SPE 103856

Multistage Approach on Pore Pressure Prediction—A Case Study in South China Sea
Yuhong Xie, CNOOC Ltd.–Zhanjiang; Shunhua Cao, Chang Liu, Gongrui Yan, Schlumberger; and Ping Yi, Jun Cai, and
Liqiang Xu, CNOOC Ltd.–Zhanjiang

Copyright 2006, Society of Petroleum Engineers

Introduction
This paper was prepared for presentation at the 2006 SPE International Oil & Gas Conference One of the drilling challenges in the South China Sea is the
and Exhibition in China held in Beijing, China, 5–7 December 2006.
overpressure environment. The abnormally high fluid
This paper was selected for presentation by an SPE Program Committee following review of
information contained in an abstract submitted by the author(s). Contents of the paper, as
pressures in the overpressured formations are extremely
presented, have not been reviewed by the Society of Petroleum Engineers and are subject to hazardous when the drill bit approaches such formations. Lost
correction by the author(s). The material, as presented, does not necessarily reflect any
position of the Society of Petroleum Engineers, its officers, or members. Papers presented at controls can result in blowouts. The pore pressure is a critical
SPE meetings are subject to publication review by Editorial Committees of the Society of
Petroleum Engineers. Electronic reproduction, distribution, or storage of any part of this paper
point of any drilling project as it influences the definition of
for commercial purposes without the written consent of the Society of Petroleum Engineers is the mud weight, casing-setting depths, casing design, riser
prohibited. Permission to reproduce in print is restricted to an abstract of not more than
300 words; illustrations may not be copied. The abstract must contain conspicuous safety margin, etc.
acknowledgment of where and by whom the paper was presented. Write Librarian, SPE, P.O.
Box 833836, Richardson, TX 75083-3836, U.S.A., fax 01-972-952-9435.
The mechanism of overpressure buildup has been
extensively studied since the late 1950s.1–7 A number of
Abstract overpressure-generating mechanisms have been proposed,
A multistage approach on pore pressure prediction was which can be divided into three main groups depending on
successfully used for a drilling program in the South China their porosity/effective-stress relationships and geo-stress
Sea. The objectives were to optimize the casing-shoe position regime: 1) Loading mechanism, also called the
by predicting the overpressure profile in the formation and to undercompaction mechanism, occurs when the rate of shale
select the optimal mud weight by forecasting the formation sedimentation is high, thus preventing fluid to escape from
pressure variation. pore space in the underlying formations; 2) Unloading
Before spudding the well, data from a surface seismic mechanism, which is caused by fluid expansion or erosional
survey and log data from a nearby well were collected to build uplift that unloads the sediment column and results in rapid
a pore pressure uncertainty model. This initial model had a pore pressure increases. The possible fluid-expansion
pore pressure profile with relatively large uncertainties. While mechanisms include aquathermal expansion, hydrocarbon
drilling, the model was updated in stages with additional generation and cracking, mineral dehydration (smectite/illite,
measurements, such as those derived from vertical seismic gypsum/anhydrite), and thermal expansion related to
profiles (VSPs), dipole sonic logs, gamma ray (GR) logs, etc. temperature increase; 3) Tectonic stress mechanism, which
These additional data significantly reduced the uncertainty of occurs in the tectonic areas when the rate of fluid escape can
pore pressure estimation. not keep up with the additional compaction caused by the
At each intermediate total depth (TD), wireline logs and tectonic stress; pore pressure will increase as a result.
VSPs were acquired. Pore pressure profile above the current Commonly, more than one group of mechanisms may be
TD was estimated using the wireline logs and VSP velocity. present within a single basin.
Pore pressure profile below the current TD was estimated Current state-of-art methods for detecting overpressured
using the look-ahead VSP inversion result. Above the current formations can be classified into at least four major categories:
TD, a cross-check validation was made by comparing the pore
pressure profiles estimated from wireline logs and VSP data. 1. Drilling methods,
The predicted pore pressure model was calibrated with the 2. Well-logging methods,
actual mud weight and a well integrity test at 3505 m. 3. Seismic methods, and
At the well-planning stage, the maximum pore pressure 4. Gravity potential field methods.
was estimated to be 1.55 g/cm3. At intermediate drilling
stages, the pore pressure was consistently estimated to be The first two categories cannot be performed unless wells
between 1.9 and 2.1 g/cm3 below a depth of 4100 m. This that penetrate the top of an overpressured zone are drilled.
prediction was later confirmed by the formation test pressure Thus, they have only limited predicting capabilities, although
value of 1.98 g/cm3 at 4226 m. they have the most accurate detecting capabilities. The gravity
All pore pressure prediction results were delivered within methods are often far too ambiguous to determine whether the
six hours after each intermediate data acquisition. The pore observed gravity anomalies are from overpressured formations
pressure prediction results were successfully used for the or massive shales. The reflection seismic method can map the
critical decision of setting the casing shoe and selecting the top of overpressure distribution in a basin with a resolution of
mud weight. around 100 m, and this resolution can be significantly
2 SPE 103856

improved by combining VSP and logging-while-drilling integrated methodology of the uncertainty model was
(LWD) data. successfully implemented in a multistage manner.
Because acoustic impedance depends on porosity through
both velocity and density, overpressure can be interpreted as a Analysis Technique
zone of anomalously low acoustic impedance. The look-ahead Conventional pore pressure prediction methods are based on
VSP inversion can be used to estimate the acoustic impedance compaction theory, which use correlations between pore
profile ahead the bit and, therefore, accurate and reliable pore pressure and some measured properties of the formation. The
pressure can be predicted for the drilling program. fundamental assumption is based upon Terzaghi’s original
The Eaton Method8 used here is based upon the same equation:
premise that when acoustic values of clean shale are read
directly from logs and plotted as a function of depth on semi- p p = σ v − σ v ' , ………………………………………(1)
log paper, there exists a normal trend line through the
normally pressured section. A deviation of the log-derived where σv and σv' are the total and effective overburden
values from this normal trend line is indicative of abnormal pressure, respectively. The total pressure σ v is the pressure
pore pressure.
These estimates can be calibrated with pressure data exerted by the weight of overlying sediments:
acquired during drilling. Because all the measurements have Z
their own uncertainties, hence the pore pressure prediction σ v = g ∫ ρ b dz , ………………………………………(2)
derived from these inaccurate data is inherently inconsistent. 0
Factors that control the accuracy of pore pressure prediction
include: uncertainties of data, deviations of the model used to where g is the acceleration of gravity, and ρ b is the bulk
describe the overpressure mechanism, the availability of density of the rock.
information and data at different stages of a project, as well as Eaton’s method8 relates the effective overburden stress
the combination of different types of data input into the model.
The uncertainty model for pore pressure prediction σ v ' directly with the acoustic or the resistivity values of clean
(Malinverno et al.9) provides a methodology that integrates shale as:
information from many different sources of data (surface
seismic, wireline or VSP, mud weights, pore pressure σ v ' = (σ v − pnorm ) × (V / Vnorm ) n , …………………(3)
measured by formation tests, fracture pressures from leakoff
tests) and predicts pore pressure and fracture gradients with where pnorm is the normal (hydrostatic) pore pressure, Vnorm
associated uncertainties. These uncertainties are controlled by is the expected normal value of formation velocity where there
the extent and accuracy of available data. In the well-planning is no overpressure, V is the measured formation velocity, and
phase, the mud-weight window—the predicted pore pressure n is the Eaton’s exponent that he original suggested to be
and fracture gradient—will have relatively large uncertainties. around 3.0 for the overpressure caused by the
While drilling, the method allows for updating the mud-weight
undercompaction mechanism.8 The Vnorm term is defined in
window in real time or in stages with additional
measurements, such as borehole seismic data, wireline logs, the normal trend line that reflects the trend in sediments of
while-drilling measurements, etc. These additional data decreasing porosity. Then the pore pressure can be computed:
typically lead to an update in the window and a reduction in its
uncertainty. It could greatly benefit the planning at critical p p = σ v − (σ v − pnorm ) × (V / Vnorm ) n . ……………(4)
times during drilling. In a multistage pore-pressure prediction project, the first
This case study describes how the integrated methodology step is to build the pore pressure model prior to spudding. It
of the uncertainty model was successfully implemented in a can be done from measurements at offset wells and from
multistage manner. Before spudding, the analysis of both surface seismic data. If both sonic log and pressure
seismic stacking velocities and the well data of a nearby well information are available at the offset wells, the pore pressure
indicated that there was a high probability of encountering an model and normal trend line can be built and calibrated. Then,
overpressured zone at the planned well location. Therefore, the the model from the offset wells can be extrapolated or
drilling program included the pore pressure prediction propagated to the proposed well location as the start model of
program with the acquisition of intermediate VSPs to pore pressure before spudding. If only surface seismic data are
determine the mud weight and the accurate placement of available, the pre-spud pore pressure model can be built, but
casing shoes. usually with a larger uncertainty.
First, we introduce the technique for pore pressure To build the pore pressure model from an offset well
prediction based on the compaction theory. Subsequently, we before spudding, the following steps are normally taken:
describe how to use an uncertainty model to get more reliable
pore pressure results and quantify the uncertainty. This 1. Audit the geological information and offset-well data.
uncertainty model is based on Bayesian approach and has the 2. Evaluate the sonic log from top to bottom for velocity
ability to combine various measurements acquired during values in the clean shales.
drilling with the prior-to-spudding pore pressure model.
Finally, the procedure is applied to a pore pressure prediction
project in the South China Sea, which illustrates how the
SPE 103856 3

3. Determine σv from density data and p p from the prior-to-spudding model) and a likelihood function
known normal pressure gradient or water salinity P[d1 , d 2 , ..., d n | V , ρ , C pore ] .
data. As pointed out by Malinverno et al.,9 the likelihood
4. Establish the normal compaction trend line. function quantifies how well any choice of V , ρ , and C pore
5. Calculate (V / Vnorm ) 3 values at various depths. fits the data acquired while drilling. The above algorithm
6. Use Eq. 4 to calculate pore pressure. suggests that an updated probability distribution of data
7. Modify the trend line by calibrating the computed needed for pore pressure prediction can be computed by
pore pressure with measured pressure information, incorporating the prior model and new data acquired during
such as the mud weights, pore pressure measured by drilling, hence more reliable pore pressure results can be
formation testing, fracture pressures from leakoff estimated.
tests, etc., until a satisfactory model for this offset
well is reached. Pre-Spudding Pore Pressure Model
8. Extrapolate the model to the designated well. In our case study, a pore pressure profile was estimated
from surface seismic and data from an offset well during the
In general this pre-spud model has a large uncertainty. To well-design phase. Both the offset well and the planned well
reduce the uncertainty and obtain a more accurate and reliable were vertical exploration wells in water depth around 50 m.
prediction, drilling information and measurements (for The distance between these two wells is about 5000 m and
example, LWD and VSP data) should be taken into account to surface seismic interpretation indicates a possible fault
update the model. To combine these data from different between the offset well and the planned well. Figure 1 shows
stages, the Bayesian approach, which can quantify and the prediction of the pore pressure profile with the proposed
propagate the uncertainties of a model, was adopted in the possible shoe depth for 9.6-in. casing.
uncertainty pore pressure model of Malinverno et al.9 The As a part of the multistage pore pressure prediction program,
final prediction is the probability distribution of pore pressure the offset well data was further analysed and a pore pressure
resulting from the combination of multistage data and the model was obtained as shown in Figs. 2 and 3. Owing to the
probability model. availability of limited data for calibration, it was difficult to
In Bayesian theorem, the probability of an event B j establish a reliable normal velocity trend line. However, the
results from the model provided a reasonable prediction for
before new information A becomes available is called “prior” both pore pressure and fracture gradient profiles below the TD
probability, or a priori probability, whereas the probability of (Fig. 3).
B j after new information A becomes available is called
Depth(m)
Time(s)

SEISMIC VELOCITY (m/s) PORE PRESSURE GRADIENT (g/cm3)

“posterior” probability, or posteriori probability. The theorem
states:
P[ A | B j ]P[ B j ]
P[ B j | A] = , ………………(5)
∑k =1 P[ A | Bk ]P[ Bk ]
n
Main Target I
Pore Pressure
where P[.] is the probability of an event, and P[ B | A] is 1.48--1.55
1.48

the conditional probability of an event B , given that A has Main Target II

Pore Pressure
occurred. In our practical application of pore pressure 1.55--1.59
1.55
predication, this rule becomes:
P[V , ρ , C pore | d1 , d 2 , ..., d n ] ∝
P[d1 , d 2 , ..., d n | V , ρ , C pore ] • P[V , ρ , C pore ] , …
9 5/8“ Casing Shoe
……………………………………………………(6)
where V is the velocity, ρ is the density, C pore is a vector of
uncertain coefficients needed in Eqs. 2–4 for pore pressure
computation, and d1 , d 2 , ..., d n denotes the additional
information obtained during drilling. The posterior probability
of P[V , ρ , C pore | d1 , d 2 , ..., d n ] quantifies the uncertainty
Fig. 1—Velocity and estimated pore pressure profiles. Blue color
in V , ρ , and C pore after the new data are available during represents the data profile at the offset well location, and red
represents the predicted profile at the new well location.
drilling. It states that this posterior probability is proportional
to the production of a prior distribution P[V , ρ , C pore ] (from
4 SPE 103856

Fig. 2—Estimated pore pressure and fracture gradient (FG)

profiles at the offset well. The overburden gradient (OBG) is also
displayed for comparison .

Fig. 2 shows the pore pressure analysis at the offset well. Fig. 4—Logs were mapped from the offset well to the new well
location.
The left-most panel shows that the GR and caliper logs were
used for lithology classification. The second and third panels
from the left show that sonic delta time (DT) and its inverse Based on the location of the new well and the local
(velocity) were used to establish the normal formation stratigraphy from seismic interpretation, the log data (GR, DT
pressure trendline. The right panel shows that the pore and bulk density (RHOB)) from the offset well were mapped
pressure profile was obtained from the sonic DT log with to the new well location as shown in Fig. 4. These mapped
calibration from the available data. Fig. 3 shows the pore logs were used to build a prior pore pressure profile at the new
pressure results predicted from the uncertainty model for the well location as the starting point for the multistage approach
offset well. on pore pressure prediction.

Pore Pressure Prediction While Drilling

According to the pre-spud pore pressure profile estimation, an
intermediate VSP was acquired at a current TD of 3450 m, just
above the onset of the predicted formation overpressure
transition. The intermediate VSP dataset was processed to
obtain the velocity profile above the current TD and the
corridor stack. A smoothed interval velocity was obtained
from the VSP transit times between 1000 and 3449 m. The
standard zero-offset VSP data-processing workflow was
applied to the intermediate VSP and a corridor stack was
obtained.
Under the constraint of the velocity profile above the
current TD, a Bayesian inversion was carried out to obtain the
acoustic impedance profile from the corridor stack. Then, the
acoustic impedance profile was converted into a velocity
Fig. 3—Estimated mud-weight window and uncertainty profile, assuming that the density and velocity is described by
distribution at the offset well.
Gardner’s equation. Figure 5 shows the process from the
corridor stack to the updated velocity profile.
SPE 103856 5

After the second intermediate VSP from 3449 to 4050 m

was acquired, it was merged with the first intermediate VSP.
With the combined dataset, more reliable inversion results
would be obtained to take the full advantage of the longer data
interval. The same prediction process was carried out and the
prediction result is shown in Fig. 7.

Velocity

Fig. 5—Inversion results from the corridor stack to the updated

velocity profile for the first intermediate VSP.

The newly derived velocity profile from the VSP inversion

was used to update the pore pressure profile. The result as
shown in Fig. 6 was available a few hours after the data
acquisition.

Fig. 7—Updated pore pressure after the second intermediate VSP.

Vertical scale in meters.

Fig. 7 shows the pore pressure prediction panel. The left

track shows the lithology classification based on the newly
acquired wireline logs. The second track shows DT from the
dipole sonic log and the trend lines from sonic and VSP data.
The third track shows DT obtained from the VSP inversion
and the trend lines from sonic and VSP data. The far right
track shows pore pressure profiles: the purple line for the VSP,
Fig. 6—Updated pore pressure after the first intermediate VSP. the dark purple line for the sonic, and the actual mud weight in
orange color.
Fig. 6 shows the pore pressure prediction panel. The left Because the dipole sonic data and other log data were
two tracks show the GR and DT mapped from the nearby well. acquired at the same time as the VSP, the velocity from sonic
The third track shows DT obtained from the VSP inversion. was used to cross check the pore pressure prediction results
The far right track shows pore pressure profiles: the blue line from the VSP data. The calibration data included the actual
for the upper limit, the red line for the lower limit, and the mud weight and a leakoff test (LOT) point at 3505 m.
dashed black line for the pre-spudding estimate. The second pore pressure prediction showed a pore
It was observed that there was a large difference between pressure profile similar to the first prediction. The predicted
the pore pressure prediction from VSP and the pre-spud pore pore pressure indicated that the high pore pressure between 1.9
pressure prediction below the current TD. The updated pore and 2.1 g/cm3 is expected below 4050 m, as shown in Fig. 7.
pressure gradient using VSP data (1000 to 3449 m) indicates The prediction was later confirmed by the formation-test
the increasing tendency of pore pressure from 1.3 g/cm3 at derived value of 1.98 g/cm3 at 4226 m, as shown in Fig. 8.
depth 3500 m up to 2.0 g/cm3 around 4250 m.
The significant difference in pore pressure between the
pre-spud prediction and while-drilling prediction was noticed
and discussed. Based on the well construction program, it was
decided to acquire the second intermediate VSP after drilling
another 600 m.
6 SPE 103856

Fm. Pre. Test confirmed the

Acknowledgments
PP Predictions from We would like to thank both CNOOC and
VSP2(vf2) & Sonic(vel) Schlumberger for permissions to publish this case
study.

References

1. Dickinson, G.: “Reservoir pressures in Gulf Coast Louisiana,”

Am. Assoc. Petrol. Geol. Bull. (1953) 37, 410.
2. Hottman, C.E. and Johnson, R.K.: “Estimation of formation
pressures from log-derived shale properties,” J. Pet. Technol.
LOT
(1965) 17, 717-723.
3. Bredehoeft, J.D. and Hanshaw, B.B.: “On the maintenance of
anomalous fluid pressures. I. Thick sedimentary sequences,”
Fm.
PreTest Geol. Soc. Am. Bull. (1968) 79, 1087-1106.
4. Burst, J.F.: “Diagenesis of Gulf Coast Clayey sediments and its
possible relation to petroleum migration,” Am. Assoc. Petroleum
Fig. 8 Comparison of predicted pore pressure and formation test
result. Geologists Bull. (1969) 53, 73-93.
5. Jones, P.H.: “Geothermal and hydrocarbon regimes, northern
Gulf of Mexico Basin,” Proceedings, First geopressured
In Fig. 8, the results of predicted pore pressure and the geothermal energy conference, University of Texas at Austin,
formation dynamic pressure test are displayed together for (1975) 15-89.
comparison. The left track shows the lithology classification 6. Hunt, J.M.: “Generation and migration of petroleum from
based on the wireline log data. The second track shows DT abnormally pressured fluid compartments,” AAPG Bull. (1990)
from the dipole sonic log and the trend line. The third track 74, 1-12.
shows DT obtained from the VSP inversion and the trend line. 7. Tigert, V. and Al-Shaieb, Z.: “Pressure seals: their diagenetic
The far right track shows pore pressure profiles: the blue line banding patterns,” Earth Science Reviews, (1990) 29, 227-240.
for the first intermediate prediction, the red line for the second 8. Eaton, B.A.: “The equation for geopressure prediction from well
logs,” paper SPE 5544 presented at the 1975 Fall Meeting of the
intermediate prediction, and the actual pressure value
Society of Petroleum Engineers of AIME, Dallas, Sept. 28 –
measured by the formation test. The comparison indicates that Oct. 1.
the pore prediction from the multistage approach accurately 9. Malinverno, A., Sayers, C.M., Woodward, M.J., and Bartman,
updated the pore pressure profile from the existing prediction R.C.: “Integrating diverse measurements to predict pore
below the current TD. pressure with uncertainties while drilling,” paper SPE90001
presented at the 2004 SPE Annual Technical Conference and
Conclusions Exhibition, Houston, Sept. 26-29.
The multistage approach for pore pressure prediction was
successfully implemented in the optimization of the drilling
program. All prediction results were delivered in time for the
decision-making and were proven accurate in comparison with
drilling results. The objectives of the pore pressure program
were achieved with a successful well reaching its target TD.
Thus, the multistage approach was a cost-effective approach in
managing the drilling program in an overpressure region.
This case study also shows that the accuracy and reliability
of pore pressure prediction was improved by using the
uncertainty model of combining additional measurements as
the drilling program progresses. Advanced warnings from pore
pressure prediction were used for scenario analysis to
minimize drilling costs and improve drilling safety. Overall,
the pore pressure prediction program yielded tremendous
savings.
VSP data from the pore pressure prediction program was
also used for the calibration of surface seismic. Multiple
applications of the same data were achieved with improved
efficiency of data usage and more values were obtained with
the same investment.