and Koestler,
A. G., Eds., Hydrocarbon seals: Elsevier Press.
Dutta, N. C., Borland, W. H., Leaney, S. W., Meehan, R., and Nutt,
W. L., 1998, Pore pressure ahead of the bit: An integrated approach:
Am. Assn. Drilling Eng. ForumonPressure Regimes inSedimentary
Basins and Their Prediction.
Dutta, N. C., Khan, M., and Gelinsky, S. G., 2001, Seismic predrill pore
pressure imaging using a deepwater rock model: Ann. Mtg. Eur.
Assn. Geosci. Eng., Extended Abstracts, 742744.
Dutta, N. C., and Ray, A., 1996, Subsurface image of geopressured
rocks using seismic velocity and acoustic impedance inversion: 58th
Annual Meeting Eur. Assoc. Geosci. Eng., Amsterdam (extended
abstract).
1997, Image of geopressured rocks using velocity and acous-
tic inversion of seismic data: 67th Ann. Internat. Mtg., Soc. Expl.
Geophs., Extended Abstracts, 19291930.
Eaton, B. A., 1968, Fracture gradient-prediction and its application in
oil eld operations: J. Pet. Tech., 10, October, 13531360.
1972, Graphical method predicts pressure worldwide: World
Oil, 182, 5156.
Fertl, W. H., 1976, Abnormal formation pressures: Elsevier.
Gardner, G. H. F., Gardner, L. W., and Gregory, A. R., 1974, Formation
velocity and densityThe diagnostic basics of stratigraphic traps:
Gjoystdal, H., and Ursin, B., 1981, Inversion of reection times in three
dimensions: Geophysics, 46, 972983.
Hottmann, C. E., and Johnson, R. K., 1965, Estimation of formation
pressures from log-derived shale properties; J. Pet. Tech., 6, June,
717722.
Hower, J., Eslinger, E. V., Hower, M. E., and Perry, E. A., 1976, Mech-
anism of burial metamorphism of argillaceous sediment. I. Miner-
alogical evidence: Geological Society of America Bulletin, 87, 725
737.
Hubbert, M. K., and Rubey, W. W., 1959, Role of uid pressures in
mechanics of overthrust faulting: Geol. Soc. Am. Bull., 70, 115166.
Hubral, P., andKrey, T., 1980, Interval velocities fromseismic reection
time measurements: Soc. Expl. Geophys. Monograph.
Huffman, A., and Castagna, J., 2001, The petrophysical basis for
shallow-water ow in the Deepwater Gulf of Mexico: The Leading
Edge, 20, 10301035.
Issler, D. R., 1992, A new approach to shale compaction and strati-
graphic restoration; Beaufort-McKenzie basin and McKenzie corri-
dor, North Canada: AAPG Bull., 8, 11701189.
Kolb, P., Collino, F., and Lailly, P., 1986, Pre-stack inversion of a 1-D
medium: Proc. IEEE, 74, 498508.
Lavergne, M., and Willm, C., 1977, Inversion of seismograms and pseu-
dovelocity logs: Geophys. Prosp., 25, 231250.
Lee, S., Shaw, J., Ho, R., and Steeb, D., 1999, Deepwater reservoir
prediction using seismic and geomechanical methods: The Leading
Edge, 18, 726728.
Lindseth., R. O., 1979, Synthetic sonic logsAprocess for stratigraphic
interpretation: Geophysics, 46, 12351243.
Magara, K., 1978, Compaction and uid migration: Elsevier.
Mallick, S., 1995, Model-based inversion of amplitude-variation with
offset data using a genetic algorithm: Geophysics, 60, 939954.
1999, Some practical aspects of prestack waveform inversion
usingagenetic algorithm: Anexamplefromtheeast Texas Woodbine
gas sand: Geophysics, 64, 326336.
Mathew, W. R., and Kelly, J., 1967, How to predict formation pressure
and fracture gradient: Oil and Gas J., 65, 92106.
McMechan, G., 1983, Seismic tomography in boreholes: Geophys. J.
Roy. Astr. Soc., 74, 601612.
Meehan, R., Nutt, L., Dutta, N., andMenzies, J., 1998a, Drill-bit seismic
technology1: Seismic information helps predict drilling hazards,
choose casing points: Oil and Gas J., May 11, 5360.
1998b, Drill-bit seismic technologyConclusion: Case histories
showreal-time information reduces uncertainty: Oil and Gas J., May
18, 5459.
1998c, Drill-bit seismic: Adrillingoptimizationtool: IADC/SPE
Drilling Conference, paper 39312.
Mora, P., 1987, Nonlinear two-dimensional elastic inversion of multi-
offset seismic data: Geophysics, 52, 12111228.
Nur, A. M., and Wang, Z., Eds., 1989, Seismic and acoustic velocities
in reservoir rocks, 1: Experimental studies: Soc. Expl. Geophys.
Eds., 1992, Seismic and acoustic velocities in reservoir rocks, 2:
Theoretical and model studies: Soc. Expl. Geophys.
Palciauskas, V. V., and Domenico, P. A., 1989, Fluid pressures in de-
forming porous rocks: Water Resources Research, 25, 203213.
Pennebaker, E. S., 1968, Seismic data indicate depth and magnitude of
abnormal pressure: World Oil, 166, 7382.
Pickett, G. R., 1963, Acoustic character logs and their application in
formation evaluation: J. Petr. Tech., 15, 659667.
Prasad, M., 2002, Acoustic measurements in unconsolidated sands at
low effective pressure and overpressure detection: Geophysics, 67,
405412.
Raiga-Clemenceau, J., Martin, J. P., and Nicoletis, S., 1988, The concept
of acoustic formation factor for more accurate porosity determina-
tion from sonic transit time data: The Log Analyst, 29, 5460.
Reynold, E. B., 1970, Predicting overpressuredzones withseismic data:
World Oil, 171, 7882.
Reynolds, E. B., 1973, The application of seismic techniques to drilling
techniques: Soc. Petr. Eng. Preprint 4643.
Reynolds, E. B., May, J. E., and Klaveness, A., 1971, Geophysical as-
pects of abnormal uid pressures, in Abnormal subsurface pressure:
A study group report 19691971: Houston Geol. Soc., 163168.
Rubey, W. W., and Hubbert, M. K., 1959, Role of uid pressure in
mechanics of overthrust faulting, II. Overthrust belt in geosynclinal
area of western Wyoming in light of uid pressure hypothesis: Geol.
Soc. Am. Bull., 70, 167205.
Smith, J. E., 1971, The dynamics of shale compaction and evaluation
of pore-uid pressure: Math. Geol., 3, 239263.
Stump, B., Flemings, P. B., Finkbeiner, T., and Zoback, M. D., 1998,
Pressure differences between overpressure sands and bounding
shales of the Eugene Island 300 eld (offshore Louisiana, U.S.A.)
with implications for uid ow induced by sediment loading: Pre-
sented at Overpressures in Petroleum Exploration Conf.
Taner, M. T., and Koehler, F., 1969, Velocity spectraDigital computer
derivation and applications of velocity functions: Geophysics, 34,
859881.
Tarantola, A., 1984, Inversion of seismic reection data in the acoustic
approximation: Geophysics, 49, 12591266.
Terzaghi, K., 1943, Theoretical soil mechanics: John Wiley and Sons,
Inc.
Traugott, M. O., 1997, Pore/fracture pressure determinations in deep
water: World Oil August, Deepwater Technology supplement.
Vernik, L., 1994, Predicting lithology and transport properties from
acoustic velocities basedonpetrophysical classicationof siliclastics:
Wyllie, M. R. J., 1957, The fundamentals of log interpretation, 2nd ed.:
Academic Press.
Wyllie, M. R. J., Gregory, A. R., and Gardner, G. H. F., 1956, Elastic
wave velocities in heterogeneous and porous media: Geophysics, 21,
4170.
1957, An experimental investigation of factors affecting elastic
wave velocities in porous media: Geophysics, 23, 459493.
Yilmaz, O., 1987, Seismic data processing: Soc. Expl. Geophys.

1527101

Загружено:

Сведения о документе

Исходное описание:

Авторское право

Доступные форматы

Поделиться этим документом

Поделиться или встроить документ

Параметры публикации

Этот документ был вам полезен?

Это неприемлемый материал?

Авторское право:

Доступные форматы

1527101

Загружено:

Авторское право:

Доступные форматы

GEOPHYSICS, VOL. 67, NO. 6 (NOVEMBER-DECEMBER 2002); P. 20122041, 34 FIGS., 4 TABLES.

Вам также может понравиться