SPE 77354

Application of Quantitative Risk Analysis to Pore Pressure and Fracture

Gradient Prediction
Q. J. Liang, SPE, IPM Schlumberger

Copyright 2002, Society of Petroleum Engineers Inc.

This paper was prepared for presentation at the SPE Annual Technical Conference and
Exhibition held in San Antonio, Texas, 29 September 2 October 2002.
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
presented, have not been reviewed by the Society of Petroleum Engineers and are subject to
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
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
for commercial purposes without th e written consent of the Society of Petroleum Engineers is
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
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.

Abstract
The present deterministic pore pressure and fracture gradient
prediction techniques simplify the input variables from a
statistical range of data to a deterministic value, thereby
producing over-simplified deterministic pore pressure and
fracture gradient results. As a consequence, we lose the ability
to quantitatively analyze risks, and we cannot quantify the
uncertainties of our pore pressure and fracture
gradient prediction.
This paper describes a methodology for quantitative risk
analysis (QRA) of pore pressure and fracture gradient
prediction and shows that applications of QRA in this area
will improve the techniques of pore pressure and fracture
gradient calculations. In addition, QRA will open a full range
of new applications in risk prediction, risks evaluation, risk
management, decision making, real-time kick and loss-return
risk monitoring, risk control, and casing design.
Introduction
The present methods of pore pressure and fracture gradient
prediction are deterministic approaches that use seismic,
logging and drilling data to calculate pore pressure and
fracture gradient. The result of the prediction is a single-line
curve for pressure and for fracture gradient as shown in Fig. 1
The sources of input variables, such as density log for
formation bulk density, sand strength analysis log for
Poissons ratio, resistivity log, seismic or sonic log, and mud
log etc, determine that the input data for pore pressure and
fracture gradient prediction have randomness in their
statistical ranges at any given depth. Therefore, the output
results we get from the input data will be statistically
distributed. A Gaussian distribution can approximate the

probabilistic distribution of pore pressure.1 It is safe to assume

that the fracture gradient is a Gaussian distribution as well.
To drill safely, we must design the mud weight to balance
formation pressure and at the same time not to exceed the
formation fracture gradient. The current deterministic
techniques design the mud weight by adding a positive safety
margin to the estimated pore pressure and a negative margin to
the fracture gradient. The bottomhole equivalent hydrostatic
pressure or equivalent mud weight (EMW) is a function of
mud weight, annular resistance, mud rheological properties,
hole cleaning, and pipe moving speed. It is easy to understand
that the actual bottomhole EMW has a range of randomness.
Study shows that the bottomhole EMW is a
Gaussian distribution.2
The deterministic methods simplify the input variables;
thereby produce simplified pore pressure and fracture gradient
results. This eliminates the ability to quantitatively analyze the
risks
involved
in
pore
pressure
and
fracture
gradient prediction.
Fig. 2 shows the typical distribution of pore pressure,
fracture gradient, and bottomhole EMW. It also shows that the
three principle output results are combinations of randomly
distributed individual input variables. There will be risks of
EMW being lower than pore pressure and greater than fracture
gradient, and it is important to know the risks involved. The
deterministic method could not provide the quantitative risk
information on the probabilities of bottomhole EMW to fall
below formation pressure or to exceed fracture gradient.
The objectives of this paper are to describe a
methodology using the QRA method to predict pore pressure
and fracture gradient, and to explain the applications of the
QRA results.
QRA Methodology
QRA has been used widely in the construction industries, and
also has been used in casing design and well planning by the
oil and gas industries.1-3
The QRA approach considers the uncertainty of each
input variable and provides comprehensive statistical
properties of pore pressure, fracture gradient, and EMW, such
as means, deviations, and probabilities. The information will
be critical for high-pressure, high-temperature (HPHT) wells
when pore pressure and fracture gradient margin is low.
As shown in Fig. 2, in the present well design method the
engineer determines a pore pressure value A, a fracture

Q. J. LIANG

gradient value C, and then chooses a mud weight value B

in between to balance the pore pressure and not to exceed
fracture gradient.
In fact, pore pressure, fracture gradient of formation and
EMW at any given depth will have a bell curve distribution
as shown in Fig. 2. It is obvious that the overlapped (shaded)
area of curve PP (pore pressure) and curve EMW represents
the ris k of under-balance or taking a kick (RK ), and the
overlapped area of curve EMW and curve FG (fracture
gradient) represents the risk of loss-return or fracturing the
formation (RF ). The QRA approach can quantitatively
determine the risks of RK and RF.
The perimeters needed for quantitatively calculating the
risks are discussed in the following sections.
Mean Value. Mean value, , is the expected value or the
weighted average. It is defined as follows:

x
=

..(1)

Standard deviation. Standard deviation, , is a measure of

dispersion or variability. It measures how closely the values of
the ransom variable are clustered around the mean value. It is
given as:

(x

) 2

EP = E2 + P2 (5)
The risk of taking kick (RK ) at any given depth can be
determined by:

RK = 1 (

EP
) ....(6)
EP

Where designates the standard normal distribution, which is

a Gaussian distribution with parameters = 0 and = 1.0.
Similarly, the risks of loss-return (RF) can be calculated
by:

RF = 1 (

FE
) (7)
FE

Where:

FE = F E ...(8)
and

.(2)

Coefficient of Variance (COV). COV quantifies the

dispersion of the standard deviation over the mean value, the
higher the COV value, the more uncertain the data.

COV =

SPE 77354

..(3)

To calculate the risk of equivalent mud weight lower than pore

pressure, or the risk of taking a kick, RK , we must first
determine the means and standard deviations of pore pressure
data, P and P , and of equivalent mud weight data, E and E .
For normally distributed pore pressure and equivalent
mud weight data, the margin between the two probability
density functions (PDFs) is also normal distribution, and has a
mean margin of:

EP = E P .(4)
and its standard deviation (EP ) is also given as:

FE = F2 + E2 (9)
Applications
The applications of QRA in pore pressure and fracture
prediction will not only improve the techniques of pore
pressure and fracture gradient calculations, but also will open
a full range of new applications in risk evaluation, risk
management, decision making, real-time risk monitoring, well
control, and casing design.
Improving Pore Pressure and Fracture Gradient
Prediction Technique. At any depth, the input variables for
calculating pore pressure and fracture gradient, such as
overburden gradient, Poissons ratio, resistivity, sonic travel
time, and penetration rate (or d exponent), have randomness
in their statistical ranges. Fig. 3 to Fig. 74-8 demonstrate the
randomness of input variables. Therefore, the output results
that obtained from the input data will be statistically
distributed. Fig. 7 shows the randomness relationship between
overburden and actual measured fracture pressure.8 Research
shows that pore pressure has a Gaussian distribution.1 It is safe
to assume that bulk density, Poissons ratio, transit time and
shale resistivity, and d exponent are all Gaussian distributed,

SPE 77354

APPLICATION OF QUANTITATIVE RISK ANALYSIS TO PORE PRESSURE AND FRACTURE GRADIENT PREDICTION

and therefore, fracture gradient can also be assumed a

Gaussian distribution.
The deterministic pore pressure and fracture gradient
prediction techniques simplify the input variables from a
statistical range of data to a deterministic value, and therefore,
produce over-simplified deterministic pore pressure and
fracture gradient results. As a consequence, we cannot
quantify the uncertainties of our pore pressure and fracture
gradient prediction, and we lose the ability to quantitatively
analyze risks.
The probabilistic properties calculated from BPs large
numbers of actual pore pressure values1 indicated that, for
development wells COV ranges from 0.01 to 0.05, and for
exploration wells COV has a range of 0.05 to 0.15. It is
obvious that input data for development wells is usually more
resourceful than for exploration wells. Therefore, the
uncertainties of predictions for development wells are lower
than for exploration wells. For wildcat wells, offset
information may not be available; the uncertainties of pore
pressure and fracture gradient predictions will be substantially
higher than exploration wells.
Further more, the bottomhole equivalent hydrostatic
pressure or EMW is a function of mud weight, annulus
resistance, mud rheological properties, hole cleanness, and
pipe movement. The actual bottomhole EMW has a range of
randomness. Study shows that the bottomhole EMW is a
Gaussian distribution.2 As shown in Fig. 2, there will be
uncertainties of EMW to balance pore pressure and not to
exceed fracture gradient. The current deterministic techniques
design the mud weight by adding a positive safety margin to
the estimated pore pressure and a negative margin to the
fracture gradient. The deterministic techniques do not provide
a way to quantitatively determine the uncertainties (or risks) of
taking a kick (RK ) or loss-return (RF).
To calculate RK and RF, sufficient logging and drilling
data must be collected to produce input variables as we do in
deterministic pore pressure and fracture gradient predictions.
The sources of data are listed as follows:
Density log for formation bulk density
Sand strength analysis log for Poissons ratio,
Resistivity log for shale resistivity,
Seismic or sonic log for sonic travel time,
Mud log,
Formation testing,
Drilling records,
Mud reports,
Bit records.
The PDF of each input variable must then be developed,
and the probability properties of each PDF (, , and COV)
can be determined. Combining the input variable values and
probability properties with deterministic methods, we can
develop PDFs for pore pressure, fracture gradient, and EMW
as demonstrated in Fig. 2, and determine the probability
properties of each PDF. By applying the probability properties
of pore pressure, fracture gradient, and EMW, we can
calculate the uncertainties (RK and RF) from Eq. 6 and Eq. 7.

This information will be particularly critical for HPHT wells

at the depths where pore pressure and fracture gradient margin
are low.
Computer programs with application of Monte Carlo
simulations or similar methods can be used to calculate the
risks. In a Monte Carlo simulation, some or all of the input
variables are treated as probability distributions rather than
numbers. The output results are also presented in the form of
probability properties.
Management Reasoning for Decision Making. QRA results
can be used by management to determine the risks involved in
a project. Well control risk is one of the major risks that we
must identify when we decide to take on a well project.
Eaton9, 10 categorized wells as good, bad, or ugly for
low risk and cheap to drill, risky and expensive to drill, or
high risk and unprofitable to drill, respectively. His
categorizations are based on a deterministic pore pressure and
fracture gradient prediction method. His classifications are
helpful in making investment decisions. However, the good,
bad, or ugly categorizations cannot provide the answer to
how good, how bad, or how ugly a project is, and
cannot provide quantified risk analysis and reasoning.
From taking a kick or loss-return point of view, RK and RF
calculated from the QRA approach can provide quantified
numbers of risks involved. For example, assume a well has
RK = 25% at a given depth where pore pressure and fracture
gradient margin is low. This indicates that one out of four
wells in this depth will take a kick. Considering the time
needed to kill the well and daily cost for the operations,
engineers and management can find out how much
contingency should be planned for the project.
Fig. 9 11 shows an example of actual pore pressure,
calculated pore pressure and equivalent mud weight. Actual
kicks took place at point A and point B. In an HPHT well it is
often to see multiple zones with low margin of pore pressure
and facture gradient. The risks of each zone must be calculated
separately, and the total risks (RK-T and RF-T ) can be obtained.
If PNKi denotes the probability of a kick not happening at
zone i, and PNFi denotes the probability of loss-return not
happening at zone i, then PNKi and PNFi can be calculated by:

PNKi = 1 R Ki
PNFi = 1 RFi
Then the total risks, RK-T and RF-T , can be calculated by:
n

RK T = 1 PNKi (10)
i =1

Where:

Q. J. LIANG
n

NKi

= PNK 1 * PNK 2 * PNK 3 * ...PNKn

i =1

and
n

RF T = 1 PNFi (11)
i =1

Where:
n

NFi

= PNF 1 * PNF 2 * PNF 3 * ...PNFn

i =1

With the quantitatively calculated risks, management can

determine how good, how bad, or how ugly the project
is from well-control and drilling-difficulty point of views, and
can make decisions base on the quantified risk analysis
and reasoning.
Real-Time Monitoring and Warning. The statistical
distributions of the pore pressure and fracture gradient of a
formation in a given area must be generated for the existing
database. During drilling operations, real-time data from
logging tools such as Logging-While-Drilling (LWD) tools
can generate real-time pore pressure data, and the real-time
data will update the statistical database. By knowing the given
statistical distributions of pore pressure and fracture gradient,
we can control the statistical properties of EMW, such as
mean and standard deviation, to reduce the risks of kicking or
loss-return. This can be achieved by real-time monitoring of
the EMW distribution with tools such as APWD* Annular
Pressure While Drilling.
For example, at the kick points A or B in Fig. 9, the PDFs
of pore pressure and real-time EMW can be presented as
shown in Fig. 8. When real-time statistical EMW data indicate
that the risk of taking a kick is too high (COV1 value is too
high), without increasing mud weight, the drilling supervisor
can reduce the risk by reducing the COV value of EMW to
COV2 (COV1 > COV2 ). The supervisor can achieve this by
reducing mud weight variation, improving mud rheological
properties, improving hole-cleaning condition, reducing
annular resistance, controlling tripping speed, or controlling
penetration rate.
There is a need for a computer program to be developed
that would calculate the probabilities of pore pressure and
fracture gradient for a given area or a project from offset data,
combined with downhole pressure tools (such as LWD and
APWD tools) that provide real-time downhole data. The
program could calculate real-time pore pressure, fracture
gradient, and EMW statistical distributions and provide
outputs as shown in Fig. 2. These real-time results could be
used to monitor the risks of kicking or loss-return. They could
also provide warning when risks exceed a predetermined level.

Mark of Schlumberger

SPE 77354

Casing Design and Shoe Depth Selection. The QRA pore

pressure and fracture gradient calculation approach can
provide much more realistic information for casing design and
shoe depth selection.
QRA application in casing design (probability casing
design method1, 2 ) is revolutionizing casing design techniques
and could reduce casing costs significantly. Pore pressure and
fracture gradient are the most important input data for casing
design. Without statistically calculating and analyzing the pore
pressure and fracture gradient, there will be no real QRA
casing design.
The current casing design method does not quantify the
risks involved in casing shoe depth selection. The best we can
do is to guess at a range of casing setting depths for
contingency, and the risk considerations are non-quantitative
and ambiguous. As demonstrated in Fig. 10, applications of
QRA in pore pressure and fracture gradient calculations will
provide quantified risk results for the chosen casing setting
depth. It can also provide real-time risk monitoring when we
try to push the casing setting depth deeper than planned
during operations.
Conclusions and Recommendations
The classical deterministic pore pressure and fracture gradient
prediction techniques provide simplified and incomplete
information. In this paper we describe a general methodology
for quantitative risk analysis of pore pressure and fracture
gradient predictions.
The proposed QRA method provides useful tools of risk
prediction, risk management, and risk control for project
management,
well
planning,
casing
design,
and
drilling operations.
The needs to develop comprehensive databases of logs,
seismic data, and formation mechanical data for each project
field are obvious and critical. Computer programs must be
developed using the databases to efficiently provide accurate
statistical results. Existing tools such as Monte Carlo
simulation programs can be used. The computing interfaces
between programs and downhole real-time LWD and APWD
pressure tools need to be developed.

SPE 77354

APPLICATION OF QUANTITATIVE RISK ANALYSIS TO PORE PRESSURE AND FRACTURE GRADIENT PREDICTION

Acknowledgments
The author wishes to thank Schlumberger for its support and
permission to publish this paper.

4.

Reference
1.

Adams,

A.J.;

Parfitt,

S.H.L.;

Reeves,

T.B.;

Thorogood,

J.L.:Casing System Risk Analysis Using Structural Reliability,

Paper SPE 25693, Presented at the 1993 SPE/IADC Drilling
Conference, Amsterdam,23-25 February
2.

Nilson, T.; Sandoy, M. and Rommetveit, R.: Risk-Based Well

Control Planning: The Integration of Random and Known
Quantities in a Computerized Risk Management Tool, Paper SPE
68447, Presented at the 2001SPE/IcoTA Coiled T ubing
Roundtable, Houston, 7-8 March.

3.

Payne, M.L. and Swanson, J.D.,: Application of Probability

Reliability Methods to Tubular Design, Paper SPE 19556, Proc,
64th SPE Conf., Oct. 1989, 373-388.

Eaton, B. A.: Fracture Gradient Prediction and Its Application in

Oilfield Operations, Paper SPE 2162, Presented at SPE 43rd
Annual Fall Meeting, Houston, 29 Sept. 2 Oct. 1968.

5.

Nomenclature
COV = coefficient of variance
f(x) = probability density function (PDF)
i = index of zone i
n = total number of zones
N = total numbers of variables
PNK = probability of kick does not happen
PNF = probability of loss-return does not happen
RK = risk of taking a kick or under balance
RKi = risk of taking kick in zone i
RK-T = total risk of taking kick
RF = risk of loss-return or fracture formation
RFi = risk of loss-return in zone i
RF-T = total risk of loss-return
xi = individual variable value
= statistical mean
E = mean value of equivalent mud weigh
P = mean value of pore pressure
F = mean value of fracture gradient
EP =mean value of safety margin between pore pressure and
equivalent mud weight
FE =mean value of safety margin between equivalent mud
weight and fracture gradient
= statistical standard deviation
E = standard deviation of equivalent mud weigh
P = standard deviation of pore pressure
F = standard deviation of fracture gradient
EP = standard deviation of safety margin between pore
pressure and equivalent mud weight
FE = standard deviation of safety margin between equivalent
mud weight and fracture gradient
= standard normal probability distribution
= the series product operator.

Hottman, C.E. and Johnson, R. K: Estimation of Formation

Pressures from Lag-Derived Shale Properties, Paper SPE 1110,
JPT (June, 1965) 717-722.

6.

Jorden, J.R. and Shirley, O.J: Application of Drilling Performance

Data to Overpressure Detection, JPT (Nov. 1966) 1387-94.

7.

Martinez, R.D.; Schroeder, J.D. and King, G.A.: Formation

Pressure Prediction with Seismic Data from the Gulf of Mexico,
Paper SPE16621,

Presented at 1987 Offshore Technology

Conference, 27-30 April, Houston.

8.

Rocha, L.A. and Bourgoyne, A.T.: A New Simple Method to

Estimate Fracture Pressure Gradient, Paper SPE28710, Presented
at 1994 SPE Intl. Petroleum Conference and Exhibition of Mexico,
Veracruz, 10-13 Oct.

9.

Eaton, B.A.: How to Use Drilling Petrophysical Data in Prospect

Evaluation, World Oil (Sept. 1995) 69-74, and (Oct. 1995) 45-54.

10. Eaton, B.A.: Using Pre-Drill Seismic and LWD Data for Safe,
Efficient Drilling, World Oil (December 1998) 51-57.
11. Plumb, R.; Edwards, S.; Pidcock, G.; Lee, D.; Stacey, B.: The
Mechanical Earth Model Concept and Its Application to High-Risk
Well Construction Projects, Paper SPE 59128, Presentation at the
2000 SPE/IADC Drilling Conference, New Orleans, 2325
February.

Q. J. LIANG

SPE 77354

Equivalent Mud Density

8

10

11

12

13

14

15

16

17

18

0
2000
4000

Fracture Gradient

Mud Weight

TVD

6000
8000
Pore Pressure

10000
12000
14000
16000

Figure 1 - Deterministic approach of pore pressure and fracture gradient prediction

E
RK

F
RF

Figure 2 - Typical distribution of pore pressure (PP), fracture gradient (FG), and bottomhole EMW

19

SPE 77354

APPLICATION OF QUANTITATIVE RISK ANALYSIS TO PORE PRESSURE AND FRACTURE GRADIENT PREDICTION

Figure 3 Example of bulk density randomness

Figure 5 Example of d exponent randomness

Figure 6 Example of sonic velocity randomness

Figure 4 Example of sonic travel time randomness

Q. J. LIANG

SPE 77354

f(x)

COV2

COV1

E
margin

Figure 7 Example of leak-off pressure randomness

Pressure

Figure 8 Typical PDFs of PP and real-time EMW

Figure 9 - Example of actual pore pressure, calculated pore pressure and equivalent mud weight

SPE 77354

APPLICATION OF QUANTITATIVE RISK ANALYSIS TO PORE PRESSURE AND FRACTURE GRADIENT PREDICTION

Figure 10 - Applications of QRA in pore pressure and fracture gradient calculations can provide quantified
risk results for the chosen casing setting depth