Documente Academic
Documente Profesional
Documente Cultură
The Society of
Petrophysicists and TUTORIAL
Well Log Analysts 351 Organic Mudstone Petrophysics, Part 3: Workflow to EsƟmate Storage Capacity
Kent Newsham, Joe Comisky, and Roland Chemali
8866 Gulf Freeway, Suite 320
Houston, TX 77017, USA
P: +1-713-947-8727
F: +1-713-947-7181 ARTICLES ͳ REGULAR SUBMISSIONS
Email: Sharon@spwla.org
Membership@spwla.org 373 ComposiƟon of the Shales in NiuƟtang FormaƟon at Huijunba Syncline and its
www.SPWLA.org Influence on Microscopic Pore Structure and Gas AdsorpƟon
Fu De-liang, Xu Guosheng, Tian Tao, Qin Jian-qiang, and Yang Fu
SPWLA Foundation. This fund
supports scholarships in higher educa on
and research in formation evaluation. 384 Accelerated Whole-Core Analysis OpƟmizaƟon With Wellsite Tomography
Please send dona ons to: SPWLA, 8866 InstrumentaƟon and Bayesian Inversion
Gulf Freeway, Suite 320, Houston, TX Alberto Mendoza, Lassi Roininen, Mark Girolami, Jere Heikkinen, and Heikki Haario
77017, USA
397 Finite-Volume ComputaƟons of Shale Tortuosity and Permeability From 3D Pore
ISSN 1529-9074 Networks Extracted From Scanning Electron Tomographic Images
PETROPHYSICS is published Mouin Almasoodi and Zulfiquar Reza
bimonthly by the Society of
Petrophysicists and Well Log Analysts 409 ApplicaƟon of a Fast NMR T1 RelaxaƟon Time Measurement to Sedimentary
(SPWLA). Subscription is included in Rock Cores
annual dues for members. Library Jonathan Mitchell and Andrea Valori
Subscription cost is $250 (6-issues,
Feb, April, June, Aug, Oct, Dec) Send
Subscription and member address 421 Reconsidering Klinkenberg’s Permeability Data
changes to the SPWLA business office. Douglas Ruth and Rasoul Arabjamaloei
Back issues if available are $35 each
to members and $45 each to library 429 DeterminaƟon of the Appropriate Value of m for EvaluaƟon of Carbonate Reservoirs
subscribers. With Vugs and Fractures at the Well-Log Scale
ADVERTISING RATES: (based on Dahai Wang and Jun Peng
one-time, full color) Full Page: $1500,
Half Page: $1200, Quarter Page: 438 Perched Water Contacts: Understanding Fundamental Controls
$1000. Additional fee for printing on Iulian N. Hulea
covers. Media kits are available by
contacting the SPWLA business office 450 Wellbore PosiƟoning While Drilling With LWD Measurements
or by visiting the SPWLA website
www.spwla.org. Responsibility: The Benny Poedjono, Dozie Nwosu, and Albert Mar n
statements and opinions expressed
in PETROPHYSICS are those of the
authors and should not be construed
as an official action or opinion of the
Society of Petrophysicists and Well
Log Analysts, Inc. © Copyright 2017
by the Society of Petrophysicists and
Well Log Analysts, Inc.
Full-Color versions of all technical arƟcles are available in the digital ediƟon of Petrophysics,
which is free to SPWLA members and can be found at Recent_Petrophysics_Journals.aspx.
About the Cover
SEM images of sample microscopic pores.
The organic pores are widely distributed,
As of Vol. 46 (1) 2005,
inorganic pores including intergranular The Society of Petrophysicists and Well Log Analysts
Petrophysics [ISSN: 1529-9074]
pores, crystal par cle pores, microcracks, is dedicated to the advancement of the science of forma on
evalua on through well logging and other forma on evalua on is indexed and abstracted in
and intragranular pore can also be seen in
techniques. SPWLA is dedicated to the applica on of these Thomson Reuters:
the images. See the ar cle by Fu et al. 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
President-Elect VP PublicaƟons
Jesus Salazar Carlos Torres-Verdin
Acous cs and Rock Physics ConocoPhillips University of Texas at AusƟn
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 EducaƟon
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 ConƟnental 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 PaƩerson 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
Nuclear Europe
Pingjun Guo, ExxonMobil Michael Webster
ProducƟon Petrophysics Ltd
Aberdeen, Scotland, UK
(+440) 7568-476931
Director-Europe@spwla.org
Petroleum Geochemistry
Drew Pomerantz, Schlumberger
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
TUTORIAL
Organic Mudstone Petrophysics, Part 3: WorkÀow to Estimate Storage Capacity
This is the third 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., 2019a). Part 2 described in
detail the many components of the workflow and how they come together to determine the storage capacity of the reservoir (Newsham et al., 2019b).
Part 3 links the petrophysical results to the production potential in terms of fractional flow and water cut and presents alternate cross-checks of the
storage properties to validate the results.
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
Storage validation using magnetic resonance and total-porosity measurement is readily visualized from Fig. 1.
dielectric logs does not replace the comprehensive All components of organic mudstone are shown juxtaposed,
core analysis integrated in the workÀow. It serves as an with solids on the left and Àuids on the right. The organic
independent cross-check of the Àuid output from the solid component, also called solid TOC, has density and
workÀow. It further helps provide consistent, portable hence neutron signatures similar to those of hydrocarbon; it is,
reliable estimation of hydrocarbon storage capacity, also however, part of the matrix. We have accounted for it in
known as “Petrophysics CPR.” The advanced technology Part 2 through correction equations modi¿ed from Passey
logging sensors provide an independent cross-check of total et al. (1990) and Schmoker and Hester (1983). The total
porosity and water saturation since neither are included in porosity from the workÀow is therefore a corrected porosity
the constrained simultaneous inversion. for TOC. Since magnetic resonance porosity does not see
this component, it should corroborate the results from the
USE OF MAGNETIC RESONANCE FOR TOTAL workÀow and validate the correction for TOC.
POROSITY In Track 5 (“Msol Phi”) in Fig. 2, we show a typical
overlay of the porosity values from the workÀow, validated
Direct measurement of total porosity by magnetic by total porosity from magnetic resonance. The good
resonance is a well-known, robust application for modern agreement suggests that our lithological model, including
logging technology (Coates et al., 1997). Since they were TOC is likely accurate.
¿rst introduced, magnetic resonance logs have been
publicized as the solution for determining free-Àuid volume USE OF DIELECTRIC LOG FOR WATER-FILLED
and permeability (Timur, 1991), Àuid viscosity (Morriss et POROSITY
al., 1994), gas saturation (Akkurt et al., 1996), Àuid typing
(Anand et al., 2017) and total porosity (Coates et al., 1997). The dielectric log was originally developed for the case
Of all these applications, total porosity is the most robust. It of freshwater mud ¿ltrate (Rmf) or fresh formation water
is the only one we consider in this tutorial. resistivity (Rw), where the classic resistivity-based Archie
Simply stated, the total magnetic resonance porosity equation could not yield accurate reliable results. The
signal is directly proportional to the number of hydrogen dielectric log would then rely on the large contrast between
atoms contained in the pore Àuids. In that sense, it is similar the dielectric permittivity of oil and the dielectric permittivity
to the neutron log, except for the fact that the neutron log of water to measure the water-¿lled porosity (Calvert and
is also affected by the hydrogen contained in the borehole, Wells, 1977; Poley et al., 1978). Table 1 illustrates the large
and in the clay matrix, and by other elements in the matrix difference in dielectric permittivity between water, oil and
and by various absorbers including chlorine and boron. various matrix minerals.
Therefore, neutron-porosity logs require borehole correction, Dielectric logs are often proclaimed as capable of
lithology correction, chlorine correction and other types of yielding an accurate water-¿lled porosity independent
corrections. One correction that the neutron-porosity log and of water salinity. This is only true for low-salinity water
the magnetic resonance porosity log have in common is that environments. The mudstones of the Permian Basin do not
for the hydrogen index (HI). (HI is equal to the number of fall in the category of “low-salinity environment” as water
hydrogen atoms per unit volume of pore Àuid, divided by the salinities are generally greater than 50,000 ppm. Statements
number of hydrogen atoms per unit volume of pure water at like “dielectric log yields a salinity independent water-
surface conditions). ¿lled porosity,” and “a water-¿lled porosity independent of
One of the main advantages of the magnetic resonance saturation exponent,” do not apply in this case. We use a
Fig. 1—Graphic representation of an organic mudstone. Total porosity Àuids include oil, gas and water contained within the organic and inorganic pore.
Fig. 2—Track 5 shows a cross-validaƟon of total porosity from the workŇow (red) with total porosity from magneƟc resonance(blue). Note that in the lithology track
(Track 3) kerogen or solid TOC is shown in purple.
Table 1—RelaƟve Dielectric Permiƫvity of Oil Water and Matrix Material for the CRIM and the modi¿ed CRIM (e.g., Wang et
al., 2018; Seleznev et al., 2006; Bittar et al., 2010). The
complex refractive index is uniquely related to permittivity
and to resistivity. Physicists chose to introduce a complex
permittivity whose real component is the traditional
dielectric permittivity and whose imaginary component is
related to resistivity. The governing equations are:
(1)
*A funcƟon of frequency, temperature and pressure.
A similar set of identities applies to the complex permittivity such as pyrite. When we apply the modi¿ed CRIM equation
of the formation as measured by the dielectric logging we simply lump the hydrocarbon with the rest of the matrix
sensor: * . The error resulting from this approximation is
into İmat
minimal.
(2) Observe that in the Permian Basin mudstones, water
conductivity is a signi¿cant component of İw*, therefore
where İ t’ is the permittivity of the formation at 1 GHz, water salinity must be taken into account when estimating
İ”= 18 and Rt is the resistivity of the formation at 1 GHz.
t Rt
water-¿lled porosity from dielectric logs. We use the
same water salinity as used within the workÀow of Part
Equation 3 is the modi¿ed CRIM equation which yields the 2. It is often based on samples from the produced water.
water ¿lled porosity, w: Another zonal parameter is the “mn” exponent, related to
the tortuosity of the water path within the rock. We base it
on local experience. It too has a bearing on the computed
(3) water-¿lled porosity. Some petrophysicists have proposed
advanced methods for automatically determining “mn”
using dispersion theory (Seleznev et al., 2006; Wang et al.,
where İmat
* is the complex dielectric constant of the matrix. 2018) however, their approaches are not part of this study.
We compute it based on the lithological data obtained from In Fig. 3, Track 7 (Bulk Volume Water), we show the
the workÀow of Part 2. It is normally a real number except in good agreement between the dielectric-log-derived water-
cases where the matrix contains some conductive minerals, ¿lled porosity (Track 7, blue curve) and the results from
Fig. 3—Cross-validation of water-¿lled porosity from the workÀow (Track 7, red curve) with water-¿lled porosity from dielectric, processed through the
modi¿ed CRIM equation (Track 7, blue curve). In Track 6, note again the good agreement between the total porosity from the workÀow (red) and the
total porosity from magnetic resonance (blue).
the workÀow described in Part 2 (Track 7, red curves). We of TOC by programmed pyrolysis is given by the generalized
also note the excellent agreement between the total porosity Eq. 5 (Law, 1999).
from the workÀow (red curve) and that from the magnetic
resonance logging sensor (blue curve) in Track 6 (Total (5)
Porosity).
where TOC is in wt%, S1 is the free hydrocarbons present
GEOCHEMICAL METHOD TO ESTIMATE OIL in the sample before the analysis, S2 is the volume of
STORAGE VOLUME hydrocarbons that formed during thermal pyrolysis of the
sample, and S4 is the residual carbon content of the rock
Modi¿ed Downey-Asquith Method sample.
Downey et al. (2011) proposed a quantitative measure
of OIP from measurements of the distillable oil from the S1 Programmed Pyrolysis
measurements in programmed pyrolysis (i.e., the Rock-Eval For the programmed pyrolysis test, the authors have
pyrolysis method). The literature is replete with documents adopted the ‘multiramp’ temperature program modi¿ed
on the use of programmed pyrolysis in oil¿eld chemistry from Romero and Sarmiento, (2015). Figure 4a shows
applications (Claypool and Reed, 1976; Espitalie et al., 1977, the temperature series beginning at 50°C ramping to 100,
1985a, 1985b, 1986; Clementz, 1979; Clementz et al., 1979; 200, 350 and ¿nally 650°C. Four FID response peaks
Jarvie and Baker, 1984; van Krevelen, 1984; Peters, 1986; results include the Sh0, Sh00, Sh1 and Sh2. The Sh0 peak
Behar et al., 2001; Maende, 2017). The method involves represents the lightest thermos-vaporized hydrocarbons
combusting crushed material and measuring the combusted (HC) released between T1 (50°C) and T2 (100°C). The Sh00
hydrocarbon components in a Àame ionization detector peak represents the thermo-vaporized HCs released between
(FID). The units of measure are milligrams of hydrocarbon T2 (100°C) and T3 (200°C). The Sh1 peak represents the
per gram of rock (mg/g). However, this unit of measure heavier thermo-vaporized HCs released between T3 (200°C)
isn’t a standard for use in the oil¿eld. The Downey Method and T4 (350°C). The Sh2 peak represents HCs issued from
provides a method for converting the units of measure to both the pyrolysis of sedimentary organic matter (OM)
barrels per acre-feet, a more commonly referenced unit of and the thermo-vaporization of the heaviest HCs detected
measure. The method allows for up-scaling cuttings or core- between T4 (350°C) and T5 (650°C). We sum the Sh0, Sh00
based tests to calculate OIP pro¿les for a given formation. and Sh1 to represent the producible hydrocarbon.
OIP estimates from programmed pyrolysis are comparable Asquith S1 Upscaling. To upscale the S1 from a point
to estimates from petrophysics, providing a lower limit representation to a continuous pro¿le, the authors adopted
of OIP due to potential volatile losses in extracting core/ the Asquith (2014, 2015) method. Asquith recognized that S1
cuttings samples from the subsurface and/or due to sample is a subset of TOC and generally is found to be correlative.
handling. The Downey Method for determining OIP is given Figure 4b is a graph showing the relationship between TOC
by Eq. 4. and S1 that is useful to upscale S1 from point to pro¿le. The
S1 pro¿le is then substituted into Eq. 4 to generate the OIP
pro¿le. Combining Eq. 1, the “Downey Method,” with the
(4)
upscaling of S1 via TOC, the “Asquith Method,” results in
the “Downey-Asquith Method” for estimating OIP from
where ȡAve is the average bulk density (g/cm3), S1Ave is the programmed pyrolysis.
average S1 (mg/g) measured from the programmed pyrolysis Sample Handling. Since the S1 is the producible
test and represents the free hydrocarbon in the sample before hydrocarbon in the sample, the authors developed a best
combustion, ȡoil is the density of oil (g/cm3) and 4965.36 practice for retaining as much of the producible hydrocarbon
is the units conversion factor explained in Downey et al. before evaporative losses. To minimize exposure at the
(2011). wellsite, cuttings are stored in isojars and whole core or rotary
In practice, the average density is obtained from the sidewall cores are sealed, stored in a climate-controlled
bulk-density log. The density of the oil is determined from container and transported to the laboratory immediately. In
equations of state referenced in Part 2. The limitation of the laboratory, detubing is performed in a controlled and
the Downey method is in the limited S1 samples available paced manner such that samples can be extracted, weighed,
within a given well. Asquith (2014, 2015) recognized that and sealed with a minimum of exposure before pyrolysis
upscaling S1 is achieved by correlation to the total organic testing. We have found that wax dipping is the best method
content (TOC), also measured by programmed pyrolysis or to seal Àuids in samples while in storage. Upon the start
via log-based correlations discussed in Part 2. The measure
(a) (b)
Fig. 4—(a) Programmed pyrolysis temperature series. (b) Asquith-type TOC vs. S1 correlation.
of testing, the samples are reweighed to account for any is especially true for high gas/oil (GOR) systems where gas
evaporated loss while in storage. will evolve from the liquid phase as samples are recovered
to surface conditions, resulting in lost gas and oil shrinkage.
Modi¿ed Downey-Asquith Method Figure 5 shows an example of the Downey-Asquith Method
The authors recognized that lost oil volumes due applied to a reservoir with a GOR of 2,200.
to extraction and handling, especially in highly volatile Track 9 in Figure 5 shows the incremental OIP
systems will subdue the S1 measured in the laboratory. This calculated from the petrophysics workÀow represented by
Fig. 5—Log pro¿le with the results of the Downey-Asquith method shown in Tracks 9 and 10 (far right). Track 9 shows the incremental OIP derived
from the petrophysics workÀow (dashed blue) and Downey-Asquith method (solid). Track 10 shows the cumulative OIP from the petrophysics
workÀow (dashed blue) and the Downey-Asquith method (solid).
the dashed blue curve and the incremental OIP estimate similar, especially in the mudstone systems. There are still
using the Downey-Asquith method represented by the solid signi¿cant losses (yellow shading) in the more permeable
curve. Track 10 (far right) is the cumulative OIP from the siltstone faces that are not fully corrected.
petrophysics workÀow and the Downey-Asquith method,
represented by the dashed and solid curves, respectively. ABSOLUTE PERMEABILITY
The bright green shading is the OIP volume retained by the
S1-based measurement. The yellow shading represents the Absolute permeability (kabs) as de¿ned by Darcy’s law
lost hydrocarbon. Generally, the S1-based OIP will be less (Hubbert, 1956; Amyx et al., 1960) relates the Àow of a
than the petrophysics-based OIP. It should be considered a single-phase incompressible Àuid to the pressure difference
lower constraint for the estimation of OIP. Conceptually, the (ǻP), viscosity (ȝ), length (L), and cross-sectional area (A)
authors have attempted to correct for the lost hydrocarbon of a porous medium:
through Eq. 6.
(7)
(6)
In addition to the assumptions of a single-phase
where Bo is the formation factor pro¿le for oil, and ȡoil is incompressible Àuid, application of Eq. 7 requires steady-
the density of oil (g/cm3) estimated from the equations of state, viscous, and laminar-Àow conditions where no
state (EOS) described in Part 2 (Newsham, 2019b), ȡb is the reactions take place that would change the Àow capacity
measured bulk density and S1 is the upscale pro¿le from of the rock through time or that consider the preferential
the Asquith Method. Both ȡb and ȡoil are considered to be wettability of one Àuid to another. Absolute permeability
at reservoir conditions, whereas S1 is measured at surface de¿nes the ability of a rock to transmit Àuids via the matrix
conditions. Conceptually, multiplying S1 by Bo corrects the and is a cornerstone in petrophysical engineering. While Eq.
S1 to reservoir conditions. Figure 6 is the same plot as Fig. 7 has a very narrow de¿nition and a handful of assumptions,
5, but with this correction. The two OIP results are very measurements of absolute permeability in the laboratory
Fig. 6—Log pro¿le with the results of the modi¿ed Downey-Asquith method, including the use of the formation volume factor to correct for volatile
losses, are shown in Tracks 9 and 10 (far right). Track 9 shows the incremental OIP derived from the petrophysics workÀow (dashed blue) and Downey-
Asquith method (solid). Track 10 shows the cumulative OIP from the petrophysics workÀow (dashed blue) and the Downey-Asquith method (solid).
often violate these rules; however, industry has instituted kgas and 1/Pm including turbulent Àow and/or double-slip
some guidelines for best practices. Klinkenberg effects (Rushing et al., 2004; Fathi et al., 2012).
Gas (e.g., helium, nitrogen, argon, CO2) is often used The unsteady-state method (USS) is also widely used
in the laboratory with a variety of equipment con¿gurations in industry now for estimating kgas and kinf using not only
to estimate absolute permeability in lieu of using a single- core plugs, but also crushed material. Jones (1972) was the
phase liquid. When an intact core plug void of fractures ¿rst to introduce this method, which uses a pressure falloff
or other discontinuities is available, two methods are technique (i.e., pressure vs. time) in conjunction with a
commonly employed: steady-and unsteady-state. The API diffusivity equation to estimate kinf, kgas, the slippage factor b,
Recommended Practices for Core Analysis RP 40 (API, and various coef¿cients related to the Forchheimer effect, for
1998) recommends using the multipoint steady-state method plugs. While many commercial laboratories and equipment
for estimating absolute permeability to gas. The viscous, manufacturers have used this technique for conventional
laminar-Àow assumption is one of the primary violations of and tight sands above 0.001 mD, it may not be suitable for
the Darcy de¿nition (also called non-Darcy effects) when organic mudstones or extremely tight sands/silts.
measuring absolute permeability to gas. When liquid Àows Another USS technique commonly used on core plugs
through a capillary or pore, there is a zero-velocity layer is the pulse method (Brace et al., 1968; Dicker and Smits,
at the solid/Àuid interphase. When a gas is used, the zero- 1988) where a pressure difference is initiated at both ends of
velocity assumption is not valid, and there is additional a plug and the resulting decay is recorded while gas diffuses
“slippage” of molecules at the solid/gas interphase (McPhee through the sample. A slight variation of this technique
et al, 2016). involves keeping the pressure constant at one end of the
The slippage phenomena of gas relative to the pore wall sample while the pressure buildup is monitored at the other
is known as the Klinkenberg effect (Klinkenberg, 1941), end (Heller et al., 2014). Despite the speci¿c experimental
and the “apparent” absolute permeability measured under setup, pressure-pulse techniques are inherently USS and a
these conditions will be greater than what is measured with variety of analytical models must be used to solve for kgas
a single, incompressible, nonreactive Àuid. The magnitude and kinf as a function of the gas properties (i.e., Z-factor and
of this slippage is a function of the mean free path of the gas viscosity) in addition to the pore volume of the sample. There
molecules at a given temperature, the mean Àowing pressure are also many instances when only a kgas value is reported
(Pm), and the radius of the pore throats through which the while ignoring kinf. While all of the above-mentioned
gas is Àowing. Many references show that an estimate of the USS methods for plugs are scienti¿cally sound and well-
absolute permeability to gas can be achieved by measuring practiced by both industry and academia, small details in
the apparent gas permeability at a variety (i.e., three or more) the experimental procedure and data analysis may present
of mean Àowing pressures (Rushing et al., 2004; McPhee systematic or even random differences in results (David et
at al., 2016). A plot of the reciprocal mean pressure (1/Pm) al., 2018). We have found that while USS techniques are
vs. the apparent gas permeability (kgas) reveals a straightline generally faster for rocks in the submicrodarcy to nanodarcy
relationship with the following form: range, it is well worth the time and effort to adhere to the API
RP 40 (1998) recommendation for multipoint steady-state
(8) measurements as a basis for kinf using gas as the nonreactive
Àowing agent.
where kinf is the Klinkenberg-corrected absolute permeability It is also worth noting that all of the various permeability
as Pmĺ, and the slope b is the slippage factor. The methods described above for plugs should be conducted
y-intercept on a plot of 1/Pm vs. kgas is equivalent to kinf, or one initially at subsurface reservoir net con¿ning stress (NCS)
can surmise that as Pmĺ then 1/Pm ĺ0. Some may refer to conditions. Subsequent measurements at higher NCS points
kinf as the equivalent liquid permeability; and for all practical should also be considered to simulate the reservoir stress
purposes is a valid estimate for our original de¿nition of depletion path during production. It is well documented
absolute permeability using Eq. 7 (Chhatre et al., 2015). that tight sand and mudstone permeability exhibit a strong
Techniques have also been developed to estimate kinf by con¿ning stress dependency, and that preferential closing of
measuring kgas at a single 1/Pm using various correlations for slot-like pores and cracks in the sample at higher con¿ning
the b term in Eq. 8 (Jones and Owens, 1980; Florence et al., stresses is the main driver (Jones and Owens, 1980; Byrnes,
2007). Additional non-Darcy factors may be present besides 1997; Sondergeld et al., 2010).
slippage that could render a nonlinear dependence between
The crushed rock permeability method pioneered by commercial and academic laboratories.
Luffel et al. (1993) and further expanded upon by Cui et al.
(2009), Tinni et al. (2012), and Civan et al. (2013) attempts PORE TO FLOW: MATCHING AND CALIBRATING
to mitigate several problems associated with measuring the TO PRODUCTION
apparent and absolute matrix permeability of mudstones
using plugs. One inherent problem is the tight nature and Our discussion about absolute permeability is important
nonconnectivity of the matrix pores which may prevent in that it serves as the basis for eventually de¿ning the
the samples from being fully gas-saturated during the preferential Àow capacity of one Àuid phase vs. another.
permeability measurement. Another issue is the presence of Almost all hydrocarbon-bearing intervals contain more
stress-relief cracks and bedding planes which may arti¿cially than one Àuid phase, so we must consider the effective
enhance the apparent matrix permeability. Crushing the permeability of each to help quantify Àow at the surface.
samples into smaller (0.5 to 3 mm) particles overcomes While gas, oil, and water are typically found as distinct
the pore access issue by increasing the surface area/volume phases within the pores of rocks in the subsurface, we will
ratio over which gas can diffuse into the matrix while also focus on oil/water systems with no free-gas component.
eliminating the effect of cracks and discontinuities that one Equation 7 focuses on a rock containing only a single phase,
may encounter with a full core plug. but when both oil and water are present, the effective Àow
The measurement principles and experimental setups for each phase is de¿ned as:
for the crushed-rock permeability method are fairly
straightforward, yet there are many small details that may (9)
lead to inconsistencies. The crushed particles are loaded and
sealed in a double-cell experimental setup where the pressure
decay is measured while gas (typically He) diffuses from the (10)
reference into the sample chamber and eventually into the
pores of the crushed material. A very attractive aspect of the
method is that measurement times vary from tens of minutes where Qo and Qw are the Àow rates at reservoir conditions,
to 1 hour. Consequently, the grain volume is also determined ko and kw are the effective permeabilities, and ȝo and ȝw are
during this measurement, which leads to the estimation of the viscosities of oil and water, respectively. It is important
the ambient porosity. Cui et al. (2009) and Tinni et al. (2012) to note that Eqs. 9 and 10 refer to the effective permeability
address some of the shortcomings of this method including to oil and water at a speci¿c value for water saturation (Sw),
the effect of dead volumes, the ratio of the sample chamber and each of those values may change rapidly as a function of
volume to the crushed particles, crushed particle size, and the varying Sw for a given ǻP, A, and L.
initial ¿ll-up pressures. Another challenging aspect of this An alternative way of incorporating the important
method is that only a small portion of the total pressure drop information expressed by Eqs. 9 and 10 is to consider the
(<<1%) is associated with diffusion into the matrix during relative permeability, or the ratio of the effective permeability
the equilibrium phase; and one must be careful to capture the for each phase to a reference permeability:
entire pressure vs. time curve over which the permeability is
calculated. Slippage effects also play an important role since (11)
the equilibrium pressures are relatively small (100 to 500
psi), and correcting this measurement back to a kinf equivalent (12)
relies on additional analytical models (Civan et al., 2013)
that many commercial laboratories do not consider. It is where kro is the relative permeability to oil, krw is the relative
not surprising that several published studies (Sondergeld permeability to water, and kref is the reference permeability.
et al., 2010; Passey et al., 2010) and numerous personal There are two common ways to de¿ne kref, the ¿rst being
communications have revealed major inconsistencies to set this equal to the kabs or kinf we de¿ned in the absolute
between different laboratories and experimental setups. permeability section. This is typically done within the
While the crushed-rock method may have limited value for context of reservoir Àow simulations when an estimate
determining the absolute permeability of liquid systems, the of the absolute permeability is known, and the individual
authors have found it useful as a relative index of reservoir contributions of each phase is then de¿ned by the shape of the
quality, particularly when comparing results from the same kro and krw curves as a function of Sw. An alternative kref used
(13) (21)
Table 2—Values Used to Calculate the Relative Permeability and fws Curves Shown in Figs. 9 and 10
Examination and comparison of the relative permeability lies in continuously updating our models and pro¿les for
curves and the resulting fws predictions in Figure 9 and Table relative permeability with the values that work best for the
2 reveals that rock type and wettability will have a strong reservoir and completion engineers trying to history-match
inÀuence on the fractional Àow. We observe that rocks with a the actual production at the surface.
strongly water-wet behavior (Fig. 9a) will have a higher Swirr Ultimately, taking these concepts from point to
and lower Sor compared to rocks with a higher af¿nity to oil pro¿le is needed to begin to understand the intervals that
(Fig. 9c). We can also just focus on the resulting fws curves are contributing oil versus water. Three methods are
for each case (Fig. 10) and observe how the predicted water readily available including estimation of static water/oil
cut varies as a function of the rock’s af¿nity for oil vs. water. ratios (WOR) from the petrophysical analysis using the
The predicted fws at Sw = 0.6 for the strongly water-wet case appropriate equations of state (EOS) to determine the oil and
is much less than the fws at Sw = 0.6 for the oil-wet case using water in place, as described in Part 2. Second, we use the
this model and the assumptions we listed in Table 2. fractional Àow at surface—what we refer to as the dynamic
This example was meant to illustrate the point that just solution, as described above—to generate pro¿les of relative
because one interval has a lower Sw compared to the other, permeability and fractional Àow at surface. Third, water-cut
there is not always a direct link to the water cut at the surface estimations from either the static or dynamic methods are
(i.e., zones with lower Sw will not always have a lower fsw). available by Eq. 23.
In reality what we tend to observe in oil-wet systems is a
much more consistent and lower downhole Sw at reservoir (23)
conditions that is closer to Swirr compared to intermediate
and mixed-wet systems which tend to have a much more Figure 11 is a pro¿le example. From left to right,
variable Sw pro¿le. This example also shows the need to the tracks are: GR correlation, resistivity, model-based
for more research and effort in de¿ning the endpoints composition, total porosity from the model (blue) compared
and relative shapes of these relative permeability pro¿les. to the independent MR total porosity (red), water saturation,
Ultimately, we have found that the current state-of-the-art WOR and water cut, relative permeability, fractional Àow at
Fig. 9—Example calculations of krw, kro, and fws for three cases (a) strongly water-wet, (b) intermediate, and (c) strongly oil-wet. The discrete points on
each curve show the calculated value at a speci¿c Sw which was chosen to be 10 s.u. units higher than Swirr.
Fig. 11—Log pro¿le showing the upscaling of relative permeability and fractional Àow at surface across the interval contributing to Àow as estimated
by the chemical extraction (Track 9). From left to right, the tracks are: Track 1, GR correlation; Track 2, resistivity; Track 3, model-based composition;
Track 4, total porosity from the model (blue) compared to the independent MR total porosity (red); Track 5, water saturation; Track 6, WOR and
water cut, Track 7, relative permeability, Track 8, fractional Àow at surface; and Track 9, the contributing interval Àow as estimated from the chemical-
extraction method.
(a) (b)
Fig. 12—(a) The histogram of the fractional Àow curve from Fig. 11 limited to the contributing interval to Àow as estimated by the chemical extraction
method. (b) Water cut from production stabilizing after 55 days and used as the match-reference for the fractional Àow estimation.
Multiple approaches for allocating production are system pressure. For example, organic mudstones, such as
currently in practice including methods involving DNA the Spraberry, Wolfcamp and Avalon (Leonard), the RE
sequencing of anaerobic bacteria, chemical ¿ngerprinting ranges from 2 to 5%, when using a per-section, per-well
of compounds and elemental isotopes and the use of an basis. Nonorganic siltstones, such as the Hoban, Bone Spring
equivalent EUR method. The two former methods generate a and Dean, range from 4 to 9% using the same per-section,
static vertical pro¿le from cuttings or core extractions across per-well reference. High GOR systems tend to have greater
zones of interest. In the case of the DNA method, groups RE than low GOR due to gas expansion drive and lower
of anaerobic bacteria ‘families’ are pro¿led. In the chemical oil viscosity yielding greater Àuid mobility. Reservoirs that
extraction method, pro¿les of combinations and ratios of have pore pressure far exceeding the bubblepoint also have
chemical compounds and their isotopes are pro¿led. In greater RE.
both methods, production streams are collected over time Use of the equivalent EUR requires knowledge of the
from which similar extractions are made and represent the type curve EUR or a near offset well EUR, generally limited
composite of the contributing pro¿le. The production stream to 1-mile long lateral wells or 1-mile equivalents, since the
‘extractions’ are then allocated, based on the static pro¿le, OIP and well density are on a per-section basis. The method
on a percentage contribution basis. The product is a pro¿le integrates the petrophysics-derived OIP from the bottom of
indicating the percentage contribution of each interval to the a contributing interval, upward stratigraphically, calculating
production stream. From this a drainage height is implied. the equivalent EUR at each step until the equivalent EUR
These methods are time consuming and expensive. matches the offset or type curve EUR. At the point of
The equivalent EUR method is an alternative approach matching EURs, the drainage height is implied. The method
that uses the petrophysics OIP in combination with is essentially a quicklook material-balance test.
knowledge of well spacing by plan or execution and an Figure 13 is an example showing the comparison
assumed range of recovery ef¿ciencies for a given facies between the OIP allocation using the equivalent EUR and
type by Eq. 24. the chemical-extraction method. From left to right, the
tracks are: Track 1, GR correlation; Track 2, resistivity;
(24) Track 3, model composition; Track 4, total porosity from
the petrophysics model (blue) and the independent MR total
porosity (red); Track 5, water saturation; Track 6, magnetic
where EUReq is the equivalent EUR, OIP is the oil-in- resonance T2 distribution; Track 7, chemical allocation;
place per section obtained from petrophysics and RE is the Track 8, OIP allocation and the cumulative OIP per bench
recovery ef¿ciency. The assumption here is the OIP is well from petrophysics. The chemical-extraction track de¿nes
constrained by use of the constrained simultaneous inversion the contributing interval to Àow. The EUR match is based
described in Part 2 (Newsham, 2019b). The well spacing on a per section basis using four wells per section and RE
is a variable but constrained by plan, drilling execution or of 6%, well within the range for tight siltstones. There is
general perception. The RE ranges by facies, Àuid type and good agreement between the chemical and OIP allocation
methods in this case.
Fig. 13—Log pro¿le comparing the OIP allocation pro¿le with the chemical extraction pro¿le which de¿ned the contributing interval to Àow. From left to
right, the tracks are: Track 1, GR correlation; Track 2, resistivity; Track 3, model composition; Track 4, total porosity from the petrophysics model (blue)
and the independent MR total porosity (red); Track 5, water saturation; Track 6, magnetic resonance T2 distribution; Track 7, chemical allocation; Track
8, OIP allocation and the cumulative OIP per bench from petrophysics. Track 9 HPV, bench ranking.
BENCH RANKING the GOR to increase, oil to decrease and increases the oil
shrinkage. Finally, incorporating the OIP Allocation method
From a ¿nancial perspective, it’s important that can yield drainage height if the well spacing and RE are
companies invest and drill their best opportunities ¿rst, well-constrained. If the RE and well spacing are unknown
whether at a basin, ¿eld or well scale. As a determinant or uncertain, OIP Allocation can provide sensitivity analysis
for which bench to drill ¿rst, there are numerous attributes to the merits of ¿eld development plans and their associated
based on the workÀow that should be considered for bench well spacing assumptions or RE. In the end, the ordering
ranking. Storage de¿nes the opportunity, so it makes sense or weighting of these attributes is company dependent.
to include an assessment of porosity, water saturation and The critical consideration is to actually incorporate these
OIP in any ranking criteria. Additionally, the water-in-place attributes in the bench-ranking exercise. Figure 14 displays
(WIP), water-oil-ratio (WOR), or conversely water cut, need many of these attributes. From left to right, the tracks show
consideration for facilities design and implantation. High the correlation curves, resistivity curves, composition,
water cuts diminish a well’s productivity and economic Poisson’s ratio in a butterÀy display, pore volumes, gas
value. Facies type will affect the range of RE. As stated shows, pressure pro¿les with the grey shading representing
in the Allocation section, organic mudstones generally an undersaturated condition where pore pressure is greater
will have lower RE ranges, typically ranging from 2 to than the bubblepoint pressure, the cumulative OIP, WIP and
5%. Low-permeability siltstones may range from 4 to GIP per bench, the WOR, the TOC content, and hydraulic
9%. Fluid pressure-volume-temperature (PVT) properties fracture models for various bench opportunities.
are important due to the impacts on mobility and drive
mechanisms. High GOR systems will generally have SUMMARY
lower Àuid viscosity yielding greater mobility. Also, these
systems bene¿t from gas-expansion drive. Pore pressure is In Part 3, we have shown examples of cross-discipline
important as a drive mechanism and understanding whether applications of the workÀow including logging, geochemical,
the reservoir is above the bubblepoint (i.e., undersaturated) reservoir engineering and production methods. The emphasis
or below the bubblepoint (saturated). As the pressure drops is on calibration to known references beyond laboratory core
below the bubblepoint, gas disassociates from the oil causing measurements to support the authors’ claim of Petrophysics
CPR: consistent, portable and reliable analysis results. mudstone and tight oil systems. It is comprehensive and
Examples of the use of total porosity from the magnetic yields consistent results satisfying Petrophysics CPR and
resonance log and the bulk-volume water from the dielectric- SEC “Reliable Technology” mandates using the “Three-Point
permittivity logs are used as an independent calibration Stance” approach.
reference to the petrophysics results. They are independent
since neither are included in the constrained simultaneous ACKNOWLEDGMENTS
inversion method as inputs. We show a modi¿ed geochemical
method developed by Downey and Asquith to estimate OIP The authors wish to acknowledge Occidental Petroleum
from cuttings or core using the S1 measured from programmed Corporation and Devon Energy for supporting the publication
pyrolysis. These results are compared to the petrophysics OIP of these tutorials. We’d also like to thank David Bell,
as a cross-check and provide a lower constraint of OIP. Some Occidental’s Manager of Rocks and Fluids team, for his
engineering applications include the use of absolute, effective contributions to the ¿gures used in the tutorials.
and relative permeability in the estimation of fractional Àow.
We also present static and dynamic comparisons of water/ NOMENCLATURE
oil ratio or water-cut to produced water after cleanup of the
hydraulic fracture Àuid. Finally, considerations for how to Abbreviations
rank bench opportunities from the various products resulting BPP = bubblepoint pressure
from the workÀow is given. CRIM = complex refractive index method
This concludes the author’s tutorial on, “Organic DFIT = diagnostic fracture injection test
Mudstone Petrophysics: WorkÀow to Estimate Storage EOS = equation of state
Capacity.” The workÀow should be considered a framework EUR = expected ultimate recovery
and is not implied as the only approach to evaluating organic EUReq = equivalent EUR
Fig. 14—Log pro¿le composite showing various (but not all) attributes used for developing a rank ordering of bench opportunities. From left to right,
Track 1, correlation curves; Track 2, resistivity curves; Track 3, composition; Track 4, Poisson’s ratio in a butterÀy display; Track 5, pore volumes;
Track 6, gas shows; Track 7, pressure pro¿les with the grey shading representing an undersaturated condition where pore pressure is greater than
the bubblepoint pressure; Track 8, the cumulative OIP, WIP and GIP per bench; Track 9, WOR; Track 10, TOC content; Tracks 11 to 13, hydraulic
fracture models for various bench opportunities.
Amyx, J.W., Bass Jr., D.M., and Whiting, R.L., 1960, Petroleum Claypool, G.E., and Reed, P.R., 1976, Thermal-Analysis Technique
Reservoir Engineering Physical Properties, McGraw-Hill. for Source-Rock Evaluation: Quantitative Estimate of Organic
ISBN: 0070016003 Richness and Effects of Lithologic Variation, AAPG Bulletin,
Anand, V., Ali, R.M., Abubakar, A.A., Grover, R., Neto, O., 60(4), 608–612. DOI: 10.1306/83D92410-16C7-11D7-
Pirie, I., and Iglesias, J.G., 2017, Unlocking the Potential of 8645000102C1865D.
Unconventional Reservoirs Through New Generation NMR Clementz, D.M., 1979, Effect of Oil and Bitumen Saturation on
T1/T2 Logging Measurements Integrated with Advanced Source-Rock Pyrolysis, AAPG Bulletin, 63(12), 2227–2232.
Wireline Logs, Petrophysics, 58(2), 81–96. DOI: 10.1306/2F918919-16CE-11D7-8645000102C1865D.
Asquith, G.B., 2014, OOIP Utilizing GEOCHEM [ECS] Data, Clementz, D.M., G.J. Demaison, and Daly, A.R., 1979, Well Site
Triple Combo Data Only, and Pyrolysis S1 Data, Permian Geochemistry by Programmed Pyrolysis, Paper OTC-3410
Wolfcamp “A” and “B” Shales, AAPG Search and Discovery presented at the Offshore Technology Conference, Houston,
Article 41406. http://www.searchanddiscovery.com/ Texas, 30 April–3 May. DOI: 10.4043/3410-MS.
documents/2014/41406asquith/ndx_asquith.pdf. Accessed Coates, G.R., Menger, S., Prammer, M., and Miller, D., 1997,
May 23, 2019. Applying NMR Total and Effective Porosity to Formation
Asquith, G.B., 2015, OOIP Utilizing GEOCHEM [ECS] Data, Evaluation, SPE-38736 presented at the SPE Annual Technical
Triple Combo Data Only, and Pyrolysis S1 Data, Permian Conference and Exhibition, San Antonio, Texas, 5-8 October.
Wolfcamp “A” and “B” Shales, Midland Basin, Texas, DOI:10.2118/38736-MS.
AAPG Search and Discovery Article 110207. http://www. Cui, X., Bustin, M.M., and Bustin, R.M., 2009, Measurements of
searchanddiscovery.com/documents/2015/110207asquith/ Gas Permeability and Diffusivity of Tight Reservoir Rocks:
ndx_asquith.pdf. Accessed May 23, 2019. Different Approaches and Their Applications. GeoÀuids, 9(3),
Behar, F., Beaumont, V., and De B. Penteado, H.L., 2001, Rock- 208–223. DOI: 10.1111/j.1468-8123.2009.00244.x.
Eval 6 Technology: Performances and Developments, Oil Dacy, J.M., 2010, Core Tests for Relative Permeability of
& Gas Science and Technology – Rev. IFP, 56(2), 111–134. Unconventional Gas Reservoirs, Paper SPE-135427 presented
DOI: 10.2516/ogst:2001013. at the SPE Annual Technical Conference and Exhibition,
Birchak, R.J., Gardner, G.C., Hipp, J.E., and Victor, J.M., 1974, Florence, Italy, 19–22 September. DOI: 10.2118/135427-MS.
High Dielectric Constant Microwave Probes for Sensing David, C., Wasserman, J., Amann, F., Lockner, D.A., Rutter, E.H.,
Soil Moisture, Proceedings of the IEEE, 62(1), 93–98. DOI: Vanorio, T., Amann, A., Billiotte, J., Reuschle, T., Lasseux,
10.1109/PROC.1974.9388. D., Fortin, J. Lenormand, R., Selvadural, P., Meredith, P.,
Bittar, M., Li, J., Kainer, G., Cherry, R., Torres, D., and McCoy, Browning, J., Mitchell, T.M., Loggia, D., Nono. F., Sarout,
D., 2010, A Modern Microwave Formation Evaluation J., Esteban, L., Davy, C.A., Louis, L., Bltnott, G., Madonna,
Sensor and its Applications in Reservoir Evaluation, Paper B, C., Jahns, E., Fleury, M., Berthe, G., Delage, P., Braun, P.,
Transactions, SPWLA 51st Annual Logging Symposium, Perth, Gregoire, D., Perrier, L., Polito, P., Jannot, Y., Sommier, A.,
Australia, 19–23 June. Krooss, B., Fink, R., Hu, Q., Klaver, J., and Clark, A.C.,
Brace, W.F., Walsh, J.B., and Frangos, W.T., 1968, Permeability of 2018, KG2B, a Collaborative Benchmarking Exercise for
Granite Under High Pressure, Journal of Geophysical Research, Estimating the Permeability of the Grimsel Grandiorite—
73(6), 2225–2236. DOI: 10.1029/JB073i006p02225. Part 1: Measurements, Pressure Dependence, and Pore-Fluid
Buckley, S.E., and Leverett, M.C., 1942, Mechanism of Fluid Effects, Geophysical Journal International, 215(2), 799–824.
Displacements in Sands, Paper SPE-942107-G, Transactions, DOI: 10.1093/gji/ggy304.
AIME, 146, 107–116. DOI: 10.2118/942107-G. Dicker, A.I., and Smits, R.M., 1988, A Practical Approach for
Byrnes, A.P., 1997, Reservoir Characteristics of Low-Permeability Determining Permeability From Laboratory Pressure-Pulse
Sandstones in the Rocky Mountains, The Mountain Geologist, Decay Measurements, Paper SPE-17578 presented at the
34(1), 39–51. International Meeting on Petroleum Engineering, Tianjin,
Calvert, T.J., and Wells, L.E., 1977, Electromagnetic China. 1–4 November. DOI: 10.2118/17578-MS.
Propagation…A New Dimension In Logging, Paper SPE- Downey, M.W., Garvin, J., Lagomarsino, R.C., and Nicklin,
6542, presented at the SPE California Regional Meeting, D.F., 2011, Quick Look Determination of Oil-in-Place in
Bakers¿eld, California, 13–15 April. DOI: 10.2118/6542-MS. Oil Shale Resource Plays, AAPG Search and Discovery
Chhatre, S.S., Braun, E.M., Sinha, S., Determan, M.D., Passey, Article 40764. http://www.searchanddiscovery.com/
Q.R., Zirkle, T.E., Wood, A.C., Boros, J.A., Berry, D.W., documents/2011/40764downey/ndx_downey.pdf. Accessed
Leonardi, S.A., and Kudva, R.A., 2015, Steady-State Stress- May 23, 2019.
Dependent Permeability Measurements of Tight Oil-Bearing Espitalie, J., Deroo, G., and Marquis, F., 1985a, Rock-Eval
Rocks, Petrophysics, 56(2), 116–124. Pyrolysis and its Applications (Part 1), Oil & Gas Science
Civan, F., Devegowda, D., and Sigal, R., 2013, Critical Evaluation and Technology – Revue IFP, 40(5), 563–579. DOI: 10.2516/
and Improvement of Methods for Determination of Matrix ogst:1985035.
Permeability of Shale, Paper SPE-166473 presented at the Espitalie, J., Deroo, G., and Marquis, F., 1985b, Rock-Eval
SPE Annual Technical Conference and Exhibition, New Pyrolysis and its Applications (Part 2), Oil & Gas Science
Orleans, Louisiana, USA, 30 September–2 October. DOI: and Technology – Revue IFP, 40(6), 755–784. DOI: 10.2516/
10.2118/166473-MS. ogst:1985045.
Espitalie, J., Deroo, G., and Marquis, F., 1986, Rock-Eval Lee, W.J., 2011, The “Reliable Technologies” Rule: What Did
Pyrolysis and its Applications (Part 3), Oil & Gas Science the SEC Intend? Paper SPE-139494, SPE Economics &
and Technology – Revue IFP, 41(1), 73–89. DOI: 10.2516/ Management, 3(1), 18–21. DOI: 10.2118/139494-PA.
ogst:1986003. Luffel, D., Hopkins, C., and Schettler, P.D., Jr., 1993, Matrix
Espitalie, J., Madec, M., Tissot, B., Mening, J.J., and Leplat, P., Permeability Measurement of Gas Productive Shales, Paper
1977, Source Rock Characterization Method for Petroleum SPE-26633 presented at the SPE Annual Technical Conference
Exploration, Paper OTC-2935 presented at the Offshore and Exhibition, Houston, Texas, USA, 3–6 October. DOI:
Technology Conference, Houston, Texas, 2–5 May. DOI: 10.2118/26633-MS.
10.4043/2935-MS. Maende, A., 2017, Wildcat Compositional Analysis For
Fathi, E., Tinni, A., and Akkutlu, I.Y., 2012, Shale Gas Conventional and Unconventional Reservoir Assessment,
Correction to Klinkenberg Slip Theory. Paper SPE-154977 HAWK Petroleum Assessment Method, Application
presented at the SPE Americas Unconventional Resources Note 052016-1, Wildcat Technologies. https://www.
Conference, Pittsburgh, Pennsylvania, USA, 5–7 June. DOI: wildcattechnologies.com/download_¿le/view_inline/224.
10.2118/154977-MS. Accessed May 22, 2019.
Florence, F.A., Rushing, J., Newsham, K.E., and Blasingame, McPhee, C., Reed, J., and Zubizarreta, 2016, Core Analysis: A
T.A., 2007, Improved Permeability Prediction Relations for Best Practice Guide, Developments in Petroleum Science, 64,
Low Permeability Sands, Paper SPE-107954 presented at the Elsevier. ISBN: 9780444635334.
Rocky Mountain Oil and Gas Technolog Symposium, Denver, Morriss, C.E., Freedman, R., Straley, C., Johnston, M., Vinegar,
Colorado, USA, 16–18 April. DOI: 10.2118/107954-MS. H.J., and Tutunjian, P.N., 1994, Hydrocarbon Saturation and
Glorioso, J.C., and Rattia, A.J., 2011, New SEC Rules: What is Viscosity Estimation From NMR Logging In the Belridge
New for the Petrophysicist?, Paper SPE-141045 presented Diatomite, Paper C, Transactions, SPWLA 35th Annual
at the SPE Middle East Oil and Gas Show and Conference, Logging Symposium, Tulsa, Oklahoma, USA, 19–22 June.
Manama, Bahrain, 25–28 September. DOI: 10.2118/141045- Newsham, K., Comisky, J., and Chemali, R., 2019a, Organic-
MS. Mudstone Petrophysics: WorkÀow to Estimate Storage
Heller, R., Vermylen, J., and Zoback, M., 2014, Experimental Capacity, Petrophysics, 60(1), 4–15. DOI: 10.30632/
Investigation of Matrix Permeability of Gas Shales, AAPG PJV60N1-2019t1.
Bulletin, 98(5), 975–995. DOI: 10.1306/09231313023. Newsham, K., Comisky, J., and Chemali, R., 2019b, Organic-
Hubbert, M.K., 1956, Darcy’s Law and the Field Equations of the Mudstone Petrophysics, Part 2: WorkÀow to Estimate Storage
Flow of Underground Fluids, Paper SPE-749-G, Transactions, Capacity, Petrophysics, 60(2), 181–207. DOI: 10.30632/
AIME, 207, 222–239. PJV60N2-2019t1
Jarvie, D.M., and Baker, D.R., 1984, Application of the Rock-Eval Ojha, S.P., Misra, S., Sinha, A., Dang, S., Tinni, A., Sondergeld,
III Oil Show Analyzer to the Study of Gaseous Hydrocarbons C., and Rai, C., 2017, Relative Permeability and Residual
in an Oklahoma Gas Well, presented at the 187th ACS Saturation Estimates for Organic-Rich Shale Samples From
National Meeting, St. Louis. Bakken, Wolfcamp, Eagle Ford and Woodford Formations.
Jones, F.O., and Owens, W.W., 1980, A Laboratory Study of Low- Paper SPE-187398 presented at the SPE Annual Technical
Permeability Gas Sands, Journal of Petroleum Technology, Conference and Exhibition, San Antonio, Texas, USA, 9–11
32(9), 1631–1640. DOI: 10.2118/7551-PA. October. DOI: 10.2118/187398-MS
Jones, S.C., 1972, A Rapid Accurate Unsteady-State Klinkenberg Passey, Q.R., Bohacs, K., Esch, W.L., Kimentidis, R., and Sinha, S.,
Permeameter, Society of Petroleum Engineers Journal, 12(5), 2010, From Oil-Prone Source Rock to Gas-Producing Shale
383–397. DOI: 10.2118/3535-PA. Reservoir—Geologic and Petrophysical Characterization of
Klinkenberg, L.J., 1941, The Permeability of Porous Media to Unconventional Shale Gas Reservoirs, Paper SPE-131350
Liquids and Gases, Paper API-41-200 presented at API presented at the International Oil and Gas Conference and
Drilling and Production Practice, New York, New York, 1 Exhibition, Beijing, China, 8–10 June. DOI: 10.2118/131350-
January. MS.
Krumm, R.L., and Howard, J.J., 2017, WorkÀow for Measurement Passey, Q.R., Creaney, S., Kulla, J.B., Moretti, F. J., and Stroud, J.
of Steady-State-to-Liquid Permeability in Nanodarcy D., 1990, A Practical Model for Organic Richness from Porosity
Reservoir Rock, Paper SCA2017-038 presented at the and Resistivity Logs, AAPG Bulletin, 74(12), 1777–1794.
International Symposium of the Society of Core Analysts, DOI:10.1306/0C9B25C9-1710-11D7-8645000102C1865D.
Vienna, Austria, 17 August–1 September. Peters, K.E., 1986, Guidelines for Evaluating Petroleum Source
Land, C.S., 1968, Calculation of Imbibition Relative Permeability Rock Using Programmed Pyrolysis, AAPG Bulletin,
for Two- and Three-Phase Flow From Rock Properties, 70(3), 318–329. DOI: 10.1306/94885688-1704-11D7-
Society of Petroleum Engineers Journal, 8(2), 149–156. DOI: 8645000102C1865D.
10.2118/1942-PA. Poley, J.P, Nooteboom, J.J., and De Waal, P.J., 1978, Use of VHF
Law, C.A., 1999, Evaluating Source Rocks, Chapter 6, in Dielectric Measurements for Borehole Formation Analysis,
Beaumont, E.A., and Foster, N.H., Editors, Treatise of The Log Analyst, 19(3), 8–30.
Petroleum Geology/Handbook of Petroleum Geology, AAPG. Romero-Sarmiento, M.F., Pillot, D., Letort, G., Lamoureux-Var, V.,
ISBN: 089181602-X. Beaumont, V., Huc, A.Y., Garcia, B., 2015, New Rock-Eval
ABSTRACT
The inÀuence of the shale composition, especially form more mesopores. The isosteric heat of methane
organic matter and clay minerals, on microscopic pore adsorption in the study area is 5.3 to 22.2kJ/mol and the
structure and gas adsorption at Huijunba syncline were standard entropy of adsorption is í98.8 to í48.8J/mol·K.
investigated by the TOC, XRD, low-pressure gas- The methane adsorption is mainly affected by the organic
adsorption measurement, SEM and isothermal adsorption matter, and the organic matter in the shale with low clay-
analysis. The results show that organic matter is the mineral content could have much more effective methane
main material basis for the micropores and clay minerals adsorption position.
INTRODUCTION (Fig. 1). The total organic carbon (TOC), X-ray diffraction
(XRD), low pressure N2-adsorption measurement, SEM
With the rapid development of hydraulic fracturing data and isothermal CH4-adsorption analysis were used to
technology and horizontal drilling technology, shale gas study the inÀuence of the organic matter and clay minerals
has gradually become an important source of clean energy. on the microscopic pore structure and CH4 adsorption, to
Unlike conventional reservoirs, nanoscale pores are widely provide some reference for future shale-gas exploration and
developed in shale, which leads to a large part of the gas development work in this region.
in the shale being adsorbed on the pore surfaces (Curtis,
2002). The organic matter and clay minerals are two key SAMPLES AND METHODS
compositions that affects the pore structure and the adsorption
characteristics (Wang et al., 2016; Xiong et al., 2017). All 22 samples used in this paper were collected from
Therefore, understanding the relationship between shale the outcrop section of the Lower Cambrian Niutitang
composition characteristics and microscopic pore structure Formation at the Huijunba syncline (Fig. 1). The lithology
is the key to study the gas-adsorption characteristics. The of the sample is mainly black carbonaceous shale, black
isothermal adsorption experiment, which is the most effective shale, black silty shale, etc.
method for evaluation of shale adsorption, is widely used The TOC analyses were measured using a CS-344
and can be applied to geological conditions by mathematical analyzer following the standard GB/T 19145-2003 of China.
models, such as Langmuir model, SDR model, SLD model The XRD analysis of full rock was performed using a Rigaku
et, al. (Tian et al., 2016; Li et al., 2018). Low-temperature Ultima IV diffractometer with Cu-KĮ radiation at 40kV and
N2 adsorption experiments, backscattering high-resolution 40mA. The scan was taken from 2 to 60° with a rate of 4°/
scanning electron microscopy (SEM) combined with argon- min and a step interval of 0.02°. For the clay-mineral XRD
ion polishing and other technical means have provided analysis, the samples powders were treated by HCl (2 to
effective support for study of the nanopore structure 3%) to remove the carbonate minerals, and H2O2 (30%) was
characteristics (Chalmers et al., 2012). used to remove the organic matter. Then the samples were
The Huijunba syncline, which has a good potential washed repeatedly with distilled water, and ammonia water
for shale-gas accumulation, is located at the center of the added and let stand to extract the suspension to centrifuge
Micangshan-Hannan paleo-uplift with structural coupling the clay mineral. The treated samples were dried at 60°C
between the Sichuan Basin and the Qinling Orogenic Belt and ground to a powder for XRD analysis. The analysis was
Manuscript received by the Editor December 16, 2018; revised manuscript received May 4, 2019; manuscript accepted May 6, 2019.
1
Key Laboratory of Coal Exploration and Comprehensive Utilization, Ministry of Land and Resources, Xi’an 710021, PR China; Fu Deliang. No.
26 Wenjing Road Xi’an, 710021, Shaanxi, PR China. E-mail address: fudl3513@foxmail.com; Tian Tao. No. 26 Wenjing Road Xi’an, 710021,
Shaanxi, PR China. E-mail address: 54497213@qq.com
2
College of Energy, Chengdu University of Technology, Chengdu 610059, PR China;
performed using a Bruker AXS D8 diffractometer with Cu- The mineralogical composition adjacent to pore structure
KĮ radiation at 40kV and 40mA. The scan was taken from 3 was identi¿ed using elemental analysis (EDS).
to 45° with a rate of 2°/min and a step interval of 0.02°. The isothermal methane-adsorption experiments were
The low-pressure nitrogen-adsorption measurements conducted on a GAI-100 high-pressure isotherm adsorption
were performed on a Micromeritics ASAP 2460 physical instrument at temperatures of 35, 60, and 85°C, at pressures
adsorption instrument. About 0.4 g of each powder sample ranging from 0.15 to 14 MPa. The samples were dried at
was automatically degassed at 110°C for 24 hours in 100°C for about 24 hours in a He atmosphere to remove the
vacuum to remove adsorbed moisture and volatile matter. moisture in the sample and loading about 110 g of dried-
The degassed samples were exposed to nitrogen at í196 °C powder sample into the experiment cell. The experiment
(with liquid nitrogen) along a series of precisely controlled system (Fig. 2) was ¿lled with He gas to remove the other
gas pressure. The pore-structure characteristics, such as gases and the sample cells were pressurized with He gas
speci¿c surface area, pore volume, and pore diameter, were up to 20 MPa for a leak test at experiment temperature
generated by instrument software which included Brunauer- for 24 hours (an accepted leakage is < 6.89×10-4 MP/h.
Emmett-Teller (BET) analysis for speci¿c surface area, and The void volume was determined by helium expansion
nonlocal density functional theory (NLDFT) for pore volume at experiment temperature individually for the methane-
and varied range of pore-size distributions (Lastoskie et al., adsorption isotherms calculation. Valves 2 and 3 were kept
1993). closed and the methane was transferred to the tank for a
The morphological and structural characteristics of the ¿xed volume and Valve 1 was closed. When the pressure
pores were observed using a SEM with Merlin Compact in the tank reached equilibrium, Valve 2, which connects
¿eld emission environment produced by Zeiss, which has an the tank and the cell, was opened. When the change in
accelerating voltage range of 20 V-30 kV, a resolution of 1.6 pressure to decreased below 6.89×10-4 MP over 5 min, the
nm@1 kV, and a magni¿cation of 12X to 2×106X. The block equilibrium pressure was assumed to have been achieved
samples (about 5×5×3 mm) were ¿rst manually polished, and the excess adsorption of methane at a given temperature
and then prepared by argon-ion cross-section polishing and equilibrium pressure could be calculated via the method
(Leica EMTIC020 mill) using an acceleration voltage of 8 previously reported (Zhang et al., 2012, Gasparik et al.,
kV, a sample current of 2.8 mA, and a milling time of 1 hour. 2014; Xiong et al., 2016).
RESULTS AND DISCUSSION isotherms for samples YSB-05, NZ-09, and FZ-14, and
the Table 2 shows the pore structure characterization.
Pore Structure Characterization and Main Controlling According to the International Union of Pure and Applied
Factors Chemistry (IUPAC), pores are classi¿ed as micropores (<2
The TOC and XRD analysis results are presented in nm), mesopores (2 to 50 nm), and macropores (>50 nm),
Table 1. Figure 3 show the low-pressure N2-adsorption additionally, the pores are generally divided into organic
Q, quartz; F, feldspar; Car, carbonate; Py, pyrite; Bar, barite; Anh, anhydrite; K, kaolinite; C, chlorite; I, illite; I/S, illite/montmorillonite mixed layer; C/S,
chlorite/montmorillonite mixed layer. “/”, not detected.
pores and inorganic pores based on the pore-forming material widely distributed, inorganic pores including intergranular
(Slatt and O’Brien, 2011). Figure 4 presents SEM images of pores, crystal particle pores, microcracks, and intragranular
sample microscopic pores. In Fig. 4 the organic pores are pore can also be seen in the images.
Fig. 3— Low-pressure N2-adsorption isotherms for Samples YSB-05, NZ-09, and FZ-14.
a: T-P, total pores, b: Mi-P, micropores (<2 nm), c: Me-P, mesopores (2 to 50 nm), d: Ma-P, macropores (>50 nm).
Micropores in the shale primarily provide speci¿c Isothermal Adsorption Characteristics and
surface area, while mesopores and macropores provide Thermodynamic Parameters
pore volume. Figure 5 show the relationship between shale Samples FZ-14, NZ-09, YSB-05 were subjected to
composition and pore SDFT (the pore surface area generated isothermal adsorption experiments and the result are shown
by the NLDFT model), the TOC and the speci¿c surface area in Fig. 7.
of micropores show a good positive correlation (Fig. 5a), Similar to previous studies, the amount of methane
indicating that organic matter is the main form of micropores. adsorption in the three samples at same temperature increased
There is a positive correlation between N2 adsorption and its with the increase of pressure and when the pressure reached
speci¿c surface area, this is also reÀected in the increase in a certain value, the adsorption amount gradually became
the TOC, the shale adsorption gas capacity increases (Tian et stable. The three parameters in the Langmuir adsorption
al., 2016; Wang, 2017). There is a slight positive correlation model (Eq. 1) and the thermodynamic adsorption principle
between the speci¿c surface area of the mesopores and the (Eq. 2) (Li et al., 2018) were used and the results were ¿tted
clay minerals, reÀecting the contribution of clay minerals to to obtain the Langmuir constant (K) and the maximum
the mesopores. Aside from the organic matter, clay minerals absolute adsorption capacity at a given temperature (nmax)
are the most important minerals that affect the content of (Table 3) and the adsorption thermodynamic parameters
adsorbed gas in the shale (Gu et al., 2018), and previous (Table 4).
study has shown that the adsorption capacity of clay minerals
has the following order: montmorillonite > I/M mixed layer
(1)
> kaolinite > chlorite > illite (Ji et al., 2012).
The pore volume would be more relevant to the free-
gas storage and Fig. 6 shows the relationship between
shale composition and pore VDFT (pore volume generated (2)
by the NLDFT model), which is similar to the surface area.
However, it’s important to note that the mesopore volume where nexcess is the excess adsorption amount obtained by
is much higher than the micropore and the macropore CH4 isothermal experiment, ȡg is density of free methane
volumes in the NLDFT method, which are limited. The VDFT at experimental temperature and pressure (which can be
shown in Figs. 6c and 6f is incomplete, which means that obtained via the US National Institute of Standards and
the macropores and mesopores provide much more storage Testing (NIST) package using the Setzmann and Wagner
space for free gas. equation (Setzmann and Wagner, 1991), ȡads is the density
Fig. 5—Crossplots showing the relationship between shale composition and the pore SDFT.
Fig. 7—Measured and Langmuir-based model calculations of excess adsorption isotherms of methane for shale.
Table 3—Langmuir Constant and Maximum Adsorption Amount of Each Sample Under Different Temperature
Table 4—Adsorption Heat and Standard Adsorption Entropy physical adsorption, the entropy produced by the conversion
of methane from free gas to adsorbed gas is about í87 J/
mol/K when the temperature is between 0 to 150°C (Xia
and Tang, 2012), while the entropy of Sample FZ-14 is
í98.8 J/mol/K (Table 4) which indicates that some methane
molecules are trapped in the micropores and have not been
adsorption (Hu et al., 2015).
The similarity between the diagram of excess adsorption
normalized to rock and to clay mineral or to TOC should
reÀect the level of the correlation between the rock and
of adsorbed methane which is always ¿tted to values of
clay and organic matter that the higher the similarity, the
424 mg/cm3 (the density of liquid methane at its boiling
temperature under 0.1 MPa), q is enthalpy of adsorption,
ǻS0 is molar entropy of adsorption, and P0 is constant
atmospheric pressure 0.1 MPa.
As the experimental temperature increases, the
adsorption amount gradually decreases, indicating that
the adsorption gas molecular activity is enhanced with
increasing temperature. The q and ǻS0 are two important
thermodynamic parameters used to describe the dependence
of Langmuir constant on the temperature at a certain
temperature. A higher q indicates a stronger adsorption force
of the gas on the surface of the shale, and a larger ǻS0 means a
weaker adsorption force of the gas on the surface of the shale
(Zhang et al., 2012). The thermodynamic parameters listed
in Table 4 show that the isosteric heat of adsorption is 5.3 to (a)
22.2kJ/mol and the standard entropy of adsorption is í98.8
to í48.8J/mol·K. and the parameters prove that Sample FZ-
14 has the strongest adsorption to methane, while Sample
NZ-09 has the weakest.
stronger the correlation. In order to compare the effects of of the three samples changes from FZ-14 > NZ-09 > YSB-
clay mineral and TOC on CH4 adsorption, the isothermal 05 to NZ-09 > FZ-14 > YSB-05, which means that as the
methane adsorption at three temperature were normalized to temperature increases the adsorption amount of Sample FZ-
the original rock (Fig. 9a1-3), clay mineral (Fig. 9b1-3) and 14 decreases faster than Sample NZ-09. This characteristic is
TOC (Fig. 9c1-3). similar to the original rock in which the methane adsorption
As Figs. 9b1-3 show a similar characteristic that the amount of Sample FZ-14 decreased much faster than the
methane adsorption is normalized to the clay mineral of the other two with increasing temperature, indicating that the
three samples at different temperature that is Samples NZ-09 organic matter is in the dominate factor in the methane
> YSB-05 > FZ-14, but is signi¿cantly different from Figs. adsorption of shale.
9a1-3, which could mean that the methane adsorption of clay
minerals is not the key factor in the methane adsorption in CONCLUSIONS
shale. Additionally, the clay-mineral content of Samples
NZ-09 and YSB-05 are relatively low but the clay-mineral The range of the thermodynamic parameters of Niutitang
normalized adsorption is higher than Sample FZ-14, which Formation shales in Huijunba syncline were calculated as
means that the adsorption of shales with low clay-mineral (1) 5.3 to 22.2kJ/mol for the isosteric heat of adsorption,
content may have a higher effective adsorption position. and (2) í98.8 to í48.8J/mol·K for the standard entropy of
It is clear from Figs. 9c1-3 that the adsorption amount adsorption.
decreases with increasing temperature. The ¿gures are very The nanoscale micropores in the study area are very
different from each other at different temperatures, which developed and the organic matter is the main material basis
range from 35 to 85°C, the order of the amount of adsorption for the formation of micropores. The surface area has two
Fig. 9—CH4-adsorption isotherms for three samples, normalized to rock (a), to clay mineral (b), and to TOC (c),
peaks in the range of pore diameter below 2 nm, which Image Analyses: Examples From the Barnett, Woodford,
provides the main sites for methane adsorption. Haynesville, Marcellus, And Doig Units, AAPG Bulletin,
The CH4 isothermal adsorption normalized to the 96(6):1099–1119. DOI: 10.1306/10171111052.
original rock, clay mineral and TOC reÀects that the organic Curtis, J.B., 2002, Fractured Shale-Gas Systems, AAPG Bulletin,
86(11), 1921–1938. DOI: 10.1306/61EEDDBE-173E-11D7-
matter is in the dominate factor in the methane adsorption of
8645000102C1865D.
shale and the adsorption of low clay-mineral content shales Gasparik, M., Bertier, P., Gensterblum, Y., Ghanizadeh, A.,
may have a higher effective adsorption position. Krooss, B.M., and Littke, R., 2014, Geological Controls
on the Methane Storage Capacity in Organic-Rich Shales,
ACKNOWLEDGMENTS International Journal of Coal Geology, 123, 34–51. DOI:
10.1016/j.coal.2013.06.010.
This research is ¿nancially supported by the Independent Gu, Y., Ding, W., Yin, S., Yin, M., and Xiao, Z., 2018,
subject of the Key Laboratory of Coal Exploration and Adsorption Characteristics of Clay Minerals in Shale,
Comprehensive Utilization, Ministry of Land and Resources Petroleum Science and Technology, 36(2):108–114. DOI:
10.1080/10916466.2017.1405031.
(ZP2019-3), the “Enterprise top innovative young talents
Hu, H., Zhang, T., Wiggins-Camacho, J.D., Ellis, G.S., Lewan,
support plan” (20180407), and Shaanxi Provincial Key
M.D., and Zhang, X., 2015, Experimental Investigation of
Research and Development Project (2017GY-150). Changes in Methane Adsorption of Bitumen-Free Woodford
Shale With Thermal Maturation Induced by Hydrous
NOMENCLATURE Pyrolysis, Marine and Petroleum Geology, 59, 114–128. DOI:
10.1016/j.marpetgeo.2014.07.029.
Abbreviations Ji, L., Zhang, T., Milliken, K.L., Qu, J., and Zhang, X., 2012,
Experimental Investigation of Main Controls to Methane
BET = Brunauer-Emmett-Teller Adsorption in Clay-Rich Rocks, Applied Geochemistry,
I/M = illite/montmorillonite 27(12), 2533–2545. DOI: 10.1016/j.apgeochem.2012.08.027.
NL-DFT = nonlocal density functional theory Lastoskie, C., Gubbins, K.E., and Quirke, N., 1993, Pore Size
SDR = supercritical Dubinin–Radushkevich Distribution Analysis of Microporous Carbons: A Density
SEM = scanning electron microscope Functional Theory Approach, The Journal of Physical
SLD = simpli¿ed local density Chemistry, 97(18):4786–4796. DOI: 10.1021/j100120a035.
TOC = total organic carbon Li, J., Zhou, S., Gaus, G., Li, Y., Ma, Y., Chen, K., and Zhang, Z.,
XRD = X-ray diffraction 2018, Characterization of Methane Adsorption on Shale and
Isolated Kerogen From the Sichuan Basin Under Pressure up
to 60 MPa: Experimental Results and Geological Implications,
Symbols
International Journal of Coal Geology, 189, 83–93. DOI:
K = Langmuir constant 10.1016/j.coal.2018.02.020.
nexcess = excess adsorption amount obtained by Setzmann, U., and Wagner, W., 1991, A New Equation of State and
CH4 isothermal experiment Tables of Thermodynamic Properties for Methane Covering
nmax = maximum absolute adsorption capacity at the Range from the Melting Line to 625 K at Pressures up to
a given temperature 100 MPa, Journal of Physical and Chemical Reference Data,
P = pressure 20(6), 1061–1155. DOI: 10.1063/1.555898.
P0 = constant atmospheric pressure (0.1MPa) Slatt, R.M., and O’Brien. N.R., 2011, Pore Types in the Barnett and
q = enthalpy of adsorption Woodford Gas Shales: Contribution to Understanding Gas
Storage and Migration Pathways in Fine-Grained Rocks, AAPG
R = nolar fas constant (8.314 J/mol/K)
Bulletin, 95(12), 2017–2030. DOI: 10.1306/03301110145.
ǻS0 = molar entropy of adsorption Tian, H., Li, T., Zhang, T., and Xiao, X., 2016, Characterization
T = temperature of Methane Adsorption on Overmature Lower Silurian–
ȡg = the density of free methane at experimental Upper Ordovician Shales In Sichuan Basin, Southwest
temperature and pressure China: Experimental Results and Geological Implications,
ȡads = the density of adsorbed methane (424 mg/cm3) International Journal of Coal Geology, 156, 36–49. DOI:
10.1016/j.coal.2016.01.013.
Wang, Y., Zhu, Y., Liu, S., and Zhang, R., 2016, Pore
REFERENCES Characterization and its Impact on Methane Adsorption
Capacity for Organic-Rich Marine Shales, Fuel, 181, 227–
Chalmers, G.R., Bustin, R.M., and Power. I.M., 2012, 237. DOI: 10.1016/j.fuel.2016.04.082.
Characterization of Gas Shale Pore Systems by Porosimetry, Wang, Z. 2017, Characterization of the Microscopic Pore
Pycnometry, Surface Area, and Field Emission Scanning Structure of the Lower Paleozoic Shale Gas Reservoir in the
Electron Microscopy/Transmission Electron Microscopy Southern Sichuan Basin and its InÀuence on Gas Content,
Petroleum Science and Technology, 35(23), 2165–2171, DOI: Yang Fu graduated from Northwest University (China)
10.1080/10916466.2017.1390682. and worked as a senior engineer at Shaanxi Coal¿eld
Xia, X., and Tang, Y., 2012, Isotope Fractionation of Methane Geology Group Co., Ltd. He is engaged in unconventional
During Natural Gas Flow With Coupled Diffusion and natural gas geological research at the Key Laboratory of
Adsorption/Desorption, Geochimica et Cosmochimica Acta,
Coal Resources Exploration and Comprehensive Utilization
77, 489–503. DOI: 10.1016/j.gca.2011.10.014.
Xiong, F., Jiang, Z., Li, P., Wang, X., Bi, H., Li, Y., Wang, Z.,
of Ministry of Land and Resources, China. He has a PhD in
Amooie, M.A., Soltanian, M.R., and Moortgat, J., 2017, geology.
Pore Structure of Transitional Shales in the Ordos Basin, NW
China: Effects of Composition on Gas Storage Capacity, Fuel,
206, 504–515. DOI: 10.1016/j.fuel.2017.05.083.
Xiong, W., Zuo, L., Luo, L., Hu, Z., and Cui, Y., 2016, Methane
Adsorption on Shale Under High Temperature and High
Pressure of Reservoir Condition: Experiments and Supercritical
Adsorption Modeling, Adsorption Science & Technology,
34(2–3), 193–211. DOI: 10.1177/0263617415623425.
Zhang, T., Ellis, G.S., Ruppel, S.C., Milliken, K., and Yang,
R., 2012, Effect of Organic-Matter Type and Thermal
Maturity on Methane Adsorption in Shale-Gas Systems,
Organic Geochemistry, 47, 120–131. DOI: 10.1016/j.
orggeochem.2012.03.012.
ABSTRACT
Our previous work con¿rmed that Bayesian inversion analysis programs. Considering carbonate core samples
enables whole-core tomography with sparse and noise- with complex pore topology, images stemming from
perturbed measurements, suitable for wellsite portable Bayesian inversion yield estimates of image-detectable
industrial scanners. The current study examines further macroporosity within 8% relative to those from customary
potential of Bayesian inversion on-site X-ray computerized medical tomograms and depict similar distribution of vugs
tomography (CT) in producing quanti¿able digital and angular bedding. Moreover, further appraisal of the
representations of core samples. To that end, workÀows performance of the proposed imaging technique, in the
using 3D tomographic images depict (1) rock heterogeneity, presence of multiphase saturation, considers a micro-CT
(2) lithological changes, (3) bedding planes, (4) fractures of an oil-bearing carbonate after waterÀooding. The intent
and nodules, and (5) scale of heterogeneity by quantifying is to con¿rm availability of a consistent CT method across
porosity regions. In so doing, this wellsite-based analysis scales, and its relevance to whole-core imaging in situations
aids in accelerating an informed assessment of whole of sponge coring or ¿ltrate invasion. At this scale, while
round core and plug-sample locations for laboratory images from Bayesian inversion generate coarsened pore
analysis. Additionally, earlier ¿rst-stage core analysis, networks, porosity and saturation estimates are within 2%
promotes timely core-log integration and can signi¿cantly compared to those from customary micro-CT.
improve ef¿ciency of subsequent laboratory sampling and
Manuscript received by the Editor November 27, 2018; manuscript accepted February 18, 2019.
1
The Alan Turing Institute, UK; The British Library, 96 Euston Road, London, NW1 2DB; amendozachavez@turing.ac.uk; mgirolami@turing.ac.uk
2
LUT University, Finland; Yliopistonkatu 34, FI-53850 Lappeenranta; lassi.roininen@lut.¿; heikki.haario@lut.¿
3
Finnos Oy, Tukkikatu 5, FI-53900 Lappeenrant; jere.heikkinen@¿nnos.¿
useful information about whole-core samples. In this study, METHOD AND WORKFLOW
our primary objective is to evaluate the potential of Bayesian
inversion wellsite whole-core tomography as a quanti¿ed Particularly when whole-core samples are required for
representation of the sample. To that end, we assess the reservoir characterization, on-site tomographic imaging
performance of the inversion technique in characterizing (1) enables accelerated preliminary integrated analysis.
rock heterogeneity, (2) lithological changes, (3) fractures Additionally, earlier sample imaging aids in designing
and nodules, (4) bedding planes, and (5) porosity regions. more informed, time-ef¿cient, and cost-effective laboratory
We show that similar information about the core can core-analysis programs. Depending on drillsite location
be inferred from sparse-measurement Bayesian inversion and logistics of transportation to a core-analysis laboratory,
tomograms and dense-measurement (base-case) medical it may take days to several weeks before beginning core
X-ray CT images of whole-core carbonate samples. At the analysis. Alternatively, if whole-core X-ray imaging is
outset, we examine two contrasting carbonate whole-core completed on-site, an informed ¿rst-stage evaluation of the
samples of complex pore topography and distinctive degrees core can be completed prior to its arrival to the laboratory
of heterogeneity. In so doing, we segment 3D tomograms (see Fig. 1). Moreover, accelerated 3D whole-core imaging
into pore and rock matrix, and subsequently ¿t pore-network enables the design of detailed sampling plans that aid in
models for quantitative comparison between inversion selecting location and orientation of plugs and whole-core
results and base-case. Similarly, we consider imaging a sample intervals (see Fig. 2).
microscale case of a waterÀooded oil-bearing carbonate A practical on-site whole-core tomographic imaging
with the intent of evaluating the inversion performance in technique is feasible with a portable X-ray source-detector
the presence of multiphase saturation. This attribute could assembly, which either rotates around a stationary sample
be useful in whole-core applications such as situations of or is ¿xed for a rotating core. While this imaging setting can
sponge whole core, or mud-¿ltrate invasion. acquire dense, full- or limited-angle, X-ray measurements,
Fig. 1—Simpli¿ed core-analysis workÀow from wellsite whole-core recovery to specialized laboratory procedures in subsamples. The top section
shows a standard situation when the ¿rst X-ray CT is done in the laboratory and the bottom section shows a proposed workÀow where X-ray CT of
whole-core samples is completed on-site with a portable industrial scanner (see Fig. 3). RCA and SCAL refer to routine core analysis and special
core analysis, respectively.
Fig. 2—3D visualization of a 0.5-m long, 5-in. diameter whole-core and potential 1.5-in. diameter plug samples, allowing accelerated detailed sampling
planning for laboratory analysis, when imaging is completed on-site.
the portability of the scanner is inversely proportional sample. To that end, we use a single-component Metropolis-
to the measurement sophistication. Consequently, we Hastings (SCMH) algorithm to estimate the posterior
consider X-ray imaging with sparse and limited-angle conditional mean (CM) and reconstruct 3D whole-core
measurements, that simpli¿es the scanning equipment, X-ray images as in Mendoza et al. (2019). In this analysis,
increasing its portability. While conventional (medical) at the expense of additional computational intensity for the
scanners produce high-resolution tomographic images inversion, we consider 91 angular projections (still very
using direct computational methods, i.e., ¿lter back- sparse, given that FBP typically uses more than 2,000)
projection (FBP), which are fast, these imaging techniques acquired in a limited interval of 180° around the sample,
require dense and noise-free measurements (Vepsäläinen with 10% noise contamination. As described by Markkanen
et al., 2014). Mendoza et al. (2019) show that statistical et al. (2019) and Mendoza et al. (2019), the assumed
tomographic imaging methods, namely Bayesian inversion, measurement con¿guration consists of a 2D fan-beam with
can reconstruct whole-core images with sparse and noise- source and detector relative distance to the rotational axis of
contaminated measurements, which can effectively detect 4 and 2, respectively.
coarse sedimentary features and provide useful information
about the rock texture and heterogeneity. In so doing, Tomographic imaging with Bayesian inversion
portable industrial scanners combined with the proposed For tomographic imaging, with discretized “unknown”
computational method can accelerate availability of valuable parameter-vector v ԹN (i.e., with N pixel/voxel values), the
information (see Fig. 3). linear inverse problem is characterized by
In this paper, we further examine the potential of on-site
Bayesian inversion (via Markov Chain Monte Carlo, MCMC) (1)
whole-core tomography as a quanti¿ed representation of the
Fig. 3—Laboratory (top-left) and industrial (bottom-left) X-ray scanning instrumentation and schematics of their corresponding measurement
principles. Right panels represent direct (top) and statistical (bottom) computational imaging methods, suitable for dense and sparse measurements,
respectively.
where m is the known, noise-perturbed measurements, h, h’ > 0 and regularization parameter ı2, we give the prior
A is the known linear forward-theory matrix, and e is the with independent and identically distributed increments (i.e.,
measurement-noise vector with known statistical properties differences between neighboring pixels) as
(refer to Mendoza et al. (2019) for construction of these
vectors and matrices). The solution of a Bayesian statistical (4)
inverse problem is the posterior probability density, which
we give as an un-normalized representation as follows: (5)
variations is critical for designing improved subsampling presence of multiple saturating Àuids. Additionally, we
and analysis plans. To that end, ¿rst, we segment the 3D assess scale-consistency of the technique, which may be
whole-core images into pore space and rock matrix using useful for upscaling digital rocks, for example. Moreover,
thresholding (where the threshold values are determined because of its statistical characteristics, Bayesian inversion
consistently for multiple samples). Second, in analogy may offer insights into quantifying the intrinsic uncertainty
to a micro-CT scale core-analysis workÀow, we extract of the petrophysical properties inferred from X-ray images,
a conventional pore network from discretization of the which we leave for future study.
pore space from reconstructed images using the algorithm
Pnextract (Raeini et al., 2017) and OpenPNM (Gostick et Whole Core X-Ray CT
al., 2016). To that end, we report visual, using ParaView Sample 1 in Fig. 4 is a tomographic image of a whole-core
(Ayachit, 2015), and tabulated comparisons between carbonate rock sample with complex pore topology resulting
the results from the inversion image and those from the from diagenetic processes (sample C04B21_Raw140keV
unknown. A last set of results, which is a microscale image, from Victor and Prodanovic (2017)). To compare medical
includes, in addition to the above, comparison of estimated X-ray CT images to corresponding Bayesian inversion
relative permeability estimates via the network simulation tomograms, we show results using sections of the samples
code PoreÀow (Valvatne and Blunt, 2003). described by Victor et al. (2017) in their dual-energy CT-
based Monte Carlo study. While Bayesian inversion was
RESULTS AND DISCUSION completed for the entire region in the medical experimental
setup, we show results for an 8.5×8.5×50 cm, rectangular
To evaluate the feasibility of obtaining valuable section (i.e., focusing on the rock sample only). To that end,
information about core samples from Bayesian inversion the 3D tomograms consist of 137×137×400 voxels, each
3D whole-core images, we compare vCM estimates to of dimensions 0.488×0.488×1.25 mm in the x, y, and z
corresponding base-case images (acquired in the laboratory coordinate directions.
domain). We assume that the unknown is represented The left panels in Fig. 4 show the medical CT image
by the base-case image, which consequently, in our (Fig. 4a) and the pore network model (Fig. 4g) stemming
formulation, restricts the resolution of the estimator to that from binarizing the tomogram into rock matrix (Fig. 4c)
of the unknown. The goal is to obtain similar information and pore-space (Fig. 4e). Accordingly, the right panels show
from the estimators using sparse, partial angle, and noise- the corresponding estimators from Bayesian inversion. The
contaminated measurements, to that from standard X-ray CT results in Figs. 4b, 4d, and 4f show that, on ¿rst inspection,
images. the tomograms reconstructed with Bayesian inversion (using
Because whole-core analysis is of special interest in sparse, noise-perturbed, and limited-angle measurements)
carbonates due to their scale of heterogeneity (see Baynum image rock heterogeneity similar to that of the base-case.
and Koepf, 1957; Honarpour et al., 2005; Mohamed et al., While the density of large vugular pores is lower in the
2010; Skinner et al., 2015; Victor et al., 2017; Dernaika et inversion results, the distribution of porosity regions is
al., 2018), at the outset, we present results from two distinct similar to that of the base-case, e.g., lower porosity planes
carbonate samples. Given the sample scale and image between 150 and 200 mm and 350 and 450 mm in the
resolution, the analysis focuses in obtaining information z direction. Additionally, similar lithological variations
about rock heterogeneity, lithological variations, bedding (dolomitization) are reproduced by relative concentrations
planes, presence of fractures and nodules, and regions of of more- and less-attenuating rock, in orange and yellow
concentrated macroporosity and vugular porosity. colors, respectively.
An additional set of results examines the performance of While detection of the above sedimentary characteristics
the inversion at the microscale in the presence of multiphase yields valuable information to quantify and compare the
saturation. Conversely, the objective of this example is not porosity attributes (captured within the sample scale and
to evaluate the potential of on-site CT at the microscale. image resolution) from the inversion image, we examine
Because micro-CT is customarily done in small rock extracted pore networks. In so doing, we construct
samples and using specialized laboratory high-resolution conventional pore-network models following techniques
instrumentation (see Ketcham and Carlson, 2001; Arns et al., designed for micro-CT analysis (Dong and Blunt, 2009;
2005; Cnudde and Boone, 2013), the use of on-site industrial Raeini et al., 2017) as a comparative metric of the inversion
scanners, at this scale, remains impractical. Accordingly, the image to the unknown and not as a meticulous description
intent of this example is to further validate the technique of the rock porosity and connectivity. To that end, Table 1
by examining its performance in smaller scale and in the lists characteristics of pore-network models constructed
Fig. 4—Sample 1 whole-core spatial distribution of the macroporosity estimated from Bayesian inversion results (right panels) vs. that of the unknown
(left panels). (g) and (h) compare pore networks extracted (in analogy to a micro-CT scale image) using the whole-core macroporosity distribution (see
Table 1). Axes show length in mm and voxel size is 0.488×0.488×1.25 mm.
Table 1—Comparison of Pseudo-Pore-Network Properties Extracted From Whole-Core Sample Images (see Figs. 4 and 6).
using medical (unknown) and inversion images. Porosity Threshold CT values for porosity were determined as,
depicts the ratio of number of voxels classi¿ed as pore space CTth = ȝ – aı. Where ȝ is the mode of the image CT density, ı
(Figs. 4e and 4f), below a threshold CT value (see Fig. 5a), is the standard deviation of a normal distribution, with mean
to those classi¿ed as matrix (Figs. 4c and 4d), above the ȝ, that best ¿ts the image density (neglecting the left tail),
threshold. Similarly, net porosity is the percent pore space and a is a scaling factor equal to 2.3 and 2.74 for inversion
calculated after network extraction via Pnextract (Raeini et and unknown images, respectively.
al., 2017); throats are cylindrical pore-to-pore connections, Sample 2 in Fig. 6 is of the same dimensions of Sample
and connection number represents the number of connecting 1 and is a section of sample C04B10_Raw140keV from
pores to a single pore. While the percent difference between Victor and Prodanoviü (2017). Characteristically different
the porosity from Bayesian inversion and base-case is 8%, from Sample 1, Sample 2 includes larger regions of
the number of pores and throats differ by 14 and 21%, subresolution vugular porosity and angular bedding planes
respectively. with large vugs. Additionally, we include a fractured section
of the core for imaging. Following the same tomographic
image CT thresholding procedure as used in Sample 1, Figs.
6d and 6f show Bayesian inversion results for rock matrix
and porosity, respectively. Accordingly, on ¿rst inspection,
porosity regions (i.e., large vugs and subresolution porosity)
and fracture geometry are similar to those depicted from the
unknown (Figs. 6c and 6e). In this case, for pore-network
extraction, we exclude the last 5 cm of the sample, which
contain the fracture. To that end, visual comparison between
the network form Bayesian inversion results (Fig. 6h) and
that of the base-case (Fig. 6g) shows that, while the location
of vugs is similar, pore size and number is signi¿cantly
different. Incidentally, Table 1 shows that the porosity (and
net porosity) values of the inversion image are very similar
(even more than in the case of Sample 1) than those of the
unknown. However, contrary to Sample 1, in this case the
number of pores and throats differ from the base-case values
by 49 and 60%, respectively.
The above comparisons show that although quantitative
characterization of the pore network is not feasible using
Bayesian inversion (or medical) tomograms from whole-
core samples (with the resolution of the images used in
this study), both medical and industrial CT images yield
similar valuable information. Successful detection of rock
heterogeneity, vugular porosity regions, bedding planes,
and fractures with Bayesian inversion enables accelerated
optimization of core analysis with measurements acquired
Fig. 5—Normalized frequency of CT values for (a) Sample 1, and (b) on-site. Additionally, reservoir rock heterogeneity inferred
Sample 2. Inversion results are in red and the unknown is in black (see from accelerated whole-core tomographic images can
Figs. 4 and 6). Dashed vertical lines are corresponding porosity cutoff
values. be valuable input for ¿rst-stage analysis based on digital
Fig. 6—Sample 2 whole-core spatial distribution of the macroporosity estimated from Bayesian inversion results (right panels) vs. that of the unknown
(left panels). (g) and (h) compare pore networks extracted (in analogy to a micro-CT scale image) using the whole-core macroporosity distribution (see
Table 1). Axes show length in mm and the voxel size is 0.488×0.488×1.25 mm.
techniques. On-site images could be used for preliminary into single, larger pores. Consequently, the extracted pore
estimates of density and atomic number (Victor et al., 2017) network, of similar net porosity, consists of fewer and larger
or a preface assessment of the effects of heterogeneity pores and pore throats, i.e., a simpli¿ed, coarsened version
on permeability following similar techniques to those of of the base-case.
Dernaika et al. (2018). To assess the impact of pore-network model coarsening,
Results con¿rm that accelerated whole-core-based stemming from use of sparse X-ray angular projections (and
¿rst-stage characterization of rock heterogeneity and possibly imaging method), on rock Àow attributes, Fig. 9
detection of coarse sedimentary features is feasible with compares estimated relative permeability curves. To that
industrial scanners combined with statistical imaging end, we consistently simulate relative permeability in an oil-
methods. However, to further assess the boundaries of the brine system using both unknown and inversion images via
of the inversion technique, it is advantageous to examine its the network-simulation code PoreÀow (Valvatne and Blunt,
performance at the microscale and high-resolution. To that 2003). In both, drainage and imbibition processes, estimated
end, (acknowledging that micro-CT resides in the laboratory values of oil, kro, and water relative permeability, krw, match
domain and with specialized scanning techniques) in a when values of water saturation, Sw, are below approximately
similar way as in the whole-core examples, we report an 42%.
example of Bayesian inversion imaging at the microscale in While the intent of this exercise is not to promote use of
the presence of multiphase saturation. sparse-measurement-based imaging for meticulous micro-
CT analysis, results show further feasibility of the inversion
Microscale X-Ray CT technique in detecting multiphase saturations and preserving
Sample 3 in Fig. 7 is a 1.2-cm cubic section of Sample2_ useful information about rock connectivity, under enough
Original from Alhammadi et al. (2018), an oil-bearing detector resolution.
mixed-wet carbonate reservoir rock after 20 pore volumes of
waterÀooding (see Alhammadi et al. (2017) for details of the SUMMARY AND CONCLUSIONS
waterÀooding conditions and wettability characterization).
The 3D tomogram under analysis consists of 600×600×600 With the conception that customary, direct, X-ray
cubic voxels of 2 ȝm size. Figure 7 compares distribution CT methods, namely FBP, require dense and noise-free
of water (Fig. 7d) and remaining oil (Fig. 7f) stemming measurements, generally acquired in a laboratory, we
from Bayesian inversion imaging to corresponding Àuid evaluated the feasibility of using simpli¿ed measurement
distributions from standard micro-CT images (Figs. 7c and principles for accelerated on-site tomographic imaging.
7e). In both, inversion and base-case, we used corresponding In particular, we con¿rmed that combination of Bayesian
thresholding determined from the image CT distribution inversion (via MCMC methods) with sparse and noise-
for segmentation into rock matrix, brine, and oil. Figure contaminated measurements, can yield valuable information
8 shows normalized density and threshold values for the prior to transporting core to a laboratory. In so doing, portable
unknown (black) and inversion results (red). A remarkable industrial scanners can successfully produce 3D tomograms
characteristic of the Bayesian inversion image is the more of core samples useful for accelerated ¿rst-stage analysis.
evident trimodal atribute of its density compared to that of Using medical CT images of two carbonate samples as a
the unknown. This characteristic aids in selecting threshold base-case, we compared basic sedimentary characteristics
values, which in this case, yield very similar porosity and inferred from these tomograms to those from Bayesian
Àuid-saturation estimates between the unknown and the inversion images and showed that:
inversion images (Table 2). x In carbonates with complex pore topology, inversion
Figures 7g and 7h show that, in general, pore size is images effectively characterize distribution of
signi¿cantly larger in the pore network extracted from macroporosity regions, large vugs, and angular
inversion results compared to that from the unknown. bedding.
Accordingly, for similar net porosity, the number of pores x Bayesian inversion images show similar proportion
is substantially lower in the case of the Bayesian inversion of lithological variations i.e., dolomitization.
estimator. While the percent variation of net porosity x Inversion yields similar values (up to 8% difference)
between the two images is less than 2%, the number of of CT-derived porosity (neglecting subresolution
pores and throats decrease in the inversion by 72 and 82%, pores) via thresholding, while differences in ¿tted
respectively (see Table 2). These results indicate that sparsity pores and pore throats were as high as 49 and 60%,
of measurements used in Bayesian inversion yields a coarser respectively. However, extracted pore networks
tomographic image that groups neighboring small pores preserve the relative distribution of large pores.
Fig. 7—Sample 3 spatial distribution of the porosity and Àuids estimated from Bayesian inversion results (right panels) vs. that of the unknown (left
panels). (g) and (h) compare extracted conventional pore networks (see Table 2). Axes show length in ȝm and voxel size is 2×2×2 ȝm.
Fig. 8—Normalized frequency of image values for Sample 3. Inversion Similarly, further assessment of the performance of
results are in red and the unknown is in black (see Fig. 7). Dashed and
dotted vertical lines are corresponding porosity and Àuid-phase cutoff
the inversion technique at the microscale in the presence of
values, respectively. multiphase saturation shows that:
x The multimodes of the density of Bayesian inversion
images are more evident, aiding basic segmentation
via thresholding.
x While CT image estimates of porosity and saturation
are within 2%, pore-network modeling shows over
70% decrease in the number of ¿tted pores and
pore throats, but with larger radius (i.e., a coarsened
description of pore network).
ACKNOWLEDGMENTS
NOMENCLATURE Alhammadi, A., AlRatrout, A., Singh, K., Bijeljic, B., and Blunt,
M.J., 2017, In Situ Characterization of Mixed-Wettability in
Abbreviations a Reservoir Rock at Subsurface Conditions, Nature: Scienti¿c
Reports, 7(1), 10753. DOI: 10.1038/s41598-017-10992-w.
3D = three-dimensional Arns, C.H., Bauget, F., Ghous, A., Sakellariou, A., Senden, T.J.,
CM = conditional mean Sheppard, A.P., Sok, R.M., Pinczewski, W.V., Kelly, J.C.,
CT = computational tomography and Knackstedt, M.A., 2005, Digital Core Laboratory:
FBP = ¿lter back-projection Petrophysical Analysis From 3D Imaging of Reservoir Core
MCMC = Markov chain Monte Carlo Fragments, Petrophysics, 46(4), 260–277.
OpenPNM = open-source pore-network modeling package Ayachit, U., 2015, The Paraview Guide: A Parallel Visualization
Application, Kitware Inc. ISBN 978-1930934306.
ParaView = open-source, multiplatform data analysis and
Baynum, R.S., and Koepf, E.H., 1957, Whole-Core Analysis
visualization application Methods and Interpretation of Data from Carbonate
PoreÀow = two-phase network modeling code Reservoirs, Journal of Petroleum Technology, 9(11), 11–15,
Pnextract = pore-network extraction code DOI: 10.2118/817-G.
RCA = routine core analysis Cnudde, V., and Boone, M., 2013, High-Resolution X-Ray
SCAL = special core analysis Computed Tomography in Geosciences: A Review of the
SCMH = single-component Metropolis Hastings Current Technology and Applications, Earth-Science Reviews,
123, 1–17, DOI: 10.1016/j.earscirev.2013.04.003.
Dernaika, M., Al Mansoori, M., Singh, M., Al Dayyani, T.,
Symbols Kalam, Z., Bhakta, R., Koronfol, S., and Uddin, Y.N., 2018,
a= scaling factor Digital and Conventional Techniques to Study Permeability
A= linear forward theory matrix Heterogeniety in Complex Carbonate Rocks, Petropysics,
CTth = threshold CT value 59(3), 373–396, DOI: 10.30632/PJV59N3-2018a6.
D(•) = probability density of • Dong, H., and Blunt, M.J., 2009, Pore-Network Extraction from
D(•|*) = conditional probability density of • given * Micro-Computerized-Tomography Images, Physical Review
E, 80(3), 036307. DOI: 10.1103/PhysRevE.80.036307.
e= measurement noise vector
Freifeld, B., Kneafsey, T., and Rack, F., 2006, On-Site Geological
h= magnitude of vertical discretization step Core Analysis Using a Portable X-ray Computed Tomographyc
h’ = magnitude of horizontal discretization step System, in Rothwell, R.G., New Techniques in Sediment Core
j= vertical discretization step Analysis, Geological Society (London) Special Publication,
j’ = horizontal discretization step 267, 165–178. DOI: 10.1144/GSL.SP.2006.267.01.12.
kro = oil relative permeability Freifeld, B., Kneafsey, T., Tomutsa, L., and Pruess, J., 2003,
krw = water relative permeability Development of a Portable X-ray Computed Tomographic
Imaging System for Drill-Site Inverstigation of Recovered
ȝ= mode of the image CT density
Core, Paper SCA2003-51 presented at the International
m= “known” measurements Symposium of the Society of Core Analysts, Pau, France,
N= number of pixels/voxels 21–24 September.
Թ= set of all real numbers Gilks, W.R., Best, N.G., and Tan, K.C., 1995, Adaptive Rejection
ı= standard deviation of the normal distribution Metropolis Sampling Within Gibbs Sampling, Journal of The
ı2 = regularization parameter Royal Statistical Society, Series C (Applied Statistics), 44(4),
Ȉ= covariance matrix 455–472, DOI: 10.2307/2986138.
Gostick, J., Aghighi, M., Hinebaugh, J., Tranter, T., Hoeh, M.A.,
Sw = water saturation
Day, H., Putz, A., 2016, OpenPNM: A Pore Network Modeling
T= transpose Package, Computing in Science & Enginering, 18(4), 60–74,
ߥ= “unknown” parameter vector DOI: 10.1109/MCSE.2016.49.
ߥCM = conditional mean estimator of the posterior Hastings, W.K., 1970, Monte Carlo Sampling Methods Using Markov
distribution Chains and TheirApplications, Biometrika, 57(1), 97–109, DOI:
10.1093/biomet/57.1.97. http://citeseerx.ist.psu.edu/viewdoc/
download?doi=10.1.1.452.6839&rep=rep1&type=pdf.
Accessed May 10, 2019.
REFERENCES Honarpour, M.M., Djabbarah, N.F., and Sampath, K., 2005,
Whole-Core Analysis—Experience and Challenges, Paper
Alhammadi, A., AlRatrout, A., Bijelijc, B., and Blunt, M.J., SPE-81575, SPE Reservoir Evaluation and Engineering, 8(6),
2018, X-ray Micro-Tomography Datasets of Mixed-Wet 460–469, DOI: 10.2118/81575-PA.
Carbonate Reservoir Rocks for In Situ Effective Contact Ketcham, R.A., and Carlson, W.D., 2001, Acquisition, Optimization
Angle Measurements, Digital Rocks Portal. DOI: 10.17612/ and Interpretation of X-Ray Computed Tomographyc Imagery:
P7VQ2G. https://www.digitalrocksportal.org/projects/151. Applications to the Geosciences, Computers & Geosciences,
Accessed May 22, 2019. 27(4), 381–400, DOI: 10.1016/S0098-3004(00)00116-3.
Lee, J., Sung, W., and Choi, J., 2015, Metamodel for Ef¿cient Earth, 122, 9804–9824, DOI: 10.1002/2017JB014408.
Estimation of Capacity-Fade Uncertainty in Li-Ion Batteries
for Electric Vehicles, Energies, 8(6), 5538–5554, DOI: ABOUT THE AUTHORS
10.3390/en8065538.
Markkanen, M., Roininen, L., Huttunen, J.M.J., and Lasanen, S.,
Alberto Mendoza is a group leader for Oil and Gas
2019, Cauchy Difference Priors for Edge-Preserving Bayesian
InversionWith an Application to X-Ray Tomography, Journal research projects with the Data-Centric Engineering Program
of Inverse and Ill-Posed Problems, 1-16. DOI: 10.1515/jiip- at the Alan Turing Institute in London, UK. He holds PhD,
2017-0048. https://arxiv.org/pdf/1603.06135.pdf. Accessed MSc, and BS degrees in Petroleum Engineering from the
May 10, 2019. University of Texas at Austin, and an MSc in Statistics from
McPhee, C., Reed, J., and Zubizarreta, I., 2015, Core Analysis: Imperial College London. From 2008 to 2014, he worked
A Best Practice Guide.Elsevier, Development in Petroleum with ExxonMobil as a Formation Evaluation Specialist in
Science, 64, Elsevier. ISBN: 9780444635334.
the USA and Russia, and during 2014 to 2016 he was a
Mendoza, A., Roininen, L., Girolami, M., Heikkinen, J., and
Haario, H., 2019, Statistical Methods to Enable Practical Petrophysical Engineer with Shell | NAM in The Netherlands.
On-Site Tomographic Imaging of Whole-Core Samples,
Geophysics, 84(3). DOI: 10.1190/geo2018-0436.1. Lassi Roininen holds the position of Associate Professor
Mohamed, S.S., Dernaika, M., Al Hosani, I., Hannon, L., (tenure track) in applied mathematics in LUT University,
Skaeveland, S., Kalam, M.Z., 2010, Whole Core Versus Finland. He is also a docent in applied mathematics at the
Plugs: Integrating Log and Core Data to Decrease Uncertainty University of Oulu, Finland, and an Academy of Finland
in Petrophysical Interpretation and STOIP Calculations,
postdoctoral researcher. He develops rigorous numerical and
Paper SPE-137679 presented at the Abu Dhabi International
Petroleum Exhibition and Conference,Abu Dhabi, UAE, 1–4 computational tools for inverse problems with applications
November. DOI: 10.2118/137679-MS. in near-space remote sensing, subsurface imaging, and X-ray
Patel, G., 2009, Computed Tomography Image Reconstruction tomography.
[slide presentation], SlideShare.net. https://es.slideshare.
net/ripjan/ct-1470963?qid=3ce2b676-e1ea-4129-93c7- Mark Girolami is the Sir Kirby Laing Professor of
6558663060bf&v=&b=&from_search=6. Accessed May 10, Civil Engineering in the University of Cambridge where he
2019.
Raeini, A.Q., Bijelic, B., and Blunt, M.J., 2017, Generalized
also holds the Lloyds Register Foundation - Royal Academy
Network Modeling: Network Extraction as a Coarse-Scale of Engineering Research Chair in Data Centric Engineering.
Discretization of the Void Space of Porous Media, Physical He is Program Director at the Alan Turing Institute where he
Review E, 96(1), 013312. DOI: 10.1103/PhysRevE.96013312. served as one of the original founding Executive Directors.
Roininen, L., Huttunen, J., and Lasanen, S., 2014, Whittle-Matern He is an elected member of the Royal Society of Edinburgh
Priors for Bayesian Statistical Inversion with Applications and previously was awarded a Royal Society – Wolfson
in Electrical Impedance Tomography, Inverse Problems and Research Merit Award. Professor Girolami was anEPSRC
Imaging, 8(2), 561–586, DOI: 10.3934/ipi.2014.8.561.
Skinner, J.T., Tovar, F.D., and Schechter, D.S., 2015, Computed
Research Fellow from 2007 to 2018.
Tomography for Petrophysical Characterization of Highly
Heterogeneous Reservoir Rock, Paper SPE-177257 Jere Heikkinen holds an MSc in applied mathematics
presented at The SPE Latin America and Caribean Petroleum from the Technical University of Lappeenranta. He is
Engineering Conference, Quito, Ecuador, 18–20 November, working with limited- and sparse-angle X-ray tomography
DOI: 10.2118/177257-MS. through scienti¿c research as a PhD student and as general
Valvatne, P.H., and Blunt, M.J, 2003, Predictive Pore-Scale manager of Finnos Ltd. -- a Finnish technology company
Network Modeling, Paper SPE-84550 presented at the SPE
Annual Teachnical Conference and Exhibition, Denver,
which develops real-time imaging systems for various ¿elds
Colorado, USA, 5–8 October. DOI: 10.2118/84550-MS. of industries.
Vepsäläinen, M., Markkanen, M., and Sundberg, P., 2014, X-Ray
Tomography of Large Objects with Limited Measurement Heikki Haario is a professor of applied mathematics
Angle, AIP Conference Proceedings, 1581, 1800–1807, DOI: and head of department at the LUT University, Finland and
10.1063/1.4865042. a part-time research professor at the Finnish Meteorological
Victor, R., and Prodanoviü, M., 2017, Dual-Energy Medical
Institute. He studies methods for inverse problems, especially
CT in Carbonate Rocks, Digital Rocks Portal, DOI:
10.17612/P74368.URL: https://www.digitalrocksportal.org/ statistical Bayesian algorithms with applications to remote
projects/102. Accessed May 22, 2019. sensing, dynamical systems as well as industrial imaging
Victor, R.A., Prodanoviü, M., and Torres-Verdín, C., 2017, and model identi¿cation.
Monte Carlo Approach for Estimating Density and Atomic
Number from Dual-Energy Computed Tomography Images
of Carbonate Rocks, Journal of Geophysical Research: Solid
ABSTRACT
Recent advancements in computational geosciences highly convoluted pathways of Àuid diffusion through
and pore-scale imaging have made it possible to extract shale reservoirs. Additionally, permeability was computed
three-dimensional (3D) pore geometries from tight rock at 10 subvolumes of the original shale sample to investigate
samples, such as shale. This study presents a viable mean the size of permeability representative elementary volume
to determine elusive transport properties of tight reservoirs (REV). Permeability computations were conducted using
using techniques from computational Àuid dynamics (CFD). a ¿nite-volume method and under the consideration of a
We present a numerical procedure to compute hydraulic transient incompressible Newtonian Àuid. Findings reveal
tortuosity in a complex 3D pore system imaged by focused that a shale sample of 7.44 ȝm3 is insuf¿cient to reach
ion beam scanning electron microscope (FIB-SEM). The permeability REV. Lastly, we carried out mesh density
effectiveness of the procedure is demonstrated on a pore sensitivities, for the ¿rst time on a shale sample, to examine
network that was extracted from a shale reservoir in the the reliability of using 3D pore networks extracted from
United States. Results suggest log-normal distribution of FIB-SEM stacks.
tortuosity with an average of 1.8, which illustrates the
Manuscript received by the Editor November 5, 2018; revised manuscript received February 4, 2019; manuscript accepted February 13 2019.
1
Devon Energy, 333 W Sheridan Ave, Oklahoma City, OK 73102; mouin.almasoodi@dvn.com.
2
Zul¿quar Reza, University of Oklahoma, Mewbourne School of Petroleum and Geological Engineering (MPGE), 100 Boyd St,
Norman, OK 73069; zul¿quar.reza@ou.edu.
paths cannot be measured experimentally. Hence, several thin porous coating covering a thicker ¿brous base layer. The
numerical studies were conducted to compute tortuosity. The morphology of the pore structure within the coating layer
most recent work was done by Saomoto and Katagiri (2015) has a signi¿cant impact on the Àow pattern and properties of
in which a 2D theoretical porous medium was constructed the ink inside the paper medium. The authors illustrate that
to compare electric tortuosity to hydraulic tortuosity using upon arrival of an ink droplet, Àuid starts invading the pore
¿nite-element analysis. Saomoto and Katagiri (2015) found space and thus displacing the air that was initially ¿lling the
that, on average, hydraulic tortuosity is 15% greater than pore space. Essentially, this process is similar to imbibition
the electric tortuosity. Similarly, the concept of tortuosity in oil and gas reservoirs.
has gained prominence in the lithium-ion battery industry as The full understanding of imbibition requires
battery manufactures are primarily interested in reducing the characterization of the pore morphology, wettability, and
electric tortuosity of porous electrodes to improve battery effect of capillarity. Furthermore, Silin et al. (2011) used
ef¿ciency and reduce waste heat (Ebner et al., 2014; Delattre 3D images generated using synchrotron-based X-ray
et al., 2018). microtomography of pore space as input for the maximal
Recent advancements in pore-scale imaging, processing, inscribed spheres (MIS) method to predict two-phase Àuid
and parallel computing have made it possible to simulate distribution in capillary equilibrium. The study showed
Àuid Àow through realistic pore geometries and investigate agreement between the computed Àuid distribution in the
pore-scale physics, such as multiphase behavior (Zuo et al., pores and experimental data. Even though the MIS method
2017), relative permeability (Zhang, 2017), and capillary is incapable of capturing the morphological detail of the
action (Ruspini et al., 2017). Computational rock physics, pore geometry, the study suggests that microcomputed
known as digital rock physics, uses numerical techniques of tomography (micro-CT) along with MIS is a viable approach
various physical phenomena to extract transport properties to study the pore-scale mechanisms of CO2 injection into
(e.g., electrical conductivity and permeability), and to gain an aquifer. Furthermore, Blunt et al., (2013) described in
insights into Àow dynamics within the reservoir. These detail the imaging of rock pore space from the nanometer
computational techniques are directly applied to the pore scale and upwards. They provide three examples to illustrate
geometry of the rock, which eliminates the need for using the range of scienti¿c problem that can be addressed. The
theoretical pore networks to represent the pore geometry. ¿rst example was the dispersion of highly heterogenous
In addition, recent developments in pore-scale imaging carbonate rocks; the second example, imaging of super
allow the detailed and more accurate mapping of 3D pore critical CO2 to illustrate the possibility of capillary trapping
geometries of rock samples. Conversely, traditional rock in geological carbon storage; and the third example focused
physics involves either empirical relationships based on the computation of relative permeability for mixed-
on experimental data, or theoretical models based on wet carbonates and discussed implications for oil¿eld
idealized microstructures (Andrä et al., 2013). Despite the waterÀood recovery. The authors concluded that pore-scale
valuable insights that conventional rock physics provides, modeling has the potential to transform our understanding of
it is challenged to capture the inherent complexity of multiphase Àow processes, improve contaminants removal,
the pore morphology needed to accurately characterize and safe carbon storage. More recently, Zapata and Sakhaee-
tight reservoirs. Several studies have demonstrated the Pour (2016) have attempted to characterize the pore space
utility of coupling the imaging technology with the well- of shale formations by using data from mercury-intrusion
established physics of Àuid dynamics (Piri and Blunt, 2005; and nitrogen-adsorption experiments. The authors were
Madonna et al., 2013; Raeini et al., 2014; Berg et al., 2016; able to distinguish pore bodies, from pore throats. In order
Mohammadmoradi and Kantzas, 2016). However, despite to account for the restrictions within the connected path of
the extensive research, there is a research gap in modeling the pore space, Zapata and Sakhaee-Pour (2016) needed to
3D porous networks extracted from tight reservoirs, such implement an acyclic pore model. Alternatively, the rock
as shale. Similarly, the impact of mesh re¿nement on sample could have been imaged to extract the pore geometry
transport properties, such as permeability, has not been which then can be directly used in the Àow simulation of the
fully analyzed until now. This is mainly due to the large mercury-injection experiment.
mesh sizes that could easily exceed 30 million cells for a Regarding the computational aspect of pore-scale
sample size of 1 ȝm3. modeling, signi¿cant efforts have been invested to verify and
Besides the oil and gas industry, there are many other validate codes of computational Àuid dynamics (CFD). Most
industries interested in computational pore-scale modeling. of the efforts are emerging from mechanical and aerospace
For instance, Aslannejad and Hassanizadeh (2017) illustrated engineering. For instance, Oberkampf and Trucano (2002)
that paper used in the printing industry usually contains a presented an extensive literature review on fundamental
topics in CFD, such as solution versus code veri¿cation, z axes, respectively. The sample porosity was computed to
model versus solution validation, and sources of error and be 4% by taking the ratio of the pore volume to the bulk
uncertainty. Despite the continuous development of new volume of the rock sample. The sample was imaged and
codes, and faster solution algorithms, legacy codes remain processed at the laboratories of the University of Oklahoma.
crucial to today’s simulation due to the extensive validation The sample was cut and mounted to an aluminum stub using
process and ¿nancial investment in such codes. In this work, a conductive carbon paste. The sample was then coated with
we address some of the gaps in recent literature, such as how Au/Pd in a Denton Vacuum Desk V sputtering system to
to develop a method to compute rock hydraulic tortuosity provide a conductive coating on the sample surface. Next,
from FIB-SEM images, explore permeability dependence the sample was prepared for 3D SEM imaging using a dual-
on mesh density, and investigate whether FIB-SEM images beam FIB-SEM system. A platinum pad was deposited on
provide enough resolution and scale to compute shale the edge of the sample to minimize curtaining effects. Data
permeability. acquisition was performed via FEI Slice N View software.
In order to use the 3D pore system for Àow simulation,
METHODOLOGY a mesh needs to be created. Mesh generation is the most
important and laborious step in the modeling process
Sample Description and Preparation (Power et al., 2003; Dennis et al., 2005; Zadeh et al., 2014).
Both connected and isolated 3D pore networks were Good quality mesh must ensure accurate representation of
extracted from a shale sample using FIB-SEM technology. the rock pore structure and allow the use of higher-order
Isolated pores were ignored since they are not connected to the numerical schemes. The meshing process started by creating
Àow domain. However, tight reservoirs are often stimulated a background mesh containing hexahedral cells. The 3D
by hydraulic fracturing, which could connect isolated pores pore network was embedded inside the background mesh.
and increase permeability. The sample minimum pore throat Cells were re¿ned near the surface of the pore geometry to
was approximately 30 nm. Figure 1 presents the isometric accurately capture the sample morphology. Finally, cells
projections of the 3D pore geometry that was used for the outside the sample geometry were removed. Figure 2 shows
tortuosity and permeability simulations; it also shows the the FIB-SEM model of the pore space (blue) prior to the
position of the inlet and outlet. The dimensions of the pore meshing process and the ¿nal mesh (gray).
geometry are 2.36, 1.89, and 1.67 ȝm along the x, y, and
Fig. 2—The imaged pore geometry is shown in blue and the mesh is shown in gray. A zoom-in view shows the cell re¿nement along one of the pore
throats. The sample has dimensions of 2.36, 1.89, and 1.67 ȝm along the x, y, and z axes, respectively.
Where P is the pressure, ȝ is the dynamic viscosity, Numerical Approach and Simulation Setup
ĺ
and V is the velocity vector. Equation 2 is a representation Numerical implementation and simulation assumptions
of Newton’s second law and it is known as the Navier- are discussed in this section. The ¿nite-volume method
Stokes equation. The right-hand side has two terms that was used to solve the mathematical model discussed in the
represent the pressure gradient and diffusion terms. The previous section. The goal of the ¿nite-volume method is
left-hand side has two acceleration terms, which are linear to represent and evaluate partial differential equations in
local acceleration and nonlinear advective acceleration. The the form of algebraic equations. Besides its computational
nonlinearity due to the advective term will be addressed via ef¿ciency, it is capable of handling unstructured grids,
an iterative numerical framework. which are needed to capture the details of the shale sample
After solving the velocity and pressure ¿elds pore morphology. Essentially, the Àow domain was
numerically, streamlines within the pore system were divided into subdomains called control volumes and then
computed by integrating the equation of motion for massless the conservation equations (mass and momentum) were
particle shown below: integrated over each control volume. Therefore, the ¿nite-
volume technique inherently satis¿es the conservation
(3) property and there is no need to impose conservativeness to
the formulation. The volume integrals were then converted
to surface integrals with the aid of the divergence theorem Figure 4 presents Àow streamlines colored by their
to arrive at the discretized form. PISO (pressure-implicit corresponding tortuosity value. Results suggest that the
with splitting of operators) was used as a procedure to solve sample has a tortuosity range of 1.1 to 3.9, and a mean of
the pressure-velocity coupling; the PISO technique was 1.7. Table 1 shows the summary statistics of the streamline
originally proposed by Issa (1986). Finally, the results were tortuosity. Figure 5 depicts a ¿ltered version that shows
presented in terms of algebraic quantities that were solved tortuosity values 2 to illustrate the highly tortuous paths
iteratively. within the tight shale sample.
The ¿nite-volume method was implemented via
openFOAM® toolbox. Details regarding the required
hardware and software along with code are provided in the
Code Availability section. The open-source code offers a
variety of schemes for discretizing the gradient, convection,
and diffusion terms. Depending on the mesh quality, the
choice of the scheme is determined. For instance, when
the mesh has good quality cells, such as full hexahedrons,
the gradient can be discretized by applying Gauss’s linear
scheme which is a central differencing scheme. Otherwise,
the gradient limiter scheme needs to be applied to ensure
that the extrapolated gradient falls within the minimum and
maximum of neighboring points.
Regarding the pore Àuid physical properties, water was
selected as the pore Àuid during the simulation; it has density
of 1,000 kg.m-3 and dynamic viscosity of 0.001 Pa.s. The
slip boundary condition was adopted at the solid boundaries
and the Àow domain was initialized by applying a pressure
gradient between the inlet and outlet of 100 psi.
RESULTS
Fig. 3—Overall workÀow outlining the steps for computing tortuosity
from pore-scale images based on streamline tracers.
Computation of Tortuosity from Streamlines
Streamlines are paths that imaginary particles would
take if they were released into the Àow stream. Streamlines Tortuosity is often estimated from empirical models,
carry signi¿cant information about the velocity ¿eld especially in the absence of laboratory experiments and
direction and magnitude at each point. We de¿ne tortuosity simulation results. One of the most commonly used empirical
a as the ratio of streamline length to the straightline distance models is the model proposed by Comiti and Renaud (1989).
between its two ends as shown in Eq. 4, where Ts is the The model represents tortuosity as logarithmic function of
streamline tortuosity, Ls is streamline tortuous length, and L porosity as show in Eq. 5.
is the streamline straightline distance between its two ends
(5)
(4)
Where is porosity and P is an empirical parameter
After solving for the velocity ¿eld that was derived in that is usually estimated by simulations, or experiments.
the mathematical model section, the equation of motion for The model satis¿es T = 1 when, = 100%, and satis¿es T
a massless particle was integrated to compute streamlines = when = 0%. Streamline-based simulation can be used
within the sample pore body. The cumulative distance to ¿ne-tune empirical models. For instance, since the shale
between the particles of each streamline was computed sample has 4% porosity and average tortuosity of 1.8, the P
to determine the tortuous length of the streamline. The value is estimated to be 0.25. Obviously, more samples need
straightline distance was also calculated for each streamline to be simulated to inform the appropriate P value for shale
by simply using the distance equation between two points. reservoirs. Additionally, our simulated tortuosity results are
Figure 3 summarizes a workÀow that shows the necessary well aligned with ¿ndings derived from the Pisani (2011)
steps to calculate tortuosity based on streamline tracers. model for spherical particles shown in Eq. 6. Based on
Pisani’s model, tortuosity is calculated to be 1.6.
Fig. 4—Transparent 3D pore geometry reveals streamlines colored by their corresponding tortuosity.
(6)
Computation of Permeability with Varying Mesh Sizes Table 2—The Number of Cells in Each Direction Along With the Final
Mesh Size After Re¿nement and Cell Removal Processes
It is imperative to accurately estimate reservoir
permeability to identify the optimal number of horizontal x, y, and z
wells and mitigate economic risks due to overcapitalization.
In this section, permeability is computed numerically
based on transient incompressible simulation, and a
mesh convergence study is conducted to investigate the
dependence of permeability on the mesh size.
Five transient simulations were run with varying mesh
sizes, as shown in Table 2. The meshes were designed such
that the minimum pore throat has at least 10 cells from wall
to wall. Simulation cases were terminated once the velocity step to avoid convergence problems. Finally, permeabilities
had stabilized. The average inlet and outlet pressures were were plotted against simulation time as shown in Fig. 7.
calculated by performing surface integration at the inlet and Results suggest lower permeability values for ¿ner meshes
outlet faces for each time step. The average velocity was during the transient period. Eventually, the ¿ve mesh sizes
obtained by conducting volume integration over the control reached comparable steady-state permeability of 480 nD
volumes that constituted the pore geometry. Consequently, after about 20 ns.
permeability values were calculated via Darcy formulation at Numerical results need to be calibrated with experimental
each time step. Similar discretization schemes were applied data once a consistent methodology for measuring shale
to all cases. Essentially, the time schemes were discretized permeability becomes available. Several studies, such
using the Crank and Nicolson (1996) method, which is a as Sondergeld et al. (2010) and Tinni et al. (2012), have
second-order accurate and implicit scheme. The gradient pointed out that standardized methods for measuring shale
calculation was performed using a least-squares approach. permeability do not exist. Similarly, Passey et al. (2010)
Moreover, the Courant number (convergence measure) conducted a comparative study on permeability measured
was calculated independently for each cell because it depends by different laboratories using crushed rock samples in
on the cell size, time step, and velocity. The Courant number which each laboratory received sample splits from the same
was maintained at < 1 by automatically adjusting the time depth interval. Passey et al. (2010) found that permeability
values reported by different laboratories varied by two to the full pore volume, as shown in Fig. 8. The subvolume
three orders of magnitude. However, more recently, there inlet and outlet pressures were calculated by performing
have been several encouraging developments to measure surface integration. The corresponding Àow velocity was
shale permeability experimentally. For instance, Krumm and obtained by performing volume integration over the volume
Howard (2017) developed a workÀow for measuring shale of the sample. Now that we know the subvolume pressure
permeability in the presence of microcracks. They have drop and Àow velocity, Darcy’s law is used to compute
integrated micro-CT scanning, and NMR technology with permeability. These steps were repeated for each sector to
a standard steady-state permeability rig for measurement obtain permeability. Figure 9 presents a bar chart with the
of the hydrocarbon Àow. Krumm and Howard (2017) claim permeability value for each volume increment. Furthermore,
that a full permeability test requires 7 to 14 days to complete. we have applied uniform volume reductions of 25, 50, and
It is worth mentioning that permeability results presented in 75% of the original volume and then evaluated permeability.
this study need to be veri¿ed with similar experimental tests. Permeability values for reductions of 0, 25, 50, and 75%
were 0.480, 0.146, 4.808, and 4.744 mD, respectively.
Analysis of Permeability Representative Elementary Looking closely at the results, two permeability groups
Volume (REV) can be distinguished (S1–S4) and (S5–S10). The average
Several statistical techniques have been attempted to permeability for each group is shown by blue horizontal
characterize the microstructure of porous media at various lines in Fig. 9. Results suggest that a shale sample of volume
length scales, such as the recent work of Adeleye and Akanji 7.44 ȝm3 is not enough to identify the REV of permeability.
(2017). In this work, we implement a computational approach Findings from recent studies, such as Al-Raoush and
to assess the dependence of permeability on the pore-volume Papadopoulos (2010) and Mostaghimi et al. (2012), have
size. The ¿nest mesh size was used to compute permeability revealed that the REV is property speci¿c, which makes the
at different volume increments. The pore volume was task of upscaling learnings from one scale to the next even
reduced by 10% successively to create 10 subvolumes. more challenging.
S-1 represents the smallest volume and S-10 represents
DISCUSSION
identify the REV of permeability. Results suggest that a COMPUTER CODE AVAILABILITY
shale sample of volume of 2.36×1.89×1.67 ȝm (7.44 ȝm3)
is within the domain of microscopic heterogeneity. In other Shale sample pore geometry will be made available
words, the observation scale is smaller than the representative on request. Tortuosity data along with the code were made
elementary volume. Emerging plasma FIB-SEM technology available at https://github.com/MouinAlmasoodi/Tortuosity.
offers the promise of larger volumes at shorter imaging time The repository also includes R script that was used for
that could be up to 50 times faster compared to conventional modeling the probability density function. Simulation cases
FIB-SEM technology (Burnett et al., 2016). were carried out on a server with 64 Xenon processors,
speed of 3.3 GHz, and 128 GB of memory.
CONCLUSIONS
from pore-scale imaging of imbibition, Advances in Water Krumm, R.L., and Howard, J.J., 2017, WorkÀow for Measurement
Resources, 90, 24–35. DOI: 1016/j.advwatres.2016.01.010. of Steady-State-to-Liquid Permeability in Nanodarcy
Blunt, M.J., Bijeljic, B., Dong, H., Gharbi, O., Iglauer, S., Reservoir Rock, Paper SCA2017-038 presented at the
Mostaghimi, P., Paluszny, A., and Pentland, C., 2013, Pore- International Symposium of the Society of Core Analysts,
Scale Imaging and Modelling, Advances in Water Resources, Vienna, Austria, 27 August–1 September 2017.
51, 197–216. DOI: 10.1016/j.advwatres.2012.03.003. Lee, Y., Barthel, S.D., Dáotko, P., Moosavi, S.M., Hess, K., and
Burnett, T.L., Kelley, R., Winiarski, B., Contreras, L., Daly, M., Smit, B., 2017, Quantifying Similarity of Pore-Geometry in
Gholinia, A., Burke, M.G., and Withers, P.J., 2016, Large Nanoporous Materials, Nature Communications, 8, 15396.
Volume Serial Section Tomography by Xe Plasma FIB Dual DOI: 10.1038/ncomms15396.
Beam Microscopy, Ultramicroscopy, 161, 119–129. DOI: Madonna, C., Quintal, B., Frehner, M., Almqvist, B.S., Tisato,
10.1016/j.ultramic.2015.11.001. N., Pistone, M., Marone, F., and Saenger, E.H., 2013,
Carman, P.C., 1937, Fluid Flow Through Granular Beds, Synchrotron-Based X-ray Tomographic Microscopy for Rock
Transactions, American Institute of Chemical Engineers Physics Investigations, Geophysics, 78(1), D53–D64. DOI:
(AIChE), 15, 150–166. Reprinted in 1997 in Chemical 10.1190/geo2012-0113.1.
Engineering Research and Design, 75, Supplement, S32–S48. Mohammadmoradi, P., and Kantzas, A., 2016, Pore-Scale
DOI: 1016/S0263-8762(97)80003-2. Permeability Calculation Using CFD and DSMC Techniques,
Comiti, J., and Renaud, M., 1989, A New Model for Determining Journal of Petroleum Science and Engineering, 146, 515–525.
Mean Structure Parameters of Fixed Beds From Pressure DOI: 10.1016/j.petrol.2016.07.010.
Drop Measurements: Application to Beds Packed With Mostaghimi, P., Bijeljic, B., and Blunt, M., 2012, Simulation of Flow
Parallelepipedal Particles, Chemical Engineering Science, and Dispersion on Pore-Space Images, Paper SPE-135261,
44(7), 1539–1545. DOI: 10.1016/0009-2509(89)80031-4. SPE Journal, 17(4), 1,131–1,141. DOI: 10.2118/135261-PA.
Crank, J., and Nicolson, P., 1996, A Practical Method for Numerical Oberkampf, W.L., and Trucano, T.G., 2002, Veri¿cation and
Evaluation of Solutions of Partial Differential Equations Validation in Computational Fluid Dynamics, Progress in
of the Heat-Conduction Type, Advances in Computational Aerospace Sciences, 38(3), 209–272. DOI: 10.1016/S0376-
Mathematics, 6(1), 207–226. DOI: 10.1007/BF02127704. 0421(02)00005-2.
Delattre, B., Amin, R., Sander, J., De Coninck, J., Tomsia, A.P., Passey, Q.R., Bohacs, K.M., Esch, W.L., Klimentidis, R., and
and Chiang, Y.-M., 2018, Impact of Pore Tortuosity on Sinha. S., 2010, From Oil-Prone Source Rock to Gas-
Electrode Kinetics in Lithium Battery Electrodes: Study Producing Shale Reservoir-Geologic and Petrophysical
in Directionally Freeze-Cast LiNi0.8Co0.15Al0.05O2 (NCA), Characterization of Unconventional Shale Gas Reservoirs,
Journal of the Electrochemical Society, 165(2), A388–A395. Paper SPE-131350 presented at the International Oil and Gas
DOI: 10.1149/2.1321802jes. Conference and Exhibition in China, Beijing, China, 8–10
Dennis, J., Fournier, A., Spotz, W.F., St-Cyr, A., Taylor, M.A., June. DOI: 10.2118/131350-MS.
Thomas, S.J., and Tufo, H., 2005, High-Resolution Mesh Piri, M., and Blunt, M.J., 2005, Three-Dimensional Mixed-Wet
Convergence Properties and Parallel Ef¿ciency of a Spectral Random Pore-Scale Network Modeling of Two- and Three-
Element Atmospheric Dynamical Core, The International Phase Flow in Porous Media. II. Results, Physical Review E,
Journal of High Performance Computing Applications, 19(3), 71(2, part 2), 026302. DOI: 10.1103/PhysRevE.71.026302.
225–235. DOI: 10.1177/1094342005056108. Pisani, L., 2011, Simple Expression for the Tortuosity of Porous
Ebner, M., Chung, D.-W., García, R.E., and Wood, V., 2014, Media, Transport in Porous Media, 88(2), 193–203. DOI:
Tortuosity Anisotropy in Lithium-Ion Battery Electrodes, 10.1007/s11242-011-9734-9.
Advanced Energy Materials, 4(5), 1301278. DOI: 10.1002/ Power, C., Navarro, J.F., Jenkins, A., Frenk, C.S., White, S.D.,
aenm.201301278. Springel, V., Stadel, J., and Quinn, T., 2003, The Inner
EIA, 2018, Investment in Tight Oil, Oil Sands, and Deepwater Structure of ȁCDM Haloes—I. A Numerical Convergence
Drives Long-Term Oil Production Growth, Today in Energy, Study, Monthly Notice of the Royal Astronomical Society,
March 9, U.S. Energy Information Administration. https:// 338(1), 14–34. DOI: 10.1046/j.1365-8711.2003.05925.x.
www.eia.gov/todayinenergy/detail.php?id=35272#. Accessed Raeini, A.Q., Blunt, M.J., and Bijeljic, B., 2014, Direct Simulations
May 9, 2019. of Two-Phase Flow on Micro-CT Images of Porous Media and
Gommes, C.J., Bons, A.-J., Blacher, S., Dunsmuir, J.H., and Tsou, Upscaling of Pore-Scale Forces, Advances in Water Resources,
A.H., 2009, Practical Methods for Measuring the Tortuosity 74, 116–126. DOI: 10.1016/j.advwatres.2014.08.012.
of Porous Materials From Binary or Gray-Tone Tomographic Ruspini, L.C., Farokhpoor, R., and Øren, P.E., 2017, Pore-
Reconstructions, AIChE Journal, 55(8), 2000–2012. DOI: Scale Modeling of Capillary Trapping in Water-Wet Porous
10.1102/aic.11812. Media: A New Cooperative Pore-Body Filling Model,
Issa, R.I., 1986. Solution of the Implicitly Discretised Fluid Flow Advances in Water Resources, 108, 1–14. DOI: 10.1016/j.
Equations by Operator-Splitting, Journal of Computational advwatres.2017.07.008.
Physics, 62(1), 40–65. DOI: 10.1016/0021-9991(86)90099-9. Saomoto, H., and Katagiri, J., 2015, Direct Comparison of
Kozeny, J., 1927, Uber Kapillare Leitung der Wasser in Boden, Hydraulic Tortuosity and Electric Tortuosity Based on Finite
Royal Academy of Science, Vienna, Proceedings Class I, 136, Element Analysis, Theoretical and Applied Mechanics Letters,
271–306. 5(5), 177–180. DOI: 10.1016/j.taml.2015.07.001.
Silin, D., Tomutsa, L., Benson, S.M., and Patzek, T.W., 2011, Zul¿quar Reza is an associate
Microtomography and Pore-Scale Modeling of Two-Phase professor in the Mewbourne School of
Fluid Distribution, Transport in Porous Media, 86(2), 495– Petroleum and Geological Engineering
515. DOI: 10.1007/s11242-010-9636-2. at the University of Oklahoma. He holds
Sondergeld C.H., Newsham K.E., Comisky J.T., Rice M.C., and
a PhD degree in petroleum engineering
Rai, C.S., 2010, Petrophysical Considerations in Evaluating
and Producing Shale Gas Resources, Paper SPE-131768 from the University of Alberta. Reza can
presented at the SPE Unconventional Gas Conference, claim more than 20 years of experience
Pittsburgh, Pennsylvania, USA, 23–25 February. DOI: in the oil and gas industry, research, and
10.2118/131768-MS. academia including Schlumberger and Chevron among
Spearing, M., and Matthews, G.P., 1991, Modelling Characteristic others, and a stint as ECLIPSE reservoir-simulator developer.
Properties of Sandstones, Transport in Porous Media, 6(1), His current research interests include conventional-and
71–90. DOI: 10.1007/BF00136822. unconventional-reservoir engineering; enhanced oil and
Tinni, A., Fathi, E., Agarwal, R., Sondergeld, C.H., Akkutlu, I.Y., gas recovery; advanced reservoir simulation and simulator
and Rai, C.S., 2012, Shale Permeability Measurements on
development; multiscale, multiphysics modeling; pore-scale
Plugs and Crushed Samples, Paper SPE-162235 presented
modeling; data analytics; ¿eld-development planning; waste-
at the SPE Canadian Unconventional Resources Conference,
Calgary, Alberta, Canada, 30 October–1 November. DOI: water disposal; and uncertainty and risk management. Reza
10.2118/162235-MS. contributed to a monograph on production-data integration.
Zadeh, S.N., Komeili, M., and Paraschivoiu, M., 2014, Mesh
Convergence Study for 2-D Straight-Blade Vertical Axis Wind
Turbine Simulations and Estimation for 3-D Simulations,
Transactions of the Canadian Society of Mechanical
Engineers, 38(4), 487–504. DOI: 10.1139/tcsme-2014-0032.
Zapata, Y., and Sakhaee-Pour, A., 2016, Modeling Adsorption–
Desorption Hysteresis in Shales: Acyclic Pore Model, Fuel,
181, 557–565. DOI: 10.1016/j.fuel.2016.05.002.
Zhang, M., 2017, Pore-Scale Modelling of Relative Permeability
of Cementitious Materials Using X-Ray Computed
Microtomography Images, Cement and Concrete Research,
95, 18–29. DOI: 101.1016/j.cemconres.2017.02.005.
Zuo, L., Ajo-Franklin, J.B., Voltolini, M., Geller, J.T., and
Benson, S.M., 2017, Pore-Scale Multiphase Flow Modeling
and Imaging of CO2 Exsolution in Sandstone, Journal of
Petroleum Science and Engineering, 155, 63–77. DOI:
10.1016/j.petrol.2016.10.011.
ABSTRACT
Nuclear magnetic resonance (NMR) offers a powerful like components in shale formations (kerogen, bitumen,
toolbox for petrophysical analysis of reservoir rocks. structural water). The longitudinal T1 relaxation time is
Laboratory measurements are often performed at low insensitive to internal gradients, but slow to measure with
frequency (2 MHz) to provide consistent spin physics to the traditional inversion-recovery pulse sequence. Here, a
downhole logging tools. The popular single-shot Carr- modi¿ed fast double-shot T1 pulse sequence is applied to
Purcell-Meiboom-Gill (CPMG) sequence for measuring the conventional rock formations, allowing these samples to
transverse T2 relaxation time is sensitive to local magnetic be studied reliably at high magnetic ¿elds. Representative
¿eld inhomogeneities (internal gradients) that scale with porosity and permeability values are recovered for a
resonance frequency. For T2 analysis of conventional selection of brine-saturated sandstone and carbonate
formations, low-¿eld magnets are considered mandatory. core plugs. This double-shot T1 measurement is readily
Recently, there has been renewed interest in operating at implementable on commercial NMR hardware appropriate
higher frequency ( 20 MHz) to provide sensitivity to solid- for laboratory or rigsite deployment.
Manuscript received by the Editor November 21, 2018; revised manuscript received April 4, 2019; manuscript accepted April 5, 2019.
1
Schlumberger Cambridge Research, High Cross, Madingley Road, Cambridge CB3 0EL, UK; Tel: +44 1223 325426;
E-mail: jmitchell16@slb.com.
2
Schlumberger Houston Formation Evaluation Center, Sugar Land, TX, USA; E-mail: avalori@slb.com.
1998). Diffusion of liquids through these gradients enhances (Washburn et al., 2008). Laboratory measurements of T1 are
the decay of coherent transverse magnetization, as measured usually based on slow pulse sequences, such as inversion
in the CPMG experiment, preventing quanti¿cation of recovery (IR), requiring multiple separate acquisitions to
liquid volumes (Straley et al., 1997). Consequently, most cover a range of longitudinal recovery times (Vold et al.,
laboratories use low-¿eld NMR instruments (typically 1968). Several fast single-shot T1 sequences have been
around 2 MHz) when measuring the diffusion-sensitive T2 proposed in the literature and reviewed elsewhere (Kingsley,
relaxation time. 1999). These methods suffer from a convolution of
Interest in shales has restored the old paradigm that parameters, with the ¿nal magnetization (signal amplitude)
“higher ¿elds are better” (Mitchell et al., 2013; Washburn, dependent on both the number of resonant nuclei and the
2014). There is a continuing trend in medical imaging and relaxation time. Recently, a double-shot (DS) T1 pulse
chemical spectroscopy to use stronger magnetic ¿elds for sequence was introduced to overcome these limitations
improved spatial resolution and chemical shift separation. (Chandrasekera et al., 2008). Although slower than a single-
In laboratory studies of shales, higher ¿elds improve the shot CPMG acquisition (by approximately a factor of two
contrast of components, such as kerogen, bitumen, and for a given number of repeat scans), the DS T1 measurement
structural water in T1-T2 correlation plots. The traditional is still an order of magnitude faster than the classic IR
assumption that the rock matrix is invisible to NMR no experiment. The DS pulse sequence was originally designed
longer applies as organic kerogen is detected with solid- for use in the second dimension of multidimensional high-
state pulse sequences (Jia et al., 2018). It is important to note ¿eld acquisitions, such as the T2-T1-į correlation experiment
that, although strong internal gradients are present in shales, (where į is the chemical shift) or the T1-T1 exchange
the geometric restriction imposed by the nanometric pores experiment. The DS pulse sequence has been implemented
on molecular motion minimizes the detrimental effects of on a benchtop spectrometer and used to generate T1-ǻȤ
diffusion (Washburn, 2014). Core-analysis companies have correlations (as a measure of rock heterogeneity) and as a
been offering routine measurements of unconventional shale rapid method of imaging with T1 contrast (Mitchell, 2014).
reservoir samples at f0 = 20 MHz (B0 = 0.5 T) since 2013 and The original experiment used magnetic ¿eld gradient pulses
various interpretation schemes have been published (Fleury requiring high-performance audio ampli¿ers considered
and Romero-Sarmiento, 2016). impractical for rigsite deployment.
Recently, there has been renewed interest in formation In this paper, a simpli¿ed DS sequence is introduced
evaluation based on rigsite analysis of drill cuttings (Georgi without the requirement for a pulsed ¿eld gradient (PFG).
and Loermans, 2018) as an alternative to costly logging The measurement is demonstrated on a high-¿eld
operations in low-tier, high-volume markets. Automated instrument, B0 = 0.25 T (10.7 MHz), and used to acquire
analysis of drill cuttings has the potential to provide T1 relaxation-time distributions for a selection of water-
near-real-time answers for production decisions and saturated sandstones and carbonates. These relaxation-time
geosteering (Nikitin et al., 2017). Rigsite NMR has existed distributions are compared to classic T2 relaxation-time
as a concept since the 1980s (Nigh and Taylor, 1984), and distributions acquired on the same samples at low frequency
numerous studies have been published comparing porosity (2.4 MHz). The DS sequence is readily implementable
measurements from cuttings, core plugs, and downhole logs at 20 MHz (or higher frequency) and so can be used to
(Meazza et al., 1996; Lenormand and Fonta, 2007; Hübner, quantitatively measure Àuids in conventional reservoir rocks
2014; Yu and Menouar, 2015; Kesserwan et al., 2017). To on hardware optimized for shale characterization.
avoid the proliferation of expensive NMR instruments for
cuttings analysis, particularly at the rigsite, where space NMR METHOD
and cost constraints are high, it would be desirable to have
a single platform appropriate for analysis of conventional The DS pulse sequence, modi¿ed for robust
(sandstones, carbonates) and shale resources. implementation on a low-¿eld instrument, is illustrated
The need to quantify the kerogen and bitumen content in Fig. 1. The longitudinal magnetization is prepared by a
of shales as reservoir quality indictors demands a high composite ʌ/2-ʌ/2 radio frequency (RF) pulse followed by a
¿eld strength (Reeder et al., 2016), with existing B0 = 0.5 series of n small-tip-angle RF pulses (Į ~ 0.2 rad). Each Į pulse
T benchtop permanent magnet technology practical for rotates a fraction of the longitudinal magnetization onto the
laboratory or rigsite (surface) deployment for analysis of core transverse plane where it is observed as a free induction decay
or cuttings. Conventional rocks can be analyzed reliably at (FID) described by the exponential time constant T2* (Hahn,
such ¿eld strengths using T1 relaxation-time distributions, as 1950). Each FID is ¿tted with a monoexponential decay
longitudinal relaxation processes are insensitive to diffusion function to determine the time-dependent signal amplitude
ROCK SAMPLES
the T1 distributions, albeit shifted on the relaxation-time consistent, and almost identical for DO, PT, and SK. The
axis. However, even at low ¿eld, the inÀuence of internal sampling of the DS decay leads to broader distributions for
gradients on the T2 measurements manifests in some of these the ES and IN samples, compared to the IR results, but the
samples (Bra, BRb, LA) as multiple discrete peaks (Mitchell integral areas under the distributions remains consistent for
et al., 2010). each sample. The T2 distributions (2.4 MHz) for the carbonate
samples are shown in Fig. 6 with the DS results (10.7 MHz)
repeated for clarity. The two distributions are qualitatively
similar for each sample, with the T2 distribution shifted to
shorter relaxation times compared to the T1 distribution.
As these carbonates exhibit a low magnetic-susceptibilty
contrast, the T2 distributions are not modi¿ed by diffusion
through internal gradients and therefore closely resemble the
T1 distributions acquired at higher ¿eld.
(a) (b)
Fig. 5—Double-shot T1 measurement of carbonate core plugs at 10.7
MHz (solid lines). Column (a) contains the raw data used to generate
the T1 distributions in column (b). The data and distributions obtained by
the IR sequence (dashed lines) are included for comparison. The area
under each distribution has been normalized to unity.
(a) (b)
Fig. 6—CPMG T2 measurement of carbonate core plugs at 2.4 MHz
(dot-dashed lines). Column (a) contains the raw data used to generate
the T2 distributions in column (b). The data and T1 distributions obtained
by the DS sequence at 10.7 MHz (solid lines) are repeated from Fig. 5
for comparison. The area under each distribution has been normalized
to unity.
the DS sequence, the total signal intensity is determined of sandstone and carbonate core plugs studied here were
directly from the zero-time amplitude of the initial FID converted to permeability using estimated constants of
as an alternative to integrating the distribution; the two C s = 8 × 106 m2 sec-2, Cc = 2.5 × 104, and ȡ1 = 2 × 10-6 m sec-1.
estimates are equivalent. Porosity values determined by Note that surface relaxivity is frequency dependent and will
helium porosimetry and NMR are compared in Fig. 9 for increase with increasing resonance frequency. The NMR
the selection of rock plugs. In core analysis, an uncertainty permeability estimates are compared to N2 gas permeability
of ±0.5 p.u. is considered acceptable on any porosity in Fig. 10. It is usual in logging applications to adjust the
measurement (Thomas and Pugh, 1989). The NMR porosity constants depending on the formation under investigation.
estimates fall within, or very near, to these bounds. The DS Here, the use of single numeric values gave permeability
sequence is therefore considered appropriate for measuring estimates within the factor ×3 range (approximately ± half
porosity at high frequency. a decade) considered acceptable for NMR analysis. OB is
an obvious outlier with k [N2] = 1 md and k [NMR] = 18 md
due to an unusually high surface relaxivity. Nevertheless,
the measured T1LM relaxation times are considered robust for
interpretation of the formation permeability with appropriate
scaling factors.
(3) CONCLUSIONS
or a modi¿ed equation for carbonates (Allen et al., 2001) The optimum nuclear magnetic resonance (NMR)
hardware for benchtop measurements of core plugs or drill
(4) cuttings depends on the reservoir formation: conventional
rocks (sandstones, carbonates) are best analyzed with
where Cs and Cc are empirical constants and ȡ1 is the low-¿eld magnets, whereas detection of kerogen in
longitudinal surface-relaxivity parameter (Brownstein and unconventional organic shale is best achieved at a higher
Tarr, 1979). The T1LM relaxation times for the selection magnetic ¿eld strength. There is interest in deploying
Precession in Nuclear Magnetic Resonance Experiments, in Magnetic Resonance, 11(4), 243–276. DOI: 10.1002/
Physical Review, 94, 630–638. DOI: 10.1103/PhysRev.94.630. (SICI)1099-0534(1999)11:4<243::AID-CMR5>3.0.CO:2-C.
Chandrasekera, T.C., Mitchell, J., Fordham, E.J., Gladden, L.F., Kleinberg, R.L., Farooqui, S.A., and Hors¿eld, M.A., 1993a, T1/
and Johns, M.L., 2008, Rapid Encoding of T1 with Spectral T2 Ratio and Frequency Dependence of NMR Relaxation in
Resolution in n-dimensional Relaxation Correlations, Journal Porous Sedimentary Rocks, Journal of Colloid and Interface
of Magnetic Resonance, 194(1), 156–161. DOI: 10.1016/j. Science, 158(1), 195–198. DOI: 10.1006/jcis.1993.1247.
jmr.2008.06.008. Kleinberg, R.L., Straley, C., Kenyon, W.E., Akkurt, R., and
Chen, Q., Marble, A.E., Colpitts, B.G., and Balcom, B.J., 2005, The Farooqui, S.A., 1993b, Nuclear Magnetic Resonance of
Internal Magnetic Field Distribution, and Single Exponential Rocks: T1 vs. T2, Paper SPE-26470 presented at the SPE
Magnetic Resonance Free Induction Decay, in Rocks, Journal Annual Technical Conference and Exhibition, Houston,
of Magnetic Resonance, 175(2), 300–308. DOI: 10.1016/j. Texas, USA, 3–6 October. DOI: 10.2118/26470-MS.
jmr.2005.05.001. Kleinberg, R.L., and Vinegar, H.J., 1996, NMR Properties of
Davies, S., and Packer, K.J., 1990, Pore-Size Distributions from Reservoir Fluids, The Log Analyst, 37(6), 20–32.
Nuclear Magnetic Resonance Spin-Lattice Relaxation Lenormand, R., and Fonta, O., 2007, Advances in Measuring
Measurements of Fluid-Saturated Porous Solids, II. Porosity and Permeability from Drill Cuttings. Paper SPE-
Applications to Reservoir Core Samples, Journal of Applied 111286 presented at the SPE/EAGE Reservoir Characterization
Physics, 67(6), 3171–3176. DOI: 10.1063/1.345396. and Simulation Conference, Abu Dhabi, UAE, 28–31 October.
Fleury, M., and Romero-Sarmiento, M., 2016, Characterization DOI: 10.2118/111286-MS.
of Shales Using T1-T2 NMR Maps, Journal of Petroleum Meazza, O., Della Martera, M., Lyne, A., Taylor, M., and Saunders,
Science and Engineering, 137, 55–62. DOI: 10.1016/j. M.R., 1996, Porosity from Cuttings: Options and Answers,
petrol.2015.11.006. Paper SCA9606 presented at the International Symposium
Georgi, D., and Loermans, T., 2018, Can We Propel Cuttings- of the Society of Core Analysts, Montpelier, France, 8–10
Based Mud Logging Into the 21st Century? SPWLA Today September.
Newsletter, 1(2), 14–15. Meiboom, S., and Gill, D., 1958, Modi¿ed Spin-Echo Method for
Graumann, R., Barfuss, H., Fischer, H., Hentschel, D., and Measuring Nuclear Relaxation Times, Review of Scienti¿c
Oppelt, A., 1987, TOMROP: A Sequence for Determining Instruments, 29(8), 668–691. DOI: 10.1063/1.1716296.
the Longitudinal Relaxation Time T1 in Magnetic Resonance Mitchell, J., 2014, Rapid Measurements of Heterogeneity in
Tomography, Electromedica, 55, 67–72. Sandstones Using Low-Field Nuclear Magnetic Resonance,
Hahn, E.L., 1950, Nuclear Induction Due to Free Larmor Precession Journal of Magnetic Resonance, 240, 52–60. DOI: 10.1016/j.
[Abstract], Physical Review, 77(2), 297–298. DOI: 10.1103/ jmr.2014.01.006.
PhysRev.77.297.2. Mitchell, J., Chandrasekera, T.C., Johns, M.L., Gladden, L.F., and
Hübner, W., 2014, Studying the Pore Space of Cuttings by NMR Fordham, E.J., 2010, Nuclear Magnetic Resonance Relaxation
and CT. Journal of Applied Geophysics, 104, 97–105. DOI: and Diffusion in the Presence of Internal Gradients: The
10.106/j.jappgeo.2014.02.016. Effect of Magnetic Field Strength, Physical Review E, 81(2),
Hürlimann, M.D., 1998, Effective Gradients in Porous Media due 026101. DOI: 10.1103/PhysRevE.81.026101.
to Susceptibility Differences, Journal of Magnetic Resonance, Mitchell, J., Chandrasekera, T.C., Holland, D.J., Gladden, L.F.,
131(2), 232–240. DOI: 10.1006/jmre.1998.1364. and Fordham, E.J., 2013, Magnetic Resonance Imaging in
Jia, Z., Xiao, L., Chen, Z., Liao, G., Zhang, Y., Wang, Z., Liang, C., Laboratory Petrophysical Core Analysis, Physics Reports,
and Guo, L., 2018, Determining Shale Organic Porosity and 526(3), 165–225. DOI: 10.1016/j.physrep.2013.01.003.
Total Organic Carbon by Combining Spin Echo, Solid Echo Nigh, E., and Taylor, M., 1984, P-KTM: Wellsite Determination of
and Magic Echo, Microporous and Mesoporous Materials, Porosity and Permeability Using Drilled Cuttings, Canadian
269, 12–16. DOI: 10.1016/j.micromeso.2017.11.049. Well Logging Society Journal, 13, 43–58.
Kenyon, W.E., 1997, Petrophysical Principles of Applications of Nikitin, A., Krugliak, Z., and Stepanov, V., and Dolan, S., 2017,
NMR Logging, The Log Analyst, 38(2), 21-43. Automated Mud Logging System as a Cost-Ef¿cient Way
Kenyon, W.E., Day, P.I., Straley, C., and Willemsen, J.F., 1988, A of Acquiring Subsurface Data: Results of the Field Trial,
Three-Part Study of NMR Longitudinal Relaxation Properties Paper SPE-187387 presented at the SPE Annual Technical
of Water-Saturated Sandstones, Paper SPE-15643, SPE Conference and Exhibition, San Antonio, Texas, USA, 9-11
Formation Evaluation, 3(3), 622–636. DOI: 10.2118/15643- October. DOI: 10.2118/187387-MS.
PA. Reeder, S.L., Craddock, P.R., Rylander, E., Pirie, I., Lewis, R.E.,
Kesserwan, H., Alshanqaiti, E., Jin, G., and Nuic, I., 2017, Porosity Kausik, R., Kleinberg, R.L., Yang, J., and Pomerantz, A.E.,
Measurements on Drill Cuttings—Comprehensive Inputs 2016, The Reservoir Producibility Index: A Metric to Assess
to Formation Evaluation Near Real-Time While Drilling, Reservoir Quality in Tight-Oil Plays from Logs, Petrophysics,
Paper SPE-188881 presented at the Abu Dhabi International 57(2), 83–95.
Petroleum Exhibition and Conference, Abu Dhabi, UAE, 13– Straley, C., Rossini, D., Vinegar, H., Tutunjian, P., and Morriss, C.,
16 November. DOI: 10.2118/188881-MS. 1997, Core Analysis by Low-Field NMR, The Log Analyst,
Kingsley, P.B., 1999, Methods of Measuring Spin-Lattice (T1) 38(2), 84–95.
Relaxation Times: An Annotated Bibliography, Concepts Thomas, D.C., and Pugh, V.J., 1989, A Statistical Analysis of the
Accuracy and Reproducibility of Standard Core Analysis, The pulses (Į ~ 0.1 rad) to drive the spin ensemble to some
Log Analyst, 30(2), 71–77. equilibrium magnetization Meq which is a nontrivial function
Valori, A., Mitchell, J., and Fordham, E.J., 2016, Digital Filters for of M0, Į, and T1. For samples with a well-de¿ned T1 (e.g.,
Low-Field NMR, Concepts in Magnetic Resonance, 46B(4), bulk water) it is possible to deconvolve these contributions
202–220. DOI: 10.1002/cmr.b.21346.
to Meq and estimate the relaxation time. For complicated
Valori, A., Hursan, G., and Ma, S.M., 2017, Laboratory and
Downhole Wettability from T1/T2 Ratio, Petrophysics, 58(4), samples (Àuid-saturated rocks), this deconvolution is
352–356. impractical. The DS pulse sequence introduces a preparative
Vold, R., Waugh, J., Klein, M., and Phelps, D., 1968, Measurement z-store pulse to TOMROP, allowing the initial magnetization
of Spin Relaxation in Complex Systems, Journal of Chemical to be alternated between -M0 and zero on successive scans.
Physics, 48, 3831–3832. DOI: 10.1063/1.1669699. Summing two scans cancels the Meq term and encodes the
Washburn, K.E., 2014, Relaxation Mechanisms in Shales, T1 relaxation time as an exponential decay. The measured
Concepts in Magnetic Resonance, 43A(3), 57–78. DOI: magnetization has the form
10.1002/cmr.a.21302.
Washburn, K.E., Eccles, C.D., and Callaghan, P.T., 2008, The
(A2)
Dependence on Magnetic Field Strength of Correlated Internal
Gradient Relaxation Time Distributions in Heterogeneous
Materials, Journal of Magnetic Resonance, 194(1), 33–40. where sin Į is a constant scaling factor on the signal amplitude
DOI: 10.1016/j.jmr.2008.05.025. and IJ1 is the longitudinal recovery delay between consecutive
Wilson, J.D., 1992, Statistical Approach to the Solution of First- RF pulses. Using the DS pulse sequence, the T1 relaxation
Kind Integral Equations Arising in the Study of Materials and time is measured in a couple of minutes (Mitchell, 2014).
Their Properties, Journal of Materials Science, 27(14), 3911– After each Į pulse the transverse magnetization decays with
3924. DOI: 10.1007/BF00545476. the exponential time constant T2* which is determined by a
Yu, Y., and Menouar, H., 2015, An Experimental Method to combination of the true transverse relaxation time T2, ¿eld
Measure the Porosity from Cuttings: Evaluation and Error inhomogeneities of the NMR instrument ǻB0, and local ¿eld
Analysis, Paper SPE-173591 presented at the SPE Production
inhomogeneities introduced by ǻȤ such that
and Operations Symposium, Oklahoma City, Oklahoma,
USA, 1–5 March. DOI: 10.2118/173591-MS.
(A3)
APPENDIX – NMR EXPERIMENTS
where it is reasonable to assume, for rocks measured at
The classic IR sequence for measuring T1 relaxation modest B0 > 100 mT, the term ȖǻȤB0 will dominate and
time stores the longitudinal magnetization on the negative the signal will be described by a monoexponential decay
z-axis with a ʌ tip angle radio frequency (RF) pulse. After a function (Chen et al., 2005). At higher magnetic ¿elds the
recovery time IJ1, the magnetization is interrogated with a ʌ/2 T2* relaxation time becomes shorter. It is necessary to wait
RF pulse and a free induction decay (FID) measured (Hahn, for the transverse magnetization to approach zero before
1950). The transverse magnetization (signal amplitude) M(t) applying the next Į pulse to avoid residual signal or unwanted
immediately after the ʌ/2 RF pulse is a function of IJ1 over coherences (echoes) in the acquisition window. In previous
successive experiments and has the form implementations of the DS sequence (Chandrasekera et al.,
2008; Mitchell 2014), pulsed ¿eld “homospoil” gradients
(A1) were applied to rapidly dephase the coherent transverse
magnetization. To avoid the use of pulsed gradients and
It is necessary to wait a recycle delay of tRD = 5×T1 between enable robust implementation on a benchtop spectrometer,
experiments to ensure quantitative signal amplitudes are the DS sequence was modi¿ed in the following ways:
recorded. Consequently, the entire IR measurement can have x The initial magnetization condition was established
a duration of an hour or more. The acquisition time can be with a composite ʌ/2-ʌ/2 RF pulse rather than a spin
reduced slightly using saturation recovery (the magnetization echo.
is prepared with multiple ʌ/2 pulses and recovers from zero x The FID generated by the composite ʌ/2-ʌ/2
rather than -M0) which removes the requirement for a long initialization pulse was acquired to determine M0
tRD (Vold et al., 1968). However, exploring IJ1 delays up to 10 explicitly.
sec over multiple acquisitions is still time-consuming. x The number n of Į pulses is restricted in each scan to
An equivalent single-shot experiment is the “T1 by ensure IJ1 » 10×T2*.
multiple read-out pulses” (TOMROP) sequence (Grauman x The delay IJ0 between the preparative composite
et al., 1987). TOMROP uses a series of n small-tip-angle ʌ/2-ʌ/2 pulse and the ¿rst Į pulse is incremented over
several (e.g., two or three) repeat acquisitions and the UK] controlled by a DRX-TCP spectrometer [Oxford
results interleaved. Instruments, UK]. The magnet was ¿tted with a 53-mm inner
x An extended eight-step RF phase cycle is diameter RF probe including an active damping feedback
recommended to eliminate residue magnetization, preampli¿er [MRF Innovations, UK] to improve the signal/
see Table A1. noise ratio (SNR). RF pulses were provided by a 500 W
ampli¿er [Tomco Technologies, Australia] and a ʌ/2 pulse
Careful processing of the DS data was required. had a duration of 20 sec at full power. To generate an Į tip
The (now obsolete) DRX-TCP spectrometer [Oxford angle pulse, the pulse power was reduced. A tip angle of Į =
Instruments, UK] used in this study only permitted a single 0.2 rad (approximately 10°) was chosen to provide a useful
acquisition phase to be de¿ned per scan. Consequently, the SNR after eight repeat scans. Each FID consisted of 128
quadrature detection phase of the initial FID differed from points with a dwell time of tdw = 10 s per point. The IJ1 delay
the phase of the subsequent FIDs, and this relative rotation was ¿xed in each acquisition based on the T2* of the sample.
was corrected at the processing stage. More sophisticated A train of n = 50 Į pulses was applied. A recycle delay of tRD
spectrometers may allow the acquisition phase to be = 5 sec was included between scans and IJ0 was incremented
altered within a single scan, removing the requirement for over three separate experiments (IJ0 = IJ1, 1.3×IJ1, 1.6×IJ1).
additional post-processing. The initial FID (observed only The eight repeat scans necessary to accommodate the phase
on alternate scans) had a different scaling factor compared cycle in Table A1 took approximately 90 sec to acquire, and
to the FIDs acquired after each subsequent Į pulse. As the tip the entire measurement had a duration of 4 minutes. Sample-
angle of the Į pulse was known, the signal amplitudes were speci¿c details of the acquisition parameters are given in
corrected according to Eq. 1. Each FID was ¿tted with a Table A2.
monoexponential decay function after correctly accounting For comparison to the DS sequence, T1 relaxation times
for time delays in the digital ¿lters (Valori et al., 2016). The were measured using the classic IR pulse sequence. The
data acquired with different IJ0 delays were interleaved. As longitudinal recovery time was incremented logarithmically
the initial FID was acquired multiple times, these data were from IJ1 = 100 ms to 5 sec over 32 separate acquisitions.
averaged to improve the estimate of M0. The data processing Eight repeat scans were summed and tRD = 5 sec. The entire
steps are illustrated in Fig. 2. Relaxation-time distributions experiment duration was 48 minutes. Also, T2 relaxation
were generated from the processed magnetization decays times were measured on the same brine-saturated rock plugs
(see Fig. 2d, × symbols) using a numerical inversion method using a 2.4 MHz Geospec magnet [Oxford Instruments,
based on Tikhonov regularization with a second-derivative UK]. CPMG decays were recorded with an echo time of tE =
smoothing operator (Wilson, 1992). 600 s and 8,333 echoes, tRD = 5 sec and eight repeat scans
The modi¿ed DS sequence was implemented on a for a total experiment duration of 80 sec.
custom 10.7 MHz permanent magnet [Laplacian Ltd,
Table A1—RF Phase Cycle for the Pulse Sequence Illustrated in Fig. 1
The two ʌ/2 pulses constitute the composite preparation pulse (applied once per acquisition). The Į pulse is applied n times per acquisition (constant
phase). Acq phase refers to the FID acquisition windows. The RF phases are incremented over repeat scans, so eight scans are required to
accommodate the entire phase cycle.
Table A2—Estimated Relaxation Times of Brine-Saturated Outcrop Rock Samples and Corresponding DS Sequence Parameters
The total experiment duration is based on eight repeat scans, three IJ0 values, and tRD = 5 sec
ABOUT THE AUTHORS patents in the ¿eld of NMR and is a reviewer for scienti¿c
and technical international journals.
Jonathan Mitchell is a Senior Research Scientist at
Schlumberger Cambridge Research, UK. He obtained a BSc
degree in 2000 and a PhD in 2003, both in physics, from the
University of Kent, UK. He was previously employed at the
University of Cambridge, UK, under a technology transfer
program bridging academia and industry. His research
interests cover all aspects of NMR related to porous media,
petrophysics, and well construction with a current focus
on drill cuttings and drilling Àuids. He has published over
90 articles in peer-reviewed journals and eight patents. In
2012, he was awarded the Giulio Cesare Borgia Prize for
contributions to the understanding of NMR in porous media;
in 2013 he was awarded the BRSG-NMRDG annual prize
for contributions to magnetic resonance; and in 2014, he was
awarded the prize for Best Young Professional Paper at the
Society of Core Analysts Annual Symposium.
ABSTRACT
The foundational paper by Klinkenberg contains a very methods: by optimizing the Darcy permeability and the
rich dataset for gas Àow in porous samples over a range of Klinkenberg coef¿cient simultaneously; by holding the
mean pressures from 1 to 2,000 kPa. Based on his data, Darcy permeability constant but optimizing the value of
Klinkenberg proposed a correlation between pressure drop the Klinkenberg coef¿cient to obtain a single value for
and Àow rate that depends on both the Darcy permeability all mean pressures; by optimizing Darcy permeability,
(the permeability at in¿nite mean pressure) and the ratio the Klinkenberg coef¿cient, and a second Klinkenberg
of a coef¿cient, now generally termed the Klinkenberg coef¿cient divided by mean-pressure-squared. The last
coef¿cient, and the mean pressure. Klinkenberg’s approach approach is successful in correlating all of Klinkenberg’s
to analyze his data was to determine the Darcy permeability data to within 5%. However, the improvements due to
at a high mean pressure, then calculate Klinkenberg the modi¿ed Klinkenberg equation are marginal and do
coef¿cients at lower values of mean pressures. He found not explain all the disagreement. For this reason, a second
that values of the calculated Klinkenberg coef¿cient dataset, published by Ash and Grove, was explored. This
remained constant for a certain range of mean pressures, but dataset, which has been largely ignored in the literature,
changed signi¿cantly at low mean pressures. Klinkenberg provides convincing evidence for Klinkenberg’s ideas,
clearly stated that his results did not show a strictly linear once the data are reanalyzed to account for shortcomings
function of effective permeability with the inverse of in the ranges of experimental pressures. Based on ideas
mean pressure—this observation has never been studied in documented by Carman for mixed viscous/ diffusive Àows,
detail. Based on an approach published by Arabjamaloei the results are used to derive estimates of an effective pore
and Ruth, Klinkenberg’s data were reanalyzed using three diameter and the tortuosity.
INTRODUCTION
suf¿ciently detailed experiment data to allow reanalysis of
Two foundational papers on low-pressure Àow in porous the results, a rare occurrence in the literature.
media, one by Klinkenberg (1941) and the other by Ash For the purpose of the present paper, three characteristic
and Grove (1960), have had very different impacts in the Àow regions are de¿ned. When pressure is very low or pore
literature. Based on data from Google Scholar at the time sizes are very small, this will be termed the “purely diffusive
of writing the present paper, the paper by Klinkenberg has Àow” region. This region is often termed the “free molecular
been cited 1,853 times while the paper by Ash and Grove Àow” region. When the pressure is very high or the pore
has been cited only 15 times. The present paper will show sizes are large, this will be termed the “purely viscous
that by combining the results from these papers with the Àow” region. This region is often termed the “Poiseuille
model documented by Carman (1956) (a similar treatment Àow” or “Darcy Àow” region. Between the two regions is
is also presented in Klinkenberg’s paper for a single straight an “intermediate Àow” region. This region is also termed
capillary), a very simple method can be derived to predict an the “Knudsen Àow”, “slippage Àow” or “Klinkenberg Àow”
effective pore diameter and tortuosity of a porous sample. region, although the Knudsen Àow region is often de¿ned to
The great strength of the two papers is that they both contain include the purely diffusive Àow region.
Manuscript received by the Editor October 26, 2018; manuscript accepted March 21, 2019.
1
Originally presented at the International Symposium of the Society of Core Analysts, Trondheim, Norway, August 27–30, 2018,
Paper SCA2018-001.
2
Department of Mechanical Engineering, University of Manitoba, Winnipeg, Manitoba, Canada
and
Here Ab is the bulk cross-sectional area and ȡ is the density.
For steady, compressible gas Àow, the mass Àow is constant
Second Klinkenberg Number (12)
along the sample but the Darcy velocity will vary with the
density, hence, pressure. Substituting Eq. 2 into Eq. 1 results
to yield
in the equation
(13)
(3)
At ¿rst sight, the second Klinkenberg number appears to
For an ideal gas and isothermal Àow be ill-behaved because as Ph ĺ Pl , this term goes to in¿nity.
However, as Ph ĺ Pl then In(Ph /Pl ) ĺ 0 which compensates.
(4)
Here the subscript denotes the conditions at the arithmetic THE KLINKENBERG RESULTS
mean pressure. Substituting into Eq. 3 and multiplying
through by P Klinkenberg (1941) included the following statement
in his paper: “Figs.1, 2 and 3 show that the apparent
permeability is approximately a linear function of the
(5) reciprocal mean pressure. The linear function, however,
is an approximation… wherein the value of the constant b
The right-hand side of Eq. 5 is a constant for steady Àow. increases with increasing pressure.”
Therefore, the equation can be integrated over the bulk To explore reasons for this behavior, the Klinkenberg
length, Lb, and from the pressure at its highest value, Ph , to data were reanalyzed using three different approaches. First,
its lowest value, Pl . The result is the data were ¿tted with Eq. 9 but assuming b2 = 0. This
resulted in values for ko and b. Second, the data were ¿tted
(6) with Eq. 9 holding the value of ko equal to the value at high
pressure and assuming b2 = 0. Third, the data were ¿tted with
Eq. 9 allowing ko, b and b2 to vary. The results for the three
Because the mean pressure, Pm, is equal to to (Ph + Pi)/2, this samples for which Klinkenberg provided detailed data are
equation can be rearranged as shown in Table 1.
(7)
- - -
Figures 1 through 3 show the errors between the ¿tted better predicted at high pressures, although not as well as for
equations and the measured values of effective permeability. Sample 1, but the error near 10 kPa is still large. Using all
For Sample 1 (Fig. 1), using both ko and b results in errors three parameters results in the best prediction. However, the
exceeding 5% at high pressures and at approximately 10 improvements are again marginal and the errors near 10 kPa
kPa. When only b is used in the ¿t, the effective permeability are still relatively large.
is well predicted at high pressures (this should occur because In summary, despite using a higher-order correlation,
this is the region used to predict ko) but the error near 10 there remains a systematic deviation in the Klinkenberg
kPa is the greatest observed. Using all three parameters data. In order to obtain further insights into this problem,
results in the best prediction. However, the improvements a second dataset, published by Ash and Grove (1960), was
are marginal and the errors at high pressures and near 10 kPa studied.
are still relatively large.
The behavior of Sample 2 (Fig. 2) is similar to that for
Sample 1. Again, using both ko and b results in the largest
errors at high pressures. When only b is used in the ¿t, the
effective permeability is better predicted at high pressures,
although not as well as for Sample 1, but the error near 100
kPa has increased. Using all three parameters results in the
best predictions. However, the improvements are again
marginal and the error near 100 kPa is still relatively large.
The behavior of Sample 3 (Fig. 3) is even more similar
to that for Sample 1. Again, using both ko and b results in the
largest errors at high pressures and approximately 10 kPa. Fig. 2—The errors for Sample 2 (Jenna Glass ¿lter). These errors are
the values predicted by the correlation equation minus the measured
When only b is used in the ¿t, the effective permeability is values, divided by the measured values.
Fig. 1—The errors for Sample 1 (Core Sample “A”). These errors are Fig. 3—The errors for Sample 3 (Core Sample “F”). These errors are
the values predicted by the correlation equation minus the measured the values predicted by the correlation equation minus the measured
values, divided by the measured values. values, divided by the measured values.
THE ASH AND GROVE RESULTS In the present analysis, two different error equations were
used
Ash and Grove (1960) reported effective permeability,
and upstream and downstream pressure data, for 10 gases (17)
on a single sample of ceramic porous media. However, the
data are not in the same form as that used in the petroleum and
literature. The Àow rate, G, was calculated using the
declining pressure in a known upstream volume using the (18)
equation
Table 5—Calculated Values for the Klinkenberg Coef¿cient and the Mean Free Path Coef¿cient
_
¥2 or 1.414, which is remarkably close to the predicted
value based on the Ash and Grove (1960) data.
At very low pressures, or for samples with very small
Àow passages, the Àow becomes purely diffusive and
viscosity effects become minimal. The onset of this behavior
is characterized by the Knudsen number, which is de¿ned as
(24)
Fig. 5—The errors for the Ash and Grove (1960) data. These errors are (25)
the values predicted by the correlation equation minus the measured
values, divided by the measured values.
As the Knudsen number becomes large, this equation goes to
The Ash and Grove (1960) work was motivated by a If experiments are conducted in the diffusive region, this
desire to determine pore properties based on Àow in capillary equation may be used to model the results. The measurement
tubes collected into a parallel tube model. The equation for of diffusive properties of porous media has great utility. As
the effective permeability (as developed in Carman (1956) pointed out by Klinkenberg in a separate paper (Klinkenberg,
but based on the earlier work of others, most notably Adzumi 1951), diffusion is an analogy for electrical conductivity in
(1937a, 1937b, 1937c)) is porous media. For a parallel tube model, the formation factor
is given by (Ruth et al., 2013)
(21)
(27)
Here is the porosity, į is an “effective” tube diameter, IJ is
the tortuosity, and Ȝࡄ is the mean free path of the gas given Therefore, once effective pore diameters and tortuosities
by Eq. 19. Comparing this equation with Eq. 9, and ignoring are determined, formation factors can be predicted without the
the second Klinkenberg term, the Darcy permeability may need to saturate the samples with an electrically conducting
be identi¿ed as liquid. The present work clearly demonstrates that when
experiments are conducted to capture and analyze Àow in
(22) both the diffusive and viscous regions, gas Àow experiments
give the results predicted from simple theories of Àow in
and the Klinkenberg coef¿cient by tubes. Therefore, they should allow calculation of meaningful
values for effective pore diameters and tortuosities. It is the
(23) opinion of the authors that diffusive experiments represent
a very important but underused opportunity to gain a much
better understanding of rock samples.
Given values for , ko and b, these equations allow
A word of caution is required. It should always be
the prediction of the effective diameter and the tortuosity.
possible, by using very low pressures, to conduct experiments
Based on the Ash and Grove (1960) data, the values for these
in moderate- to high-permeability samples that range
variables are į = 1.91×10-6 ± 0.14 m and IJ = 1.46 ± 0.11. The
over the diffusive, intermediate, and viscous Àow regions.
value for the diameter compares favorably with values that
However, for low-permeability samples, it may be dif¿cult
were calculated by Ash and Grove (1960) based on a number of
to perform experiments in the viscous Àow region without
different approaches (0.41×10-6 to 2.84×10-6 m). The medium
using very high pressures. Therefore, the potential to use this
used for this study was a ceramic, which would be expected
technique to determine pore structure on tight samples needs
to have a uniform and systematically packed structure. As
further investigation.
such, a reasonable expectation for the tortuosity would be
10.1130/0016-7606(1951)62[559:ABDAEC]2.0.CO;2.
Ruth, D.W., Lindsay, C., and Allen, M., 2013, Combining
Electrical Measurements and Mercury Porosimetry to Predict
Permeability, Petrophysics, 54(6), 531–537.
ABSTRACT
Water saturation is one of the a few key reservoir by logs. However, carbonate formations are usually
properties, if not the most important one, for pay-zone heterogeneous due to the existence of vugs and fractures,
identi¿cation and, hence, reserves determination. The oil and the value of m measured from core plugs or even whole
industry uses resistivity logs, with help of porosity logs, core does not represent the value of the formation volume
to calculate water saturation based on the Archie equation. investigated by logs. The frequently asked question in
This is because the resistivity log has the deepest depth of the industry is: “How can we get the value of m if core
investigation and largest difference in log readings between values cannot represent the reservoir?” and “What is
hydrocarbon (HC) and water among all existing logs, and the appropriate value of m for evaluation of carbonate
the Archie equation is a simple and reliable equation with reservoirs with vugs and fractures?”
only three parameters: porosity (or cementation) exponent The purpose of this paper is to answer the above
m, saturation exponent n and formation water resistivity questions. We ¿rst explain why the values of m from
Rw. As we can accurately measure Rw from water samples core plugs and whole core are not suitable for evaluation
or calculate it in wet zones, the two remaining critical of carbonate reservoirs by using the scale model of core
parameters for water saturation calculation are m and n. In plugs, whole core and logs. We then propose a new concept
this paper, we focus on m only; we will discuss n in a future for obtaining the appropriate value of m for evaluation of
paper. carbonate reservoirs with vugs and fractures at the well-
In homogeneous rock, such as most sandstone log scale based on Monte Carlo simulation with a triple-
formations, the value of m measured from core plugs porosity model of matrix, vugs and fractures.
represents the value of the formation volume investigated
Manuscript received by the Editor March 25, 2019; manuscript accepted April 29, 2019
1
Southwest Petroleum University, College of Earth Science and Technology; No.8 Xindu Road, Chengdu, China, 610500;
pengjun@swpu.edu.cn; lishi6226@163.com
SCALE OF CORE PLUG, WHOLE CORE AND LOG Combining Figs. 1 and 2, we can obtain a simpli¿ed scale
model (see Fig. 3) to show the relative scale of core plug,
In homogeneous rock, such as most sandstone formations, whole core, formation volume investigated by resistivity
the value of m measured from core plugs represents the log and the distribution of fractures and vugs within the
formation volume investigated by well logs. But, since reservoir.
carbonate formations are often heterogeneous, due to the From Figs 1 to 3 one can easily understand that the value
existence of vugs and fractures, the value of m measured of m obtained from core plugs or whole core cannot represent
from core plugs or even whole core does not represent the the formation volume investigated by logs. As mentioned
formation volume investigated by logs. Figure 1 shows a previously, the equations of even the triple-porosity model
model illustrating the relative volumes of investigation of also cannot be used to get m due to the dif¿culty of obtaining
core plugs, whole core and well logs. the porosities of fractures and vugs. Consequently, this paper
proposes a new concept to determine the value of m for
evaluation of carbonate reservoirs with fractures and vugs.
METHODOLOGY
Fig. 3—Simpli¿ed scale model showing the relative scale of core plugs, whole core, formation volume investigated by well logs and the fractures and
vugs within the formation volume or rock.
The Effect of Nonconnected Vugs on m assumed a dual-porosity system that consists of matrix and
Nonconnected vugs are not connecting or touching each nonconnected vugs. Equation 1 can be used to simulate the
other directly, but they are connecting to the matrix and, effect of nonconnected vugs on m by forcing the porosity of
hence, the system. So, nonconnected vugs are still connected fractures equal to zero, in this case. Figure 6 shows the effect
to the reservoir system and have the capability of storing of nonconnected vugs on m. It is obvious that presence of
and transporting oil and gas. To demonstrate the effect of nonconnected vugs causes m to increase from mb.
nonconnected (or nontouching) vugs on m, the formation is
Fig. 4—Petrophysical volume model for the triple- porosity system in carbonate.
Fig. 5—Effect of fractures on m in a dual-porosity system (matrix + Fig. 6—Effect of nonconnected vugs on m in a dual-porosity system
fractures), with mb = 2.0 and mf = 1.0. The value of m decreases with (matrix + nonconnected vugs),with mb = 2.0. The value of m increases
increasing of porosity of fractures. The presence of fractures results in with increasing of porosity of the nonconnected vugs. The presence of
m < 2. nonconnected vugs results in m > 2.
The Combined Effect of Fractures and Nonconnected case of matrix and fractures only (Fig 7a) and the effect of
Vugs on m nonconnected vugs on m in case of matrix and nonconnected
Figure 7 is formed by combining Figs. 5 and 6. From vugs only (Fig. 7b).
Fig. 7, one can observe the effect of fractures on m in the
(a) (b)
Fig. 7—(a) Effect of fractures on m in a dual-porosity system (matrix + fractures), where mb = 2.0 and mf =1.0. (b) Effect of nonconnected vugs on m
in a dual-porosity system (matrix + nonconnected vugs), where mb = 2.0.
Typically, both fractures and vugs exist in carbonate represent the matrix and the value of m of the matrix has
reservoirs. To determine the combined effect of fractures a very narrow range with an average close to 2.0.
and nonconnected vugs on m, we use Eq. 1 to simulate the x The value of m obtained from whole core also does not
effect of fractures and nonconnected vugs together on m, represent reservoir volume investigated by resistivity logs
which is illustrated in Fig. 8. and it cannot be used for formation evaluation directly.
From Fig. 8, one can see that the range of m is much x The dual-porosity models (matrix + fractures; matrix
narrower when both factures and nonconnected vugs are + nonconnected vugs) previously developed are
present, compared with the cases of only fractures or only fundamental models that can help petrophysicists to
understand the theoretical effect of fractures or vugs on
nonconnected vugs shown in Fig. 7. One can determine the
m. Those models show that the value of m varies over
value of m of a reservoir with both fractures and vugs if the
a large range (see Fig. 7). However, those models are
porosity of fractures and the porosity of nonconnected vugs
not suitable for real formation evaluation as carbonate
can be obtained from logs. reservoirs often contain both fractures and vugs.
x Triple-porosity models (matrix + fractures + nonconnected
A New Concept for Obtaining the Appropriate Value of vugs) developed by researchers in recent years are very
m at the Well-Log Scale useful models. These models can be used to determine
From the discussion to this point, one can conclude the the value of m for formation evaluation if one can manage
following facts for carbonate reservoirs: to determine the porosity of fractures and the porosity
x The value of m obtained from core plugs does not of nonconnected vugs within the formation volume
represent the reservoir volume investigated by resistivity investigated by resistivity logs. Unfortunately, up to the
logs, and it cannot be used for formation evaluation present, it is almost impossible to use even modern logs
directly. In practice, during the acquisition of core plugs, to obtain a reliable porosity of fractures and a reliable
many core plugs are sampling matrix, some of them are porosity of nonconnected vugs within the formation
sampling matrix with some vugs, and a few of them are volume investigated by resistivity logs. However, triple-
sampling matrix with microfractures. As a result, the porosity models do demonstrate one important thing,
value of m measured from core plugs can vary over a large that is, the value of m varies over a much narrower range
range (1.0 to 7.3) depending on the positions of the core when both fractures and vugs are present in the reservoir
plugs. Based on the coring and core-analysis experience (see Figs. 7 and 8).
of authors at least, if carefully selected, core plugs can
Fig. 8—Combined effect of fractures and nonconnected vugs on m in a triple-porosity system (matrix + fractures + nonconnected vugs), where mb =
2.0 and mf = 1.0.
From the discussions above, it seems that there most-likely value at each simulation case. While hundreds
currently no way to determine the value of m for evaluation of cases are simulated, Fig. 9 shows one example from those
of carbonate reservoirs. This paper proposes a new concept simulation cases, which includes the input (matrix porosity,
to determine the appropriate value of m of the formation fracture porosity, nonconnected vug porosity) and the value
volume investigated by resistivity logs through Monte Carlo of m modeled by Monte Carlo simulation.
simulation using the triple-porosity model.
Consider a formation volume investigated by resistivity RESULTS
logs (see Figs. 1 and 3). The value of m of the matrix within Table 1 lists the simulated values of m for the cases of
the formation volume is set to 2.0 (close to 2.0 in the real various combinations of matrix porosity, fracture porosity
case). Adding fractures (or connected vugs) to the volume and nonconnected vug porosity. For various cases, the
randomly will result in m < 2.0. Adding nonconnected vugs values of m of reservoir volume investigated by resistivity
to the volume randomly will result in m > 2.0. Based on the logs with both randomly distributed fractures and vugs are
triple-porosity models, the Al-Ghamdi et al. (2011) model is all very close to 1.99 (average: 1.99, variation range: 1.94 to
selected here, one can use Monte Carlo simulation to obtain 2.02). From Table 1, two key points are stand out: (1) The
the value of m of the volume with randomly distributed value of m slightly increases with increasing matrix porosity,
fractures and vugs at various combinations of matrix indicating that fractures have a greater impact than vugs on
porosity, fracture porosity and nonconnected vug porosity. m when matrix porosity is low, while vugs have a greater
To get a more realistic simulation result, the variation ranges impact than fractures on m when matrix porosity is high; (2)
of matrix porosity, fracture porosity and nonconnected the values of m are all close to 1.99, indicating that 1.99 can
vug porosity are set within their minimum value 0.0 and be used as the value of m without knowing anything about
maximum value 1.0, and each of those porosities has a the target reservoir.
Fig. 9—Monte Carlo simulation at the combination case of כf = 0.005 v/v, כv = 0.05 v/v and כb = 0.15 v/v. Input includes (a) כf minimum value = 0.0 v/v,
maximum value = 1.0 v/v, most-likely value = 0.005 v/v, probability distribution =triangle); (b) כv minimum value = 0.0 v/v, maximum value = 1.0 v/v,
most-likely value = 0.05 v/v; and (c) כb minimum value = 0.0 v/v, maximum value = 1.0 v/v, most-likely value = 0.15 v/v. In all three cases the green
triangle represents the probability distribution. (d) m values from Monte Carlo simulation using the Al-Ghamdi et al. (2011) triple-porosity model (the
result or the median value of m = 1.98).
Table 1—Monte Carlo Simulated Values of m for Various Combination of Matrix Porosity, Fracture Porosity and Nonconnected Vug Porosity
0.01 0.03 0.05 1.96 0.01 0.03 0.05 1.96 0.02 0.03 0.05 1.95
0.01 0.03 0.10 1.97 0.01 0.03 0.10 1.97 0.02 0.03 0.10 1.97
0.01 0.03 0.15 1.98 0.01 0.03 0.15 1.98 0.02 0.03 0.15 1.98
0.01 0.03 0.20 1.99 0.01 0.03 0.20 1.99 0.02 0.03 0.20 1.99
0.01 0.03 0.25 1.99 0.01 0.03 0.25 1.99 0.02 0.03 0.25 1.99
0.01 0.03 0.30 2.00 0.01 0.03 0.30 2.00 0.02 0.03 0.30 2.00
0.01 0.03 0.35 2.00 0.01 0.03 0.35 2.00 0.02 0.03 0.35 2.00
0.01 0.03 0.40 2.00 0.01 0.03 0.40 2.00 0.02 0.03 0.40 2.00
0.01 0.05 0.05 1.96 0.01 0.05 0.05 1.96 0.02 0.05 0.05 1.96
0.01 0.05 0.10 1.98 0.01 0.05 0.10 1.97 0.02 0.05 0.10 1.98
0.01 0.05 0.15 1.98 0.01 0.05 0.15 1.99 0.02 0.05 0.15 1.98
0.01 0.05 0.20 1.99 0.01 0.05 0.20 2.00 0.02 0.05 0.20 1.99
0.01 0.05 0.25 2.00 0.01 0.05 0.25 2.00 0.02 0.05 0.25 2.00
0.01 0.05 0.30 2.00 0.01 0.05 0.30 2.00 0.02 0.05 0.30 2.00
0.01 0.05 0.35 2.00 0.01 0.05 0.35 2.00 0.02 0.05 0.35 2.00
0.01 0.05 0.40 2.01 0.01 0.05 0.40 2.00 0.02 0.05 0.40 2.00
0.01 0.08 0.05 1.98 0.01 0.08 0.05 1.97 0.02 0.08 0.05 1.97
0.01 0.08 0.10 1.98 0.01 0.08 0.10 1.98 0.02 0.08 0.10 1.98
0.01 0.08 0.15 2.00 0.01 0.08 0.15 1.99 0.02 0.08 0.15 1.99
0.01 0.08 0.20 2.00 0.01 0.08 0.20 2.00 0.02 0.08 0.20 1.99
0.01 0.08 0.25 2.00 0.01 0.08 0.25 2.00 0.02 0.08 0.25 2.00
0.01 0.08 0.30 2.01 0.01 0.08 0.30 2.00 0.02 0.08 0.30 2.00
0.01 0.08 0.35 2.01 0.01 0.08 0.35 2.01 0.02 0.08 0.35 2.01
0.01 0.08 0.40 2.01 0.01 0.08 0.40 2.01 0.02 0.08 0.40 2.01
0.01 0.10 0.05 1.99 0.01 0.10 0.05 1.98 0.02 0.10 0.05 1.97
0.01 0.10 0.10 1.99 0.01 0.10 0.10 1.99 0.02 0.10 0.10 1.99
0.01 0.10 0.15 2.00 0.01 0.10 0.15 2.00 0.02 0.10 0.15 2.00
0.01 0.10 0.20 2.00 0.01 0.10 0.20 2.00 0.02 0.10 0.20 2.00
0.01 0.10 0.25 2.00 0.01 0.10 0.25 2.01 0.02 0.10 0.25 2.00
0.01 0.10 0.30 2.01 0.01 0.10 0.30 2.01 0.02 0.10 0.30 2.01
0.01 0.10 0.35 2.02 0.01 0.10 0.35 2.01 0.02 0.10 0.35 2.01
0.01 0.10 0.40 2.02 0.01 0.10 0.40 2.01 0.02 0.10 0.40 2.01
0.01 0.15 0.05 1.99 0.01 0.15 0.05 1.99 0.02 0.15 0.05 1.99
0.01 0.15 0.10 2.00 0.01 0.15 0.10 2.00 0.02 0.15 0.10 2.00
0.01 0.15 0.15 2.01 0.01 0.15 0.15 2.01 0.02 0.15 0.15 2.01
0.01 0.15 0.20 2.01 0.01 0.15 0.20 2.01 0.02 0.15 0.20 2.01
0.01 0.15 0.25 2.02 0.01 0.15 0.25 2.02 0.02 0.15 0.25 2.02
0.01 0.15 0.30 2.02 0.01 0.15 0.30 2.02 0.02 0.15 0.30 2.02
0.01 0.15 0.35 2.02 0.01 0.15 0.35 2.02 0.02 0.15 0.35 2.02
0.01 0.15 0.40 2.02 0.01 0.15 0.40 2.02 0.02 0.15 0.40 2.02
ABSTRACT
Building realistic and reliable subsurface models (for comparable rock quality). A signature of perching is
requires detailed knowledge of both the rock and Àuids that between two sampled wells, the hydrocarbon (HC)
involved. While the hydrocarbon volume estimation has a column is in hydraulic communication while two mobile
profound impact on the viability of a development, next water pools seem to be disconnected. The fundamental
to the permeability, saturation-height models, free-Àuid controls that lead to the formation of perched contacts
levels (FWLs) and the hydraulic communication have a are studied and shown to be the rock quality and relative
signi¿cant role in determining the recoverable reserves. permeability. The perching effect is not going to feature
When in different parts of the same ¿eld different in poor quality rocks (submillidarcy permeability)—the
free-Àuid levels (leading to different Àuid contacts for effects would be visible only for a considerable barrier
the same rock quality) are identi¿ed, the lateral hydraulic height. Regarding transition zones, the results showing no
communication at the ¿eld level can be challenged. This signi¿cant difference are expected above the perched zone
aspect is of importance since the hydrocarbon volume when compared to the unconstrained parts of the ¿eld.
distribution impacts the recovery factor. At the same time Field observations and dynamic simulations are used to
building and initializing a model based on different FWL identify the perching controls. A clear distinction is shown
positions (zero capillary pressure) is challenging. between capillary pressure and buoyancy. The fundamental
Perched water contacts are the result of water entrapment assumption that the capillary pressure can be calculated by
(behind barriers for lateral Àow) during hydrocarbon using the height above FWL is shown to be de¿cient when
migration in the reservoir, as a result leading to locally water becomes immobile.
elevated FWL or hydrocarbon-water contact position
Manuscript received by the Editor July 12, 2018; revised manuscript received February 1, 2019; manuscript accepted February 7, 2019.
1
Parts of this work originally presented at the SPWLA 59th Annual Logging Symposium, London, England, UK, June 2–6, 2018, Paper K.
2
Shell Global solutions International BV, Kessler Park 1, 2288 GS, Rijswijk, The Netherlands; Iulian.Hulea@shell.com
While the phenomenological description in the above Table 1—Parameters Varied as Sensitivities Cover a Wide Range of
Properties
cases appears to be accurate, the rock or Àuid microscopic
parameters controlling the effect are not identi¿ed (Weber, Pce
1997). Also, in terms of predictability it is not obvious how
rock permeability affects perching. Should we expect the
effect to be exacerbated or on the contrary, to be absent as
rock quality is changed? In some simulations there is no assumed relationship between entry
pressure and permeability (see Eqs. 1 and 3).
METHODOLOGY
The model size is 1,000×1,000×1,000 m in the x, y, and z
To study the effect of water perching, dynamic directions. To facilitate the trapping conditions in the models,
simulations have been carried out. The framework model the central part of the model (Zone 2) is isolated laterally by
of the rock is shown in Fig. 3, where a constant porosity two barriers (see Vrolijk et al., (2016) on possible barriers
and permeability is assumed through individual models. origins) and the bottom is sealed to Àow. Zone 1 and 3 are
Porosity/permeability can be varied between simulations in communication, while the injection happens at the top of
according to Table 1. In the vertical direction, the models the model where all three zones are communicating laterally
have a 5-m resolution, much smaller than the difference in (Fig. 3a).
FWL we aim to investigate (tens to hundreds of meters.) The oil-water-rock interaction is governed by Brooks-
Corey saturation-height function (Hulea and Nicolls, 2012):
(1)
(2)
(b) with Swc ¿xed at 0.2 and resulting in the graphs shown in Fig.
4. A Swc > Swi is a necessary condition to ensure the water
becomes immobile at a given capillary pressure for a few Nw
(Corey parameter) values.
The simulations are carried out starting at Sw = 1 and
slowly injecting 1/3 of the model (pore volume) with oil.
Once the injection has stopped, we wait for equilibrium
another 100X the injection time leading to total simulations
Fig. 3—The (3D) models built are homogeneous with constant porosity
times representing up to tens of thousands of years. The
(hence permeability). Zone 2 is isolated from Zone 1 and 3 in a lateral
sense between the barriers shown. The bottom of Z1 and Z3 have a FWL position at the end of the simulation (per zone) is
pressure boundary open for water Àow while Zone 2 is isolated, allowing determined as the pressure-gradient intersection between
Àuid exchange only at the top. Each one of the zones is monitored the hydrocarbon- and water-pressure gradients, as shown in
by one vertical well (Zone 1 by Well 1, etc.). During the simulations,
saturations, pressures, and relative permeability are monitored along Fig. 5.
the three synthetic wells. (b) a cross section of the model at the end of
the simulations with the ¿nal saturations showing a perching example
(and the corresponding FWLs).
(a) (b)
Fig. 4—Examples of relative permeability and capillary pressure curves used in this work. (a) Relative permeability curves used in the simulations with
the water Corey exponent (see Eq. 2) ranging from < 1 to > 5 covering a wide range while the No is kept constant at 2. Sensitivities have been carried
out for No = 2 and No = 5 without an impact on the simulation outcome. (b) The Sw vs. depth resulting by coupling Pce and permeability according to
Eq. 3 and using two Npc values (1.25 and 2.5 are used for these examples). For the same color, the effect of Npc is seen while by comparing colors
the effect of entry pressure becomes visible.
RESULTS
(a)
(b)
(c)
(b)
Fig. 6—The fundamental controls on perching extracted from a dataset
Fig. 7—(a) FWL position with respect to barrier position as identi¿ed
where the entry pressure (Pce), and capillary pressure shape factor
at the end of the simulation for all the simulations vs. those leading
(Npc) are varied independently and with equal probability. The factors
to perching. A FWL2 closer to the barrier than 400 m is correlated to
controlling perching are (a) the capillary pressure shape factor (Npc)
perching (this number is speci¿c to the injected volume). (b) The FWL in
and (b) entry pressure (Pce). Blue shows input number of values, while
Zone 2 vs. FWL in Zone 1 showing a wide range of values, perching is
in orange only those leading to perching. (c) A crossplot of Pce vs. Npc
observed only for the orange dots (deviations from a 1:1 line).
showing the combined values leading to perching.
One interesting aspect is that for perching to occur, as the perching would decrease as the FWL in the trap gets deeper.
capillary pressure shape factor increases, the entry pressure This is highlighted in Fig. 7a.
should decrease, otherwise the effect disappears (Fig. 6c). Where perching does occur, a microscopic explanation
One other observation is that as the FWL in the perched of how the effect develops is offered by examining one of
zone can be found deeper (high entry pressure) the ratio of the simulations in more detail. It rapidly becomes obvious
cases where perching appears vs. the total number of cases that the water permeability is the limiting factor in the
decreases. As the entry height increases, the height interval water escape route (Fig. 8). It is worth noting that at the
(as compared to FWL) required to immobilize the water highest point of the barrier the water relative permeability is
increases, hence, accommodating more HC in the trapped essentially zero under equilibrium—this is the critical place
zone. So, for a given injected HC volume the chances of in the subsurface where the water escape path is choked.
DISCUSSION
, (3)
This example (Fig. 8) highlights how the process
leading to water perching is directly controlled by relative
permeability. In addition, we can compare the transition hence, entry pressure being an indicator of rock quality, ı
zone signature between the perched zone and the rest of the being the oil-water interfacial tension, k is permeability, is
reservoir. By overlaying the Sw vs. depth pro¿les we observe porosity, and ș the contact angle.
the saturation pro¿les do not differ between the normal and
perched zones (for same rock quality (Fig. 9).
(a) (b)
Fig. 12—(a) The porosity-permeability data input colored according to a perched discriminator. (b) The number of perching cases observed vs.
permeability showing the lower the permeability the lower the chance of perching occurring.
Concerning the link between rock quality and perching relevant for modeling perched contacts in subsurface models.
(Fig. 12), a decrease in rock permeability would mean the As shown by simulations, the saturation pro¿les do not
FWL can go deeper compared to the Àow barrier reducing differ between the perched zone and the rest of the reservoir,
the chances of perching (Figs. 7 and 12). Figure 12 shows assuming same rock quality (see Fig. 9). The transition zone
that as permeability is reduced (with all other conditions does not get compressed—irrespective of rock quality, hence
remaining constant) the likelihood of perching is strongly opening the possibility that the observation in the Grane ¿eld
reduced. is caused by rock quality differences between the two wells
An important result of this work touches upon the (details in Fig. 2). Rock quality might be reÀected in Npc
expected transition above the perched zone, an aspect differences, not necessarily in entry pressure/permeability.
Another important ¿nding of this work lies in a To study this inconsistency in modeling capillary
deeper understanding of capillary and buoyancy forces in pressures, in the simulations we have monitored not only
the reservoir. Capillary pressure is the pressure difference the formation pressure but additionally the pressure per
between the two phases: phase. In the cases where perching does not happen, the
simulation outcomes are as expected. Above the FWL the Pc
Pc = Poil — Pwater (4) increases equaling buoyancy with a prede¿ned (by density)
gradient (see Fig.14). Note that above the FWL both the
oil and water have a prede¿ned gradient of 0.22 and 0.45
This is easily calculated by extrapolating the water psi/ft, respectively. For cases where perching does happen
and HC gradients, but in the case of perched contacts as (Fig. 15), the simulation results are not straightforward.
described here, it leads to an apparent inconsistency. Based While the formation pressures are comparable, a peculiarity
on the assumption the pressure follows prede¿ned gradients, appears in the water-phase pressure. As soon as it becomes
the capillary pressure at reservoir top in Well 1 (with the immobile it does not follow the expected water pressure but
deep contact) would be different than at Well 2 (with the tracks instead the HC pressure, following the HC gradient
perched contact) but this would lead to different pressures (see Fig. 15d vs. Fig. 14d). This phenomenon starts at the
and, hence, crossÀow; which is not observed—the system depth corresponding to krw = 0, this becoming the maximum
is in equilibrium. The capillary pressure at the reservoir Pc point. Shallower than this point, Pc is not increasing. This
top should be the same irrespective of what is underneath, is well understood from the point of view that saturation
perched or not (contrasting the approach of Adams, 2016), and Pc are interdependent, an immobile wetting phase (Sw = Swc)
see Figure 13. For clarity, buoyancy is: translates into a saturated capillary pressure. Capillary
pressure change without water saturation alteration is not
possible. The same effect is observed along Well 1 as well
Pb (5)
as along Well 2, the capillary pressure corresponding to the
perched zone saturates at the edge of the barrier. Above that
With ǻp the density difference between water and HC. point, the capillary pressure does not change. The change
By assuming equilibrium between capillary pressure and starts immediately under the edge in the perched zone while
buoyancy, a discontinuity in Pc could be expected between outside the perched area it will occur deeper (see Fig. 16).
the FWL (under the perched contact) and the barrier edge (a The fact that between the two zones we have a barrier,
¿rst surface where the Pc could be expected to vary between hence, no communication means both the capillary pressure
the zones de¿ned by Wells 1 and 2 (see Fig. 13). and formation pressure can differ across the barrier. Above
the barrier, there is no such discontinuity.
Fig. 14—Fluid properties are monitored along two wells. Well 1 penetrates normal aquifer, while Well 2 sits in a zone susceptible to perching (see
Figs. 1 and 13). (a) Water saturation is plotted alongside the barrier, (b) phase pressure, (c) capillary pressure, (d) Àuid gradients, and (e) water
relative permeability.
Fig. 15—Fluid properties are monitored along two wells. Well 1 penetrates normal aquifer, while Well 2 sits in a zone susceptible to perching (see
Figs. 1 and 13). (a) Water saturation is plotted alongside the barrier, (b) phase pressure, (c) capillary pressure, (d) Àuid gradients, and, (e) water
relative permeability.
Fig 16—(a) Saturation vs. depth for a perching case showing the capillary pressure (b) development along the two wells. The capillary pressure
reaches a maximum, between the two wells, a discontinuity is only seen inside the barrier. The calculated buoyancy is plotted along the same
trajectory. (c) The capillary pressure difference as compared to calculated buoyancy is highlighted, leading to signi¿cant discrepancy at the reservoir
top.
We must emphasize the fact that signi¿cant confusion primary drainage assumptions the enabler for perching
persists between capillary pressure and buoyancy, usually is water permeability at the top of the barrier. One way to
they are assumed to be in equilibrium. When increasing the parameterize this is via the saturation, the entry pressure
buoyancy (by moving further away from the FWL), more being the most important control combined with the capillary
water is forced out of the rock until it becomes immobile. pressure shape factor. These two parameters determine (by
Above that depth the capillary pressure and buoyancy determining how thick the transition zone is) the HAFWL
do not appear to be equal anymore. This result can be where water becomes immobile.
explained since the buoyancy assumes the phase exerting Perching is predicted to affect poor rock to a lesser extent
the Archimedes force should be mobile. The underlying than good rock. Good and poor refer here to permeability—
assumption of the Archimedes force is that the displaced to date, cases reporting perching show unambiguously good
phase exercises the buoyant force. Hence, when that phase rock (hundreds of millidarcy or higher permeability).
cannot be displaced we cannot alter buoyancy. When that A clear distinction is shown between capillary pressure
phase becomes immobile the buoyancy concept becomes and buoyancy. The fundamental assumption that the capillary
redundant. Hence, there is no contradiction between the pressure can be calculated by using the HAFWL is shown to
capillary pressure and buoyancy, buoyancy simply does not be de¿cient when water becomes immobile.
apply when water becomes immobile. So, when dealing with
perching we need to use the concept of maximum capillary ACKNOWLEDGEMENTS
pressure.
At a high level, the relevance of the results obtained The author would especially like to acknowledge Harm
is especially around addressing connected hydrocarbon Dijk’s critical contribution and to express his gratitude
volumes with the potential of having a smaller well spacing. to Shell’s permission to publish this work. In addition, in
By establishing a perched contact’s existence and control setting up the simulations Bastiaan Huisman and Bram
mechanism, the hydrocarbon connectivity is not challenged, Timmerarends have been instrumental. Igor Kim’s input
we might even be able to extract meaningful properties, such is also acknowledged as well as signi¿cant manuscript
as barrier height, knowing the rock properties. The transition- improvements by Frank Wolters, Jean-Paul Koninx, Ap van
zone signature allows us to address a possible hydrocarbon- der Graaf, Martin Kraaijveld, Mike Dean, John van Wunnik,
volume reduction method and water production prediction Ove Bjorn Wilson and Shehadah Masalmeh.
by avoiding well drilling in the minima of a permeable
formation.
Signi¿cant confusion persists around perching contacts, NOMENCLATURE
other sources of variability of FWL across a ¿eld—like
hydroactive aquifers—are considered (Boya-Ferrero et al., Abbreviations
2012). While pure perching cases (assuming an equilibrium FWL= free-water level
state) involve no HC below the FWL, the hydroactive HAFWL= height above free-water level
aquifers are associated to a certain extent with both FWL HC= hydrocarbon
differences at ¿eld level and residual HC. A pure perching SHM= saturation-height model
contact (under the equilibrium assumption) implies the
absence of HC below FWL. In a complex reservoir history, it
could be envisaged that by tilting the reservoir (changing the Symbols
FWL versus barrier height) residual HC may appear below k = permeability
the FWL. krw = water relative permeability
In this work homogeneous models have been considered, Npc = capillary pressure Brooks- Corey
heterogeneous models are the subject of other work as well parameter (Eq. 1)
as studying hysteresis (among other relevant effects) when Nw = relative permeability Corey parameter (Eq. 2)
looking beyond primary drainage. Pc = capillary pressure (Eq. 1)
Pce = entry pressure (Eq. 1)
CONCLUSIONS Sw = water saturation
Swi = Brooks-Corey parameter (Eq. 1)
This work has been shown to add a predictability Swc = Brooks- Corey connate water saturation (Eq. 2)
dimension to perching, a dimension not addressed before.
It has been demonstrated via simulations that under
ABSTRACT
Driller’s depth has always been the reference run on several projects around the world. Environmental
measurement used when logging while drilling (LWD), corrections were applied in various applications to improve
calculated using the sum of pipe-strap measured while the the accuracy of the depth and demonstrate the signi¿cance
pipe is on surface. However, environmental corrections of the correction for reservoir development. The intention
must be applied to the driller’s depth to account for the of this paper is to raise awareness of the impact of LWD
dynamic changes the pipe undergoes while in the borehole. depth errors and existing approaches for correcting them.
These include dynamic mechanical changes due to drilling Case studies are presented to demonstrate the bene¿ts
activities, changes in the wellbore pro¿le, torque, drag, derivable from applying depth corrections. In one case, the
friction factor, and temperature. They all result in LWD placement of the pressure and sample points provided the
depth being shallower than the absolute depth. most accurate TVD possible. In another case, the corrected
Understanding the position objectives, assigning the measurement enabled determination of where to set the
depth uncertainty and environmental modeling to predict casing depth to within the expected rat hole.
the magnitude of depth correction should be applied in Applying depth correction allows for accurate mapping
advance, along with the surveying technique to be used. of the geological markers, reservoir tops, sand continuity,
These are critical components in the prejob analysis to and Àuid contacts as well as for setting casing and other
ensure the position objectives can be met prior to drilling. drilling applications in offshore deepwater and extended-
Over the years, depth correction has successfully been reach drilling worldwide.
INTRODUCTION new. It was ¿rst discussed by Reistle and Sikes (1938) who
identi¿ed the challenges to accurately obtaining drillpipe and
Knowing the wellbore position and its associated wireline depth measurements. The introduction of the electric
uncertainty is fundamental for optimal wellbore placement, wireline cable tension measurement in the early 1970s,
understanding the geological structure, long-term reservoir comprising integrated dual-wheel electronic measurements
management, avoiding hazards while drilling, and preventing alongside the minicomputer, resulted in the adoption of the
wellbore collision in crowded subsurface environments. wireline depth as the primary depth measurement in the late
Without predicting the position and its uncertainties while 1970s. It took time for the drillpipe depth to catch up, and in
drilling, certain activities can’t be performed reliably with a the early 2000s the introduction of drillpipe environmental
high degree of con¿dence, these include: corrections improved the LWD depth accuracy. However,
x geosteering, positioning the wellbore relative to the in the 80 years since 1938, the industry has still not fully
geological marker, or drilling thin reservoirs. adopted the application of environmental corrections and is
x geometrically or positioning the wellbore as planned only using them on a case-by-case basis.
trajectory for a desired vertical and lateral spacing in In the last few decades, LWD and measurement while
conventional and unconventional reservoir drilling (MWD) has played a key role in primary logging
x wellbore-to-wellbore correlation during exploration and surveying in the ¿eld of exploration, including
or ¿eld development phase. development in conventional and unconventional reservoirs
that were previously dominated by wireline. MWD is the
The quest for an accurate depth measurement is not ¿rst measurement available in real time to drill ahead and
Manuscript received by the Editor February 16, 2018; revised manuscript received January 22, 2019; manuscript accepted January 24, 2019.
1
Schlumberger, 225 Schlumberger Drive, Sugar Land, TX 77478; poedjono1@slb.com
1
Schlumberger, 5599 San Felipe, Houston, TX 77056; nnwosu@slb.com
1
Schlumberger, 3600 Briarpark Drive, Houston, TX 77042; amartin5@slb.com
position the wellbore accurately while drilling. Tables 1 and 2 show the four combinations of correlation
This section will cover position accuracy, uncertainty, coef¿cients allowed to ensure the correlations are de¿ned
drilpipe-stretch modeling, LWD depth measurements, below.
tracking, depth vs. time, environmental corrections and
examples. Table 2 —Combinations of Correlation Coef¿cients
DEPTH UNCERTAINTY
Table 1—The ISCWSA Error Model Values at 1ı for MWD/LWD and Wireline Depth
Fixed Fixed
WELLBORE-POSITION ACCURACY AND The vertical position, TVD, is reliant mainly on the
UNCERTAINTY accuracy of the inclination measurement. Accelerometers
are used to calculate inclination, relying on the earth’s
Improvements to the vertical and lateral position certainty gravity that has varying degree of accuracy. Methods exist
in 3D space with their associated uncertainties, has been for improving inclination to meet the tighter TVD objective
discussed by Williamson (2000) and Grindrod et al. (2016). (Williamson, 2000; Stockhausen and Lesso, 2003; Bordakov
To de¿ne wellbore position, the survey, must be acquired et al., 2007; Monterrosa et al, 2016).
at certain depth intervals, ±100 ft or less, depending on the The vertical and lateral positions are calculated by
positional objective. Survey is de¿ned depth, inclination and interpolating the depth, inclination and azimuth using the
azimuth. The inclination and azimuth are measured using the minimum curvature (Fig. 1) or continuous surveying method
most common surveying tool, MWD. (Fig. 2). The most common azimuth is acquired using
magnetometers in the MWD tool.
Fig. 1—This shows slide rotate pattern vs. calculated position using the minimum curvature method. The actual TVD position could be shallower or
deeper than the actual position depending on the slide/rotate or rotate/slide sequence.
Fig. 2—A plot of a static survey (blue squares) vs. continuous inclination (red circles) and captures the rotate/slide drilling activities.
There are various geomagnetic models available with Modeling using the ¿nite-element method (FEM)
different accuracies (Figs. 3 and 4), as well as update intervals solves a more general mechanical behavior of the drillpipe
(Russell et al., 1995; Williamson et al., 1998; Stockhausen equilibrium inside a 3D borehole (Lesso et al., 1999). This
and Lesso, 2003; Poedjono et al., 2010, 2012a, 2012b, is where the drillpipe and BHA are considered an assembly
2013, 2014, 2018; Mohammed and Terpening, 2012; Maus of tubular components with different mechanical and
et al., 2012). The existence of improvements in inclination geometrical characteristics.
and azimuth measurements underscores the importance of The total stretch is determined by the sum of all the
applying environmental corrections to depth to enhance 3D element displacements in the axial direction by considering
space description. the external forces acting upon each element. This include
gravity, Archimedes force, normal reaction force, axial and
DRILLPIPE-STRETCH MODELING circumferential friction, and weight on bit (WOB).
The only parameter not modeled in the drillpipe
Modeling drillpipe environmental-effect factors allows examples (Figs. 5 and 6) is temperature gradient. As
for a better understanding of what will be seen and gives temperature increases, the stretch will also increase such
a prediction of what to expect in real time during drilling that, depending on how much the temperature increases, this
operations. Drillpipe stretch is affected by several factors, could greatly increase the amount of stretch seen versus the
including wellbore trajectory type, pipe length, friction, model prediction.
buoyancy from the mud, drillpipe size, connection, type and
the bottomhole assembly (BHA) and temperature.
Fig. 3—The relative accuracy of the global model. The contribution to the geomagnetic ¿eld is from the main ¿eld (blue) and crustal ¿eld (green).
Fig. 4—A comparison of the advanced survey processing and SAG correction along with ellipse of uncertainty (yellow) vs. the surveying technique
using HDGM (blue).
Fig. 5—Graph showing that the wellbore pro¿le along with the drilling activity affect the drillpipe stretch.
Fig. 6—Graph showing the stretch for various types of drillpipe along with the drilling activity in a vertical wellbore.
Modeling Results
Fig. 7—Visualization of the various wellbore pro¿les used in the A vertical wellbore is the worst-case scenario for
modeling. drillpipe stretch when drilling on bottom, reaming out, and
rotating off bottom. However, it is not the worst case when
Fig. 8—Top, 5-in. drillpipe with typical BHA. Bottom, 5-in. and 6.625-in. tapered string used in the modeling.
tracking accuracy. This is often done whenever a new cause discrepancies between the real-time and recorded-
drum of drawworks drill line is installed, each time the rig mode depth include delay in data transmission due to the
performs a slip-and-cut operation of the drill line, or when data rate, mud type and viscosity, and pipe length. These
the depth is not tracking as expected. Tracking the block factors together affect how fast the data are transmitted
position using a well-calibrated system and using the in-out in real time and, consequently, could put the data at an
slip algorithm, one can track the bit depth and translate it incorrect depth. Even when real-time depth is transmitted
to a depth measurement, using the pipe tally to cross-check correctly, a downhole clock drift can occur that would cause
for any gross error. Any discrepancy detected needs to be asynchronization of surface and downhole clock, hence,
resolved prior to continuing drilling. putting the recorded data at an incorrect depth. (Dashevskiy
Many rigs may have surface sensors already installed et al., 2008). An example is when a tool is initialized on a
as part of the rig equipment or installed by a “third party,” different date than the surface depth-¿le initialization, as
therefore, the rig is the sole provider of the depth-tracking shown in Fig. 10. An error of this type is usually corrected
measurements that is then transmitted to various users via prior to merging the depth and data by applying time offset
WITS or WITSML. to the surface depth and time ¿le.
Pipe-Tally Tracking
Typical LWD depth tracking includes setting the bit
depth at a certain depth before the bit leaves the casing or
liner shoe, and then tracking the bit depth the entire time
in the open hole, up to the point when it reenters the casing
and liner during pull-out of hole. During the up and down
movement of the traveling block, the LWD depth may be off
when compared to the driller’s depth at the end of a stand.
The LWD engineer would adjust, usually in the unit of inches,
and record corrections and offsets on a depth-tracking sheet.
Any adjustments made while drilling reÀects immediately
on the real-time depth. The depth-tracking sheet is used
during processing of memory data (recorded mode) to ensure
the memory data are on depth with real time. However, for
the rig that provides depth-tracking measurements, then no (a)
cross-check of adjustment is possible. The LWD engineer
would have to rely on the depth tracking as provided.
LWD Depth Measurement Errors Table 5—Cause and Effect of Sources of Error on Drillpipe Length
When Below the Rotary Table and Not in Slips
LWD depth measurements are performed at the surface,
where the conditions are not the same as downhole due
to temperature and pressure changes—consequently, the
depth measurement at surface does not represent downhole.
The differences caused by dynamic drilling environments
and changes to the wellbore pro¿le, introduce a dynamic
error in the measurement often making the driller’s depth
shallower than the actual along-hole depth. The dynamic
errors are quanti¿able and must be corrected to improve
the accuracy of the depth measurement. The LWD depth is
also prone to static and systematic errors that may include
equipment, environment, calibration and human errors
(Reistle and Sikes, 1938). Consequently, the LWD engineer
therefore often compares the LWD depth with the driller’s S-shaped well. They also showed that downhole friction,
depth at the end of each joint or stand. and makes minor WOB and heave are next in order of magnitude to the error
adjustments to the depth tracking. On a Àoating rig equipped contribution on stretch. While these errors are acknowledged
with a heave-compensation system, it is common to observe by the industry, they are often ignored, or at best a block-
effects of residual heave and tide on LWD images which if shift correction is used, to correct them using wireline
uncorrected, creates noise on LWD images. depth as the reference. As a recommended practice, the best
method is to apply environmental corrections to the drillpipe
DRILLPIPE-STRETCH CORRECTION and the wireline, then reconcile the difference and assign
uncertainties, as shown in Table 1.
The drillers’ s depth is the sum of the tape measure of the Implementing environmental correction applies the
drillpipe plus BHA components taken at surface conditions. principle that major contributors of pipe stretch in order
However, it is also known that there are inherent errors of magnitude are mechanical and temperature. This is
associated with the driller’s depth due to the environmental undertaken using input of the wellbore trajectory, geometry,
conditions and the physical changes that the pipe and BHA, surface sensor, temperature thermal pro¿le, and torque
components undergo while in the hole, which produce errors and drag. A calibration of torque-and-drag model is necessary
that are unaccounted for. These effects may result in LWD and completed using actual sensor measurements made in
depth being shallower that the absolute depth. In the Gulf different drilling modes and parameters selection; average
of Mexico, depth stretch of up to 70 ft has been observed, sliding-friction factor, average pipe wear, mud weight and
especially in wellbores up to 30,000 ft depth. effective block weight, as shown in Fig. 11. Correction is
Several published papers have discussed the applied to every BHA and drillpipe run in hole. The result is
environmental correction of associated static and a new dynamic corrected depth ¿le that is merged with the
dynamic errors, as well as the impact of the errors on LWD time data to generate corrected LWD depth.
LWD measurement (Lesage et al.,1988; Chia et al., 2006;
Bordakov et al., 2007; Dashevskiy et al., 2008). The main Heave and Tidal Effect
sources of drillpipe depth error identi¿ed are drillstring On Àoating rigs, the effects of heave and tide on the rig
weight, temperature of the pipe, pressure, friction effect do cause changes in the depth. Although such rigs are usually
of pipe on the borehole wall, buckling of drillpipe when in equipped with a heave compensator, sometimes residual
compression, weight on bit, torque on the pipe, and heave, motion (up and down movement) of the rig gets propagated
on Àoating rigs (Chia et al., 2006; Bordakov et al., 2007). into LWD depth systems and causes the data to be placed
Some of the factors have positive effects on the pipe length intermittently at incorrect depth. When the heave effects are
while other have negative effects, hence, some parts of the present, it can be easily seen on LWD images. Figure 12
logs are stretched and some shrink. Table 5 lists the cause shows heave effect on the LWD image and improvements of
and effect of these major sources of error on drillpipe length. the image after ¿ltering out the heave effect. Tidal effects are
Bordakov et al. (2007) demonstrated that drillpipe not usually as severe as heave, but still need to be accounted
weight and thermal expansion contribute up to 80% of for using the location tidal chart.
the total correction seen in a 5,000-m (~16,400 ft) MD
Fig. 11—Torque-and-drag model calibration using surface sensor measurements and parameters selection to correct the LWD depth.
Fig. 13—Time-lapse logs obtained in three passes (P1 to P3) showing depth offsets between each of the different measurements (resistivity, neutron,
and density) resulting from using different drilling parameters.
and to ensure it stayed within the expected rat hole of 40 ft. Case 2
(±25%) from TD. Since the rig did not have the capability Figure 15 presents gamma-ray logs from four wells in
to pull back casing, it was necessary to have an accurate the Gulf of Mexico plotted with driller’s depth shallower
openhole depth in order to determine the actual length of than the stretch-corrected depth. The correction varied
casing to run. Using the 12.25-in. BHA tally, survey data, from 55 to 69 ft in ~30,000 ft of hole, depending on the
surface time data and temperature pro¿le, depth stretch at trajectory. It is important to note that the stretch correction
the openhole LWD total depth was calculated to be ~13 ft. does not alter LWD data, but the LWD depth will be shifted
With this information, a more accurate openhole depth was downwards. The depth correction was applied to assess well
determined and used to plan the length of casing to run. The compartmentalization risk and oil/water contacts across the
casing assembly (made of drillpipe and casing) stretch at ¿eld. This was done by placing pressure and sample points
the corrected openhole total depth was initially modeled to on the most accurate TVD depth possible obtained after
be approximately 18 ft using a friction factor of 0.22. The applying depth correction on the original LWD depth. LWD
result was used with the landing window to back-calculate depths were stretch corrected, then pressure and sample points
the uncorrected landing depth. While running the 11.75-in. were shifted to the corrected depth. The excess pressure plot
casing, the model was updated in semireal time with actual shown in Fig. 16 shows Well 4 to be in a totally different
torque-and-drag surface time data. At 80% of the casing pressure regime than Wells 1 to 3. NMR data also indicated
run, the actual stretch of uncorrected driller’s casing depth that the sand quality was very different. Similar work was
was calculated to be 17.69 ft. (within 2% of the model). reported by Cribbs (2009) at a AAPG Technology Workshop;
The casing was subsequently landed within a window of he showed that a stretch correction of approximately 60 ft
approximately 48 ft (Fig. 14). reduced the risk of compartmentalization.
Fig. 14—Determination of casing landing depth for a deepwater Gulf of Mexico well to within 40 ft (±10 ft) of expected rat hole.
Fig. 15—Gamma-ray logs from four wells from the same ¿eld in the Gulf of Mexico and drilled with different BHAs. Gamma-ray logs plotted with
drillers depth are shown in blue; the stretch-corrected logs are plotted in green.
Fig. 16—Excess pressure plot of four wells in the same ¿eld in the Gulf of Mexico. Well 4 seems to be in a totally different pressure regime than Wells
1 to 3.
CONCLUSIONS
Case 3
Bordakov et al. (2007) have shown that the dynamic The industry is making discoveries and drilling in areas
corrected depth ¿le has application in improving LWD and formations where depth accuracy is now becoming more
image quality. They showed a 20% correction in stratigraphic important in order to maximize production and optimize
thickness computed from two dips picked 52 ft apart from drilling time. Consequently, LWD/MWD is playing a
same image log before and after depth correction (see signi¿cant role in “getting it right the ¿rst time.” Lateral
Fig.17). wellbores are being geosteered in longer boreholes in thinner
Fig. 17—LWD depth correction showing a slight difference in true stratigraphic thickness. Top, LWD density image. Bottom, blue lines are dip planes
and green lines are borehole trajectory (from Bordakov et al., 2007).
layers where depth placement means the difference in a Formation Evaluation by Utilizing Dynamically Corrected
wellbore being economic or not. The density of wellbores Drilling-Derived LWD Depth and Continuous Inclination
being drilled in unconventional ¿elds is increasing— and Azimuth Measurements, Paper SPE-109972, SPE
drilling stacked laterals and “wine rack” patterns, wellbore Reservoir Evaluation & Engineering, 12(1), 137–148. DOI:
10.2118/109972-PA.
placement (vertical, lateral and depth) are essential to
Brooks, A.G., Wilson, H., Jamieson, A., McRobbie, D., and
maintain the desired spacing to optimize production—thus, Holehouse, S.G., 2005, Quanti¿cation of Depth Accuracy,
reducing the risk of borehole collisions is a greater concern Paper SPE-95611 presented at the SPE Annual Technical
to ensure the residual risk is maintained at a non-HSE level Conference and Exhibition, Dallas, Texas, USA, 9–12
while maintaining production. October. DOI: 10.2118/95611-MS.
It is important for the industry to track depth accurately, Chia, C.R., Laastad, H., Kostin, A.V., Hjortland, F., and Bordakov,
implement depth environmental corrections, and assign G.A., 2006, A New Method for Improving LWD Logging
uncertainties to reconcile LWD and wireline depths. As the Depth, Paper SPE-102175 presented at the SPE Annual
industry and ¿elds mature, it is imperative that the accuracy Technical Conference and Exhibition, San Antonio, Texas,
USA, 24–27 September. DOI: 10.2118/102175-MS.
of wellbore placement increases, and reconciling the depth
Cribbs, B., 2009, Practical Wellbore Formation Test Interpretation,
is a major part of improving accuracy. Drillpipe modeling
AAPG Search and Discovery Article 120009. http://www.
results shows different error magnitudes while drilling and searchanddiscovery.com/documents/2009/120009cribbs/
while tripping out. The model shows drillpipe error to be ndx_cribbs.pdf. Accessed May 8, 2019.
highest in vertical wellbores while drilling and lowest in a Dashevskiy, D., Dahl, T., Brooks, A.G, Zurcher, D., Lofts, J.C.,
horizontal well. During trip out, a 3D-S well pro¿le showed and Dankers, S., 2008, Dynamic Depth Correction to Reduce
the worst case, while a vertical well showed the least. Time- Depth Uncertainty and Improve MWD/LWD Log Quality,
lapse logging provides the opportunity to capture data Paper SPE-103094, SPE Drilling & Completion, 23(1), 13–
necessary to effectively correct the drillpipe stretch in any 22. DOI: 10.2118/103094-PA.
drilling activities, i.e., running in hole, on bottom drilling, Grindrod, S.J., Clark, P.J., Lightfoot, D.J., Bergstrom, N., and
Grant, L.S., 2016, OWSG Standard Survey Tool Error Model
pulling out of hole, etc.
Set for Improved Quality and Implementation in Directional
When the drillpipe environmental stretch correction Survey Management, Paper IADC/SPE-178843 presented
is properly modeled and applied, the accuracy of depth at the IADC/SPE Drilling Conference and Exhibition, Fort
measurement increases greatly, allowing for more precise Worth, Texas, USA, 1–3 March. DOI: 10.2118/178843-MS.
wellbore placement. Lesage, M., Falconer, I. G., and Wick, C. J., 1988, Evaluating
Examples demonstrate the impact of depth correction Drilling Practice in Deviated Wells With Torque and Weight
on wellbore position and reservoir evaluation. Accurate Data, Paper SPE-16114, SPE Drilling Engineering, 3(3), 248–
wellbore positioning for better reservoir exploitation, 252. DOI: 10.2118/16114-PA.
positioning the well correctly and optimizing drill time, Lesso, Jr., W.G., and Chau, M.T., 1999, Quantifying Bottomhole
Assembly Tendency Using Field Directional Drilling Data
preventing collision, correcting vertical, and horizontal
and a Finite Element Model, Paper SPE-52835 presented
spacing, etc., are potential applications of drillpipe stretch
at the SPE/IADC Drilling Conference, Amsterdam The
corrections. It is therefore necessary to environmentally Netherlands, 9–11 March. DOI: 10.2118/52835-MS.
correct drillpipe stretch on every drilled wellbore. Maus, S., Nair, M.C., Poedjono, B., Okewunmi, S., Fairhead,
D., Barckhausen, U., Milligan, P.R., and Matzka, J., 2012,
ACKNOWLEDGMENTS High-De¿nition Geomagnetic Models: A New Perspective
for Improved Wellbore Positioning, Paper IADC/SPE-
The authors wish to express appreciation for and 151436 presented at the IADC/SPE Drilling Conference and
acknowledgement of the Bureau of Safety and Environmental Exhibition, San Diego, California, USA, 6–8 March. DOI:
Enforcement (BSEE) and Schlumberger for their permission 10.2118/151436-MS.
Mohammed, N.C., and Terpening, M.E., 2012, High-Con¿dence
to publish the concepts and methods used and the material
Vertical Positioning for Extended Reach Wells, Paper IADC/
contained in this paper. The authors would also like to SPE-151441 presented at the IADC/SPE Drilling Conference
thank and acknowledge the SPWLA technical reviewers for and Exhibition, San Diego, California, USA, 6–8 March.
reviewing the draft manuscript. DOI: 10.2118/151441-MS.
Monterrosa, L.C., Rego, M.F., Zegarra, E., and Lowdon, R., 2016,
Statistical Analysis Between Different Surveying Instruments
REFERENCES to Understand the Reliability of MWD/RSS High Resolution
Surveys and its Effect in Well Trajectory Characterization,
Bordakov, G.A., Kostin, A.V., Rasmus, J., Heliot, D., Laastad, Paper IADC/SPE-178830 presented at the IADC/SPE Drilling
H., and Stockhausen, E.J., 2007, Improving LWD Image and Conference and Exhibition, Fort Worth, Texas, USA, 1–3