Petrophysics Vol. 60 No. 2 April 2019

PETROPHYSICS
Best Papers of the 2018 SCA Interna onal Symposium
Vol. 60, No. 2 April 2019

PETROPHYSICS
April Vol. 60, No. 2
Contents
178 Le er to the Editor
179 From the Editor
180 Petrophysics Peer Reviewers for 2018
The Society of
Petrophysicists and TUTORIALS
Well Log Analysts 181 Organic Mudstone Petrophysics, Part 2: Workflow to Es mate Storage Capacity
Kent Newsham, Joe Comisky, and Roland Chemali
8866 Gulf Freeway, Suite 320
Houston, TX 77017, USA 208 Introduc on to Resis vity Principles for Forma on Evalua on: A Tutorial Primer
P: +1-713-947-8727 David Kennedy and Fredy Garcia
F: +1-713-947-7181
Email: Sharon@spwla.org
Membership@spwla.org ARTICLES BEST OF THE SOCIETY OF CORE ANALYSTS 2018 INTERNATIONAL
www.SPWLA.org SYMPOSIUM
229 Pore-Scale Insights on Trapped Oil During Waterflooding of Sandstone Rocks of
SPWLA Foundation. This fund Varying We ability States
supports scholarships in higher educa on Hélène Berthet, Mathilde Hebert, Sandra Barbouteau, Prisca Andriamananjaona,
and research in formation evaluation. and Richard Rivenq
Please send dona ons to: SPWLA, 8866
Gulf Freeway, Suite 320, Houston, TX 240 Uncertainty Quan fica on in Image Segmenta on for Image-Based Rock Physics in
77017, USA a Shaly Sandstone
James Howard, Sam Lin, and Shawn Zhang
ISSN 1529-9074
255 A Review of 60 Years of NMR We ability
PETROPHYSICS is published Andrea Valori and Benjamin Nicot
bimonthly by the Society of
Petrophysicists and Well Log Analysts 264 A New Waterflood Ini aliza on Protocol With We ability Altera on for Pore-Scale
(SPWLA). Subscription is included in Mul phase Flow Experiments
annual dues for members. Library Qingyang Lin, Branko Bijeljic, Samuel C. Krevor, Mar n J. Blunt, Maja Rücker,
Steffen Berg, Ab. Coorn, Hilbert van der Linde, Apostolos. Georgiadis, and Ove B. Wilson
Subscription cost is $250 (6-issues,
Feb, April, June, Aug, Oct, Dec) Send 273 In-Situ Satura on Monitoring (ISSM)— Recommenda ons for Improved Processing
Subscription and member address Jules Reed and Arjen Cense
changes to the SPWLA business office.
Back issues if available are $35 each 283 Review of the Intercept Method for Rela ve Permeability Correc on a Variety of
to members and $45 each to library Case Study Data
subscribers. Jules Reed and Jos Maas
ADVERTISING RATES: (based on
one-time, full color) Full Page: $1500,
297 Monitoring Core Measurements With High-Resolu on Temperature Arrays
Half Page: $1200, Quarter Page: James J. Howard and Keith C. Hester
$1000. Additional fee for printing on
covers. Media kits are available by ARTICLES REGULAR SUBMISSIONS
contacting the SPWLA business office
or by visiting the SPWLA website 306 How the Invasion Zone Can Contribute to the Es ma on of Petrophysical Proper es
www.spwla.org. Responsibility: The From Log Inversion at Well Scale?
statements and opinions expressed Thibaud Vandamme, Emmanuel Caroli, and Serge Gra on
in PETROPHYSICS are those of the
authors and should not be construed 326 Loading Effects on Gas Rela ve Permeability of a Low-Permeability Sandstone
as an official action or opinion of the F. Agos ni, P. Egermann, L. Jeannin, E. Por er, F. Skoczylas, and Y. Wang
Society of Petrophysicists and Well
Log Analysts, Inc. © Copyright 2017 335 Borehole Acous c Imaging Using 3D STC and Ray Tracing to Determine Far-Field
by the Society of Petrophysicists and Reflector Dip and Azimuth
Well Log Analysts, Inc. Nicholas Benne , Adam Donald, Sherif Ghadiry, Mohamed Nassar, Rajeev Kumar,
and Reetam Biswas
About the Cover Full-Color versions of all technical ar cles are available in the digital edi on of Petrophysics,
Capillary desatura on curves obtained which is free to SPWLA members and can be found at Recent_Petrophysics_Journals.aspx.
on water-wet miniplugs of sandstone
rocks (Fontainebleau, Bentheimer) by As of Vol. 46 (1) 2005,
micro-CT imaging successfully compared The Society of Petrophysicists and Well Log Analysts
with tradi onal core-scale experiments. Petrophysics [ISSN: 1529-9074]
is dedicated to the advancement of the science of forma on
On the right is an imaged volume of the evalua on through well logging and other forma on evalua on is indexed and abstracted in
Fontainebleau sample at Swi = 18%. See techniques. SPWLA is dedicated to the applica on of these Thomson Reuters:
the ar cle by Berthet et al. for details.
techniques, to the explora on and exploita on of gas, oil and Scien fic Cita on Index Expanded
other minerals. PETROPHYSICS publishes original contribu ons
Journal Cita on Report–Science
on theore cal and applied aspects of forma on evalua on;
par cularly well logging and petrophysics. Current Contents–Physical, Chemical,
PRINTED IN USA ON ACID FREE PAPER and Earth Sciences
April 2019 PETROPHYSICS 176

PETROPHYSICS The Society of Petrophysicists and Well Log Analysts
Editor Board of Directors 2018–2019
Carlos Torres-Verdin President VP Finance
University of Texas at Aus n Zhipeng “Z” Liu Jennifer Market
(+1) 512-471-4216 Kinder Morgan Lloyd’s Register
cverdin@aus n.utexas.edu Houston, TX, USA Houston, TX, USA
(+1) 713-369-8059 (+1) 713-302-8325
Associate Editors President@spwla.org VP-Finance@spwla.org
President-Elect VP Publica ons

Jesus Salazar Carlos Torres-Verdin
Acous cs and Rock Physics ConocoPhillips University of Texas at Aus n
Richard Coates, Halliburton Houston, TX, USA Aus n, TX, USA
(+1) 281-293-5237 (+1) 512-471-4216
President-Elect@spwla.org VP-Publica ons@spwla.org
VP Technology VP IT
Core Analysis and Laboratory Petrophyiscs James Hemingway Mehrnoosh Saneifar
Gary Beck, Cardinal Geosciences Consultant BHP Petroleum
Manitou Springs, CO, USA Houston, TX, USA
(+1) 281-433-5170 (+1) 832-600-4046
VP-Technology@spwla.org VP-InfoTech@spwla.org
Electromagne cs VP Educa on
Fei Le, Baker Hughes Katerina Yared
SM Energy
Highlands Ranch, CO, USA
(+1) 720-431-7482
VP-Educa on@spwla.org
Forma on Tes ng
Mayank Malik, Chevron
REGIONAL DIRECTORS
N. America 1 Middle East/Africa/India
Adam Haecker Shouxiang (Mark) Ma
Integrated Forma on Evalua on Con nental Saudi Aramco
Chengbing Liu, Saudi Aramco Oklahoma City, OK, USA Dhahran, Saudi Arabia
Wilfrido Solano, Chevron (+1) 979-587-1061 (+966) 3874 6931
Director-NA1@spwla.org Director-MEA@spwla.org
N. America 2 Asia/Australia
Doug Pa erson Rick Aldred
Integrated Forma on Evalua on and Case Studies Consultant Petrophysicist
Baker Hughes
Hesham El-Sobky, ConocoPhillips Houston, TX, USA Queensland, Australia
(+1) 713-879-4056 (+610) 408-453-351
Director-NA2@spwla.org Director-Asis-Aus@spwla.org
Magne c Resonance La n America Execu ve Director

Nadege Bize-Forest Sharon Johnson
Lalitha Venkataramaman, Schlumberger
Schlumberger SPWLA
Wim Looyes jn, Shell Houston, TX 77017
Rio de Janeiro, Brazil
(+552) 197 45 45 772 (+1) 713-947-8727
Director-LA@spwla.org sharon@spwla.org
Nuclear Europe
Pingjun Guo, ExxonMobil Michael Webster
Produc on Petrophysics Ltd
Aberdeen, Scotland, UK
(+440) 7568-476931
Director-Europe@spwla.org
Petroleum Geochemistry
Drew Pomerantz, Schlumberger
Well and Reservoir Surveillance

Gerardo Cedillo, BP
Managing Editor Publica on Manager

Stephen Prensky Anna Tarlton
(+1) 301-593-4966 InkSpot Prin ng
Data-Driven Analy cs sprensky@gmail.com 2301 S. Shaver
Chicheng Xu, Aramco Services Pasadena, TX 77502, USA
(+1) 713-472-1100
orders@inkspotprin ng.com
177 PETROPHYSICS April 2019

TO THE EDITOR
To the Editor of Petrophysics,
I would like to make a comment in regards to the ¿ne paper by Frost and Quinn, “Data Preconditioning for Predictive
and Interpretive Algorithms: Importance in Data-Driven Analytics and Methods for Application,” Petrophysics, 2018, 59(6),
873–890.
One of the most disastrously wrong petrophysical interpretations (not by Frost and Quinn) I confronted in my career was
partially caused by the inappropriate application of a “Reconstruction and Prediction Algorithm” to a density log in a case
of critically stressed geomechanics conditions where the borehole in shales failed and the sands suffered drilling-induced
fractures. There were also natural subvertical fractures and a fresher Rw condition which caused an “elite” evaluation team
to falsely turn a modest volume of 20 kppm NaCl water into millions of barrels of oil. Lessons learned that may be added to
Frost and Quinn’s paper include:
x Beware geomechanical conditions where both borehole failure in shales and drilling-induced fracturing in sands
can occur. Some refer to this as “critically stressed borehole trajectories.” Where more than one systematic source
of borehole degradation is present, each systematic condition may require its own family of prediction algorithms.
x Although the reconstruction and prediction algorithm for the density log worked nicely for the elite team for its
intended design of breakout in shale, how to appropriately apply the results of such algorithms throughout the entire
wellbore must be reviewed objectively and competently by humans.
Michael J. Manning

FROM THE EDITOR
This new issue of Petrophysics is devoted to highlighting the best technical contributions presented at the 2018
International Symposium of the Society of Core Analysts (SCA). The SCA vetted and selected the ¿ve papers for this issue,
which were subsequently subjected to the high-standards of the Petrophysics review process to comprise the special section.
Additionally, we are including the second of a three-part tutorial on Petrophysics of Unconventional Resources, the
¿rst part of which was extremely well received by many of our readers. A second tutorial, written by Kennedy and Garcia,
provides valuable insights and physical intuition about Archie’s equation for those who are interested in learning the inner
workings of such an important and trascendental equation.
Finally, there are three additional papers that were received through the regular Petrophysics submission and review
process, all comprising very interesting and practical material to read and digest.
We look forward to receiving your comments and contributions while we begin to assemble the June 2019 issue of
Petrophysics. Thanks in advance and keep reading!
Sincerely,
Carlos Torres-Verdín, Ph.D., Professor
Brian James Jennings Memorial Endowed Chair in Petroleum and Geosystems Engineering
Department of Petroleum and Geosystems Engineering
University of Texas at Austin
cverdin@mail.utexas.edu

Petrophysics Peer Reviewers for 2018
Petrophysics and SPWLA extend our sincere appreciation Forsyth, Dave Proett, Mark
and a thank you to the individuals who provided at least one Gao, Guizhing Pruno, Stefano
review of a submitted technical paper during 2018. Many Gaur, Stuti Quinn, Terrance H.
of our reviewers contributed more than one review. It is Gerard, Matt G. Radtke, R.J.
through the dedication of these individuals to the journal and Gong, Bo Raizada, Shashank
to the profession that enables Petrophysics to maintain high Gu, Ming Ramaswami, Shyam
standards. The time commitment these volunteers put in to Gulick, Byron Ramos, Juan
read and comment on the submitted manuscripts not only Guo, Weijun Rasmus, John
improves the quality and lasting value of the manuscripts Hamza, Farrukh Richardson, Will
themselves, but it also increases the technical rigor of future Han, Xiaogang Rushing, Jay Alan
research in the ¿eld. Holmes, Kory Salazar, Jesus
Horkowitz, John Sanchez-Ramirez, Jorge
Statistics for 2018: 106 papers accepted for review: 49 Howard, James J. Saneifar, Mehrnoosh
accepted for publication; 26 rejected; 17 returned to Hu, Guoyu (David) Se, Yegor
authors for revision; 10 under review; 4 withdrawn. We Hu, Qinhong Singer, Philip
also published three special issues that consisted largely of Huang, Shan Skelt, Chris
invited papers. Ijasan, Olabode Sullivan, Michael J.
Jiang, Han Sun, Boquin
Editor Beck, Gary Jones, Paul Theologou, Paul
Carlos Torres-Verdin Belanger, Dave Katahara, Keith Thern, Holger
Bennett, Nicholas Kausik, Ravinath Tracadas, Philip
Associate Editors Bilodeau, Bruce Kersey, David Valenza, John J
Vivek Anand Blyth, Matthew Krumm, Robert Lee Wang, Haijing
Gary Beck Bolt, Harold Lang, Xiaozheng Wang, Hanming
Gerardo Cedillo Bowers, Mark Li, Hao Wang, Jiaxin
Richard Coates Cade, Chris Li, Weichang Wang, Ruijia
Hesham El-Sobky Cao Minh, Chanh Liang, Lin Wang, Weihua
Pingjun Guo Chen, Dingding Liang, Yu Webster, Michael
Fei Le Chen, Ji Liu, Chao Willis, Mark
Chengbing Liu Chen, Jiefu Liu, Chao Wright, Milly
Wim Looyestijn Chen, Songua Liu, Shujie Wu, Xianyun
Mayank Malik Cheng, Min Loermans, Ton Xie, Harry
Drew Pomerantz Chi, Lu Looyestijn, Wim Xie, Hui
Wilfrido Solano Clerke, Edward Louis, Laurent Xu, Chicheng
Christian Stoller Coates, Richard Luo, Maolin Xu, Guangping
Lalitha Venkataramanan Coles, Darrell Ma, Shouxiang (Mark) Xu, Haomin
Hanming Wang Connolly, Paul Maalouf, Elsa Xu, Libai
Jianghui Wu Craddock, Paul Malik, Mayank Xu, Rui
Chicheng Xu Cychosz, Katie Maloney, Dan Xu, Yankai
Adrian Zett Deng, Tianqi Merkel, Richard Yang, Qinshan
Adrian Zett Dong, Chengli Michael, Eric Zett, Adrian
Donovan, Glenn Misra, Siddharth Zhang, Hao
Technical Reviewers Dria, Dennis Mohapatra, Avinash Zhang, Xiaowen
Akkut, Ridvan Dumont, Hadrien Mullins, Oliver Zhang, Yinxi
Aldred, Rick Elkington, Peter Myers, Gary Zhao, Jiajun
Arns, Christoph Hermann Endo, Takeshi Needham, Riley Zhou, Tong
Ashby, Michael Falzone, Sam Ostermier, Richard
Badruzzaman, Ahmed Ferreira, Elton Pan, Wen
Basu, Saptaswa Fitz, Dale Pemper, Richard
Bateman, Richard Flaum, Mark Pineda, Wilson

PETROPHYSICS, VOL. 60, NO. 2 (APRIL 2019); PAGES 181–207; 25 FIGURES; 3 TABLES. DOI: 10.30632/PJV60N2-2019t1
TUTORIAL
Organic Mudstone Petrophysics, Part 2: WorkÀow to Estimate Storage Capacity
Kent Newsham1, Joe Comisky2, and Roland Chemali1
This is the second of a three-part tutorial describing a workflow for evaluating unconventional resources including organic mudstones and tight
siltstones. Part 1 reviewed the unique challenges and provided an overview of the proposed workflow (Newsham et al., 2019). Part 2 describes
in detail the many components of the workflow and how they come together to determine the storage capacity of the reservoir. Part 3 links the
petrophysical results to the production potential in terms of fractional flow and water cut and will present alternate cross-checks of the storage
properties to validate the results.
INTRODUCTION formation volume factor (Bo), oil viscosity (ȝo), bubblepoint

pressure (Pb) and gas density (ıg). We’ll review the impact
As stated in Part 1, one of the most important functions pressure and temperature have on the Àuid property response
that the petrophysicist provides is the estimation of accurate of the density and acoustic logs. All of these elements are
storage properties. However, when the authors survey the combined within the equations of state (EOS), of which
range of workÀows used to estimate the storage capacity many are available in the public domain (Standing, 1952;
of these complex systems, we ¿nd a wide range of options. Lasater, 1958; Vasquez and Beggs, 1980; Al Marhoun, 1988;
Solutions can vary from simple deterministic to more Petrosky and Farshad, 1998). However, Àuid properties
complex probabilistic approaches. Whatever the method, the tend to vary widely and the authors recommend collecting
objective should be the same: to provide consistent, portable Àuid samples and performing pressure-volume-temperature
hence reliable estimation of hydrocarbon storage capacity,
(PVT) testing on their Àuids samples to create speci¿cs
also known as “Petrophysics CPR.” As mentioned in Part
project EOSs. Since most of the author’s reservoirs are
1, estimation of hydrocarbon storage is more than just the
oil systems, and for brevity, discussion of Àuid properties
calculation of porosity and water saturation. In this tutorial,
we will describe a workÀow that has been successfully used and storage computations will be con¿ned to black and
to evaluate thousands of wells in the Permian Basin with volatile oils. Discussion of gas-condensate, wet-gas and
great consistency. The authors have nearly 100 wells with dry-gas systems requires much more detail on absorbed gas,
core data to calibrate the workÀow. We will show examples adsorbed gas, lost-gas estimates and Langmuir isotherms not
of the workÀow’s portability by highlighting examples covered in this tutorial. Mavor provides an overview of gas-
from the Midland Basin, the Texas Delaware Basin and in-place (GIP) estimation (Mavor, 1994, 2014; Mavor and
the New Mexico Delaware Basin. We will show how every Nelson, 1997).
property measured in core matches to log-based pro¿les The workÀow requires a minimum suite of logs that
using a combination of deterministic and the constrained includes gamma ray (GR), deep and shallow resistivity,
simultaneous solution methods. The workÀow also is found neutron, bulk density and photoelectric responses. A
to be reliable in other basins throughout the world, however, spectral gamma-ray log is preferable but not required. If
the examples will be con¿ned to the Permian Basins. any logs are of pre-1980s vintage, then an acoustic log is
In Part 1, Fig. 4, the authors showed the Organic a necessary addition to the suite. If any elastic-properties
Mudstone WorkÀow. In Part 2, we’ll provide details of estimates are intended, then a dipole acoustic log is
each workÀow component beginning with a description of required. However, in most of the Permian formations we
the “statics” estimation. The statics refer to those properties get a reliable shear slowness from a long-spaced monopole
that are intrinsic to the rock, such as mineral composition, acoustic log. Additional advanced logs for consideration
especially bulk clay and pyrite, organic content, formation include the magnetic resonance, dielectric, and elemental
salinity and electric properties. Accurate estimation of these (compositional) tools. In Part 3 of the tutorial, the authors
properties leads to consistent calculations of total porosity (‫׋‬t) will discuss these logs in more detail regarding their use to
and total water saturation (Swt). The “dynamics” refer to those validate the workÀow storage estimates.
properties that change with depth, such as gas/oil ratio (GOR), One of the key concepts in the workÀow is to constrain
gas gravity (Ȗg) temperature and pressure. These include the the most extreme components through rock calibration before
1
Occidental Petroleum Corp., 31827 Edgewater Dr. Magnolia, TX 77354; Kent_newsham@oxy.com
2
Devon Energy Corporation, 333 W Sheridan Ave, Oklahoma City, OK 73102; joe.comisky@dvn.com
3
Occidental Petroleum Corp, 5 Greenway Plaza, Suite 110, Houston, TX 77046; Roland_Chemali@oxy.com

invoking the use of an inversion method. Without some form THE STATICS: ESTIMATION OF BULK CLAY
of constraint, inversions can provide nonunique solutions,
especially if the system of equations is underdetermined, The volume of clay is calculated deterministically
i.e., the number of petrophysical unknowns is larger than the and calibrated against cuttings and/or core composition
number of logs and core data available. System components measured from various sources, such as X-ray diffraction
that have extreme endmember response can also perturb the (XRD), Fourier-transform infrared spectroscopy (FTIR)
results. In organic mudstone systems, total organic content or X-ray Àuorescence (XRF). However, XRF will require
(TOC) and pyrite are common. Both have rather extreme some form of stoichiometry to convert elemental yields to
density and neutron responses. TOC is treated as part of the mineral weights. Clay volumes are estimated from a bulk
framework rock but has Àuid like responses. Pyrite is a very rock and bulk clay perspective. Estimating clays species
dense mineral with low neutron porosity. Figure 1 presents is not attempted. The clay mixture measured from cuttings
a density-neutron crossplot showing the relative positions and/or core is modeled using nuclear modeling applications
of various organic-mineral responses and the position of available to the industry to simulate the neutron, density,
pyrite. The transparent oval represents the typical range of photoelectric, and sigma responses to various mineral
response for an organic-mudstone reservoir, clay being on mixtures through a range of pressure, temperature and Àuid
the far right boundary of the oval. The numerous response conditions. For this reason, the primary logs for estimating
lines are quartz, calcite, dolomite and the pure clay-mineral the volume of clay are density and neutron because they can
responses, from left to right. Clearly, a small amount of be directly calibrated to the rock. The GR log is a secondary
TOC or pyrite can go a long way in inÀuencing the overall clay indicator since there are no published GR response
log response of a complex mixture of minerals. For this values for various minerals. Hence the GR is subsequently
reason, we choose to deterministically estimate TOC, clay calibrated to the density-neutron clay index. The GR is less
and pyrite external to the simultaneous inversion. Critically, inÀuenced by “standoff” conditions, i.e., borehole rugosity,
we calibrate these properties to cuttings and/or core. These and serves as a backup index. The minimum scalar value of
properties are input into the inversion as logs, but with a the two indexes are used as the ¿nal estimate of clay volume.
high level of con¿dence, such that their scalar properties
are maintained in the solution. Hence, the output of the Generate Lithology-Response Lines Using Nuclear
inversion is “constrained” by making the values of TOC, Models
clay, and pyrite entering the inversion the same as the values The calculation of the clay volume requires the a priori
coming out. This bounds the system of equations to solve knowledge of the response of the logging sensors to the
for only the remaining de¿ned properties, such as quartz, clay minerals. This is achieved through the use of forward-
feldspar, carbonate, total porosity and grain density. In this modeling software. The authors know of three programs
way, the authors will demonstrate the consistency of the in existence. These include Nuclear Modeling (NMOD),
solution such that the use of compositional tools are limited developed by Ralph Wiley and Jay Patchett (Wiley and
in current appraisal programs. Pachett, 1990) while at Amoco (Now BP); Schlumberger’s
Nuclear Parameter (SnuPar) modeling program (Edmundson
and Raymer, 1979); and the University of Texas at Austin
Petrophysical and Well-Log Simulator (UTAPWeLS)
(Mendoza et al., 2010a, 2010b; Ijasan et al., 2013). Forward-
modeling programs are used for predicting the response of
bulk density, photoelectric, neutron porosity and sigma cross
section for a range of minerals and trace elements. It is used in
this workÀow to generate lithology-response lines of various
mineral mixtures, including clays. The modeling software
can account for temperature, pressure, water saturation and
salinity. Figure 1 shows the neutron- and density-response
lines for a range of clay minerals generated using one of
these nuclear-modeling applications.
Fig. 1—A density-neutron crossplot with mineral response lines for

various clays and minerals.

Use Reservoir-Speci¿c Nuclear-Response Lines for Clay intrinsic property for a particular mineral and can be found
Estimation in the standard log chartbook reference tables. Be aware that
The objective of the nuclear models is to generate many laboratories may provide a volume percent conversion
discrete density- and neutron-response lines for a mixture of but may lack either the TOC and/or porosity required for a
clays and matrix minerals representative of a given reservoir. full conversion from weight to wet volume percent. Table 1
We use XRD, FTIR or other sources of rock composition to is an example of the impact porosity causes on the volume
calibrate to cuttings and/or core. When working with these conversion, all other input values being the same. In this
composition measurements, the unit of measure is weight example, the same sample is converted using increasing
percent. Most nuclear models require mineral input in values of porosity, signi¿cantly altering the mineral volume
volume percent. Conversion of the minerals from weight to distribution.
volume percent is given by Eq. 1. Figure 2 is a graphic workÀow showing how we use the
various compositional components in modeling the matrix
WetVol% = (DryWt% × (1í ࢥt) × (Bulk Rock ȡgd)/(Mineral ȡgd), (1) and clay lines. To construct the clay-response lines, the clay
fraction from the compositional data needs to be standardized
into unity space and converted to volume percent. In Fig.
where the DryWt% is the measured value from XRD, FTIR 2, an average mix of chlorite, kaolinite, illite, and smectite
or equivalent, ȡgd is the grain density of either the bulk rock are modeled through a range of porosity. This results in the
or mineral and ‫׋‬t is the measured total porosity. Routine or clay-mix line of the neutron-density crossplots in Figs. 2 and
crushed rock analysis provides a source of the bulk-rock 3. The same process is applied to the matrix components
grain density and porosity. The mineral grain density is an resulting in the matrix-mix lines in Fig. 2 and 3.
Table 1— Conversion to Bulk Volume% Including Total Porosity

Fig. 2—Shows a graphic workÀow describing the process of fractionating the compositional data into constituent matrix and clay components as input
into a nuclear model to create density and neutron matrix- and clay-response lines. These lines provide a bounded solution for estimating clay volume
using Eq. 2, also represented graphically from the ternary solution of Figs. 2 and 3. The resultant clay volume is cross-checked against core data.
Table 1 shows how variation in porosity alters the

weight-to-volume conversion distribution. The same sample
is represented but converted from weight to volume using
increasing porosity as input into Eq. 1
A mixture of calcite, dolomite, sandstone, orthoclase,
albite and pyrite yields the matrix-mix line. These lines
provide a bounded range of log response for this reservoir.
A ternary solution provides an estimation of clay volume by
Eq. 2. From Figs. 2 and 3, the ternary endpoints are easily
de¿ned, including the zero-porosity matrix point, the Àuid
point and the projection of the wet-clay trend from the
zero-porosity matrix point to the clay matrix-response line
such that the log data are bounded within the triangle. The
clean-matrix line is de¿ned as the projection from the zero-
porosity matrix point to the Àuid point. The diagonal red
Fig. 3—A density-neutron crossplot showing the ternary solution for
clay volume. The matrix- and clay-mixture lines were estimated from
a nuclear model using XRD data. The green points are the log data.
Note how the matrix- and clay-mixture lines bound the log data. The red
diagonals are constant clay-volume lines.

lines of Fig. 3 represent the lines of constant clay volume clay index is the primary clay index because we are able to
as de¿ned within the ternary solution. Use of the model calibrate to cuttings and/or core. As the GR responds to the
response lines provides a consistent method for de¿ning the natural radiation of the earth, there are no published mineral
ternary solution used in estimating clay volume from density response endpoints. Hence, the gamma-ray log is considered a
and neutron (D-N) logs. secondary method and is matched to the D-N index by Eq. 3.
(2)
(3)
where DenCl1 and NeuCl1 are the D-N values de¿ning the
zero-porosity matrix point. DenCl2 and NeuCl2 are the D-N where GR is the measured log value, GrClean is the
log values de¿ning the trajectory of the Àuid point. Neutron minimum GR, and GrClay is the maximum GR that provides
is the neutron log response and density is the bulk-density a match. The ¿nal clay volume is the minimum of the two
log response. NeuClay is the neutron value for wet-clay point methods. Figure 4 is a clay-volume plot showing the GR and
and DenClay is the bulk density value for wet-clay point. D-N pro¿les for a well in the Wolfcamp. The tracks from left
to right include: formation interval, GR with minimum and
Estimation of Minimum Clay Using Combined Indexes maximum vertical lines for estimating the GR clay index,
The ¿nal clay-volume estimation comes from a the D-N pro¿le, the D-N clay index with XRF and XRD
combination of the D-N and GR clay indexes. The D-N clay volume points measured on cuttings samples (green
Fig. 4—A clay-volume plot showing the GR and D-N pro¿les for a well in the Wolfcamp. The tracks from left to right: Track 1, formation interval; Track
2, GR with minimum and maximum lines indicated; Track 3, neutron and density logs; Track 4, clay volume estimated from the D-N ternary solution
(Fig. 2); Track 5, overlay of the clay volumes estimated from GR and D-N; Track 6, clay volume ¿nal estimate, taking the lesser value from Tracks 4
and 5, compared with cuttings FTIR and XRF results.

and blue, respectively), the overlay of the GR and D-N clay (5)
indexes and the minimum clay index with the XRF and
XRD clay points. The spectral GR is preferred but is often
not available. Note that the GR endpoints are highly varied Ro = Rwno/‫׋‬t2, (6)
to match the D-N clay index and the cuttings results. This
emphasizes the reasoning for the use of the D-N clay index as
the primary, especially when cuttings and/or core control in (7)
not available. As long as the clay mineralogy doesn’t greatly
vary, the use of the response lines as a bounding solution
for estimating clay from the D-N will provide consistent where Rt is the formation resistivity in ȍ·m, Ro is the
estimates of clay volume. formation resistivity when 100% water saturated, Rwno is the
water resistivity in a nonorganic interval, ‫׋‬t is total porosity,
LOM is the level of maturity and Vro is the observed vitrine
THE STATICS: ESTIMATION OF TOTAL ORGANIC reÀectance, and TS is a linear scaler. LOM is not a parameter
CONTENT (TOC) AND PYRITE in common usage. Equation 7 is provided to convert Vro to
LOM. Vro estimates are made by observing the frequency
There are several methods for estimating total organic of occurrence of vitrinite in thin sections. However,
carbon (TOC). This workÀow uses a combination of three programmed pyrolysis is more commonly used to estimate
differing methods including Faust (1953), Schmoker and both Vre, the vitrinite reÀectance equivalent and TOC.
Hester (1983), and Modi¿ed Passey (Passey et al., 1990).
Again, we use multiple indexes, as no single index provides Faust Method
a perfect solution. The authors have found that the average The Faust Method is essentially an acoustic overlay
of the TOC indexes provide the best correlation to cuttings technique. It overlays the measured sonic slowness with
and /or core TOC results. Furthermore, all indexes are a synthetic sonic slowness based on resistivity. TOC is
parameterized to provide the ability to tune each index to a equivalent to the difference in the measured and synthetic
measured reference. Again, calibration to key wells where slowness using Eq. 8 (Faust, 1953).
TOC has been measured from cuttings, rotary sidewall cores
(RSWC) or whole core is essential.
(8)
Density Regression (Schmoker Method)
The density regression method is a variation of the where Dt is the compressional sonic log, Fd is a TOC divisor
Schmoker and Hester (1983) method where the key parameters defaulted to a value of 10 and DtFaust, the Faust coef¿cient
of the equation are adjustable by the petrophysicist. The only given by Eq. 9.
log required is density. The original formulation by Schmoker
and Hester is modi¿ed to allow for tuning by Eq. 4.
(9)
(4)
where Z is the true vertical depth and AFaust is a variable but
defaults to 0.17.
where Acoef is a constant with the default value of 154.497
and B is a constant with a default value of 57.261. In general, the average of the Modi¿ed Passey, Faust
and Schmoker methods yields the best results across a range
Modi¿ed Passey Method of cuttings, rotary sidewall cores and whole-core plug data.
Modi¿ed from the original method published by Figure 5 displays the comparison to cuttings-based TOC
Passey et al., (1990), the “Modi¿ed Passey” method uses (discrete red points): Track 1, clay volume; Track 2, zone
a Ro overlay technique. One of the key modi¿cations is that intervals; Track 3, Passey Ro overlay; Track 4, Modi¿ed
any porosity curve can serve as input to estimate Ro, the Passey; Track 5, Schmoker; Track 6, Faust; Track 7, index
resistivity of 100% water-saturated rock using the Archie comparison; and Track 8, average TOC from all three
equation. The Ro curve is overlain to the recorded deep indexes. When no core or cuttings data are available, it is
resistivity (Rt) curve in a nonorganic reservoir interval. The good practice to compare the different approaches and adjust
resistivity overlay is equivalent to the ǻlogR portion (LogRt the parameters such that each index agrees with the other.
– LogRo) of the Passey method and is described in Eqs. 5, 6, Note that the application allows for disengaging various
and 7. indexes by interval to improve the average TOC result.

Fig. 5—Comparison between the various TOC methods and core analysis. Consistency is achieved by adjusting the various parameters governing
these equations and/or not using speci¿c indexes within speci¿c intervals. Track 1, clay volume;Track 2, zone intervals; Track 3, Passey Ro overlay;
Track 4, Modi¿ed Passey; Track 5, Schmoker; Track 6, Faust; Track 7, index comparison; and Track 8, average TOC from all three indexes.
Estimation of Pyrite THE STATICS: TOTAL POROSITY AND TOTAL

Organic content and pyrite both form in reducing WATER SATURATION
environments. So it is reasonable to assume that a correlation
relating TOC to pyrite exists. Figure 6 shows a simple Equation 10 is the Archie equation (Archie, 1942) that
linear relationship between TOC and pyrite for a Wolfcamp states the formation resistivity Rt is directly proportional
interval. The actual correlations may vary from formation to the resistivity of the connate water Rw, and inversely
to formation, but once a correlation is established, it can be proportional to the porosity ‫׋‬t raised to a power m, and to the
used in well without core control. As will be discussed in a water saturation Swt raised to a power n:
later section on the constrained inversion method, the pyrite
correlation is used as a ¿ne-tuning parameter to balance the
solution of other minerals other than bulk clay. (10)
where Rw is the resistivity of the formation water, a is

generally set to a value of 1, n is the water saturation
exponent, m is the porosity exponent, ‫׋‬t is the total porosity,
and Swt is the total water saturation.
The porosity exponent, m, reÀects the pore tortuosity
(Archie, 1942; Wyllie and Gregory, 1953). The saturation
exponent, n, reÀects the distribution of the Àuids within
the pores and the wettability of the matrix (Keller, 1953;
Anderson, 1986).
In conventional reservoirs, numerous variations of
Archie’s equations were developed during the last several
decades to account for the conductivity of the clays present
Fig. 6—Correlation between TOC and pyrite content for an organic
mudstone.
in the formation. (Simandoux, 1963; Poupon and Leveaux,

1971; Waxman and Smits, 1968; Clavier et al., 1984). where ‫׋‬clay is the total porosity of the clay, ȡdryclay is the bulk
They have been successfully applied to a large array of density of the zero-porosity clay response line, ȡwetclay is the
conventional oil and gas bearing formations worldwide. bulk density of the wet-clay response, ȡÀ is the bulk density
In unconventional reservoirs, and particularly in the of the Àuid, Vclay is the volume of clay, ‫׋‬e is effective porosity
emerging organic-mudstone plays, the workÀow of this and ‫׋‬t is total porosity.
tutorial uses the total porosity and the total water saturation A large uncertainty in assessing the value of the wet-clay
concepts. This approach is better suited than the effective- porosity can lead to signi¿cant error in the effective porosity.
porosity (‫׋‬e) model for evaluating unconventional organic Speci¿cally, Fig. 7a shows a D-N crossplot illustrating two
mudstone for the following reason. An assumption of different options for the projection of the wet-clay point. In
‫׋‬e in conventional reservoirs is that ‫׋‬e is connected and this example, the dry-clay density is 2.825 g/cm3 and the
producible. The estimation of ‫׋‬e is related to clay volume. two wet-clay densities are 2.644 and 2.53 g/cm3, repectively.
In organic mudstone systems, this is a Àawed assumption, If the Àuid density is 1.0 g/cm3, the resulting porosity of
as the organic content has a large impact on the presence, the clay is 0.099 and 0.162, respectively. This can lead to
volume, and connectivity of porosity, as does clay content, a signi¿cant range in the calculation of effective porosity
yet organics are not accounted for in the “classic” clay-based depending upon the amount of clay in the system. The
effective porosity and effective water saturation models. uncertainty in estimating effective porosity and the lack of
Furthermore, most laboratories provide core measurements an effective-porosity calibration reference from core leads
of total porosity and total water saturation. The Gas the authors to use a total-porosity and dry-clay model, as
Research Institute (GRI) and benchtop magnetic resonance described in Eq. 10.
(MR) methods measure only total water saturation and total
porosity. Hence, calibration to core data is further reason for
using a total porosity and total water saturation method for
organic mudstones.
The total-porosity and dry-clay model is also preferred
over the effective-porosity and wet-clay models because
estimation of the clay-bound water can be problematic. For
the Permian Basin, clay species are dominated by illite. There
are only minor concentrations of kaolinite and chlorite, and
even less smectite. The illite is mostly fully illitized and well
ordered, with a Reichweite classi¿cation of R3 (Jagodzinski,
1949; Watanabe, 1988). The very low (<15%) ratio of
smectite to illite is a measure of this classi¿cation. R3 illite has
little to no interlayer water remaining as it has been expelled
via the illitization process. Only nonconductive hydroxyls (a)
(OH) remain in the interlayers (Moore and Reynolds, 1989).
However, the clay-response line inferred from the neutron-
density crossplot clearly indicates associated porosity. Clay
minerals have tremendous surface area. In a fully illitized,
water-wet system much of the associated water is surface
water, i.e., capillary-bound water.
The uncertainty inherent to the wet-clay model is
illustrated as follows. The D-N readings at the wet-clay point
are used for clay-volume calculations and for estimating
effective porosity. Equations 11 and 12 compute the effective
porosity using a wet-clay method that includes bound water.
(11) (b)
Fig. 7—(a) shows a D-N crossplot with multiple options for the wet-clay-
point determination. (b) shows the same plot but with the trajectory lines
associated with estimating total or effective porosity as a function of dry
(12) clay or wet clay, respectively.

Dry-Clay Model Option preserved core plug, in the as-received state, without the need
Combining the Archie total-porosity (Archie-‫׋‬t) model to extract and separate the Àuids. This new technology uses
with the use of dry clay as an input eliminates the potential complex resistivity measurements at varying frequencies
error associated with wet clay. From the previous discussion (Hanna et al., 2014). Laboratory technicians place the core
on nuclear modeling, the dry-clay point is the zero porosity plug between electrodes in a resistivity cell and measure the
on the clay-mix D-N response line. Since the D-N values complex impedance at different frequencies, then compute
of the dry clay are ¿xed by the clay-response model, then it Rw using a specialized algorithm. No Àuids are introduced or
becomes much easier to use a dry-clay model and this dry- removed from the core sample during or prior to conducting
clay point to estimate total porosity. Figure 7b shows the the procedure.
trajectory lines associated in estimating total and effective
porosity from the dry- and wet-clay points, respectively. In The Classic Method of Electrical Property Measurement
using a dry-clay model, only total porosity is relevant. The The Archie electrical properties of m and n can be
combination of a dry-clay and Archie-‫׋‬t water saturation measured on core samples by the formation factor (FF)
model eliminates the need for clay porosity (‫׋‬clay) and clay and resistivity index (RI) methods, respectively. Equation
resistivity (Rclay) input. Furthermore, the bulk volume product 13 de¿nes FF and is the proportionality factor between the
is equivalent whether one chooses an effective versus total formation’s resistivity at 100% water saturation (Ro) and the
porosity and water saturation model. The authors ¿nd that connate water resistivity (Rw). RI is the ratio between Rt and
the total porosity, total water saturation method yields more Ro given by Equation14.
consistent and repeatable results.
(13)
THE STATICS: ESTIMATION OF ARCHIE
PARAMETERS
(14)
Two approaches for determining the parameter inputs
into Eq. 10 are by (a) direct measurement of the Àuid and
rock properties, or (b) indirectly by use of a Pickett plot Saturation of the clean and dry core plug occurs with a
(Pickett, 1966). We are most concerned about the Archie laboratory brine of known salinity concentration. The Rw of
parameters Rw, m, and n. the brine is also determined at a test temperature consistent
There are several methods to measure Rw. The ¿rst and with required test conditions. Sample preparation for the
most direct consists of taking a water sample from production, determination of m includes placement in a hydrostatic core
and simply measuring its resistivity. The water sample must cell and increasing con¿ning stress to reservoir conditions.
be positively identi¿ed as coming from the speci¿c formation The test brine is Àowed through the pore space while
of interest, and must not be commingled with production continuously monitoring resistance and temperature. Based
from other zones. Another method for determining Rw starts on the sample’s speci¿c dimensions, an estimate of Ro
by bringing a core to the surface, extracting the Àuids from occurs once the resistance measurement has fully stabilized.
a core plug, then measuring the resistivity of the extracted Equation 15 provides the solution for m and Figure 8a show
water. Specialized laboratories perform this measurement the graphic solution for m.
as part of a routine core analysis. An interesting variation
of this method consists of determining the water resistivity (15)
by applying a new type of measurement directly on the
(a) (b)
Fig. 8—(a) The traditional graphic method for determining cementation exponent m. (b) The traditional graphic method for determining the saturation exponent, n.

In the graphical solution of Figure 8a, multiple samples Description of the Modi¿ed Electrical Properties Test
are combined. A power law regression is used to ¿t formation Method
factor (FF) and porosity (‫ )׋‬while forcing the intercept to In tight unconventional formations, and especially in
unity for selected samples. The slope of the regression ¿t organic mudstones, the above laboratory methods for m
de¿nes m, where the intercept a = 1.00. and n determination are time consuming and not feasible
The determination of the saturation exponent n on for nanodarcy rocks. Extracting all Àuids and impregnating
the same core plugs as above can be achieved through an the core with brine is a challenge. To circumvent these
extension of the previous series of steps. Beginning with the dif¿culties, the authors have recently proposed a new
100% brine-saturated state, each sample is desaturated in method (Newsham et al., 2018), better adapted to ultralow
order to get several Rt measurements at several known partial permeability formations. In the new method, there is no
water saturations. The test uses a water-wet porous plate at cleaning, drying or resaturating the core-plug sample with
the downstream end of the sample. This porous plate allows brine, as is done with the classical method. Since all tests use
wetting-phase production while retaining the nonwetting samples in as-received (AR) condition, the use of alternative
phase. Introduction of a low pressure, nonwetting phase Àuid methods is necessary to estimate parameters, such as Rw,
occurs at the upstream end. Measurements of the produced Ro, Swt, and ‫׋‬t. These parameters are used with bulk Rt
water properties include resistance, weight and temperature. measurements of samples to ultimately de¿ne m and n. The
Produced water weight, resistance and temperature continues advantage of the new method is the ability to maintain AR
for a series of several pressures steps to drain the sample to, or wettability conditions in the core samples without requiring
near, irreducible water saturation. Following the completion the long process of wettability restoration, especially in low-
of the drainage steps, Dean-Stark core analysis veri¿es the permeability rocks.
remaining unproduced water volume. A material balance Figure 9 shows a workÀow diagram of the Modi¿ed
between remaining water and produced water validates Electric Properties Method. Using AR core material, in-
the water saturations at each pressure. After conversion of situ formation water resistivity, total porosity, total water
resistance to resistivity (Rt), the saturation exponent, n, is saturation and bulk rock resistivity are measured. The
obtained using Eq. 16. Figure 8b is a plot of resistivity index sample is desaturated using a high-speed centrifuge to
(RI) versus Sw used to de¿ne the water saturation exponent, approximately 600 and 1,200 psi of pressure drainage
n, from the slope of a power law, regression ¿t of RI and Sw without using a displacing Àuid, referred to as ¿rst and
while forcing the intercept to unity. second spin, respectively. A log-log crossplot of bulk Rt and
Swt, provides an estimate of Ro using a power law, regression
(16) analysis, as indicated by Eq. 17 and presented in Fig. 10.
(17)
Fig. 9—The workÀow diagram of the Modi¿ed Electric Properties tests.

Pickett-Plot Option
Another well-known method for determining Rw and m,
but not n, is the Pickett plot (Pickett, 1966). The Pickett plot
is a log-log plot of porosity versus resistivity, as shown in
Fig. 12. It represents a graphic solution of Eq. 10, the Archie
equation. Resistivity is provided from either induction or
laterolog measurement. Total porosity is determined from the
constrained simultaneous inversion. This leaves Rw, and the
electrical properties of m and n as required input, which may
vary by formation. If a portion of the formation is known to
be water bearing, a line ¿tting the corresponding lowermost
points will have a slope representative of m and an intercept
Fig. 10—Graphic solution of Ro from the AR and centrifuge spins. Rt and
representative of Rw. However, in organic mudstones it is
Swt results showing second centrifuge spin results as the orange circle. generally not possible to identify the nonhydrocarbon-
bearing intervals. In this case it is necessary to enter Rw from
an independent measurement, then ¿t the slope of the Ro line
Combining Ro with the measured Rw in Eq. 13 yields FF. bounding the data to determine m.
Furthermore, combining Ro with the measured Rt in Eq. 14
yields RI. In Fig. 11a, the slope of a power-¿t line between
the intercept a = 1 and the AR FF de¿nes m. In Fig. 11b, the
slope of a power-¿t solution for the points with the intercept
being 100% Swt de¿nes n. The multiple points in Figure 11b
are the AR point and the subsequent desaturation points.
Fig. 12—A Pickett plot, the graphic solution of the Archie equation.
THE DYNAMICS: THE IMPACT OF PORE

PRESSURE AND FLUID PHASE ON FLUID
PROPERTIES
(a)
This section describes how Àuid phase impacts the Àuid

properties of liquid-hydrocarbon systems across a range of
pressure, volume and temperature (PVT), API, gas gravity
and gas/oil ratios (GOR). The calculated Àuid properties
include Àuid viscosity and formation volume factor (Bo),
which have a direct impact on Àuid mobility and oil-in-place
(OIP) estimation. Fluid property values are very important
when using density and sonic as input curves in the
constrained simultaneous inversion, especially in a reservoir
with high GOR.
Phase Behavior
(b)
The phase behavior of petroleum Àuids at various
Fig. 11—(a) The log-log graphic solution for m. Figure 11b shows the pressures and temperatures is important in the petrophysical-
log-log graphic solution for n.

CPR process for two main reasons: (1) Fluid types affect the (courtesy Jiasen Tan, Oxy): oil, gas condensate, wet gas, and
input log responses due to their variable hydrogen content, dry gas. The classi¿cation of these Àuid types depends on
and (2) the hydrocarbon storage capacity, in particular the relative location of the subsurface reservoir and surface
formation volume factor, is strongly tied to the EOS and separator pressure and temperature conditions to the phase
PVT properties. We consider four reservoir Àuid types that diagrams of Figs. 13 and 14 (Mavor, 2014; Whitson and
are summarized by the phase diagrams shown in Fig. 13 Brule, 2000).
(a) (b)
(c) (d)
Fig. 13—(a) Phase behavior of oil (courtesy Jiasen Tan). (b) Phase behavior of gas condensate. (c) Phase behavior of wet gas. (d) Phase behavior
of dry gas.
Fig. 14—Phase diagram showing the cricondenbar and cricondentherm thresholds (Mavor, 2014) (modi¿ed from Whitson and Brule, 2000).

Several de¿nitions in relation to Figs. 13 and 14 are from the oil. There is an additional pressure and temperature
provided here. The critical point for a multicomponent drop when oil is collected in the stock tank such that only
mixture represents the speci¿c pressure and temperature a fraction of the total volume is now in liquid form. Gas-
at which the physical differences of the vapor and liquid condensate systems (Fig. 13b) are initially in gaseous phase
phases are indistinguishable. This critical point also de¿nes at pressures that place it outside of the two-phase region
the boundary of the bubblepoint and dewpoint along the and at temperatures above the critical temperature, hence
saturation envelope. The bubblepoint curve (left of the it is 100% at reservoir conditions (Point A). As pressures
critical point in Fig. 13) de¿nes the boundary between 100% decrease in the reservoir from Point A to A’, the two-phase
single-phase liquid and the two-phase (liquid and gas) region region is intersected at the dewpoint curve and liquid begins
at temperatures below the critical temperature. The dewpoint to condense. Liquid formation may increase along the line
curve (right of the critical point in Fig. 13) de¿nes the A’ to B in a process known as retrograde condensation.
boundary between the 100% gas and the two-phase region For gas-condensate systems, the pressure and temperature
below the critical temperature. The family of lines inside the of the surface separator are such that only a fraction of
two-phase region is termed isovol lines and represent the the total produced volume is in liquid form. Finally, wet-
total liquid percent relative to the total volume. From Fig. gas (Fig. 13c) and dry-gas (Fig. 13d) behavior is much
14, the cricondentherm and cricondenbar are the de¿ned as more straightforward since no phase changes occur in the
the highest temperature and pressure, respectively, at which reservoir, regardless of the pressure drop at isothermal
the gas and liquid phases can coexist along the saturation conditions. This is because the reservoir temperature is
envelope. beyond the cricondentherm and the depletion path always
The four Àuid types each follow a unique path through lies outside of the two-phase region. In the wet-gas system
the phase diagram as pressure decreases in the reservoir due of Fig. 13c, the line from original reservoir conditions to the
to depletion at isothermal conditions, and while the Àuid separator conditions will result in liquid dropout forming
undergoes additional pressure and temperature changes outside of the reservoir within the production system. In
as it Àows to the surface separator. For black-oil systems, dry-gas systems (Fig. 13d), the depletion path from original
Point A in Fig. 13a represents initial reservoir pressure and reservoir conditions to surface conditions never results in
temperature (Tr) conditions. This point is in the 100% liquid a phase change due to pressure and temperature changes.
region at temperatures below the critical temperature, and Again, we’ll focus on the phase behavior of oil systems
at pressures above the bubblepoint curve, also known as the within this tutorial.
undersaturated region. The line A to A’ represents pressure Since PVT phase envelopes are not always available,
depletion at a constant temperature until the bubblepoint McCain (1990) developed ¿eld rules for de¿ning the various
curve is reached, at which point the ¿rst free gas in the Àuid phases. Table 2 shows the McCain (1990) ¿eld rules to
reservoir evolves from the oil and the system is considered describe various Àuid phases from black oil to dry gas. From
saturated. Additional changes from Point A’ to the separator Table 2, GOR provides the greatest ¿delity as to Àuid phase,
pressure and temperature at the surface releases more gas followed by stock-tank liquid color.
Table 2—McCain (1990) Field Identi¿cation of Fluids
* Heptane-Plus cutoff is de¿ned here for engineering purposes as opposed to regulatory de¿nitions.

Equations of State (21)

Key Àuid properties, such as Bo, viscosity and Pb are
described in series of equations tuned to speci¿c ranges of
Àuid properties. These equations are referred to as equations
of state (EOS). All of these properties are dynamic in that (22)
they vary with pressure, temperature and depth and have
enormous impact on the estimate of storage and Àow
capacity, as seen in Eqs. 18 and 19. Bo is in the denominator
(23)
of both equations. Changes in viscosity inÀuence the Àuid
mobility ratio, the ratio of effective permeability to viscosity,
hence also impact Àow rate. GOR and API gravity have the
greatest impact on Bo, followed by pore pressure and gas (24)
gravity. However, pore pressure has a large impact on Àow
capacity. Accurate estimation of Bo and Pb are critical to
accurate estimates of original oil in place (OOIP). If the
reservoir pore pressure is above the bubblepoint pressure, (25)
the system is considered undersaturated. Reservoirs where
the pressure is signi¿cantly above bubblepoint pressure are
considered ideal. (26)
EOSs are typically derived from laboratory PVT tests of
Àuid samples for speci¿c PVT conditions that match reservoir
conditions. Downhole Àuid-sampling programs provide the
Eqs. 27 and 28 are the Standing’s correlations (Standing,
best source of reservoir Àuid samples. However, downhole
1952) for bubblepoint pressure (Pb) and Bo.
samples are expensive and less frequently available then
are surface recombined samples, where both the gas and
oil components are gathered and then recombined under (27)
laboratory conditions emulating reservoir conditions. The
EOSs take the general form shown in Eqs. 21 through 26.
(18) (28)
(19) where ȖAPI is the gravity of stock-tank oil, Ȗg is the gas gravity
(air = 1), Ȗo is the stock tank oil density (g/cm3), GOR is gas/
oil ratio (scf/STB), T is the reservoir temperature (°F), Pi
is the initial reservoir pressure (psia), Pb is the bubblepoint
(20) pressure (psia), Boi is the formation volume factor at Pi (bbl/
STB), Bob is the formation volume factor at Pb (bbl/STB), Co
is the unsaturated oil compressibility (1/psi), ȡo is the live-oil
density (g/cm3) and ȝo is the live-oil viscosity (cP).
Where A is the area (acres), H is the reservoir thickness Commonly used EOSs were evaluated for their accuracy
(ft), ‫׋‬t is total porosity, Keff is effective permeability, Swt is to estimate Bo and Pb for Permian oils from over 30 PVT
the total water saturation, Bo is the initial formation volume samples. Table 3 presents the error analysis in comparing
factor (bbl/STB), PI is the productivity index, Q is the various public domain EOS correlations for bubblepoint
Àow rate, Pi is the initial reservoir pressure (psi), Pfbh is the pressure (Pb) and formation volume factor (Bo) to the the
Àowing bottomhole pressure (psi), ȝo is the viscosity (cP), Re PVT measurements. Clearly, this indicates that generally
is the drainage radius, Rh is the wellbore radius, and Į is a available EOSs may not provide enough accuracy in the
conversion factor. estimation of Bo and Pb. Hence the authors recommend that
operators pursue laboratory PVT analysis on a range of
produced Àuids and build project-speci¿c EOSs.

Table 3—Error Analysis of Public-Domain EOS Corrections (Courtesy Jiasen Tan)
Density and Acoustic Response to Phase Behavior Pore Pressure Calibration and Estimation From Logs
Fluid phase has a signi¿cant impact on the Àuid Pore pressure calculation plays an important role in
response of the density and acoustic logs. Depending on estimating hydrocarbon in place. In most basins of North
pressure, temperature and the GOR, the Àuid responses for America, pore pressure is not constant and varies as a
bulk density and compressional sonic can vary signi¿cantly function of depth, organic maceral type, organic maturity,
from the standard default values of 0.85 g/cm3 and 190 ȝs/ft., and seal ef¿ciency. The authors have observed pore pressure
respectively. Batzle and Wang (1992) published equations variations ranging from 0.44 psi/ft to as high as 0.90 psi/ft in
for estimating compressional slowness, density and elastic a given interval. For this reason, we estimate pore pressure
properties for a range of Àuid properties. Han and Batzle pro¿les from logs calibrated to a pore pressure reference.
(2000) updated the correlations for a greater the range of In the organic-mudstone environment, the best source for
GORs. Figure 15 shows two models; one for a dead-oil estimating pore pressure is from the diagnostic formation
system (Fig. 15b) with a GOR equal to 56 scf/STB and API injectivity test (DFIT) acquired in either vertical pilot holes
gravity of 33°, and a volatile oil (Fig. 15a) with a GOR of or at the toe position of a lateral well. The best logs for
2,246 scf/STB and API gravity of 49.9°. For the dead-oil estimating pore pressure are the density, acoustic log and
example, the bulk density and compressional slowness are resistivity. We use a basic Eaton (1975) function to estimate
essentially equal to the default values mentioned above. For pore pressure calibrated to DFIT values of pore pressure.
the volatile (live) oil, the bulk density and slowness are 0.56 Equation 29 is the formulation for the acoustic log but the
g/cm3 and 360 ȝs/ft., respectively. By comparison, these are same form of the equation is used for the resistivity.
signi¿cantly different properties based on a change in Àuid
phase. Providing accurate Àuid response inputs will have an
impact on the estimation of total porosity. (29)
(a) (b)
Fig. 15—Batzle and Wang (1992) and Han and Batzle (2000) Àuid-response model results for a volatile oil system (a) and a dead oil system (b).

where Pp is the pore pressure, ıv is the overburden stress apply a trend curve to capture the gross pore pressure
gradient (psi/ft.), ın is the normal pore pressure gradient variations with depth for use in the storage calculations. In
(psi/ft.), ¨Tn is the slowness response of the normal pressure Track 5, this pressure trend is the red dashed curve.
trend, ¨Tm is the measured acoustic slowness and x is the
Eaton coef¿cient. The overburden stress is estimated by CONSTRAINED SIMULTANEOUS INVERSION
integrating the bulk-density log from surface to total depth
of the well. Where the density isn’t available Eq. 30 can be The workÀow relies on a constrained simultaneous
used to de¿ne the overburden stress as de¿ned by Traugott inversion method in order to minimize the nonuniqueness of
(1997). a free-running inversion. This nonuniqueness is particularly
problematic when available logs are limited to a triple-
(30) combo. Former colleagues of the authors provide a very
good description of this method (Vosburgh et al., 2013).
Essentially, the method simultaneously solves tool-response
where W is the water depth (ft), D is the true vertical depth equations and constrains tool-response parameters (Mayer
(ft), A is the air gap (ft), ȡave is the average sediment density and Sibbit, 1980). Constraint is achieved by entering as input,
(g/cm3). TOC (in wt%), Vclay (calibrated to wt%) and Àuid properties,
Figure 16 shows a pore pressure analysis using a sonic previously computed outside of the simultaneous inversion.
log. Pore pressure estimates are limited to the higher clay- These external inputs serve as “ground truth” by forcing
bearing intervals since clay is thought to be less elastic the TOC and clay values input into the inversion to equal
and hence, retains more of the maximum stress imparted the values exiting the inversion. A linear function between
at a given depth. The GR threshold is used to limit which pyrite and TOC serves as the solution for pyrite internal to
acoustic points are used in the pore pressure estimate. The the inversion. Alternatively, pyrite can be estimated external
normal pressure trend is de¿ned in the “Sonic Model” track to the inversion and input as a curve similar to the TOC and
of Fig. 16. Equation 29 is applied to the acoustic offset from clay volume (see Fig. 6). Note that in many instances the
the normal pressure trend. The Eaton coef¿cient, x, is used Àuid parameters entered as curves are signi¿cantly different
as a tuning parameter to calibrate the pore pressure and from the default values of any given software application,
gradient to the DFIT measurement (black square). Since the especially for hydrocarbons. The typical default values for
pore pressure result has the tendency to be noisy, the authors density and acoustic response to oil may be signi¿cantly in
Fig. 16—A pore pressure pro¿le analysis using the sonic log to calibrate a DFIT. Track 1, gamma ray; Track 2 depth, Track 3, sonic model; Track 4,
pore pressure gradient; Track 5, fracture pressure gradient.

error, especially as the volatility of the oil increases with an note that Eq. 18 is speci¿cally designated here for black-oil
increasing GOR. The conversion from weight percent inputs systems at temperatures less than the critical temperature of
to volume percent outputs occurs as part of the inversion. the reservoir Àuid (Fig. 13a). Referring to Fig. 4 in Part 1
Figure 17 shows a typical inversion matrix. The section to of this tutorial, the use of a gas-storage-capacity workÀow
the left de¿nes the tool-response equations for use within is required for systems at temperatures above the critical
the inversion. The section to the right de¿nes the endpoint temperature. Retrograde condensate, in addition to wet-
minerals and Àuid responses for use in the solution of and dry-gas systems, use the gas formation volume factor
simultaneous equations. (Bg) as opposed to Bo, and the total hydrocarbon Àuids
The method solves a set of normalized linear or nonlinear entrained within the system are determined by considering
equations normalized by a con¿dence weighting factor. The the produced liquid yield in bbl/MMSCF. Additionally, gas
equations are solved as a system of equations, i.e., matrix systems present a unique challenge when part of the total
algebra. The resultant volumes are standardized so that they gas in place is absorbed within the pores of the organic
add up to 1.0. Synthetic logs are reconstructed from the matter. Mavor provides an overview of gas-in-place (GIP)
computed volumes. The total normalized error is calculated estimation (Mavor, 1994, 2014; Mavor and Nelson, 1997).
from the difference between the input log and results using The reservoir thickness (H) is meant to represent the
Eq. 31. total rock containing hydrocarbons in storage. This de¿nition
is much more inclusive than other de¿nitions of “net pay”
(Worthington and Consentino, 2005; Worthington, 2010;
(31)
Bust et al., 2011). The authors recognize that net pay is a
dynamic term and changes not only with reservoir pressure
where Crvi is the ith input curve value, CrvReci is the ith and temperature conditions, but also with the economics of
reconstructed curve from the mineral volumes, CrvToli is the the commodity. OIP estimates are intended to aid in de¿ning
ith input curve tolerance; NumCrvs is the number of input the opportunity. Hydrocarbon Àow via the PI of Eq. 20 pays
curves in the model. the bills.
The process is run through many iterations, varying Finally, the total water in place (OWIP) is determined
the mineral volumes until a minimum error between the by replacing the (1 í Swt) term in Eq. 12 with simply Swt.
synthetic logs and the measured logs is achieved. The total gas in place (OGIP) dissolved in the oil at initial
reservoir conditions is determined by the product of the OIP
ESTIMATION OF STORAGE VOLUMES and producing GOR.
Storage volumes are determined from the classic OIP EXAMPLES SHOWING THE WORKFLOW
equation (Eq. 18). “PETROPHYSICS CPR”
The three petrophysical properties (‫׋‬t, Swt, and Bo) in
this tutorial have been thoroughly de¿ned and their values One of the primary objectives of the workÀow was
ascribed by the workÀow described earlier. The authors to honor our theme of “Petrophysics CPR”; consistent,
Fig. 17—The interface for the matrix solution used in the Constrained Inversion mode. TOC, clay and Àuid parameters previously computed by
deterministic methods are entered as curve input with a 0.01 con¿dence parameter. Here, VCL is dry wt%, and XVCL is bulk vol%.

portable and reliable results. The best way to demonstrate x Track 3 – Resistivity: deep and shallow logs.
this is by example. We will show examples from various x Track 4 – Nuclear: Bulk density, compensated
basins within the Permian. We’ll “ground truth” the neutron, photoelectric logs and core bulk density,
petrophysical calculations with core-based measurements where available.
and some independent cross-checks using advanced logs, x Track 5 – Inversion: Constrained simultaneous
such as the magnetic resonance (MR) and dielectric logs. inversion mineral distribution pro¿le showing bulk
The core-based examples are paired examples; the mineral clay, quartz-feldspar, bulk carbonate, pyrite and total
composition match followed by the storage match. As porosity.
discussed in previous sections, XRD, XRF and FTIR are x Track 6 – Composition: Mineral composition tool
the methods used to determine the core-based mineral results include bulk clay, quartz-feldspar, bulk
composition. Core measurement of total porosity and total carbonate and pyrite.
water saturation is determined using either the crushed-rock x Track 7 – Bulk clay: Continuous curve is from the
method or benchtop MR. The authors will show examples inversion. The discrete points are from core
from the New Mexico Delaware Basin, the Texas Delaware x Track 8 – QF: Quartz-feldspar
Basin and the Midland Basin. x Track 9 – Bulk carbonate: Includes limestone and
Figure 18 shows an example from the New Mexico dolomite, if calculated.
Delaware Basin. A similar plot format is used for all the x Track 9 – Pyrite
following examples. All mineral volumes are converted x Track 10 – TOC: Total organic carbon in weight
from weight to volume percent. The track sequence is percent. The inversion converts to volume percent.
x Track 1 – Correlation: GR; well log and core, where x Track 11 – Output equation: Grain density.
available. x Track 12 – Total water saturation.
x Track 2 – Depth. x Track 13 – Total porosity.
Fig. 18—Core-log comparison of the mineral content from the New Mexico Delaware Basin. The discrete points are from core. Discrete black points
are sourced from cuttings and/or core.

Figure 19 displays the storage properties for the same x Track 9 – Water/Oil: Water/oil ratio estimated from
example shown in Figure 18. A similar plot format used for OIP calculations, formation pressure, temperature
all examples. The mineral distributions are converted from and formation volume factor.
weight to volume percent. The track sequence is x Track 10 – Inversion: Constrained simultaneous
x Track 1 – Correlation: GR; well log and core, where inversion mineral distribution pro¿le showing bulk
available. clay, quartz-feldspar, bulk carbonate, pyrite and total
x Track 2 – Depth. porosity.
x Track 3 – Resistivity: Deep and shallow resistivity x Track 11 – Composition: Mineral composition tool
logs. results includes bulk clay, quartz-feldspar, bulk
x Track 4 – Nuclear: Bulk density, compensated carbonate and pyrite.
neutron, photoelectric logs and core bulk density,
where available. Figures 20 and 21 are examples from the Texas Delaware
x Track 5 – Total Sw: Total water saturation with discrete Basin. Figure 20 shows the mineral distribution from the
black core points. ECS tpp; rather than a LithoScanner pro¿le. Figure 21 shows
x Track 6 – Total Porosity: Total porosity with discrete the storage properties pro¿le: there are two Phase Properties
black core points. tracks showing the water/oil ratio, formation pressure,
x Track 7 – Total Porosity: Bulk volume hydrocarbon temperature and formation volume factor matches to DFIT
and water. Also shows the MR total porosity match (discrete points) results and PVT results (thick black, blue
to the inversion. and green lines) in the portion of the well. Figures 22 and 23
x Track 8 – OIP/GIP: Oil, water and gas volumes are examples from the Midland Basin. Figure 22 shows the
estimated on a per section basis. mineral distribution and Fig. 23 shows the storage properties
pro¿le. Discrete black points are sourced from core.
Fig. 19—Core-log comparison of the storage properties for the same well shown in Fig. 18, which is from the New Mexico Delaware Basin. The
discrete points are from core. Discrete black points are sourced from cuttings and/or core.

Fig. 20—Core-log comparison of the mineral content from the Texas Delaware Basin.
Fig. 21—Core-log comparison of the storage properties from the Texas Delaware Basin.

Fig. 22—Core-log comparison of the storage properties from the Midland Basin.
Validation Using the “Three-Point Stance” In Figure 23, the mineral composition is validated by
The authors use multiple independent cross-checks comparing the compositional tool results to the inversion.
to validate the constrained simultaneous inversion results Track 6, MR total porosity, displays a good match between
including the use of measured core data, MR total porosity, the MR (red curve) and inversion total porosity (blue curve).
compositional logging tools and the dielectric log, when Track 7, dielectric BVW, shows good agreement between
available. As presented in Figs. 18 through 23, core the dielectric BVW (red curve) and inversion BVW (blue
measurements provide a validation of mineral composition, curve). In Figure 24, the ‫׋‬t track (Track 5) shows a good
total porosity, total water saturation and grain density. Within match between the MR (green curve) and inversion total
the total porosity track of each storage properties example, porosity (red curve). In the same ¿gure, Tracks 6 and 7 are
the MR total porosity is matched to the total porosity from the BVW and water saturation (Sw) tracks, respectively. In
the constrained simultaneous inversion. The MR porosity each track, the red curve is from the inversion. In the SWT
is not used within the inversion solution, hence, it provides track (Track 7), the dielectric water saturation is estimated
an independent cross-check of the inversion results. The from the ratio of the dielectric BVW and the inversion total
authors do not use the compositional tools as input to the porosity. The comparison is somewhat ‘noisy’ but the overall
inversion, thus providing an independent comparison of the trends match well. In Figure 25, the track arrangement is
mineral distribution. The dielectric log provides a measure the same as in Figure 24. However, the blue curves are
of the bulk volume of water (BVW). Again, the dielectric from the inversion and the red curves from either the MR
log is not used within the solution. It is used for comparison or dielectric BVW. The PORT track (Track 6) compares
to the inversion BVW. Furthermore, an independent check MR and inversion total porosity. The BVW track compares
of total water saturation occurs by dividing the inversion or the inversion BVW to the dielectric BVW. The SWT track
MR total porosity into the dielectric BVW. Figures 23, 24 compares the inversion Swt to the dielectric Swt computed
and 25 are examples from various ¿elds and basin where the from the ratio of the dielectric BVW and the inversion total
MR and dielectric BVW provide a quality assurance check porosity.
on the inversion results.

Fig. 23—The matches between the inversion and independent measures of mineral, total porosity and BVW.
Fig. 24—Good matches between the inversion and independent measures of mineral, total porosity and BVW.

Fig. 25—Good matches between the inversion and independent measures of total porosity and BVW.
SUMMARY We discussed the rationale for using a total porosity

and water saturation approach. Determination of Archie’s
In Part 2, we have described in more detail the protocol parameters were reviewed by use of the Pickett plot and core-
used within the organic-mudstone workÀow with emphasis based methods. A discussion of Àuid properties highlights
on calibration to known references, such as laboratory core the need for accurate EOSs from PVT tests so appropriate
measurements and/or independent log-based measurements, values of Bo are used for estimating OIP. Variations in GOR
to support the authors’ claim of Petrophysics CPR; consistent, and API gravity cause signi¿cant variation in the density
portable and reliable analysis results. We showed how clay and acoustic Àuid responses well beyond the default values,
volume estimates are directly calibrated to the rock by use of usually applied for dead oils. The authors stress the need
nuclear models that yield density, neutron and photoelectric for ¿eld-speci¿c EOSs for best results. A brief discussion
response lines useful as a bounding solution in crossplots. of log-based pore pressure shows that the Eaton equation
The average of multiple log-based TOC indexes yields the can be successfully applied in mature basins, but requires
best match to core and/or cuttings measures of TOC. A linear calibration, typically to DFIT pore pressure results. Finally,
relationship between TOC and pyrite provides the ¿nal the authors showed multiple examples of how well the
external constraint to the simultaneous inversion method. inversion results match to rock composition, total porosity
Clay volume, TOC and pyrite are the mineral outliers where and total water saturation across a range of basins. However,
“a little bit goes a long way” towards the response impact on we did focus the discussion to “oily” systems rather than
the log responses. It’s for this reason that the authors have “gassy” systems for brevity. The workÀow should be
taken the approach of estimating these properties external considered a framework and is not implied as the only
to the simultaneous inversion by ‘locking these responses approach to evaluating organic-mudstone and tight oil
down” deterministically and then importing the results as systems. It is comprehensive and yields consistent results
“log responses” into the simultaneous inversion. This is satisfying the Petrophysics CPR mandate.
essentially how the inversion is constrained; the values of In Part 3, the authors will redirect the discussion
clay, TOC and pyrite going into to the inversion have the towards crossdisciplinary methods that complement the
same values exiting the inversion. workÀow. We’ll discuss some geochemistry and engineering

applications including the use of programmed pyrolysis Dt = compressional acoustic log slowness
to estimate OIP and the fractional Àow analysis to better DenClay = density log value for the wet-clay point
de¿ne water prone intervals. We’ll tie the storage results DenCl1 = density log value de¿ning the zero-porosity
to production and discuss forecasting of WOR or produced matrix point
water cut. We’ll expand on the use of advanced logs, such as DenCl2 = density log value de¿ning the trajectory of the
MR and dielectric logs to serve as independent cross-checks Àuid point
of the storage properties and provide additional examples in DryWt% = value from XRD or FTIR
support of Petrophysics CPR. Fd = TOC divisor in Faust equation
FF = resistivity formation factor
NOMENCLATURE GR = gamma-ray log value
GrClay = gamma-ray log value in 100% clay
Abbreviations GrClean = gamma-ray log value in 0% clay
AR = as received GV = grain volume
BVW = bulk volume water H = reservoir thickness
DFIT = diagnostic fracture injection test Keff = effective permeability
D-N = bulk density and neutron logs LOM = level of maturation
EOS = equation of state m = porosity exponent in Archie’s equation
FTIR = Fourier-transform infrared spectroscopy n = saturation exponent in Archie’s equation
GOR = gas/oil ratio NeuClay = neutron log value for the wet-clay point
GR = gamma ray NeuCl1 = neutron log value de¿ning the zero-porosity
MR = magnetic resonance matrix point
NMOD = Amoco nuclear modeling program NeuCl2 = neutron log value de¿ning the trajectory of
OGIP = original gas in place the Àuid point
OIP = oil in place NumCrvs = number of input curves in the model
OOIP = original oil in place Pb = bubblepoint pressure
OWIP = original water in place Pfbh = bottomhole Àowing pressure
PVT = pressure, volume, temperature Pi = initial reservoir pressure
SnuPar = Schlumberger nuclear parameter program PI = productivity index
TOC = total organic carbon Pp = pore pressure
UTAPWeLS = University of Texas at Austin Petrophysical PV = pore volume
and Well-Log Simulator Q = Àow rate
WIP = water-in-place RE = recovery ef¿ciency
XRD = X-ray diffraction Rhob = density log value
XRF = X-ray Àuorescence RI = resistivity index
Re = drainage radius
Symbols Rclay = clay resistivity
A= area Rt = formation resistivity
Acoef = coef¿cient in the Schmoker method Ro = formation resistivity if fully water saturated
AFaust = variable in Eq. 9 Rh = wellbore radius
B= constant in Eq. 4 Rw = formation water resistivity
Bo = formation volume factor Rwno = water resistivity of a nonorganic interval
Bob = formation volume factor at SWT = total water saturation
bubblepoint pressure T = temperature in °F
Boi = formation volume factor at initial TOCFaust = total organic carbon from Faust equation
reservoir pressure TOCPass = total organic carbon from Passey equation
BV = bulk volume TOCSchm = total organic carbon from Schmoker equation
Co = unsaturated oil compressibility TS = linear scalar
Crvi = ith input curve VclDN = clay volume from the density-neutron
CrvReci = ith reconstructed curve from mineral volumes crossplot
CrvToli = ith input curve tolerance VclGR = clay volume from the gamma-ray log

Vre = vitrinite reÀectance equivalent Eaton, B.A., 1975, The Equation for Geopressure Prediction from
VRO = observed vitrinite reÀectance Well Logs, Paper SPE-5544 presented at the Fall Meeting of
W= water depth the Society of Petroleum Engineers of AIME, 28 September–
x= Eaton coef¿cient (Eq. 29) 1-October. DOI: 10.2118/5544-MS.
Edmundson, H., and Raymer, L. L., 1979, Radioactive Logging
Z= true vertical depth
Parameters for Common Minerals, The Log Analyst, 20(5),
38–47.
Ȗg, = gas gravity Faust, L.Y., 1953, A Velocity Function Including Lithologic
Ȗo = stock-tank oil density Variation, Geophysics, 18(2), 271-288. DOI:
ȖAPI = gravity of stock-tank oil 10.1190/1.1437869.
ǻTm = acoustic slowness Han, D., and Batzle, M., 2000, Velocity, Density and Modulus
ǻTn = acoustic slowness response of the normal of Hydrocarbon Fluids—Empirical Modeling, SEG
pressure trend Technical Program Expanded Abstracts 2000, 1867–1870.
oil viscosity DOI:10.1190/1.1815793.
ȝo =
Hanna, R., Griffen, T., Jr., Martin, P., Lee, R., and Dacy, J., 2014,
ȡave = average sediment density
Systems and Methods for the Determination of Formation
‫׋‬clay = clay total porosity Water Resistivity and Conductivity, US Patent No. 8,843,319
ȡdryclay = dry-clay density B2, granted September 23, 2014.
ȡÀ = Àuid density Ijasan, O., Torres-Verdín, C., and Preeg, W.E., 2013, Fast Modeling
ȡg = gas density of Borehole Neutron Porosity Measurements with a New
ȡgd = grain density Spatial Transport-Diffusion Approximation. Geophysics,
ȡo = live oil density 78(3), D151–D168. DOI: 10.1190/geo2012-0433.1.
ȡwetclay = wet-clay density Jagodzinski, H., 1949, Eindimensionale Fehlordnung in Kristallen
ın = normal pressure und ihr EinÀuss auf die Rontgeninterferenzen. I. Berechnung
des Fehlordnungsgrades aus der Rontgenintensitaten:
ıv = overburden pressure
Acta Crystallographica, 2(4), 201–207. DOI: 10.1107/
‫׋‬e = effective porosity S0365110X49000552.
‫׋‬t = total porosity Keller, G.V., 1953, Effect of Wettability on the Electrical Resistivity
‫׋‬clay = wet-clay porosity of Sand, Oil and Gas Journal, 31(4), January 5, 1953, 62–65.
Lasater, J. A., 1958, Bubble Point Pressure Correlation, Paper
SPE-957-G, Journal of Petroleum Technology, 10(5), 65–67.
DOI: 10.2118/957-G.
REFERENCES Mavor, M.J., 1994, Recent Advances in Coal Gas-Well Openhole
Well Completion Technology. Paper SPE-26211, Journal of
Al-Marhoun, M.A., 1988, PVT Correlations for Middle East Crude Petroleum Technology, 46(7), 587–593. DOI: 10.2118/26211-
Oils, Paper SPE-13718, Journal of Petroleum Technology, PA.
40(5), 650–666. DOI: 10.2118/13718-PA. Mavor, M., 2014, Reservoir Fluid Properties Required for Low-
Anderson, W.G., 1986, Wettability Literature Survey—Part 3: The Permeability Oil Reservoir Analysis, AAPG Search and
Effects of Wettability on the Electrical Properties of Porous Discovery Article No. 80363. http://www.searchanddiscovery.
Media. Paper SPE-13934, Journal of Petroleum Technology, com/documents/2014/80363mavor/ndx_mavor.pdf. Accessed
38(12), 1371–1378. DOI: 10.2118/13934-PA. March 11, 2019.
Archie, G.E., 1942, The Electrical Resistivity Log as an Aid in Mavor, M.J., and Nelson, C.R. 1997, Coalbed Reservoir Gas-In-
Determining Some Reservoir Characteristics, Transactions, Place Analysis, Gas Research Institute Topical Report GRI-
AIME, 146, 54–62. DOI: 10.2118/942054-G. 97/0263.
Batzle, M., and Wang, Z., 1992, Seismic Properties of Pore Fluids, Mayer, C., and Sibbit, A., 1980, Global, A New Approach to
Geophysics, 57(11), 1396–1408. DOI: 10.1190/1.1443207. Computer-Processed Log Interpretation, Paper SPE-9341
Bust, V.K., Majid, A., Oletu, J.U., and Worthington, P.F., 2011, The presented at the SPE Annual Technical Conference and
Petrophysics of Shale Gas Reservoirs: Technical Challenges Exhibition, Dallas, Texas, USA, 21–24 September. DOI:
and Pragmatic Solutions, Paper IPTC-14631-MS presented 10.2118/9341-MS.
at the International Petroleum Technology Conference, McCain, W.D., 1990, Properties of Petroleum Fluids, 2nd ed.,
Bangkok, Thailand, 15–17 November. DOI: 10.2523/IPTC- PennWell Publishing. ISBN: 978-0878143351.
14631-MS. Mendoza, A., Torres-Verdín, C., and Preeg, W.E., 2010a, Linear
Clavier, C., Coates, G., and Dumanoir, J., 1984, Theoretical Iterative Re¿nement Method for the Rapid Simulation of
and Experimental Bases for the Dual-Water Model for Borehole Nuclear Measurements, Part 1—Vertical Wells:
Interpretation of Shaly Sands, Paper SPE-6859, Society Geophysics, 75(1), E9–E29. DOI: 10.1190/1.3267877.
of Petroleum Engineers Journal, 24(2), 153–168. DOI: Mendoza, A., Torres-Verdín, C., and Preeg, W.E., 2010b, Linear
10.2118/6859-PA. Iterative Re¿nement Method for the Rapid Simulation of

Borehole Nuclear Measurements, Part 2—High-Angle Interstrati¿ed Mineral and the Diagram for its
and Horizontal Wells: Geophysics, 75(2), E79–E90. DOI: Identi¿cation, Clay Science, 7, 97–14. DOI: 10.11362/
10.1190/1.3335953. jcssjclayscience1960.7.97. Reichweite Illite Classi¿cation
Moore, D.M., and Reynolds, R.C., 1989, X-Ray Diffraction and Referenced.
the Identi¿cation and Analysis of Clay Minerals, Oxford Waxman, M.H., and Smits, L.J.M., 1968, Electrical Conductivities
University Press. ISBN: 978-0195951704. in Oil-Bearing Shaly Sands, Society of Petroleum Engineers
Newsham, K., Comisky, J., and Chemali, R., 2019, Organic- Journal, 8(2), 107–122. DOI: 10.2118/1863-A.
Mudstone Petrophysics: WorkÀow to Estimate Storage Wiley, R., and Patchett, J.G., 1990, CNL Neutron Porosity
Capacity, Petrophysics, 60(1), 4–15. DOI: 10.30632/ Modeling, a Step Forward, The Log Analyst, 31(3), 133–149.
PJV60N1-2019t1. Whitson, C.H., and Brule, M.R., 2000, Phase Behavior, SPE
Newsham, K., Hanna, R., Lee, R., Whitney, C., and Chemali, Monograph Series 20. ISBN: 978-1-55563-087-4.
R., 2018, As-Received Core Electrical Properties Tests for Worthington, P.F., 2010, Net Pay-What Is It? What Does It
Determining Formation Factor FF and Resistivity Index Do? How Do We Quantify It? How Do We Use It? SPE
RI, Paper SPE-191476 presented at the 2018 SPE Annual Reservoir Evaluation & Engineering, 13(5), 812–822.
Technical Conference and Exhibition, Dallas, Texas, USA, DOI:10.2118/123561-PA.
24–26 September. DOI: 10.2118/191476-MS. Worthington, P. F., and Cosentino, L., 2005, The Role of Cut-
Passey, Q.R., Creaney, S., Kulla, J.B., Moretti, F. J., and Stroud, J. offs in Integrated Reservoir Studies, Paper SPE-84387, SPE
D., 1990, A Practical Model for Organic Richness from Porosity Reservoir Evaluation & Engineering, 8(4), 276–290. DOI:
and Resistivity Logs, AAPG Bulletin, 74(12), 1777–1794. 10.2118/84387-PA.
DOI:10.1306/0C9B25C9-1710-11D7-8645000102C1865D. Wyllie, M.R., and Gregory, A.R., 1953, Formation Factors of
Petrosky, G.E., Jr., and Farshad, F.,1998, Pressure-Volume- Unconsolidated Porous Media: InÀuence of Particle Shape
Temperature Correlations for Gulf of Mexico Crude Oils., and Effect of Cementation, Journal of Petroleum Technology,
Paper SPE-51395, SPE Reservoir Evaluation & Engineering, 5(4), 103–110. DOI: 10.2118/223-G.
1(5), 416–420. DOI: 10.2118/51395-PA.
Pickett, G.R., 1966, A Review of Current Techniques for
Determination of Water Saturation From Logs, Paper SPE- ABOUT THE AUTHORS
1446, Journal of Petroleum Technology, 18(11), 1425–1433.
DOI: 10.2118/1446-PA.
Poupon, A., and Leveaux, J., 1971, Evaluation of Water Saturation Kent Newsham is Chief
in Shaly Formations, The Log Analyst, 12(4), 3–8.
Petrophysicist for Unconventional
Simandoux, P., 1963, Mesures Diélectriques en Milieu Poreux,
Application a la Mesure des Saturations en Eau, Etude du
Resources and Director of Petrophysics at
Comportement des Massifs Argileux, Revue de I’Institut for Occidental Oil and Gas Corporation in
Français du Pétrole, Supplementary Issue, 193–215. A Houston, Texas. His prior employer was
version in English was published in 1982 in the SPWLA Shaly Apache Corporation. At Apache, Kent
Sand Reprint Volume. held the position of Global Director of
Schmoker, J., and Hester, T., 1983, Organic Carbon in Bakken Petrophysics and Distinguished Advisor
Formation, United States Portion of Williston Basin, AAPG before his promotion to Chief of Staff. Prior to joining
Bulletin, 67(12), 2165–2174. Apache in 2003, Newsham worked in various petrophysicist
Standing, M.B., 1952, Volumetric and Phase Behavior of Oil Field
roles of increasing responsibility for Anadarko Petroleum,
Hydrocarbon Systems, Reinhold Publishing. Reprinted by
Amoco, and Unocal from 1981 to 1992.
SPE in 1977. ISBN: 978-0895203007.
Traugott, M. 1997, Pore Pressure and Fracture Pressure Kent has authored or coauthored numerous technical
Determinations in Deepwater, World Oil, 281(8), 68– papers and patents. In 2004, Kent codiscovered one of the
70, Deepwater Technology Supplement. http://www. largest gas ¿eld complexes in North America located in
hxrdrillingservices.com/wp-content/uploads/WorldOil.pdf. the Horn River and Liard Basins of NE British Columbia.
Accessed March 11, 2019. Kent is a graduate from Pennsylvania State University
Vasquez, M., and Beggs, H.D., 1980, Correlations for Fluid Physical with a BS degree in Earth Science-Geology and a minor
Property Prediction, Journal of Petroleum Technology, 32(6), in Marine Science. Newsham is also a graduate of the
968–970. DOI: 10.2118/6719-PA. Amoco Petrophysics Center of Excellence, a masters-based
Vosburgh, E.T., Djordjevic, O.A., and Rushing, J.A., 2013,
program at Amoco Production Research. Mr. Newsham
Liquids-Rich Resource Play Characterization Using Well Log
Analysis Calibrated with Rock Properties from Drill Cuttings,
has established the Newsham-Farris-Apache Petrophysics
Paper URTEC-1620306 presented at the Unconventional Endowment at the University of Oklahoma, and at Penn
Resources Technology Conference, Denver, Colorado, USA, State, the Newsham Petroleum Geosystems Endowment
12–14 August. DOI: 10.15530/URTEC-1620306-MS. and the Newsham-Comisky Endowment in Petroleum
Watanabe, T., 1988, The Structural Model of Illite Smectite Engineering.

Joe Comisky is the discipline lead

and technical advisor for petrophysics at
Devon Energy Corporation in Oklahoma
City. He specializes in the integration of
various geologic and engineering data,
particularly core and log measurements.
Joe trained in geophysics, geology, and
engineering at the Pennsylvania State
University earning a BS and MS in Geosciences. He began
his career working various tight gas and deepwater projects
for Anadarko Petroleum and ConocoPhillips in Houston.
Joe joined the petrophysics team at Apache Corporation
in 2005 where he helped to develop integrated workÀows
for describing both conventional and unconventional rock,
pore, and Àuid systems for their international and domestic
regions. He also performed research and graduate work
at the University of Oklahoma focused on integrating
NMR, capillary pressure, geomechanics, and mineralogy
measurements for various Apache projects. Currently he is
advising and mentoring the staff at Devon Energy in all of
their US basins.
Roland Chemali is Consultant

Petrophysicist for Occidental Oil and
Gas Corporation in Houston, Texas. In
the years prior, Roland was Halliburton
Technology Fellow and Chief
Petrophysicist. He served as President
of the Society of Petrophysicists and
Well Log Analysts (SPWLA) in 2012–
2013. He was also Distinguished Lecturer for the Society
of Petroleum Engineers (SPE) in 2010–2011. For several
decades Roland Chemali dedicated his activity to resistivity,
magnetic resonance and dielectric logging physics
and interpretation. He received the SPWLA Technical
Achievement Award in 1997. Roland earned engineering
degrees from the Ecole Polytechnique of Paris and from the
French Petroleum Institute and a Master’s in Mathematics
from Louisiana State University. He has coauthored over 70
papers and patents.

PETROPHYSICS, VOL. 60, NO. 2 (APRIL 2019); PAGES 208–227; 6 FIGURES; 1 TABLE. DOI: 10.30632/PJV60N2-2019t2
TUTORIAL
Introduction to Resistivity Principles for Formation Evaluation: A Tutorial Primer
David Kennedy1 and Fredy Garcia2
PROLOGUE into resistivity space, Archie’s power law model is revealed

as an approximation to the limiting cases. Thus, from the
The standard model for relating bulk formation conductivity formulation, an intuitive understanding of the
resistivity to porosity and water saturation was introduced Archie model emerges. Moreover, the conductivity model
to the petroleum industry in 1941; it remains the industry can be derived in at least three different ways, each yielding
standard to this day. The model was discovered empirically different insights into formation conductivity.
by means of graphical analysis. Basically, G.E. Archie
discovered that when the logarithm of formation resistivity INTRODUCTION
factor was plotted against the logarithm of porosity the
resulting trend could be ¿tted by a straight line. A similar This article is intended as a tutorial for novice formation
relationship was discovered connecting the logarithms evaluators. Our object is not to teach resistivity interpretation
of resistivity index and water saturation. When these two techniques, but to provide insight into the most commonly
power laws are combined into a single equation, it can be used model; namely, the Archie model for connecting the
solved for water saturation (which is not observable from a brine saturation of a porous rock to its bulk resistivity.
borehole) in terms of bulk formation resistivity, interstitial Although our article is backed by the scholarship required of
brine resistivity, and porosity (all of which can be estimated a Petrophysics article, the editor has granted us permission
from observations made in boreholes). This revolutionized to withhold citations from the text (as in a textbook) in the
log interpretation. There has always been a problem with interest of readability; we will supply a list of the references
the model in terms of its “explainability”. That is, it cannot that we have used in the usual place following the end of
be derived in any straightforward way from accepted the article. Also, we shall strive for clarity over brevity; if
¿rst principles of physics. It does not contradict any ¿rst this means repeating ourselves, it is for pedagogical effect.
principle, but neither does it seem to follow ineluctably from (As has been said since Roman times, “Repetitio est mater
them. However, since the model works, most formation studiorum”, repetition is the mother of learning.) Our
evaluators have memorized the relationships that follow target audience is practitioners of formation evaluation
from the model and simply “get used to them”. That remains newly introduced to the subject. Our goal is that a serious
the situation to this day. However, there is a path around student having worked through this article should have a
this obstacle to understanding formation resistivity at a clear understanding of why rocks conduct the way that they
fundamental level, and that way forward is to abandon the conduct based upon intuitively obvious ¿rst principles,
resistivity formulation in favor of its reciprocal, conductivity. rather than a rote memorization. First principles can be
It is surprising that such a seemingly trivial change could stated verbally. An example from Isaac Newton would be
open a new vista into the relationships among formation “acceleration is proportional to applied force and inversely
electrical properties. A conductivity formulation permits the proportional to mass”. However, to be useful in analysis,
asking of questions about how a formation’s conductivity ¿rst principles must also be expressible as mathematics;
should respond to changes not only in brine conductivity, e.g., a = F/m. A formula such as this is called a “model”; a
but also in the fractional amount of brine in a formation, and successful understanding of the physical world will be based
its geometrical con¿guration. By answering these questions upon internalizing the ¿rst principle (which should be, more
in an obvious way, and with some analysis of data taken in or less, obvious), not on memorizing a formula (which might
the laboratory, an intuitively obvious model explaining bulk be complicated). The mathematical level required for a basic
formation conductivity emerges. The model is not the same understanding of the electric properties of rocks is at the level
as the Archie model. However, when certain parameters are of secondary school algebra including the transformations
taken to their limiting values, and the model is converted of powers to roots (and vice versa), and the relationships
1
Texas A&M University, Department of Petroleum Engineering, david_kennedy@qed-petrophysics.com; david_kennedy@tamu.edu
2
Ecopetrol S.A., Carrera 7 #32-42, 7th Floor, San Martin Building, Bogotá - Colombia; fredyru.garcia@ecopetrol.com.co

between logarithms and exponents. Archie’s Model

The literature, almost without exception, introduces Archie’s model is based upon trends de¿ned by
the topic of formation resistivity with some variant of G.E. resistivity observations that he recorded in two ¿gures in his
Archie’s model for water saturation. That is, the topic is 1942 paper (see Fig. 1). These equations describe the trends
introduced by citing relationships, such as suggested by Archie’s own, and four additional, datasets.
Archie discovered that for the rocks which he studied the
ratio of a rock’s bulk resistivity, R0, to its interstitial brine
or resistivity, Rw, is a constant, F, an invariant property of the
rock. That is, R0/Rw = F; Archie expresses this as R0 = FRw
in Archie Eq. (1). Archie dubs F the “formation resistivity
with little or no preamble and little or no discussion of the factor.” The formation resistivity factor for different rock
provenance of the formulas. This is to put the end of Archie’s samples varies; Archie found by inspection of his plot
own chain of reasoning at the beginning of an explanation. of logarithm of F versus logarithm of porosity, ĳ, that the
Archie himself eschews any discussion of physics or curve variation of F with porosity is described by the trend F = 1/ĳm
¿tting, but he at least describes the results of his trend analysis (Archie’s Fig. 2 and Archie‘s Eq. (2)). Equating F in the
of the data that led him to this formulation by beginning at de¿nition of formation resistivity factor with F from trend
the beginning. We shall follow Archie’s example. analysis, Archie discovered a relationship between bulk rock
Archie’s model is the oldest quantitative log interpretation resistivity and porosity, R0 = Rw ĳím (Archie’s Eq. (3)). (We
technique; its literature begins in the 1940s. So it is expected note for our readers that Archie uses ș to denote porosity;
that much has been written on resistivity interpretation since we substitute the modern convention and use ‫( ׋‬introduced
that time. As far as exposition is concerned two approaches by Wyllie and Rose) in its script form ĳ. Another notational
are followed: (1) the equations of the Archie model are difference is Archie’s use of R and S where the modern
written down and justi¿ed as empirical discoveries without conventions are Rt and Sw.)
further explanation, or (2) the explanation begins with a cube
or cylinder of conductive material and the Archie model is
shown to be consistent with Ohm’s law. Even the literature
that attempts to put the Archie model on a theoretical
footing restricts itself to the use of Ohm’s law in terms of
resistivity. The problem is that these methods do not explain
the quirkiness of the Archie model. As we shall show in
this article, the restriction of explanations to the “resistivity
domain” prevents these efforts from achieving conceptual
clarity. Suf¿ce to say, consulting the “usual” sources of
information on formation resistivity—porosity—water
saturation provides only a single, and incomplete, point of
view.
Archie’s 1942 seminal paper contains only six equations.
A tutorial purporting to elucidate the electrical behavior of
rocks should begin with an introduction to the Archie model
and these six equations, and we shall do so. However, we
urge the student to download the original Archie paper from
the SPE; its copyright has expired and it is available without
charge at OnePetro. Archie’s paper is often cited, but judging
by claims attributed to Archie that he never made, seldom
Fig. 1—Archie’s ¿gures de¿ning the trends F = 1/ĳm and I = 1/Swn that
read. In the interest of good scholarship, if you should ever are the basis of his model and much of modern formation evaluation
author an article that relies on Archie’s work, you should (modi¿ed from Archie, 1942)
review what he says before you cite him. If you rely on
formation evaluation folklore, or even the citations of other In addition, Archie analyzed four datasets available in
scholars, for information, you are likely to misattribute literature published in the years 1936 through 1938. This is
something to Archie. really an amazing contribution as the presentation of these

data in the four papers is very different; at ¿rst glance they successful. The more sophisticated question of how to derive
do not appear to be the same thing. Archie summarizes these the Archie law from ¿rst principles was not even asked for
data in a plot analogous to his formation resistivity factor- 40 years following Archie’s 1941 presentation.
porosity plot and bases his next observation on the trend Pabitra Sen was among the ¿rst, if not the ¿rst, to ask this
followed by the four datasets in this plot. Archie writes his question in the literature. His 1980 SPE conference paper,
fourth equation (which is really two equations), Sw = (R0/Rt )1/n The Dielectric and Conductivity Response of Sedimentary
and Rt = R0Swín (Archie’s Eq. (4)). The ratio Rt /R0 is now Rocks, includes a section entitled “Derivation of Archie’s law
known as the “resistivity index”, although that term does not from First Principles.” Although this indicates that, 39 years
appear in Archie’s paper; nor does an equation I = Rt /R0 (now following Archie’s introduction of resistivity interpretation,
used as the de¿nition of resistivity index), nor does I = Swín, an interest in this topic was emerging among researchers, the
explicitly appear. Archie’s contribution here is to discover “¿rst principle” that Sen refers to is an “effective-medium
that the resistivity index is predicted by a power law of water theory” formula that many researchers would not elevate to
saturation. Finally, noting that n = 2 approximately, Archie a “¿rst principle”, and the “Archie law,” which he derives
writes two additional versions his of Eq. (4) explicitly using has an exponent of 3/2; this is 1/2 too small to qualify as
a square root, viz., applying to most Archie reservoir rocks with exponents
close to 2.
To make progress in intuitively understanding the
(Archie’s Eq. (5) and (Archie’s Eq. (6)). physical basis of the Archie model the resistivity formulation
of the problem must be abandoned in favor of a formulation
based upon conductivity. This might sound like a trivial
That is the full explanation of Archie’s contribution as he change in point of view, but it turns out to make this
introduced it in 1941. It seems an obvious step given that seemingly intractable problem as simple as it should be.
he has written down Archie Eq. (4) as Rt = R0 Sw-n to then A ¿nal word is to be said before leaving the discussion
substitute Archie Eq. (3) R0 = Rwĳ-m to produce of Archie’s model. Our intention here is not to disparage
Archie or his model. His model is robust for the rocks where
, it is applicable and has been in use every day for over 75
years, including by your authors. Few artifacts or ideas from
1941 are still in use without modi¿cation today; Archie’s
the combined equation probably most cited today as the law is a survivor and as a practical matter is unlikely to
Archie “law”; but, it is a step Archie himself did not take. be supplanted any time soon, if ever. However, in order to
It is worth noting that Archie did not coin, and does really understand why it works, it is necessary to step out of
not use, the misnamed “cementation exponent” nor does he the “resistivity domain” and into the “conductivity domain”.
introduce the factor a, (also poorly named) now called the We shall now take this step.
“tortuosity” factor. These misnomers are later inventions of
others. Finally, Archie relies entirely upon trend analysis for Conductivity in Brine-Saturated Archie Rocks
his discovery. Archie offers neither physical explanations, The paragraphs to follow will show that it is possible
nor derivations from ¿rst principles. The questions raised to predict the electrical behavior of a certain class of rocks
in the minds of many who would study and use the Archie known as “Archie rocks” from a set of (perhaps arguable)
model, such as “why is it a power law?”, and “why are the self-evident ¿rst principles. Due to its empirical nature,
exponents approximately equal to 2?” are not even asked, let Archie’s model best describes only these Archie rocks;
alone answered. so, although the log-derived Àuid saturation model under
Archie’s presentation to the AIME in Dallas was in the this approximation is uncertain in many cases, it has been
fall of 1941; by the time his paper appeared in 1942, the historically applied to almost every kind of rock, with mixed
United States was at war and civilian research was largely success. This method of evaluation has been universally
moribund during the war years. When research resumed accepted mainly due to a lack of serviceable alternatives.
following the war, at least two publications attempted to The attributes of Archie and non-Archie rocks are listed in
explain the Archie law in terms of circuit theory principles Table 1.
framed in terms of resistance, but these were not really

Table 1—Effect of Pore Geometry on Electrical Rock Type
wet
Moldic
Archie formulated his analysis in terms of resistivity. These principles combine into a “triple-product theorem”
This was a natural consequence of well logs being recorded which states:
in terms of resistivity, and measurements on core plugs
also being reported in terms of resistivity. However, and . (1)
unfortunately, the resistivity formulation of the electrical
behavior of rocks does not lend itself to an intuitive grasp of This could be taken as a reasonable, almost self-evident,
the relationships involved. working hypothesis. However, we call this a “theorem”
The empirical nature of Archie’s equation and any because it is capable of proof (or is falsi¿able if the proof
ambiguity associated with it can be avoided by using a can be found to be in error). The proof is:
relatively simple, straightforward and general approach Proposition (I) is taken as axiomatically self-evident.
to understanding the conductivity of rocks. It is founded If the conductivity of pore-¿lling brine in an Archie rock
upon an analysis of the geometrical con¿guration of the is doubled (or halved), the bulk conductivity of the rock is
conducting brine in the rock. Based on this premise, the concomitantly doubled (or halved). The same would be true
internal architecture of the rock can be explained by the for any fractional change in brine conductivity producing the
introduction of an explicit geometrical factor, E0. same fractional change in bulk rock conductivity.
Reformulated in terms of conductivity, the relationships Proposition (II) requires a demonstration of its meaning
become, if not trivial, at least possible to grasp with a for clari¿cation, especially for readers conditioned by prior
minimum of mental effort. We begin by stating our premises. exposure to the Archie model to regard bulk conductivity as
A brine-saturated Archie rock comprises two phases: being approximately proportional to ĳ2. Consider the model
nonconducting mineral grains and conducting brine with core plug containing the single sinuous channel, as shown in
conductivity ıw. By de¿nition, for Archie rocks, all of the the upper panel of Fig. 2. The core plug will exhibit a bulk
bulk rock conduction is due to the conducting brine in the conductivity. Next, consider the similar core plug in the lower
void space between the mineral grains. panel which is identical to the plug in the upper panel except
We now assert three principles regarding the bulk that the sinuous channel is duplicated. The dimensions and
conductivity of an Archie rock. The conductivity model rests shape of the second sinuous channel are exactly the same as
upon these principles. the dimensions and shape of the ¿rst. This is the meaning
of “constant pore geometry.” Obviously, the porosity and
The bulk conductivity, ı0, of an Archie rock (or similar bulk conductivity of the plug in the lower panel are double
medium) is proportional to: the porosity and bulk conductivity of the plug in the upper
I. Conductivity of the brine ıw, ı0 v ıw panel. The reader might well object that these plugs are very
II. Amount (i.e., fractional volume) of brine (for constant “unrocklike” in their geometry. To answer this objection, the
¿ve-core-plug model of Fig. 3 is introduced, and explained
pore geometry) ĳ, ı0 v ĳ
in its caption. The conclusion with this more rocklike model
III. Geometry of brine as expressed by a function of pore is the same as with the sinuous channel model, and also
geometry E0, ı 0 v E0 validates Proposition (II).

tube extending end-to-end through the core plug, in which

case the bulk conductivity is a maximum (equal to ımax = ıwĳ
as will be shown below). For distributions of brine in an
(actual) Archie rock, a more detailed development will lead
to a functional form for the geometrical factor E0.
In the ¿rst place, the reader will note that bulk rock
conductivity cannot be less than zero or more than the
maximum conductivity permitted by the amount of brine
available (i.e., ımax = ıwĳ). This means that the range allowed
for E0 must be 0 d E0 d 1. Since a detailed description of a
bulk rock sample containing tens of thousands to millions
of pores and pore throats will be forever intractable, direct
computation of E0 except, perhaps, for very small samples,
is impractical. However, note that all of the quantities in the
triple-product formula except E0 are measurable, and that
when Eq. 1 is solved for E0, it is computable by means of
Fig. 2—In the upper panel a core plug is modeled as a cylinder with
the plug outer dimensions area A and length l, and the void space, or . (2)
porosity contained in a single sinuous tube of constant cross-sectional
area A1 and length l1. In the lower panel the sinuous tube is duplicated,
which doubles both the porosity and the bulk conductivity of the core
plug while the pore geometry is held constant. So, E0 is computable for any single real core plug. Since
ı 0 can vary from 0 to ı max = ı wĳ, the limiting conditions
0 d E0 d 1 are satis¿ed.
E0 is the ratio of a rock’s actual conductivity, ı0, to its
theoretical maximum conductivity if all of its brine were to
be collected into a single through-going cylindrical tube,
ıwĳ. This ratio thus measures the “electrical ef¿ciency” of
the brine phase of the rock and the notation for it is chosen
partly in view of this fact.
From Eq. 2 an interesting observation follows. We have
ı0/ıw = ĳE0=1/F or, equivalently F=1/(ĳE0) = (1/ĳ).(1/E0). This
shows that a plot of F versus ĳ is equivalent to the product
of two plots: a plot of 1/ı versus ĳ multiplied by a plot of 1/E0
versus ĳ. Plotting 1/ĳ versus ĳ on log-log paper will result
in a straight line with slope í1. This amounts to a built-in
correlation in every F versus ĳ plot. Obviously, there is no
information regarding the rock in a 1/ĳ versus ĳ plot, so it
Fig. 3—Consider a model based upon two quarter-cylinders cut from
is the remaining plot of E0 versus ĳ that contains all of the
two core plugs taken in Archie rocks, one with zero porosity and the
other with 30% porosity. Next imagine that these two quarter-cylinders pore geometric information. Since the slope of the product
are replicated so that there are four of each kind. The pore systems F versus ĳ is approximately í2, this suggests that the slope
within the quarter cylinders are the pore systems of the original rock. of 1/E0 versus ĳ will also be approximately í1, or E0 | ĳ. In
Now imagine putting these cylinders together in ¿ve con¿gurations
having composite porosities of 0, 7.5, 15, 22.5, and 30% porosity. Our other words, to eliminate the inÀuence of porosity from F,
claim is that the conductivities of these rocklike cylinders will vary in it must be multiplied by ĳ. That is to say, a plot of ĳF = 1/E0
direct proportion to (i.e., to the ¿rst power of) the porosity. versus ĳ would remove the inÀuence of porosity from the
plot, leaving only the inÀuence of pore geometry. It is the
built-in correlation of 1/ĳ versus ĳ, together with 1/E0 | 1/ĳ,
Proposition (III) is also trivially self-evident if that produces m | 2; i.e., about half of the slope value m
“unrocklike” geometrical distributions of the void-¿lling (when m | 2) is due to correlating 1/ĳ with ĳ!
brine are permitted. The limiting cases are two: there is no Since E0 is a pore geometric factor, to proceed further
end-to-end connection of the brine, in which case the bulk it must be determined how E0 varies with porosity. Since in
rock conductivity is zero; or, the brine occupies a cylindrical real rocks (as opposed to models; e.g., the sinuous tube) pore

geometry and porosity must vary concomitantly, we propose To summarize, we have the triple-product theorem
to observe how E0 varies with porosity. Conductivity of the ı0 = ıwĳE0 and the observation that E0=a0ĳ + b0. Combining
bulk rock should vary with cross-sectional area of the void these results yields ı0 = ıwĳ(a0ĳ + b0)= ıw (a0ĳ2 + b0ĳ) | ıwĳ2.
space perpendicular to current Àow direction. Expressed as For comparison, Archie’s model in conductivity terms
a fraction of the total cross-sectional area of a rock sample, expresses this relationship as ı0 = ıwĳm with m | 2. Archie’s
this is proportional to porosity. Given an ensemble of core model is based upon a trend analysis of resistivity-porosity
plugs over a range of porosity, a correlation of the pore data uninformed by any physics. We have shown that,
geometrical factor and porosity can be obtained. To illustrate expressed in conductivity terms and speci¿cally accounting
this process, we apply the suggested procedure to Archie’s for a geometrical factor at the outset, a formula similar to
Nacatoch sandstone data. The results are shown in Fig. 4. the Archie model emerges from three proportionalities that
bulk rock conductivity must obey, and the discovery by
observation the relationship between pore geometry and
porosity is linear in Archie rocks. With this new insight, it
is readily seen why Archie’s m must | 2 (i.e., a0ĳ2 + b0ĳ Ł ĳm
where the single adjustable parameter m must account for
the same variation found in the two parameters a0 | 1 and
b0 | 0 with the exponent on the a0 term being exactly 2). If
that is considered a mystery solved, it introduces the new
mystery of why a0 | 1 and b0 | 0. This question is further
elaborated in the discussion section.
We have already mentioned that
, or . (3)
This illustrates a de¿ciency of the Archie model in

that the formation resistivity factor combines the effects of
pore geometry and porosity. In sympathy with Archie and
Fig. 4—The data in this ¿gure is from Archie’s Fig. (2) (see Fig. 1 in other pioneers working with a resistivity formulation for
this tutorial) for the Nacatoch sandstone. It is converted to E0 using
the electrical properties of rocks, R0 = FRw seems to be a
Eq. 2; E0 = 1/(Fĳ). Note that the trend is linear, and that the slope is
approximately 1 and the intercept approximately zero. Also note that perfectly rational approach to the problem. However, in this
the three points “northwest” of the main grouping of data are somehow formulation it is not obvious how F can be broken into a
different from the other data, probably taken in a different kind of rock, factor dependent upon porosity and another factor dependent
perhaps with fractures parallel to the direction of current Àow. If these
three points (out of 74) are omitted, the slope of trend is even closer to on pore geometry. Indeed, nothing in the resistivity domain
1 (1.0336) and the intercept closer to zero (í0.0112) with a coef¿cient of even prompts the question. However, formulated in the
determination R2 = 0.7633. conductivity domain in terms of our three ¿rst principles,
porosity and pore geometry are introduced as separate factors
a priori, before any data is examined. Thus ı0 = ıwĳE0 and it
From the results in Fig. 4, E0 is correlated to porosity by is noticed subsequently that a formation conductivity factor
E0=a0ĳ + b0. Note that a0 | 1 and b0 | 0. These approximate can be de¿ned as f = ĳE0 = 1/F in terms of reciprocal F. So,
values for a0 and b0 hold not only for Archie’s Nacatoch it is recognized that ı0 = fıw in the conductivity domain is
sandstone dataset, but are present in all of the Archie rock the equivalent of R0 = FRw in the resistivity domain. Viewed
datasets observed to date. from the resistivity domain it is hard to see how the leap to
Our plot indicates that E0 approximately equals porosity. an explicit pore geometry factor can be made, and indeed, it
This is interesting since it establishes that, for example, never has been.
for a rock specimen with 20% porosity only about 20% of As a ¿nal observation, writing the Archie model in
the bulk porous medium conducts most of the electrical terms of conductivity, ı0 = ıwĳm , we know that the left
current, regardless that in an Archie rock the total porosity side of the formula (i.e., bulk conductivity) depends upon
is interconnected. With increasing porosity transmissibility pore geometry. However, on the right side neither ıw nor ĳ
increases and tortuosity decreases making the rock an depend upon pore geometry. By process of elimination, the
increasingly ef¿cient conductor. pore geometry must therefore be contained in m. This has

been known for a long time, but it is not quite the whole Substituting this observed trend into the triple-product
story. Factoring the Archie model into the triple product theorem, ıt = ıw (Swĳ)Et , yields
theorem terms, ı0 = ıwĳm = ıwĳĳm-1 = ıwĳE0. The term-by-
term comparison sets the geometrical factor to E0 = ĳm-1 ; (7)
suggesting that there is more to the geometrical factor in the
Archie model than m alone can account for.
this beautiful quadratic equation in Swĳ has Àowed naturally
To summarize the main idea introduced to this point, the
from our three ¿rst principles. The solution for Sw is
bulk conductivity of a brine-saturated Archie rock depends
elementary:
only on properties of the brine. Bulk rock conductivity is
proportional to: (I) conductivity of the brine; (II) amount
. (8)
(or fractional volume) of the brine; (III) the geometrical
con¿guration of the brine determined by pore geometry.
Obviously, pore geometry must vary as porosity varies. It is
In the case of Eq. 6, from observation it is not necessarily
observed that the postulated geometrical factor is correlated
true that Įt | 1 and ȕt | 0. However, assuming that this is so,
with porosity, and expressible as a linear function of porosity.
the equation reduces to
Conductivity in Hydrocarbon-Bearing Archie Rocks
, (9)
Archie rocks are water-wet. When (nonconducting)
hydrocarbon displaces brine in an Archie rock, it moves
into the “centers” of the pores and pore throats. A ¿lm of which is the Archie model with m = n = 2. This article began
conducting brine will remain on the mineral grain surfaces with this same equation and the observation that the Archie
in the rock. The things that change with the emplacement of model for the electrical behavior of rocks is often introduced
hydrocarbons are the amount of brine (which is reduced) and using this equation, and the further observation that no
the geometry of the brine (which presumably becomes more physical explanation for the relationship is forthcoming
complicated or, at least, is changed). The reduced amount in those presentations. Even when the Archie model is
of brine is accounted for using a coef¿cient, Sw, to express introduced as
the reduction from the maximum amount of brine, ĳ, so Swĳ
gives the fraction of brine in the bulk rock. The change in , (10)
brine geometry is accounted for by an additional geometrical
factor, et. Thus, ı0 = ıwĳE0 is modi¿ed to
there is no physical explanation as to why the conducting
, phase, Swĳ should be broken into two factors and each given
(4) its own, potentially different, exponent. We can ask how,
physically, would this make sense?
where Et = E0et gives the geometrical factor for the partially Of course, the actual explanation of the need for, and
hydrocarbon-saturated Archie rock. Note that this retains use of, two distinct exponents is that the porosity exponent,
the form of, indeed is a more general form for, the triple- m, in the Archie model is determined in an experiment
product theorem. No new proof or additional argumentation using a number of brine-saturated core plugs in one plot,
is required. The amount of brine is reduced and its geometry while the saturation exponent, n, is determined in a different
is changed, but that is all. In this case ıt, ıw, and Swĳ are experiment (or a series of experiments, one for each core
known from measurements. Et is determined from plug) where brine saturation is varied from Sw = 1.0 to values
as low as possible in the experiment. The observation that for
. (5) many of these determinations that frequently m | n | 2 is fairly
strong evidence in favor of the triple-product theorem’s
correctness since it accounts for the observation in a natural
When Et is determined for an ensemble of core plugs way.
over a range of Swĳ, the trend suggested has always been As you might imagine, there is a great deal of additional
observed to be linear, described by interesting and informative and perhaps controversial
discussion that could follow on this topic. However,
, (6) our object of providing a basic introduction to electrical
properties of rocks has been met. This article is aimed at
new students of formation evaluation. In their mind’s eye,

if the student can visualize that the bulk rock conductivity resistivity of the material formed in a cylinder or prism of
depends upon three properties of the interstitial brine, its length l and cross-sectional area A. Ohm published in 1827
conductivity, its amount, and its geometry (this much is (100 years before the ¿rst well log was acquired). However,
pure logic) and also remember that it is observed that the to determine ȡ for a material one ¿rst constructs the cylinder
geometrical factor is approximately equal to the fractional of known geometry and then measures V and I with a
brine volume, then it is easy to write out voltmeter and an ammeter from which follows ȡ = V/I × A/l.
The point is that use of resistance and resistivity is an
, (11) accident of history, the result of an arbitrary choice, and is in
no way fundamental. In an alternative universe Ohm might
have written
from which, by elementary algebra, distributing the exponent
and substituting the Archie parameters for 2, then
, (14)
. (12)
where g is the electrical engineer’s notation for conductance
and ı is the physicist’s notation for conductivity. Referring
recovering Archie’s resistivity model. Then follow all of to Fig. 5, for the core plug having volume ߥ =Al containing
Archie’s results; viz.: the single sinuous tube pore of cross-sectional area A1 and
length l1, the measured voltage drop and current would be
. (13) exactly the same for both cases; only the dimensions of the
conductor vary. So, equating the conductance of the bulk
medium (i.e., the core plug) and the sinuous channel, g0 = g1
and substituting the conductivity expressions, then
We shall continue in the next section by showing that
the results above are consistent with more conventional
methods of derivation. , (15)
Triple-Product Theorem from Ohm’s Law and Circuit

and this is solved for the equivalent conductivity of the core
Theory
plug as
In its role as a tutorial targeted at novice formation
evaluators it will be interesting to visit some topics of
. (16)
perennial interest in petrophysics and formation evaluation.
Let us begin with a derivation of the triple-product formula
from Ohm’s law. In our derivation using Ohm’s law, in
deference to the customary notation used in electrical
engineering, we shall let R denote resistance and ȡ denote
resistivity. Ohm formulated his law in terms of electrical
circuits. His voltage sources were bimetal junction
thermocouples of known potential; he measured current with
a galvanometer. His circuit elements were “standard” wires
of ¿xed material and diameter whose resistance varied with
the length of the wire. He announced his result in a formula
relating current to the ratio of voltage and resistance; i.e.,
I = V/R in modern notation. A more modern formulation is Fig. 5—The triple-product model can be derived from Ohm’s law by
R = V/I which indicates that the current through a circuit equating the conductance of the sinuous channel to the conductance of
element known as a resistor adjusts in reaction to the applied the model plug, substituting for conductivity and geometry, and solving
voltage so that the ratio remains a constant, R, dependent on for the model plug equivalent conductivity and grouping terms into brine
conductivity, fractional brine volume, and geometric factors.
the electrical properties and geometry of the material. This
relationship is typically expressed as R = ȡl/A where ȡ is the

The quantity in the denominator is recognized as the and Rose,1950; Winsauer et al., 1952). Putting all of the
de¿nition of tortuosity, ߬. In a rock-model where the void reduction in conductance in the tortuosity term tends to
space is collected into a single through-going cylinder, the give it a misleadingly large value. However, there is ample
porosity is expressible as ĳ = A1/A where A1 is the area of the evidence that tortuosity in Archie rocks should not be, and
through-going cylinder, or A1 = ĳA. In actual rocks the area in fact is not, a large number but in the neighborhood of 2 or
exposed in cross section is not equal to the area of the cross less. Thus if
section of the through-going cylinder, but is proportional to
it since the exposed void space in a random cross section of .
(22)
an Archie rock will be less than the area of the through-going
cylinder. This is expressed as A1 =ĮĳA where 0 Į 1. We then the transmissibility is on the order of ߬ĳ.
call Į transmissibility. Thus To conclude this discussion, we make good on our
promise to show that ımax = ıw ĳ. We noted in our derivation
, (17) of the triple-product theorem from Ohm’s law that
ı 0 = (A1/A/l1/l)ıw where A1 is the cross-sectional area of a
which upon cancelation of the A terms, and grouping the sinuous tube of length l1 embedded in a cylinder of cross-
other terms becomes sectional area A and length l. The denominator represents
tortuosity, a number ߬ > 1; however, in the limit l1 o l, that
. (18) is, as the sinuous tube becomes straight, then ߬ o 1 and the
right side approaches A1/A Ł ĳ (A1/A)/(l1/l). Thus, ı0ĺĳıw
is the maximum value that the bulk conductivity can assume.
Note that this expression has the form of the triple-product
theorem if the geometrical factor is de¿ned as Connections to Percolation Theory
Solid-state physicists commenced the study of
. (19) conductivity of alloys and mixtures of materials at the
atomic level in support of the development of transistors
and integrated circuits in the 1950s. In mixtures of
Since E0 is by de¿nition a geometrical factor, and since
equidimensional conductive and nonconductive spheres they
Į/߬ is a function of ratios of areas and lengths, it seems the
found there is a de¿nite ratio of conducting to nonconducting
requirements of the triple-product theorem are met. Thus
spheres at which conduction commences. They dubbed the
value of this ratio the “percolation threshold.” For spheres,
, (20) conduction begins when the ratio of conducting spheres to
total spheres is, roughly, 1:3. In the 1970s geophysicists
or reordering the factors on the right side, ı0 = ıwĳE0. The studying the conductivity of the whole earth applied
triple-product theorem is thus consistent with (and can be percolation theory to geophysical systems, ¿rst publishing
derived using) Ohm’s law from circuit theory. the formula
The form E0 = Į/߬ invites some discussion. The
factorHFJDKKAS, where GHK is the average tortuosity , (23)
of the medium, by its de¿nition is d 1. The parameter
Į describes the “transmissibility” of the pore system to
electrical current, and is also a number d 1. Thus, E0 is d 1. where r | 2. (We have modi¿ed the notation from the original
However, we also have from observation that E0 = a0 ĳ + b0. article to conform with our notation.) The parameter ĳ‫ ׇ‬is
Thus the percolation threshold parameter. By de¿nition ı0 /ıw = 0
for ĳ < ĳ‫ ;ׇ‬in other words, the bulk conductivity of the rock
. (21) is zero when the porosity is less than a critical porosity, or
percolation threshold. Percolation thresholds in rocks are
much lower, approaching zero, than in sphere mixtures
In most formulations all of the reduction in conductivity and other models studied by physicists, (e.g., cubic resistor
(increase in resistivity) due to the presence of mineral grains lattices). We observe that Eq. 23 is not formally derived from
in the brine has been attributed to the tortuosity term; the ¿rst principles, but rather an empirical model consistent with
transmissibility term has been ignored even in publications the data presented in the source article and its references.
where its existence is explicitly acknowledged (e.g., Wyllie Combining Eq. 1 and E0 = a0ĳ + b0 gives

ı0 /ıw = ĳE0 = ĳ(a0ĳ + b0). We now apply boundary theory requires plotting observations to determine a0 and b0.
conditions to this formula. The boundaries are the percolation Practical limits of porosity for conventional reservoir rocks
threshold where bulk conductivity is zero, and 100% porosity are roughly from to 10 to 40% porosity. Extrapolation from
where bulk conductivity equals ıw and ı0 /ıw = 1. We call this range to minimum (e.g., zero) conductivity on the left
these limits the left and right boundaries, respectively. Begin and maximum conductivity on the right requires imposition
by factoring a0 from the parenthesis: of a functional form for the ¿t that may not be capable of
¿tting observations as well as if no boundary conditions are
. (24) imposed. The paper by Kennedy (2016) is devoted to this
issue.
Noting that at the left boundary ı0 /ıw = 0 when ĳ = íb0/a0 , Stepping Beyond Archie Rocks: Parallel Conduction
then b0/a0 = ĳ‫ ׇ‬by de¿nition. At the right boundary Model
Possibly the ¿rst use of a conductivity formulation for
, (25) describing the electrical properties of rocks is made in the
paper Electrical Conductivities in Oil-Bearing Shaly Sands
by M. Waxman and L.J.M. Smits (1968). They propose to
So model a “shaly sand” as two resistors (with resistances r1
and r2) in parallel. In terms of resistance, the equivalent
. resistance r0 of the two resistors would be
(26)
, (28)
Thus
or
. (27)
. (29)
In this form, the geometrical-factor theory and triple-

product theorem are seen to be consistent with percolation
Expressed in term of conductance where g = r í1 the
theory, and indeed E0 is seen to be de¿nable in terms of a
corresponding relationship is
percolation threshold. However, comparison of Eqs. 23 and
27 reveals a difference in that Eq. 23 has (approximately)
two factors (ĳ í ĳ‫)ׇ‬/(1 í ĳ‫ )ׇ‬whereas Eq. 27 has only one. . (30)
Since we are interested in derivations from ¿rst principles,
this difference leads us to re-examine our list of three ¿rst
As you can see, Eq. 30 is considerably simpler than Eq.
principles. We notice that the second principle is not quite
29. We saw in Eq. 14 that conductances and conductivities
correct: it should be stated in terms of ĳ í ĳ‫ׇ‬, that is:
are connected by g = ı(A/l). Making this substitution
The bulk conductivity of an Archie rock (or similar medium),
ı0, is proportional to: . (31)
II. amount of connected and conducting brine (for constant

pore geometry) ĳc = (ĳ í ĳ‫ ;)ׇ‬i.e., ı0 v ĳc . This system of conductors is illustrated in Fig. 6. A brief
digression into the notation of the Waxman-Smits article
This would also change the computation of the will be helpful for our purposes here and for those readers
geometrical factor to Ec = (ı0/ıw) / (ĳ о ĳ‫)ׇ‬. who may read the Waxman-Smits paper. In the notation
We shall not pursue the implications of this correction used in Waxman-Smits, Eq. 30 for conductance is rendered
to our geometrical-factor theory further in this venue. Crock = Cc+ Cel (W-S Eq. (1)) where the c subscript denotes
Suf¿ce to say that any practical consequences would be clay conductance and the el subscript denotes electrolyte
minimal due to the small magnitude of the percolation conductance. The solution of Eq. 31 for ı0 is written as
threshold parameter. Practical use of the geometrical factor C0 = xCe + yCw (W-S Eq. (2)) where x and y are called

“appropriate geometric constants.” The e subscript denotes or in Waxman and Smits terms
cation exchange conductivity and the w subscript denotes
brine conductivity. (Note that in the Waxman-Smits notation , (36)
both conductance and conductivity are (confusingly)
denoted by C; more conventional notation uses the separate where the “appropriate geometrical factors” of W-S are
symbols g and ı, respectively, for these different quantities.) identi¿ed with x = (Į1/߬1)ĳ1 and y = (Į2/߬2)(ĳ í ĳ1).
However, Waxman and Smits do not provide a derivation Although we shall not follow them there, Waxman and
of the constants x and y. They just assert that there are such Smits then assert “We assume next that the electric current
constants. We shall now show that a form for these constants transported by the counterions associated with the clay
is readily derived in the parallel conduction model. travels along the same tortuous path as the current attributed
to the ions in the pore water,” and by this assumption they
set x = y = (F*)í1, where by F* they mean “the shaly sand
formation resistivity factor”; presumably this means
. (37)
Re = 1/Ce denotes exchange cation resistivity; this could vary

from a low value up to in¿nity (e.g., if the clay volume is zero,
although a problem with the Waxman-Smits model is that
it lacks an explicit volumetric partitioning of porosity into
an interstitial brine volume and a cation exchange, or clay,
Fig. 6—A cartoon representation of the parallel conduction model where
the channels have both different geometries and different conductivities. volume). In the limit of zero clay volume F* = F; otherwise
F* > F. According to their assumptions, the Waxman-Smits
model will be valid when (Įe/߬ e)ĳe = (Įw /߬w)(ĳ í ĳe) = (F*)-1
Performing the algebra to isolate the bulk conductivity on is satis¿ed. In the case of (Įe /߬ e) = (Įw /߬w) or Ee = E w (i.e.,
the left side of the equation in Eq. 31, then “the same tortuous path”) then ĳe = ĳw and the porosities
are implicitly equally apportioned between the conducting
, (32) phases. Otherwise (i.e., if x z y), then Ew/Ee = ĳe/ĳw or
ĳe = Ew/Eeĳw meaning if Ee > Ew, then ĳe < ĳw which might
seem to be the case for an authigenic clay coating grains in
and noting that the fractional cross-sectional areas are a predominately brine-¿lled void space. The Waxman-Smits
not equal to, but less than, the porosities of the individual requirement that x = y = 1/F* places severe constraints upon
channels (i.e., Ai=Įi௒ĳiA where 0 Įi 1) by the same the rocks where it can be applied with accuracy.
arguments as used to justify Eq. 17 above,
CONCLUSIONS
, (33)
Discussion
The cornerstone of the Archie model is the de¿nition of
where we note that the total porosity of the plug is sum of
the formation resistivity factor, F = R0 /Rw. This answers
the individual channel porosities, so that ĳ2 = ĳ – ĳ1. Then,
the question of how bulk rock resistivity depends upon a
canceling the area ratios and grouping the terms, gives
particular rock and its interstitial brine resistivity, R0 = FRw;
the model connects brine resistivity to bulk rock resistivity
, (34) through a rock-dependent factor but is silent as to how
brine volume and brine geometry separately inÀuence R0.
We must resort to the laboratory to discover the connection.
and noting that the ratios contain only geometrical terms, However, even the observation that F = ĳím from the lab
then Ei = Įi /Ĳi and does not explicitly break down into a volume fraction and a
geometrical factor.
, (35)

In contrast, the geometrical factor representation focuses speculate that since the cross-sectional area normal to current
on how the bulk rock conductivity is directly proportional to Àow is one of the main controllers of how much current is
the three properties of the conducting phase: conductivity, passed, to the degree that effective tortuosity approaches 1,
fractional volume, and geometry; i.e., ı0 = ıwĳE0, our triple- then transmissivity approaches porosity in value. This may
product theorem. Comparison to the Archie model expressed be evidence of our claim that tortuosity in fact is a number
in conductivity terms, ı0 = ıw/F reveals that 1/F = ĳE0 approaching 1 for Archie rocks.
which is obviously a formation conductivity factor; i.e., The development above does not take into account
ı0 = fıw. Then f, and thus 1/F, is seen to be the product of formation conductivity (or resistivity) anisotropy. Real rocks
the two factors, porosity and geometric factor. It is true that are almost invariably anisotropic, and the triple-product
we must still retire to the laboratory to discover a correlation theorem is readily extensible to anisotropic rocks. Archie’s
between E0 and ĳ; however, E0 is an explicit geometrical model can also be extended to anisotropic rocks. We will
factor directly correlated to porosity. develop anisotropy in the next article of this tutorial.
The correlation of the geometrical factor to porosity is
E0 = a0ĳ + b0. In contrast, it is not clear how to partition Concluding Thought
ĳm into a volume fraction and a geometrical factor. It would Our purpose in this article has been to introduce and
be hard to guess that the partition is ĳĳmí1, and indeed, the illuminate the physical processes responsible for conduction
problem was never solved in this way, but always by putting in rocks. We have framed the problem in three distinct ways:
the entire contribution from geometry into m. a derivation from three physical ¿rst principles; a derivation
We have shown that the triple-product theorem is from Ohm’s law in conductivity terms; a derivation from
consistent with Ohm’s law by deriving the theorem from the brine conductivity in terms of ion concentration (see
law. This formulation also leads to the expression of E0 as Appendix). All three methods lead to the triple-product
the ratio of transmissibility to tortuosity. In the appendix we theorem formula, which is then trivially transformed to the
offer a third derivation of the triple-product theorem in terms Archie model in certain limiting cases. Since the physical
of ionic conductivity. The triple-product theorem also gives ¿rst principles are very nearly self-evident, and lead to
explicit form to the geometrical factors used in the parallel formulas that produce the Archie model directly, there is
conductivity models employed for shaly sand interpretation. every reason to embrace them. Empiricism is still required
One of the mysterious aspects of the Archie model is the to determine the geometrical dependence, but in the triple-
closeness of m and n to 2. In terms of the Archie model derived product theorem formulation the dependence on geometry is
from empirical observations this is just accepted as what the explicit, not being shoved into parameters m and n because
data dictate. At ¿rst glance the triple-product theorem seems there is nowhere else to put the dependence. This ful¿ls our
to resolve this mystery, since ı0 and ıt are quadratic in ĳ and mission.
Swĳ, respectively, with dominant terms having exponents We have often paused to wonder why formation
exactly equal to 2. However, for the triple-product theorem evaluation has resisted progress when compared to other
the empirical step is to determine how E0 and Et depend upon 20th century technical developments. Heisenberg (1925)
ĳ and Swĳ, respectively, using crossplots similar to Archie’s and Schrodinger (1926) published the foundations of
use of crossplots to determine m and n. The resulting quantum mechanics just prior to 1927, the year that the ¿rst
correlations E0 = a0ĳ + b0 and Et = at Swĳ + bt , where a0 and well log was acquired in France. In the intervening 90 years
at | 1, and b0 and bt | 0 , have served only to shift the mystery quantum mechanics, invented to explain atomic physics,
from the Archie parameters to the geometrical factor theory has been applied secondly to nuclear physics and then to
parameters. The observations restrict the values of the a and the physics of protons and neutrons as particles comprised
b parameters to approximately 1 and 0, respectively, leading of still smaller particles, quarks. Neil Armstrong set foot on
to E0 | ĳ and Et | Swĳ ; substitution of these equivalencies the moon just 65 years following the ¿rst powered Àight.
into the triple-product theorem formulas reduces them to Given these examples (and there are many others) one may
the Archie model with m = n = 2. Our analysis provides an well wonder why in the 78 years since Archie announced his
answer to the question of the Archie exponents being nearly empirical model corresponding progress has not been made
equal to 2, but it does so by shifting the question to why the in formation evaluation theory. Is formation evaluation more
geometrical factors E0 and Et should be so nearly equal to the dif¿cult than physics’ “standard model” or the engineering
brine volumes (i.e.; ĳ and Swĳ) in Archie rocks. challenges of space Àight? Perhaps the answer lies in that in
As a ¿nal observation as to the relationships E0 = a0ĳ + b0 physics and the engineering of Àight thousands of scientists
and E0 = Į/߬ | ĳ and why the electrical ef¿ciency of the and engineers in hundreds of universities and government-
Archie rock should approximately equal the porosity, we can supported agencies and companies were engaged in

competitions to decipher the workings of matter, whether REFERENCES AND NOTES

of atoms or airfoils. Conversely, in formation evaluation
only a handful of thoughtful people, distributed over several Pioneering Papers
competing commercial organizations and distributed over Archie, G.E., 1942, The Electrical Resistivity Log as an Aid in
seven decades in time, were dedicated to thinking about Determining Some Reservoir Characteristics, Paper
conductivity in reservoir rocks. Archie’s model works SPE-942054-G, Transactions, AIME, 146, 54–62, DOI:
10.2118/942054-G. https://www.onepetro.org/journal-paper/
so well in so many cases managers had little incentive to
SPE-942054-G
allocate resources to what seemed to them to be a non-
problem. Moreover, management goals tended (and still Archie based his relationship between water saturation and
tend) to change annually, so only a very few lucky scientists resistivity index on data available in tables and graphs
(e.g., Wyllie at Gulf, and Archie at Shell) were able to devote found in the following four articles
years-long effort to the problem. Jakosky, J.J., Hopper, R.H., 1937, The Effect of
The thought processes of researchers were so entrenched Moisture on the Direct Current Resistivities
in the resistivity formulation that even when they were of Oil Sands and Rocks, Geophysics, 2(1), 33–54.
trying, they were not able to overcome the barrier posed DOI: 10.1190/1.1438064.
by the resistivity formulation of the problem, namely that Leverett, M.C., 1938, Flow of Oil-Water Mixtures
through Unconsolidated Sands, Paper SPE-
the formation resistivity factor is the product of (reciprocal)
939149-G, Transactions, AIME, 132, 149–171. DOI:
porosity and a (reciprocal) geometrical factor. Further, even 10.2118/939149-G.
though the role of brine cross-sectional area was recognized Martin, M., Murray, G.H., and Gillingham, W.J, 1938,
in print by some of the pioneers (Wyllie and Rose, 1950; Determination of the Potential Productivity of Oil-
Winsauer et al., 1952), their models focused wholly on Bearing Formations by Resistivity Measurements,
tortuosity to the exclusion of transmissibility. To the best of Geophysics, 3(3), 258–272. DOI: 10.1190/1.1439502.
our knowledge, Dr. David Herrick, at the Amoco research Wyckoff, R.D., and Botset, H.G., 1936, The Flow of
laboratory in the 1980s, was the ¿rst to completely rethink Gas-Liquid Mixtures Through Unconsolidated
the problem without reference to prior models. His thinking Sands, Physics, 7(9), 325–345. DOI: 10.1063/
1.1745402.
was the source of the ideas embodied in the geometrical
Guyod, H., 1952, Electric Well Logging Fundamentals, Part 12,
factor theory, and were independently con¿rmed by the in Fundamental Data for the Interpretation of Electric Logs,
work of Professor C.F. Berg in 2012. Well Instrument Developing Co., p. 76. (Probable original
It is our hope that this exposition will inspire its readers source is Oil Weekly, 115(38), Oct. 30, 1944.)
to “think out of the box” in terms of resistivity interpretation. This article is the source of the coinage “cementation
Although the subject here is mostly con¿ned to conventional exponent”.
Archie rocks, the shaly sand problem might be pro¿tably Owen, J. E., 1952, The Resistivity of a Fluid-Filled Porous Body,
revisited. The application of unconventional thinking to Paper SPE-952169-G, Journal of Petroleum Technology, 4(7),
unconventional resources is surely the way to progress in 169–174. DOI: 10.2118/952169-G.
the future. We hope that this article will trigger a renewed This article is the earliest reference that we are aware
our that uses a in the formula F = a/ĳm.
and vigorous, even if contentious, discussion of this topic. It
Sen, P.N., 1980, The Dielectric and Conductivity Response of
will only be through the disputation of ideas, new and old, Sedimentary Rocks, Paper SPE-9379 presented at the SPE
that will lead to progress and, perhaps, consensus among Annual Technical Conference and Exhibition, Dallas, Texas,
formation evaluators as to the form a new model will take. USA, 21–24 September. DOI: 10.2118/9379-MS.
Sen has a section entitled “Derivation of Archie’s Law
ACKNOWLEDGEMENTS from First Principles.” However, a more accurate title
for the section might have been “Derivation of a
The fundamental ideas developed in this article originated Resistivity Power Law from an Effective Medium
in the fertile brain of Dr. David Herrick. Recently and Approximation.”
Wyllie, M.R.J., and Rose, W.D., 1950, Some Theoretical
independently, our colleague Professor Carl Fredrik Berg
Considerations Related to the Quantitative Interpretations of
developed the identical concept with a rigor acceptable to the Physical Characteristics of Reservoir Rock from Electric
the physics community. This article would not have been Log Data, Paper SPE-950105-G, Journal of Petroleum
attempted without the invitation and encouragement of Technology, 2(4), 105-118. DOI: 10.2118/950105-G.
Professor Carlos Torres-Verdin. This paper contains the ¿rst attempt that we know of to
derive the Archie model from theoretical principles.
Interestingly, all of the necessary principles are indeed
discussed, in particular the role of cross-sectional area

normal to the direction of current Àow, and its 4863–4868. DOI: 10.1029/JB079i032p04863.
proportionality to porosity. Wyllie and Rose then Although the formula was never adopted into
proceed to not use this fact in their analysis and thereby petrophysical use, these authors were ¿rst to apply
miss the opportunity to have discovered the geometrical percolation theory percepts to conduction in rocks.
factor theory and triple product theorem. Perhaps their work went unnoticed in formation
Wyllie, M.R.J., 1952, Role of Clay in Well Log Interpretation, evaluation because they were geophysicists publishing
Clays and Clay Minerals, 1(1), 282–305. DOI: 10.1346/ in the Journal of Geophysical Research rather than
CCMN.1952.0010125. petroleum engineers or geologists publishing is SPE or
We include this paper as one of interest for the shaly AAPG venues.
sand problem because it indicates pretty clearly that
Wyllie understood all of the issues 16 years before Geometrical Factor Theory Papers
the Waxman-Smits publication appeared. Perhaps The exposition of the Geometrical Factor Theory, ¿rst published
arguably, his analysis is even more cogent when pore in 1993 and 1994, evolved over 18 years as its inventors realized
geometrical aspects are concerned. For example, its implications, eventually culminating in its formulation as the
Wyllie writes Cwa = Cf + Cw = Cf + Cc/F. followed by Triple-Product Theorem. Prof. C.F. Berg, a physicist then working
“It may be noted that in Eq. 6 Cs/Fƍ may be substituted at Statoil, independently and with mathematical rigor, reinvented
for C f.” In other words, Cwa = Cs /Fƍ+ Cc/F. “Here Cs is the GFT in 2012.
the actual conductivity of the conductive solids and Fƍ
their formation factor.” This is equivalent to Waxman’s Berg, C.F., 2012, Re-Examining Archie’s law: Conductance
and Smit’s Eq. 3 but with different formation factors Description by Tortuosity and Constriction, Physical Review
(i.e., pore geometries) for the two conductive phases. E, 86(4)., DOI:10.1103/PhysRevE.86.046314.
We can speculate that Wyllie carried this no further as Herrick, D.C. and Kennedy, W.D., 1993, Electrical Ef¿ciency: A
there is no obvious way to separately determine F and Pore Geometric Model for the Electrical Properties of Rocks,
Fƍ. Still, it would have been pro¿table to have opened a Paper HH, Transactions, SPWLA 34th Annual Logging
discussion of the issue. Symposium, Calgary, Canada, 13–16 June.
Winsauer W.O., Shearin, H.M., Masson P.H., and Williams, M., Herrick, D.C. and Kennedy, W.D., 1994, Electrical Ef¿ciency—A
1952, Resistivity of Brine-Saturated Sands in Relation to Pore Pore Geometric Theory for Interpreting the Electrical
Geometry, AAPG Bulletin, 36(2), 253–277. Properties of Reservoir Rocks, Geophysics, 59(6), 918–927.
This paper makes extensive reference to the role of DOI: 10.1190/1.1443651.
conductor cross-sectional area in discussion. It then Herrick, D.C. and Kennedy, W.D., 2009, A New Look at Electrical
fails to use it in its analysis. It is also the source of the Conduction in Porous Media: A Physical Description of Rock
“Humble formula”; i.e., F = 0.62 / ĳ2.15. Conductivity, Paper BB, Transactions, SPWLA 50th Annual
Waxman, M.H., and Smits L.J.M., 1968, Electrical Conductivities Logging Symposium, The Woodlands, Texas, 21–24 June.
in Oil-Bearing Sands, Paper SPE-1863-A, SPE Journal, 8(2), Kennedy, W.D., Herrick, D.C., 2012, Conductivity Models for
107–122. DOI: 10.2118-1863-A. Archie Rocks, Geophysics, 77(3), WA109-WA128. DOI:
This paper is, perhaps arguably, the seminal paper for 10.1190/geo2011-0297.1.
resistivity log interpretation in clay-bearing rocks. Kennedy, W. D., 2016, Conducting Connected Porosity: A
Its longevity is remarkable considering its derivation Concept for Unifying Resistivity-Porosity Models, Paper
contains very questionable assumptions regarding U, Transactions, SPWLA 57th Annual Logging Symposium,
pore geometry and its lack of volume-weighting for its Reykjavik, Iceland, 25–29 June.
parallel conductive components.
Clavier, C, Coates, G., Dumanoir, J., 1984, Theoretical Service Company Literature
and Experimental Bases for the Dual-Water Model for Schlumberger, 1972, Log Interpretation: Volume I – Principles.
Interpretation of Shaly Sands, Paper SPE-6859, SPE Journal, Notes: Chapter 1. Formation Factor and Porosity. Here we ¿nd
24(2), 153–168. DOI: 10.2118/6859-PA. the ¿rst mention of formation resistivity factor. There is no
This paper initially appeared in 1977 as a conference discussion of causality, but the article begins with “It has
paper and in a quite different version as a peer reviewed been established experimentally that the resistivity of a clean
paper. As far as its physics and electrochemistry is formation … is proportional to the resistivity of the brine
concerned, it is basically the same as in the Waxman- with which it is fully saturated.” A misattribution to Archie
Smits paper; however, these authors correct the follows, “Archie proposed the formula F = a/‫׋‬m where m
Waxman-Smits failure to partition the pore volume is the cementation factor.” In fact, Archie (1942) (and the
into interstitial brine and exchange cation brine. reference is cited) proposed only F = 1/‫׋‬m, without any use
Unfortunately, the Waxman-Smits assumption that the or mention of a, and Archie does not refer to a “cementation
geometry of the pores and the surfaces is the same is factor”. This latter coinage is from H. Guyod writing in 1944.
retained in the dual-water model. The earliest reference that we have found to a is in an article
Shankland, T.J., and Waff, H.S., 1974, Conductivity in Fluid- by Owens, 1952.
Bearing Rocks, Journal of Geophysical Research, 79(32), Water Saturation. In this section Archie’s Eqs. 4 and 5 are

listed. The only justi¿cation is that “Archie determined

experimentally …” Actually, Archie deduced his Eqs. 4 from .
the literature.
Invasion, Vertical Saturation Gradients and Vertical Fluid This is a parallel conductivity model where Vx is a fractional
Migration. There is a nice sketch of invasion pro¿les provided volume (or a term depending upon fractional volume) of a
on page 3. The accompanying discussion is in very general conductive phase identi¿ed with the x subscript, and C is a
terms. Some of its claims would be hard to follow for a person term “related to water saturation.” Translated into conductivity
without a reservoir engineer’s background in fractional Àuid terms gives
Àow.
,
Anisotropic Formations. Anisotropy is brieÀy mentioned but
attributed mainly to shale and mudcake. Laminated shale-
sand is mentioned as being anisotropic without elaboration. and substituting Waxman-Smits parameter notation, Cx = BQv,
Chapter 14. Resistivity Interpretation. Sw2/F = 1/G*, and C = F*/ Swn+1* = 1/Sw then
In spite of the title of this chapter, it is mainly about which
log should be used in which circumstances, and which charts ,
(in a separate Chart Book volume) should be employed in
interpretation. Step resistivity pro¿les are mentioned but their which is the Waxman-Smits formula except that it includes
limitations are not discussed except for the case of a resistivity volume weights for the two conductive phases.
annulus.
Chapter 15. Determination of Saturation (Clean Formations). There are several other service company publications equivalent
The Archie Formula. The section begins “All water saturation to the Schlumberger Principles volumes, including from Gerhart
determinations from resistivity logs in clean formations with Owens International, Welex, Halliburton, Atlas (in several forms:
homogenous intergranular porosity are based on Archie’s Dresser Atlas, Western Atlas, Baker Hughes, etc.). None of
formula or variations thereof.” Once again, there is no these publications depart from a conventional discussion of the
discussion of any formation resistivity (or conductivity) development of formation resistivity, and we do not review them
physics, but merely a recitation of formulas and the logs that separately for this summary.
should be obtained to have proper parameters to plug into the
formulas. Books: Petrophysical Theory and Tool Physics
Chapter 16. Shaly Formations Doveton, John H., 2014, Principles of Mathematical Petrophysics,
Introduction. The “Thomas-Stieber” model with its Laminated- International Association for Mathematical Geology Studies
Dispersed-Structural shale distribution is presented in picture in Mathematical Geology 9, Oxford University Press. ISBN:
form. The publication of Principles – Volume 1 proceeds the 978-0199978045.
Thomas-Stieber publication, so we assume that Thomas and Doveton’s Chapter 1 is a review of formation resistivity as
Stieber drew on the pictures in this volume in the production it developed historically, beginning with a formula from
of their analytical model. J.C. Maxwell’s Treatise on Electricity and Magnetism. He
Laminated Sand-Shale Simpli¿ed Model. This presents the does point out the distinction between the empirically-based
parallel conduction model for laminated sand and shale resistivity formulations and physics-based conductivity
formations as a formula. It is correct but there is no discussion formulations on the very ¿rst page of his book. In the
of how it arises from the physics of the medium. remainder of the chapter the resistivity models are introduced
Dispersed Shale Simpli¿ed Model, Shaliness and Cation more or less in chronological order of their publications. This
Exchange & Total Shale Model. Again, there is no discussion book, and its Chapter 1 provide an excellent place for novice
of conduction mechanisms in these sections, only formulas. formation evaluators to begin their continuing education on
The only “surviving” technique mentioned in these articles is formation resistivity methods.
the Waxman-Smits model. We have addressed the limitations Ellis, D.E., Singer, J.M., 2007, Well Logging for Earth Scientists,
of the Waxman-Smits parallel conduction model in our main 2nd Edition, Springer. ISBN: 978-1402037382.
article on parallel conduction. This book has three sections on resistivity interpretation
Schlumberger, 1989, Log Interpretation Principles/Applications. Chapter 3 (Basic Resistivity and Spontaneous Potential),
Notes: This is the most recent Schlumberger offering, now 30 years Chapter 4 (Empiricism: The Cornerstone of Interpretation),
old. Although the organization of the content has changed and and Chapter 23 (Saturation and Permeability Estimation).
been added to, much of it remains identical to the 1972 Log Equation (23.1) is Swn = Rw/(Rtĳím). These authors mention
Interpretation Volume I – Principles. Additions have been the triple-product theorem (Eq, 4.12) in their discussion in
made to the shaly sand resistivity/conductivity interpretation the section “Alternative Models.” This book should be on the
discussion. For example, “There are many formulas that shelf of every scientist planning to contribute, or contributing,
relate resistivity to water saturation in shaly sands. Most are to formation evaluation.
generally of the form: Hearst, J.R., Nelson, P.H., and Paillett, F.L., 2000, Well Logging
for Physical Properties, 2nd Edition, John Wiley & Sons, Ltd.
ISBN: 978-0471963059.

This book attempts to relate all of well-log interpretation, Unfortunately, much of the story of the historical development
insofar as possible, to the basic underlying physical of formation evaluation is not established fact, but based upon
principles. It also has three sections that discuss resistivity peoples’ notions of “how it had to have been” which in many
and its interpretation. First is a brief mention in the section cases is not “how it happened”. This is so common that Barbara
on Mixing Rules in the ¿rst chapter. Its Chapter 5 (Electrical Anderson (Scientist at Schlumberger-Doll Research, Editor of
and Magnetic Methods) provides (in 100 pages) a thorough Petrophysics (2004–2005), and President of SPWLA (1994–
discussion of hardware and interpretation physics. A 1995) coined the term “folklore” to describe it. Bateman’s
subsection titled “Theoretical Understanding of Archie’s book devotes Chapter 6, section 6 to the development of the
Equation” (p. 75) provides a discussion based upon the same Archie model. In the lead-up to a key result Bateman states:
work that supports this paper. The authors ¿nally return to “Eventually, Archie found that laboratory measured values of
resistivity interpretation in Chapter 14 (Saturation). F could be related to the rock porosity by an equation of the
form F = a/ĳm.” Well, actually Archie never employs a in any
Books: Petrophysical Practice of his four publications. Later in the text (Chapter 26, page
Asquith, G., and Krygowski, D., 2004, Basic Well Log Analysis, 462) Bateman expands: “where a is a constant (the Archie
2nd Edition, AAPG Methods in Exploration Series, 16. ISBN: constant) …” In fact, a is usually termed the “tortuosity
978-0891816676. factor”, and as just mentioned, Archie never uses any form other
This book is very much oriented as to the “how to” aspects than F = 1/ĳm. In section 6.6.3 Bateman reports “Archie’s
of formation evaluation using logs. Archie’s model relating experiments showed that the saturation of a core could be
formation resistivity, brine resistivity, porosity, and water related to its resistivity.” We do not doubt that Archie made
saturation using parameters m and n is introduced beginning (or paid for) many such measurements during his long career;
on page 3. There is discussion of the model’s formula but however, the relationship he reported in his 1941 presentation to
discussion of the physical basis of the model is not present the AIME and subsequent 1942 publication (i.e., Sw = (Rt /R0)1/n,
except for an explanation of the relationship between and Archie’s Eqs. (4), (5) and (6)) are all based upon data to
resistivity and resistance. The author’s claim that “G.E. Archie be found in four papers cited by Archie from the literature of
… presented a paper … which set forth the concepts used as a the late 1930s; Archie reports no experiments that he made to
basis for modern quantitative log interpretation” overstates the validate this relationship. Otherwise, the Bateman presentation
scope of the Archie model. Archie puts forth no “concepts”; is conventional and does not attempt any derivations for the
only formulas. Asquith and Krygowski also commit the Archie model.
common error of attributing the a in F = a/‫׋‬m to Archie. There Dewan, J.T., 1983, Essentials of Modern OpenǦHole Log
is some further discussion in Chapter 5 (Resistivity) which is Interpretation, PennWell Publishing Company. ISBN: 978-
primarily devoted to logging instrument hardware concepts, 0878142330.
and Chapter 7 (Log Interpretation) but there is no discussion Dewan’s treatment of resistivity uses the arti¿ce of brine-
as to how the formulas relate to ¿rst principles. and sand-¿led cubes to motivate a discussion of formation
Bassiouni, Z., 1994, Theory, Measurement, and Interpretation of resistivity factor. However, his subsequent introduction of
Well Logs, SPE Textbook Series 4. ISBN: 978-1555630560. the Archie model is to just write down the relationships and
Bassiouni begins his book with a chapter on The Electrical claim they are justi¿ed by “general principles” which are not
Resistivity of Rocks. Section 1.4 is devoted to a derivation elaborated.
of the formation resistivity factor. The technique used is to Helander, D.P., 1983, Fundamentals of Formation Evaluation,
equate the resistance of a core plug to the resistance of the OGCI Publications. ISBN: 978-0930972028.
brine in the core plug, both based upon the dimensions and Helander attempts a rigorous justi¿cation of the Archie model
resistivity of the plug. The argument is that to keep the volume on his pages 63 and 64. His ¿nal result is F = ߬2/ĳ. where ߬
of brine V = ĳAL constant in view of the longer-than-L lengths is tortuosity de¿ned in the conventional way as Le/L. This
of the conductive paths, then the cross-sectional area must be would, of course, mean that ĳ | 1/߬ 2, a remarkable result.
modi¿ed accordingly; i.e., V = ĳAL(Le/Le) = ĳA(L/Le) Le = This would certainly seem to warrant some discussion, but
(ĳA/߬)Le. As Bassiouni puts it: “The cross-sectional area of there is none. However, this result is sometimes cited in the
the equivalent water volume, Ac, has to be ĳA(L/Le).” But this literature, and Helander’s derivation of the result is a good
is not so. If the ¿nal formula is to reÀect the effect not only source for understanding the result. Helander does not number
of tortuosity, but also the effect of reduced cross-sectional his equations, but the third equation in the left column of page
area then we can take V = ĮĳA(L/Į) = ĮĳALec as a more, or 64 is the source of this result. It is an equation for the volume
at least equally, valid parameterization of the brine volume. of brine in a core plug, Vcw = AcLĳ , where Ac is the cross-
Then using this parameterization leads to F = (߬/Į)/ĳ = 1/ sectional area of the plug, L the length, and ĳ the porosity. He
(E0ĳ) rather than Bassiouni’s result F = ߬2/ĳ which uses two then multiplies this quantity by 1 = Le/Le and identi¿es Ac(L/Le)ĳ
factors of tortuosity and none for transmissibility of the cross- as the cross-sectional area of an equivalent plug of the same
sectional area. volume with length Le. He then uses the cross-sectional area
Bateman, Richard M., 2012, Openhole Log Analysis and Formation so de¿ned in his expression for the equivalent resistance of
Interpretation, 2nd Edition, Society of Petroleum Engineers. the brine in the core. This leads with a little algebra to his
ISBN: 978-1613991565. ¿nal result. However, what the factor L/Le < 1 does is lead to

a reduction in effective cross-sectional area (which is needed been researching formation evaluation for over a decade. On
for the derivation); the same result could have achieved with page 2 of his book he opines: “In many ways it unfortunate
the introduction of a coef¿cient Į < 1 on the porosity; i.e., that conductivity was not selected in place of resistivity as
Įĳ. With this substitution (i.e., Į/Į = 1 rather than Le/Le = 1) the standard term in the early days of electric logging. …
Helander’s result would have been F = (߬/Į)/ĳ = 1/E0ĳ. Conductivity logs, although identical in general form to
Hilchie, D.W., 1978, Applied Openhole Log Interpretation for resistivity logs, would somewhat simplify the equations now
Geologists and Petroleum Engineers, self-published by D.W. used in log interpretation. It is probably too late to upset
Hilchie. the entire terminology of logging, but it is still sometimes
Introduction to resistivity and saturation analysis by use of the easier to speak and think in terms of conductivities instead
Archie model. There is no attempt to justify the model. of resistivities. … The word conductivity implying as it
Hilchie, D.W., 1989, Advanced Well Log Interpretation, self- does the ability to conduct electric current, is a particularly
published by D.W. Hilchie. convenient term when the mechanism of current Àow in rocks
Introduction to resistivity and saturation analysis by use of is considered.” Wyllie’s derivation of formation resistivity
the Archie model. There is no attempt to justify the model. factor F parallels the discussion in his 1950 paper with
Although the title has changed from “Applied” to “Advanced”, W.D. Rose. He states in that text “Any real rock … has a
the treatment of Archie’s resistivity model remains unchanged. conducting area A perpendicular to the direction in which a
Peters, E.J., 2012, Advanced Petrophysics, v. 1, Geology, Porosity, resistance measurement is made which is proportional to the
Absolute Permeability, Heterogeneity, and Geostatiscis, Live rock porosity, ĳ.” [emphasis added] However, he does not
Oak Book Company, use this proportionality in his derivation. For rock resistance
In spite of its title, the treatment of the Archie model in this he writes rRock = RLe /ĳ (where Le is the length of the current
book is completely conventional. path through a sample of length L) rather than rRock = RLe /Įĳ
Rider, M., and Kennedy, M., 2011, The Geological Interpretation (which includes a proportionality factor Į and forms the ratio
of Well Logs, 3rd Edition, self-published by Rider-French of this quantity to the bulk resistance of the saturating brine,
Consulting, Ltd. ISBN: 978-0954190682. rSolution = RL, his result is F = (Le /L)/ĳ, where Le /L is by de¿nition
Although this book devotes its chapter 5 to Resistivity Logs, tortuosity ߬. Had he included the porosity proportionality
only 1/2 of one column on page 105 is devoted to repeating constant, his result would have been F = (߬/Į)/ĳ, or in our
the Archie model equations, presented with water saturation notation F = 1/E0ĳ. Wyllie’s next equation is a statement of
on the left side. There is no connection made to physics, or Archie’s de¿nition of formation resistivity factor, FRw = R0.
even to trendline ¿tting. Had Wyllie and Rose included the proportionality of cross-
Tittman, J., 1986, Geophysical Well Logging, excerpted from sectional area to porosity in their de¿nition of F, they would
Methods in Experimental Physics, 24: Geophysics, Academic have invented geometrical factor theory, Rw/ĳE0 = R0, in
Press. ISBN: 01206913900. 1950, or Wyllie could have done so in his book a decade later.
Tittman’s book is mainly concerned with logging hardware, Unfortunately, those opportunities were lost. From this point
and is a good resource for logging instruments up to its date of (page 13) in the text, and beyond, Wyllie is conventional in
publication. It’s treatment of resistivity theory is conventional his explanations.
and cursory.
Tiab, D., and Donaldson, E.C., 1996, Petrophysics: Theory and Internet Sources
Practice of Measuring Reservoir Rocks and Fluid Transport One might expect oil¿eld professional societies to be the
Properties, Student Edition, Gulf Professional Publishing. authoritative sources for this kind of material; however, in practice
This book is a comprehensive compendium on (as its title they are of limited use, not very different from the print sources
promises) petrophysics. It is not a “how-to” book on log cited above. Some are cited below.
analysis, and its treatment of formation resistivity is purely Wikipedia (https://en.wikipedia.org/wiki/Archie%27s_law). There
conventional. It does, however, include a discussion on is no derivation to be found here, but the model is given in its
“Theoretical Formula for FR” based upon a 1976 paper by C.P. conductivity form as Ct = CwĳmSwn/a
Rosales, “Generalization of the Maxwell Equation Formation
Resistivity Factors,” Paper SPE-5502-PA, published in JPT. AAPG Wiki (https://wiki.aapg.org/Archie_equation). This
We have not had time to review the reference for our present reference gives the Archie model solved for Sw
article, but suf¿ce to say under most conditions the Rosales (i.e., S wn = Rw/Rtĳm )
formula reduces to F = a/ĳm which is not a surprise.
Wyllie, M.R J., 1963, The Fundamentals of Well Log Interpretation, and with no explanation of the physics. This is the type
3rd Edition, Academic Press. example of why resistivity in rocks is so poorly understood by
Wyllie’s book may have been the ¿rst book published so many who use logs.
(in English, at least) on log interpretation. Wyllie was a
research scientist at Gulf Oil Corporation. In 1963 he had SPE Petrowiki (https://petrowiki.org/Water_saturation_
determination). There is a very thorough discussion of water
saturation determination, but as far as the use of the Archie

model is concerned, the only equation is Sw = (R0/Rt)1/n. with a denotes the bulk properties of the sand-¿lled cylinder.
separate equation for R0, F = R0/Rw. In general, J0 is less than Jw because n0 < nw by the brine
volume ratio, and v0 < vw since the end-to-end speed of the
APPENDIX: TRIPLE-PRODUCT THEOREM FROM ions is reduced due to mobility restrictions placed upon the
ELECTROLYTE CONDUCTIVITY ions by the sand-grains acting as obstacles to Àow, reducing
the cross-sectional area (transmissibility) and increasing the
In the main text of this article the triple-product theorem streamline lengths (tortuosity).
was developed from a statement of three self-evident ¿rst
principles and also by use of Ohm’s law. In the former case
the physical property of conductivity is implicitly assumed
to exist; in the latter case conductivity is de¿ned by ratios
of electrical and geometrical properties; viz., I/V and l/A
respectively. Neither of these methods considers the actual
mechanism of conduction; i.e., mobile charges. We offer a
¿nal discussion of charges in motion that also leads to the
triple-product theorem.
A fundamental relationship for transport is Qvol = Av Fig. A.1—A brine-¿lled cylinder.
where Qvol is the vector volumetric Àow rate with units of
m3/sec, A is the area normal to Àow direction, and v is the
speed of the Àow. If there is something in the volume, say
particles, then the particle Àow rate would be Qpart=nAv
where n is particle density in particles/m3. If the particles
are charged (i.e., ions), then the charge Àow rate will be
Qcharge =nqAv where q is the charge per ion. Note that the
unit is charge per second. Since this is electric current, let
us change to the conventional notation for electric current,
Fig. A.2—A sand-¿lled cylinder suffused with the same brine.
I, and I =nqvA where I is the current vector in particles (or
ions) per second, n is the density of particles (ions/volume),
q is the charge per ion, |v| is the speed of the ions, and A The general forms of the equations are not useful unless
is the cross-sectional area of the Àow. Dividing by A, the the speed of the ions is known. This may not be a convenient
cross-sectional area, converts the left side to current density; formula when the speed would be dif¿cult to measure. Since
J = nqv, where J is current density, is the fundamental the transport is induced by an applied electric ¿eld, it is
equation relating current density to amount of charge passing convenient to convert the fundamental equation to a form
through an area A in one second. The units are included in including electric ¿eld strength. To do this the speed vector
square brackets below to illustrate how they combine. is multiplied by 1 in the form of applied electric ¿eld divided
by applied electric ¿eld magnitude; i.e., for the brine
(A.1) , (A.2)
and for the bulk rock volume

You could well imagine the current density through a
brine-¿lled cylinder (Fig. A.1) given by Jw = nw qvw where the , (A.3)
subscript w is the convention used to denote brine properties.
The same brine suffusing a similar, but sand-¿lled, cylinder
(Fig. A.2) will exhibit different properties. In particular, the where the ratios |vw|/|E| = ȝw and |v0|/|E| = ȝ0 are called
amount of charge in each unit volume would be reduced to a “mobilities”. Their units are (speed per unit of E ¿eld
fraction of the charge in the brine, the fraction being equal to the strength) or ((m/sec) /(volt/m)). Values for mobilities are
volume ratio of the brine in the sand-¿lled cylinder to the brine speci¿c to each different kind of brine and must be measured
volume in the brine-¿lled cylinder; in other words, its porosity. in a laboratory, but once known, can be applied to any known
In the sand-¿lled cylinder J0= n0 qv0 where the 0 subscript solution.

So now we have Jw=nwqȝwE and J0=n0qȝ0E; a term by term Research and operations, Baker-Hughes, PathFinder, and
comparison with Ohm’s law shows that J0=n0qȝ0E = ı0E and Southwestern Energy. David has been an SPWLA member
Jw=nwqȝwE = ıwE. Form the ratio of these quantities as for 40 years and served as V.P. Publications and Editor of
Petrophysics from 1999–2002, V.P. Technology in 2009,
and President (etc.) from 2014–2016. David is inventor or
coinventor on 6 U.S. Patents, and has published about 45
(A.4)
articles as author or coauthor in conference proceedings
. and refereed journals. His major research interest has been
formation conductivity interpretation and electromagnetic
The factors of q and E are the same in numerator and logging instrumentation. David has been an educator all
denominator; thus they cancel and his life, having taught courses at the USMA at West Point,
courses in computer science, physics, materials science, and
, electrical engineering at San Francisco Bay Area private
(A.5)
and junior colleges. He is now a professor in the Petroleum
Engineering Department at Texas A&M University
or, to put it succinctly
instructing graduates and undergraduates in formation
, evaluation. Before all of that, David served as a Lieutenant
(A.6) of Infantry in Vietnam where he received both a Purple Heart
and Bronze Star with V device.
where porosity is identi¿ed with the ion density ratios in the Fredy R. García R., is an
two cylinders and where E0 (not to be confused with electric Exploration Petrophysicist Professional,
¿eld E) is identi¿ed with the mobility ratio. Basically, it is the grade I at Ecopetrol S.A., the largest
ratio of the time it would take for an ion to traverse the sand- petroleum extraction company in
¿lled cylinder with its complicated pore space to the time it Colombia. In this position, Mr. García
would take the same ion to traverse the unobstructed brine- has had experience in the Geosciences
¿lled cylinder. Thus, the triple-product theorem, ı0 = ıwĳE0, and Exploration Department with
is recovered from a comparison of charge Àow rates. The openhole logging interpretation for all
Archie model would follow experimentally by comparing E0 exploratory basins in Colombia and the Caribbean offshore,
for a number of sand-¿lled cylinders of differing porosity. including the Gulf of México, Venezuela and the North Sea
This discussion is simpli¿ed for heuristic purposes; an (United Kingdom). In addition, he has acquired expertise
actual electrolytic solution would have ions of both positive in problematic reservoirs: clean sands with heavy oil/fresh
and negative charge, each type with its own concentration, water, turbidites, tight gas sands, marls and shaly laminated
charge and mobility. Inclusion of these details would sandstones.
complicate the argument without altering the conclusion. Mr. García earned his BSc, in Petroleum Engineering
at the America University Foundation in his native town of
Bogotá, Colombia. He also holds a distinction degree for the
ABOUT THE AUTHORS MSc, in Petrophysics and Formation Evaluation at Aberdeen
University, Scotland.
David Kennedy began a career in His professional experience has Àuctuated between
the logging industry in 1973 following the academic and the technical world. In the ¿rst realm, he
earning a BS in Physics at Georgia Tech. has been a TA for the course “Formation Evaluation and
He entered the industry as a Schlumberger Well Logging Interpretation” at the National University of
¿eld engineer in California and Alaska, Colombia, a teacher for the course “Production Engineering
staying with Schlumberger for ¿ve years. and Open Hole Logging Interpretation” at the America
Following that, Dave returned to school University Foundation, a course developer in Basic and
and earned Masters’ degrees in Physics Intermediate Petrophysical Engineering and Oral Presenter
and Earth Sciences at the University of Texas at Dallas, with and a visiting geoscientist for the AAPG. As part of his
further studies at UC Berkeley. In his career he has worked technical expertise, he has worked as a Development
in one or another capacity at Arco, Sohio Research and Petrophysicist in Gran Tierra Energy, a Reservoir Engineer
operations, Mobil Research and operations, ExxonMobil and Petrophysicist in Vetra Exploration and Production,

as a Well Completions Engineer at Schlumberger and a

Production Engineer at Perenco.
An active member of the SPWLA and the AAPG, Mr.
García has won the Imperial Barrel Award Competition of
the AAPG with the exposition “Hydrocarbon Prospects in
Cooper Eromanga Basin – Australia PEL Asset 106,” as
well as a professional recognition from Ecopetrol based on
the petrophysical interpretation of oil¿eld discoveries in
offshore basins of the Colombian northern coastline. He has
also given oral presentations at SPWLA conventions in the
United States and Colombia.
Like Brian May, BSc and ARCS in Physics of the
Imperial College and lead guitar of the English rock group
Queen, Mr. García also holds a special place in his activities
for the study of the stars and music, being a member of
the Colombian Association of Amateur Astronomers and
a pro¿cient drummer for different bands in Colombia and
Scotland.

Best Of
The Society Of
Core Analysts
2018 International
Symposium

PETROPHYSICS, VOL. 60, NO. 2 (APRIL 2019); PAGES 229–239; 10 FIGURES; 1 TABLE. DOI: 10.30632/PJV60N2-2019a1
Pore-Scale Insights on Trapped Oil During WaterÀooding of Sandstone Rocks of

Varying Wettability States1
Hélène Berthet2, Mathilde Hebert2, Sandra Barbouteau2, Prisca Andriamananjaona2, and Richard Rivenq2
ABSTRACT
Capillary desaturation experiments are combined with to ensure a high-quality segmentation.

high-resolution microtomography imaging to understand Global oil saturation results show how the wettability
the impact of wettability on the global and local distribution impacts the shape of capillary desaturation curves, in
of Àuids in the pore space of sandstone outcrops. Small particular, the existence of a critical capillary number. In
cylindrical rock samples are cored, imaged in dry state the nonwater-wet experiments, oil saturation is controlled
then successively prepared at irreducible water saturation by a large, highly-connected oil cluster percolating from the
before steps of waterÀood. Several samples also go through inlet to the outlet of the sample. Such results are important
a wettability-alteration phase in order to expand the range for pore-scale Àow modeling strategy and validation. We
of wettability conditions: namely, oil-wet to mixed-wet. demonstrate that the wettability is not always uniformly
WaterÀooding is done at various capillary numbers and distributed along the core despite of the use of classical
injected brine volumes, depending on the case. The entire wettability-alteration protocols, highlighting potential
rock is imaged at voxel resolutions of typically 2 or 4 ȝm, biases in traditional SCAL tests.
INTRODUCTION For strongly water-wet systems, the residual oil saturation,

Sorw, remains constant in the lowermost range of capillary
Experimental waterÀooding on cores drained with numbers up to a critical number above which oil starts to be
reservoir oil have long been used by reservoir engineers, as mobilized with Sorw decreasing. For nonwater-wet systems,
a key tool to study the impact of capillary forces and viscous experimental works reported in the literature show a more
forces on the amount of oil that remains microscopically gradual change of oil saturation with increasing capillary
trapped in a well-swept porous medium (Chatzis and number or critical values of these numbers signi¿cantly
Morrow, 1984; Chatzis et al., 1988). This information different from the strongly water-wet case (Mohanty and
is essential for the oil and gas industry to design ef¿cient Salter, 1983; Morrow et al., 1988; Jadhunandan et al., 1995;
production processes as well as EOR techniques that will Masalmeh, 2012; Humphry et al., 2014).
target mobilization of the trapped oil. Most of capillary CDC are the averaged results of local microscopic
desaturation studies were carried out on water-wet systems, processes that take place at the pore-scale. It is therefore
though most of the oil ¿elds are oil-wet to intermediate- very important to study the oil saturation in a more local
wet. Capillary desaturation curves (CDC) used to represent manner. Experimental biases, such as capillary end effects,
the evolution of oil saturation with respect to the capillary preferential Àow pathways, also need to be veri¿ed to
number (balance of viscous and capillary forces) may greatly ensure high experimental quality. One technique used by oil
vary with the system wettability, as recently suggested by companies is to obtain saturation pro¿les using radiographs
Humphry et al. (2014) and Masalmeh (2012). Pore-network of the core during the waterÀooding (Puyou et al., 2017).
characteristics (pore geometry and mineralogy) also appear Recent advances in X-ray tomography can greatly help
to have an impact on the shape of the capillary desaturation us characterize the trapping patterns with much higher
curves. precision, at the pore-scale, at various stages of the capillary
Manuscript received by the Editor November 11, 2018; manuscript accepted January 8, 2019.
1
Originally presented at the International Symposium of the Society of Core Analysts, Trondheim, Norway, August 27–30, 2018,
Paper SCA2018-036.
2
Total SA, Avenue Larribau 64018 Pau Cedex, France; helene.berthet@total.com; mathilde.hebert@total.com; sandra.barbouteau@total.com;
prisca.andriamananjaona@total.com; richard.rivenq@total.com

Pore-Scale Insights on Trapped Oil During WaterÀooding of Sandstone Rocks of Varying Wettability States
desaturation process (Youssef et al., 2015; Wildenschild et macroscopic point of view (capillary desaturation curve) to
al., 2013; Armstrong et al., 2014; Oughanem et al., 2015). local investigation of the trapped oil phase.
This is achieved by visualizing the distribution of Àuid
phases within the pore space of a small core plug (typically a MATERIALS AND METHODS
few millimeters wide and centimeters long) during stationary
stages of a water-based Àood. Similar to what has been done We chose to work on Fontainebleau and Bentheimer
in the past with respect to the macroscopic characterization sandstone outcrops as they have been widely used in
of capillary desaturation curves, pore-scale investigations petrophysics studies (Oughanem et al., 2015). Their
using X-ray tomography focused ¿rst on strictly water–wet advantage also stands in their pore sizes that are large
cases and only recently results were reported on nonwater- enough for good quality X-ray imaging. Figure 1 shows
wet cases, without imaging the entire sample (Singh et al., region-of-interest tomographic images on both rocks, at very
2016; Blunt, 2017). high resolution done on Total’s microtomography system
In this work, we carry out capillary desaturation (ZEISS Versa 520T). The main mineralogy difference is the
experiments on sandstones at various wettability conditions presence of clay in the Bentheimer rock.
combined with X-ray imaging in order to visualize and For each experiment, a sample of approximately 4
analyze the trapped oil phase within the pore structure. The mm in diameter and 30 mm in length is mounted in a
entire rock sample is imaged at the end of each waterÀood microÀuidic Àow-cell developed at Total. This apparatus
(static conditions) at micrometer voxel size to well-resolve constitutes a microÀuidic Hassler cell that can be positioned
the pore structure. Submicron resolution imaging is also on the sample stage of the microtomography scanner
carried out to better see details of the distribution of Àuids enabling high-resolution 3D imaging as well as precisely
according to the wettability. controlled microÀuidic Àow of oil or brine at high pressure
The water-wet test is an interesting case to start with and temperature (up to 100 bar, 100°C). For the experiments
in order to compare with literature results. We focus on described in this paper, the operating conditions were 20
three other experiments for which the rock wettability was bars of internal pressure, 40 bars con¿ning pressure, and
altered towards an oil-wet system. Our analysis is based on 28°C temperature (temperature inside the microtomography
the comparison between water-wet and oil-wet cases, from a scanner).
(a) (b)
Fig. 1—Images of our (a) Fontainebleau, and (b) Bentheimer samples obtained from high-resolution region-of-interest microtomography (image
resolutions 590 and 409 nm/voxel, respectively). Pores appear dark, grains appear bright. The scale bar (top right in each ¿gure) indicates 100 m.

Berthet et al.
The wettability alteration (not a wettability restoration) saturation, Swi, is reached at a capillary number of 10-5. See
is obtained by heating the sample at 90°C for two weeks, Table 1 for details on the Àuids used. The samples used for
following a recipe developed at Total speci¿c to the crude the oil-wet experiments are directly oil saturated (Swi = 0,
oil and the outcrop rock Bentheimer used. The temperature not ¿eld-representative) before going through the wettability
inside the Àow cell is regulated, as well as the pressure of the change step. We then perform a series of successive
oil phase, throughout the entire time. The Àow cell is kept waterÀoods and scans of the entire sample. For Fontainbleau
inside the microtomography scanner. A similar technique water-wet experiment, the waterÀoods consist in injecting
was applied in Humphry et al. (2014). 1.5 pore volumes (PV) of brine, at increasing Àow rates
The cell is connected to a microÀuidic Àow circuit (capillary number). The oil-wet experiments differ as, for a
dimensioned for the precise Àow of crude oil, brine and given Àow rate, we have several steps of brine injections,
solvents in pressure and temperature (mainly computer- using increasing volumes of brine. Then the Àow rate is
controlled syringe pumps, sensors, valves from CETONI, increased, and we again Àow at steps of brine volumes. For
https://www.cetoni.de/). The Àow rates used in these example, in the ¿rst oil-wet experiment, we successively
experiments ranged between 0.84 L/min up to 5 mL/min. injected 0.24 then 1 then 4 PV of brine, etc.
We also designed our Àow cell so that we could control the After each waterÀood, we perform a series of tomographic
Àow circuit while continuously acquiring radiographs of the scans to capture the Àuid distribution in the entire sample.
sample and the inlet/outlet tubings. We are therefore able to The voxel resolution varies between 2 and 4 m depending
detect the arrival of brine just before it enters the sample in on the operational constraints. Over all, Àooding steps are
order to measure our dead volumes, then detect it as it Àows never longer than a few hours, accelerating to only tens
out of the sample to obtain the brine breakthrough point. of seconds for the highest Àow rates. The scanning time is
Table 1 details the experimental conditions of each however of the order of 20 hours for a 4-m voxel resolution
experiment analysed in this paper. The capillary number image, reaching 2 days for a 2-m voxel resolution scan.
(Ca) is computed with the injected Àuid viscosity, Àow rate During this waiting time, the internal pressure of the Àooding
and measured brine/decane or brine/crude oil interfacial phase is regulated using the microÀuidic Àow system.
tension: Ca = ȞQ/Ȗ. The experimental results discussed in this paper are
from the Fontainebleau water-wet experiment and the
EXPERIMENTAL PROCEDURE Bentheimer altered-wettability experiments. Though the
rocks are very similar, we performed a capillary desaturation
The entire sample, mounted in the Àow cell, is ¿rst on the Bentheimer rock in water-wet conditions (unaltered
scanned dry at 2-m voxel size. We then vacuum-saturate wettability) as well, to con¿rm the Fontainebleau behavior.
it with brine and drain it with oil until irreducible water
Table 1—Experimental Conditions for the Water-Wet and the Three Oil-Wet Experiments
Fontainbleau

Fig. 2—Processing workÀow from the raw 3D image to local analysis of Àuids distribution in the pore space. The third panel shows only the oil voxels
in the image, color-coding each connected ganglion of oil. Similarly, each individual pore is assigned a given label (color) in the pore-space partitioning
phase based on geometrical analysis (right-end panel).
IMAGE PROCESSING We then ¿lter and segment each dataset into two or
three phases: oil, brine and grains, using a region-growing
The processing of each scan follows a multistep segmentation method from ANU (Sheppard et al., 2004).
workÀow (Fig. 2) and is done using Total’s supercomputer Global saturations and saturation pro¿les are obtained from
and ANU’s image processing software MANGO (https:// the segmented images. In order to study the pore-space
physics.anu.edu.au/appmaths/capabilities/mango.php). Àuid occupancy and the Àuid connectivity, we perform a
The region of interest is extracted from the raw data (3D partitioning step of the pore space to identify individual pores,
reconstructed volume) to mask the sleeve around the sample and a partitioning of the oil phase to identify disconnected
and the inlet and outlet tubings. We optimize the masking oil clusters.
so that the sample volume is almost entirely kept in the
processed image.
(a) (b)
Fig. 3—Capillary desaturation curves for the water-wet Fontainebleau experiment (red diamond), compared to Chatzis and Morrow (1984). Oil
saturation is normalized by saturation after the ¿rst waterÀood. Black-square data points correspond to the water-wet Bentheimer test performed to
con¿rm similar behavior between the two outcrops Bentheimer and Fontainebleau. The inset graph shows the CDC curves for the large capillary
numbers. (b) Section of the imaged volume at Swi (18%), corresponding to the entire Fontainebleau sample diameter (2 m/voxel).

Berthet et al.
RESULTS using a distance equal to the REV divided by the sample

surface area. Imaging the entire sample enables to detect if
Evolution of the Oil Saturation, at the Sample Level some regions of the rock are affected by boundary effects,
Our ¿rst experiment, in water-wet conditions, and if saturations should be computed in a restricted volume
constitutes a unique dataset at the pore-scale level, as seven only.
steps of waterÀoods were realized and fully imaged at a Figure 4 also shows an image from the reconstructed
voxel resolution of 2 m. Figure 3 shows the evolution of volume extracted in the middle plane, as well as a zoom on a
the global oil saturation with respect to the capillary number. few pores to illustrate the local distribution of oil and brine:
Oil saturation is simply obtained by counting the voxels the curvature of the oil/brine interface clearly indicates a
assigned to the oil phase during the segmentation step of the water-wet system. This image also shows that the trapped
processing. We normalized the Sor value by the Sor* (=38%), oil can be disconnected in drops in the pore matrix.
saturation obtained after the ¿rst waterÀood. The agreement We now focus on the ¿rst oil-wet experiment with a
with the literature is excellent: plateau of saturation before a macroscopic view of the oil-saturation evolution, with respect
critical capillary number of 3.10-5. to the injected brine volume (Fig. 5). Oil recovery increases
Oil-saturation pro¿les along the sample length are rapidly at start of the waterÀood (though less than in the
interesting illustrations of the complexity of the untrapping water-wet case), with 40% of the oil mobilized in less than
and trapping of oil at the pore-scale. Figure 4 shows ¿ve 2 PV injected. As we continue injecting, the oil saturation
pro¿les, including one at the end of the saturation plateau continues to drop but much more slowly, without any sign
(blue), and three as the oil saturation signi¿cantly decreases. of stabilization. This slower regime could correspond to the
The smoothed curves were obtained in two steps. First, we thinning of oil layers along the grains, which are visible on
computed the porosity representative elementary volume the image of Fig. 5. Note also the curvature of the brine-
(REV) (Blunt, 2017) from the dry image: 3 mm3. Then, oil interface and the presence of brine droplets in pores,
we smoothed the saturation pro¿les with a moving average illustrating that the sample is oil-wet.
(a) (b)
Fig. 4—(a) Evolution of the oil saturation along the sample length. The bottom image shows an image extracted in the middle plane of the sample.
(b) The image shows the local distribution of Àuids in a water-wet case obtained by imaging at 800 nm/voxel.

(a) (b)
Fig. 5—(a) Evolution of oil saturation with respect to the injected brine volume, in the ¿rst oil-wet experiment. The inset shows the oil saturation pro¿le
along the sample length. Note that the end-effect, expected in more oil-wet systems, is restricted to the last millimeters of the sample only. (b)The
image was obtained from a scan of the sample after the last waterÀood, at 1-ȝm/voxel resolution.
The capillary desaturation curves of the oil-wet

experiments are presented in Fig. 6. The different data
points for a given capillary number correspond to the several
injections at increasing volumes of brine. The graph should
therefore be read from left to right and top to bottom when
applicable. In the third experiment, we were able to detect
the breakthrough point (second data point of the graph). The
saturation computed numerically (voxel count) compared
very well with the injected brine volume.
The second experiment displays a continuous decrease
without a critical capillary number. In the third test, the oil
saturation steadily decreases until it stabilizes when Ca = (a)
10-5 to 10-4 before a signi¿cant drop down to 20%. Note on
Fig. 6b that we reach very large quantities of injected brine
due to the successive waterÀoods.
Though similar wettability-change techniques were
used between the three experiments, the recovery is much
higher for the third test: 28% at Ca = 10-5 compared to just
above 40% in the second test.
Connectivity of the Oil Phase

We now focus on the connectivity of the oil phase by
processing the segmented data further. Two oil voxels are
considered connected if they are in contact through faces,
vertices, or edges (26 degrees of connectivity). In Fig. 7, we
visualize the oil connectivity by looking only at the oil phase (b)
and color-labeling each independent oil cluster. Three steps Fig. 6—(a) Capillary desaturation curves for the oil-wet experiments
of the water-wet experiment and the ¿rst oil-wet experiment (Bentheimer rock). (b) Oil saturation versus the cumulative volume of
injected brine.

Berthet et al.
are illustrated. In the water-wet test, we observe a lot of oil waterÀood and for the four experiments is plotted in Fig.
fragmentation due to snap-off events in the constrictions of 8. Three trends can be observed: (1) When the rock is
the pore space (Roof, 1970). From the ¿rst waterÀood (<1 PV water-wet, talking about a largest cluster is meaningless
injected) to the fourth one (at the critical capillary number), as it only counts for a few percent of the trapped oil, after
the 18,000 blobs of oil fragment into 59,000 clusters. The the second waterÀood. (2) On the contrary, in the oil-wet
largest blob corresponds to 0.3% of the oil-in-place. In the experiments the largest cluster, after six waterÀoods, still
¿rst oil-wet experiment, represented in the lower row, the contains more than 80% of the remaining OIP (see also Fig.
oil stays highly connected within a cluster percolating from 4). (3) An intermediate behavior is found for the two other
the inlet to the outlet of the core. Note also the shape of the oil-wet experiments. We observe a large connected cluster at
oil blobs: for the water-wet rock, oil blobs have rounded the beginning of the desaturation that gets fragmented into
borders as they are surrounded by brine in the pores. In the clusters counting for a few percent of the oil in place. As
second case, the oil is in contact with the grains, giving them explained before, we maximize our chances of considering
rougher borders. two voxels connected as we take into account contact
As shown in Fig. 7, it is interesting to follow the between vertices. Nevertheless, noisy images and resolution
proportion of oil contained in the largest oil cluster for all limit could lead to disconnected segmented voxels that are
desaturation experiments. Its evolution, waterÀood after actually connecting.
Fig. 7—Evolution of the connectivity of the oil phase depending on the wettability and the rock. The same slice extracted from the imaged volume is
displayed after each waterÀood. Brine and grain voxels are black. Each group of connected oil voxels, constituting an oil blob, is identi¿ed by a colored
label. In the water-wet case, oil gets more and more disconnected, contrary to the oil-wet case. OIP = oil in place.

Figure 9 presents the evolution of the oil ȕ1 with

cumulative brine injected volume, for the four experiments.
In order to remove the effect of the pore-space connectivity
(which may vary between Bentheimer and Fontainebleau
rocks), we normalize the value by the pore space ȕ1,
computed on the dry image at the same resolution. This value
is typically very high, as many loops can be found within the
pore network. In the water-wet case, the oil ȕ1 drops rapidly
and remains low throughout the experiment. In the oil-wet
case, it remains close to one, quantitatively con¿rming the
presence of a well-connected oil cluster. The normalized ȕ1
stay in between these two extreme situations in the other two
experiments, with the same slowly decreasing trend as the
water injections are multiplied.
Fig. 8—Oil contained in the largest oil cluster present after each
waterÀood step, in percentage. WW is water-wet, OW is oil-wet.
The critical behavior of the largest cluster size occurs

at macroscopic capillary numbers of 10-7 in the water-wet
experiment, 10-6 in the third oil-wet experiment, and 10-5 in
the second oil-wet experiment.
High connectivity of the oil phase, or, on the contrary,
oil trapping as isolated droplets in the pore space may be
quanti¿ed using the topology invariants “Betti numbers”
(Herring et al., 2013). Given an object (i.e., the oil phase),
the Betti numbers will correspond to the number of isolated
components of this object (the number of disconnected oil Fig. 9—Oil ȕ1 normalized by the pore space ȕ1, as waterÀoods are
clusters, ȕ0), the number of loops within the object (ȕ1), and carried out, for all wettability cases. Note that the capillary number
the number of entrapped objects of the second phase: brine is increased stepwise at different cumulative brine injected volumes
across the experiments.
droplets fully surrounded by oil (ȕ2). Focusing on the second
Betti number ȕ1, we computed it for each step of our water-
wet and altered-wettability experiments, in the entire image
sample volume. Let’s consider three situations: Combining results from Figs. 8 and 9 for the OW2
1. The oil is distributed as trapped spherical droplets in and OW3 experiments, we understand that the oil, though
the center of the pores. disconnecting into smaller clusters, remains in complex
2. The oil is entirely contained in one cluster connected tortuous shapes within the pore space (stable ȕ1 values)
across the pore network. possibly due to its af¿nity with the rock surface.
3. Some of the oil is trapped in the center of the pores, “Oil-wet” experiments 1, 2 and 3 were all performed
some of the oil is well connected. using the same wettability-change protocol. We con¿rmed
the wettability change, as explained earlier, by tracking
In Case 1, ȕ1 = 0, as each spherical droplet has a ȕ1 of the curvature sign of the oil/brine interface. However, the
zero. In the second case, ȕ1 will be a very large number, differences observed between the two CDC as well as the
tending to the ȕ1 of the pore space. In the third case, ȕ1 will differences in connectivity behavior led us to revisit the
vary according to the balance between trapped droplets and “raw” data: the imaged volumes. We navigated through the
the connected cluster. thousands of images, waterÀood by waterÀood, and were
One should note that ȕ1 calculation in the able to ¿nd images in which the distribution of Àuids changes
microtomography images is extremely sensitive to the completely from a pore to the next one, as illustrated in Fig.
image resolution, similar to the determination of connected 10. Curvature of the oil/brine interface changes sign at very
components by image-processing algorithms. small distance. We couldn’t ¿nd similar images in the ¿rst

Berthet et al.
Fig. 10—CT Slices extracted from oil-wet experiments OW2 and OW3 in which the curvature of oil blobs changes sign from one pore to the next.
Similarly, oil can be seen in the middle of a pore, surrounded by brine (water-wet like), as well as along pore grains (oil-wet like).
oil-wet experiment, which we consider more uniformly oil- highlights large differences of Àuid behavior depending
wet. It is also worth noting that these results are not due to on the wettability. In oil-wet experiments, the trapped
a change of wettability over time, though large quantities of oil remains highly connected, and loss of connectivity
brine were injected during long experiments. Heterogeneous (quanti¿ed in particular by the Betti numbers) seems to be a
wettability obtained after wettability alteration has previously clear indicator of water-wettability. Differences between all
been reported for other rock-Àuid systems (Aspenes et al, three oil-wet desaturation experiments show the dif¿culty in
2003). reproducing experiments involving wettability changes. We
A heterogeneously distributed wettability with patches also demonstrate that wettability-change methods may not
of water-wet and mixed to oil-wet grain surfaces is expected lead to a homogeneous change everywhere in the pore space.
to greatly impact the desaturation experiment. It could All the highlighted microscopic signatures of wettability
explain the oil fragmentation observed in experiments lead to capillary desaturation curves that greatly vary
OW2 and OW3 (local events). Also, the intermediate between water-wet and oil-wet sandstones. Quantitative
ȕ1 connectivity behavior would be the signature of the measurements of contact angles, at the pore scale, will be
proportion of water-wet to oil-wet regions. An automatic very useful to further detail our analysis of these unique
contact-angle measurement tool would be useful to quantify datasets, in particular to understand the heterogeneous
this wettability heterogeneity in terms of location and distribution of wettability.
difference in angle value.
NOMENCLATURE
CONCLUSIONS
Abbreviations
In this work, a series of capillary desaturation
experiments was conducted on sandstone outcrops and 3D = three-dimensional
combined with X-ray microtomography imaging of the CDC = capillary desaturation curve
entire rock samples to capture the distribution of oil and MW = mixed-wet
brine in each pore, at each waterÀood step. Wettability OIP = oil in place
was altered from water-wet towards oil-wet by leaving the OW = oil-wet
sample saturated with reservoir oil, at high temperature PV = pore volume
for two weeks. One sample was left water-wet. A detailed REV = representative elementary volume
analysis of the trapped oil phase and through its connectivity WW = water-wet

Masalmeh, S., 2012, Impact of Capillary Forces on Residual Oil

Symbols
Saturation and Flooding Experiments for Mixed- to Oil-Wet
Ca = capillary number Carbonate Reservoirs, Paper SCA2012-11 presented at the
Sorw = residual oil saturation International Symposium of the Society of Core Analysis,
Swi = initial water saturation Aberdeen, Scotland, UK, 27–30 August.
Sor* = residual oil saturation following the ¿rst waterÀood Mohanty, K., and Salter, S., 1983, Multiphase Flow in Porous
Ȟ = brine viscosity Media: III. Oil Mobilization, Transverse Dispersion and
Q = brine Àow rate Wettability, Paper SPE-12127 presented at the SPE Annual
S = sample cross section Technical Conference and Exhibition, San Francisco,
California, USA, 5–8 October. DOI: 10.2118/12127-MS.
Ȗ = brine/oil interfacial tension
Morrow, N., Chatzis, I., and Taber J., 1988, Entrapment and
ȕ1 = second Betti number Mobilization of Residual Oil in Bead Packs, Paper SPE-
14423, SPE Reservoir Engineering, 3(3), 927–934. DOI:
ACKNOWLEDGMENTS 10.2118/14423-PA.
Oughanem, R., Youssef, S., Bauer, D., Peysson, Y., Maire, E., and
Vizika, O., 2015, A Multi-Scale Investigation of Pore Structure
The authors would like to thank Dr. G. Hamon, Dr.
Impact on the Mobilization of Trapped Oil by Surfactant
M. Chamerois and Dr. F. Nono for useful discussions. We
Injection, Transport in Porous Media, 109(3), 673–692. DOI:
are grateful to G. Quenault, J.P. Chaulet, R. Farwati and 10.1007/s11242-015-0542-5.
R. Meftah for their contributions to the experimental and Puyou, G., N’guyen, M., and Savin, S., 2017, CXBOX: An
processing work. Innovative Tool for Fluid Dynamic Quanti¿cation During
CoreÀoods, Paper SCA2017-037 presented at the International
REFERENCES Symposium of the Society of Core Analysts, Vienna, Austria,
27 August–1 September.
Roof, J.G., 1970, Snap-Off of Oil Droplets in Water-Wet Pores,
Armstrong, R.T., Georgiadis, A., Ott, H., Klemin, D., and Berg,
Paper SPE-2504, Society of Petroleum Engineers Journal,
S., 2014, Critical Capillary Number: Desaturation Studied
10(1), 85–94. DOI: 10.2118/2504-PA.
With Fast X-Ray Computed Microtomography, Geophysical
Singh, K., Bijeljic, B., and Blunt, M.J., 2016, Imaging of Oil
Research Letters, 41(1), 55–60. DOI: 10.1002/2013GL058075
Layers, Curvature and Contact Angle in a Mixed-Wet and
Aspenes, E., Graue, A., and Ramsdal, J., 2003, In Situ Wettability
a Water-Wet Carbonate Rock, Water Resources Research,
Distribution and Wetting Stability in Outcrop Chalk Aged in
52(3), 1716-1728. DOI: 10.1002/2015WR018072.
Crude Oil, Journal of Petroleum Science and Engineering
Sheppard, A.P., Sok, R.M., and Averdunk, H, 2004, Techniques
39(3–4), 337–350. DOI: 10.1016/S0920-4105(03)00073-1.
for Image Enhancement and Segmentation of Tomographic
Blunt, M.J., 2017, Multiphase Flow in Permeable Media: A Pore-
Images of Porous Materials, Physica A, 339(1–2), 145–151.
Scale Perspective, Cambridge University Press. ISBN: 978-
DOI: 10.1016/j.phsa.2004.03.057.
1107093461.
Wildenschild, D., and Sheppard, A.P., 2013, X-Ray Imaging and
Chatzis, I., Kuntamukkula, M.S., and Morrow, N.R., 1988, Effect
Analysis Techniques for Quantifying Pore-Scale Structure
of Capillary Number on the MIcrostructure of Residual Oil in
and Processes in Subsurface Porous Medium Systems,
Strongly Water-Wet Sandstone, SPE Reservoir Engineering,
Advances in Water Resources, 51, 217–246. DOI: 10.1016/j.
3(3), 902–912.
advwatres.2012.07.018.
Chatzis, I., and Morrow, N.R., 1984, Correlation of Capillary
Youssef, S., Peysson, Y., Bauer, D., and Vizika, O., 2015, Capillary
Number Relationships for Sandstone, Paper SPE-10144,
Desaturation Curve Prediction Using 3D Microtomography
Society of Petroleum Engineers Journal, 24(5), 555–563.
Images, Paper SCA2015-008 presented at the International
DOI: 10.2118/10114-PA.
Symposium of the Society of Core Analysts, St John’s
Herring, A.L., Harper, E.J., Andersson, L., Sheppard, A.P., Bay,
Newfoundland and Labrador, Canada, 16–21 August.
B.K., and Wildenschild, D., 2013, Effect of Fluid Topology
on Residual Nonwetting Phase Trapping: Implications for
Geologic CO2 Sequestration, Advances in Water Resources, ABOUT THE AUTHORS
62, Part A, 47–58. DOI: 10.1016/j.advwatres.2013.09.015.
Humphry, K.J., Suijkerbuijk, B.M.J.M., van der Linde, H.A., Hélène Berthet is a senior reservoir
Pieterse, S.G.J., and Masalmeh, S.K., 2014, Impact of engineer and research scientist at Total.
Wettability on Residual Oil Saturation and Capillary She launched the Digital Rock Physics
Desaturation Curves, Petrophysics, 55(4), 313–318. lab within Total, developing multiphase
Jadhunandan, P.P., and Morrow, N.R., 1995, Effect of Wettability microÀuidic experiments on cores
on WaterÀood Recovery for Crude-Oil/Brine/Rock Systems, combined with X-ray microtomography
Paper SPE-22597, SPE Reservoir Engineering, 10(1). DOI:
imaging to better understand oil and brine
`10.2118/22597-PA.
transport mechanisms at the pore-scale. She graduated from

Berthet et al.
ESPCI ParisTech and Ecole Polytechnique with MS degrees Àuid Àow in porous media, digital rock physics and reservoir
in physics and Àuid mechanics and obtained her PhD in Àuid simulation. He joined Total in 1981, after graduating from
mechanics from ESPCI in 2012. ESPCI ParisTech.
Mathilde Hebert is a reservoir

engineer performing research
investigations in Digital Rock Physics at
Total. Her work focuses on multiphase
microÀuidic experiments using X-ray
imaging to characterize Àuid Àow at
thr pore-scale in reservoir rocks. She
obtained a MS from the Ecole Nationale
Supérieure de Géologie and a MS in reservoir engineering
from Heriot Watt University. Her previous work includes a
two-year reservoir engineer position at Total E&P Congo.
Sandra Barbouteau is a researcher

in Total. She currently works within
the Digital Rock Physics team in
France focusing on 3D advanced image
processing. She has previously worked
on topics like carbonate and multiscale
imaging. She graduated from ESPCI
ParisTech and obtained her PhD in
passive seismic monitoring and petro-acoustics from the
French Petroleum Institute IFPEN and Grenoble University.
Prisca Andriamananjaona is a PhD

student at the Université de Pau et des
Pays de l’Adour, Pau, France, working
closely together with the Digital Rocks
Physics team of Total, Pau, France. Her
research focuses on the in-situ wettability
characterization, and more precisely on
the manifestation of the Amott wettability
indices at the pore-scale perspective by microcomputed
tomography imaging. In 2016 she obtained her master’s
level degree in Complex systems – Complex Àuids and
Divided media at the University Pierre et Marie Curie and
the University Paris Diderot in Paris, France.
Richard Rivenq is a Senior

Research Reservoir Engineer with Total.
His ¿rst research focus with Total was the
development of chemical EOR processes
and its application in the ¿eld. He then
held a number of operational reservoir
engineering positions related to ¿eld
development and production in various
countries, mainly in Europe and Africa. He is currently
leading research projects dedicated to the fundamentals of

PETROPHYSICS, VOL. 60, NO. 2 (APRIL 2019); PAGES 240–254; 12 FIGURES; 4 TABLES. DOI: 10.30632/PJV60N2-2019a2
Uncertainty Quanti¿cation in Image Segmentation for Image-Based Rock Physics in a

Shaly Sandstone1
James Howard2, Sam Lin3, and Shawn Zhang3
ABSTRACT
Image processing of high-resolution 3D images to assignment in any pore-grain segmentation strategy. Two
create digital representation of pore microstructures for segmentation strategies, a binary segmentation with a
image-based rock physics simulations remains a highly linear-threshold and a machine-learning (ML) approach
subjective enterprise, despite the seemly precision to two-phase segmentation, are employed with different
associated with improving imaging resolutions and UQ parameter space. The contribution of resolvable
intensive parallel computations. The decisions on how macropores in these samples, and their spatial distributions
to identify pore space, both macro- and micropores, and with regard to pore-lining clay mineral with unresolvable
various mineral components remain very much dependent microporosity, are iteratively studied over the de¿ned UQ
upon user choices and biases. This study demonstrates parameter space, and cross-validated by independent NMR
how uncertainty can be quanti¿ed for a highly subjective and MICP measurements. The pore structure extracted
segmentation process. A set of shaly sand samples with from these different iterations was the basis of simulations
signi¿cant amounts of authigenic chlorite/smectite that for basic petrophysical properties. Upon cross-validation
lines larger pores was tested to identify uncertainty of simulated results with measured core properties, a UQ
quanti¿cation (UQ) requirements associated with image- framework is proposed to assess the differences between
processing steps, segmentation in particular. Much of the different measurements from three angles: sampling,
the porosity in these coarse-grain samples is associated numerical and physical.
with subresolution micropores that complicates their
INTRODUCTION by the image resolution where the narrowest constrictions

between observed pores that fall below resolution are not
There has been considerable recent attention directed always included in any transport calculations. In contrast,
to the problem of how image-based rock physics (IBRP) a PNM based on skeletonization of the image volume
can reduce uncertainty in SCAL measurements by running leads to a geometrical characterization of pore bodies with
multiple iterations of a core-analysis simulation that connecting throats of ¿nite length and volume that is often
help evaluate the relative importance of various input constructed from information additional to the images.
parameters (Schembre-McCabe and Klamath, 2017). There The nonuniqueness of the PNM construction process leads
is less discussion on how the image-processing procedures to uncertainties in the subsequent simulation of transport
contribute to the uncertainty of the calculated properties properties that are dif¿cult to quantify (Idowu et al., 2014).
(Schembre-McCabe et al., 2011; Idowu et al., 2013). The In addition to accuracy and measurement-independency
importance of characterizing the connected pore space is considerations, direct use of imaging data without PNM
often reduced to the distinction between using a hypothetical simpli¿cation allows the quanti¿cation of uncertainties in
pore-network model (PNM) of pores and throats versus only terms of choices made on gray-scale intensity values.
the image-based pore volume where the distinction between Experimental uncertainties are often expressed as an
pores and throats is less explicit. The latter is restricted error bar. Computational physics results, however, are
Manuscript received by the Editor October 28, 2018, revised manuscript received February 22, 2019; manuscript accepted March 18, 2019.
1
Origially presented at the International Symposium of the Society of Core Analysts, Trondheim, Norway, August 27–30, 2018,
Paper SCA2018-034.
2
DigiM Solution LLC, 2526 S. Delaware Pl, Tulsa OK, 74114, james.howard@digimsolution.com
3
DigiM Solution LLC, 67 S. Bedford St., Suite 400W, Burlington, MA 01803, sam.lin@digimsolution.com; shawn.zhang@digimsolution.com

Uncertainty Quanti¿cation in Image Segmentation for Image-Based Rock Physics in a Shaly Sandstone
further challenged with coming up with an error bar due inverted into a relaxation time distribution with 30 time
to the much greater number of parameters involved (Wu et points between 0.1 msec and 8 sec using a version of the
al., 2013; Al-Taie et al., 2014). Many numerical tools never Butler-Reeds-Dawson solution to the non-negative least-
pass the validation and veri¿cation phase. An earlier study squares problem (Dunn, et al., 2002). Regularization for the
illustrated where computational tools are used to quantify inversion was set to a constant value for all samples, which
the uncertainties in experiments, which in turn con¿ne the reÀected the similar signal/noise values for each of the sets
parameter space of the numerical domain (Zhang et al., of stacked echo trains.
2004). In this study, we illustrate how a highly interactive Mercury-injection porosimetry measurements were
and seemingly subjective task, such as image segmentation, acquired on 2 to 3 g of the ~5-mm cleaned chips for samples
can have its uncertainty quanti¿ed. Uncertainties in IBRP A to C using an AutoPore 9220 (Micromeritics) instrument
occur on at least three other steps (1) image ¿ltering, (2) to acquire data between 1.5 and 60,000 psi. The intrusion
segmentation post-processing, and (3) simulation. While the curve was corrected for very slight, < 0.005 cm3, intrusion
uncertainty quanti¿cation framework presented in this paper at low pressures that occurs because of surface roughness.
is applicable, we will leave the discussion to a different Intrusion pressure was converted to pore-throat radius with
project to con¿ne the scope of this paper. the Washburn equation and default parameters for the Hg/air
system; contact angle of 140o and a surface tension of 480
SAMPLES AND METHODS dyne cm-1.
Micro-CT images were collected on the cleaned
Four samples were selected from a 40-m interval of miniplugs in the dry state on a Versa 510 instrument (Zeiss)
reservoir sandstone noted for a pore-lining interstrati¿ed to acquire 4.0- and 2.0-ȝm resolution images. The data were
chlorite/smectite clay mineral that affected various recorded as 16-bit unsigned integers. The absence of images
petrophysical properties including permeability and acquired at different saturation states limited the detection
electrical conductivity. The mineralogy of this interval was of subresolution features. The images were processed with
60 to 70 wt% quartz, 15 to 20 wt% feldspars, mostly Na- the instrument’s software and exported in *.tiff format.
plagioclase, and 10 to 15 wt% clay minerals, the authigenic Image quality was good enough that no post-processing was
chlorite/smectite. Diagenetic siderite cement occluded used to remove noise or sharpen edges. This may affect the
some of the pores in these samples. Average grain size for effectiveness of a basic binary segmentation with linear-
these samples was 110 ȝm, with broad distributions skewed threshold approach, but does not inÀuence the machine-
towards the smaller sizes. learning (ML) strategy, since it takes advantage of a broader
Several types of samples were prepared for each for the range of information associated with each image voxel. The
four samples. Standard core plugs, 3.75-cm diameter, 3.7- to total-image package consisted of 1,000 stacked images in
4.0-cm length were cleaned and dried for porosity and gas- the Z direction, 992 × 1,014 voxels in the x- and y-directions.
permeability measurements. These plugs were then pressure These image stacks were cropped to 400 × 400 × 400 and
saturated with a 0.1 N NaCl brine in preparation for NMR and 680 × 680 × 900 volumes to remove edge effects and poor-
electrical conductivity measurements. Smaller plugs, 4-mm quality images due to cone-angle artifacts at the ends of the
diameter and 6- to 8-mm length were drilled from remnant sample. The cropped volumes were centered on the sample’s
material adjacent to the routine core plugs for use in the center such that the middle slice for each cropped volume
micro-CT tests. Small chips of cleaned remnant material ~5- was still the sample’s middle.
mm diameter were selected for mercury injection capillary The cropped images were segmented with a ML engine
pressure (MICP) measurements. optimized for a proprietary image-management cloud (Witten
Electrical conductivity was measured on the brine et al., 2017). Speci¿cally, for this project, a random-forest
saturated samples, Cw = 10.5 mS/cm at room temperature, algorithm was used to combine a large number of statistical
with a two-electrode setup in a Hassler cell with a con¿ning results extracted from a set of ¿lters, edge detection and
pressure of 250 psi. The measurement stabilized for 5 other feature extraction algorithms that evaluated nearest-
to 10 min before a ¿nal reading was collected. The brine neighbor and next-nearest-neighbor relationships for each
conductivity was divided by the measured conductivity to voxel (Zhang, et al., 2019). The workÀow started with
calculate formation factor. a set of seed voxels selected by the user directly from the
Low-¿eld NMR relaxation measurements on the image’s 2D training area to represent the pores and grains. In
saturated routine core plugs were acquired on a 0.047- addition to gray-scale intensity and its gradient, a collection
T benchtop spectrometer out¿tted with a high-Q probe of statistical measures was computed on the training data that
with a tau value of 300 ȝsec. The stacked echo train was included ¿rst and second derivatives of intensity pro¿les,

Howard et al.
along with sensitivity information on the neighboring voxels mercury/air capillary pressure (Young-Laplace equation
and the detection of edges. Iterative re¿nement of the seed solved with a morphological method), with a workÀow
voxels on the small training image ensured critical features based upon a previous validation effort (Byrnes et al., 2017).
associated with pores and collective grains were identi¿ed Size distributions for any segmented phase, primarily the
and correctly segmented. The segmentation procedure used resolved macropores in this study, were calculated with a
on the training image was then automatically applied to maximal-ball/digital-sphere approach on the connected
the full 2D ¿eld of view and subsequently to the full 3D 3D volume (Delerue, et al., 1999). Sphere size growth is
stack. To ensure accuracy, the default number of statistical stopped when it touches the edge of adjacent different phase,
measures in the basket is 15 with the ability to turn off certain i.e., grains in this study, so the resulting size distributions
features. Furthermore, the training image can be multiple represent minimal pore widths.
2D cross sections at different locations and orientations. The transport-property simulations involved a binary
Both considerations demanded signi¿cant computing that system of grains and pores, of which only the resolved
a desktop workstation cannot meet. A high-performance macropores contributed to the calculation. Each simulation
computing cluster was designed to optimize both image required an initial estimate for the value of the input parameter
processing and image-based numerical simulation (Zhang, of interest, the resolvable pores. The initial value of the input
et al., 2019). parameter was more critical for the calculation of absolute
A binary linear-threshold segmentation was applied values, e.g., permeability, than for calculation of values that
to the same image volumes used in the ML-segmentation were normalized, e.g., electrical conductivity. The input
effort in order to allow for direct comparison of results. The parameter for grains was not activated, so the simulation
initial threshold value was evaluated visually on a single only considered the connected pore space. It was possible to
slice before submitting the entire set of slices for processing. de¿ne a very small value to the grain voxels, but this affected
A second, more subjective approach was performed by the computational ef¿ciency and required suf¿ciently more
incrementally increasing the threshold value on the image computing time. The choice of initial permeability input also
volume and calculating the image-based porosity. was important in reducing the number of computational steps
The image resolution of 2 and 4 ȝm made it dif¿cult required for convergence. The choice of initial conditions
to characterize subresolution pores associated with clay or input parameter values affect the success of the ¿nite-
minerals, and thereby any subsequent calculations of volume method solution of Navier-Stokes by reducing the
petrophysical properties (Liu et al., 2018). Consequently, a potential for large gradients between computation steps.
decision was made to combine the observable macropores There are a number of approaches to estimating permeability
with subresolution pores associated with the clay mineral from different measures of pore geometry, most of which
lining for the purpose of this study. The highly subjective are based on the Kozeny-Carman model (Arns et al.,
segmentation operation was repeated several times by two 2005). While these were useful to illustrate the connection
different operators with different backgrounds in geoscience between Àow properties and pore geometry, the actual input
and image processing. Each training exercise was saved and parameters to the Darcy Àow simulation model had to be
applied to a segmentation run, which in turn was used in adjusted to account for the resolved pore space in the image
simulations of various petrophysical properties. There was volume (Liu, et al., 2018).
no effort to train the image segmentation with core-derived For electrical conductivity, that input parameter for
porosity and clay-mineral abundance values. the pore space reduces to the brine conductivity, which,
A number of petrophysical properties were calculated when divided into the calculated conductivity, returns the
from these segmented images using direct numerical unitless formation factor. For the permeability simulation
simulations on the voxels (Zhang et al., 2011; Byrnes et al., the initial estimate of permeability for the pore component
2017, 2018). A spatial distribution representative elementary was dependent upon a correlation between pore size and
volume (REV) was calculated to con¿rm that the sample simulated permeability values. In this instance the value of
volumes were representative of the larger samples used the input value affected the calculated result, though there
for the conventional core measurements. The calculated was a linear correlation between input and output values that
properties included total volume fractions, pore-size could be related to connected image porosity and tortuosity
distributions, electrical conductivity (Ohm’s law solved with (Fig. 1). All image processing, quantitative analysis and
¿nite-volume method), effective permeability (multiphase image-based direct numerical simulations were conducted
Darcy’s equation solved with ¿nite-volume method), and with a cloud computing platform.

porosity and permeability observed in this interval. The

actual miniplug cut from the Sample D interval, however, cut
through a tight stringer of rock where all of the visible pore
space was ¿lled with the siderite cement and therefore was
not representative of the larger core plug used for original
petrophysical properties.
The micro-CT images of each sample illustrated
similarities and differences among the four samples (Fig.
2). The images taken from the central xz-plane of the
4-ȝm resolution 3D volume showed the variation in grain
(lighter features) and resolvable pore (dark areas) sizes
for the different samples. Samples A and B had distinctly
larger grains than Sample C. Sample D had no resolvable
Fig. 1—Comparison of input permeability value for segmented pore pores at 4-ȝm or at 2-ȝm resolution as the pore space was
component in a ML-based two-phase segmentation and the calculated ¿lled with a high-density siderite cement. Any measurable
permeability. The two points at input = 10 represent two image volumes,
a 400 × 400 × 400 cube and a larger 680 × 680 × 900 volume.
core properties for Sample D were controlled by pores
smaller than image resolution, hence its exclusion from the
subsequent workÀow. Smaller patches of siderite cement in
Uncertainty analysis can be applied to different steps in Samples A and C were oriented in the horizontal plane and
the workÀow of IBRP studies. This study focused on the error had some impact on transport property heterogeneity.
associated with subjective segmentation of the resolvable The NMR results showed a broad, bimodal distribution
pores in the image from the rest of the space occupied by of pores for the three samples, A, B and C, with roughly
grains. Uncertainty associated with binary segmentation with equivalent pore volume for each mode (Fig. 3). Conventional
a linear-threshold approach was evaluated by comparing interpretation of these T2 relaxation time distributions
image porosity extracted at different threshold levels and indicated that the component faster than 10 to 20 msec
identifying the threshold value of maximum change in the corresponded to microporosity while the slower component
resolvable porosity. Uncertainty associated with the ML represented the larger macroporosity. The conversion of
approach was evaluated by comparing the results of multiple relaxation times to a length scale used to compare with
segmentation runs on the same sample volume by different other estimates of pore-size distribution required the linear
operators with differing experiences and training. This study correlation term that represents the surface relaxivity (Dunn
did not consider issues associated with image quality and the et al., 2002). The standard approach uses the throat-size
importance of ¿ltering and noise removal. This study also distribution from MICP to match the NMR distribution.
did not evaluate the selection of values for input parameters For these samples, this method generated surface relaxivity
to various simulations of transport properties, though an values of 20 ȝm sec-1, within the range associated with
example was provided for the permeability simulation. sandstones. Comparison of pore versus throat sizes is not the
Finally, the uncertainty associated with the laboratory best way to optimize surface relaxivity calculations, but the
measurements used in this study was not addressed here. prevalence of MICP data in the literature has controlled this
discussion (Howard and Kenyon, 1992).
RESULTS Throat-size distributions generated from the MICP
drainage curve showed that for these samples the dominant
Routine core measurements for Samples A to C revealed throat size was in the range of the resolution of the micro-CT
a high-porosity sandstone, between 0.14 and 0.18, with a images, ~4 ȝm (Fig. 3). The distributions were predominantly
range of permeability values between 3 and 60 mD (Table unimodal, though Sample A had the suggestion of a second
1). Electrical conductivity for these three samples ranged mode of much smaller throat sizes than the dominant
between 0.10 and 0.23 mS/cm, which converted into mode. Note that the modal throat diameter was larger than
formation factor values of 45 for Sample B and 90 to 95 for the median size reported in Table 1. The long tails in the
Samples A and C. The saturation exponent for Sample B was distributions skewed towards the smaller throat sizes (higher
2.2 while Sample A was 2.3, and Sample C was 2.4. intrusion pressures) shifted the median value towards a
Sample D, with a porosity of 0.07 and 0.5 mD smaller size.
permeability, was selected to represent the lower range of

Howard et al.
Table 1—Basic Petrophysical Properties
(a) (b)
(c) (d)
Fig. 2—Representative micro-CT images of the four samples in this study, central xz-plane of the 3D volume. Images acquired at 4-ȝm resolution on
4-mm diameter cylindrical miniplugs in the dry state. (a) Sample A, (b) Sample B, (c) Sample C, and (d) Sample D.

(a) (b)
Fig. 3—(a) NMR T2 relaxation time distributions show predominantly bimodal distribution of pore sizes. (b) MICP throat-size distributions for samples
in this study. Vertical dashed-line represents the 4-ȝm resolution of the micro-CT images.
The results from the highest porosity and permeability (Fig. 4). The operators had different strategies to select seed
Sample B were the focus of the analysis of the micro-CT points for the ML-based segmentation of the low-intensity
images and subsequent simulations. All of the quantitative pore space (Table 2). Some chose short traces with only a
segmentation results and subsequent simulations on these few voxels while other runs included a larger number of
three samples were generated from ML-based segmentation. voxels. While the mean intensity value for each run was
Sample B was characterized by CT intensities that ranged roughly the same, the larger range of values sampled with
from 2,000 to 10,000 units, dominated by a mode around the longer traces produced a slightly broader histogram of
7,000 and a distinct shoulder on the distribution around 4,000 segmented macropores (Fig. 4).
Fig. 4—Histograms of segmented macropores from different runs compared to the original sample intensity distribution.

Howard et al.
Table 2—Seed Properties for Porosity in Sample B (4-ȝm Resolution Images)
Subtle features (arrow in Fig. 4) that represented pores Representative segmentation and simulation results
and pore throats close to the 4-ȝm image resolution could for each sample showed that the “best” image porosity of
only be picked up via iterative seeding. Note that all voxels the 4-ȝm resolution images was signi¿cantly less than
to the right of the dashed vertical threshold cutoff line in Fig. core-based results (Table 3). In general, the image-based
4 would be assigned as grain using a binary linear-threshold- macroporosity value from the 4-ȝm resolution images was 40
based method. to 70% less than the value from routine core measurements.
Quantitatively, different segmentations resulted in Image porosity for Sample B from different segmentation
different estimates of pore volume. Run 128 generated an iterations ranged from 0.103 to 0.132, which was 60 to
image porosity of 0.132, higher than the 0.103 and 0.110 70% of the total porosity of 0.175. The proportion of the T2
values for Runs 118 and 123. The difference in segmentation distribution’s slow component was ~50% of the total signal
of a larger pore volume was driven by a broader range of intensity, which suggested that the image segmentation
seed values, not the mean value of the selected voxels. captured most if not all of the larger macropores. The
Larger number of voxels from iterative, improved seeding connected macroporosity extracted from the segmentation is
increased computation time, but allowed the user to pick up approximately 90% of the image porosity, which factored
subtle features. It was apparent that additional seeding from into successful simulation of various transport properties.
Run 128 increased segmentation accuracy (Fig. 5). The pore-size distributions generated from the micro-
CT images were limited at the lower end by the resolution
of the measurement. The different segmentation efforts
did not generate large differences in calculated pore-size
distributions (Fig. 6). This was due, in large part, to the
similarity in observed pore volume and the limited range of
sizes detected by the micro-CT images. This was illustrated
by the different segmentation runs on Sample B and their
calculated pore-size distributions. The two segmentation
runs on 4-ȝm dataset, Runs 118 and 123, had no signi¿cant
difference in the calculated pore-size distribution and the
median size. The pore-size distribution associated with Run
(a) (b) 128 captured some larger pores as the segmentation included
more porosity that allowed the maximal-ball approach
to increase sphere size by a voxel or two. An additional
segmentation run on the 2-ȝm resolution, small volume
(400 × 400 × 400) had a distinct shift towards smaller pore
sizes (Fig. 7). The range of sizes was the same, roughly 1.5
orders of magnitude that illustrated how calculated pore-size
information was dependent upon image resolution. While
the lower-limit of detected pore widths is smaller for the
2-ȝm resolution images, the calculated distribution also lost
information at the larger end of the distribution. The limited
(c) (d)
range of observed pore sizes was illustrated by the calculated
Fig. 5—Comparison of image segmentation generated by different pore-size distributions for the three samples. Sample C had a
seeding strategies for Sample B, 4-ȝm resolution, 680 × 680 × 900 distinctly smaller pore size that reÀects the smaller grain size
volume, slice 450. (a) Raw CT; (b) Run 118, phi = 0.103; (c) Run 123,
phi = 0.110; (d) Run 128, phi = 0.132. observed with this sample relative to the other two samples.

Table 3—Simulation Results for Samples A to C
Z-Axis
(a)
Fig. 7—Comparison of pore-size distributions generated from 4- and
2-ȝm resolution 3D volumes for Sample B. Each size distribution covers
approximately one order of magnitude range in pore width, with the
2-ȝm resolution results starting at 10 ȝm and the 4-ȝm results starting
at 20 ȝm.
The overlay of the image-based pore-size distribution

and the slow component of the T2 distribution was observed
for all samples and illustrated for Sample B (Fig. 8). The
NMR relaxation times were converted to pore size by
applying a surface relaxivity value of 170 ȝm sec-1, so that the
slow-relaxation component matched the size distribution of
the resolved macropores obtained from the 4-ȝm resolution
micro-CT image. The MICP-based throat-size distribution
(b) was 1.5 orders of magnitude smaller than the pore-body
Fig. 6—Comparison of pore-size distributions generated from different
sizes computed from micro-CT images. The MICP-based
segmentation runs on 4-ȝm resolution images on Sample B (a) and for pore-throat-size distribution roughly corresponds to the size
the three samples (b). associated with the faster relaxation component based on the
same linear conversion of time to length scale.

Howard et al.
Fig. 8—Comparison of image-based pore-size distribution for Sample B (4-ȝm resolution) with NMR and MICP results converted to a common length
scale.
Calculation of transport properties using a ¿nite-volume patterns perpendicular to the z-axis that reduced calculated
method was dependent upon the amount of connected pore permeability by a factor of 2 to 3 compared to the measured
volume relative to the total-image porosity. Connected permeability on the adjacent routine core plug.
porosity values for each segmentation run for every sample Calculated electrical conductivity values were
in this study was > 90% of the total-image porosity value converted into formation factor by normalizing against the
(Table 3). These results were not seriously affected by the input conductivity value for the Àuid-¿lled macroporosity,
size of the computed volume (680 × 680 × 900 versus 400 × essentially brine conductivity. These calculated formation
400 × 400) nor by the resolution of the image (4 vs. 2 ȝm). factors are approximately twice the value obtained from
The calculated permeability values were dependent upon the core measurements on the adjacent routine core plugs. This
input permeability value for the porosity, which limited its difference reÀects the difference in core porosity, which
value as a predictive tool since this unitless selection was includes a signi¿cant microporosity component, and the
somewhat arbitrary. image porosity that captures only the larger macropores. The
The calculated Darcy permeability for Sample B was resultant formation factor and image porosity determined
85.2 mD in the z-axis, and 47.2 and 65.5 mD in the x and y the cementation exponents for these samples to be between
directions, respectively, when the input parameter was set at 1.9 and 2.5 (Table 3). The simulated electrical properties
5,000. This input parameter represented the permeability of for Samples A to C were compared with the measurements
a single voxel classi¿ed as a pore, therefore in the simulation on the adjacent large core plugs (Fig. 9). The shift in the
of the entire image volume this input value is reduced by the routine core-plug results towards larger porosity yet reduced
presence of nonpore voxels, the degree of pore connectivity formation factor generated a position on the crossplot with
and pore pathway tortuosity. This choice of input parameter similar slope as the simulated data. Replacing the total-
value for all three samples resulted in reasonable matches image porosity with connected-image porosity reduced the
to the measured permeability on the adjacent routine core scatter of points around a cementation exponent of 2.0. In
plugs. The calculated permeability for all three samples general, the calculated conductivities for Samples B and C
was slightly lower in the z-direction than in the x and y were higher in the z-direction than in the x and y directions,
axes. No large-scale laminations were observed in the though the increase was less than 10%. Sample A with its
micro-CT images for Samples B and C that might create distinct cementation pattern was less conductive in the z-axis
anisotropic Àow patterns that would inÀuence the simulation by a factor of 2.
results. In contrast, Sample A had distinct cementation

(a) (b)
Fig. 9—(a) Electrical conductivity simulation results for multiple iterations for Samples A to C compared to single-point measured core results. (b)
Comparison of macropores from image porosity and connected image porosity, 4-ȝm resolution images, illustrates importance of pore connectivity
in solving for transport properties.
Calculated MICP curves for these samples illustrate the micro-CT images and the presence of signi¿cant pore-¿lling
limitations of IBRP methods. The small range of pore sizes cement. The somewhat larger pores and throats in Sample
and explicitly resolved pore throats observed in the 4-ȝm C generated a calculated entry pressure that was more
resolution images was reÀected in the range of calculated agreement with the measured MICP curve, but the absence
pressures in the Hg/air system (Fig. 10). The 3D sample of small-pore information limited its value at higher capillary
volumes used in the simulations represent approximately pressures. Only with Sample B did the observed porosity
one-third of the sample size used in the actual MICP system generate a MICP curve that showed good agreement
measurement, and have idealized surfaces that reduce beyond the entry pressure and through part of the transition
artifacts from intrusion of surface-associated pores and zone, though at higher pressures the low Sw behavior was
openings observed in experiments. The calculated MICP incomplete. Calculations with the 2-ȝm resolution images
curve for Sample A did not begin to match the measured were no better in that while they included smaller pores and
entry pressure, in large part due to most of the pores (and throats associated with slightly higher pressures, they lost
throats) being smaller than the resolvable pores in the information with the larger pores.
(a) (b) (c)

Fig. 10—Comparison of MICP curves generated from experiment (red) and image-based calculations (blue) for the (a) low permeability (Sample A),
(b) high permeability (Sample B), and intermediate permeability (Sample C) samples.

Howard et al.
This project assumed ¿xed imaging resolution, but at the expense of adding a signi¿cant contribution of
image preprocessing, segmentation post-processing and single-voxel pores within the grains (Fig. 11). The dramatic
simulation, and that no true reference value existed. The increase in segmented porosity as the threshold exceeded
focus was on image segmentation of intergranular porosity 5,700 was caused in part by these single-voxel features
and the uncertainties generated by threshold segmentation within the grains. These features would likely be reduced if
and supervised ML segmentation in identifying intergranular preprocessing noise ¿ltering was applied to the image. As a
porosity. reference, the gray-scale range of the grain phase based on
Porosity was calculated with varying threshold gray- a single quartz particle was between 4,200 and 8,500, with
scale values that ranged between 4,000 to 7,000 for the a standard deviation of 1,230. The gray-scale range of pore
4-ȝm resolution images of Samples A, B and C (Zhang phase on a selected intergranular pore was between 2,300
et al., 2011). An increased threshold value improved the and 5,700, with a standard deviation of 980.
resolution of the intergranular pore space for Sample B,
Fig. 11—ML segmentation versus threshold segmentation for Sample B, 4-ȝm resolution images. (a) Gray-scale image with intensity range 945-
15,000; (b) ML segmentation, Run128; (c) threshold 4,500; (d) threshold 4,800; (e) threshold 5,100; (f) threshold 5,400; (g) threshold 5,700; (h)
threshold 6,000.

The comparison of resolved porosity as a function of

threshold value indicated two linear slopes, a lower slope
at gray-scale value of 5,200 or smaller (Fig. 12, line 1), the
other higher slope at gray-scale values of 6,200 or higher
(line 2). The gray-scale values between 5,200 and 6,200
represent the uncertainty zone for threshold segmentation.
The intersection of the two linear slopes de¿ned the location
of a perpendicular line from the point of maximum inÀection
in the intensity threshold curve at Point 3. The slopes 1 and
2 deviated from the curve at Points 4 and 5, respectively.
Points 3, 4 and 5 de¿ne the threshold porosity, and its upper
and lower limits (Table 4).
The threshold values measured with this method agree
well with the core porosity value. However, this good
agreement is misleading since the goal of this thresholding
exercise was to segment out intergranular porosity. The
thresholding workÀow, however, did not differentiate Fig. 12—Relationship between image porosity and intensity for all three
samples.
microporosity from intergranular porosity, hence the two are
lumped into the same uncertainty zone. The good agreement
indicates that the threshold segmentation is effective in
estimating a bulk porosity with averaged effect from to allow higher resolution properties to be distributed.
microporosity, but erroneous in de¿ning the distribution of
the porosity. Consequently, using this segmentation to make DISCUSSION
any rock physics study is risky at best, and wrong most often.
Microporosity has distinctive physical properties that has to Validation of any segmentation process can be
be treated differently. This is reÀected by the large range of accomplished in a variety of ways. The simplest involves
uncertainties for all three samples. a visual inspection of the segmented results against the
The computation of ML-based segmentation resulted original image. While this approach can be done on an
in a statistical probability for each voxel. This probability image-by-image basis, this is not a satisfactory method
function was used to estimate uncertainties from ML for large stacks of images. This approach also lacks a
segmentation. These results illustrate how ML segmentation quantitative measurement that can be used to compare
reduced uncertainty, particularly on the boundary between studies. A second validation approach is to compare static
intergranular porosity and mineral grains. Uncertainties were properties extracted from the images, such as porosity or
slightly larger when differentiating intergranular porosity pore-size distribution, with physical measurements from
from microporosity. comparable core samples. The limitation to this approach
It is also important to note that the reference value was lies in the resolution of the images where signi¿cant features,
also subjective to the particular workÀow. The segmentation such as subresolution micropores or clay-sized grains, are
strategy can be very different for intergranular porosity not captured in the image or at least cannot be extracted
characterization and for multiscale upscaling. In the latter successfully. The most common example is the recognition
case, for example, porosity phase will need to be segmented in many studies that image porosity is always less than
into a volume fraction that is higher than the total porosity, core porosity. A third validation approach is to compare the
Table 4—Porosity and its Uncertainties Assessed With the Proposed Framework

Howard et al.
simulated results of dynamic properties, such as Àuid Àow, from different measurement methods make this ground truth
electrical or thermal conductivity, against laboratory results. unreliable in many cases. Furthermore, due to resolution
Again, the limitation to this approach lies in the inability to requirement, micro-CT and other high-resolution imaging
capture all of the critical features from the image to de¿ne a techniques often image a smaller subsample (or a subvolume
well-characterized model. The loss of information about the of the original sample), where the corresponding physical
smallest, or largest, feature can affect the simulations (Liu measurement does not exist.
et al., 2018). In the case of Darcy’s law Àuid Àow and the
accompanying permeability constant, the loss of information CONCLUSION
from the smallest pores has limited impact on the measured
or calculated Àuid Àow since Àow is dominated by the largest An outstanding challenge of IBRP is its uncertainty
pores. This is less the case with electrical conductivity where associated with various steps in the process, which limits
surface conduction effects, especially in the smallest pores, its adoption for practical engineering decision-making. Lack
contribute signi¿cantly to the total conductivity. of systematic effort to quantify uncertainty, and sometimes,
The IBRP workÀow involves the selection of multiple even recognize the importance of uncertainty, further
parameters associated with segmentation and simulation of underscores the gap that needs to be ¿lled by the research
transport properties. The decision to choose one parameter and development community. Understanding the uncertainty
for an early step inÀuences the parameters in the later quanti¿cation is required for both the developers of IBRP
steps. Often these choices made within the workÀow are technology and application scientists/engineers who use it.
arbitrary, such as the decision to emphasize segmentation In this project, a set of shaly sand samples with signi¿cant
for intergranular macroporosity that in turn determines what amount of authigenic chlorite/smectite that lined the larger
value was used to initiate the permeability calculation. In this pores was tested to identify uncertainty quanti¿cation (UQ)
study, the micro-CT images resolved intergranular porosity requirements associated with image-processing steps. We
in the effort to evaluate preliminary transport potential. focused on segmentation in particular.
Unresolved microporosity, which inÀuence storage and Images from a conventional micro-CT were limited in
saturation, would have required different segmentation the range of features that can be extracted. In this study, the
considerations. Several steps to evaluate uncertainty of size range was less than two orders of magnitude. The lower
resolvable intergranular porosity were followed in this size limit was de¿ned by image resolution of the instrument
study. Evaluation of the uncertainty of the microporosity and the upper limit controlled by sample size. Even with
was compounded by the absence of higher resolution data, these constraints, the image volumes in this study contained
such as those obtained with scanning electron microscopy ~2×105 pores, which was suf¿cient to provide a statistical
or synchrotron tomography, that made it much harder to representation of pores in these samples. The signi¿cant
quantity. This study also chose not to separate intergranular clay-mineral volume found lining the pores of these samples
porosity from microporosity by using an independent has a major impact on petrophysical properties because of
estimate of porosity partitioning, such as can be extracted its high surface area and interparticle porosity, yet it could
from NMR relaxation measurements. While this second not be evaluated in this instance due to the limitations in
approach to evaluate the uncertainty associated with the images. The strategy in this exercise was to focus on the
distinguishing two types of porosity has its adherents, the larger intergranular porosity and to leave the micropores
choice here was to focus on a combined porosity distinct associated with the clay minerals to a later study. As much
from the grains. as half of the pore volume in these sandstone samples was
Another challenge of uncertainty quanti¿cation was associated with subresolution micropores that complicated
the lack of an absolute ground truth. Often, in the ML their assignment in any pore-grain segmentation strategy.
segmentation literature, this is de¿ned by a manually Threshold segmentation can misleadingly reach porosity
segmented image that is compared with algorithm- matching the core-analysis data. The petrophysical properties
based segmentations. In the case of porosity, a helium- derived therein, however, will be completely erroneous. The
pycnometer-based measurement on a clean and dried core uncertainty framework presented in this paper reÀected the
plug is often the used as this ground truth, but comparison large, over 100% uncertainty.
with images is limited by their resolution. Ideally, a core ML segmentation limits uncertainty to 30% or lower.
plug would be imaged in its entirety at multiple resolutions, The impact of individual operator can also be quanti¿ed,
and then analyzed with multiple laboratory tests. Even if it which is within the 30% uncertainty range. It clearly offers
were possible to match imaging volumes with laboratory more Àexibility and does a better job dealing with pore-grain
measurement on the same sample, uncertainties that arise interfaces.

It is important to emphasize that segmentation is only URTEC-2901840 presented at the Unconventional Resources
one step among many in the IBRP workÀow. The uncertainty Technology Conference, Houston, Texas, USA, 23–25 July.
of the workÀow strongly dependent upon image quality, DOI: 10.15530/URTEC-2018-2901840.
image modality, image ¿ltering/preprocessing, segmentation Delerue, J., Perrier, E., Yu, Z., and Velde, B., 1999, New
Algorithms in 3D Image Analysis and Their Applications
post-processing, and the numerical simulation technique. As
to the Measurements of Spatialized Pore Size Distributions
the value of imaging is dictated by resolution, in order to in Soils, Physics and Chemistry of the Earth, Part A: Solid
extend the pore-size information from image-based methods Earth and Geology, 24(7), 639–644. DOI: 10.1016/S1464-
in order to match the 3 to 4 orders of magnitude range 1895(99)00093-9.
associated with NMR methods, other imaging tools must be Dunn, K-J., Bergman, D., and LaTororraca, G., 2002, Nuclear
added. The most obvious inclusion is SEM (2D and/or 3D), Magnetic Resonance: Petrophysical and Logging
where upscaling workÀow and combining various datasets Applications, Pergamon Press. ISBN: 978-0444542533.
are applicable (Byrnes et al., 2018). Howard, J., and Kenyon, W., 1992, Determination of Pore-
Size Distribution in Sedimentary Rocks by Proton Nuclear
Magnetic Resonance, Marine and Petroleum Geology, 9(2),
139–145. DOI: 10.1016/0264-8172(92)90086-T.
NOMENCLATURE
Idowu, N., Nardi, C., Long, H., Øren, P., and Bondino, I., 2013,
Improving Digital Rock Physics Predictive Potential for
Abbreviations Relative Permeabilities from Equivalent Pore Networks, Paper
IBRP = image-based rock physics SCA2013-017 presented at the International Symposium of
MICP = mercury injection capillary pressure the Society of Core Analysts, Napa Valley, California, USA,
ML = machine learning 16–19 September.
NMR = nuclear magnetic resonance Idowu, N., Nardi, C., Long, H., Varslot, T., and Øren, P., 2014,
Effects of Segmentation and Skeletonization Algorithms on
PNM = pore-network model
Pore Networks and Predicted Multiphase Transport Properties
REV = representative elementary volume of Reservoir Rock Samples, Paper SPE-166030, SPE
UQ = uncertainty quanti¿cation Reservoir Evaluation & Engineering, 17(4) 473–483. DOI:
10.2118/166030-PA.
Symbols Liu, T., Jin, X., and Wang, M., 2018, Critical Resolution and Sample
Size of Digital Rock Analysis for Unconventional Reservoirs,
SW = water saturation Energies, 11(7), 1798, DOI: 10.3390/en11071798.
Schembre-McCabe, J., and Kamath, J., 2017, Using Digital Rock
Technology to Quality Control and Reduce Uncertainty in
Relative Permeability Measurements, Paper SCA2017-014
REFERENCES presented at the International Symposium of the Society of
Core Analysts, Vienna, Austria, 28 August–1 September.
Al-Taie, A., Hahn, H.K., and Linsen, L., 2014, Uncertainty Schembre-McCabe, J., Salazar-Tio, R., Ball, G., and Kamath, J.,
Estimation and Visualization in Probabilistic Segmentation, 2011, A Framework to Validate Digital Rock Technology,
Computers & Graphics, 39, 48–59. DOI: 10.1016/j. Paper SCA 2011-028 presented at the International Symposium
cag.2013.10.012. of the Society of Core Analysts, Austin, Texas, USA, 18–21
Arns, C., Knackstedt, M., and Martys, N., 2005, Cross-Property September.
Correlations and Permeability Estimation in Sandstone, Witten, I, Eibe, F., Hall, M., and Pal, C., 2017, Data Mining:
Physical Revuew E, 72(4 part 2) 046304. DOI: 10.1103/ Practical Machine Learning Tools and Techniques, 4th Edition,
PhysRevE.72.046304. Morgan Kaufmann Series in Data Management Systems,
Byrnes, A., Zhang, S., Canter, L., and Sonnenfeld, M., 2017, Elsevier. ISBN: 978-0128042915.
Application of Integrated Core and 3D Image Rock Physics Wu, J.C., Halter, M., Kacker, R.N., Elliott, J.T., and Plant, A.L.,
to Characterize Niobrara Chalk Properties Including 2013, Measurement Uncertainty in Cell Image Segmentation
Relative Permeability With Bound Water Effect, Paper Data Analysis, NIST Report NISTIR 7594. DOI: 10.6028/
URTEC-2670963 presented at the Unconventional Resources NIST.IR.7954. https://nvlpubs.nist.gov/nistpubs/ir/2013/
Technology Conference, Austin, Texas, USA, 24–26 July. NIST.IR.7954.pdf. Accessed March 20, 2019.
DOI: 10.15530/URTEC-2017-2670963. Zhang, S., Byrnes, A., Jankovic, J., and Neilly, J., 2019,
Byrnes, A., Zhang, S., Canter, L., and Sonnenfeld, M., 2018, Management, Analysis and Simulation of Micrographs with
Application of Integrated Core and Multiscale 3D Image Rock Cloud Computing, Microscopy Today, 27(2), 26–33. DOI:
Physics to Characterize Porosity, Permeability, Capillary 10.1017/S1551929519000026.
Pressure, and Two- and Three-Phase Relative Permeability Zhang, S., Klimentidis, R.E., and Barthelemy P., 2011, Porosity
in the Codell Sandstone, Denver Basin, Colorado, Paper and Permeability Analysis on Nanoscale FIB-SEM

Howard et al.
Tomography of Shale Rock, Paper SCA2011-080 presented at

the International Symposium of the Society of Core Analysts,
Austin, Texas, USA, 18–21 September.
Zhang, S., Zabusky, N.J., Peng, G., and Gupta, S. 2004, Shock
Gaseous Cylinder Interactions: Dynamically Validated
Initial Conditions Provide Excellent Agreement Between
Experiments and Numerical Simulations to Late-Intermediate
Time, Physics of Fluids, 16(5), 1203–1216. DOI:
10.1063/1.1651483.
ABOUT THE AUTHORS
James J. Howard is a technical advisor to DigiM. He

has worked in research and technology groups in the service
sector, operating companies and academia throughout his
career. He has a PhD in geology.
Sam Sheng Lin is a Software Development Manager

with DigiM. As one of the chief architects of cloud-based
image processing platform, Sam leads the design and
development of DigiM I2S. Sam also champions AI-based
image processing since he joined DigiM in 2015. Sam holds
a computer science degree from Qingdao University of
Science and Technology.
Shawn S.A. Zhang is founder and managing partner

of DigiM. Shawn graduated from Rutgers University with
a PhD in Computational Physics and a minor in Computer
Engineering. Prior to starting Boston-based DigiM in 2014,
Shawn held senior positions at Fluent, Ansys and FEI, and
played instrumental roles in software development and
management of sales and application teams. Shawn has
several patents and hundreds of publications and public
presentations.

A Review of 60 Years of NMR Wettability1

Andrea Valori2 and Benjamin Nicot3
ABSTRACT
Wettability is a key parameter in the development of faster, allowing much faster turnaround of laboratory
an oil¿eld as it strongly affects oil saturations, capillary measurements, and can be measured in-situ downhole,
pressures, electrical properties, relative permeabilities with the result of the measurement being available in real
and oil recovery. Despite attempts made to evaluate time. These extreme advantages fueled the research on the
wettability downhole, the standard methods to quantify it topic of NMR wettability despite the above-mentioned
are still laboratory based; the two most commonly used are dif¿culties.
Amott-Harvey (AH) and US Bureau of Mines (USBM). There are at least three main NMR parameters
These techniques are expensive and very time-consuming, measurable downhole: T1, T2 and diffusion; with additional
requiring a sample to be retrieved from the well and information extractable from the correlation between these
analyzed in the laboratory. In several cases, the results are three. Wettability affects all of these parameters, and the
obtained late and only after several decisions regarding the correlation between them. This means that there is not a
reservoirs had to be made, without this important piece of single way to extract wettability information from NMR
information. data, but there are different options.
It is ubiquitously recognized that nuclear magnetic Here, we review 60 years of literature on the topic
resonance (NMR) is very sensitive to the strength of the of NMR and wettability, from the ¿rst experimental
Àuid-rock interactions, and therefore has been considered observations in the 1950s to the most recent advancements.
as a good candidate for wettability determination since Also, this work aims at presenting strengths and limitations
the 1950s. The NMR signal, however, is also sensitive of the techniques being developed nowadays, to help the
to several other Àuid and rock properties, for example audience make the best choice for each speci¿c case. In
viscosity and pore-size distribution, making the practical this paper, we discuss both laboratory- and log-based
extraction of wettability information from NMR data not applications, although we place greater emphasis on
straightforward. NMR has, however, two considerable laboratory-based applications.
advantages compared to AH and USBM: it is much
INTRODUCTION complexity arises, is in the quanti¿cation and measurement

of this property. For extremely simple and perfect geometries,
Wettability is a key parameter for a reservoir, it strongly such as a capillary tube or a smooth and Àat solid surface, the
affects residual oil saturations, imbibition capillary pressure contact angle can be de¿ned. This angle is an extremely good
curve, electrical properties, relative permeabilities and oil and fundamental quanti¿cation of the wetting properties of
recovery. The knowledge of wettability is crucial to correctly the surface. Still, geometrical imperfection and roughness of
interpret petrophysical properties, to correctly model the the surface may cause the measurement to be uncertain in
reservoir, and ultimately to make reservoir management practice. In the case of complex microscopic geometry, or
decision, which can make the difference between success or solids with multiple phases, as is the case of porous media
failure of a reservoir (Anderson, 1987). and rocks, the contact angle becomes impractical to measure
The de¿nition of wettability in principle is simple: “the accurately, and therefore loses its reliability.
ability of a liquid to maintain contact with a solid surface in Wettability strongly affects the Àow of Àuids in the
the presence of a third phase.” This is a simple and correct rock, as well as the distribution geometry and connectivity
de¿nition. However, it is totally qualitative. In fact, where the of the Àuids in the rock. Because resistivity measurements
Manuscript received by the Editor November 20, 2018; manuscript accepted December 20, 2018.
1
Paper SCA2018-017.
2
Schlumberger SDCR, Dahrhan, Saudi Arabia
3
TOTAL CSTJF, Avenue Larribau, Pau, France

A Review of 60 Years of NMR Wettability
are dependent on the connectivity of the Àuid phases different ways. This spectrum of possibilities leads to the
(speci¿cally water), resistivity is also dependent on fact that there is not a single NMR wettability technique
wettability. Simply described, wettability is one of several for which “one size ¿ts all”. Rather, there are several
critical factors that affect the Archie equation (Eq. 1) different techniques, and each one has its own strengths and
parameter n, which represents the variation of resistivity as limitations. The aim of this paper is to give the reader an
the saturation changes. This is equivalent to saying how the overview of these techniques developed over six decades
resistivity changes for different wettability values at constant of studies, without any claim of completeness, therefore we
saturation. refer to the references for details, where real examples and
case studies are shown.
(1)
INDUSTRY STANDARD (USBM and A-H)
where Sw represents the water saturation, Rw is the resistivity The A-H imbibition and USBM tests, or a combination
of the water phase, ‫ ׋‬is the porosity of the rock, and Rt is of both, are commonly used in the industry and widely
the the resistivity of the sample. The cementation factor, accepted for use with core-plug samples. In the A-H test, a
m, expresses how the resistivity is a function of porosity. sample at irreducible water saturation placed into a water-
Resistivity measurement is the oldest downhole logging ¿lled tube spontaneously imbibes water over a period of
technique, and still brings a lot of value, especially for its time. Then the sample is placed in a Àow cell and water is
wide range of depth of investigation (DOI). forced through, with the additional oil recovery noted. The
The oil industry developed robust laboratory techniques sample is now at residual oil saturation and the process
to extract meaningful wettability information from rock core is repeated with an oil-¿lled imbibition tube and then an
plugs, such as the US Bureau of Mines (USBM) and Amott- oilÀooding apparatus. The results of the A-H test range
Harvey (A-H) wettability indices. The same is not true for between +1 (strongly water-wetting) and –1 (strongly oil-
downhole applications, and, inferring wettability downhole wetting). In a USBM test, a centrifuge spins the core sample
is far from being a well-established answer product in the oil at stepwise increasing speeds starting at irreducible water
industry (Abdallah et al., 2007). saturation to residual oil saturation and then to another
Nuclear magnetic resonance (NMR) is well known to be series of measurements. The measurement range extends
extremely sensitive to the surface interactions between Àuids from + (strongly water-wetting) to – (strongly oil-
and solids. However, in an NMR measurement, the effects wetting), although most measurements are in the range of
of surface interactions are always averaged with the bulk +1 to í1. The centrifuge method is fast, but the saturations
properties. For this reason, NMR becomes an effective tool must be corrected because the centrifuge induces a nonlinear
for surface properties characterization only when the ratio of capillary pressure gradient in the sample. Even though the
surface of interaction to the volume of the Àuid (S/V) becomes USBM and A-H wettability indices do not always match,
large enough. This is exactly the case of porous media in they are accepted in the oil industry, considering there are no
general, and rocks in particular, which are the concern of the alternatives. These are laboratory techniques that cannot be
oil industry. NMR is an incredibly rich technique. In general applied downhole considering the required measurements.
terms, it spans from high-¿eld spectroscopy for molecular
characterization of compounds (used by chemists) all the NMR SENSITIVITY TO WETTABILITY
way to earth-¿eld measurements for underground aquifers,
passing through relaxation analysis used for materials and The ¿rst publications we could ¿nd that NMR relaxation
food characterization, Àow and diffusion measurements and for liquids is dependent on the surface-to-volume ratio date
fast-¿eld cycling for determination of molecular motion. back to 1956, and can be found in several articles in the
Although only a minimum subset of this extreme Bulletin of the American Physical Society (Brown, 1956;
portfolio is applicable to the landscape of oil industry, still, Korringa, 1956; Torrey, 1956). This initial work addressed
in petrophysics, the experimentalist has a wide choice of the relaxation enhancement of the Àuid on the surface layer
parameters and experiments. On one side, this freedom in terms of molecular dynamics only, without discussing the
introduces complexity and potential pitfalls if not carefully solid-Àuid interaction characteristic of the porous media,
managed, on the other end, this freedom allows optimization such as rocks. The key aspect in this work, that will be
of the experimental setup to maximize the value in different relevant for the application of NMR in porous media, is the
scenarios. Most, if not all NMR experiments are sensitive idea of the “fast exchange regime,” which happens when the
to surface interactions, and therefore wettability, but in diffusion across the Àuid body is fast relative to the relaxation

Valori and Nicot
time. In this case, despite the presence of two populations of where T2 is determined by several parameters, and wettability
molecules with different relaxation times (one for the surface is only one of the parameters, together with Àuid viscosity,
molecules and one for the bulk molecules), the measured rock relaxivity and pore-size distribution. This means that
relaxation time is a single value equal to the average of the extracting reliable and quantitative wettability information
relaxivities, weighted by the volume of molecules on the from a relatively simple T2 measurement requires several
surface (S·İ) and the volume of molecule in the bulk VíS·İ, additional pieces of information. This information can only
where S is the surface of the pore, İ the thickness of the be available in the laboratory, and downhole application
surface layer affected by the enhancement of relaxation, and of this technique is usually only done on a qualitative and
V the total volume of the pore: relative basis.
The great advantage of this technique is that allows
(2) discrimination of two conditions with the same overall
wettability, but where in one case the small pores are water-
wet and the big pores oil-wet or vice versa. The same applies
Strictly speaking, the volume of the molecule in the for overall saturation but oil in small pores and water in large
bulk is VíS·İ; however, since V >> S·İ, VíS·İ § V. Note that pores or vice versa. The underlying principle is to consider
Eq. 2 is valid for a single Àuid wetting the pore surface, and Eq. 3 as a function of pore size, where Sr-o is the “effective”
there is no mention of wettability alteration or multiple Àuid surface of interaction between oil and rock (which is taken
saturation yet. as proxy for pore-size dependent oil-wetness) and VporeSw is
The ¿rst report in the literature that discusses the the effective volume of water in the pore of given size (which
relationship between NMR relaxation parameters and leads to the pore-size dependent saturation). The intrinsic
wettability is Torrey (1956). Interestingly, publications relaxivity parameters, ȡw and ȡo, are also free parameters,
about measurement of wettability by NMR (Brown and Fatt, but only its ratio affects the model (Al-Muthana et al., 2012).
1956) appeared in the literature before the measurement of The practical way of applying this technique (Fig. 1) is
pore-size distribution (PSD) and microstructure properties, to build a forward model based on the different parameters
despite PSD being the main application nowadays. We brieÀy affecting T2, and then inverting it with a numerical method
summarized the differences, advantages and disadvantages to determine the wettability and saturation functions.
among the different NMR-based wettability techniques in
a previous paper (Valori et al., 2017) where we presented
the T1/T2 ratio technique. In this review, however, we will
go more into detail of each technique and focus on logging
applications.
T2-BASED WETTABILITY
The ¿rst advanced wettability technique we present in

this review is the one based on T2 only. It is based on the Fig. 1—Forward model underlying the T2-based wettability technique.
concept presented in the previous section, but moved a long The green boxes are measurements available (inputs) and the blue box
way forward since the early applications in the 1950s. The represents the inverted data.
main improvements of the recent techniques (Freedman et
al., 2002; Looyestijn and Hofman, 2006; Looyestijn, 2007; T2 VERSUS SATURATION
Al-Muthana et al., 2012, 2013) compared with the early
applications, are that more than a single pore size and Àuid A different way of exploiting Eq. 3 to extract wettability
saturation, and therefore T2 component, is considered. is to make measurements at different saturation states (but
the same wettability condition). This technique can be, in
a sense, considered a simpli¿cation and derivation of the
T2-based technique. In this case, the effective Àuid volume,
(3) V, changes with saturation, while the effective surface S
does not. For a nonwetting Àuid, the effective surface, S, is
equal to zero (Sr-o = 0 for 100% water-wet) and therefore
the second member of the addition in Eq. 3 vanishes, thus

making the observed T2 of the Àuid equal to the bulk value,

and independent from saturation. Examples of applications
of this technique can be found in the literature (Howard,
1998; Al-Mahrooqi et al., 2003; Johannesen and Steinsbø,
2006). Although mostly qualitative, this technique has the
potential to be applicable downhole, particularly when a tool
having multiple depths of investigation (DOI) covering a
range across the invasion pro¿le, is available.
RESTRICTED DIFFUSION
A more recent work focused on a technique applicable

downhole is based on restricted diffusion (Minh et al.,
2015). Restricted diffusion for a speci¿c Àuid appears on
experimental data when the size of the pore where the Àuid
is stored is smaller than the expected free diffusion path
for the speci¿c experimental parameters used. This means
that, within some physical limitations imposed by relaxation
parameters and Àuid bulk diffusion coef¿cient, it is possible Fig. 2—Dependency of the shape of the Padé line as a function of
relaxivity (rho) and tortuosity (m) (from Minh et al., 2015).
to adjust the experimental parameters to be able to see
restricted diffusion for different pore sizes. The expected
free-diffusion path for a molecule can be predicted from
displacement, free-diffusion coef¿cient and diffusion time
viscosity (determined by salinity for the water phase) and
are related, knowing two allows the third to be determined.
temperature. Combining the knowledge of the expected free-
Typically, the two known parameters are td and D, and the
diffusion parameter and the presence of restricted-diffusion
information extracted from restricted diffusion is a pore size
phenomena in the experimental data allows extraction of
derived from NMR data, but independent from relaxivity.
information on the true pore size, independent of relaxivity.
Note that, when the free-diffusion path is much longer than
Note that, because the expected free-diffusion path increases
the pore size, the data become dependent to the connectivity
to much more than the pore size, the information embedded
between pores (tortuosity, m), rather than the pore size.
in the restricted-diffusion data becomes the interconnectivity
Practically, the starting dataset for this type of analysis is
of the pores or “tortuosity”, which can be related to the
a diffusion-T2 dataset (D-T2) with visible effects of restricted
resistivity parameter m.
diffusion. The data dependency along the two axes are as
The Padé model (Hurlimann et al., 1994) combines
follow:
apparent relaxivity (rho), tortuosity (m), pore size (r) and
x diffusion: D0 and pore size
free-diffusion coef¿cient (D0) to predict the relationship
x T2: Pore size and apparent relaxivity.
between T2 relaxation time and measured diffusion (D). This
appears on the D-T2 dataset as a curved line Fig. 2. From the
The free-diffusion coef¿cient D0 can be calculated, or
apparent relaxivity, the wettability can be estimated (Minh et
at least estimated, from temperature and viscosity (for oil)
al., 2015):
or salinity (for water). Using the Padé model (Hurlimann
et al., 1994), the pore-size dependency can be eliminated
(4)
to determine the apparent relaxivity and from this the
wettability (Minh et al., 2015).
When a Àuid is in a closed environment (such as a The practical way of applying the workÀow in Minh et
pore in rocks), the maximum displacement is capped, and al. (2015) is to visually determine the location of water and
therefore is ǻx2. This means that, increasing the diffusion oil on a D-T2 map and ¿t restricted-diffusion lines (Fig. 2) to
time td over a certain value (dependent on the size of the the Àuid signals. This allows determination of the effective
environment) the linearity in Eq. 4 breaks. This is exactly relaxivity of oil and water, and from this, a determination of
the case called “restricted diffusion.” Since maximum wettability (see Minh et al., 2015, for a detailed mathematical
derivation).

Valori and Nicot
The limitations of this technique are the following: et al., 2005) and rocks of different mineralogy can be found
x It assumes uniform relaxivity across the pore spectrum. in the literature (Mitchell et al., 2009). Other authors refer
x It assumes that we are able to correctly measure the increased T1/T2 ratio to wettability effects in conventional
the diffusion coef¿cient for all relaxation times. rocks (Guan et al., 2002; Katika et al., 2017) longitudinal,
Limitations in measuring the diffusion coef¿cient T1, and transverse, T2, relaxation times should in principle be
depends on the technique used (i.e., pulsed ¿eld similar. However, internal magnetic ¿eld gradients related
gradients or constant gradients) but there is also a to minerals can shorten T2, as compared to T1, provided the
dynamic limit. saturating Àuid has high af¿nity to the solid. Consequently,
the T1/T2 ratio should quantify the af¿nity between the
T1 FREQUENCY DISPERSION—FIELD CYCLING mineral and wetting pore Àuid, so we estimate wettability
from logging data by comparing the T1/T2 ratio of oil and
There is an entire branch of NMR, called fast-¿eld water peaks in the reservoir zone to the T1/T2 ratio in the
cycling (FFC), NMR dispersion (NMRD), or even NMR water zone. We tested the hypothesis on core samples and
relaxometry, which may cause confusion with the ¿xed- used the predicted wettability to successfully determining
¿eld relaxometry that is based on the measurement and the elastic bulk modulus of samples containing oil and water.
interpretation of the dispersion of the relaxation time T1. In order to investigate the T2-shortening, we performed 1D
The demanding experimental conditions for this technique and 2D NMR experiments on samples of chalk, kaolinitic
make the required equipment complex and bulky and the sandstone, and chloritic greensand, saturated either with
sample size relatively small. All these conditions restrict this water, oil or oil/water at irreducible water saturation. The 1D
technique to laboratory applications and make the technique NMR experiment involved determination of T2 spectrum,
impossible to apply downhole. However, the theory whereas the 2D NMR experiments included determination
connecting motion at the microscale with NMR properties of T1-T2 and D-T2 maps, where D is the intrinsic diffusion
is general to any NMR application in petrophysics and coef¿cient. T2 spectra show that in all water-saturated
porous media, and therefore also useful for understanding samples, surface relaxation dominates; in oil-saturated chalk
downhole data. The theory of T1 dispersion is useful to and kaolinitic sandstone, bulk relaxation dominates; whereas
explain the T1/T2 ratio discussed below, which does have T2 of oil-saturated greensand shows surface relaxation in the
downhole applications. Further recommended literature on part of the spectrum representing chlorite. In all samples
the theory of T1 dispersion include Godefroy et al. (2001), with irreducible water saturation, water shows surface
Guan et al. (2002), Korb et al. (1997, 2014), McDonald et al. relaxation, whereas oil shows bulk relaxation. In line with
(2005), Mitchell et al. (2009), Singer et al. (2013), Nicot et this observation D-T2 maps of these samples show ¿eld-
al. (2015), and Katika et al. (2017). gradient effects in the oil, but not in the water indicating that
Overall, the T1 dispersion in frequency is an indicator of the water is trapped between solid and oil due to restricted
the molecular motions at the pore wall, therefore the higher diffusion. A T2 shortening will increase the T1/T2 ratio, so we
the dispersion, the higher the liquid/solid interactions at the use the T1/T2 ratio obtained from T1-T2 maps as a measure
pore wall, the stronger the wetting. of Àuid-mineral af¿nity. By this measure, the chalk shows
If this technique is incredibly helpful to understand the high af¿nity for water, the kaolinitic sandstone has no clear
physics involved in the dynamics of Àuids at liquid/solid preference for oil or water, whereas chloritic greensand
interfaces (i.e., wettability) it is quite complex to implement. shows different behavior for small and large pores. Small
Hopefully, it is interesting to note that a T1/T2 experiment pores (fast-relaxing components, or unconventional (Singer
performed at a single frequency is a good proxy for the T1 et al., 2013; Nicot et al., 2015), where the porosity in the
dispersion curve. It is therefore interesting to study T1/T2 as kerogen is expected to be strongly oil-wet.
a proxy for wettability. The simplest, but a fundamental explanation of the
Reports on the fact that the T1/T2 ratio depends on the increased T1/T2 ratio for wetting Àuid can be derived from
effective surface relaxivity or wettability are abundant in studies on the phenomenon of T1 dispersion. A detailed step-
literature. Many authors tie this parameter to the intrinsic by-step explanation is presented in Valori et al. (2017), here
properties of the surfaces, ¿rst and foremost the density of we will report only the schematic for the conceptual steps
paramagnetic particles. Applications on cement (McDonald (Fig. 3).

Fig. 3—Schematic of the conceptual steps allowing explanation of the increased T1 /T2 ratio for wetting Àuids from T1 dispersion data (from Valori et
al., 2017).
T1/T2 RATIO One drawback is that T1 measurements are intrinsically

more time-consuming than T2 since they require waiting
The most recent technique discussed in this review is for polarization. Furthermore, only a single point on the
based on the ratio between the two relaxation parameters T1 recovery curve can be detected for each scan, whereas
T1 and T2. It is well established in the NMR ¿eld that both several points on the T2 decay can be acquired for each scan.
these relaxation times are determined by the characteristic T2 is, therefore, always much better sampled, and therefore
molecular dynamics. The connection between dynamics and de¿ned, then T1. Within this physical limit, new NMR tools
relaxation times dates back to the work of Bloembergen, are making great progress towards improved determination
Purcell, and Pound (BPP theory) in 1948 (Bloembergen et al., of T1-T2 correlation on the move.
1948). BrieÀy, for molecules with fast anisotropic motion,
the two relaxation times T1 and T2 are equal, and therefore
T1/T2 = 1. For a molecule with complex or slow dynamics,
T1 and T2 become different, diverging as the motion gets
slower and slower. In fact, for crystal and solid materials,
T2 is typically extremely short and T1 is very long, several
orders of magnitudes apart. Note that in NMR jargon, “fast”
and “slow” are always related to the precession of the spins
(Larmor frequency). Translating the description into the
time domain, which is often a more immediate visualization,
“fast” dynamics means motion for which the characteristic
correlation time Ĳc is much shorter than the spin precession
period T = 1/f, where fL is the Larmor frequency.
In a recent paper (Valori et al., 2017), we used a
laboratory based workÀow to demonstrate how the oil-phase
T1/T2 ratio is strongly correlated with the industry standard
USBM wettability index (Fig. 4). This technique is based
only on relaxation times, and not diffusion measurements.
This makes the experimental part of the technique simple Fig. 4—Correlation between the average T1/T2 ratio for the oil phase and
and robust—ideally suited for downhole applications. the measured wettability index USBM* the asterisks indicate that the
USBM values were renormalized in the í1 to +1 interval rather than the
usual í to + (from Valori et al., 2017).

Valori and Nicot
The complexity into the applicability of the T1/T2 NOMENCLATURE

technique downhole is the separation of the two Àuids. In
fact, only one phase at the time must be analyzed. Using Abbreviations
the workÀow suggested for the laboratory in Valori et al. AH = Amott-Harvey wettability index
(2017) is not really feasible downhole for two reasons. First, NMR = Nuclear Magnetic Resonance
doping the water phase is laborious and requires complete USBM = United States Bureau of Mines wettability
invasion for the entire investigated volume, which is not index
always veri¿ed when the formation is tight or the mudcake USBM* = USBM index modi¿ed to be in the range [-1;1]
buildup very effective. Secondly, the T1/T2 ratio for the oil
may deviate from unity due to intrinsic bulk-oil properties.
Symbols
Interpreting this deviation as due to wettability would lead to
incorrect conclusions. A more robust approach would be to D= diffusion coef¿cient
interpret the deviation from unity of the water phase, which D0 = free diffusion coef¿cient
is expected to always have T1/T2 = 1 for ideal nonwetting M= cementation factor
conditions. Because it is not possible to dope the oil phase, Rt = resistivity measured on a given sample
other “dopant-free” techniques are required to separate the Rw = water resistivity
Àuids. A workÀow based on diffusion was presented in S= surface
Valori et al. (2017). An alternative approach can be the use So = oil saturation
of a visual recognition technique of the kind presented in Sr-o = effective surface of interaction between oil
Anand (2017). and pore wall
Sr-w = effective surface of interaction between water
CONCLUSIONS—COMPARISON, STRENGTHS and pore wall
AND LIMITATIONS Sw = water saturation
td = diffusion time
This paper provides an overview of the different T1 = NMR longitudinal relaxation time
techniques and methods in the literature used to extract T2 = NMR transverse relaxation time
wettability information from NMR data. Some techniques, V= volume
for example ¿eld cycling, have virtually no applicability ȡ= relaxivity
downhole due to the equipment required; others, such as the ‫=׋‬ porosity
one based on T2, requires extremely good knowledge of the
Àuids and rock sample, knowledge that is only available in REFERENCES
the laboratory. In our opinion, the most promising techniques
having downhole applicability are the ones based on Abdallah, W., Buckley, J.S., Carnegie, A., Edwards, J., Herold, B.,
saturation, on T1/T2 ratio, and on restricted diffusion. Each Fordham, E., Graue, A., Habashy, T., Seleznev, N., Signer,
one of these techniques has strengths and limitations, which C., Hussain, H., Montaron, B., and Ziauddin, M., 2007,
are summarized in Table 1. Shales have not been discussed Fundamentals of Wettability, Schlumberger Oil¿eld Review,
in this paper, since the entire concept of wettability in shales 19(2), 44–61.
Al-Mahrooqi, S., Grattoni, C.A., Moss, A.K., and Jing, X.D.,
is very debatable if not meaningless, with oil in oil-wet pores
2003, An Investigation of the Effect of Wettability on NMR
and water in water-wet pores. Characteristics of Sandstone Rock and Fluid Systems, Journal
of Petroleum Science and Engineering, 39(3–4), 389–398.
DOI: 10.1016/S0920-4105(03)00077-9.
Table 1—Strengths and Limitations of the Different Techniques Applicable Downhole

Al-Muthana, A.S., Hursan, G.G., Ma, S.M., Singer, P.M., Nicot, Symposium of the Society of Core Analysts, Trondheim,
B., Valori, A., Ali, F., and Bachman, N., 2013, Methods for Norway, 12–16 September.
Determining Wettability from NMR, US Patent 9,405,037, Katika, K., Saidian, M., Prasad, M., Fabricius, I.L., 2017,
granted August 2, 2016. Low-Field NMR Spectrometry of Chalk and Argillaceous
Al-Muthana, A.S., Hursan, G.G., Ma, S.M., Valori, A., Nicot, B., Sandstones: Rock-Fluid Af¿nity Assessed from T1/T2 Ratio,
and Singer, P.M., 2012. Wettability as a Function of Pore Size Petrophysics, 58(2), 126–140.
by NMR, Paper SCA2012-031 presented at the International Korb, J.-P., Nicot, B., Louis-Joseph, A., Bubici, S., and Ferrante,
Symposium of the Society of Core Analysts, Aberdeen, G., 2014, Dynamics and Wettability of Oil and Water in Oil
Scotland, UK, 27–30 August. Shales, Journal of Physical Chemistry C, 118(40), 23212–
Anand, V., 2017, Novel Methodology for Accurate Resolution 23218. DOI: 10.1021/jp508659e.
of Fluid Signatures From Multi-Dimensional NMR Well- Korb, J.P., Whaley-Hodges, M., and Bryant, R.G., 1997,
Logging Measurements, Journal of Magnetic Resonance, Translational Diffusion of Liquids at Surfaces of Microporous
276, 60–68. DOI: 10.1016/j.jmr. 2017.01.007. Materials: Theoretical Analysis of Field-Cycling Magnetic
Anderson, W.G., 1987, Wettability Literature Survey—Part 6: The Relaxation Measurements, Physical Review E, 56, 1934–
Effects of Wettability on WaterÀooding, Paper SPE-16471, 1945.
Journal of Petroleum Technology, 39(12), 1605–1622. DOI: Korringa J., 1956, Interpretation of Data on Nuclear Spin
10.2118/16471-PA. Relaxation in Liquids for Large Surface-to-Volume Ratios,
Bloembergen, N., Purcell, E.M., and Pound, R. V, 1948, Relaxation Bulletin of the American Physical Society, Series II, 216.
Effects in Nuclear Magnetic Resonance Absorption, Physical Looyestijn, W.J., 2007, Wettability Index Determination From
Review, 73, 679–712. DOI: 10.1103/PhysRev.73.679. NMR Logs, Petrophysics, 49(2), 130–145.
Brown, R.J.S., 1956, Measurements of Nuclear Spin Relaxation Looyestijn, W., and Hofman, J., 2006, Wettability-Index
of Fluids in Bulk and for Large Surface-to-Volume Ratios, Determination by Nuclear Magnetic Resonance, Paper SPE-
Bullettin of the American Physical Society, Series II, 216. 93624, SPE Reservoir Evaluation & Engineering, 9(2), 12–
Brown, R.J.S., and Fatt, I., 1956, Measurements of Fractional 15. DOI: 10.2118/93624-PA.
Wettability of Oil Fields’ Rocks by the Nuclear Magnetic McDonald, P.J., Korb, J.P., Mitchell, J., and Monteilhet, L., 2005,
Relaxation Method, Paper SPE-743-G presented at the Fall Surface Relaxation and Chemical Exchange in Hydrating
Meeting of the Petroleum Branch of AIME, Los Angeles, Cement Pastes: A Two-Dimensional NMR Relaxation
California, 14–17. Society of Petroleum Engineers. DOI: Study, Physical Review E, 72, 011409. DOI: 10.1103/
10.2118/743-G. PhysRevE.72.011409.
Freedman, R., Heaton, N., Flaum, M., Hirasaki, G.J.J., Flaum, Minh, C.C., Crary, S., Singer, P.M., Valori, A., Bachman, N.,
C., and Hürlimann, M., 2002, Wettability, Saturation, Hursan, G.G., Ma, S.M., Belowi, A., and Kraishan, G., 2015,
and Viscosity Using the Magnetic Resonance Fluid Determination of Wettability From Magnetic Resonance
Characterization Method and New Diffusion-Editing Pulse Relaxation and Diffusion Measurements on Fresh-State
Sequences. Paper SPE-77397 presented at the SPE Annual Cores, Paper III, Transactions, SPWLA 56th Annual Logging
Technical Conference and Exhibition, San Antonio, Texas, Symposium, Long Beach, California, USA, 18–22 July.
USA, 29 September–2 October. DOI: 10.2118/77397-MS. Mitchell, J., Hürlimann, M.D., and Fordham, E.J., 2009, A Rapid
Godefroy, S., Korb, J.P., Fleury, M., and Bryant, R.G., 2001, Measurement of T1/T2: The DECPMG Sequence, Journal
Surface Nuclear Magnetic Relaxation and Dynamics of Water of Magnetic Resonance, 200(2), 198–206. DOI: 10.1016/j.
and Oil in Macroporous Media, Physical Review E, 64, 1–13. jmr.2009.07.002.
DOI: 10.1103/PhysRevE.64.021605. Nicot, B., Vorapalawut, N., Rousseau, B., Luis, F., Hamon, G.,
Guan, H., Brougham, D., Sorbie, K.S., and Packer, K.J., 2002, and Korb, J., 2016, Estimating Saturations in Organic Shales
Wettability Effects in a Sandstone Reservoir and Outcrop Using 2D NMR, Petrophysics, 57(1), 19–29.
Cores From NMR Relaxation Time Distributions, Journal of Singer, P.M., Rylander, E., Jiang, T., McLin, R., Lewis, R.E., and
Petroleum Science and Engineering, 34(1–4), 35–54. DOI: Sinclair, S.M., 2013, 1D and 2D NMR Core-Log Integration
10.1016/S0920-4105(02)00151-1. in Organic Shale, Paper SCA 2013-018 presented at the
Howard, J.J., 1998, Quantitative Estimates of Porous Media International Symposium of the Society of Core Analysts,
Wettability From Proton NMR Measurements, Magnetic Napa Valley, California, USA, 16–19 September.
Resonance Imaging, 15(5–6), 529–533. DOI: 10.1016/S0730- Torrey, H.C., 1956. Theory of Nuclear Spin Relaxation of Liquids
725X(98)00060-5. for Large Surface-to-Volume Ratios, Bullettin of the American
Hurlimann, M.D., Helmer, K.G., Latour, L.L., and Sotak, Physical Society Series II, 216.
C.H., 1994, Restricted Diffusion in Sedimentary Rocks. Valori, A., Ali, F., and Abdallah, W., 2018, Downhole Wettability:
Determination of Surface-Area-to-Volume Ratio and Surface The Potential of NMR, Paper SPE-190332 presented at the
Relaxivity, Journal of Magnetic Resonance, Series A., 111(2), SPE EOR Conference at Oil and Gas West Asia, Muscat,
169–178. DOI: 10.1006/jmra.1994.1243. Oman, 26–28 March. DOI: 10.2118/190332-MS.
Johannesen, E., Steinsbø, M., Howard, J., and Graue, A., 2006, Valori, A., Hursan, G., and Ma, S.M., 2017, Laboratory and
Wettability Characterization by NMR T2 Measurements in Downhole Wettability From NMR T1/T2 Ratio, Petrophysics,
Chalk, Paper SCA2006-039 presented at the International 58(4), 352–365.

Valori and Nicot
ABOUT THE AUTHORS
Andrea Valori is NMR team leader for Interpretation

Engineering at the Schlumberger engineering centre
in Houston. He holds a PhD in NMR physics from the
University of Surrey (UK) and worked for eight years in
the Schlumberger research center in Saudi Arabia. His work
focuses on NMR applications to porous media (cement while
at the University, and rocks since joining Schlumberger). His
present work is focused on the processing and integration
of laboratory and downhole data to improve petrophysical
interpretation. Andrea authored several papers and patents in
the ¿eld of NMR and is reviewer for scienti¿c and technical
international journals.
Benjamin Nicot, now petrophysicist at Total E&P,

is involved in formation evaluation. He holds a PhD in
NMR from IFP, France and worked for seven years for
Schlumberger as an NMR senior research scientist, and
a petrophysics domain champion. Specialized in NMR,
both laboratory and log, he is also involved in formation
evaluation techniques in general, and log interpretation.

A New WaterÀood Initialization Protocol With Wettability Alteration for Pore-Scale

Multiphase Flow Experiments1
Qingyang Lin2, Branko Bijeljic2, Samuel C. Krevor2, Martin J. Blunt2, Maja Rücker2, Steffen Berg2,3, Ab. Coorn3,
Hilbert van der Linde3, Apostolos. Georgiadis3, and Ove B. Wilson3
ABSTRACT
In the context of digital rock analysis, pore-scale In this paper, a methodology for initializing multiple
imaging of multiphase Àow experiments using X-ray small rock samples to the same connate water saturation
microtomography can be used to obtain fundamental and wettability state has been developed by adopting
insights into pore-scale displacement physics. This best SCAL practices, namely the porous-plate method or
provides a basis to better calibrate numerical pore-scale centrifugation using crude oil, followed by aging. We drill
simulators, or it can be used to understand local Àuid multiple small plugs from a full-size SCAL core sample,
distributions, while simultaneously measuring average without losing capillary continuity with the base of the
properties, equivalent to a traditional SCAL experiment. original sample. In the example presented, for Bentheimer
Imaging studies in the literature have historically sandstone, the initial saturation was established using
been conducted on small water-wet plugs, using kerosene, centrifugation. The experiment is designed to prevent a
or another re¿ned oil, as the non-wetting phase. Prior nonuniform saturation pro¿le in the small plugs. We use
to conducting waterÀood experiments, the initial water in-situ imaging to determine the water saturation after
saturation has been established by dynamic Àooding. The primary drainage and show that it is indeed uniform
disadvantage with this is that a nonuniform saturation across the sample with a value consistent with larger-
pro¿le is established due to the capillary end effect. This scale SCAL measurements and the measured mercury-
will result in a higher average initial water saturation injection capillary pressure. We also show that a signi¿cant
compared with, for instance, standard SCAL techniques, wettability alteration had occurred by measuring in-situ
such as the porous-plate method or centrifugation. contact angles.
INTRODUCTION (Andrew et al., 2014; AlRatrout et al., 2017; Scanziani et

al., 2017). These Àow studies have generally used water-wet
Flow experiments using X-ray microtomography (also rocks, where the initial water saturation was established by
called micro-CT) are commonly used to gain insights Àooding, and the waterÀood experiments were conducted
into rock properties and pore-scale displacement physics with inert model Àuids. Recently, some experiments have
(Armstrong et al., 2012; Berg et al., 2013; Aghaei and been conducted with crude oil to obtain a representative
Piri, 2015; Rücker et al., 2015; Khishvand et al., 2016b; reservoir wettability (Alhammadi et al., 2017; Bartels et al.,
Lin et al., 2016, 2017, 2018; Schlüter et al., 2016; Blunt, 2017a). This requires preparation of small plugs, typically 5
2017; Gao et al., 2017; Reynolds et al., 2017; Singh et al., to 12 mm in diameter and around 25 mm long.
2017) to calibrate numerical simulators (Dong and Blunt, The procedure by which samples are prepared,
2009; Blunt et al., 2013; Bultreys et al., 2015; Koroteev including oil injection to connate water saturation (Swc), is
et al., 2014; Scheibe et al., 2015; Norouzi Apourvari and expected to have a major impact on aging. The initial water-
Arns, 2016; Raeini et al., 2017) or to directly determine ¿lm thickness in the pore space affects the aging potential
wetting properties by measuring the in-situ contact angles of a given crude oil-rock-brine combination (Suijkerbuijk
Manuscript received by the Editor December 4, 2018; revised manuscript received February 13, 2019; manuscript accepted February 14, 2019.
1
Paper SCA2018-032.
2
Imperial College London, London, SW7 2AZ, United Kingdom; q.lin11@imperial.ac.uk; b.bijeljic@imperial.ac.uk; s.krevor@imperial.ac.uk;
m.blunt@imperial.ac.uk; m.rucker15@imperial.ac.uk
3
Shell Global Solutions International BV, The Netherlands; Steffen.Berg@shell.com; Ab.Coorn@shell.com; Hilbert.VanDerLinde@shell.com;
a.georgiadis@shell.com; O.Wilson@shell.com

A New WaterÀood Initialization Protocol With Wettability Alteration for Pore-Scale Multiphase Flow Experiments
et al., 2013; Bartels et al., 2017b). With thick water ¿lms An initial water saturation can be established by
or layers, the rock is expected to remain more water-wet the porous-plate method, centrifugation, or by dynamic
(Anderson, 1986), preventing any direct interaction between displacement (McPhee et al., 2015). Most oil companies
crude oil and rock. For thinner water ¿lms, polar crude-oil recommend the porous-plate method or centrifugation as
components may either diffuse through the water and adsorb the standard for SCAL. Dynamic Àooding is not usually
on the rock, or the ¿lm becomes unstable due to attractive recommended, since the capillary end effect is likely to lead
electrostatic forces between rock/brine and brine/crude oil to a nonuniform saturation distribution in the sample.
interfaces, which brings crude oil directly into contact with The porous-plate method is a direct technique, where a
the surface (Anderson, 1986). During a primary drainage water-saturated core and a ceramic plate are installed in an
process, the water distribution is controlled by the capillary isostatic core holder under stress (Kennedy, 2015; McPhee
pressure (Brady et al., 2015). For low (connate) water et al., 2015). The purpose of the ceramic plate is to prevent
saturation, the ¿lm thickness is controlled by the disjoining the nonwetting phase being produced during primary
pressure, as illustrated in Fig. 1. Depending on the salinity drainage. Primary drainage is conducted by imposing a
of the water phase and other parameters, the water-¿lm capillary displacement pressure limited by the entry pressure
thickness ranges between 100 and 1 nm (Dickinson et al., of the ceramic plate. The only disadvantage of the method
2016), which is below the imaging resolution of micro- is that it is time-consuming. This approach should not be
CT scanners (which is about 1 m). Such thin ¿lms have mistaken with the large porous-plate technique where
practically negligible hydraulic conductivity compared to multiple samples are placed on a large porous plate in a
the corner menisci (layers) which are much thicker and have pressure chamber without stress (Kennedy, 2015; McPhee et
long-range connectivity. The corner menisci provide the al., 2015). The centrifugation method (Hassler and Brunner,
main hydraulic conductivity for the water phase at low water 1945) is an alternative technique in which oil is forced to
saturation (Mohanty et al., 1987). Àow into a core at a speci¿ed rotational speed. This method
is fast and provides an effective method to reach connate
water saturation for permeable core plugs. However, for
low-permeability rocks, there may be a noticeable saturation
pro¿le in the core plug at the end of the experiment (McPhee
et al., 2015). For the dynamic displacement method, oil is
injected at constant pressure or rate to displace water in the
sample. The main disadvantage is that drainage coreÀoods
are affected by the capillary end effect, which cannot easily
be prevented, resulting in a nonuniform and higher average
initial saturation than other methods (McPhee et al., 2015).
In micro-CT Àow experiments, dynamic displacement
by high Àow-rate Àooding is currently the prevailing
approach for pore-scale imaging. The initial wetting-phase
(a) saturation is established by injecting the nonwetting phase
(crude oil) into the core at an incrementally increasing Àow
rate (Khishvand et al., 2016a), followed by in-situ aging to
restore rock wettability inside the core holder at reservoir
conditions for about four weeks (Alhammadi et al., 2017).
The main limitations for this method are two-fold. Firstly,
the established initial water saturation is not necessarily the
connate water saturation. Furthermore, the distribution of
initial water saturation, from inlet to outlet end, is nonuniform
in most cases. This is mainly caused by the capillary end
effect where the wetting phase is retained near the outlet
of the sample. Consequently, water ¿lms and layers are too
thick to allow for ef¿cient aging (Brady et al., 2015) and the
wettability distribution may not be representative, particularly
(b)
tending to understate the degree of wettability alteration. The
Fig. 1—(a) Water-layer thickness controlled by capillary pressure. results from micro-CT Àow experiments therefore cannot be
(b) Water-¿lm thickness controlled by disjoining pressure.

Lin et al.
directly compared with SCAL measurements. Secondly, this added potassium iodide (KI). In this study, the concentration
method is time-consuming since it occupies Àooding-loop of KI is approximately 3.5 wt%. The oil phase was decalin
equipment throughout Àooding and aging. (cis and trans decahydronaphthalene mixture with a dynamic
In this paper, a methodology for initializing small plugs viscosity of 3 mPas at 20 qC, supplied by Alfa Aesar). We
for Àooding experiments has been developed by adopting performed a steady-state waterÀood experiment to assess the
best SCAL practices. Multiple millimeter-scale plugs wettability alteration. The viscosity of the oil, ȝo, was three
were drilled from a 38-mm (1.5-in.) core without breaking times higher than the brine.
capillary continuity between the distinct smaller plugs
and the base of the core. The bene¿t of this approach is a Table 1—Ion Concentration for the Brines Used
systematic and repeatable way of establishing Swc for all the
samples. This protocol is aligned with SCAL practice, which
can be used as validation for the pore-scale experiments. In
addition, it becomes possible to prepare a large number of
samples for micro-CT Àow experiments given the rock is
homogeneous. By deploying this protocol, up to 45 micro-
CT samples (up to 15 small cores for each 38-mm core) can
be prepared simultaneously in the same centrifuge batch
(three 38-mm cores for each batch), followed by aging. A
strict protocol needs to be followed with respect to mounting/
dismounting the small plugs in sleeves and core holders prior Preparation of the Small Plugs
to waterÀooding experiments in the micro-CT apparatus. Standard-size core samples were cleaned and saturated
The only disadvantage of the new initialization protocol with formation brine. Permeability and porosity can also be
is that dry scans of the small plugs cannot be obtained at measured on these samples. The micro-CT samples were
the beginning of the experiment. If a dry scan is needed to drilled to approximately 2.4-cm length from a larger core 3.8
assist segmentation, or needed as an input for modelling, it cm in diameter and 4 cm long. The diameter of the micro-
needs to be obtained at the end of all Àow experiments after CT cores used in this study was 6 mm. Figure 2 shows the
recleaning. samples, which remained attached to the base of the original
The water saturation was measured by imaging. The sample.
saturation was uniform along the core, with an average value
of 8%, consistent with independent SCAL experiments.
We also show that a signi¿cant wettability alteration had
occurred by measuring in-situ contact angles. In contrast,
dynamic Àooding in similar samples resulted in an average
saturation of 14%.
MATERIALS AND METHODS
Rock Samples and Fluid Properties

The rock sample used in this study was Bentheimer
sandstone (containing 98% quartz, 1% kaolinite/chlorite
and 1% microcline), the average porosity was 0.24, and the
brine permeability was 2.4 Darcy. For the initial primary
drainage step, the brine was comprised of deionized water
with various salts added, see Table 1. The sample was aged
using crude oil from a producing ¿eld in the Middle East,
supplied by Shell Global Solutions International BV.
The aged small cores were then extracted from the
larger core under crude oil (see later). The crude was
replaced by re¿ned oil, followed by waterÀooding. For
these experiments a doped brine was used to enhance the Fig. 2—Drilled small plugs. The plugs were 2.4 cm long and remained
attached to the base of the sample which retains capillary continuity
attenuation contrast between Àuid phases for imaging; we during primary drainage.

Then the entire arrangement was Àooded with crude oil 2. Verifying that the average Swc from the small core is
(McPhee et al., 2015). In this study, centrifugation was used. consistent with Swc from SCAL experiments, or with
Since the small plugs were attached to the core base during mercury-injection measurements.
centrifugation, nonuniformities in saturation were largely
con¿ned to the 1.6 cm at the base of the rock that was not In addition, to demonstrate that wettability alteration
part of the small plugs. For low-permeability material it is occurred, a steady-state waterÀood experiment was
therefore important to break capillary continuity as soon as performed on a small plug sample, and compared to an
centrifugation is completed, before continuing with aging, to experiment where initial conditions were established by
prevent redistribution caused by releasing capillary pressure. dynamic Àooding without aging (Gao et al., 2017).
In this case, using high-permeability Bentheimer, this was
not necessary. Experimental Procedure for WaterÀooding Experiments
The experiment was conducted using the following
Primary Drainage procedure:
Speed design for centrifugation, or capillary 1. A micro-CT core (6 mm in diameter and 24 mm in
displacement pressure design for the porous-plate method length) was taken from the larger core. The sample
is based on converting mercury-injection capillary pressure was easily broken off from the larger core. The small
(MICP) measurements to equivalent oil/brine systems. In sample was then loaded into a micro-CT core holder
this study, two capillary displacement pressures were used surrounded by crude oil.
for the centrifuge. The centrifuge speeds are listed in Table 2. A con¿ning pressure of 2 MPa was applied and
2. From mercury-injection tests and SCAL experiments, it maintained within the cell to compress the Viton
was demonstrated that connate water saturation is reached sleeve around the core sample to avoid Àuid bypass.
at a capillary pressure of approximately 20 kPa. Therefore, 3. With a back pressure of 3 MPa, decalin was injected
a maximum speed of 3,200 rev/min, which represents an to replace crude oil in the system.
imposed capillary pressure of 60 kPa, was suf¿cient to reach 4. Decalin was injected at a Àow rate of 0.03 ml/min (fw
Swc. = 0). After reaching steady state, which was indicated
The crude oil injected into the system (TAN = 0.09 by differential pressure transducer measurements,
mgKOH/g, TBN = 0.270 mg/g, SARA analysis: Sat = 44.0, scans were taken to obtain the initial water saturation.
Aro = 44.0, Res = 9.69, and Asp = 2.31) had a density of 5. Brine and oil were both injected at the same time
0.85 g/cm3 and a viscosity around 8 to 10 mPa.s at room with a water fractional Àow fw = 0.5, with a total Àow
temperature. rate of 0.03 ml/min. Injection continued until steady
state was achieved as indicated by a stable pressure
Aging differential. Scans were taken after steady state was
After completing centrifugation, the entire rock reached.
assembly with multiple micro-CT cores was submerged 6. The differential transducer used in this study was a
in the same crude oil. The sample was kept in crude oil at Keller PD-33X, with an accuracy of ±0.3 kPa. A back
3-MPa pressure and 80qC for four weeks to complete aging. pressure of 3 MPa was applied throughout the entire
experiment.
Table 2—Centrifuge Speeds and Saturation for Primary Drainage
All the scans were taken using a Zeiss Versa 510 with
a Àat-panel detector. Before waterÀooding the entire sample
was imaged at a voxel size of 6.6 Pm. At fw = 0.5, a scan with
a smaller voxel size (3.58 Pm) was taken to characterize the
wettability. The wettability characterization was compared
with an unaged sample at steady state, again with fw = 0.5.
VALIDATION OF THE INITIALIZATION Initial Saturation

PROTOCOL The results of the image segmentation and the saturation
pro¿le for each slice (6.6 m per slice) of the oil phase for the
This initialization protocol was validated using the following sample prepared following the procedure above are shown in
criteria: Fig. 3. A uniform saturation can be observed and the average
1. Con¿rming by imagining that Swc is uniform along the water saturation for the sample is 0.082 (computed from the
entire small plug length.

Lin et al.
entire volume). This compares well with the average for all
the small cores plus the larger core base of 0.078 obtained
after centrifugation. Figure 4 shows the comparison between
the average Swc from the microcore with MICP.
Using the Àooding method to establish initial saturation
in a small core of Bentheimer gave an average value of 0.14
(Gao et al., 2017) which, as expected, is higher than that
established with our new protocol. The value we obtain
is, however, similar to that found on larger core samples
following standard SCAL protocols: in three replicate
experiments on entire 38-cm cores Swc values of 0.094, 0.080
and 0.068 were obtained, consistent with our values.
Fig. 4—Comparison between the average Swc (0.082) from the small
core with the scaled MICP. We imposed a capillary pressure of 60 kPa
in the centrifuge experiments.
Comparison of Results With and Without Aging

Figure 5 shows 2D gray-scale images extracted from a
1,0003-voxel cube image after applying a nonlocal means
edge preserving ¿lter for both aged and unaged samples.
The scans were taken after steady state was reached with a
fractional Àow of 0.5. It can be observed that, as expected,
in the unaged, water-wet case, the large pores are mainly
occupied by the nonwetting oil phase. The brine mainly
remains in the small pores. In the aged sample, the oil phase
is seen in both small and large pores. A negative capillary
pressure (Pc = Pnw – Pw, where Pnw is the pressure in the
nonwetting phase and Pw is the pressure in the wetting phase)
is inferred from the curvature of the brine and oil interface,
which indicates the wettability of the rock surface is oil-wet
(a) (b) (c) in many places. Parts of the rock surface still remains water-
wet, suggesting mixed-wet conditions.
Fig. 3—Tomographic image of a Bentheimer plug initialized to Swc, Figure 6 shows in-situ contact-angle measurements for
following the protocol described in this paper. (a) Gray-scale two-
dimensional (2D) cross section of the three-dimensional (3D) image of the highlighted region in Fig. 5b. Here we used the manual
the sample after applying a nonlocal means edge preserving ¿lter. (b) contact-angle method (Andrew et al., 2014). We see values
Segmented oil phase shown in blue. (c) The saturation per slice along greater than 90°, indicative of a signi¿cant wettability
the direction of Àow. The thickness for each slice is 6.6 m.
alteration, meaning that the sample preparation procedure
was effective. In contrast, for the unaged, water-wet sample,
the contact angles are all below 90°.

(a) (b)
Fig. 5—Gray-scale 2D cross sections of 3D images after applying a nonlocal means edge preserving ¿lter. The dimension of both images is 1,0003
voxel cubed and the voxel size is 3.58 m. In both ¿gures, the dark phase represents oil, the intermediate gray phase is brine and the bright phase
represents rock grains. The images are taken after steady-state waterÀooding as a fractional Àow of 0.5. (a) Unaged, water-wet case. (b) Images of an
aged sample, following the protocol in this paper. Some regions with contact angles larger than 90° with negative capillary pressure (from interfacial
curvature) in the brine phase are highlighted.
CONCLUSIONS
A novel methodology for preparing multiple core plugs saturation. We also con¿rmed that a substantial wettability
to the same desired representative initial saturation has been alteration had occurred through measuring in-situ contact
developed. The methodology is based on following best angles and observing negative interfacial curvatures during
SCAL practices for establishing Swc, i.e., the porous-plate waterÀooding.
method or centrifugation. The main concept is based on The main bene¿ts of deploying the new initialization
drilling several millimeter-scale plugs with approximately protocol can be summarized as follows:
2.4 cm length from a larger standard SCAL core sample x It secures a uniform saturation distribution as a
without breaking capillary continuity. starting point for waterÀooding, consistent with
The initialization protocol has been validated by SCAL requirements.
con¿rming a uniform saturation distribution along the entire x It is possible to prepare a large number of samples
sample length. Furthermore, the average water saturation simultaneously. Establishing Swc by centrifugation
is similar to that of the whole assembly of small plugs takes approximately 2 to 3 days, while the porous-
and the base, and on independent experiments on intact plate method takes approximately 4 to 5 weeks. After
full-sized cores. Dynamic aging leads to a higher initial this, core plugs need to be aged for four weeks.
Fig. 6—In-situ contact angles for the highlighted region in Fig. 5b were measured showing that the wettability of the rock surface had been altered. In
the ¿gures, the dark phase represents oil, the intermediate grey phase represents brine and the bright phase represents rock grains.

Lin et al.
x Applying the methodology for homogenous rocks Linking Pore-Scale Interfacial Curvature to Column-Scale
means that a sizeable number of small cores have the Capillary Pressure, Advances in Water Resources, 46, 55–62.
same initial properties. This gives unique possibilities DOI: 10.1016/j.advwatres.2012.05.009.
with respect to integration of different types of Bartels, W., Rücker, M., Berg, S., Mahani, H., Georgiadis, A.,
Fadili, A., Brussee, N., and Coorn, A., van der Linde, H., Hinz,
experiments in research.
C.,Jacob, A., Wagner, C., Henkel, S., Enzmann, F., Bonnin,
x Micro-CT Àooding units are not occupied for A., Stampanoni, M., Ott, H., Blunt, M., and Hassanizadeh,
preparation and aging of the cores. This means that S.M., 2017a, Fast X-Ray Micro-CT Study of the Impact of
it is possible to conduct more Àow experiments in Brine Salinity on the Pore-Scale Fluid Distribution During
research studies. WaterÀooding, Petrophysics, 58(1), 36–47.
Bartels, W.B., Rücker, M., Boone, M., Bultreys, T., Mahani, H.,
ACKNOWLEDGEMENTS Berg, S., Hassanizadeh, S.M., and Cnudde, V., 2017b, Pore-
Scale Displacement During Fast Imaging of Spontaneous
This work is part of the Digital Rock program between Imbibition, Paper SCA2017-005 presented at the International
Symposium of the Society of Core Analysts, Vienna, Austria,
Shell and Imperial College. We gratefully acknowledge
27 August–1 September.
Shell for ¿nancial and material support and permission to
Berg, S., Ott, H., Klapp, S.A., Schwing, A., Neiteler, R., Brussee,
publish this work. N., Makurat, A., Leu, L., Enzmann, F., Schwarz, J.-O.,
Kersten, M., Irvine, S., and Stampanoni, M., 2013, Real-
NOMENCLATURE Time 3D Imaging of Haines Jumps in Porous Media Flow,
Proceeding of the National Academy of Sciences, 110(10),
Abbreviations 3755–3759. DOI: 10.1073/pnas.1221373110.
SARA = saturates, aromatics, resins, and asphaltene Blunt, M.J., 2017, Multiphase Flow in Permeable Media: A Pore-
SCAL = Special Core Analysis Scale Perspective, Cambridge University Press, ISBN: 978-
1316145098, DOI: 10.1017/9781316145098.
TAN = total acid number
Blunt, M.J., Bijeljic, B., Dong, H., Gharbi, O., Iglauer, S.,
TBN = total base number
Mostaghimi, P., Paluszny, A., and Pentland, C., 2013, Pore-
Scale Imaging and Modelling, Advances in Water Resources,
Symbols 51, 197–216. DOI: 10.1016/j.advwatres.2012.03.003.
fw = fractional Àow of the wetting phase Brady, P.V., Morrow, N.R., Fogden, A., Deniz, V., Loahardjo, N.,
Swc = connate water saturation and Winoto, 2015, Electrostatics and the Low Salinity Effect
Pc = capillary pressure in Sandstone Reservoirs, Energy & Fuels, 29(2), 666–677.
DOI: 10.1021/ef502474a.
Bultreys, T., Van Hoorebeke, L., and Cnudde, V., 2015, Multi-
REFERENCES Scale, Micro-Computed Tomography-Based Pore Network
Models to Simulate Drainage in Heterogeneous Rocks,
Aghaei, A., and Piri, M., 2015, Direct Pore-to-Core Up-Scaling of Advances in Water Resources, 78, 36–49. DOI: 10.1016/j.
Displacement Processes: Dynamic Pore Network Modeling advwatres.2015.02.003.
and Experimentation, Journal of Hydrololgy, 522, 488–509. Dickinson, L.R., Suijkerbuijk, B.M.J.M., Berg, S., Marcelis,
DOI: 10.1016/j.jhydrol.2015.01.004. F.H.M., and Schniepp, H.C., 2016, Atomic Force Spectroscopy
Alhammadi, A.M., AlRatrout, A., Singh, K., Bijeljic, B., and Blunt, Using Colloidal Tips Functionalized With Dried Crude Oil:
M.J., 2017, In Situ Characterization of Mixed-Wettability in a A Versatile Tool to Investigate Oil–Mineral Interactions,
Reservoir Rock at Subsurface Conditions, Scienti¿c Reports, Energy & Fuels, 30(11), 9193–9202. doi:10.1021/acs.
7, 10753. DOI: 10.1038/s41598-017-10992-w. energyfuels.6b01862
AlRatrout, A., Raeini, A.Q., Bijeljic, B., and Blunt, M.J., 2017, Dong, H., and Blunt, M.J., 2009, Pore-Network Extraction From
Automatic Measurement of Contact Angle in Pore-Space Micro-Computerized-Tomography Images, Physical Revew
Images, Advances in Water Resources, 109, 158–169. DOI: E, 80(3, Part 2), 036307. DOI: 10.1103/PhysRevE.80.036307.
10.1016/j.advwatres.2017.07.018. Gao, Y., Lin, Q., Bijeljic, B., and Blunt, M.J., 2017, X-Ray
Anderson, W.G., 1986, Wettability Literature Survey–Part 1: Microtomography of Intermittency in Multiphase Flow
Rock/Oil/Brine Interactions and the Effects of Core Handling at Steady State Using a Differential Imaging Method,
on Wettability, Paper SPE-13932, Journal of Petroleum Water Resources Research, 53(12), 10274–10292. DOI:
Technology, 38(10), 1125–1144. DOI: 10.2118/13932-PA. 10.1002/2017WR021736.
Andrew, M., Bijeljic, B., and Blunt, M.J., 2014, Pore-Scale Hassler, G.L., and Brunner, E., 1945, Measurement of Capillary
Contact-Angle Measurements at Reservoir Conditions Using Pressures in Small Core Samples, Paper SPE-945114-G,
X-Ray Microtomography, Advances in Water Resources, 68, Transactions, AIME, 160, 114–123.
24–31. DOI: 10.1016/j.advwatres.2014.02.014. Kennedy, M., 2015, Practical Petrophysics, Developments in
Armstrong, R.T., Porter, M.L., and Wildenschild, D., 2012, Petroleum Science, 62, Elsevier. ISBN: 978-0444632708

Khishvand, M., Akbarabadi, M., and Piri, M., 2016a, Micro- Geophysical Research Letters, 42(10), 3888–3894. DOI:
Scale Experimental Investigation of the Effect of Flow Rate 10.1002/2015GL064007.
on Trapping in Sandstone and Carbonate Rock Samples, Scanziani, A., Singh, K., Blunt, M.J., and Guadagnini, A., 2017,
Advances in Water Resources, 94, 379–399. DOI: 10.1016/j. Automatic Method for Estimation of in Situ Effective Contact
advwatres.2016.05.012. Angle From X-Ray Micro Tomography Images of Two-Phase
Khishvand, M., Alizadeh, A.H., and Piri, M., 2016b, In-Situ Flow in Porous Media, Journal of Colloid and Interface
Characterization of Wettability and Pore-Scale Displacements Science, 496, 51–59. DOI: 10.1016/j.jcis.2017.02.005.
During Two- and Three-Phase Flow in Natural Porous Media, Scheibe, T.D., Perkins, W.A., Richmond, M.C., McKinley, M.I.,
Advances in Water Resources, 97, 279–298. DOI: 10.1016/j. Romero-Gomez, P.D.J., Oostrom, M., Wietsma, T.W.,
advwatres.2016.10.009. Serkowski, J.A., and Zachara, J.M., 2015, Pore-Scale and
Koroteev, D., Dinariev, O., Evseev, N., Klemin, D., Nadeev, Multiscale Numerical Simulation of Flow and Transport in a
A., Safonov, S., Gurpinar, O., Berg, S., van Kruijsdijk, C., Laboratory-Scale Column, Water Resources Research, 51(2),
Armstrong, R., Myers, M.T., Hathon, L., and de Jong, H., 1023–1035. DOI: 10.1002/2014WR015959.
2014, Direct Hydrodynamic Simulation of Multiphase Flow Schlüter, S., Berg, S., Rücker, M., Armstrong, R.T., Vogel,
in Porous Rock, Petrophysics, 55(4), 294–303. H.-J., Hilfer, R., and Wildenschild, D., 2016, Pore-Scale
Lin, Q., Al-Khulai¿, Y., Blunt, M.J., and Bijeljic, B., 2016, Displacement Mechanisms as a Source of Hysteresis for Two-
Quanti¿cation of Sub-Resolution Porosity in Carbonate Phase Flow in Porous Media, Water Resources Research,
Rocks by Applying High-Salinity Contrast Brine Using 52(3), 2194–2205. DOI: 10.1002/2015WR018254.
X-Ray Microtomography Differential Imaging, Advances Singh, K., Menke, H., Andrew, M., Lin, Q., Rau, C., Blunt, M.J.,
in Water Resources, 96, 306–322. DOI: 10.1016/j. and Bijeljic, B., 2017, Dynamics of Snap-Off and Pore-Filling
advwatres.2016.08.002. Events During Two-Phase Fluid Flow in Permeable Media,
Lin, Q., Bijeljic, B., Pini, R., Blunt, M.J., and Krevor, S., 2018, Scienti¿c Reports, 7(1), 5192. DOI: 10.1038/s41598-017-
Imaging and Measurement of Pore-Scale Interfacial Curvature 05204-4.
to Determine Capillary Pressure Simultaneously With Relative Suijkerbuijk, B., Kuipers, H., Van Kruijsdijk, C., Berg, S., Van
Permeability, Water Resources Research, 54(1), 7046–7060. Winden, J.F., Ligthelm, D.J., Mahani, H., Almada, M.P.,
DOI: 10.1029/2018WR023214. Van den Pol, E., Joekar Niasar, V., Romauka, J., Vermolen,
Lin, Q., Bijeljic, B., Rieke, H., and Blunt, M.J., 2017, Visualization E.C.M., and Al-Qarshubi, I.S.M., 2013, The Development of
and Quanti¿cation of Capillary Drainage in the Pore a WorkÀow to Improve Predictive Capability of Low Salinity
Space of Laminated Sandstone by a Porous Plate Method Response, Paper IPTC-17157 presented at the International
Using Differential Imaging X-Ray Microtomography, Petroleum Technology Conference, Beijing, China, 26–28
Water Resources Research, 53(8), 7457–7468. DOI: March DOI: 10.2523/IPTC-17157-MS.
10.1002/2017WR021083.
McPhee, C., Reed, J., and Zubizarreta, I., 2015, Best Practice ABOUT THE AUTHORS
in Coring and Core Analysis, Chapter 1, Developments in
Petroleum Science, 64, 1–15, Elsevier. DOI: 10.1016/B978- Qingyang Lin obtained his PhD in 2015 and continued
0-444-63533-4.00001-9. as a Research Associate in the Department of Earth Science
Mohanty, K.K., Davis, H.T., and Scriven, L.E., 1987, Physics of
and Engineering at Imperial College London. His main
Oil Entrapment in Water-Wet Rock, Paper SPE-9406, SPE
research interests are using imaging techniques, in particular
Reservoir Engineering, 2(1), 113–128. DOI: 10.2118/9406-
PA. X-ray microtomography, to characterize and understand
Norouzi Apourvari, S., and Arns, C.H., 2016, Image-Based multiphase Àow in porous media with applications in
Relative Permeability Upscaling From the Pore Scale, improved oil recovery and carbon dioxide storage.
Advances in Water Resources, 95, 161–175. DOI: 10.1016/j.
advwatres.2015.11.005. Branko Bijeljic is a Senior Research Fellow in the
Raeini, A.Q., Bijeljic, B., and Blunt, M.J., 2017, Generalized Department of Earth Science and Engineering at Imperial
Network Modeling: Network Extraction as a Coarse-Scale College London. He has a PhD in chemical engineering
Discretization of the Void Space of Porous Media, Physical from Imperial College. His research interests are in
Review E, 96(1), 013312. DOI: 10.1103/PhysRevE.96.013312.
multiphase Àow and reactive transport in porous media, with
Reynolds, C.A., Menke, H., Andrew, M., Blunt, M.J., and Krevor,
S., 2017, Dynamic Fluid Connectivity During Steady-State
applications in oil recovery, carbon storage and contaminant
Multiphase Flow in a Sandstone, Proceeding of the National transport.
Academy of Sciences, 114(31), 8187–8192. DOI: 10.1073/
pnas.1702834114. Samuel Krevor is a Senior Lecturer in the Department
Rücker, M., Berg, S., Armstrong, R.T., Georgiadis, A., Ott, H., of Earth Science & Engineering at Imperial College
Schwing, A., Neiteler, R., Brussee, N., Makurat, A., Leu, London. Previously he was a postdoctoral scholar at
L., Wolf, M., Khan, F., Enzmann, F., and Kersten, M., 2015, Stanford University. He received his PhD in Environmental
From Connected Pathway Flow to Ganglion Dynamics,

Lin et al.
Engineering from Columbia University. His research group Apostolos Georgiadis joined Shell in 2012, as
investigates fundamental and applied aspects of subsurface Researcher in the Rock & Fluid Physics team. He obtained his
Àuid Àow. PhD in 2011 and continued as Post-Doctoral Associate, both
in Chemical Engineering under the Shell-Imperial College
Martin Blunt is Shell Professor of Reservoir research program on Clean Fossil Fuels. His research has
Engineering at Imperial College London. Previously he focused on capillary phenomena and phase behavior in the
was a faculty member in the Department of Petroleum context of EOR and CCS processes, and more recently on
Engineering at Stanford University, and has also worked as pore-scale physics. His current involvement encompasses
a reservoir engineer for BP. He has a PhD in physics from Digital Rock activities, speci¿cally focusing on bespoke
Cambridge University. He performs theoretical, numerical surface and surface-Àuid interaction characterization
and experimental research on multiphase Àow in porous techniques, in context of porous displacement processes
media with application to improved oil recovery and carbon under varying wetting states.
dioxide storage.
Ove Bjorn Wilson is an experienced SCAL Senior
Maja Rücker is a Research Associate in the Chemical Reservoir Engineer/ SCAL Subject Matter Expert with
Engineering Department at Imperial College London. Her 25 years’ experience in the oil industry. He has a PhD in
research focus is on wettability and wettability-alteration Reservoir Engineering from NTNU in Norway. He has 10
effects on two-phase Àow in porous media. She received her years’ experience working with ¿eld development studies
PhD in Petroleum Engineering in a joint project of the Rock (FDP) in Oman and Malaysia. He is currently supporting
& Fluid Physics team at Shell Global Solutions International OUs and JVs from his position in Shell. In parallel with this,
B.V. and Imperial College London. In 2014, she obtained her he is involved in research related to digital rock technology.
master’s level degree in Geology at the University of Mainz,
Germany.
Steffen Berg is a senior researcher at Shell Global

Solutions International B.V. in the Netherlands. His main
research interests range from the fundamental aspects of
multiphase Àow in porous media to CO2 sequestration
and enhanced oil recovery. He holds a master’s degree in
materials science from the University of the Saarland and
a PhD in physics from the University of Mainz / Max
Planck Institute for Polymer Research at Mainz, Germany.
After a postdoc at Princeton University he joined Shell as
a research scientist. He is currently also a visiting reader in
the Earth Science & Engineering and Chemical Engineering
departments at Imperial College London.
Ab Coorn is a research technician at Shell Global

Solutions International B.V. in the Netherlands. He is
involved in a wide range of research activities ranging from
core analysis to catalysis research. He is responsible for
operating the ȝCT and medical CT scanners.
Hilbert van der Linde is a senior associate researcher at

Shell Global Solutions International B.V. in the Netherlands.
His research mainly focuses on Special Core Analysis
experiments, including multiphase Àow and capillary
pressure. He holds a bachelor’s degree in applied physics
from Rijswijk University of Applied Sciences.

PETROPHYSICS, VOL. 60, NO. 2 (APRIL 2019); PAGES 273–282; 7 FIGURES. DOI: 10.30632/PJV60N2-2019a5
In-Situ Saturation Monitoring (ISSM)— Recommendations for Improved Processing1

Jules Reed2 and Arjen Cense3
ABSTRACT
In-situ saturation monitoring (ISSM), using X-rays or inappropriate processing and interpretation methods. This
gamma rays, has become a common method to determine paper ¿rst considers the stipulated 1 to 2 s.u. accuracy and
Àuid saturations in commercial coreÀood experiments. The the necessary signal-to-noise ratio, i.e., counts required, to
most common method in commercial laboratories entails achieve this; as well as providing a suggested approach,
1D saturation measurements as a function of core-plug where plausible, to correct saturation data compromised by
length and of experimental time. Laboratories often employ incorrect calibration scans. It also considers the uncertainties
ISSM as the only method of determining Àuid saturations, in use of ISSM production volumes in determining
assuming an almost infallible accuracy of 1 to 2 saturation unsteady-state relative permeability; speci¿cally, pre- and
units (s.u.). However, as for all measurement methods, post-breakthrough data and the assumptions surrounding
there are possible sources of uncertainty in ISSM data. selection of breakthrough from Àood-front scans. In
Previous papers have discussed some of these uncertainties, addition, ISSM pro¿les are often used in coreÀood
such as X-ray drift, and inappropriate calibration scans or simulation of relative permeability to aid correlation of
changes to core or Àuid properties during testing. Despite the capillary end effect; incorrect data processing may
this evidence, some laboratories continue to use ISSM compromise this correlation. The paper considers several
measurements alone, assuming negligible uncertainty. sources of error in ISSM data and provides a recommended
In the authors’ experience, uncertainties not only exist approach to acquisition, processing and interpretation of
in measurement errors, but also may be introduced by ISSM data for calculation of Àuid saturations.
INTRODUCTION temperature and pressure variation, radiation-source age and/

or degradation, power Àuctuations, core-plug heterogeneity,
In-situ saturation monitoring (ISSM) was introduced to core-plug location displacement and random noise. It is
the oil and gas industry by Boyer et al. (1947), with suggested recommended that, before instigating core-analysis ISSM
improvements by Morgan et al. (1950) and supported by measurements, the equipment and test procedures should be
further experimental data by Geffen and Gladfelter (1952). optimized to minimize external errors and uncertainties, for
Since this time, in general, laboratory equipment has example:
developed and improved to reduce some of the potential x Scan position – Repeatable return and near-exact
uncertainties: radiation sources and detectors are more stable positioning of the scanning measurement unit is an
(possibly due to more stable power sources), core holders essential precursor to gaining saturation accuracy.
have been developed using lower attenuation materials The scanning unit moves based on calibrated full and
and/or with thinner walls to reduce background noise, fractional revolutions of a stepper motor. A linear
temperature-regulated equipment reduces large Àuctuation encoder, to double-check ¿nal position and adjust to
in attenuation due to temperature, etc. However, despite match the desired target position, greatly improves
improvements, there remains potential for uncertainty in the ability to perform scans at nearly the exact same
attenuation measurements from various factors, including scan position, within microns.
considerations regarding an appropriate doping scheme, x Doping scheme – Whether doping a Àuid phase is
Manuscript received by the Editor November 6, 2018; manuscript accepted January 3, 2019.
1
Paper SCA2018-007.
2
Lloyd’s Register, Kingswells Causeway, Prime Four Business Park, Aberdeen, UK, AB15 8PU; Jules.reed@lr.org
3
A/S Norske Shell, Stavanger, Norway

In-Situ Saturation Monitoring (ISSM)— Recommendations for Improved Processing
necessary to increase Àuid attenuation differences, and (5) percentage of the core diameter captured within
which phase to dope, what concentration of dopant the X-ray (or Ȗ-ray) stream. The paper recommends that,
to use—all based on the requirements to achieve in addition to the recommendations by Coles et al. (1995),
high saturation accuracy (within ± 1 to 2 s.u.) as ISSM data should be supported by an alternative method
well as considering the possible chemical impact of of determining intermediate (i.e., during coreÀooding)
introducing foreign material. saturations, as well as veri¿cation of endpoint saturation(s)
x Temperature and pressure effects – Linear X-ray by an alternative method. Despite evidence to the contrary,
absorption coef¿cients vary with temperature and some laboratories continue to assume negligible uncertainty
pressure as a function of density changes. Calibration for ISSM data and employ only ISSM saturation, without
scans should be performed under the same conditions alternative veri¿cation.
as the test, else a temperature and pressure correction In the authors’ experience, ISSM saturation
scheme must be developed for scans performed under uncertainties are not only introduced from mechanical,
different experimental conditions. experimental, physical and/or chemical variances, but may
x Temperature effect on the test rig – Start and end also be introduced through inappropriate data processing
positions of a core plug in a test rig may change when and interpretation methods. Few literature articles have
the rig is heated because of thermal expansion of considered these uncertainties. First and foremost, it is
materials within the test rig. One should ensure that essential that suf¿cient X-ray (or Ȗ-ray) counts are collected
calibration or baseline scans are performed under test at each location to ensure that the measurement variance
conditions. is within a predetermined percentage of the attenuation
x Detector processing time – Total detector counting contrast between the Àuids (i.e., usually 1 to 2%), this may
time (for a scan position) consists of live-time (when not always be the case. The method of determining which
counts are recorded) and dead-time (processing scan locations are extraneous (i.e., which locations are not
time). Roughly speaking, when emergent intensity is measuring core-plug data) may lead to errors in average
high (for example, when scanning a dry core plug), saturation calculations and may introduce uncertainty into
dead-time may too be high. As intensity is decreased, the process of correlating capillary pressure and relative
because of the presence of a Àuid or doped Àuid phase, permeability in coreÀood simulations.
dead-time is reduced. Saturation calculations should ISSM is regularly performed in combination with the
be based on counts per live-time (i.e., time that is use of composite (stacked) core. Composite cores are used
corrected for processing time) rather than counts per for two reasons: (1) to attempt to reduce capillary end
real time (i.e., time that includes processing time or effects or to improve pressure measurements (increase
time taken to resolve pulse accumulation events). differential pressure), by increasing sample length; and
x Detector energy discrimination – Ideally one will (2) to improve volumetric measurements, by increasing
calculate intensities by summing intensities over sample pore volume. The use of composite cores does not
a range of detector channels rather than the entire directly impact ISSM measurements (except to increase the
range of the detector. This excludes detector-speci¿c number of scan positions and hence, the overall scanning
anomalies that have little to do with changes in Àuid time, required), but does introduce procedural challenges: to
contents of the sample. ensure homogeneity of each individual sample in the stack
and to guarantee capillary contact between the samples.
Coles et al. (1995) recommend some techniques to Discontinuities at sample connections can often be observed
improve accuracy for X-ray systems: (1) scan a ¿xed reference in ISSM pro¿les.
material immediately prior each core scan and use the value
to calibrate each scan; and (2) use slow warm-up times for GENERAL ACCURACY CONSIDERATIONS
X-ray tubes to improve source stability and extend their life.
Use of a reference material scan enables corrections also The use of ISSM techniques to determine the saturation
for gamma-ray systems—for temperature drift and source pro¿le in coreÀooding experiments should always be
degradation. Cense et al. (2014), in addition to temperature considered, if budget allows. It is the only method to con¿rm
variance, noted other potential uncertainties: (1) changes to that a core sample behaves as a homogeneous rock and to
the rock matrix during testing; (2) errors in calibration scans observe the capillary end effect. If a core sample consists
(i.e., the scans performed at 100% saturation of the individual of multiple rock types, different saturations will prevail
Àuid phases); (3) limited attenuation difference between the during a coreÀood, and different areas in the sample may
Àuid phases; (4) component transfer between Àuid phases; have different relative permeability. If one assumes that the
sample is homogeneous in the interpretation, varying rock

Reed and Cense
types will be ignored, and the result will be an upscaled

relative permeability and/or residual oil value, which may
be valid for the core sample but not for the rock types. (2)
To allow for good in-situ saturation measurements, one
needs an experiment where only the saturation changes, and
other experimental conditions remain stable. This means
that: where ISw refers to the number of counts measured at the
1. The core sample does not dissolve in the Àuids used detector at a saturation Sw, Sw = 0 refers to the calibration
and there is no precipitation—even if minor amounts scan with the core fully ¿lled with oil/gas, Sw = 1 refers to
of minerals leach out or precipitate—this has a huge the calibration scan with the core fully ¿lled with water. The
impact on the saturation calculation due to the much ref superscript refers to the reference scan, which can be
higher X-ray/gamma-ray absorption of rock minerals made on a separate detector. The reference scan is made at
compared to Àuids, the same time as the measurement of the absorption of the
2. The core sample should remain in the same place. core holder (Fig. 1).
Some laboratories remove samples from the setup
to clean them. Placing these samples back to their
original position is extremely dif¿cult. Even a slight
mismatch in position may result in a different rock
volume being acquired in the X-ray/gamma-ray beam.
This has a huge impact on the saturation calculation
for the same reason mentioned in Item 1,
3. the calibration scans of the core fully Àooded with
oil (or brine) should be done on core samples that are
100% ¿lled with oil (or brine). Often, it is not easy to
con¿rm whether all residual Àuids have been cleaned
out before Àooding the samples with the calibration
Àuid. A sign that this was indeed a problem is that
saturations will be below zero or above one. A
mitigation will be to repeat the cleaning cycle and Fig. 1—Schematic top view of the X-ray source, the core sample, the
detector and the reference detector.
remeasure the calibration scans,
4. The temperature of the detector must be constant. As
X-ray detectors are extremely sensitive to changes Scanning times need to be suf¿ciently long to allow for
in temperature, the detector should be at a constant precise saturation interpretation. But what is long enough?
temperature during the experiment. Since this is In an ISSM experiment using an X-ray source, stable
a nontrivial task, one can use reference scans to saturation at the end of a steady-state fractional Àow step
compensate for temperature Àuctuations in the room, usually is determined from the average of all slices along the
5. The intensity of the source must be constant. If the sample length. If the scanning time is too short, the error in
intensity varies over the duration of the experiment, saturation may be larger than desired. In the example shown
the calibration scans will not reÀect representative in Fig. 2, the error at slice 26 mm initially was 3.5%, and it
conditions. Again, this issue can be mitigated using took about 6 hours to acquire statistically suf¿cient scans
a reference scan to compensate for source intensity before the saturation error was reduced to less than 2% at
Àuctuations. individual slices, while scanning the whole core plug. This
applies to all slices in the core and thus, it can be shown that
If using a reference scan to account for source intensity the error in the average saturation of the core is of the same
variance, the standard equation to calculate saturation from magnitude.
counts (Eq. 1) is revised (Eq. 2). It requires no complex calculation to check whether the
error in saturation is reduced to an acceptable level: simply
plot saturation from individual slices (in practice this will
(1) be counts, as one usually does not have the calibration
scans available at this experimental stage) and calculate the
standard deviation. With some experience, the calibration

levels of an oil-¿lled and water-¿lled core, thus, error in sample heterogeneities will be observed as differences in
saturation can be estimated. Especially in cases where saturation along the length of the core plug, creating peaks
residual oil/gas is determined, one needs to wait suf¿ciently and troughs in the saturation pro¿le. The peaks and troughs
long to attain higher accuracy. Merely Àooding 1.5 pore occur at the same position in every scan, but minor shifts in
volumes in a couple of hours is not good enough. their positions indicate loss of positional control. In such a
case, the method for data processing described in this paper
SAMPLE LENGTH DETERMINATION will not work; a more complex approach will be required,
though any approach could be unviable.
Core plugs are commonly loaded into cylindrical rubber During data acquisition, either a voltage is applied
sleeves between two end-stems usually composed of high- (X-ray) or a slot or pinhole opened (Ȗ-ray) to allow radiation
grade steel or corrosion resistant alloy, with high attenuation to be emitted from the source towards the detector. Radiation
coef¿cient (Fig 3a). An X-ray (or Ȗ-ray) source is situated on reaches the detector through a slot, which most often
one side of the core with a detector linked directly opposite transcribes as a 2-mm slice. The source/detector unit is then
and both attached to the same mobile unit on a motorized moved 2 mm and counts recorded from that next 2-mm slice.
track. The unit can move backwards and forwards, or up and As indicated in Fig. 3b, the number of counts accumulated
down, depending on equipment orientation, but the source through the end-stems is lower than those accumulated
and detector are always in the same relative position, i.e., through the core plug. There is interference from the end-stems
detector directly opposite source. The positional accuracy of into the measurements at the boundary between end-stem and
the motorized unit typically ensures that scans are performed core plug. The interference causes some potential uncertainty
at near-exact positions, though occasionally this may fail. regarding where the sample begins and ends. This uncertainty
Positional failures are rare but usually evident from minor leads to loss of data at the core-plug boundaries.
location shifts in “peak-trough” behavior. For instance,
(a) (b)
Fig. 2—(a)Saturation (blue), average saturation (green) and standard deviation (red) at a single slice at the end of a steady-state step. (b) The initial
error in saturation is 3.5%, but it is halved to 1.7% after measuring the saturation for about 6 hours.

Reed and Cense
(a) (b)
Fig. 3—(a) Example image of core plug loaded inside ISSM core holder. (b) Example X-ray counts at scan locations across the system length.
The approach used by many laboratories to determine locations (1 to 6 and 37 to 44) will not be carried forward
the plug boundaries is indicated in Figs. 4a through 4c. The into calculations of saturation. Only the data between
total (or base) count measurements are acquired, and plug- these inÀection points (locations 7 to 36) will be used to
ends are determined by using either the initial scan (Fig. 4a) determine saturation, without reviewing the full dataset.
only or all scans (Fig. 4b), as the point of inÀection from After calculating saturation, any unusual variance near the
high counts towards lower counts, as indicated —locations selected ends may result in additional attrition, e.g., the ¿rst
7 and 36 along the sample length. Data from extraneous location in Fig. 4c, length = 0.248 cm may be negated.
(a) (d)
(b) (e)
(c) (f)
Fig. 4—(a) Example of base counts, scan locations 1 to 44. (b) Lab selects sample-ends from base counts: locations 7 to 36. (c) Laboratory-reported
saturation data: locations 7 to 36 stretched to caliper length. (d) Saturation calculated from all locations indicating alternative end selection: locations
8 to 41. (e) Selection of sample-ends from (d). (f) Improved saturation data, actual sample length—not forced to caliper length.

Some laboratories will assume that the full length of

the sample has been acquired, stretching the data to match
caliper length. In this process, caliper length will be divided
by the number of scanned locations, to produce a scan-
interval length. Scan location numbers are renumbered from
1 (hence, 7 to 36 becomes 1 to 30). In the example shown,
Fig. 4c, caliper length = 7.43 cm, sample scan locations =
30, producing a scan interval of 0.248 cm. The renumbered
locations are multiplied by interval length to produce the
sample length (0.248 to 7.43 cm). As seen, there are no data
between 0 and 0.248 cm, but data incorrectly exists to the
full length of the core plug.
A suggested improvement to this common approach, is
¿rst to calculate saturations employing all location data and Fig. 5—The production end-face showing the different saturation pro¿les
derived from different data processing approaches. At Pc = 0, using
plot as a function of location, limiting saturation (y-axis) the common laboratory approach Sw = 0.62, using the recommended
between 0 and 1, as shown in Fig. 4d. This exhibits locations approach extrapolated Sw = ~0.4.
at the extremities to have nonsensical, nonphysical data,
below 0 or above 1, obviously invalid. It provides a more
WATER-BREAKTHROUGH SELECTION
appropriate method to determine sample end-faces (here
shown to be locations 8 and 41) where saturation becomes
Many laboratories employing ISSM during unsteady-
relatively constant and homogeneous. This approach can
state (USS) relative permeability coreÀoods, use only ISSM
often extend the number of selected locations (usually by
as the basis of determining saturation change. Figure 6a
one or two locations, i.e., 2 to 4 mm), but sometimes by
provides a typical example of saturation pro¿les acquired
several locations, as in the case shown in Fig. 4e, where four
during such a Àood. The plot shows the original Swi-state
scan locations (0.8 cm) were added. Figure 4f provides the
(blue line) and the progressing Àood front (light-gray lines)
¿nal reinterpretation of Sw as a function of length. Compared
at increasing time steps. Average Sw from these pro¿les is
to the laboratory interpretation, there is a missing interval
often used towards JBN style calculations (Johnson et al.,
at the inlet end (close to zero), due to the decrease in Sw
1959) of relative permeability, but incorporate error, since
between locations 7 and 8, (approaching a10-s.u. decrease)
each slice has been acquired at a different time point and not
and deemed impacted by end-stem interference. This interval
the time used for determining average Sw. The magnitude of
was not removed by the laboratory and can be observed in
this error varies depending upon aspects such as, Àow rate,
Fig. 4c. These different approaches can produce signi¿cant
scan number, scan time, front stability, etc. Appropriate time
variance in calculated saturation. In the presented example,
corrections and/or numerical simulations are recommended
there was an average difference of 3 s.u. and maximum of 6
to synchronize these data before use in interpreting relative
s.u.
permeability. Figure 6a also shows the pro¿le most often
As previously stated, saturation data are not expected to
selected as the point of water breakthrough; the pro¿le
extend the full sample length due to end-stem interference.
when Sw begins to change at the production face (indicated
The practice of stretching saturation data to the full caliper
by the dark, dashed line, “Lab-selected BT”), the thought
length also impacts the observed capillary end effect. In
being that since water has begun to change at this face,
fact, the missing few millimeters at the production end are
breakthrough must have occurred and both oil and water
those most heavily inÀuenced by capillary effects. We would
will now be produced. However, there are two main errors in
not recommend this practice. We recommend that length
this approach: First, there are unobserved data in the last few
calculation be determined by using the standard slice interval
millimeters and the ¿nal location data may not correspond
of the equipment (usually 2 mm). Most often approximately
to the actual production face; second, this selection does
2 to 3 mm may be lost from each end. In the example
not account for capillary effects, and a potential that water
core plug (Fig. 5), the difference between saturation at the
saturation may ¿rst build at the production face prior to
end-face was observed as Sw = 0.62, using the laboratory
actual breakthrough. This case is indicated by the dark-gray,
approach, and Sw = 0.4, using the suggested approach.
continuous line (“Actual BT”) and may correspond to a
water-wet case.

Reed and Cense
(a) (c)
(b) (d)
Fig. 6—(a) ISSM pro¿les showing typical BT selection versus actual BT pro¿le. (b) Example showing the ISSM pro¿les at water breakthrough, for a
range of wetting states, water-wet through oil-wet. (c) Relative permeability input to (a). (d) Capillary pressure input to (a).
The point of water breakthrough depends on variables,

such as mobility ratio, stability of Àood front, heterogeneity,
wettability, etc. Assuming appropriate sample selection
and quality control, the major variables determining
breakthrough in a coreÀood will be mobility ratio (viscous
forces) and wettability (capillary forces). At the production
face, viscous displacement forces approach zero and are less
able to overcome nonwater-wet capillary entry pressures.
Thus, production face saturation will be strongly inÀuenced
by wettability close to Pc = 0. (a)
Figure 6b shows the variance in possible saturation
pro¿les at the point of breakthrough for different wetting
conditions (wetting conditions were represented using
a variety of relative permeability, Fig. 6c, and capillary
pressure, Fig. 6d, inputs). For the water-wet case, as the
Àood front approached the production-boundary end, and
prior to water breakthrough, water saturation accumulated,
in conjunction with increasing differential pressure, until
spontaneous imbibition saturation was achieved. For the oil-
wet case, water breakthrough was almost immediate upon
water reaching the production end, since there were minimal
or no spontaneous water imbibition forces. (b)
Fig. 7—(a) Unsteady-state coreÀood production data indicating different
breakthrough times, ISSM derived and actual. (b) Analytical relative
permeability curves (JBN or Jones-Roszelle) derived from the different
BT times.

In many water-oil imbibition coreÀoods the experimental ISSM SATURATION CORRECTION

time for volumetric oil production (Vo) and differential
pressure (dP) may be offset due to time to displace dead McPhee et al. (2015) recommend that saturation
volumes and outlet Àow lines: inlet dead volume for dP; uncertainty should be within ± 3 s.u., and that saturation
inlet + outlet dead volume and separator Àowline volume should be veri¿ed by at least one additional measurement
(or piping to other volumetric measurement vessel) for Vo. technique. ISSM saturation veri¿cation was strongly
Although there are various methods to align Vo and dP to a recommended in Cense et al. (2014), and we would reiterate
mutual timeline, a common method is merely to align the the need to corroborate Àuid saturation after coreÀooding,
breakthrough point, most often by correcting dP-time to since ISSM saturation data are not infallible.
align with Vo-time using the dP point of inÀection, usually If appropriate sample selection has been performed,
maximum dP, as the dP-breakthrough point. However, if best practice experimental procedures followed, and quality
Vo breakthrough has been selected incorrectly, as described controls implemented throughout testing, there is generally
above and shown in Fig.6a, disparity is created. Figure 7a excellent correlation between ISSM and other methods
shows an example case, where breakthrough was chosen of determining saturation; particularly at experimental
incorrectly from ISSM data (the ¿rst scan to exhibit Sw endpoints (test initiation and completion). There may be
increase at production end) and dP time was offset to match small differences in intermediary saturations during the
this breakthrough time. Subsequent analytical methods steady-state method because of system dead volumes and
to calculate relative permeability, such as JBN or Jones- subsequent small errors introduced by the assumptions
Roszelle (Jones and Roszelle, 1978), using this post- about how Àuids segregate throughout pipework and valves
breakthrough, time-correlated data will calculate incorrect at speci¿c fractional Àow rates. Thus, in correctly controlled
relative permeability data, as depicted in Fig. 7b. This steady-state coreÀoods, the ISSM saturation data from
occurs because a portion of prebreakthrough oil production the intermediary fractional Àow rates may be considered
is included in the interpretation, producing erroneous data at as more accurate than the intermediary volumetric data.
unusually low saturation values. Although the difference in It is therefore essential to verify the endpoint saturations
adjusted time does not appear signi¿cant, it has a signi¿cant to validate these intermediary data. The recommended
impact on the interpreted data, particularly saturation. approach to verify saturation, is to ensure that there is full
A recommended approach is to correct Vo and dP control of saturation data throughout the preparatory stages
timelines independently, and quality check the point of ahead of the coreÀood: saturating the core with formation
breakthrough, preferably using an alternative method to water and establishing initial water saturation conditions, and
ISSM data: throughout the coreÀood itself, by verifying ¿nal saturation;
1. Vo time: most often measuring ¿nal water saturation (Swf).
a. It is essential to know the actual Àow rate (either From initial knowledge of the sample pore volume,
by using volumetrically calibrated pumps, or checks should be in place to ensure the sample is fully
by determining Àow rate from the initial linear ¿lled with formation water during the saturation process.
production data, where the gradient Vo/time = actual Initial water saturation (Swi) must be veri¿ed, either by
Àowrate (Qact). gravimetric or volumetric methods (or preferably both).
b.It is essential to have measured the dead volumes There are a variety of methods of establishing Swi, but the
of the system (i.e., production Àuid held within preferred method would be an individual porous-plate with
Àowlines that cannot be bypassed/displaced prior to net con¿ning stress, that allows production volumes to be
test initiation). measured directly into a graduated vessel (usually a glass
c. Vo-time must then be corrected for dead volumes burette/pipette). Sample weights should be measured before
and regression performed to ¿nd the time intercept (fully water saturated) and after (at Swi), if possible, to
at Vo = 0. compare Swi volumetrically and gravimetrically.
2. dP-time:
From knowledge of the inlet dead volume and actual
Àow rate, dP-time offset is given by tdP = tmeas í (volumetric) (3)
(DVin»Qact), where tdP is the corrected dPtime, tmeas is
measured test time (from commencement of Àow),
DVin is the inlet dead volume and Qact is the actual
Àow rate, as determined during Vo-time offsets. (gravimetric) (4)
3. Verify that Vo and dP breakthrough are consistent.

Reed and Cense
where Vp is pore volume (cm3), Vw is the produced water

volume (cm3), WSw=1 is fully saturated weight (g) (with (6)
formation water), WSwi is the weight at Swi (g), and is the
density difference of the Àuid pair (g/cm3). These data
should be compared against ISSM derived average Swi data where is the ISSM based Swi, is the ISSM based residual oil
(± 3 s.u.). saturation, is corrected saturation, is the true Swi and is the
After the coreÀood, the ¿nal ISSM water saturation true residual oil saturation.
should be con¿rmed by an additional measurement technique:
volumetric production data, Karl Fischer titration, miscible CONCLUSIONS
dispersion analysis, Dean-Stark extraction, etc. Production
volumes captured into a graduated and calibrated separator In-situ saturation monitoring (ISSM) by attenuation of
are often an excellent additional measurement and may be X-rays or Ȗ-rays, may not always provide accurate saturation
used, not only to verify ¿nal saturation, but intermediary data. The following recommendations are suggested:
saturations also; although, volumetric error may be x Optimize readings by including reference scans,
introduced by small system leaks, mass transfer (if Àuids are to correct for variance in test conditions (hence
not fully immiscible), droplet retention in the pipework, etc. attenuation variance).
Karl Fischer titration is an excellent method for determining x Obtain several measurements at steady-state
small water volumes, but may be susceptible to error from conditions (at least ¿ve), calculate standard deviation
(1) incomplete extraction of the water volume during the (stdev) and, if necessary, continue measuring until
solvent injection stage of the process, (2) inadequate mixing stdev is below 0.02 before continuing to the next
and sampling of the solvent/hydrocarbon/water efÀuent stage of testing.
blend, (3) chemical interference if sodium iodide is used, etc. x Calculate saturation from all location data (including
Miscible dispersion could be used to determine the saturation nonsample data) and use saturation versus scan
of the mobile Àuid phase, but could incorporate signi¿cant location to determine inlet and outlet end faces.
error dependent on the heterogeneity of the sample and/ x Do not stretch data to the measured caliper length,
or Àuid Àow path. Dean-Stark extraction is usually an ill- but merely use the equipment slice interval (usually
advised method for con¿rming water volume from ISSM 2 mm) to calculate from scan number to plug length.
analyses, because the core plug must be removed from the x Do not use ISSM to determine water breakthrough;
equipment, and replaced exactly to the same location and an alternative method should be employed
orientation, which is dif¿cult to accomplish. The impact of x Synchronize production and pressure times by using
removal and replacement is dependent upon the accuracy to measurements of system volumes and accurate Àow
which this can be accomplished, the attenuation differences rates to determine offsets
between the different components (core holder, sleeve, core, x Verify the production and pressure timelines to ensure
Àuids, etc.) and core-plug heterogeneity. corroboration at water breakthrough.
Once all these data are collated, it should be considered x Verify saturation at the test endpoints by alternative
which of the data are the more accurate at initial and ¿nal methods (e.g. volumetric or gravimetric data, Karl
conditions, for instance Swi and Swf, respectively. If ISSM Fischer titration, miscible dispersion, or Dean-Stark
data is observed to exhibit signi¿cant variance from other extraction).
veri¿able data, the source of the error should be scrutinized,
considering the potential sources of error for ISSM data, ACKNOWLEDGEMENTS
as previously outlined. However, the endpoint saturations The authors would like to thank the reviewers for their
deemed to be accurate might be used to correct ISSM data useful input.
to determine saturation pro¿les that can be useful, at least
qualitatively. First, calculate normalized water saturation NOMENCLATURE
(Swn) based on the ISSM endpoints, then denormalize based
on the true endpoints: Abbreviations
BT = breakthrough (for imbibition process, this is water
breakthrough)
(5) ISSM = in-situ saturation monitoring
JBN = Johnson-Bossler-Naumann method to determine
relative permeability

stdev = standard deviation Coles, M.E., Muegge, E.L., Auzerais, F.M., Frulla, P., and Kantzas,
s.u. = saturation units A., 1995, The Use of Attenuation Standards for CT Scanning,
USS = unsteady state Paper SCA-9513 presented at the International Symposium of
the Society of Core Analysts, San Francisco, California, USA,
12–14 September.
Symbols
Geffen, T.M., and Gladfelter, R.E., 1952, A Note on the X-Ray
dP = differential pressure Absorption Method of Determining Fluid Saturation in Cores,
DVin = inlet dead volume Paper SPE-952322-G, Journal of Petroleum Technology, 4(3),
= number of X-ray (or Ȗ-ray) counts measured and Transactions AIME, 195, 322.
by a detector through a core containing water Johnson, E.F., Bossler, D.P., and Naumann, V.O., 1959, Calculation
saturation, Sw of Relative Permeability From Displacement Experiments,
= measured X-ray counts at 0% water saturation Paper SPE-1023-G, Transactions, AIME, 216, 370–372.
= measured X-ray counts at 100% water saturation Jones, S.C., and Roszelle, W.O., 1978, Graphical Techniques
= reference X-ray counts measured concurrently for Determining Relative Permeability From Displacement
Experiments, Paper SPE-6045, Journal of Petroleum
with ISw
Technology, 30(5). DOI: 10.2118/6045-PA.
= reference X-ray counts measured concurrently McPhee, C., Reed, J. and Zubizarreta, I., 2015, Core Analysis: A
with ISw=0 Best Practice Guide, Developments in Petroleum Science, 64,
= reference X-ray counts measured concurrently Elsevier. ISBN: 978-0-444-63533-4.
with ISw=1 Morgan, F., McDowell, J.M., and Doty, E.C., 1950, Improvements
Pc = capillary pressure in the X-Ray Saturation Technique of Studying Fluid Flow,
Qact = actual injection Àow rate Paper SPE-950183-G, Journal of Petroleum Technology, 2(7),
Sor = residual oil saturation and Transactions, AIME, 189, 183. DOI: 10.2118/950183-G.
= residual oil saturation (based on ISSM data)
= true residual oil saturation ABOUT THE AUTHORS
Sw = water saturation
Sw corr = corrected water saturation Jules Reed has over 25 years’
Swf = ¿nal water saturation experience in core analysis and is
Swi = initial water saturation coauthor of the book, Core Analysis:
= initial water saturation (based on ISSM data) A Best Practice Guide. Jules is a Past
= true initial water saturation President of the Society of Core Analysts
Swn = normalized water saturation (SCA) and currently serves on the board
tdP = inlet dead volume corrected test time for as Director – North & East Europe. His
differential pressure experience was gained working for Core
tmeas = measured test time (from commencement of Àow) Laboratories, Corex and as Chief Engineer for ResLab,
Vo = volume of oil (produced) (Weatherford Laboratories). His areas of focus have largely
Vp = pore volume been on pragmatic data interpretation and implementation
Vw = volume of water (produced) for wettability, capillary pressure and dynamic reservoir
WSw=1 = sample weight 100% saturated with formation properties, including; relative permeability and laboratory
water feasibility studies for EOR (particularly solvent and chemical
WSwi = sample weight at Swi injection and low salinity waterÀooding).
ǻȡ = density difference between the Àuid pair
Arjen Cense works as a senior
reservoir engineer in A/S Norske Shell.
REFERENCES He is subject matter expert on SCAL in
Shell and he has been serving on the board
Boyer, R.L., Morgan, F., and Muskat, M., 1947, A New Method of the Society of Core Analysts (SCA)
for Measurement of Oil Saturation in Cores, Paper SPE- as Young Professional Director for the
947015-G, Transactions AIME, 170, 15.
past six years. He has been working on
Cense, A.W., Van der Linde, H.A., Brussee, N., Beljaars, J.,
and Schwing, A., 2014, How Reliable is In Situ Saturation integrated reservoir modeling, production
Monitoring (ISSM) Using X-Ray? Paper SCA2014-009 forecasting, capillary pressure, relative permeability and low
presented at the International Symposium of the Society of salinity water Àooding. He holds a MSc and a PhD in physics
Core Analysts, Avignon, France, 8–11 September. from Eindhoven University of Technology.

Review of the Intercept Method for Relative Permeability Correction a Variety of Case
Study Data1
Jules Reed2 and Jos Maas3
ABSTRACT
A novel method of measuring steady-state relative the-art equipment, or for its cost-saving potential. However,
permeability, called the intercept method (IM), was before employing this new method, it was important to
recently introduced. The IM entails a modi¿cation of a the authors to investigate its validity across a wider range
standard steady-state procedure that incorporates multiple of rock properties, sample dimensions and wetting states
total Àow rates at each fractional Àow rate. The objective experienced in commercial special core analysis laboratory
of the method is to measure data at each fractional Àow rate (SCAL) coreÀood experiments. This study thus draws on
that will permit simple analytical calculations to correct a variety of relative permeability curves (and supporting
differential pressure (hence relative permeability) and data) from various global core studies, originally derived
saturation data for the effects of capillary pressure. The IM by typical relative permeability methods plus coreÀood
is intended to provide a corrective technique without the simulation. From these data, we use SCORES (an open-
need for additional supportive analyses, such as capillary source coreÀood simulation software) to simulate the
pressure and in-situ saturation monitoring (ISSM), or as an expected results of multiÀow-rate steady-state experiments
alternative approach to the current considered best practice and use the IM to derive and compare the corrected relative
of numerical coreÀood simulation, which generally permeability curves. Results highlight criteria under which
requires the speci¿ed additional data. the method does not provide fully corrected data. The paper
Consequently, the IM is of interest to the global industry explores these criteria in more detail.
in regions and/or laboratories that do not possess state-of-
INTRODUCTION a changing environment, such as water inÀux during

production because of water injection or because of an
Core analysis is designed to provide quantitative active aquifer. In a hydrocarbon reservoir, due to the large
information of reservoir properties, from limited available length scales, these properties are mostly independent of one
material, that can be used to aid interpretation of more widely another; however, in laboratory-based core analysis these
available but more qualitative measurements, such as seismic properties exhibit mutual interference due to the small length
data and log measurements. The reservoir properties of scales. Mutual interference means that static and dynamic
interest to petroleum scientists, can be split largely into two properties are inextricably linked during laboratory testing,
main categories: static properties, such as capillary pressure i.e. capillary pressure measurements are impacted by the
(Pc), and dynamic properties, like relative permeability (kr) relative permeability of the Àuids and relative permeability
(McPhee et al., 2015). Static properties describe reservoir measurements are inÀuenced by capillary pressure (Gray
endpoints achieved through geological timescales, such as and Hassanizadeh, 1991).
connate water saturation driven by hydrocarbon migration, To achieve static capillary pressure conditions in
or after substantial human intervention, such as expected laboratory tests, Àuids must ¿rst be displaced, e.g., for
¿nal saturations in fully swept layers or regions; while primary drainage, water must be displaced from the fully
dynamic properties describe the movement of Àuids during water-saturated core plug under the inÀuence of pressurised
Manuscript received by the Editor November 6, 2018; revised manuscript received March 11 2019; manuscript accepted March 22, 2019.
1
Paper SCA2018-030.
2
Lloyd’s Register, Kingswells Causeway, Prime Four Business Park, Aberdeen, UK, AB15 8PU; Jules.reed@lr.org
3
Independent Consultant; jgmaas@euronet.nl

Review of the Intercept Method for Relative Permeability Correction a Variety of Case Study Data
hydrocarbon. Fluid displacement rates change dependent the core analysis process to ensure that relative permeability
upon relative permeability at speci¿c saturation values: e.g., and capillary pressure data are performed on correlated,
as water saturation decreases, relative water permeability representative samples (Maas and Schulte, 1997; McPhee et
decreases and hence, effective water permeability decreases, al., 2015).
slowing water production rates and extending test time. It is The intercept method (Gupta and Maloney, 2014)
essential in static-property analysis, to allow suf¿cient time provides a potential alternative method to acquire relative
to achieve static (or near-static) conditions, else error will permeability, without the additional capillary pressure or
be introduced to the results. Thus, static-property test times ISSM data, using simple analytical calculations to derive
will be impacted by relative permeability and the results relative permeability by correcting for CEE.
will be impacted by the actual length of time employed by
laboratories, compared to what should be required to extend DISCUSSION OF PHYSICAL PRINCIPLES OF THE
to static conditions (Leverett, 1941; Kokkedee et al., 1996; METHOD
Maas et al., 2016).
Relative permeability experiments are impacted by To understand the IM, we will ¿rst review the physics
capillary pressure, such that Àuid saturations are determined of the CEE. In a Àooding experiment, just outside the end-
as a function of pore-throat (and/or pore) radii, core wettability face of the core plug, the meniscus between two escaping
and the balance of Àuid pressures during Àow conditions. Àuids has little or no curvature, because the Àuid collection
Hence, the inÀuence of capillary pressure becomes a system in the end Àange of the apparatus has channels of
function of differential pressure (dP) which is decreasing a dimension much larger than the pore sizes in the plug.
from the inlet face (largest viscous displacement force for Consequently, the capillary pressure just outside the plug
the given test conditions) and tending to zero at the outlet will be close to zero, i.e., the pressures in the two phases
(production) face, where capillary forces will dominate. will be nearly identical. Because of pressure continuity, the
This results in an outlet-face saturation which will always be pressures of the two phases near the exit face just inside the
determined by spontaneous wetting forces and a potentially plug will then also be nearly identical.
changing saturation with sample length towards the inlet- Inside the core plug, the capillary pressure curve links
face (increasing dP), giving rise to the capillary end effect the difference in phase pressure to a local saturation. So,
(CEE). During relative permeability analysis, the impact of near the exit, the local saturation is likely to be almost a
capillary pressure will vary dependent upon several aspects constant value given by Pc = 0, throughout the experiments.
of the test, such as the method, samples length, viscosity Upstream in the core plug the saturation values likewise are
ratio, average saturation, Àow rates, etc. (Leverett, 1941; determined by the balance between viscous and capillary
Hinkley, 1986; Qadeer et al., 1988; Nordvedt et al., 1997; forces.
Maas and Schulte, 1997). Near the entry face, saturation is driven by the injected
Various methods have been considered to attempt to fractional water Àow (fw), while viscous forces and capillary
reduce these effects: using scaling criteria to increase the forces are keeping their balance. The capillary pressure
ratio of viscous-to-capillary forces; increasing Àow rates, linked to the saturation, “asked for” by the governing fw
increasing sample length (Rapoport and Leas, 1953; Batycky curve, cannot surpass the viscous pressure drop in the more
et al., 1981) using pressure taps along the sample attempting mobile phase. If it would, a negative pressure gradient
to keep the CEE outside so that unaffected data lie within the would come about in the displaced phase at the entry face.
pressure capture region (e.g., Chen and Wood, 2001; van der In effect, any Àooding experiment, whether steady-state or
Post et al., 2000) using core pieces as endstems attempting an unsteady-state (Welge, 1952) experiment, is limited in the
to capture the end effect within the core pieces; but CEE maximum capillary pressure it can probe due to limitations
may not be suf¿ciently minimized or fully removed by these in the laboratory. High Àow rates will interfere at some point
methods, particularly where composite cores are used, which with the integrity of the core plug due to the migration of
may exhibit CEE occurring at core intersections (Morse et ¿nes, or the plug may just break down at a high pressure
al., 1947; McPhee et al., 2015). drop. This is the reason that true residual oil or connate water
Simulation methods, that derive relative permeability can never be achieved in a Àooding experiment.
by accounting for capillary pressure and matching test Gupta and Maloney (2014) suggest that with increasing
measurements, can be dif¿cult and require additional data total Àow rate (Qt), at a given constant fractional Àow,
inputs: capillary pressure and preferably in-situ saturation the saturation distribution in the core plug will maintain
monitoring (ISSM). The use of additional input data requires an approximately constant average saturation. With the
stringent quality-control processes throughout all stages of saturation at the entry point being constant because it is

Reed and Maas
governed by a constant fw, the saturation pro¿le in the plug Kingdom Continental Shelf (UKCS) sandstone, mixed-wet;
gets compressed in shape, but the total saturation range Case 2, core from a South American sandstone, slightly
does not change. Based on that assumption they derive two water-wet; Case 3, core from UKCS sandstone, mixed
correction procedures: one to correct the pressure drop for to water-wet; Case 4, core from UKCS sandstone, Brent
the impact of capillary forces, and another one to correct the formation, slightly oil-wet; Case 5, Middle East carbonate,
average saturation (as observed from material balance) for slightly water-wet. Table 1 provides the core-plug and Àuid
the impact of the capillary forces on the saturation pro¿le properties input to the simulations. The table also includes
(Gupta and Maloney, 2014, 2015). the subsection, “Corresponding Plug for Pc Data”, which
It is not a priori clear when this approximation of the provides details of the representative “sister” core plugs
water saturation being a constant in the end-effect zone may used to measure capillary pressure (Pc), to constrain possible
break down. For that reason, we have tested the method on variables in the coreÀood history matching process. The
cases with a large variety of wettability, i.e., of curvature of Corey model (Corey, 1954) was used to describe relative
the capillary pressure function. permeability (kr) and the Skjaeveland model (Skjaeveland
et al., 2000) was used to describe capillary pressure curves.
CASE STUDIES The Corey and Skjaeveland et al. model parameters for each
case also are provided in Table 1, and plotted on a combined
A review of the IM was performed by simulating in graph in Fig. 1.
SCORES—an open-source coreÀood simulation software For each case (Case 1 to Case 5) relative permeability
(Maas et al., 2011). Simulations were performed as a multirate had been derived by simulation of experimental data together
steady-state experiment using input relative permeability with capillary pressure from a representative plug (from the
and capillary pressure data from several case studies, on same rock type and with similar properties (see Table 1).
core plugs of varying reservoir properties and wettability, The simulation history-match data of Cases 1 through 5 are
and various Àuid properties. The samples represent a variety provided in Figs. 2 through 6. Table 1 also provides the total
of dimensions often used in commercial laboratory analysis, Àow rates used for each case. Please note that the bracketed
from below 5 cm and up to 9 cm on single plugs, and a values are those used as an additional ¿nal fraction (fw = 1),
composite core. Although no low-porosity (< 0.215) samples as a “bump Àood” where deemed necessary, attempting to
were investigated, there was a wide permeability range, thus decrease the impact of CEE. A bump Àood is performed by
reservoir quality index. increasing the total Àow rate, which increases the differential
Six cases were reviewed: Case 0, the original data from pressure and reduces the impact of the CEE length.
Gupta and Maloney (2014); Case 1, core from a United
Fig. 1—Summary of relative permeability and capillary pressure data for the case study samples, input to intercept method simulations.

Table 1—Summary of Core Properties, Fluid Properties, Relative Permeability and Capillary Pressure Inputs
Note: The bracketed total Àow rates were those used at fw = 1, in order to increase viscous forces and reduce the inÀuence of CEE.
Fig. 2—Case 1 experimental data with simulated history match.

Reed and Maas

Case 0 dependent on the laboratory equipment and permeability

Figure 7 shows the ¿nal results of repeating the analysis value (McPhee et al., 2015). However, an impractical
by Gupta and Maloney (2014). Figure 7a shows the resultant time of at least 1,000 years was required to achieve stable
relative permeability data. The lines are the input relative conditions during the ¿nal fractional Àow, fw = 1, owing to
permeability curves (krw and kro); open circles are the results the extremely low oil relative permeability (< 10-10) in the
obtained by conventional (Darcy) calculations and ¿lled ¿nal few saturation units, but achieving stable conditions
diamonds are the results using the intercept method (IM). was discovered to be essential for the intercept method
and (Fig. 7b) shows the saturation pro¿les (saturation as a (IM) to successfully derive corrected relative permeability
function of sample length). The data eventually provided a and saturation close to residual conditions. Initial Case
match to the input relative permeability. Expected Sor was 0 simulations were performed using typical laboratory
0.20; IM predicted 0.203, well within the recommended experimental timescales (approximately 15 days) for a
laboratory-measurable saturation error, ±3 saturation units steady state analysis. These initial simulations resulted in a
(s.u.) (McPhee et al. 2015). Krw endpoint was expected to be good match for saturation, within laboratory error (± 3 s.u.)
0.65; IM predicted between 0.637 and 0.647, a maximum 2% but failed to match krw at during the ¿nal fractional Àow
error (again, well within measurable laboratory permeability rate (fw = 1); with permeability errors from +8% to -38%.
error, which varies between approximately < ±5% and ±25% Expected permeability error for 15 mD is approximately 5%.
(a) (b)
Fig. 6—Case 0 relative permeability curves (a): lines = input kr curves, open circles = analytically calculated kr (Darcy), ¿lled diamonds = IM kr;
saturation pro¿les (b); colors = individual fw steps, line style = different total Àow rates (Qt) (as per Fig. 9b).

Reed and Maas
Case 1 American sandstone. Relative permeability results are

Figure 8a shows the results of Case 1 relative permeability shown in Fig. 10, again showing excellent IM-corrected
curves from UKCS mixed-wet sandstone. The legend for all data, though again failing to extrapolate to true residual oil
the Case ¿gures: The lines are the input relative permeability saturation, since the CEE (captured within the sample length
curves; open circles are the results that would be obtained for all previous fractional Àow rates) is not captured within
from conventional (Darcy) calculations and ¿lled diamonds the sample length at fw = 1. Expected residual oil was 0.14,
are the results using the intercept method (IM). Figure 8 while Àooding only achieved 0.215—an error of 7 s.u. IM
data show dependable (within laboratory error) corrected calculations failed to extrapolate endpoint krw (expected 0.4),
data through all fractions, except at the ¿nal fraction, fw = 1 which ranged between krw’ = 0.246 to 0.511; for -39 to +28%
(dispersed open circles from approx. Sw = 0.80 – 0.87. Figure error, where the expected measurement error would be ±10
8b shows saturation pro¿les, where each total Àow rate (Qt) to 15%.
is plotted using different line formatting (each fractional
Àow rate is denoted by different line colors). The plot Case 3
indicates that the CEE has been captured within the sample Case 3 (see Fig. 11) was a water to mixed-wet, high
length for most of the saturation pro¿les, except at fw = 1 (the permeability (almost 3,000 mD) composite core plug, created
grey upper four lines). At fw = 1, the CEE extends beyond the from four individual samples from a single, homogeneous
length of the sample so that regression of dP versus Qt will whole-core preserved section. Results exhibited good
not produce a correct offset, thus true kr and true Sw cannot be agreement until Sw increased above approximately 0.70,
extrapolated. As an improved example of this phenomenon, which required unusually high fw rates (0.999). As can
Fig. 9 shows two different plots of saturation pro¿les: Fig. be observed in the saturation pro¿les, the CEE was not
9a showing an example capturing the entire CEE within captured within the sample length for the ¿nal two fractions,
the sample length, where all total rate saturations converge despite the sample being 32-cm long. It must be noted that
to a single saturation at the injection end (length = 0); and simulations in SCORES do not account for potential multiple
Fig. 9b which has not captured the CEE within the sample internal end effects, i.e., between the four individual samples
length, i.e. saturations do not converge to a single saturation of the composite stacked sample. Residual oil saturation is
at length = 0. not achieved: expected Sor = 0.18, IM-calculated Sor = 0.22 (a
4 s.u. error). Endpoint krw was not achieved: expected krw’ =
Case 2 0.25, IM krw’ ranged from 0.16 to 0.36; a -36 to +49% error
Case 2 is a slightly water-wet plug from a South for a sample with ± 25% laboratory measurable error.
(a) (b)
Fig. 8—Case 1: (a) relative permeability curves – lines = input kr curves, open circles = analytically calculated kr (Darcy), ¿lled diamonds = IM kr; (b)
saturation pro¿les – colors = individual fw steps, line style = different total Àow rates (Qt) (as per Fig. 9b).

(a) (b)
Fig. 9—Case 0 (a) and Case 3 (b) saturation pro¿les for each total Àow rate at fw = 1. Case 0 showing CEE effectively captured within the sample
length (i.e., saturations resolving to a unique value at the injection face), whereas Case 3 showing unresolved saturation pro¿les.
(a) (b)
Fig. 10—Case 2 relative permeability curves (a) and saturation pro¿les (b).
(a) (b)

Reed and Maas
Initial attempts to simulate Case 3 failed due to Case 5

insuf¿cient time provided to achieve steady-state conditions Case 5 (Fig. 14) is a slightly water-wet carbonate from
during certain fractional Àow rates, particularly at high fw the Middle East. IM calculations produced a good correlation
values. Through this study, it was discovered that if steady to the expected relative permeability curves. However, there
state had not been suf¿ciently reached (stable fractions), the is a small mismatch in krw at the higher saturations—IM krw is
IM calculations did not work well for the unstable fractions. slightly lower than expected, but within ± 5% error expected
Times and Àow rates fw and Qt) were adjusted and optimized in measurements for this permeability range. However,
to achieve the ¿nal simulations presented. However, Qt was residual oil saturation was correctly regressed (0.21). It is
limited to 480 cm3/h, since many commercial laboratories worth noting that this is the only case (of those studied)
use pumps with limited Àowrates, and very high Àow rates where true residual oil saturation was achieved, even after
may be detrimental to the core material, possibly promoting increasing total Àow rates during the fw = 1 fraction, (total
¿nes or grain migration, or turbulent Àow. Time optimization Àow rates were increased to attempt to reduce the CEE impact
for Case 3 implies that 20 days would be required, merely to and thus, improve the correlations). In practical terms, it
complete fw = 1 at all four fractional Àow rates (see Fig. 12b). may not always be possible to increase laboratory total Àow
An earlier simulation of Case 3 (Fig. 12a) shows that fw = 1
rates: for instance, differential pressure limits, pump limits,
does not stabilize within a coreÀood lasting 23 total days.
¿nes migration (and other clay sensitivity issues), etc., may
Subsequent simulations were performed changing various
prohibit further increase.
parameters (e.g., test time, Qt and fw values) to attempt to
optimise the steady-state conditions.
DISCUSSION
Case 4
Case 4 (Fig. 13) was a slightly oil-wet sandstone. IM The intercept method is a very interesting tool for
calculations produced results (¿lled diamonds) not quite determining relative permeability data, and was a good
matching the expected relative permeability curves (lines) at method to predict relative permeability under many
saturations above 0.5. IM calculations appear to overestimate conditions. However, it can fail when the CEE is not
kro and underestimate krw at these higher saturations, although captured within the length of the core plug, i.e., the capillary
they are a much-improved correlation over those that would regime extends beyond the injection face. This potential
be derived from standard Darcy steady-state analytics drawback was described in the original paper by Gupta
(open circles). The open circles can be observed to fall in and Maloney (2014), and was the impetus for this review;
four distinct curves (dotted lines), each relating to one of since, in the experience of the authors, many commercial
the four total Àow rates. Residual oil saturation was not coreÀoods use short core plugs, with signi¿cant capillary
achieved: Sor = 0.22 compared to expected 0.05, a 17 s.u. artifacts observed in saturation pro¿les measured by in-situ
error. Endpoint krw was not achieved: krw’ = 0.386 to 0.711 saturation monitoring (ISSM).
compared to expected 0.75 (up to 49% error).
(a) (b)
Fig. 12—Example of production data (Case 3) using different run times to attempt to achieve steady-state conditions, thus optimize the IM results.
The ¿nal three rates in the shorter run (a) have not achieved steady state, additional time was used for the next simulation (b). However, note the ¿nal
rate in the longer run (b) - (between 44 and 54 days) still may not have achieved steady state.

(a) (b)
(a) (b)
Fig 14—Case 5 relative permeability curves (a) and saturation pro¿les (b).
Failure to capture the CEE in the sample can derive IM calculations are easy to implement, using graphical
from fractional Àow rates that have not achieved steady-state regression analysis as the functional process. Table 2
conditions (stable saturation and stable differential pressure). provides data results from standard Darcy calculations
Laboratories should ensure that steady state is achieved by (left-hand side (blue)) and IM calculations (right-hand side
plotting production volumes and differential pressure as (green)), from selected fractional rates of Case 3. The main
a function of time (most appropriately logarithm time or functional regressions performed are:
square-root time). Various scales should be employed for
graphical axes when determining stability. Failure to capture dP vs. Qt
CEE may also derive from inadequate viscous forces, i.e., Qt
is too low and a laboratory should consider increasing Qt for Sw/(1-ȕ) vs. ȕ/(1-ȕ),
that current and future fractional rates.
where ȕ is a CEE length-factor term introduced by Gupta
and Maloney (2014).

Reed and Maas
Table 2—Case 3 Data Table Showing Test Parameters and Darcy Analytical Results
Figure 15 shows examples of these functional regressions fractions are observed as nonunique values. For example,
from three different fractional Àow rates. The left plot shows for fw = 0.999, the average krw is 0.0903 with a maximum
an example where IM predicted kr, at low fw (0.1), with an error of 16.7%. However, it must be noted that average value
excellent correlation. In Table 1 at this fraction (and other is not the correct relative permeability (regression of just the
fractions that correctly predicted kr), a unique value of krw higher Qt rates may improve the prediction). Observation
and kro can be observed for each Qt , except fw = 0.99 which of non-unique IM krw and/or kro can be used as a device to
exhibits a small error (± 0.0001mD = ± 0.25 %). The middle determine whether additional stabilisation time might be
and right-hand plots show examples when IM has failed required or an additional, higher Qt may be required.
to predict (fw = 0.999 and fw = 1). The tabular krw at these
(a) (b) (c)

Fig.15—Case 3 - example of Qt versus dP and Sw/(1-ȕ) versus ȕ/(1-ȕ) at four total Àow rates for three of the fractional Àow rates (fw = 0.1, 0.9999 and
1), indicating the breakdown of the regression, particularly in dP at higher fw, hence breakdown of the intercept method

CONCLUSIONS Symbols
ǻPI = difference between differential pressure
The intercept method (IM) is a good prediction of with and without CEE
relative permeability for a wide range of wetting states, ao = Skjaeveland correlation factor for oil
relative permeability curves and capillary pressure, over a aw = Skjaeveland correlation factor for water
wide range saturation values. However, for certain systems, Co = Skjaeveland exponent for oil
the validity of IM degrades at higher water saturations, most Cw = Skjaeveland exponent for water
often for the krw curve. Corr. dP = corrected differential pressure (corrected
The successful application of IM is dependent on for CEE impact)
capturing the full CEE within each Àow rate at each fractional dP = differential pressure
Àow step. dP-measured = measured differential pressure
The validity of IM calculations at speci¿c fractional dP-sim = simulated differential pressure
Àows is dependent upon each total Àow rate achieving frac. = fraction
steady-state conditions at that fractional rate step. This is fw = fractional Àow to water
particularly necessary at early and late fractions. fw# = # indicates the fractional Àow step number
It is easy for a laboratory to assess whether results in a steady-state experiment, e.g., fw2 is the
correctly predict saturation and relative permeability at each second fractional Àow rate to water
fractional Àow rate, i.e., results should be unique for each k = permeability
total Àow rate at that fw. If not, the steady state criteria and/or keo = effective permeability to oil
total Àow rates should be reviewed. kew = effective permeability to water
In such circumstances, it may be necessary to increase kr = relative permeability
total Àow rate to achieve viable inputs for deriving correct kref = reference permeability (upon which kr is
values, however, laboratory limits and/or core sensitivities based)
may preclude the practicality of this. kro = relative permeability to oil
Residual oil saturation cannot be guaranteed by kro = endpoint (maximum) relative permeability
performing these experiments. It thus remains, that residual to oil
oil saturation by purely Àooding techniques may not be krw = relative permeability to water
realized. krw’ = endpoint (maximum) relative permeability
It is strongly recommended to design the experiments to water
using a Àow simulator to assess what fractional Àows and No = Corey exponent for oil
Àow-rate programming should be used to achieve reasonable Nw = Corey exponent for water
results. Pc = capillary pressure
Qt = total Àow rate
ACKNOWLEDGEMENTS Qw = water Àow rate
Qw# = # indicates the water Àow rate step number
The authors would like to thank Dan Maloney and in an unsteady state experiment, e.g. Qw1 is
Robin Gupta for allowing us to use their original data and the ¿rst water injection rate
inputs, to begin our review of this interesting method. RQI = reservoir (rock) quality index;
Sor = residual oil saturation
NOMENCLATURE Sw = water saturation
Sw-sim = simulated water saturation
Abbreviations Swf = ¿nal water saturation
avg = average Swi = initial water saturation
bump = indicates a bump Àow rate (i.e. an increased Àow rate v/v = volume per volume (fractional volume
to reduce the length scale of the capillary end effect) ratio)
CEE = capillary end effect Vo = volume of oil produced
IM = Intercept Method Vo-measured = measured volume of oil produced
ISSM = in situ saturation monitoring Vo-sim = simulated volume of oil produced
PV = pore volume ȕ = capillary end effect length factor (per Gupta
SCAL = special core analysis (laboratory) and Maloney, 2014)
s.u. = saturation units ‫ = ׋‬porosity
UKCS = United Kingdom continental shelf

Reed and Maas
REFERENCES Journal, 46, 109, 194

Nordtvedt, J.E., Ebeltoft, E., Iversen, J.E., Sylte, A., Urkedal,
Batycky, J.P., McCaffrey, F.G., Hodgins, P.K., and Fischer, D.B., H., Vatne, K.O., and Watson, A.T., 1997, Determination of
1981, Interpreting Relative Permeability and Wettability from Three-Phase Relative Permeabilities From Displacement
Unsteady-State Displacement Measurements, Paper 9403, Experiments, Paper SPE 36683, SPE Formation Evaluation,
SPE Journal, 21(3), 296–308. DOI: 10.2118/9403-PA. 12(4), 221–226. DOI: 10.2118/36683-PA.
Chen, A.L., and Wood, A.C., III, 2001, Rate Effects on Water-Oil Qadeer, S., Dehghani, K., Ogbe, D.O., and Ostermann, R.D., 1988,
Relative Permeability, Paper SCA2001-19 presented at the Correcting Oil/Water Relative Permeability Data for Capillary
SCA International Symposium, Edinburgh, Scotland, UK, End Effect in Displacement Experiments, Paper SPE-17423
17–19 September. presented at the SPE California Regional Meeting, Long
Corey, A.T., 1954, The Interrelation Between Gas and Oil Relative Beach, California, USA, 2325 March. DOI: 10.2118/17423-
Permeabilities. Producers Monthly, 19(1), 38–41 MS.
Gray, W.G., and Hassanizadeh, S.M., 1991, Unsaturated Flow Rapoport, L. A., and Leas, W. J., 1953, Properties of Linear
Theory Including Interfacial Phenomena, Water Resources WaterÀoods, Paper SPE-213-G, Journal of Petroleum
Research, 27(8),1855–1863. DOI: 10.1029/91WR012609. Technology, 5(5), 139–148. DOI: 10.2118/213-G.
Gupta, R., and Maloney, D., 2014, Intercept Method—A Novel Skjæveland, S.M, Siqveland, L.M., Kjosavik, A., Hammervold
Technique to Correct Steady-State Relative Permeability Thomas, W.L., and Virnovsky, G.A., 2000, Capillary Pressure
Data for Capillary End Effects, Paper SPE-171797, SPE Correlation for Mixed-Wet Reservoirs, Paper SPE-60900,
Reservoir Evaluation & Engineering, 19(2), 316–330. DOI: SPE Reservoir Evaluation & Engineering, 3(1), 60–67. DOI:
10.2118/171797-PA. 10.2118/60900-PA.
Gupta, R., and Maloney, D., 2015, Applications of the Intercept van der Post, N., Masalmeh, S.K., Coenen, J.G.C., van der Gyp,
Method to Correct Steady-State Relative Permeability for K.H., and Maas, J.G., 2000, Relative Permeability, Hysteresis
Capillary End-Effects, Paper SCA2015-001 presented at the and I-Sw Measurements on a Carbonate Prospect, Paper
SCA International Symposium, St John’s Newfoundland and SCA2000-07 presented at the SCA International Symposium,
Labrador, Canada, 16–21 August. Abu Dhabi, UAE,
Hinkley, R.E., and Davis, L.A., 1986, Capillary Pressure Welge, H.J., A Simpli¿ed Method for Computing Oil Recovery by
Discontinuities and End Effects in Homogeneous Composite Gas or Water Drive, Paper SPE-124-G, Journal of Petroleum
Cores: Effect of Flow Rate and Wettability, Paper SPE- Technology, 4(4), 91–98. DOI: 10.2118/124-G.
15596 presened at the SPE Annual Technical Conference and
Exhibition, New Orleans, Louisiana, USA, 5–8 October. DOI:
10.2118/15596-MS. ABOUT THE AUTHORS
Kokkedee, J.A., Boom, W., Frens, A.M., and Maas, J.G., 1996,
Improved Special Core Analysis: Scope for a Reduced Jules Reed has over 25 years’
Residual Oil Saturation, Paper SCA9601 presented at the experience in core analysis and is
SCA International Symposium, Montpellier, France, 8–10
coauthor of the book, Core Analysis:
September.
Leverett, M.C., 1941, Capillary Behavior in Porous Sands, Paper
A Best Practice Guide. Jules is a Past-
SPE-941152-G, Transactions, AIME, 142, 152–169. DOI: President of the Society of Core Analysts
10.2118/941152-G. (SCA) and currently serves on the board
Maas, J.G., Flemisch, B., and Hebing, A., 2011, Open Source as Director for North and East Europe.
Simulator DUMUX Available for SCAL Data Interpretation, His experience was gained working
Paper SCA2011-08 presented at the SCA International for Core Laboratories, Corex and as Chief Engineer
Symposium, Austin, Texas, USA, 18–21 September. www. for ResLab, (Weatherford Laboratories). His areas of
jgmaas.com. focus have largely been on pragmatic data interpretation
Maas, J.G., and Schulte, A.M., 1997, Computer Simulation of and implementation for wettability, capillary pressure
Special Core Analysis (SCAL) Flow Experiments Shared and dynamic reservoir properties, including; relative
on the Internet, Paper SCA-9719 presented at the SCA
permeability and laboratory feasibility studies for EOR
International Symposium, Calgary, Alberta, Canada, 8–10
September.
(particularly solvent and chemical injection and low-salinity
Maas, J.G., Springer, N., and Hebing, A., 2016, Relative waterÀooding).
Permeability Effects Overlooked in MICP Measurements
Transition Zones Likely to be Smaller, Paper SCA2016-013 Jos G. Maas retired from Shell in
presented at the International Symposium of the Society of 2010 after holding a variety of positions
Core Analysts, Snowmass, Colorado, USA 21–26 August. over his 33 years of service. His activities
McPhee, C., Reed, J., and Zubizarreta, I., 2015, Core Analysis: A centered on reservoir engineering,
Best Practice Guide, Developments in Petroleum Science, 64, including experimental research in the
Elsevier. ISBN: 978-0444635334. laboratory in the Netherlands. He has
Morse, R.A., Terwilliger, P.L., and Yuster, S.T., 1947, Relative been the manager of Shell SCAL R&D
Permeability Measurements on Small Samples, Oil & Gas

(focusing on the measurement of remaining versus residual

oil saturation, low-salinity Àooding, SCAL for EOR),
worked on CO2 EOR (e.g., at Shell’s Bellaire Technology
Center in Houston), was manager of Geothermal R&D, and
has been Shell’s technical lead for CO2SINK (a European
CO2 sequestration demo project). Maas is a Past-President of
the Society of Core Analysts (SCA). Presently he is enjoying
an active retirement in which he is teaching SCAL and EOR,
and consulting, among others, as an Associate of PanTerra
Geoconsultants. Jos Maas studied experimental physics at
Amsterdam University and obtained his PhD in 1976. He
is the proud recipient of the SCA 2011 Darcy Award for
Technical Achievement.

Monitoring Core Measurements With High-Resolution Temperature Arrays1

James J. Howard2 and Keith C. Hester3
ABSTRACT
While distributed temperature sensing (DTS) has ¿ber optic in place so that the sensed signal was temperature
become a commonly used tool in reservoir studies, the only and did not include any strain effects. The OBR was
technology has not seen widespread use in SCAL projects. set up to acquire a temperature array every 30 seconds
Most core-scale experiments attempt to control temperature during the test at 5-mm spacings. The core holder was
at a constant value rather than monitor temperature changes placed in a MRI instrument that provided additional spatial
within a sample during a test. High-resolution temperature information on hydrate formation during the test that was
arrays are available that measure changes in temperature compared with the OBR results. Initial hydrate formation
of 0.1oC at 1-mm resolution. The optical backscatter resulted in a several degrees increase in temperature at the
reÀectance (OBR) ¿ber senses both temperature and inlet end of the cell that with time, progressed down the
strain that can be separated through experiment design length of the cell. The temperature array and MRI images
and signal processing. These OBR ¿bers are sensitive both showed the nonuniform nature of hydrate formation
enough to monitor temperature changes associated with and subsequent dissociation of the hydrate when N2 was
endo- and exothermic chemical reactions associated with injected into the cell as a permeability enhancement step.
mineral dissolution/precipitation, or Àuid-front movements The faster response of the OBR array compared to the time
in steam-assisted gravity drainage of heavy-oil tests. An required to acquire MRI images provided additional detail
example of the use of a distributed temperature array is in the sequence of hydrate formation and dissociation
in the monitoring of natural-gas-hydrate formation and during CH4-CO2 exchange. The limitation to the OBR array
dissociation in a sandpack as CO2 is exchanged with the was that it only sensed temperature Àuctuations proximal
naturally occurring CH4 in the hydrate structure. A ¿ber- to the ¿ber as a function of the hydrate system’s thermal
optic array was placed within a narrow-diameter PEEK tube conductivity.
as the sandpack was constructed. The PEEK tube held the
INTRODUCTION de¿ned heats of formation that translate to measurable

temperature changes observed during formation and
Measurable temperature changes are associated with a dissociation (Makogon, 1997; Sloan and Koh, 2007). Most
number of geochemical reactions in porous media that occur of the observations on temperature change were collected
with phase changes connected with mineral precipitation on simple water-gas-hydrate systems measured in bulk.
and dissolution. While thermodynamic data may indicate This study reports measurements of temperature changes as
large temperature changes with an idealized geochemical hydrates formed and dissociated in a porous medium.
reaction, the actual measured temperatures are often much A series of laboratory experiments designed to support
smaller because of volume and heat-Àow constraints. Most the ¿eld trial of CO2-CH4 exchange mechanisms investigated
experiments monitor a single averaged temperature of the the role of injection gas composition on the effectiveness of
reaction vessel and miss out on details of the geochemical the injection process and the actual exchange. The success
processes that occur in heterogeneous porous media. of the Ignik Sikumi No. 1 ¿eld trial in 2011–2012 was in
Multiple temperature probes are one way to capture some of part due to these laboratory tests (Schoderbek et al., 2013).
this detail. These experiments followed the design of earlier studies
Natural-gas hydrates are characterized by well- that formed methane hydrate in porous media, introduced
Manuscript received by the Editor October 28, 2018; revised manuscript received February 4, 2019; manuscript accepted February 11, 2019.
1
Origially presented at the International Symposium of the Society of Core Analysts, Trondheim, Norway, August 27–30, 2018, Paper
SCA2018-014.
2
Department of Geosciences, University of Tulsa, Tulsa OK, 74104; james.jennings.howard@gmail.com
3
Eni E&P, Strada Statale 9 Via Emilia 1, San Donato Milanese (MI) 20097, Italy; keith.hester@eni.com

Monitoring Core Measurements With High-Resolution Temperature Arrays
CO2 into the system while using MRI methods to monitor ports on the end-plugs to connect to a series of pumps that
the progress of the formation and dissociation (Stevens et controlled the introduction of various Àuids to the sample.
al., 2008; Baldwin et al., 2009). The laboratory experiments These pumps monitored the volumes of injected Àuids,
became more elaborate as the interest shifted from simple including water, methane and CO2, and their pressures. A
formation and dissociation rates to questions involving the separate dedicated pump controlled the con¿ning pressure
permeability of the porous media and the stability of the on the sample. The temperature of the sample was managed
hydrate-bearing sediment (Howard et al., 2011; Hester et by controlling the temperature of the NMR-insensitive liquid
al., 2011; Alemenningen et al., 2016). These experiments (Fluorinert, 3M) that Àowed in the space between the core
included a series of sensors attached to the sample holder, holder and the sample’s con¿nement sleeve that also handled
(e.g., electrodes, ultrasonic transducers) that measured the con¿ning pressure. A temperature probe was located in
electrical conductivity and sonic velocities along with the this open space between sleeve and core holder that monitored
standard monitoring of pressure, temperature, and Àuid the overall temperature of the system (Fig. 1).
volumes, for the duration of the test. The MRI images A sandpack was used for these experiments rather than
collected during many of these early experiments illustrated the pieces of sandstone core that were the focus of the ¿rst
nonuniform hydrate formation and dissociation throughout tests of the CO2-CH4 exchange mechanism. A mold using
the porous media. Only in rare cases, under special conditions, TeÀon shrink-wrap tubing was placed on top of one of the
was hydrate formation observed to follow a simple front- core holder’s end-pieces and ¿lled with ~100 grams of
like displacement down the length of the core. For several Ottawa F110 sand. This sand has an average grain size of 100
of these new experiments an additional temperature sensor ȝm and is used in a number of hydrates experiments (Waite
was added to the design beyond the thermocouples attached et al., 2011). A small diameter (1.5-mm OD, 0.5-mm ID)
to the sample holder. A ¿ber-optic array was placed within PEEK tube was positioned in the center of the mold with one
the porous media and used to monitor small changes in end inserted through one of the ports in the end-piece. The
temperature at distinct points along the length of the sample sand was added in small amounts with intervals of shaking
during hydrate formation and exchange processes. the mold to encourage closest packing of the sand grains,
especially around the PEEK tube. Once the mold was ¿lled
EXPERIMENTAL DESIGN AND SAMPLES with sand and a volume of water added to the sand, the top
end-piece of the core holder was secured to the shrink-wrap
The basic design was centered around an MRI- mold. Once the core holder was assembled and the various
compatible core holder that had suf¿cient inlet and outlet Àuid lines connected to their pumps, the ¿ber-optic line was
Fig. 1—Schematic of laboratory setup of a multiport MRI-compatible cell positioned within the MRI instrument. The magnetic ¿eld around the MRI
required all pumps, temperature control, and additional sensors be located several meters from the core holder. The OBR ¿ber optic was threaded
through a PEEK tube positioned in the sandpack constructed in the high-pressure MRI cell.

Howard and Hester
threaded through the dedicated inlet port, through the PEEK progress of the experiment. Most of the water was converted
tube in the sandpack and then out through its dedicated line to hydrate and the initial hydrate saturation was 72%.
in the outlet end-piece. It is possible to use the ¿ber-optic The effective porosity was 11% and the permeability was
line threaded through a PEEK tube in a consolidated rock 30 mD. After the hydrate formation slowed down, a short
sample, either by drilling a small hole in the core plug or by stage of water injection increased the total MRI signal.
using core halves ¿tted with a spacer. The latter approach This additional water ¿lled most of the air-¿lled pores with
was used in a number of earlier experiments on gas-hydrate liquid and created a situation that more closely simulated an
formation in porous media (Stevens et al., 2008). actual hydrate reservoir. The N2 injection stage was intended
As with earlier experiments, MRI images were collected to dissociate some of the hydrate and increase the overall
throughout the test as a way to monitor changes in Àuid and permeability of the sandpack. The increase in MRI signal
hydrate saturations. The MRI instrument (2.0-T Varian) was resulted from the loss of about half of the initial hydrate. The
sensitive to hydrogen nuclei in a liquid or dense gas state ¿nal stage was the introduction of a CO2-N2 gas mixture that
(e.g., water, methane at 8.4 MPa), but the limitations of that resulted in the formation of a CO2 hydrate and the production
instrument’s probe made it insensitive to hydrogen nuclei in of some methane. Separate analyses of the produced gas
the solid state (ice, hydrate). composition by gas chromatography veri¿ed the exchange
A distributed temperature sensor (DTS) probe was (Schoderbek et al., 2013).
used to monitor temperature changes along the length of
the sandpack during the hydrate formation and exchange
processes. The optical backscatter reÀectance (OBR) ¿ber-
optic cable is sensitive to changes in both temperature and
stress, the latter that might occur if the cable moved in the
sample during the test. One consequence of using the PEEK
tube to thread the ¿ber through the sample was that once the
core holder was positioned in the MRI instrument it did not
move for the duration of the test. Data acquisition parameters
were de¿ned to collect stations every 0.5 cm at a time interval
of 30 s during the entire experiment. The data acquisition
was restarted at major transition points in the experiment as
a data-management precaution, but had the downside that
sometimes the internal calibration was altered. Absolute
temperature values therefore are sometimes in error, though
the relative changes in temperature are internally correct.
Fig. 2—Overview of experiment that started with initial hydrate formation
The sensor has a sensitivity of 0.1oC. through the ¿nal step of injecting a CO2/N2 gas. MRI intensity indicates
the amount of free water in the sample that has not converted to hydrate.
RESULTS
The sandpack had an initial water saturation of
The assembled sandpacks had a length of roughly 11 cm approximately 60%, with the remaining pore volume ¿lled
and a diameter of 5.05 cm, creating a bulk volume of 225 cm3. with air. As methane was injected at the inlet end and the
The pore volumes ranged from 90 to 100 cm3, resulting in a temperature of the cell cooled from room temperature
porosity of approximately 40%. The initial water saturation, to 4oC, hydrate began to form. Temperature curves from
created by adding water during the construction stage, the front and back platen in the core holder along with a
ranged from 58 to 75%. Initial permeability of the water- temperature collected at the center of the sandpack provided
saturated sandpack prior to cooling was 200 to 400 mD. The several insights into the hydrate nucleation process (Fig. 3).
formation of hydrate dropped that permeability down to 30 The volume of methane consumed as it ¿rst dissolved in the
mD. Permeability of the system remained above 10 mD even cold water, then followed by actual hydrate formation, was
after the addition of more water to the remaining gas-¿lled captured by the pump connected to the sample holder. The
pores and the introduction of CO2 that formed even more small heat of methane solution at 4 hours was observed in the
hydrate. temperature sensor located in the core, but not at the platen
The test included a period of methane-hydrate formation ends. The second event of methane consumption represented
after the cell was cooled to 4oC (Fig. 2). MRI signal intensity the formation of hydrate that was captured with a several
averaged over the entire sample was used to monitor the degrees increase in temperature monitored in the core and

Fig. 3—Methane consumption (blue curve) along with temperature sensors in the core and at the end-pieces de¿ne several stages in hydrate
formation.
at the end-pieces. A third temperature spike at 14 hours

corresponded to a decrease in the methane consumption rate
as hydrate formation reached an end.
The combination of MRI images and the pro¿le
temperature curves provided more insights into the hydrate-
formation process than could be discerned from the gas-
consumption curve (Fig. 4). The MRI image at the earliest
stage of the hydrate formation (top image) is dominated
by “warm” colors that indicate the presence of sensible
hydrogen (i.e., water). As hydrate formed during the cooling
the MRI responded to a reduced signal intensity as water
was converted to hydrate, as illustrated with “cooler” colors
on the images in Fig. 4a (2nd through 4th images). The initial
image has a small spot in the lower portion of the sample, (a)
approximately one-third the distance from the inlet (left
side) that suggests the region where hydrate formation
started. The subsequent images show that hydrate formation
was localized on the bottom of the sandpack and slowly
moved upwards during the several days of initial hydrate
formation. The temperature pro¿les of the sandpack were
averaged over the time required to collect the MRI images,
approximately 2 hours. The pro¿le for the initial image (0 hr)
was uniform along the length of the sandpack with a slight
hint of temperature increase at 5 to 6 cm. The subsequent
temperature pro¿les had signi¿cant temperature increases at
the inlet end of the cell, increasing from 1.4 to 2.4oC as more (b)
hydrate was formed. It is important to note that the region
over which the temperature probe responds is unknown, so Fig. 4—(a) MRI images of sandpack during initial hydrate formation
collected over a two-day period (top to bottom) along with (b) temperature
direct comparisons of the 2D images and the temperature pro¿les collected at the same time. Hydrate formed initially at bottom of
pro¿les have limited usefulness, and that at this point should image in (a) (inlet), as represented by blue (cooler colors).
be used to illustrate general trends.

Howard and Hester
Injection of N2 was included in the experiment to remove The N2-injection step was followed by the injection
hydrate from some of the pores, especially at the inlet end of a N2-CO2 gas mixture over the following 36 hours. The
of the core. This resulted in the dissociation of some of the CO2 formed new hydrate and exchanged with methane in
hydrate, accompanied by a reduction in temperature and the hydrate structure, releasing additional methane. The hydrate
release of some of the methane from the hydrate (Fig. 5). The formation resulted in increased temperatures along the
starting point of this stage was de¿ned by the temperature length of the core, with the larger change taking place closer
pro¿le 1c-01 and the top most MRI image. The N2 injection to the inlet end (pro¿les 80 and 98 hr). MRI images collected
increased the sample permeability from the 30 mD created during the N2-CO2 injection show the changes in hydrate
by the initial hydrate formation to 55 mD. The dissociation saturation and the preferential dissociation near the inlet
of hydrate during N2 injection over the next 64 hours is end during the N2 injection as the total hydrate saturation
illustrated by the reduced temperature pro¿les (54 and 66 increased. The ¿nal hydrate saturation was approximately
hr) and the return of free-water signal in the MRI images 80% and the measured permeability was 10 mD after the
(2nd and 3rd images). The images suggest more hydrate completion of the N2-CO2 injection. Methane was collected
dissociation closer to the inlet end of the sandpack, while at the outlet end and the mechanism of CH4-CO2 exchange
the temperature pro¿les show a smaller drop in temperature was veri¿ed (Schoderbek et al., 2013).
compared to the starting point. These 2D slices of the entire A second experiment investigated the thermal effects
3D sandpack volume were selected as the centermost (16 associated with depressurization below the hydrate stability
of 32) and often did not represent the total distribution of (Fig. 6). Pressure was dropped suddenly from the initial
hydrate and free-water phases in the sample. 1,200 psi to 600 psi and then an additional drop of 400 psi
down to 200 psi over an 18-hour period. The MRI intensity
measurements were collected at an interval of 13 minutes.
There was a large drop in temperature associated with the
drop in large initial drop in pressure that was attributed to
gas expansion. As the pressure was slowly dropped from
600 psi, the temperature continued to drop at a much
slower rate. There was a lag of roughly 4 hours before there
was suf¿cient MRI signal to show the dissociation of the
hydrate into its constituent water and methane phases. The
temperature stabilized halfway through the pressure decline
and while the averaged MRI intensity was still growing.
MRI pro¿les were acquired with very fast scans for this test
compared to the 2.5-hour scans used in the initial test, so
the signal quality was not as good, but suf¿cient to observe
general trends.
(a)
DISCUSSION
The increase in temperature of several degrees centigrade

that was observed during hydrate formation agreed with an
earlier study that monitored hydrate formation in a reactor
vessel ¿lled with sediment (Liu et al., 2011). That experiment
placed several thermocouples at various depths in the
sediment mixture and monitored temperature during cooling
and hydrate formation. That experiment also observed very
rapid temperature transients during the hydrate formation
that matched the temperature events recorded in this study
(Fig. 3).
(b)
Fig. 5—(a) MRI images and (b) temperature pro¿les collected at start of
N2 injection (0 hr, top image) and the dissociation that accompanied the
injection (2nd and 3rd images, 54 and 66 hr). This was followed by N2-CO2
injection that reformed hydrate (4th image, 80 hr) to an endpoint (bottom
image, 98 hr).

(a) (b)
Fig. 6—(a) One-day depressurization experiment that followed temperature and MRI intensity trends after pressure was dropped below stability
pressure and then slowly ramped downward. (b) MRI 2D slices of the depressurization experiment collected at the beginning of dissociation (top),
when the temperature stabilized (middle) and at the end of the depressurization period (bottom).
The combination of imaging and temperature The images collected for this study were limited by
measurements for monitoring hydrate growth and exchange the capabilities of the older-generation instrument and the
reactions is complementary, each provides a distinct view demands of trying to capture water signals with their long
of the process and the mechanisms. MRI imaging is ideal T1 recovery times. A test that followed these temperature
because it can easily distinguish between free water and experiments showed how adjustments to the image
hydrate, though it is limited in that it cannot resolve exchange acquisition parameters could improve the resolution of the
reactions that occur in the hydrate state. Previous work images (Fig. 7). The key adjustment was to dope the water
showed that within the time frame of an MRI image, 15 to used in the test with CuSO4, which dramatically decreased
180 minutes, the uniformity of the image during exchange the T1 relaxation time constant of the water from 1.3 s to 50
of CO2 for CH4 indicated that large-scale dissociation was ms. This allowed for a signi¿cant reduction in recovery time
not the mechanism of exchange. This is not to imply a solid- for the measurement, meaning more scans per unit of time.
state reaction, since the reaction rates were far too fast for The second adjustment was to increase the applied magnetic
that process; rather, localized and rapid dissociation and ¿eld gradient, which resulted in more voxels per unit
reformation that took place at times much faster than MRI volume. In this study the original scan created a transverse
time resolution allowed. slice of 32×32 voxels. As the resolution was improved to
Temperature measurements also are sensitive to changes 64×64 and ultimately to 128×128 voxels, greater detail is
in state of the water-hydrate system, though perhaps not as seen in the images. The pore space ¿lled with free water is
robust as MRI imaging. The time constant for temperature represented by “brighter” colors, while the hydrate-¿lled
measurements is much faster, essentially instantaneous. pores are darker. The improved resolution highlighted the
The introduction of a ¿ber-optic line into the reactor cell nonuniform nature of the actual hydrate distribution in these
or sediment volume allows for high spatial resolution, fast- sandpacks.
time-acquisition temperature data. The small perturbations
in the temperature array eventually can be matched with
even higher resolution MRI imaging.

Howard and Hester
(a) (b) (c)

Fig. 7—Comparison of a single transverse slice in a hydrate-bearing sandpack captured at (a) 32×32 resolution, (b) 64×64, and (c) 128×128.
Increased signal intensity denoted by brighter colors represents free water. Hydrate-bearing regions are darker.
The temperature pro¿le along the length of the sandpack small internal changes in the sample provide useful
has the potential to provide insights into the kinetics of hydrate information on geochemical reactions.
formation in a porous medium, especially if the sandpack
can be monitored throughout by imaging technologies to CONCLUSIONS
detect the presence of hydrate or its loss due to dissociation.
High-resolution micro-CT or MRI methods are available to A high-resolution optical ¿ber used to monitor
generate snapshots of the sample pore and grain volumes. temperature Àuctuations associated with geochemical
The limitation lies in the positioning of the ¿ber optic within reactions in porous media can be added to the experimentalists’
the sandpack. While it provides detailed information along toolbox. This approach generates spatially resolved thermal
the length of the sample, its radial resolution is quite limited. data that can be used in turn to evaluate heat-Àow conditions
The temperature pro¿le is essentially a 1D measurement in in an experimental setup. The experiment design has to
a 3D volume. Improvements in imaging technology should include a means to limit movement of the ¿ber during the
allow for improved comparison of the temperature- and test so that it responds only to temperature Àuctuations and
hydrate-monitoring measurements. not strain. The combined temperature measurements and
The combination of laboratory techniques used in MRI images show that hydrate formation and dissociation
this experimental study on natural-gas-hydrate stability is nonuniformly distributed along the length of the sample,
and exchange can be applied to a number of special core thus providing new insights not readily obtainable by
analysis methods associated with more general reservoir other experimental techniques. The combination of spatial
characterization studies (Table 1). Temperature is often held temperature measurements and sequential imaging opens up
constant as an external control, but even in these situations, new opportunities to measure heat Àow and heats of reaction
in a wide variety of low-temperature diagenetic scenarios.
Table 1—Comparison of Laboratory Techniques Used in This Study and How They Relate to General Issues of Reservoir Characterization and
Management

ACKNOWLEDGEMENTS transfer_related_to_gas_hydrate_formation_dissociation.pdf.
Accessed February 13, 2019.
The authors wish to acknowledge Jim Stevens for Makogon, Y.F., 1997, Hydrates of Hydrocarbons, PennWell
his laboratory expertise in the design and execution of Publishing. ISBN: 978-0878147182.
Schoderbek, D., Farrell, H., Hester, K., Howard, J., Raterman, K.,
this study. Additional laboratory assistance provided by
Silpngarmlert, S., Martin, K.L., Smith, B., and Klein, P., 2013,
Marisa Rydzy, Elizabeth Krukowski, Lisebet Gravely and ConocoPhillips Gas Hydrate Production Test Final Technical
numerous summer interns. Early hydrate work was done in Report October 1, 2008–June 30, 2103, U.S. Dept. Energy
collaboration with staff and students from the University Report DE-NT0006553. https://www.netl.doe.gov/sites/
of Bergen, Arne Graue, Geir Ersland, Jarle Husebo, Knut default/¿les/netl-¿le/nt0006553-¿nal-report-hydrates_0.pdf.
Arne Birkedal and others. Funding for the Ignik Sikumi ¿eld Accessed February 13, 2019.
project was provided by U.S. DOE (Award DE-NT0006553), Sloan. E.D., and Koh, C.A., 2007, Clathrate Hydrates of Natural
JOGMEC and ConocoPhillips. Gases, 3rd Edition, CRC Press, Taylor & Francis Group. ISBN:
978-0849390784.
NOMENCLATURE Stevens, J., Baldwin, B., Graue, A., Ersland, G., Husebø, J., and
Howard, J., 2008, Measurements of Hydrate Formation in
DTS = distributed temperature sensing Sandstones, Petrophysics, 49(1) 67–73.
MRI = magnetic resonance imaging Waite, W., Santamarina, J., Chong, S., Grozic, J., Hester, K.,
NMR = nuclear magnetic resonance Howard, J., Kneafsey, T.J., Lee, J.Y., Nakagawa, S., Priest, J.,
OBR = optical backscatter reÀectance Rees, E., Koh, C., Sloan, E.D., and Sultaniya, A., 2011, Inter-
PEEK = polyether ether ketone Laboratory Comparison of Wave Velocity Measurements in
a Sand Under Hydrate-Bearing and Other Set Conditions,
SCAL = special core analysis
Paper 195, Proceedings, 7th International Conference on Gas
T1 = spin-lattice relaxation rate Hydrates (ICGH 2011), Edinburgh, Scotland, UK, 17–21 July.
https://www.pet.hw.ac.uk/icgh7/papers/icgh2011Final00195.
REFERENCES pdf. Accessed February 13, 2019.
Almenningen, S., Juliussen, H., and Ersland, G., 2016, ABOUT THE AUTHORS
Permeability Measurements on Hydrate-Bearing Sandstone
Cores with Excess Water, Paper SCA2016-082 preseted at James J. Howard is a visiting research associate at
the International Symposium of the Society of Core Analysts, the University of Tulsa and a technical advisor to DigiM
Snowmass, Colorado, USA, 21–26 August. Solution, LLC. He has worked in research and technology
Baldwin, B., Stevens, J., Howard, J., Graue, A., Kvamme, B., groups in the service sector, operating companies and
Aspenes, E., Ersland, G., Husebø, J., and Zornes, D., 2009, academia throughout his career. He has a PhD in geology.
Using Magnetic Resonance Imaging to Monitor CH4 Hydrate
Formation and Spontaneous Conversion of CH4 Hydrate to
CO2 Hydrate in Porous Media, Magnetic Resonance Imaging,
Keith C. Hester is a Flow Assurance Specialist at
27(5) 720–726. DOI: 10.1016/j.mri.2008.11.011. Eni S.a.P. based in Milan. He received a PhD in chemical
Hester, K., Howard, J., and Stevens, J., 2011, Composition Studies engineering from the Colorado School of Mines as part of
to Determine Rate and Extent of CO2 Exchange in a Hydrate- the natural gas hydrates program. He held a post-doctoral
Bearing Core, Paper 397, Proceedings, 7th International position at Monterey Bay Aquarium Research Institute
Conference on Gas Hydrates (ICGH 2011), Edinburgh, before joining the hydrates research group at ConocoPhillips
Scotland, UK, 17–21 July. https://www.pet.hw.ac.uk/icgh7/ in 2008, where he was the onsite technical advisor for the
papers/icgh2011Final00397.pdf. Accessed February 13, 2019. Ignik Sikumi pilot project in 2010–2011. He joined Eni E&P
Howard, J., Hester, K., Stevens, J., and Rydzy, M., 2011, in 2012.
Ultrasonic Velocity Measurements During Experimental CH4
Hydrate Formation and During CO2 Exchange, Paper 396,
Proceedings, 7th International Conference on Gas Hydrates
(ICGH 2011), Edinburgh, Scotland, UK, 17–21 July. https://
www.pet.hw.ac.uk/icgh7/papers/icgh2011Final00396.pdf.
Accessed February 13, 2019.
Liu, B., Pang, W., Peng, B., Sun, C., and Chen, G., 2011, Heat
Transfer Related to Gas Hydrate Formation/Dissociation,
Chapter 24 in Dos Santos Bernardes, M.A., Editor,
Developments in Heat Transfer, InTech, Shanghai, China.
http://cdn.intechopen.com/pdfs/19903/InTech-Heat_

Regular
Submissions

How the Invasion Zone Can Contribute to the Estimation of Petrophysical Properties
From Log Inversion at Well Scale?1
Thibaud Vandamme2, Emmanuel Caroli3, and Serge Gratton4
ABSTRACT
In a conventional formation evaluation process, at each depth. At the end of this step, we get an insight
the mud-¿ltrate invasion in the near-wellbore region is of the number of petrofacies and the correlation between
considered a bias that requires a well-log correction before permeabilities and porosities inside each of these. Second,
any petrophysical evaluation. The developments presented the inversion in itself is carried out in the hydrocarbon zone
in this paper show that the invasion zone is a valuable by exploiting the grouping from the ¿rst step. The vertical
source of information to estimate dynamic properties capillary equilibrium is added and updates permeabilities
that generally come only from core measurements, such (absolute and relative) as well as capillary pressure models
as permeability, relative permeabilities, capillary pressure for each facies.
curves and formation factor. In the context of this paper, we present a vertical
In this approach, the invasion process is not well and consider a radial oil-based mud invasion. We
simulated in itself, as it would lead to a very unstable also assume isotropic petrophysical parameters. The ¿nal
inverse problem within the time frame of the logging. results are compared to all available sources of data, such
On the contrary, it considers the Àuids in the invaded as NMR, WFT and cores for permeabilities, formation
domain as radially equilibrated and solves the Àuid factor and capillary pressure curves.
distribution governed at ¿rst-order by capillary pressures. The ultimate added value of such an approach is to
Due to the multimodality of the inverse problem and bridge static and dynamic petrophysical parameters from a
the uncertainties related to the mud-¿ltrate parameters, single source of data: logs. It provides a reliable ¿rst guess
the invasion zone is jointly inverted with the vertical of petrophysical and reservoir parameters at an early stage
capillary equilibrium at ¿eld-scale describing the vertical of the well evaluation. It also ensures an overall consistency
water saturation pro¿le in the reservoir for each facies. of the formation model for the whole range of facies and
The following workÀow is then used: First, the invasion is Àuid con¿gurations. The technique can even help in the
solved in the water intervals while inverting the resistivity formation heterogeneity and petrophysical upscaling when
logs. The resolved parameters are the local volume of run in a multiwell con¿guration.
¿ltrate, pseudopermeabilities and cementation factors
INTRODUCTION read entirely in either one of these regions at a time. When

this assumption is not possible, corrections are made to give
Well logs have a wide range in their depths of investigation one value in the Àushed zone (e.g., Rxo) and one value in the
around the wellbore: from a few centimeters for density, virgin zone (e.g., Rt).
neutron porosity, and shallow resistivity to more than one The Àushed zone is therefore generally considered
meter for deep resistivity. Conventional interpretations often as a bias to be removed and corrected during the log
make a piston-effect hypothesis. The near-wellbore is divided preprocessing. In the case of resistivity logs, invasion
into two zones with homogeneous parameters inside each of is used as the basic concept for inversion of the near-
them. Fluid substitution in the invaded zone is not modeled wellbore and virgin domains. Some dynamic indicators of
by physical laws but simply by volume balance: VolUWAT + permeability can be retrieved out of the resistivity pattern
VolUHC = VolXWAT + VolXHC. Each log is then considered to but this approach remains mainly empirical, qualitative and
Manuscript received by the Editor April 26, 2018; revised manuscript received November 19, 2018; December 10, 2018.
1
Originally presented at the SPWLA 58th Annual Logging Symposium, Oklahoma City, Oklahoma, USA, June 17–21, 2017, Paper JJJ.
2
MODIS, 4 Rue Jules Ferry 64000 Pau, thibaud.vandamme@external.total.com
3
TOTAL, Avenue Larribau 64000 Pau, emmanuel.caroli@total.com
4
ENSEEIHT, 2 Rue Charles Camichel 31000 Toulouse, serge.gratton@enseeiht.fr

How the Invasion Zone Can Contribute to the Estimation of Petrophysical Properties From Log Inversion at Well Scale?
generally does not consider the mudcake effect and the Àuid costly. The reason is the deterministic forward simulation
redistribution in the wellbore. Subsequently, all reference and of the whole sequence of invasion, with second-order
quantitative dynamic information is coming in the formation phenomena, while logs are generally not sensitive to the
evaluation process from core laboratory measurements and sequence of invasion as they are acquired most generally
production data (DST); these two sources of information after equilibrium between mud ¿ltrate and the native Àuids.
are both local and discrete, expensive when compared to Furthermore, the exact knowledge of the sequence of
logs and consequently quite rare over a ¿eld. The dynamic invasion is practically impossible in any classical logging
data require to be distributed vertically with logs along well situation. Logs are generally run at inappropriate time: too
sections thanks to statistical and upscaling techniques. This soon for LWD (the invasion has just started) or too late for
requires complete and unbiased dataset covering all reservoir WL (the invasion is nearly ¿nished). To overcome such
facies. Such a condition is generally dif¿cult to achieve at restrictions and get a stable mathematical problem, the main
the early stages of the ¿eld appraisal and pushes the dynamic novelty of the approach presented in this paper, along with
evaluation late in the ¿eld study, after core and production the workÀow developed that does not require any input
tests synthesis. However, unexploited data reside in the logs. or initialization from core data, considers the Àuids in the
As long as the Àuid distribution in the near-wellbore region invaded domain as radially equilibrated and solves the Àuid
can be properly resolved, the invasion process that obeys distribution as governed at ¿rst-order by capillary pressures.
dynamic rules could lead to retrieval of in-situ dynamic and The ¿rst part of this paper presents the modelization of the
petrophysical properties for every reservoir facies. The only problem based on this assumption of radial equilibrium.
condition is that invasion happens in the domain covered by The second part shows the methodology used to make a
the log readings. In order to achieve that, an interpretation complete petrophysical inversion on a ¿eld case. The third
process needs to solve the physical equations describing part presents and discusses the inversion results.
the mud-¿ltrate invasion. Thus, it may be possible to get
access very early in the ¿eld evaluation process to dynamic MODELIZATION OF THE PROBLEM
properties. The bene¿t of this could be to better design
core acquisition and DSTs. In addition, this would be an The central part of the log petrophysical inversion
essential piece of information for the upscaling phase, since consists in the invasion model and the physical description of
logs represent an intermediate scale between laboratory the Àuid equilibrium. Note that the formulations developed
experiments and well tests. in this paper are for an oil-based mud (OBM) as this type
This paper presents a method to estimate dynamic of mud still represents the majority of drilling conditions.
and petrophysical parameters, provided some physical It also ensures the direct access to drainage curves in the
hypotheses and conditions on invasion pattern are respected. water-bearing intervals.
Such an approach has been attempted already in the past. Wu
et al. (2004, 2005), based on the model built by Chenevert
Modelization of the Mud-Filtrate Invasion
and Dewan, (2001) on the invasion Àow rate, developed a
When invading a water-bearing zone, mass conservation
¿nite-difference invasion simulator by coupling Àow and
for water and mud gives:
mudcake equations. Alpack et al., (2006, 2011) used this
work to invert permeabilities from resistivity logs. Salazar
(1)
and Torres-Verdin (2008) made a sensitivity analysis on
petrophysical and mudcake parameters to evaluate their
effects on resistivity measurements. Finally, Heidari and (2)
Torres-Verdin. (2012) attempted a complete petrophysical
inversion with estimation of permeabilities and capillary
pressure curves from resistivity and nuclear logs. The use of where subscript w stands for water and o for oil (OBM in
multiphysics measurements helped to reduce nonuniqueness such a case), ȡ is density, S is saturation and q Àow rate.
in estimation of dynamic and petrophysical properties. Xu Darcy’s law gives Àow rates for each phase:
et al. (2012) used the invasion simulator to introduce a new
method for hydraulic rock typing based on conventional (3)
logs.
All these attempts suffer from several pitfalls. The
solution of the inverse problem is generally not unique (4)
and the simulation of the mud-¿ltrate invasion is quite

Vandamme et al.
where k is permeability, kr relative permeability, ȝ viscosity Capillary and Relative Permeability Formalisms
and P pressure. For the considered system, the saturations Among the large number of capillary functions that
are necessarily related each other by: exist in the literature (Brooks and Corey, 1966; Van
Genuchten, 1980; Kosugi, 1996; Xu et al., 2013), the Brook-
(5) Corey formalism was selected because it is widely used in
the reservoir domain and easy to implement. It expresses
saturation function of capillary pressure (Brooks and Corey,
Capillary pressures are introduced to express the pressure
1966) as follows:
differential between water and oil:
(12)
(6)
where Pd is the displacement pressure, Swirr the irreducible
After this conventional formalism, additional water saturation and Ȝ the pore-size distribution index.
hypotheses are used in order to simplify the problem. Our Relative permeability for the nonwetting phase is selected
approach considers the Àuids in the invaded domain as from the Burdine formalism (Burdine et al., 1953)
radially equilibrated and governed at ¿rst-order by capillary
pressures only. Thus, the compressibility term is neglected. (13)
Furthermore, gravitational effects are not taken into account
and the temperature is supposed to be radially constant. With
these hypotheses, Eqs. 1 to 6 give the following system in a Where Sw* is normalized saturation and Pc is capillary
cylindrical geometry: pressure. Combinations of Burdine and Brooks-Corey
formalisms come to:
(7) (14)
Such choices to express capillary pressures and relative

(8)
permeabilities do not impact signi¿cantly the inverted water
saturation. Different functions have been tested and have
not changed the nature of the results. The model that best
Bound constraints are given by: ¿ts a Pc and kr dataset representative of a study case is the
one to use; the approach presented in this paper is indeed
compatible with any type of petrophysical laws.
(9)
Modeling Hypotheses
(10) After series of forward modeling in reservoir sections,
the ¿ltration process can be summarized, as in Fig. 1. During
the mud-¿ltrate invasion, different regimes of ¿ltration occur
where Pform is the static formation pressure and q is the (Allen et al., 1991):
incoming mud-¿ltrate Àow rate. Finally, the following initial x First, the mudcake thickens by aggregation of the
condition completely determines the problem: solid particles contained in the mud. This is the ¿rst
static-¿ltration phase. During this phase, Àow rate is
(11) proportional to ;
x Then, mud circulation in the well annulus erodes the
mudcake that cannot thicken anymore. This is the
dynamic ¿ltration. During this phase, the Àow rate
can be considered as constant;
x Finally, after the end of circulation, a second static-
¿ltration phase occurs, until the mudcake becomes
gradually impermeable.

The bound and initial constraints are given by:
(17)
Sw(t = 0) = 1 for a water zone
This is a nonlinear partial differential diffusion equation

whose resolution is quicker than solving the system of Eqs. 7
to 11. For good reservoirs, the assumption of uniform water
pressure around the wellbore at the time of logging is quite
reasonable. Indeed, Fig. 2 shows different water saturation
pro¿les obtained for a 0.2 V/V porosity reservoir after 100
hours of ¿ltration and different permeabilities. Computations
are done with Eqs. 7 to 11 on the one hand (blue), and Eqs.
Fig. 1—Synthetic Àow-rate history as obtained for reservoir formations. 16 and 17 on the other hand (green). The differences between
the two formalisms can be observed only for permeabilities
In this paper, the ¿rst static-¿ltration phase is neglected less than 1 mD, hence reservoirs of low petrophysical
as generally limited to a few hours in most simulations. The quality and potentially poorly invaded. But for all permeable
most effective sequences for ¿ltration are both the dynamic formations (t 1 mD), the simpli¿cation is equivalent to the
and the ¿nal static ¿ltration. They are mostly dictated by full system of equations, with faster numerical performances
the exposure time of a formation and linked to the rate of and easier to combine with external constraints.
penetration, distance to TD and the duration of circulation
after drilling.
With these hypotheses, the dynamic-¿ltration Àow rate,
qcirc, is the only unknown. At a given depth, the cumulative
rate of penetration to TD and the duration of circulation
after drilling provides Tcirc, while the date of acquisition of
logs gives the maximum duration of the static ¿ltration that
constrains Tstat . qcirc , is then related to the total volume of
mud-¿ltrate invasion with:
(15)
where Vf is the total volume of mud ¿ltrate. Thus, once Vf

is known, the Àow-rate history is also de¿ned. Note that if
one of the simpli¿cations made above is not valid, it will
translate into a bias in the equivalent apparent dynamic Àow Fig. 2—Water saturation pro¿les for different permeabilities, computed
rate, qcirc, used only as bounding constraint to the Àow model. either with Eqs. 7 to 11 (without simpli¿cation) or Eqs. 16 and 17 (with
Finally, if the near-wellbore region is considered at simpli¿cation).
equilibrium, the radial water pressure does not evolve with
radius and it comes to the ¿nal expression used in this paper: Petrophysical Constraints
Sw
From Eq. 16 and after expressing the Pc terms with
the Brook Corey formalism, one can identify a diffusion-like
(16) equation with coef¿cient D as follows:
(18)

Vandamme et al.
which gives: interfacial tension of 480 dyne/cm. Normalization with

the reservoir interfacial tension has to be made when this
(19) relationship is used.
Log Modeling
When solving the diffusion equation in the invaded
In the next step, the radial water saturation pro¿le
zone, K and Pd will play the same role, therefore, the
obtained from the diffusion equation is converted into a
parameter that will be effectively inverted is neither K nor Pd
resistivity pro¿le thanks to a classical Juhasz (Juhaz, 1981)
but the product K . Pd , which will be considered as a single
and Waxman-Smits (Waxman and Smits, 1968) formalism:
parameter in what follows and thus be denoted KPd . K and Pd
require an external constraint to be individually solved. Such
a constraint is obtained through an empirical relationship of
this kind:
, (20)
where a1 and a2 are computed, thanks to an analysis carried

on an internal laboratory dataset made of 114 validated
measurements that cover a wide range of sandstone and
carbonate reservoirs. A pretty good correlation can be Where Vcl is the volume of wet clay, ‫׋‬t is total porosity,
obtained between KPd and permeability (cf. Fig. 3) with a ‫׋‬cl is clay porosity, m is cementation factor, n is saturation
correlation ratio R2 = 0.95 and coef¿cients a1 = í0.545 and exponent, a is tortuosity factor and St is total saturation.
a2 = 1.39. In a new step, the R(r) pro¿le is used to simulate resistivity
logs. For the time being, to simplify the problem in a ¿rst
attempt, induction tool-sensitivity functions have been used.
Such functions are tool dependent; they represent the radial
response of the ¿ve resistivity measurements as obtained
by a sensitivity analysis based on a forward modeling of
the effective tool in the same petrophysical context as the
study case. Another approach would have been a full tool
modeling to ensure a ¿ne coupling with the petrophysical
inversion and to solve the effect of bed boundaries at the
same time. This solution, too costly in terms of computation
time, has not been implemented for this proof of concept as
it would not signi¿cantly change the petrophysical value of
the results provided the application keeps away from bed
boundaries and thin beds.
The cumulative response of the resistivity tool is
expressed as follows:
Fig. 3—K vs. KPd for 114 validated laboratory measurements ranging
(22)
from clastics to carbonates. For later reference, red squares and green
diamonds are the points coming from the two study cases used in this
paper. Parameters ai, bi and ci have been inferred from resistivity
simulations with Maxwell equations for a decade comprised
between 1 and 10 ȍÂm. Resistivity logs are then obtained
Other correlations between K and capillary pressure with:
parameters have been referenced in the literature and can
be used also (Purcell,1949; Swanson, 1981; Huet, 2005).
It should be noted that a1 and a2 have been computed for (23)
parameters obtained in laboratory condition, with an

where Ri are the simulated resistivities, i = 1 to 5 for AT10, porosity around 22 to 25 p.u. and a slight clay content never
AT20, AT30, AT60 and AT90. exceeding 5%. Calcite cementation may occur locally and
reduces porosity to almost zero in meter-thick beds. In
contrast, the levee facies is composed of thinner sand layers
(from 5 m down to decimeter thick), are a bit more shaly (up
to 15%) and less porous (18 to 20 p.u. effective porosity).
Petrophysically, such facies are equivalent to channel sands,
simply more degraded (same textural fabric with higher clay
content). Multiple hydrocarbon pools could be proven—the
oil is light and ranges between 0.6 and 0.7 g/cm3, 1,000 to
2,200 scf/bbl and 30 to 40°API. Water salinity is around 10
to 12 g/L all along the section. All facies could be recognized
as either oil- or water-bearing at the well path.
Conventional quad-combo logs were acquired in LWD
and wireline. LWD did not show any signi¿cant invasion
whereas wireline resistivities, logged days after end of
drilling, did. Some extra logs were also run, such as NMR
and pressure tests with a dedicated mobility procedure.
Finally, note that all rock material is coming from rotary
Fig. 4—Sensitivity functions for induction resistivity logs.
sidewall cores, depth-located thanks to imagery.
Petrophysical Inversion WorkÀow

Such a forward model can potentially simulate all logs, All developments presented here after are run over a
including nuclear logs. However, the shallow and redundant thick uniform oil- and water-bearing interval for which a
readings of the neutron and density, as well as their poor batch of laboratory measurements from plugs is available.
sensitivity to saturation (in case of liquid hydrocarbon, The petrophysical inversion is carried out in three main
excluding sigma capture), revealed to be inef¿cient in the steps.
inversion process. Consequently, the rest of the paper will Step 1: Inversion of the Water Zone. The water zone is
focus only on resistivities. the ¿rst interval to be inverted in case of OBM when a clear
In the last step, a classical log interpretation is run resistivity contrast exists (to get stable inversion) in order to
to initiate the model with porosity, mineral volumes and capture ¿rst drainage curves for the OBM/formation-water
uncertainties (with a stochastic approach) as well as a ¿rst Àuid system. The objective is to iterate petrophysical and
guess of saturations (Àushed-zone saturation. Sxo, and virgin- dynamic parameters at each depth individually by solving
zone saturation, Sw). These results are used with a default set the following problem:
of petrophysical parameters to simulate the ¿ve resistivity
logs and start the inversion process. The objective is then
to update and optimize the petrophysical guess (Pc, kr per (24)
facies) to ¿t the resistivity distribution in the invaded zone
for all depths.
where x is the vector containing the unknown parameters,
THE PETROPHYSICAL INVERSION LB and UB are lower and upper bound constraints, W is the
matrix containing the uncertainties on the log measurements,
Two case studies were selected for this paper to illustrate h is the forward model (as described in the section before),
the speci¿cities and capabilities of the method (cf. Fig. 5). and y is the vector containing the logs. To ensure the well-
posedness of the problem, only three degrees of freedom are
Case Study 1 considered in the inversion:
The well was drilled with OBM. It penetrated multiple x One related to the petrophysical parameters (Swirr, Ȝ
clastic turbiditic sequences made of stacked massive channels or KPd).
and lateral levees with locally some debris Àows. The x One related to the volume of ¿ltrate which has entered
system is sealed internally by massive hemipelagites. When the formation (qcirc, Tcirc or Tstat).
in the channel axis, the net is almost 100%, with an effective x One related to the resistivity equation (Rw, m or n).

Vandamme et al.
Fig. 5—Case Study 1 layout: log data and interpretation.
First, in our case study, a NMR log is available and which we can compute qcirc. Finally, Àuid sampling and logs
constrains irreducible water saturation for each depth. Ȝ, K provide some values of Rw, which are interpolated for all
and Pd remain unknown. In the inversion, Ȝ is ¿xed initially levels. n is also determined, thanks to plug measurements.
at an arbitrary value and the petrophysical unknown K Pd . The third unknown in this case is the cementation factor, m.
Ȝ will be re-estimated during the joint inversion of the The unknown vector to be inverted at each depth is x = (KPd,
hydrocarbon zone. Second, all resistivity logs are acquired Vf , m).
very quickly after the end of circulation (24 hours). In our The constrained minimization problem is solved for each
approach, we assume the radial equilibrium to be reached depth with a sequential programming algorithm. Gradient is
when logs are acquired. The time of logging is considered approximated with ¿nite differences. The problem is stable
an upper limit for the end of static ¿ltration. Since the and does not depend on the starting point. Two hundred
formation is of good permeability, we suppose that the depth-levels could be inverted in the water zone of the case
mudcake became impermeable quickly and that Àuids could study. Results for formation factor and K Pd are shown Figs.
equilibrate 24 hours after the end of circulation. Tcirc is 6 and 7 as a function of effective porosity (not inverted,
given at each depth, thanks to the ROP curve. The second computed from conventional volumetric optimization
unknown is, consequently, the volume of ¿ltrate, from approach).

the same value for all the depth levels since the petrofacies
under consideration is quite homogenous and uniform.
Nevertheless, the absolute value of the Ȝ parameter cannot
be derived from the radial inversion process alone over a
water zone as there are not enough degrees of freedom left in
the system. Figure 8 shows two results of inversion for two
different values of Ȝ. The same distribution can be obtained
with different KPd.
Fig. 6—Inverted Formation factor in the water zone function of log-

derived effective porosity. Green line represents the equation for
formation factor with m =2: F = .
Fig. 8—Inverted KPd in the water zone function of effective porosity for
for Ȝ = 1.5 (blue points) and Ȝ = 0.6 (red stars).
A multiplicative factor Į = 4 is used to scale the inverted

KPd onto the core measurements. This value, without core
measurements, would remain unknown in the inversion
process until capillary pressures are sized and matched to
the vertical saturation pro¿le over a hydrocarbon zone.
Consequently, the radial inversion needs to be vertically
coupled with a hydrocarbon interval to get fully calibrated,
Fig. 7—Inverted KPd in the water zone function of log-derived total and to deliver absolute KPd (cf. next section with ¿eld-scale
porosity (blue points), KPd from plugs measurements (black squares) constraints).
and Į.KPd from plugs measurements, where Į is a scalar factor allowing
KPd from plugs to be in line with inverted permeabilities (here, Į = 4). However, even if not properly adjusted yet, the inverted
KPd result obtained at each depth displays a relationship of
the following form:
The formation factor (Fig. 6) derived from inversion
con¿rms the cementation factor of m = 2 used for the
conventional log interpretation. In addition, a clear (25)
correlation appears between KPd and the effective porosity
(cf. Fig. 7): KPd grows linearly on a semilog scale with Results for Ȝ = 1.5 are given in the following table:
porosity in the same proportions as for plug data. The
distribution of dynamic properties is globally captured by
the radial inversion within one facies of the water zone.
However, the result was obtained for an arbitrary choice
of pore-throat size index, Ȝ of 1.5. It is quite reasonable to take

Vandamme et al.
Table 1—Coef¿cients of the KPd Regression With Static Parameters

From Inversion of the OBM Invasion in the Water Zone (Over a Unique
Facies) (R2 = 0.77).
Fig. 10—Example of capillary curves inverted within a same petrofacies

(same WWJ function) with variable clay content and porosity (blue: ‫ = כ‬19.5
p.u., Vcl = 11.5%, K = 400 mD, red: ‫ = כ‬17.7 pu, Vcl = 19.4%, K = 170 mD).
Consequently, at the end of the ¿rst inversion step on the

training water zone, each WWJ facies function is potentially
de¿ned and links formation parameters between each other.
One can note however that the a1 coef¿cient depends on Ȝ.
Fig. 9—Inverted K Pd for Ȝ = 1.5 in the water zone as a function of
effective porosity, clay volume and irreducible saturation (R2 = 0.77). Thus, KPd is still underdetermined and constrained by the
value of Ȝ, which is supposed to be unique per facies.
The purpose of the next section is now to test the facies
This apparent relationship, not constrained in itself in functions with the hydrocarbon-bearing zones to ensure the
the inversion, is not fortuitous and captures the main drivers correct prediction of water saturation along the column at
of the differential invasion behavior of the rock facies at log ¿eld scale versus the resistivity derived Sw. This new phase
scale. Such a relationship gets a close form with a Wooddy- can also update the petrophysical model and solve additional
Wright-Johnson expression (WWJ) (Søndenå, 1992): unknowns.
Step 2: Inversion of the Hydrocarbon Zone. After the
(26) inversion of the radial capillary equilibrium depth by depth
in the invaded domain (water zone), a new constraint is added
to the system with the hydrocarbon zone where the same Pc
It depicts the link between static and dynamic functions should explain the vertical capillary equilibrium at
parameters: the four coef¿cients, ai, capture the internal ¿eld scale. This latter is expressed the following way:
textural organization of the facies that controls its dynamic
properties, hence its ability to get invaded. Each facies is
then fully characterized by such a function, a unique textural (27)
continuum with varying clay content and porosity (Fig. 10).

Where H is the Heaviside function, ȡw is the density of

water, ȡhc is the density of the hydrocarbon, z is the vertical
depth and FWL is the free water level.
The vertical saturation pro¿le is then computed with
Eq. 12, where Pc is given by Eq. 27, and compared to the
saturation derived directly from the deep resistivity from
classical Archie-based approaches. In case of thick transition
zone with movable water, log resistivities are inverted to
ensure a proper deep resistivity evaluation. Irreducible
saturations are derived from NMR and used as a constraint.
Finally, the WWJ facies function trained on the water zone
is used to get displacement pressures: for a given facies and Fig. 11—Inverted saturation pro¿le (green) vs. saturation derived from
Ȝ, displacement pressures are computed with Eqs. 20 and 25. resistivity (blue) with internal facies variation with larger clay content and
saturation overprint (green).
Thus, the two parameters that can be inverted from the
hydrocarbon zone are Ȝ and FWL. The inversion is made by
minimizing the following cost function at each depth: Step 3 – Petromodel Update and Parameter Decoupling.
All petrophysical parameters are calibrated and can be used
(28) to update the overall model. The Ȝ parameter from the
hydrocarbon zone leads to revise K and Pd at each depth in
the water interval thanks to Eq. 20. Then, in the hydrocarbon
section, variations on the Ȝ parameter are allowed to get a
perfect match with saturation. The vertical saturation is
Where x = (Ȝ, FWL), S is the saturation model and Sw
then transformed for each depth into independent Pd and K
is the saturation derived from resistivity. A Gauss-Newton
parameters thanks to Eqs. 12, 27, and 20.
algorithm is used to solve the minimization problem. The
Global results after the three inversion steps are given in
problem is well-posed and has a unique minimum over a
Figs. 12 and 13.
wide range of value for Ȝ and FWL.
The inversion result is displayed in Fig. 11, with
saturation computed by inverting the vertical capillary
equilibrium (green) and the one directly inferred from
resistivity (blue). The inverted value for Ȝ is uniformly 0.7,
which is quite different than the arbitrary value taken in
the inversion of the water zone (Ȝ = 1.5). The Ȝ parameter
does not vary much in the hydrocarbon zone and con¿rms
the hypothesis of unique facies over the section. In addition,
the full inversion has captured both, the vertical saturation
distribution and the heterogeneities overprint inside the
petrofacies (between X264 and X268 m, green tag in Fig.
11). It demonstrates that the water-zone invasion model
is compliant with the vertical saturation pro¿le within the
hydrocarbon zone, that the facies along this section is the
same (with simply internal porosity-clay content variations)
and that the classical Archie-approach is Pc-compliant in
this example. It indirectly con¿rms the choices made by the
interpreter to use uniform m and n exponents, as well as the Fig. 12—Inverted permeability within hydrocarbon zone function of total
clay correction model. porosity (blue circles) along with plug permeabilities (red squares).

Vandamme et al.
Plug permeabilities are in line with inverted characterization process at the well and ¿eld scales.
permeabilities, both, in trend and in absolute values.
The internal facies heterogeneity is also captured by the
inversion with the same dispersion of permeabilities with
plugs (cf. dashed lines). Within a facies, such a spread is
explained by the gradual change of the rock fabric (cf. green
interval in Fig. 11 located in Fig. 12 around the green dashed
line), probed by water saturation, and caused by clay plugs (a)
at pore throats that signi¿cantly increase irreducible water
saturation.
The result for capillary pressure is shown in Fig. 13
for two different depths (X861 and X862.7 m). Inverted
capillary pressure curves ¿t laboratory results from plugs,
at least for saturation greater than 0.4 V/V. A sensitivity
analysis by adding Gaussian noise on Eqs. 20 and 25
estimates uncertainties on the Pc function and does con¿rm
the ¿t obtained from inversion alone. The mis¿t observed for
low water saturations can be explained by:
(b)
x The use of MICP core data that do not reÀect real
effective pore throats at high pressures.
x The capillary pressure model is not fully adapted.
Effective capillary pressure cannot be modeled by a
unique Brooks-Corey formalism for the selected facies.
x The training model is mostly sensitive to low Pc.
Indeed, the inversion is trained mostly on the invasion Fig.13—Inverted capillary pressure: median (in red), 25 and 75
percentiles (dash lines), 5 and 95 percentiles (solid lines) along with
behavior over the water zone primarily driven by the
measured capillary pressure (black points). Results for depths (a)
capillary entry pressure. X261 m, and (b) X262.7 m.
x The hydrocarbon zone contributed to the inversion
over a thin interval. The hydrocarbon-bearing zone
thickness does not exceed 15 m, hence does not MODEL EXTENSION AND DISCUSSION OF
contribute to the inversion beyond 0.5 bar. RESULTS
All that put together, it appears that the inversion Extension to a Heterogeneous Clayey Reservoir
method cannot estimate capillary pressures beyond the range The log-based petrophysical inversion requires a
probed by logs at the well path. The approach is only able to measureable invasion pattern from resistivity logs. This
translate into Pc the main saturation driver: entry pressure if condition is easily achieved in massive, thick and nearly clay-
the model is dominated by the water zone, Swirr if the training free reservoirs, as in the case study of the previous section,
section is made of a long hydrocarbon interval at irreducible but would be more challenging in other contexts, like shaly
saturation and the Ȝ exponent, if the interval is covered by intervals or poor reservoirs. Indeed, invasion should at least
a thick transition zone. The mis¿t with real Pc curves is not generate a separation between a minimum of two resistivity
a bias, but simply translates the lack of sensitivity of the curves (depending on the location of the invasion front),
log data to some parameters. We are con¿dent the approach compliant with a monotonic invasion front and exceeding
preferably ¿ts the main Pc drivers. the uncertainties determined from tool modeling and used to
To balance such a limitation, the model could bene¿t weight the inversion. Consequently, any invasion shallower
from being applied within the same facies in a multiwell or than the depth of investigation of the shallowest resistivity
multireservoir con¿guration to ensure a uniform exposure or deeper than the deepest one cannot be resolved. The ¿rst
and sensitivity to all petrophysical parameters. Conversely, condition keeps the method out of the domain of LWD
this method is a unique opportunity to evaluate the exact application. In addition, apart from separation, the approach
added value of each source of data (log and core), their mainly relies on the quality of resistivity curves, especially
domain of redundancy and missing areas, to the facies shallow and medium, that drive the shape of the invasion

Fig. 14—Case Study 2 layout: log data and interpretation.
front. Logging conditions (tool centering and wellbore

shape) are of paramount importance on the inversion and
require a dedicated preprocessing to properly validate the
curves, correct them from skin and wellbore effects and,
if necessary, exclude some curves from the inversion. Any
error on one of the resistivity curves will result in a bias on:
x KPd for shallow resistivity
x Volume of ¿ltration for medium resistivities
x The m exponent for deep resistivity.
In order to test the robustness of the approach and express

its real added value, the method was run over a clay-rich
reservoir interval that has poor chance to be cored and with
a reduced invasion pattern. Another independent reservoir
interval from the same well, not thicker than 20 m, contains
up to 15% clay and is made of interbedded levels (Fig. 14).
Fig. 15—Inverted K Pd in the water zone function of log-derived effective
The overall petrophysical properties are of lower quality porosity.
with slightly lower permeabilities and porosities.
The inversion was run with a unique Ȝ over the 4.5-m For Ȝ = 1.5, the WWJ ai coef¿cients obtained from
thick water-bearing zone (Step 1), as shown in Fig. 15. inversion are (R2 = 0.71):

Vandamme et al.
Table 2—WWJ Coef¿cients Obtained From Independent Inversion in a

Clayey 4.5-m Thick Water Zone
In Fig. 16, the inversion results are compared to the

previous reservoir section in terms of KPd. The same
continuum of inverted petrophysical properties is still
applicable for porosities higher than 22 p.u. but the trend
diverges for porosities lower than 21 p.u. with increase of the
KPd ratio (for same Ȝ parameter). This, irrespective of Ȝ, can
be due to better permeabilities and/or higher displacement
Fig. 17—Inverted (green) and resistivity-derived (blue) water saturations
pressures (hence lower reservoir properties). along with NMR-derived irreducible water saturation (red).
The inverted permeabilities over the hydrocarbon zone

are compared to the two laboratory measurements from
sidewall cores (at depths X809 and X814.5 m) (cf. Fig.
18). They both lie roughly within the inverted k-‫ ׋‬trend.
However, the inversion highlights two permeability trends
(black and green dashed lines) not captured by the plug
measurements. Furthermore, all points around the green
dashed line (lower k-‫ ׋‬trend) are located within the X809
and X810 m interval (cf. green tag in Fig. 17). Both, dynamic
properties and saturations lead to the same conclusion that
the interval above X810 m is composed of a different facies
of lower reservoir properties that cannot be inverted with
same parameters.
Fig. 16—Inverted K Pd in the water zone (with unique Ȝ ) function of log-

derived effective porosity for the two intervals of the case study: interval
1, massive reservoir of the previous section (blue), and Interval 2, thin
clayey reservoir of Case Study 2 (red).
To choose between these two opposite scenarios,

inversion results for saturation in the hydrocarbon zone
are displayed in Fig. 17. The interval is clearly more
heterogeneous than the previous one. Some levels have high
clay content with signi¿cant irreducible water saturations.
Saturation in the X820 to X810 m interval ¿ts global
resistivity and the transition zone, in particular. However, a
larger mis¿t remains between X804 and X810 m (green tag).
Different reasons can explain such a behavior: nonadapted Fig. 18—Inverted permeability within the hydrocarbon-zone function of
Archie parameters or change of petrofacies (inversion to be total porosity (blue circles) along with plug permeabilities acquired on
run with at least two facies over the interval). plugs (red squares).

This conclusion is consistent with the result obtained in To illustrate this issue of upscaling, instead of comparing
terms of capillary pressures. Indeed, one single Ȝ parameter inverted permeabilities with plugs, results were plotted against
cannot ¿t optimally the data from two plugs; the parameter the two long-buildup permeabilities acquired at X809 and
was left free in the hydrocarbon zone and converged towards X818 m after sampling (Fig. 20). Due to the high uncertainty
distinct values at 0.7 and 0.4, respectively, above a below on the net thickness, obtained by different sources (imagery,
X810 m. After that adjustment, the inverted capillary pressure log, sedimentology), the permeabilities are displayed as a
curves match the two plug measurements below a certain Pc range instead of a single value. The two samples illustrate
level (see Fig. 19). The match is of same quality as for the the two clear facies behaviors: the sample at X818 m was
Case Study 1 with equivalent remarks: good reproduction run over the deeper facies. This facies is in line with the
of entry pressure and dominant pore throat but poor ¿t at continuum encountered in the ¿rst case study (black dashed
high Pc values due to both MICP and log-data limitations line) with simply a higher clay content and consequently
(insuf¿cient interval thickness and depth of invasion). To go downgraded dynamic behavior (permeability not exceeding
a step further, two curves were added on the graph: the mean 32 mD while an order of magnitude greater when clean).
capillary curve obtained for each subfacies above and below The second sample at X809 m was run over a distinctly
X810 m. The inverted capillary pressure at X815 (blue) is different facies, out of the ¿rst continuum, as proved by the
very close to the mean capillary curve of the deeper subfacies different Pc model. The shift in the permeability trend (green
(X840 and X810 m (black dashed line) while the result at dashed line) is captured by the log inversion while not by the
X809 m (red) is halfway between the capillary curves of the scarce plug data. Inversion results are indeed quantitatively
two subfacies. Indeed, this plug is close to the X810 m limit compliant with the large-scale permeabilities. The post-
and cannot be accurately resolved by the log data from which sampling buildup has a depth of investigation of about 10 m
average Pc curves at resistivity-scale are extracted. This around the wellbore over a 2- to 10-m thick interval. Such a
illustrates one of the aspects of the petrophysical inversion: scale is at the same order of magnitude as the deep resistivity
it captures dynamic and petrophysical properties at well vertical resolution. All petrophysical properties derived from
scale. Any mismatch with plug data is consequently not only log inversion are consequently averaged and upscaled over
a question of model but can also highlight an upscaling issue such a log resolution. Consequently, the ¿t between inversion
in case of heterogeneity. and sampling permeabilities in the k-‫ ׋‬domain illustrates that
the heterogeneity size of the formation of this case study falls
in between plug and resistivity scales.
Fig. 19—Inverted capillary curves for depths X815 (in blue) and X809 (in
red), measured curves for the same (black points) with mean capillary
pressure curves for the two facies of the study interval (green dashed
line for X804 to X810 m and black dashed line for X810 to X820 m).
Fig. 20—Inverted permeability within hydrocarbon-zone function of
porosity (blue circles) along with range of values for permeabilities
acquired with post-sampling long buildups (black and green lines).

Vandamme et al.
Case Study 2: Extension of the Method to Multifacies of the form: log(K Pd) = a1 + a2 . ‫ ׋‬+ a3 . Swirr. The parameters
The inversion workÀow was tested in a more challenging of the WWJ laws for each of the facies are shown in Table
context with clearly distinct facies. Porosity varies between 3 (in this case, the parameter Ȝ was set to an arbitrary value
5 and 20%. The ¿rst facies has good permeabilities (10 to of 1 and the correlation between the inverted laws and
100 mD) while the second one is degraded (0.1 to 10 mD). parameters, as shown in Fig. 24.
In addition, both facies are not uniformly distributed in
the interval: Facies 1 is mainly present in the hydrocarbon
zone while Facies 2 is located in the water zone. The full
inversion approach that consists of joining the radial and
vertical inversions over water and hydrocarbon zones is then
strongly challenged by the asymmetry of the problem.
The processing was launched following the procedure
described in the previous sections. Irreducible saturation is
estimated from NMR and a unique Ȝ is used primarily to
run the inversion. The raw results in terms of inverted KPd
and m are displayed on Fig. 21 to 23. Two different facies
appear with distinct cementation factors at 2.5 and 2.2, and
similarly with K Pd factors. Two different behaviors are
noticeable on the KPd í ‫ ׋‬graph, although the emerging
trends are even coarser on the KPd í Swirr crossplot. It
suggests that permeabilities are driven at ¿rst order by the
irreducible water saturation and not only porosity. Moreover,
the inverted K Pd parameters do not show a correlation with
the clay volume, Vcl. The WWJ law used in this case is thus
Fig. 22—Inverted KPd within the water zone along irreducible saturation.
Fig. 21—Inverted formation factors within the water zone (blue points).
The red line corresponds to the mean m for Facies 1 and the green line Fig. 23—Inverted KPd within water zone along effective porosity.
to the mean m for Facies 2.

(a) (b)
Fig. 24—WWJ law for (a) Facies 1, (b) Facies 2.
For a single facies, the vertical evolution of the capillary chose to use the estimated permeabilities from the buildup
pressures along the transition zone is used to calibrate the interpretation of the formation pretests (WFT, after Àuid
absolute Ȝ factor. Since the two facies in this study case do viscosity and Àow regime hypotheses). These values are
not show well-de¿ned transition zones and do not cover oil compared to the inverted permeabilities using an arbitrary
and water in equivalent proportions, the process of coupling value of Ȝ = 1, as shown in Fig. 25. As before, the inversion of
radial and vertical inversions is no longer possible, hence the water zone only accounts for permeabilities in a relative
no inversion of the mean Ȝ associated with each facies and way. A correction factor is then applied to the initial choice
absolute permeabilities. External constraints then must be of Ȝ so that the cloud of inverted permeabilities provides a
integrated to the problem in order to solve the average Ȝ for global ¿t of the WFT interpretation. The optimum Ȝ is 0.7
each facies. Based on the data available for this well, we (Fig. 26).
Table 3—WWJ-Coef¿cients Obtained for the Inversion of the Two

Facies
Fig. 25—Inverted permeabilities for a mean Ȝ of 1 (blue points) and

interpreted permeabilities from WFT (red squares).

Vandamme et al.
Fig. 26—Inverted permeabilities for a mean Ȝ of 0.7 (blue points) and Fig. 27—Inverted permeabilities within water and hydrocarbon zones
interpreted permeabilities from WFT (red squares). (blue points) compared to laboratory permeabilities (red squares).
The average Ȝ of each facies are thus calibrated, thanks to displayed on Fig. 28. Good agreement is observed for any
the pressure measurements carried out along the well. This saturation above 50%, as well as for the entry pressures.
calibration is legitimate as these pressure measurements The spread tends to become higher for low saturations. This
cover roughly the same scale as the electrical logs (average behavior is similar to what was obtained for the previous
radius of investigation of the order of a few meters). The study case and same observations apply here. In this
WWJ-type laws for each of the two facies are now completely particular case, this difference is external to the inversion
determined in relative and absolute. Final results with since it comes from the irreducible NMR saturation versus
inverted permeabilities are shown in Fig. 27 and compared the MICP measurements at high pressure. In addition, some
to laboratory measurements. capillary pressure measurements indicate the presence of a
The laboratory data lie on the same trend as the inverted multimodal pore network. This aspect cannot be accounted
parameters for each of the two facies. This result is even for in our approach because the formalism of Brooks-Corey
more interesting as the Ȝ values used for the estimation were is only made to describe a unimodal network. Formalisms for
calibrated with the pressure measurements independently multiple pore networks exists but were not incorporated in
from the laboratory data. There is thus an overall consistency the model presented in this paper. Xu et al. (2013) proposed
in the results obtained between the different scales: from a six-parameter model to describe bimodal pore networks.
the smallest scale (laboratory measurements performed The use of such models would be an improvement to our
on plugs) to the largest (pressure measurements) via the approach since all facies laws could be generated from a
intermediate scale with logs. Note that the low laboratory combination of canonical pore networks. This aspect could
permeabilities in Fig. 27 are the measurements carried over a only be assessed by adding some more constraints to the
nonreservoir facies that was not considered in the inversion. problem such as considering the invasion in the hydrocarbon
Comparison between inverted capillary pressures zone.
and measured capillary measurements for four depths is

(a) (b)
(c) (d)
Fig. 28—Inverted capillary pressure: median (in red), 25 and 75 percentiles (dashed lines), 5 and 95 percentiles (solid lines) along with measured
capillary pressure (black points) for the four depths of the case study.
CONCLUSIONS method; in-depth quality control and processing should be

run prior to any inversion to select and correct the input
The method presented in this paper proposes a complete curves for artifacts and shoulder-bed effects. A minimum of
petrophysical inversion of well logs by exploiting the mud- two resistivity curves is necessary with a separation larger
¿ltrate invasion. It can provide a reliable ¿rst guess of than the tool uncertainty to launch the processing. The ideal
petrophysical and reservoir parameters at an early stage of con¿guration to run the method is an equal exposure of
the well evaluation over a consistent approach that solves each facies to be solved in a water and oil zone. If not, the
petrophysical equations with logs. It can also ¿ll the gap approach should be run in multiwell mode or constrained
between core and test measurements in order to constraint with external data, such as formation tests or core data.
the upscaling of reservoir parameters. More than simply The inversion is robust for all granular systems, clastic
estimating independent properties, the approach builds or carbonate. Further developments are, however, necessary
a complete set of petrophysical laws for each reservoir to extend the approach to multiple pore systems and address
facies. As of the time of publication, the method presented complex carbonates. However, when going to water-based
in this paper has been tested on clastic reservoirs with oil- mud or OBM invasion in hydrocarbon zones, the fundamental
based mud and array induction tools. Note that an external formalism should be entirely reviewed to include miscible
source for irreducible saturation is required. Moreover, the mixtures and imbibition equations. Indeed, the context of a
quality of the resistivity logs is a critical condition to the water ¿ltrate entering an oil-bearing reservoir is of interest

Vandamme et al.
as, if properly solved, it potentially allows the prediction of Ȝ= pore-size distribution index
some important imbibition properties mandatory for ¿eld ȝ= viscosity (cP)
simulation and production forecast in case of water injection. ȡ= density (kg/m3)
‫=׋‬ effective porosity (V/V)
ACKNOWLEDGEMENTS ‫׋‬cl = clay porosity (V/V)
‫׋‬t = total porosity (V/V)
We acknowledge Total for releasing the well results.
Our thanks also go to the R&D reservoir program leader, Subscripts
David Gourlay, as well as to Total’s expert in mathematics o = oil
Peppino Terpolilli, for their continuous support all along the w = water
development of the paper.
NOMENCLATURE REFERENCES
Abbreviation Allen, D., Auzerais, F., Dussan, E., Goode, P., Ramakrishnan, T.S.,
Schwartz, L., Wilkinson, D., Fordham, E., Hammond, P., and
Williams, R., 1991, Invasion Revisited, Schlumberger Oil¿eld
Review, 3(3), 10–23.
FWL = free water level, (m) Alpak, F.O., Habashy, T.M., Abubakar, A., Torres-Verdin, C.,
p.u. = porosity unit, (0 to 100) and Sepehrnoori, K., 2011, A Multiplicative Regularized
LB = lower bound constraints Gauss-Newton Algorithm and its Application to the Joint
UB = upper bound constraints Inversion of Induction Logging and Near-Borehole Pressure
V/V = volume ratio (0 to 1) Measurements, Progress in Electromagnetics Research B,
29(29), 105–138. DOI: 10.2528/PIERSB10090502.
Symbols Alpak, F.O., Torres-Verdín, C., and Habashy, T.M., 2006,
a= tortuosity factor Petrophysical Inversion of Borehole Array Induction Logs:
gravity of earth (m . sí2) Part 1—Numerical Examples, Geophysics, 71(4), F101–F119.
g=
DOI: 10.1190/1.2213358.
H(·) = Heaviside function: H(x<0) = 0, H(xt0) = 1
Brooks, R.H., and.Corey, A.T., 1966, Properties of Porous Media
K= permeability (mD) Affecting Fluid Flow, ASCE Journal of the Irrigation and
kr= relative permeability (mD) Drainage Division, 92(2), , 61–90.
m= Archie cementation factor Burdine, N.T., 1953, Relative Permeability Calculations From
n= Archie saturation exponent Pore Size Distribution Data, Paper SPE-225-G, Journal of
P= pressure (bar) Petroleum Technology, 5(3), 71–78. DOI: 10.2118/225-G.
Pform = formation pressure (bar) Chenevert, M.E., and Dewan, J.T., 2001, A Model for Filtration of
Pc = capillary pressure (bar) Water-Base Mud During Drilling: Determination of Mudcake
Pd = displacement pressure (bar) Parameters, Petrophysics, 42(3), 237–250.
product between permeability and displacement Kosugi, K.I., 1996, Lognormal Distribution Model for Unsaturated
KPd =
Soil Hydraulic Properties, Water Resources Research, 32(9),
pressure (mD.bar)
2697–2703. DOI: 10.1029/96WR01776.
q= Àow rate (m3/s) Heidari, Z., and Torres-Verdín, C., 2012, Estimation of Dynamic
qcirc = Àow rate for dynamic ¿ltration (/s) Petrophysical Properties of Water-Bearing Sands Invaded
R= resistivity (ȍÂm) With Oil-Base Mud From the Interpretation of Multiple
S= saturation (V/V) Borehole Geophysical Measurements, Geophysics, 77(6),
St = total saturation (V/V) D209–D227. DOI: 10.1190/geo2012-0006.1.
S*w = normalized saturation (V/V) Huet, C.C., 2005, Semi-Analytical Estimates of Permeability
Swirr = irreducible saturation (V/V) Obtained from Capillary Pressure, unpublished MS thesis,
duration of dynamic ¿ltration (s) Texas A&M University, College Station, Texas, USA. http://
Tcirc =
www.pe.tamu.edu/blasingame/data/0_TAB_Grad/TAB_
Tstat = duration of dynamic + static ¿ltration (s)
Grad_Thesis_Archive/MS_036_HUET_Caroline_Thesis_
Vcl = volume of clay (V/V) TAMU_(Dec_2005).pdf. Accessed March 13, 2019.
Vf = volume of ¿ltrate () Juhasz, I., 1981, Normalized Qv—the Key to Shaly Sand Evaluation
W= uncertainties matrix on the log measurements Using the Waxman-Smits Equation in the Absence of Core
x = unknown parameters Data, Paper Z, Transactions, SPWLA 22th Annual Logging
y= vector of logs measurements Symposium, Mexico City, Mexico, 23–26 June.

Purcell, W.R., 1949, Capillary Pressures—Their Measurement ABOUT THE AUTHORS

Using Mercury and the Calculation of Permeability Therefrom,
Paper SPE-949039-G, Journal of Petroleum Technology, 1(2), Thibaud Vandamme graduated from INP-ENSEEIHT.
39í48. DOI: 10.2118/949039-G. He is currently a PhD student with Total and CERFACS. His
Salazar, J.M., and Torres-Verdín, C., 2008, Quantitative
research interests are petrophysics, numerical optimization
Comparison of Processes of Oil- and Water-Based Mud-
Filtrate Invasion and Corresponding Effects on Borehole
and data assimilation.
Resistivity Measurements, Geophysics, 74(1), E57íE73.
DOI: 10.1190/1.3033214. Emmanuel Caroli graduated from Ecole Normale
Søndenå, E., 1992, An Empirical Method for Evaluation of Supérieure, Ecole des Mines de Paris and the IFP School in
Capillary Pressure Data, in Worthington, P.F., and Chardaire- physical geology, petroleum geology and sedimentary basin
Riviere, C., editors, Advances in Core Evaluation III, modeling. He joined the Total group in 2003 as a mineral
Reservoir Measurement, Gordon and Breach Science geochemist. After several positions abroad in exploration
Publishers, 129í146. ISBN: 978-2881249044. operations, in 2009 he was appointed at the scienti¿c
Swanson, B.F., 1981, A Simple Correlation Between Permeabilities headquarters of Total in Pau as Formation Evaluation
and Mercury Capillary Pressures, Paper SPE-8234-G,
specialist. He is now mostly devoting his time to R&D in the
Journal of Petroleum Technology, 33(12), 2498í2504. DOI:
10.2118/8234-G. formation evaluation domain, software developments and
Van Genuchten, M.T., 1980, A Closed-Form Equation for Predicting training.
the Hydraulic Conductivity of Unsaturated Soils, Soil Science
Society of America Journal, 44(5), 892í898. DOI: 10.2136/ Serge Gratton is a professor at the University of
sssaj1980.03615995004400050002x. Toulouse, at Institut National Polytechnic of Toulouse and
Waxman, M.H., and Smits, L.J.M., 1968, Electrical Conductivities in the IRIT laboratory. He is also involved in the CERFACS-
in Oil-Bearing Shaly Sands, Paper SPE-1863-A, SPE Journal, IRIT joint laboratory. He is author of more than 80 papers
8(2), 107–122. DOI: 10.2118/1863-A. in international, peer-revised journals, his expertise includes
Wu, J., Torres-Verdin, C., Sepehrnoori, K., and Delshad, M., 2004,
Numerical Optimization, Data Assimilation and computing
Numerical Simulation of Mud-Filtrate Invasion in Deviated
Wells, Paper SPE-87919, SPE Reservoir & Engineering, 7(2),
aspects. He is member of the editorial board of the OMS
143í154. DOI: 10.2118/87919-PA. journal on optimization and organizer of international
Wu, J., Torres-Verdín, C., Sepehrnoori, K., and Proett, M.A., 2005, conferences in optimization.
The InÀuence of Water-Base Mud Properties and Petrophysical
Parameters on Mudcake Growth, Filtrate Invasion, and
Formation Pressure, Petrophysics, 46(1), 14í32.
Xu, C., and Torres-Verdín, C., 2012, Saturation-Height and
Invasion Consistent Hydraulic Rock Typing Using Multi-Well
Conventional Logs, Paper KK, Transactions, SPWLA 53rd
Annual Logging Symposium, Cartagena, Colombia, 16í20
June.
Xu, C., and Torres-Verdín, C., 2013, Pore System Characterization
and Petrophysical Rock Classi¿cation Using a Bimodal
Gaussian Density Function, Mathematical Geosciences,
45(6), 753í771. DOI: 10.1007/s11004-013-9473-2.

Loading Effects on Gas Relative Permeability of a Low-Permeability Sandstone

F. Agostini1, P. Egermann2, L. Jeannin2, E. Portier3, F. Skoczylas1, and Y. Wang1
ABSTRACT
This paper deals with a study of the gas relative sensitive to con¿nement and that the residual gas saturation
permeability of tight sandstones under loading. Speci¿c (through permeability “jail”) increases with loading. This
experiments have been designed and the experimental observation represents an additional source of complexity
measurements obtained show that, not only the absolute in the evaluation of low-permeability sandstone gas
permeability but also the gas relative permeability are reservoirs.
INTRODUCTION also induce a redistribution of the water phase in the pore

network. An understanding of the effective permeability to
Low-permeability sandstone gas reservoirs, also called gas as a function of both water saturation and loading is then
tight reservoirs, are generally considered stress-sensitive of interest to fully evaluate tight sand reservoirs.
reservoirs. Numerous laboratory tests under single-phase In this study, speci¿c experiments have been designed
Àow have shown that the absolute permeability of these to characterize these effects for sandstone samples of a
reservoir rocks decreases strongly with con¿nement. This low-permeability gas reservoir. The dedicated experimental
dependence on con¿nement is attributed to the existence setup and protocol put in place make it possible to accurately
of joints and interfaces in tight rocks, which close when control both the saturation of the sample and the loading.
loading increases, as pointed out by Walsh and Brace (1984) The full measurements obtained on a series of eight samples
and Warpinski and Teufel (1992). (from 1.5 D down to 0.08 PD) show that both the absolute
As two-phase Àow experiments are rather time consuming permeability and the gas relative permeabilities evolve with
and tricky for low-permeability sandstone, gas relative con¿nement variations. The relative gas permeabilities
permeabilities are generally estimated without con¿nement decrease while the residual gas saturations increase with
variations in SCAL studies. Various measurements show higher loading values.
that the gas relative permeability rapidly drops for a water These results introduce an additional complexity in the
saturation of 0.3 to 0.4 and becomes negligible at higher evaluation of low-permeability sandstone gas reservoirs.
saturations (Thomas and Ward, 1972; Byrnes, 1997; Shanley On the one hand, the evolution of the endpoints of relative
et al., 2004). A saturation region is indeed observed, where permeability, and mainly of the residual gas saturation, make
water and gas permeabilities are practically zero: this the assessment of the recoverable volume more uncertain.
phenomenon is called “permeability jail” in the petroleum On the other hand, a decrease of gas relative permeability
literature (Ward and Morrow, 1987; Chowdlah, 1988; under higher loading during depletion implies less favorable
Shanley et al., 2004; Cluff and Byrnes, 2010). However, production forecasts.
the relative gas permeabilities may also be modi¿ed under The paper is organized in two main parts: The ¿rst section
loading evolution with the reservoir depletion. At the pore is dedicated to a description of the speci¿c experimental
scale, some pathways allowing the Àow of the gas are likely setups and protocols used for tight rocks. The second part
to be blocked, either by the presence of water or by the closure presents the experimental results related to a speci¿c tight
of joints due to loading. In addition, the closure of joints may gas ¿eld.
Manuscript received by the Editor September 16, 2018; manuscript accepted October 31, 2018.
1
Univ. Lille, CNRS, Centrale Lille, FRE 2016 – LaMcube – Laboratoire de mécanique multiphysique multiéchelle, F-59000, Lille, France;
franck.agostini@centralelille.fr; frederic.skoczylas@centralelille.fr; yi.wang@centralelille.fr
2
Storengy, 12 rue Raoul Nordling - Djinn - CS 70001 92274 Bois Colombes Cedex; patrick.egermann@storengy.com;
laurent.jeannin@storengy.com
3
Neptune Energy International S.A. 9-11 Allée de l’Arche – Tour EGEE – 92400 Courbevoie, France; eric.portier@neptuneenergy.com

EXPERIMENTAL DATA Gas Permeability

Although the gas permeability of tight sandstone is
Porosity low compared to conventional sandstone, it is nevertheless
The total apparent volume, Vtotal, of a sample is measured suf¿ciently high to be measured with quasistationary Àow
by hydrostatic weighing. The dry mass, mdry, is obtained techniques (Skoczylas, 1996; Meziani and Skoczylas, 1999).
by oven-drying the sample at 105ºC. Mass stabilization is Gas permeability, Kg, is measured using a uniaxial steady-
assumed using a mass variation criterion, i.e., when the mass state gas-Àow apparatus (Fig. 1). The apparatus consists of
variation is less than 0.01g/week. a con¿ning cell, which can reach con¿ning pressures. Pc, as
The bulk density, Uapp, is then determined as the ratio of high as 100 MPa, together with a gas injection device. The
the dry mass, mdry, and the total apparent volume: gas used in the current study is 99%-pure argon.
A constant injection gas pressure. Pi, is applied on one
(1) end of the sample, while the other end is at atmospheric
pressure, P0. The quasisteady-state-Àow method consists of
measuring the average gas volume Àow rate, Qm, injected
Water porosity of a sample is de¿ned as follows: in the sample during a slight decrease, ǻPi, of injection
pressure, Pi, in a tank of known volume, Vt, connected at the
entrance of the sample (see Fig. 1). The mean injection gas
(2)
pressure is then equal to
where msat is the water-saturated mass, and ȡw is water density. Pm = Pi í (ǻPi /2). (4)
The same mass stabilization criterion is used to determine
the water-saturated mass, msat, the samples being fully water-
saturated under vacuum at room temperature. Note that Using Darcy’s law, gas permeability, Kg, can be estimated
water porosity measurements have been compared with He by:
and MIP porosity for quality-check purpose.
(5)
Sample Saturation
Intermediate water saturation, Sw, of a sample is related
to its mass, m, by the following equation: where ȝ is the dynamic viscosity of the gas, L the length of
the sample and S the cross section of the sample. Steady-
(3) state Àow is assumed when two successive gas permeability
measurements (at least 30 minutes apart) lead to a difference
The saturation method developed by Rilem Technical smaller than 3%.
Committees (1999) for concrete samples was used to
establish the initial water content. The dry sample is ¿rst
water saturated until the required mass, m, (Eq. 3) is reached.
The sample is then sealed using three layers of self-adhesive
aluminum and one layer of paraf¿n and kept in a climatic
chamber at 40ºC for at least 14 days. The time lapse allows
water distribution homogenization within the sample. It
has been veri¿ed that homogenization times longer than
14 days do not lead to changes in measured effective gas
permeability.
The gas permeability measurement may lead to changes
in the water saturation because of the production of water
(due to mechanical loading and gas pressure pushing the
water contained in the biggest pores). Therefore, the water
saturation indicated in the following is the average saturation
before and after permeability tests. A maximum variation of
1 to 2% of pore space has been observed. Fig. 1—Test apparatus for gas permeability.

Agostini et al.
Effective Permeability and Relative Permeability

In this study, gas permeability measurements have not
been corrected from Klinkenberg effect (Klinkenberg, 1941).
All the presented effective and “intrinsic” permeability
in this article are apparent permeability. Since all the
permeability measurements have been performed using the
same static gas pressure, Pi, the comparison of apparent and
“intrinsic” permeability can then be considered as relevant. Fig. 2—Size and cross-section view of Sample 2335.
To make the notations more readable, the term apparent will
be considered implicit in the following. All samples were collected in the “tight reservoir” facies
In order to measure the effective gas permeability at of the glaciogenic Ordovician formation. These are quartzitic
a given water saturation, the sample is ¿rst saturated and sandstones with little detrital clay content that varies from
conditioned according to the protocol described in the 0.3 to 12%, with rare heavy minerals and no feldspar. These
previous section. The quasisteady-state-Àow method is then rocks were deposited during the late Ordovician ice-house
used to measure the effective permeability to gas. conditions period, in a glacial to subglacial environment
The relative gas permeability for a given liquid saturation (Tournier et al., 2010). They experienced continuous burial
state, denoted Kr,g(Sw), is estimated by dividing the effective during the Paleozoic, reaching depths as deep as 4,500 m and
gas permeability at the saturation of the sample, Keff,g, by the temperatures of more than 160°C during the Carboniferous;
intrinsic gas permeability measured in dry state, Ki,g: leading to substantial porosity decrease by quartz overgrowth
cementation or chemical compaction diagenetic processes.
They were then uplifted during Hercynian orogeny and
(6)
have remained in a relatively stable structural position to the
present, and are now lying at depths ranging from 2,000 to
The experimental protocol aims at keeping measurement 2,500 m.
uncertainties as low as possible. The Àow-rate measurement The samples are cylinders of 37-mm diameter and 60-
using the pressure variation in a constant volume is accurate, mm long (Fig. 2). Water porosity and “intrinsic” (dry) gas
and usually used to calibrate commercial Àowmeter permeability, Ki,g, at 3-MPa con¿ning pressure have been
apparatus. Moreover, the possibility of using various measured (Table 1) for all samples. The water porosity varies
tank volumes. Vt (Fig. 1) allows to maintain the Àow-rate between 1.5 and 5.0%, while the gas permeability ranges
measurement accuracy for a wide range of gas permeability. between 0.6 and 6 10–17 m2, con¿rming that these samples can
The main source of measurement error is attributed to the be considered tight sandstones (see for example, Albrecht
variation of temperature during measurement. Therefore, and Reitenbach, 2015; Zhu et al., 2015; Gao and Lee, 2016).
adequate temperature-regulation equipment has been Measurements of the pore-size distribution (and cross-checks
installed and allows obtaining reproducible measurements of porosity) were obtained through the mercury-intrusion
with an accuracy of about ±2%. porosimetry (MIP) technique. The volume of the end-trim
samples is of the order of 2 to 3 cm3. Given the low porosity
RELATIVE PERMEABILITIES OF A TIGHT of the tight sandstones, the volume of mercury injected into
SANDSTONE FIELD some samples is very small and the experiments must be
interpreted with caution. Nevertheless, a fair agreement was
Reservoir Description and Characterization globally obtained between the MIP porosity and the sample
The samples investigated in this study came from four water porosity independent measurements considering that
cored wells located in a gas ¿eld of the Sbaa Basin, SW the end-trim properties can vary from the sample from
Algeria. This ¿eld presents a large range of rock types, from which it is extracted (Table 1). The MIP porosity appears
conventional fair permeability-medium porosity to “tight systematically higher than the sample porosity measured
reservoir” with low to very-low permeability and porosity. through water saturation under vacuum.

Table 1—Sample Porosity and Intrinsic Gas Permeability
Porosity
ki,g
Gas Relative Permeability at Low Con¿nement This sensitivity to saturation is mainly attributed to
Concerning tight sands, numerous experimental the pore-network structure composed of intergrain pores
studies have been carried out to estimate the gas relative connected by tight joints or cracks with an opening of the
permeability (Ward and Morrow, 1987; Byrnes, 2008; Wang order of 0.01 to 1 ȝm (Cluff and Byrnes, 2010). The gas
et al., 2012; Fu et al. 2015), which is generally estimated relative permeability of tight sandstone seems mainly
without con¿nement or at low con¿ning pressure. These dominated by the effect of these microcracks and joints,
results show a strong decrease in relative gas permeability which control the Àow in the porous network. As water
for a low water saturation value, typically of the order of 0.3 saturation increases, the microcracks, which correspond to
to 0.4. the smallest connected pathways, are the ¿rst to be water
All the samples considered in this study show this trend saturated. Then some pathways originally allowing the Àow
at 3 MPa of con¿ning pressure (Fig. 3). For a water saturation, of the gas are likely to be blocked by the presence of water.
Sw ‫׽‬0.2, the gas relative permeability is reduced by more The strong decrease of relative gas permeability, when
than 40% and for Sw ‫׽‬0.4 the gas relative permeability compared to more permeable sandstones, reÀects the fact
reduction reaches a factor of 10 for most of the samples. that the gas percolating network disappears rapidly, as water
saturation increases.
Increasing loading pressure is suspected to easily close
tight joints and cracks and also induce a redistribution of
the Àuid phases in the pore network. Thus, the gas relative
permeability may also depend on con¿nement. Nevertheless,
this effect on tight-gas relative permeability has not been
widely studied, probably due to the experimental dif¿culties
inherent to the measurements on tight samples. The next
section evaluates gas relative permeability at different
loading pressure.
Gas Relative Permeability as a Function of Con¿nement

The gas relative permeability is de¿ned as the ratio of
the effective gas permeability for a given saturation, Sw, and
con¿nement, Pc, and the dry permeability under the same
Fig. 3—Relative gas permeability below 3 MPa of con¿ning pressure. con¿nement:
(7)

Agostini et al.
Fig. 4—Relative gas permeability under con¿ning pressures of 3 and 40 MPa.
Figure 4 highlights the dependence of the gas relative permeability normalized by the intrinsic dry permeability at
permeability on the mechanical loading for the eight 3 MPa. denoted Kerg(Sw,Pc;3 MPa):
samples studied. Relative gas permeability curves are shown
for two different con¿ning pressures of 3 and 40 MPa. For
(8)
all samples, apart sample 4458, gas relative permeability
is reduced with increasing con¿ning pressure and the gas
residual saturation (through permeability “jail”) is also For sandstones with low permeability, “permeability
shifted (the saturation end point of gas relative permeability jail” is identi¿ed in the literature as the range of saturation
under loading at 40 MPa arises at a lower water saturation where the gas and water relative permeability are less than
than for the case loaded at 3 MPa). 2% (Shanley et al., 2004; Cluff and Byrnes, 2010). Figure
To better view the combined effect of both saturation 5 shows the effective relative gas permeability at different
and loading pressure, we de¿ne an effective gas relative loading and the 2% criterion motivated by ”permeability
jail” studies.

Fig. 5—Keff,g(Sw,Pc )/Ki,g(3 MPa) and “2% permeability jail” criterion (solid line).
Table 2—Threshold Saturation for “Permeability Jail”

These experimental results show that the relative
permeability for tight rocks is sensitive to loading. The
usual approximation in reservoir studies, which consists in
considering only the inÀuence of the loading on the absolute
permeability and not on the relative permeabilities may then
be a signi¿cant source of error.
Residual or Critical Gas Saturation Under Low

Con¿nement Pressure (3 MPa)
Sticking to the objective of ¿nding reliable and easily
measurable parameters that allow predicting the behavior of *Threshold saturations are estimated from the shape of the curve.
sandstones rocks partially saturated and under mechanical
loading, we focus in this section on the lowest con¿ning A ¿rst attempt to predict this threshold saturation is to
pressure (i.e., 3 MPa). We assume that 3 MPa of con¿nement correlate water porosity and the 2% threshold saturation
does not signi¿cantly modify the porous network when under low con¿nement (Fig. 6). However, it appears that
compared to the stress-free state. there is no correlation between critical saturation and water
For each sample, the “threshold saturation” is de¿ned porosity. The porosity accessible to water represents only the
by the saturation of “permeability jail” occurrence (Table amount of porous volume connected and does not provide
2). These threshold saturations cover a large range of values enough information on the connectivity of the porous
depending on the sample (Table 2). This dispersion should network or on the size of the pore threshold. The last two
be related to the differences in porous structure between characteristics are much more susceptible to be linked to the
samples. “permeability jail” phenomenon.

Agostini et al.
when the size of the pore opening which controls the

permeability is close to the pore size of the APEX, the
porous network loses its connectivity. Indeed, a fairly
good correlation between the pore size associated to the
appearance of “permeability jail” (see Table 4) with APEX
pore size (Fig. 7) is obtained (“permeability jail” pore size
around 80% of APEX pore size).
Fig. 6—Critical saturation below 3 MPa of con¿ning pressure as a

function of water porosity.
The mercury-intrusion porosimetry (MIP) provides a

measurement of the porous structure and connectivity at low
con¿nement pressure (for example, Nooruddin et al., 2014;
Guise et al., 2017). As there is a fairly good correlation
between the porosity measured by mercury porosimetry
and that measured with water, we assume that there is also
a correlation between mercury and water saturations for
Fig. 7—Correlation between pore size of “permeability jail” at 3 MPa and
a given saturated pore size. The mercury injected into the
APEX pore size.
porous network is a nonwetting Àuid and begins by ¿lling the
larger pores. We can then use mercury-porosimetry results
to interpret and explain threshold-saturation dispersion. On Thus, the MIP curves could be used before any relative
one hand, using the critical saturations of “permeability jail” permeability measurements to predict the water saturation
(Table 2) and the pore-size distribution obtained by mercury likely to cause a hydraulic cutoff at low con¿nement
porosimetry, it is possible to estimate the critical pore size pressure, as shown in Fig. 7.
(Table 3) associated to the threshold saturation, for which However, keeping in mind that the objective is to predict
the pore network no longer percolates gas, i.e., “permeability the apparition of “permeability jail” in in-situ conditions, the
jail” has been reached. pore-network characteristics for higher con¿ning pressure
need to be known in order to predict saturation threshold at
Table 3—Pore Size at the Critical Saturation of “Permeability Jail” Below this con¿ning pressure. A perspective to better characterize
3 MPa Con¿ning Pressure threshold saturations under loading would be to measure
capillary pressure under loading (Guise et al., 2017) or to
estimate it using pore network modeling.
Table 4—Pore Size at the APEX
On the other hand, APEX represents the pore inlet

diameter for which porous network connectivity is reached
(Swanson, 1981) and may be graphically determined using
MIP results (Table 4). According to the APEX de¿nition,

CONCLUSIONS ȡw = water density

‫׋‬MIP = MIP porosity
Tight reservoirs, due to their high capillary pressures, have ‫׋‬w = water porosity
a very wide transition zone. Uncertainty about gas mobility may
be responsible for incorrect evaluation of the recovery factor.
A non-negligible uncertainty in the evaluation of the tight-gas REFERENCES
potential is related to the dependence of gas mobility on both
loading and water saturation. Albrecht D., and Reitenbach, V., 2015, Investigations on Fluid
This experimental study shows that tight-gas relative Transport Properties in the North-German Rotliegend Tight
permeability is very sensitive to con¿nement. It may then be Gas Sandstones and Applications, Environmental Earth
recommended to estimate tight-gas relative permeabilities under Sciences, 73(10), 5791–5799. DOI: 10.1007/s12665-015-
loading, even if it remains time-challenging in the case of a 4322-x.
reservoir study. Byrnes, A.P., 1997, Reservoir Characteristics of Low-Permeability
Sandstones in the Rocky Mountains, The Mountain Geologist,
These experimental results have been interpreted in terms
34(1), 39–51.
of connectivity of the porous network, which is dominated by
Byrnes, A.P., 2008, Issues With Gas Relative Permeability in Low-
microcracks and joints. Below 3 MPa con¿nement pressure, the Permeability Sandstones, in Cumella, S.P., Shanley, K.W.,
threshold saturation for a relative gas permeability of 2% was and Camp, W.K., Editors, Understanding, Exploring, and
successfully correlated with the APEX pore size determined Developing Tight-Gas Sands, 2005 Vail Hedberg Conference,
from MIP results. This correlation remains to be con¿rmed under AAPG Hedberg Series 3, 63–76. ISBN: 978-1629810324.
higher loading and it may be useful to have a characterization DOI: 10.1306/13131050H31483.
of the pore-network properties under loading to predict the Chowdlah, P., 1988, InÀuence of Water-Desaturation Technique and
threshold saturation under in-situ conditions. This will be the Stress on Laboratory Measurement of Hydraulic Properties
subject of future work using either correction of MIP curves or of Tight Sandstones, Paper SPE-15210, SPE Formation
pore-network modeling allowing for loading evolution. Evaluation, 3(4), 679–685. DOI: 10.2118/15210-PA.
Cluff, R.M., and Byrnes, A.P., 2010, Relative Permeability In Tight
NOMENCLATURE Gas Sandstone Reservoirs—The “Permeability Jail” Model,
Paper LLLL, Transactions, SPWLA 51st Annual Logging
Symposium, Perth, Australia, 19–23 June.
Abbreviations:
Fu, X., Agostini, F., Skoczylas, F., and Jeannin, L., 2015,
MIP = mercury-intrusion porosimetry
Experimental Study of the Stress Dependence of the Absolute
and Relative Permeabilities of Some Tight Gas Sandstones,
Symbols International Journal of Rock Mechanics and Mining
Keff,g, = effective gas permeability Sciences, 77, 36–43. DOI: 10.1016/ijrmmms.2015.03.005.
Gao, H., and Li, H.A., 2016, Pore Structure Characterization,
Kerg = effective gas relative permeability normalized Permeability Evaluation and Enhanced Gas Recovery
by the intrinsic dry permeability at 3 MPa
Kg = gas permeability Techniques of Tight Gas Sandstones, Journal of Natural
Gas Science and Engineering, 28, 536–547. DOI: 10.1016/j.
Ki,g, = intrinsic gas permeability jngse.2015.12.018.
Guise, P., Grattoni, C., Allshorn, S, Fisher, Q.J., and Schiffer, A.,
Kr,g, = relative gas permeability
2017, Stress Sensitivity of Mercury Injection Measurements,
L= length of the sample Paper SCA2017-011 presented at the SCA International
mdry = dry mass Symposium, Vienna, Austria, 27 August–1 September.
msat = water-saturated mass Klinkenberg L.J., 1941, The Permeability of Porous Media
Pc = con¿ning pressure to Liquids and Gases, Paper API-41-200, Drilling and
Pi = injection pressure Production Practice, American Petroleum Institute.
P0 = atmospheric pressure Meziani, H., and Skoczylas, F., 1999, An Experimental Study of
Pm = mean injection gas pressure the Mechanical Behaviour of a Mortar and of its Permeability
Qm = average gas volume Àow rate Under Deviatoric Loading, Materials and Structures, 32(6),
cross section of the sample 403–409.
S=
Nooruddin, H.A., Hossain, M.E., Al-Yousef, H., and Okasha, T.,
Sw = water saturation 2014, Comparison of Permeability Models Using Mercury
Vt = tank volume Injection Capillary Pressure Data on Carbonate Rock Samples,
Vtotal = sample total apparent volume Journal of Petroleum Science and Engineering, 121, 9–22.
DOI: 10.1016/j.petrol.2014.06.032.
ȝ = dynamic viscosity of the gas
Rilem Technical Committees, 1999, RILEM TC 116-PCD:
ȡapp = bulk density

Agostini et al.
Permeability of Concrete as a Criterion of its Durability, Patrick Egermann is currently

Materials and Structures, 32(4), 174–179. Senior reservoir engineer and energy
Shanley, K.W., Cluff, R.M., and Robinson, J.W., 2004, Factors solutions coordinator with Storengy
Controlling Proli¿c Gas Production From Low-Permeability (Engie af¿liate). He is currently involved
Sandstone Reservoirs: Implications for Resource Assessment,
in various reservoir engineering topics
Prospect Development, and Risk Analysis, AAPG Bulletin,
88(8),1083–1121. DOI: 10.1306/03250403051. covering underground gas storage,
Skoczylas, F., 1996, Ecoulements et Couplages Fluide-Squelette geothermal and heat storage applications.
dans les Milieux Poreux: Etudes Expérimentales et He holds MS degrees in reservoir
Numériques, Université de Science et Technologie de Lille. engineering from IFP school and in hydraulic/Àuid mechanics
Swanson B.F., 1981, A Simple Correlation Between Permeabilities from ENSEEIHT and a PhD degree in Earth Sciences and
and Mercury Capillary Pressures, Paper SPE-8234, Journal Environment from the University of Toulouse.
of Petroleum Technology, 33(12), 2498-2504. DOI:
10.2118/8234-PA. Laurent Jeannin is Senior
Thomas, R.D., and Ward, D.C., 1972, Effect of Overburden
geomechanics and reservoir engineer.
Pressure and Water Saturation on Gas Permeability of Tight
He joined Storengy in 2016, where
Sandstone Cores, Paper SPE-3634 Journal of Petroleum
Technology, 24(2), 120–124. DOI: 10.2118/3634-PA. he is currently working for various
Tournier, F., Pagel, M., Portier, E., Wazir, I., and Fiet, N., 2010, underground gas storage and geothermal
Relationship Between Deep Diagenetic Quartz Cementation projects. He holds a MS degree (EMSE)
and Sedimentary Facies in a Late Ordovician Glacial and a PhD in Physics (ENSLyon).
Environment (Sbaa Basin, Algeria), Journal of Sedimentary
Research, 80(12), 1068–1084. DOI: 10.2110/jsr.2010.094. Eric Portier is Senior reservoir
Walsh J.B., and Brace W.F., 1984, The Effect of Pressure on geologist and engineer with 23 years of
Porosity and the Transport Properties of Rock, Journal of experience in the Oil & Gas industry
Geophysical Research, 89(B11), 9425–9431. DOI: 10.1029/
with Gaz de France, Engie and now
JB089iB11p09425.
Wang, Y., Chen, Z., Morah, V., Knabe, R.J., and Appel M., 2012, Neptune Energy. He was involved in
Gas Phase Relative Permeability Characterization on Tight ¿eld developments as well test engineer
Gas Samples, Petrophysics, 53(6), 393–400. and sedimentologist, in onsite well-
Ward, J.S., and Morrow, N.R., 1987, Capillary Pressure and Gas test supervision, interference tests and
Relative Permeabilities of Low-Permeability Sandstone, well-test pressure transient analyses, core description,
Paper SPE-13882, SPE Formation Evaluation, 2(3), 345–356. sedimentological and diagenesis studies, and special core
DOI: 10.2118/13882-PA. analyses. He is also working as supervisor of various R&D
Warpinski, N.R., and Teufel, L.W., 1992, Determination of the projects aiming at understanding parameters controlling
Effective-Stress Law for Permeability and Deformation in
reservoir quality distribution and its prediction.
Low-Permeability Rocks, Paper SPE-20572, SPE Formation
Evaluation, 7(2),123–131. DOI: 10.2118/20572-PA.
Zhu, P., Lin, C., Wu, P., Fan, R, Zhang, H., and Pu, W., Frédéric Skoczylas is full professor
2015, Permeability Prediction of Tight Sandstone in Centrale Lille. He has obtained his
Reservoirs Using Improved BP Neural Network, Open PhD in 1988 and post-doctoral degree
Petroleum Engineering Journal, 8(1), 288–292. DOI: in Centrale Lille in 1996. His research
10.2174/1874834120150625E010. is conducted in the LaMCube – CNRS-
FRE 2016 (multiphysics and multiscale
ABOUT THE AUTHORS mechanic laboratory).
Franck Agostini is assistant Yi Wang obtained his PhD in

professor (accredited to supervise Centrale Lille (LaMcube – CNRS-
research) in Centrale Lille. He has FRE 2016) in 2016. He now works as
obtained his PhD in 2006 and postdoctoral Senior Professional Manager of Civil
degree in Centrale Lille in 2018. His Engineering for New Town Holdings in
research is conducted in the LaMCube Suzhou City in China.
– CNRS-FRE 2016 (multiphysics and
multiscale mechanic laboratory).

PETROPHYSICS, VOL. 60, NO. 2 (APRIL 2019); PAGES 335–347; 14 FIGURES. DOI: 10.30632/PJV60N2-2019a10
Borehole Acoustic Imaging Using 3D STC and Ray Tracing to Determine Far-Field
ReÀector Dip and Azimuth1
Nicholas Bennett2, Adam Donald3, Sherif Ghadiry4, Mohamed Nassar4, Rajeev Kumar5, and Reetam Biswas6
ABSTRACT
A new sonic-imaging technique uses azimuthal image by providing dip and azimuth for each event that can
receivers to determine individual reÀector locations be used in further downstream boundary or discontinuity
and attributes, such as the dip and azimuth of formation characterization.
layer boundaries, fractures, and faults. From the ¿ltered A ¿eld example from the Middle East is presented
waveform measurements, an automated time pick and in which a carbonate reservoir was examined using this
event-localization procedure is used to collect possible technique and subsequently integrated with wellbore
reÀected arrival events. An automated ray-tracing and images to provide insight to the structural geological
3D slowness time coherence (STC) procedure is used to setting, which was lacking seismic data due to surface
determine the raypath type of the arrival event and the constraints. Structural dips were picked in the lower zone
reÀector azimuth. The angle of incidence of the reÀected of the main hole and used to update the orientation of
arrival is related to the relative dip, and the moveout in stratigraphic formation tops along the well trajectory. 3D
3D across the individual sensors is related to the azimuthal surfaces were then created and projected from the main
orientation of the reÀector. This information is then used to hole to the sidetrack to check for structural conformity.
produce a 3D structural map of the reÀector, which can be One of the projected surfaces from the main hole matched
readily used for further geomodeling. the expected depth of the formation top in the sidetrack
This new technique addresses several shortcomings in but two were offset due to the possible presence of a fault.
the current state-of-the-art sonic-imaging services within This was con¿rmed by parallel evaluation of the azimuthal
the industry. Similar to seismic processing, the current sonic-imaging data acquired in the main hole that showed
sonic-imaging workÀow consists of iteratively testing an abrupt change in the relative dip of reÀectors above
migration parameters to obtain a 2D image representing and below the possible fault plane using the 3D STC and
a plane in line with the desired receiver array. The image ray tracing. Dip patterns from both wells showed a drag
is then interpreted for features, which is often subjective effect around the offset formation tops, further con¿rming
in nature and does not directly provide quantitative results the presence of a fault. A comparison of the acquired
for the discrete reÀections. The technique presented here, borehole images pinpointed the depth and orientation of
besides providing appropriate parameter values for the the fault cutting both wells to explain the depth offset of the
migration workÀow, further complements the migration projected 3D formation top surfaces.
INTRODUCTION near-wellbore region from ¿ltered sonic waveforms. This

approach has been helpful for formation evaluation and
The approach to sonic imaging developed through the other purposes, because such images typically have much
works of Esmersoy et.al. (1998), Haldorsen et. al. (2006), higher resolution than what can be obtained via surface
Hirabayashi (2016), Hornby (1989), Tang (2004), Yamamoto seismic or vertical seismic pro¿le (VSP) measurements that
et al. (1999), and others has been to use migration, particularly employ a lower source frequency, though, of course, the
Kirchhoff migration and its variations, to form images of the aperture of the acoustic array and higher source frequency in
Manuscript received by the Editor July 19, 2018; revised manuscript received September 17, 2018; manuscript accepted September 20, 2018.
1
Originally presented at the SPWLA 59th Annual Logging Symposium, London, England, UK, June 2–6, 2018, Paper JJJJ.
2
Schlumberger-Doll Research, 1 Hampshire Street, Cambridge, MA 02139 USA; nnbennett@slb.com
3
Schlumberger, 4-8 Nicolae Titulescu Street, 011141 Bucharest, Romania; donalda1@slb.com
4
Schlumberger, 25 Masr Helwan Road, Maadi, Cairo, Egypt; sghadiry@slb.com; mnassar@slb.com
5
Schlumberger Middle East S.A., Abqaiq Road, Dhahran, P.O. Box 2836, Al-Khobar 31952, Kingdom of Saudi Arabia; rkumar9@slb.com
6
Jackson School of Geosciences, The University of Texas, 2305 Speedway Stop C1160, Austin, TX 78712-1692; reetam@utexas.edu

Borehole Acoustic Imaging Using 3D STC and Ray Tracing to Determine Far-Field ReÀector Dip and Azimuth
the range of 1 to 10 kHz only allow for comparatively more AUTOMATED TIME PICK
localized imaging near the well track. That being said, there
has been increasing desire to use sonic imaging to help build To determine a set of candidate arrival events, which
geological models of the near-wellbore region for purposes may or may not correspond to reÀectors near the wellbore, we
of structural geology and discrete fracture network modeling employ an automated time-picking procedure that consists
and other purposes. of a tau-P transform (Diebold and Stoffa, 1981) followed
One approach to converting the sonic-imaging migration by an event-localization procedure. These are applied
image to a structural model of the subsurface would be to to the ¿ltered waveform measurements in the common
perform feature extraction from these migration images, offset domain. For our ¿eld-data example, the waveform
either manually or automatically. However, it is not possible measurements were acquired using a sonic tool with 13
to extract reÀector dip and azimuth directly from these stacked rings of receivers equally spaced along the axis of the tool
2D migration images. In addition, sonic-imaging migration sonde each with eight receiver sensors equally distributed
images are known to frequently contain artifacts and be noisy azimuthally. The waveforms were then ¿ltered using a series
due to the relatively weak amplitudes of reÀected signals, of simple median ¿lters to reduce the interference of the
and there are often signi¿cant challenges in formulating direct borehole modes. The ¿ltered waveforms recorded
these migration images. Important migration parameters by the second receiver in the seventh receiver ring of our
include (a) the appropriate choice of velocity model requires acoustic tool are shown in Fig. 1. The automated time-pick
determining the raypath type of the reÀected arrivals, (b) the procedure is applied to the waveforms recorded by all eight
azimuth of the 2D imaging plane requires determining the receivers in the seventh receiver ring, as a reÀection will
azimuth of the impinging reÀected wave¿elds, and (c) the sometimes appear more prominently to receivers facing the
appropriate choice of migration aperture requires knowledge corresponding reÀector than to receivers facing away from
of the reÀector’s relative dip. Though sonic-imaging experts the reÀector.
may have some intuition for the appropriate values for these The user con¿gures the tau-P processing according to
and other important parameters, an iterative approach to the characteristics of the events observed or sought in the
testing different combinations of migration parameters often ¿ltered waveforms. First, and most importantly, the user
results in a labor- and compute-intensive workÀow and to chooses the time domain in which to conduct this portion
uncertainty as to the subjectivity of the imaging results. of the workÀow. For example, the user can choose to look
To improve the current state-of-the-art sonic-imaging for reÀections coming between the onset of the direct
workÀow and to construct digital models from sonic- compressional and direct shear arrivals (indicated by the
imaging measurements, we propose an automated procedure wavy blue and orange lines in Fig. 1) if there is an interest
that detects and characterizes the reÀected arrivals within in fractures or other features relatively close to the wellbore.
the ¿ltered wave¿eld and then maps those features to their Alternatively, the user may wish to look for reÀections
corresponding reÀectors. This automated procedure consists arriving after the Stoneley from reÀectors further from
of (a) a time-picking procedure that forms a collection of the wellbore. Second, the user can also specify the events’
candidate arrival events, followed by (b) an automated ray steepest moveout as a function of measured depth. The
tracing and 3D slowness time coherence (3D STC) (Bennett, default value is 400 ȝs/ft, where 0 ȝs/ft refers to events that
in press) procedure that characterizes the candidate events in would appear horizontal in Fig. 1. Finally, the user divides
terms of their raypath type, azimuth, and the position of the the entire measured-depth interval into shorter intervals (the
corresponding reÀectors relative to the well track and their default length is 100 ft) and applies the tau-P transform to
relative prominence. These 3D maps of the reÀectors provide waveforms in each subinterval. In our ¿eld-data example,
a sonic-imaging-based digital subsurface model that can be we chose to apply the tau-P transform in 50-ft windows
used for downstream processing. In addition, the automated looking for events coming after the compressional whose
procedure provides recommended migration parameters and steepest moveout was 200 ȝs/ft.
a helpful QC mechanism for the ¿nal migration image. The tau-P processing itself proceeds by locating a subset
We introduce our proposed workÀow in the context of of peaks in the tau-P domain according to their relative
a ¿eld-data example from the Middle East where seismic height while prescribing a minimum separation distance
data were not available due to surface constraints. The between the peaks to avoid an excessive number of peaks
sonic-imaging-based digital model provided insight to the for any one particular waveform feature. The parameters that
structural geological setting and was readily integrated with control this peak selection are customized for sonic logging
the interpretation from borehole-image measurements.

Bennett et al.
measurements and typically not exposed to the user. The monopole source, and 0.8 ms, when using a lower-frequency
results of applying this tau-P processing is a collection of dipole source. Portions of these waveform corridors appear in
line segments such as those shown in green in Figure 1. Fig. 1. The waveform values within two of the time-window
To determine the location of the arrival event along these corridors from Fig. 1 are displayed in Fig. 2. Recording time
line segments, an event-localization procedure is performed. within the time-pick corridor is shown along the left vertical
Along each tau-P line segment, a time-window corridor is axes, and measured depth along the time pick is shown along
created, typically of duration 0.2 ms, when using an 8-kHz the horizontal axes.
Fig. 1—Filtered waveforms shown in common offset domain for the second receiver in the seventh receiver ring with results of tau-P (green lines) and
event-localization procedures (black boxes).
Fig. 2—Background images show wave¿elds resampled along tau-P line segments for events 4 (a) and 204 (b). The yellow dots plotted on the black
curves L(j) indicate the measured depth location of the corresponding event along the tau-P line segment. The score refers to the event’s peak value
Ec(jmax)coh(jmax).

The event-localization procedure ¿nds the measured- peak score value Ec(jmax)coh(jmax), determines its relative
depth window for the arrival event within this corridor by shading. These log displays can be very helpful for discerning
creating a function that appears as the superimposed black whether an event appears consistently on multiple azimuths.
curve in Fig. 2. Let the waveform values within this time
window be w(i,j) with i being the time index and j being the
measured depth index. De¿ne a coherent energy Ec(j), total
energy Ei(j), and coherence functions coh(j) as
, (1)
, (2)
. (3)
Here T is the duration of the time window (0.2 ms or 10

samples in this case, as the waveforms have a 20-ȝs sampling
rate), and M is the length of a short window in measured
depth along the horizontal axis of the subpanels that is much (a)
shorter than the 50-ft length of the tau-P domain. The forms
of Eqs. 1 to 3 follows the general form of classical STC of
Kimball and Marzetta (1984). The black curves overlaid on
top of the resampled wave¿eld and graphed against the right
vertical axes in Fig. 2 are normalized versions of the curves
L(j) = Ec(j)coh(j) whose two components are intended to
emphasize coherent events with high energy.
If we designate max L as the maximum value that
L attains, and jmax the measured depth index where that
maximum is attained, then the event-localization procedure
consists of ¿nding the last index jstart < jmax for which
L(j) < F max L as well as the ¿rst index jstop > jmax for which
L(j) < F max L where 0 < F <1 is a user de¿ned parameter
whose default value is 0.15. The two indices jstart and jstop are
(b)
shown as yellow dots along the curves L(j) in Fig. 2 and
determine the portion of the time-window corridor used Fig. 3—Log summary view (a) of automated time-picking procedure
results for the eight receivers in the seventh receiver ring. The red box
to highlight the event in Fig. 1. The peak value, Ec(jmax) highlights the log summary for the results shown in Fig. 1. The vertical
coh(jmax), is de¿ned to be the score of the event and is also positions of the shaded blocks correspond to the measured depths of
used to rank the localized arrival events. Figure 2 shows corresponding event time picks as illustrated by dashed vertical line in
(b). The shading of each box refers to the relative prominence or score
the event-localization procedure results for the 4th and 204th of the corresponding arrival event.
highest ranking events with peak values 10.73 and 2.47,
respectively.
Although the black-box waveform overlay in Fig. 1 is AUTOMATED RAY TRACING AND 3D STC
helpful to draw the user’s attention to events that may be
present, the log summary shown in Fig. 3a is also helpful A ray-tracing and 3D STC procedure for characterizing
for quickly scanning through the processing results. The a single arrival event identi¿ed by a manual time pick, as
vertical axis is the event measured depth, and the horizontal well as a much more detailed discussion of 3D STC are
axis represents the index of the sensor in the seventh ring of presented in Bennett (in press). Here we show the results of
receivers where the event is observed. An event’s measured using that procedure in the context of our Middle East ¿eld-
depth in the log display is illustrated by the dashed vertical data example where many candidate arrival events identi¿ed
black line in the cartoon shown in Fig. 3b, and the event’s by the automated time-picking procedure are evaluated.

Bennett et al.
Fig. 4—Single sensor view of ¿ltered waveforms and overlay of automated ray tracing and 3D STC procedure results. The cyan color of the boxes
identi¿es these arrival events as PP reÀections. The time pick for event 24, shown in green, is used for the ray-tracing inversions in Fig. 5. The two
horizontal black bars along the time pick identi¿es the portion of the wave¿eld used for 3D STC analysis in Fig. 6.
The same ¿ltered waveforms shown in Fig. 1 now appear each measured depth position along the time pick.
with an overlay produced by the automated ray-tracing and Besides an estimate of the reÀector’s relative dip, the ray-
3D STC procedure in Fig. 4. The waveform overlay encodes tracing inversion also provides an estimate of the inclination
the ray-path type of the arrival event. Ray-path type refers angle between the axis of the tool sonde and the impinging
to the propagation mode (compressional or shear) from the reÀected wave¿eld at the receiver array, as measured from
source to the reÀector and then from the reÀector to the the source direction for each measured depth station. For
receiver array. We typically consider PP and SS reÀected the ray paths highlighted by the dashed lines in Figs. 5a and
arrivals as well as PS and SP mode-converted arrivals that Fig. 5b, the estimated inclination angles are 78° and 18°,
typically arise from fractures. PP and SS reÀected arrivals respectively. Since the mis¿t between the time pick and the
are coded with a cyan and orange hue, respectively, while PS inverted travel times are quite similar (0.4% for both PP and
and SP mode-converted events are coded with a magenta and SP ray-path types), we observe that the ray-tracing inversion
brown hue, respectively. Figure 4 shows that the events from (i.e., the time pick) cannot, by itself, be used to distinguish
our ¿eld-data example are mostly PP reÀections, though a whether event 24 is PP or SP, only that we may expect that
few SS reÀections and PS refracted arrivals are present at the incidence angles of their corresponding P waves at the
other measured depth locations. receiver array should differ by around 60°. We note that it is
We review the ray-tracing and 3D STC processing for unlikely that event 24 is either PS or SS, as their ray-tracing
event 24 that appears as the event with the green centerline in inversion errors are much higher (16 and 36%, respectively).
Fig. 4 near measured depth (MD) X145 ft. For each possible To assess the consistency of the ray-tracing inversion
ray-path type, we ¿rst conduct a ray-tracing inversion of the results with the portion of the wave¿eld designated by the
time pick for the position of the reÀector relative to the well time pick, we compute 2D slices of the wave¿eld’s 3D STC
track. The ray-tracing inversion results for the PP and SP representation at each measured depth station along the time
ray -path types are shown in Fig. 5 where the well track lies pick, examples of which are shown in Fig. 6. These two
along the horizontal axis, and the ray paths are shown from slices correspond to the measured depth station identi¿ed by
the source (gray circle) to the reÀector (black star) and then the two horizontal black bars in Fig. 4 and the dotted-line ray
from the reÀector to the receiver sensor (black square) for paths in Fig. 5.

Fig. 5—Ray-tracing inversion results for the time pick shown in green in Fig. 4 (a) PP, and (b) SP ray-path type interpretations. Gray circles and black
squares correspond to the measured depth positions of the source and seventh receiver ring, respectively, along the time pick, and the black stars
represent the estimated reÀector positions.
Fig. 6—3D STC representation in spherical coordinates for (a) PP, and (b) SP ray-path type interpretations of arrival event indicated in Fig. 5 for the
measurement station with the dashed ray paths, and also indicated in Fig. 4 with the two horizontal black bars along time pick No. 24.
3D STC is a relative of the classical STC of Kimball and type, and is obtained from the P and S slowness logs for this
Marzetta (1984, Eq. 3), except that we use a 3D slowness well; ș is shown along the vertical axis for a short range of
vector s = (sx, sy, sz) and the relative 3D positions of the inclination angles centered on the value estimated from the
individual receiver sensors ȡj,k = (rtoolcos Įk, rtool sin Įk, zj í z1 ) ray tracing; and the full range of azimuth angles, ĳ, is shown
to formulate our coherent energy estimator along the horizontal axis where 0° represents the propagation
of the wave¿eld towards the north direction. Further details
(4) of the construction and properties of this 3D slowness time
representation can be found in Bennett (in press).
We parameterize our 3D STC representation in spherical The peak in Fig. 6a located at azimuth ĳ = 66° and
coordinates coh(Ĳ, sr, ș, ĳ) where Ĳ comes from the time inclination angle ș = 76° is evidence of a P arrival near the
pick; sr is either P or S slowness, depending on the ray-path 78° inclination angle predicted by the ray-tracing inversion

Bennett et al.
in Fig. 5a, whereas Fig. 6b shows little evidence of a P and Jupp, 1999) in both the azimuth and inclination-angle
arrival near the 18° inclination angle predicted by the SP directions to assess the consistency of the 3D STC peak
ray-tracing inversion in Fig. 5b. Since the tool diameter locations. Each ray-path type interpretation is ranked using a
is shorter than the wavelength of the reÀected wave¿eld, combination of the ray-tracing inversion data mis¿t, angular
we cannot hope to solve directly for the reÀector azimuth. dispersion of the 3D STC peaks, and the sum of the 3D
However, as can be observed in the Fig. 6a, imprecise STC peak amplitudes. The highest-scoring interpretation is
azimuth information is available from waveforms at a single deemed to be the automated ray-path type interpretation for
measured depth location. In fact, the appendix of Bennett (in the event, and the score itself is used to give a measure of
press) shows that in the particular case of event 24, which is relative prominence for the candidate arrival event.
a PP reÀection with incidence angle near 80°, the precision We can summarize the information derived from this
of the azimuth estimate is approximately ±(4 Bsrrtool)-1 ؆ second automated procedure in a variety of ways besides the
±ͶͲ° where sr = 85 ȝs/ft is the compressional slowness, waveform overlay shown in Fig. 4. Figure 8a shows a log
and B = 26 kHz is the bandwidth of the acoustic source. view like the one shown in Fig. 3 except now the hue of the
The precision of the inclination-angle estimate in this case markings indicating the arrival events is determined by the
is approximately ±(2ABsr)-1 ؆±͵° where A = 72 in. is the ray-path type. (Optionally, we can modulate the saturation
length of our receiver array. of these markings according the relative prominence of the
We can compute these 2D slices of the wave¿eld’s 3D event.) This can be helpful in assessing whether an event
STC representation as in Fig. 6 at all measured depth stations with a particular ray-path type appears consistently between
along the time pick to assess how consistently the predicted consecutive receiver sensors around the circumference of
arrival event appears. Figure 7 shows the 3D STC peak the tool sonde, as does the event near MD X070 ft.
locations for all the tool’s measured depth positions along The log shown in Fig. 8b shows arrival events color-
the time pick for the PP and SP ray-path interpretations. coded according to ray-path type as a function of measured
Note that the vertical axis indicates the inclination angles depth and the 3D STC azimuth. This is very helpful for
made relative to the inclination angle predicted by the understanding the local geometry of the reÀectors around
corresponding ray-tracing inversion result. Our ¿nal the wellbore. In this case, we observe that the reÀectors lie
estimate of 3D STC azimuth for each ray-path type comes in a single 2D plane with azimuth of approximately 75°.
from averaging the corresponding azimuth angle estimates The arrival events themselves are indexed according to their
derived from the wave¿eld at each measurement station score, and these indices correspond to the event indices
along the time pick. We use angular dispersion (Mardia shown in Fig. 4.
Fig. 7—Scatter plot of 3D STC peak locations estimated along event No. 24’s time pick for (a) PP, and (b) SP ray-path type interpretations. Relative
inclination angle is the difference between the inclination angle predicted by the ray-tracing inversion and the estimated 3D STC peak inclination
angle.

Fig. 8—Log views of arrival event measured depth versus (a) nominal receiver azimuth, and (b) versus 3D STC azimuth. The hue of the color
corresponds to the event’s highest scoring ray-path type interpretation with PP and SS reÀectors and PS and SP refractors indicated by cyan, orange,
magenta, and brown colors, respectively. The indices of the events in (b) correspond to the arrival events shown in Fig. 4.
One can derive similar log panels for other quantities, the mean 3D STC peak azimuth, as in Fig. 7a, and the well-
such as reÀector relative dip. Additionally, one can use the survey information, one can also derive a 3D model of the
well survey to convert reÀector dip and 3D STC azimuth reÀectors along the well trajectory without having to perform
into a tadpole display of true dip and azimuth, as shown in any migration, as shown in Fig. 10.
Track 3 of Fig. 9. In this case, the tadpoles were classi¿ed
as bed boundaries or fractures based largely on their true MIGRATION RESULTS AND QC
dip. We note that while PS and SP mode-converted arrivals
can often be associated with fractures, interpreting PP and Although constructing migration images from ¿ltered
SS reÀections as bed boundaries or fractures often requires sonic waveforms is a challenging task, the results of the
considering the well trajectory and interpretations from other automated processing described above provide the ray-path
measurements. Indeed, these tadpole logs derived from sonic type and recommended parameter settings for sonic-imaging
imaging can be very instructive when compared to tadpole migration workÀows as well as a QC that associates features
logs derived from borehole image logs and other sources. in the waveform measurements with features in the migration
In this example, a relative-dip change observed in the sonic image and vice versa. For our Middle East example, the
reÀectors was corroborated with a possible fault picked appropriate velocity model for our migration should be
from the borehole image in the main hole. Furthermore, the PP, because nearly all the reÀected-arrival events are PP
same fault plane was observed in the borehole image of the reÀections. The azimuth of the 2D imaging plane should be
sidetrack at the same depth level, adding con¿dence to the approximately 75° because most of the reÀections have 3D
fault interpretation. Borehole image quality around faults is STC azimuth values of 75 or 255°. A relatively conservative
sometimes affected by washouts or fault gouge. Picked dips choice for the migration aperture should be about 15 ft
are consequently erratic, and structural dip computation is because the reÀector’s relative dip is approximately 25 to
dif¿cult in those zones because structural analysis requires 35°. We, in fact, derive an estimate of the migration aperture
the use of consistent and in-sequence dips. Sonic reÀectors from ray-tracing diagrams, like the one shown in Fig. 5.
represent dips with a deeper depth of investigation compared The results of these migration parameter selections can be
to borehole images, which allowed for a computed structural observed in Fig. 11 where we have overlaid 2D reÀectors
dip of 27.5°. derived from the automated ray-tracing inversion and 3D
From the 2D mapping of a reÀector relative to the well STC workÀow whose indices correspond to arrival events
trajectory computed using the ray-tracing inversion, as in identi¿ed in the ¿ltered wave¿eld.
Fig. 5a, the estimate of the reÀector azimuth computed from

Bennett et al.
Fig. 9—(a) Migrated image showing change in relative dip of reÀectors along well trajectory at about X980 ft. (b) Log display of borehole image dips
(Track 2) and reÀector 3D STC azimuth (Track 3) as a function of measured depth. The possible fault plane was manually picked from the borehole
image.
Fig. 10—ReÀectors along well track computed from well survey, 2D

reÀector positions estimated from ray-tracing inversion, as in Fig. 5,
and reÀector azimuths estimated from the mean 3D STC peak-azimuth Fig. 11—2D migration result with overlay of reÀectors produced from
estimates, as in the Fig. 7a. ReÀector indices correspond to waveform- ray-tracing inversion and 3D STC azimuth estimates. Feature indices
arrival event indices shown in Fig. 4. correspond to arrival-event indices shown in Fig. 4.

FURTHER FIELD-DATA INTERPRETATION 24° mean dip and 236° mean strike angles from the borehole
image data in Fig. 13a, as compared to 29° mean dip angle
The stereonet plots shown in Figs. 12 and 13 provide and 251° strike coming from the 3D STC reÀectors in Fig.
a quantitative comparison of dip and azimuth information 13b.
derived from the resistivity image log data and from the 3D Individual 3D STC reÀectors are shown with the
STC reÀectors. migrated monopole image in Fig. 14. The interpreted
The plots highlight the close agreement in the mean formation top from the resistivity image data is shown in
strike for the fractures and structural bedding features. The green and coincides with the 3D STC reÀector at the bed
open-fracture stereonet plot in Fig. 12a shows that the mean boundary, where a signi¿cant acoustic impedance is present.
strike of the fractures from the borehole images is North Comparing the near-wellbore structures from the borehole
160° compared with North 155° for the corresponding 3D images with the far-¿eld reÀectors provided through new
STC reÀectors shown in Fig. 12b. 3D STC workÀow and the migration provides the necessary
The structural-bedding stereonet plots in Fig. 13 show a information to con¿rm the structural setting.
Fig. 12—Stereonet plots for the open fractures from (a) the image log data, and (b) the 3DSTC representing the near- and far-¿eld, respectively.
Fig. 13—Stereonet plots for the bedding from the (a) the image log data, and (b) the 3DSTC representing the near- and far-¿eld, respectively.

Bennett et al.
Fig. 14—Display of the 3D STC reÀectors (blue-green disks) and formation top derived from borehole image log (black disk) alongside the migrated
monopole image next to the wellbore.
SUMMARY used by downstream modeling workÀows, which allows

sonic-imaging results to be more directly integrated with the
Sonic imaging has long offered the potential to provide interpretations provided by other measurements and thus,
relatively high-resolution images of the near-wellbore region to determine more reliably the geologic structures along the
that cannot be provided by lower-frequency VSP or seismic well track.
measurements. However, the simplicity of the traditional
ACKNOWLEDGMENTS
sonic-imaging workÀow of ¿rst ¿ltering and then migrating
acoustic measurements strongly belies the underlying Many thanks to Erik Wielemaker, Nobuyasu
interpretation challenge and has meant that this potential Hirabayashi, Gabriela Martinez, Sudipan Shasmal, Peter
has largely remained unrealized. Indeed, identifying and Schlicht, Takeshi Endo, Martyn Beardsell, Smaine Zeroug,
characterizing candidate arrival events observable within Sandip Bose, Asbjoern Johansen, Tetsushi Yamada, and
¿ltered waveform measurements recorded along hundreds Israa Adil Suliman Salim.
or thousands of feet in measured depth according to their
ray-path type, reÀector azimuth, coherence, and relative NOMENCLATURE
dip presents a huge challenge for interpreters. Further,
the standard outcome of the traditional sonic-imaging A = 72 in., (the length of the receiver array)
B = 26kHz (bandwidth of the acoustic source)
workÀow—a static 2D image of the formation along the
coh(j) = coherence of waveform samples w(i, j) in
well trajectory—is not readily consumable by downstream
window of length M centered at index j
digital geological modeling workÀows.
Ec(jmax)coh(jmax ) = score for automated time pick event
We present a new automated sonic-imaging workÀow Ec(j) = coherent energy of waveform samples
intended to address these challenges by locating and w(i,j) in window of length M centered at
characterizing candidate arrival events within the ¿ltered index j
waveform measurements using a combination of tau-p, ray Ec (Ĳ,s) = coherent energy of wave¿eld with arrival
tracing, and 3D STC processing. The results of this workÀow time Ĳ and propagation direction s
include logs of reÀector dip, azimuth, and ray-path type; Ei (j) = total energy of waveform samples w(i,j) in
2D and 3D visualizations of the reÀectors along the well window of length M centered at index j
trajectory; speci¿c guidance for the sonic-imaging migration F= fraction of ࣦ(jmax) used in de¿ning arrival
parameters; and a helpful QC that associates waveform event measured depth location
features with features in the migration image. These 2D and i= time index of waveform samples w(i, j)
3D maps of the reÀectors and associated logs can be readily j= measured depth index of waveform samples
w(i, j)

jmax = point where ࣦ(j) attains its maximum value Borehole Acoustic Array Data, Geophysics, 49(3), 274–281.
jstart = last index less than jmax for which DOI: 10.1190/1.1441659.
ࣦ(j)<F maxࣦ Tang, X.M., 2004, Imaging Near-Borehole Structure Using
jstop = ¿rst index “greater than” j_max for which Directional Acoustic-Wave Measurement, Geophysics, 69(6),
ࣦ(j)<F maxࣦ 1378–1386. DOI: 10.1190/1.1836812.
rtool = radius of cylindrical receiver array Yamamoto, H., Haldorsen, J.B.U., Mikada, H., and Watanabe,
M = number of samples in coherency analysis S., 1999, Fracture Imaging From Sonic ReÀections and Mode
Conversion, SEG Technical Program Expanded Abstracts,
window
148–151. DOI: 10.1190/1.1820802.
s r= = wave¿eld slowness vector
s = (sx,sy,sz ) = 3D slowness vector in Cartesian
coordinates
T = number of time samples in time window corridor ABOUT THE AUTHORS
w(i, j) = waveform values sampled within time window
corridor along arrival event time pick Nicholas Bennett is currently
wj,Įk (t) = trace recorded by (j,k)th receiver sensor a Principal Research Scientist at
xj,k = (rtoolcos Įk, rtool sin Įk, zj í z1 ) = 3D positions of Schlumberger-Doll Research Center,
receiver array’s sensors Cambridge, Massachusetts, where he has
zj = measured depth offset of receiver array’s jth receiver been working since completing his PhD
ring in Mathematics from Yale University
Įk= azimuth of cylindrical receiver array’s kth nominal in 1997. Nick’s main activities involve
azimuth acoustics, nuclear magnetic resonance,
ș = inclination angle made by tool sonde axis and and electromagnetic measurements.
slowness vector s
Ĳ = arrival time of wave¿eld event
Adam Donald is the Technical
ĳ = azimuth angle slowness vector s makes with axis of
Director for Geomechanics & Acoustics
the tool sonde
at Schlumberger Wireline Headquarters,
ࣦ(j) = Ec(j)coh(j) denotes function used for searching for
based in Abu Dhabi, UAE. He has been
event along arrival event time pick
with Schlumberger for 20 years in various
roles with technology development and
REFERENCES interpretation of borehole acoustics data
for applications in formation evaluation,
Bennett, N., in press, 3D Slowness Time Coherence, Geophysics. geophysics and geomechanics.
Diebold, J.B., and Stoffa, P.L., 1981, The Traveltime Equation,
TauǦP Mapping, and Inversion of Common Midpoint Data, Sherif Ghadiry completed his BSc
Geophysics, 46(3), 238–254. DOI: 10.1190/1.1441196. degree in 2004 from Einshams University
Esmersoy, C., Chang, C., Kane, M., Coates, R., Tichelaar, B., and worked for almost two years as a
and Quint, E., 1998, Acoustic Imaging of Reservoir Structure wellsite geologist. He then moved to
from a Horizontal Well, The Leading Edge, 17(7), 940–946.
Schlumberger where, for over 12 years,
DOI: 10.1190/1.1438075.
Haldorsen, J., Voskamp, A., Thorsen, R., Vissapragada, B., Sherif has gained experiences with
Williams, S., and Fejerskov, M., 2006, Borehole Acoustic many different job assignments. Sherif
ReÀection Survey for High Resolution Imaging, SEG is currently Associate Domain Champion
Technical Program Expanded Abstracts, 314–318. DOI: of Petrophysics in Egypt.
10.1190/1.2370182.
Hornby, B.E., 1989, Imaging of Near-Borehole Structure Using Mohamed Nassar is a Senior
Full Waveform Sonic Data, Geophysics, 54(6), 747–757.
Geologist for Schlumberger Software
DOI: 10.1190/1.1442702.
Hirabayashi, N., 2016, ReÀector Imaging Using Trial ReÀector and Integrated Solutions in Egypt. He joined
Cross Correlation: Application to Fracture Imaging for Sonic Schlumberger as a Borehole Geologist
Data, Geophysics, 81(6), S433–S446. DOI: 10.1190/geo2016- in 2011 after receiving his BSc degree
0046.1 in petroleum geology at the University
Mardia, K., and Jupp, P., 1999, Directional Statistics, Wiley, 17– of Calgary, Alberta. His current role as
19. ISBN: 978-0471953333. the Geology Team Lead focuses on the
Kimball, C., and Marzetta, T., 1984, Semblance Processing of

Bennett et al.
integration of borehole images and petrophysical analyses in

clastic reservoir models.
Rajeev Kumar obtained his

Master’s Degree in applied geophysics
from Indian Institute of Technology,
Roorkee, in 2006. After completing his
degree, he joined the Data and Consulting
Services department of Schlumberger.
During his career, which has spanned
over 11 years, he has worked mainly on
borehole seismic, microseismic, and sonic imaging. He has
worked in India and Qatar before moving to Saudi Arabia,
where he is working as SIS Geophysics Team Lead.
Reetam Biswas received an

MSc from the Indian Institute of
Technology, Kharagpur, and is a graduate
student at the University of Texas at
Austin. His research interests include
transdimensional seismic inversion and
full-waveform inversion.


Petrophysics Vol. 60 No. 2 April 2019

Загружено:

Сведения о документе

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

Petrophysics Vol. 60 No. 2 April 2019

Загружено:

Авторское право:

Доступные форматы

PETROPHYSICS

Best Papers of the 2018 SCA Interna onal Symposium

Vol. 60, No. 2 April 2019

April 2019 PETROPHYSICS 176

President-Elect VP Publica ons

Magne c Resonance La n America Execu ve Director

Well and Reservoir Surveillance

Managing Editor Publica on Manager

177 PETROPHYSICS April 2019

April 2019 PETROPHYSICS 178

179 PETROPHYSICS April 2019

April 2019 PETROPHYSICS 180

Kent Newsham1, Joe Comisky2, and Roland Chemali1

INTRODUCTION formation volume factor (Bo), oil viscosity (ȝo), bubblepoint

181 PETROPHYSICS April 2019

Fig. 1—A density-neutron crossplot with mineral response lines for

April 2019 PETROPHYSICS 182

Table 1— Conversion to Bulk Volume% Including Total Porosity

183 PETROPHYSICS April 2019

Table 1 shows how variation in porosity alters the

April 2019 PETROPHYSICS 184

185 PETROPHYSICS April 2019

April 2019 PETROPHYSICS 186

Estimation of Pyrite THE STATICS: TOTAL POROSITY AND TOTAL

where Rw is the resistivity of the formation water, a is

187 PETROPHYSICS April 2019

April 2019 PETROPHYSICS 188

189 PETROPHYSICS April 2019

Fig. 9—The workÀow diagram of the Modi¿ed Electric Properties tests.

April 2019 PETROPHYSICS 190

THE DYNAMICS: THE IMPACT OF PORE

This section describes how Àuid phase impacts the Àuid

191 PETROPHYSICS April 2019

April 2019 PETROPHYSICS 192

Table 2—McCain (1990) Field Identi¿cation of Fluids

193 PETROPHYSICS April 2019

Equations of State (21)

April 2019 PETROPHYSICS 194

Table 3—Error Analysis of Public-Domain EOS Corrections (Courtesy Jiasen Tan)

195 PETROPHYSICS April 2019

April 2019 PETROPHYSICS 196

197 PETROPHYSICS April 2019

April 2019 PETROPHYSICS 198

199 PETROPHYSICS April 2019

April 2019 PETROPHYSICS 200

201 PETROPHYSICS April 2019

April 2019 PETROPHYSICS 202

SUMMARY We discussed the rationale for using a total porosity

203 PETROPHYSICS April 2019

April 2019 PETROPHYSICS 204

205 PETROPHYSICS April 2019

April 2019 PETROPHYSICS 206

Joe Comisky is the discipline lead

Roland Chemali is Consultant

207 PETROPHYSICS April 2019

David Kennedy1 and Fredy Garcia2

PROLOGUE into resistivity space, Archie’s power law model is revealed

April 2019 PETROPHYSICS 208

between logarithms and exponents. Archie’s Model

209 PETROPHYSICS April 2019

April 2019 PETROPHYSICS 210

Table 1—Effect of Pore Geometry on Electrical Rock Type

211 PETROPHYSICS April 2019

tube extending end-to-end through the core plug, in which