Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
Copyright 2007, Society of Petroleum Engineers ducible water saturation is established for these computer-
This paper was prepared for presentation at the 2007 SPE Annual Technical Con- generated rocks. We compare calculated permeability val-
ference and Exhibition held in Anaheim, California, U.S.A., 11-14 November 2007. ues of computer-generated rocks and laboratory measure-
This paper was selected for presentation by an SPE Program Committee follow- ments of core samples to those estimated from different em-
ing review of information contained in an abstract submitted by the author(s). pirical approaches, such as Tixier, Coates, and Timur mod-
Contents of the paper, as presented, have not been reviewed by the Society of
Petroleum Engineers and are subject to correction by the author(s). The material, els. It is found that the latter models cannot be applied to
as presented, does not necessarily reflect any position of the Society of Petroleum general cases of clastic rocks even if their free parameters
Engineers, its officers, or members. Papers presented at SPE meetings are subject
to publication review by Editorial Committees of the Society of Petroleum Engi- are adjusted to fit core measurements. Our simulations
neers. Electronic reproduction, distribution, or storage of any part of this paper also show that spatial distributions of clay minerals and ir-
for commercial purposes without the written consent of the Society of Petroleum
Engineers is prohibited. Permission to reproduce in print is restricted to an ab- reducible water play a fundamental role in establishing an
stract of not more than 300 words; illustrations may not be copied. The abstract accurate correlation between permeability, porosity, and
must contain conspicuous acknowledgment of where and by whom the paper was
presented. Write Librarian, SPE, P.O. Box 833836, Richardson, Texas 75083-3836 irreducible water saturation. Specific deterministic equa-
U.S.A., fax 01-972-952-9435. tions must be established for rock formations that exhibit
distinct grain-size distributions, clay types, structural clay
distributions, and grain cementation.
Abstract
Introduction
Permeability is one of the most important, most spa-
tially variable, most uncertain, and hence least predictable Permeability governs the displacement of fluids through
transport properties of porous media. Various empiri- the pore space of porous media. It is one of the most impor-
cal models, such as Tixier’s, Timur’s and Coates’ equa- tant and least predictable transport properties of porous
tions, are widely used to quantify permeability from well- media in reservoir characterization. Permeability is usually
log calculations of porosity and irreducible water satura- evaluated from core samples and/or well tests. However,
tion. However, these models do not explicitly include the core samples and well-test data are often only available
role played by rock structure, spatial fluid distribution in from few wells in a reservoir while well logs are available
the pore space, wettability, or clay mineral distribution on from the majority of wells. Therefore, accurate and reliable
permeability. We present a pore-scale approach to inves- evaluation of permeability from well-log data embodies a
tigate the influence of these factors on the permeability of significant technical and economic advantage.1
clastic rocks for explicit pore geometries of brine-saturated Various empirical models have been proposed to infer
granular rocks. permeability from well-log data, based on calculations of
Synthetic pore-scale models are constructed to repre- porosity, water saturation, capillary pressure, and forma-
sent granular sands with variable grain-size distributions. tion resistivity factor.2–6 Permeability is estimated via cor-
These models include the structural effects of compaction, relations among other rock petrophysical properties. In
cementation, and distribution of dispersed hydrated clay many cases, there may exist deterministic relationships
minerals. Irreducible water is geometrically distributed on among these properties, but such correlations usually are
grain surfaces of the synthetic rocks. Permeability is calcu- empirically derived for a given formation in a given area,
lated from lattice-Boltzmann flow simulations. A nonlin- are often statistical in nature and, therefore, cannot be
ear relationship between permeability, porosity, and irre- applied to general cases.7
2 T. TORSKAYA, G. JIN, AND C. TORRES-VERDÍN SPE 109878
Tixier:3
φ6
k = 62.5 2 , (2)
Swi
Timur:5
φ4.4
k = 8.58 2 , (3)
Swi (c) Compaction (d) Cementation
and Figure 1— Schematic diagram of the numerical rock genera-
tion processes: (a) grain generation, (b) grain sedimentation,
Coates:7 (c) compaction, and (d) cementation.
φ4 (1 − Swi )2
k = 4.90 4 , (4)
Swi
cannot be reproduced in the laboratory. As an alterna-
where the unit of permeability is Darcy (D). The units of
tive to both experimental determination of rock properties
porosity and irreducible water saturation are expressed in
and petrophysical correlations, we make use of a numerical
terms of fraction of bulk volume and pore space volume,
approach11, 12 in this paper.
respectively. Despite their widespead use, existing mod-
els used to calculate permeability from porosity and irre- We construct synthetic pore-scale models to repre-
ducible water saturation do not explicitly include the role sent granular sands with variable grain-size distributions.
played by rock structure, grain geometry, grain-size distri- These models include the structural effects of compaction,
bution, wettability, and spatial distribution of irreducible cementation, and distribution of dispersed hydrated clay
water in the pore space. minerals. A D3Q19 lattice-Boltzmann algorithm12 is used
Laboratory studies have shown that permeability de- to simulate viscous flow of the single-phase fluid in the
pends on a long list of parameters: porosity, pore size and pore space of computer-generated rock samples. Perme-
shape, pore size distribution, clay content, fluid type, and ability is calculated directly from the simulated velocity
saturation – a nearly overwhelming complexity.7, 8 The field using Darcy’s equation. We compare the permeabil-
objective of this paper is to investigate the influence of ity calculated from computer-generated rocks and labo-
rock microstructure and spatial distribution of clay min- ratory measurements of core samples to those estimated
erals on the permeability of clastic rocks for explicit pore from different empirical models, such as those introduced
geometries of brine-saturated granular rocks. Hydrated by Tixier, Coates, and Timur.
clay minerals are responsible for the presence of excess ir-
reducible water whose spatial distribution in the pore space Numerical Methods
further conditions the geometry of pore throats.9, 10 There- In general, sedimentary rocks originate by deposition of
fore, presence of clays in sands can substantially affect grains followed by compaction and cementation. The lat-
the relationship between porosity, irreducible water satura- ter two processes determine the final pore-space geometry
tion, and permeability in ways that depart from standard and connectivity. A physics-based depositional model11
parametric models such as those of Tixier,3 Timur,5 and serves to reconstruct natural sedimentary rocks, and gen-
Coates7 (see Eqs. (2)-(4)). erates 3D images of the pore space at an arbitrary de-
The macroscopic properties and detailed microstructural gree of spatial resolution. This model provides detailed
information of a porous medium can be obtained from ex- microstructure of the rock, and makes it possible to cal-
periments. However, the associated experiments are of- culate the steady-state velocity field for single-phase fluid
ten time-consuming and expensive. In addition, routine flow. Figure 1 illustrates the simulation procedures used
laboratory testing is not easily applied to damaged core to contruct numerical rock samples with this method.
material or drill cuttings and often reservoir conditions Dispersed clay minerals in sandstones are of three mor-
SPE 109878 PORE-LEVEL ANALYSIS OF THE RELATIONSHIP BETWEEN POROSITY, IRREDUCIBLE WATER . . . 3
0.8 0.8
0.6 0.6
0.4 0.4
0.2 0.2
0.0 0.0
0.1 0.2 0.3 0.4 0.5 0.0 0.1 0.2 0.3 0.4
(a) PU growth (b) PT growth Grain radius (mm) Grain radius (mm)
(a) Type 1 (b) Type 2
Figure 2— Schematic diagram of different growth patterns of
pore-lining clay minerals (blue color) on grain surfaces (gray 0.8 0.8
color). Clay minerals distributed uniformly on grain surfaces
grow with different growth rates: (a) a uniform rate in all 0.6 0.6
directions, and (b) a growth rate larger in the narrow pore 0.4 0.4
region than in other regions.
0.2 0.2
0.0 0.0
0.05 0.10 0.15 0.20 0.25 0.30 0.20 0.25 0.30 0.35
phological types: (1) pore lining, (2) pore bridging, and Grain radius (mm) Grain radius (mm)
(3) discrete particles.13, 14 Pore-lining clays are attached (c) Type 3 (d) Type 4
to pore walls to form a relatively continuous and thin clay
uniformly in all directions on grain surfaces; and (2) PT Grain radius (mm)
(e) Type 5
growth: clay minerals preferentially grow in narrow grain-
contact regions. Figure 3— Grain-size distributions used to generate different
Clay shell (region of blue color in Figure 2), which is numerical rock samples. Total volume of all grains with the
composed of exceedingly fine-grained particles, often ex- same size is expressed in terms of fraction of total solid volume
hibits very high porosity. Water held in this shell is respon- in the rock. Only one rock sample is constructed from each
sible for presence of irreducible water saturation. Different of the types 1, 2, and 4, respectively. Type 3 is used to
volumes of the clay shell correspond to different values of generate two rock samples with different porosities. There
irreducible water saturation. In addition, because pores are 371 samples with different porosities and clay morphology
constructed from the same grain-size distribution, type 5.
and pore channels are equally small, for all practical pur-
poses clay shells exhibit zero permeability. Therefore, we
model clay shells as highly porous, but impermeable layers different degrees of compaction and cementation. Total
in numerical simulations. A no-flow boundary condition is volume of all grains with the same size is expressed in terms
enforced at fluid-solid interfaces. of fraction of total solid volume in the rock. Only one rock
We use the D3Q19 lattice-Boltzmann algorithm12, 17 to sample originates from each of the grain-size distributions
simulate viscous flow of a single-phase fluid in the pore of types 1, 2, and 4, respectively. They are labeled as A,
space of computer-generated rock samples. In this pa- B, and E, respectively. Two additional rock samples, C
per, we use the improved incompressible lattice-Boltzmann and D, are constructed from the same grain-size distribu-
model18 together with a pressure boundary condition19 tion, type 3, except that they exhibit different values of
enforced on the inlet and outlet faces. Simulations are porosity. The grain-size distribution, type 5, is used to
performed individually for each of the x−, y−, and generated 371 rock samples with different porosities and
z−directions. Absolute permeability is derived directly clay morphology, labeled as samples F. Initial properties
from the simulated velocity field using Darcy’s equation. of these samples are listed in Table 1. Note that, due to
Effective absolute permeability is defined as the arithmetic the number of samples F, we did not list their properties
average of the calculated absolute permeabilities in the x−, in Table 1.
y−, and z−directions. Only effective permeability is con- We assume that clay minerals in samples A-E grow uni-
sidered in this paper. formly in all directions on grain surfaces and exhibit PU
growth patterns, while clays in samples F form accord-
Results and Analysis ing to two different patterns: PU growth and PT growth.
Five different grain-size distributions, shown in Fig- Water held in clay shells is responsible for the saturation
ure 3, are used to generate numerical rock samples with of irreducible water. Due to the fact that clay shells are
4 T. TORSKAYA, G. JIN, AND C. TORRES-VERDÍN SPE 109878
k (Darcy)
Sample φt (%) Ra (mm) k (Darcy)
A 27 0.42 98.1 5
B 34 0.21 82.6
C 29 0.24 45.4 0
D 40 0.24 199.0 0.4
E 25 0.25 40.1 0.4
F - 0.10 - 0.2
0.2
0.0 0.0 φ
Swi
3
10 (a) PU pattern
Sample A
Sample B
Sample C
Permeability (Darcy)
2 15
10 Sample D
Sample E
k (Darcy) 10
1
10 5
0
0
0.4
10 0.4
0.0 0.2 0.4 0.6 0.8 0.2
0.2
Irreducible water saturation, Swi
0.0 0.0 φ
Figure 4— Variations of calculated permeability for different Swi
values of irreducible water saturation, Swi (samples A-E). (b) PT pattern
2 2
10 10
this study, Eq. (5) SIM
−2
10
−2
10
−4
−2 0 2
10 −4 −2 0 2
10 10 10 10 10 10 10
Permeability, Darcy (LBS) Permeability, Darcy (Numerical/Core samples)
(a) PU pattern
Figure 7— Permeability from lattice-Boltzmann simulations
2
10 on computer-generated rock samples (SIM) and laboratory
this study, Eq. (6) measurements on core samples (EXP), together with the cor-
Tixier 1949, Eq. (2) responding values calculated via correlation, Eq. (6).
Timur 1968, Eq. (3)
Permeability, Darcy (COR)
0 Eq. (4)
10
−2
10
−2 0 2 −2
10 10 10 10
Permeability, Darcy (LBS)
(b) PT pattern
−4
Figure 6— Calculated permeabilities from lattice-Boltzmann 10 −4 −2 0 2
simulations (LBS) compared to those calculated with differ- 10 10 10 10
ent empirical models (COR), Eqs. (2)-(6), for all computer- Permeability, mD (EXP)
generated rock samples with different clay-growth patterns (a)
PU pattern, and (b) PT pattern. Figure 8— Measured permeabilities from core samples (EXP)
compared to those estimated with different empirical models
(COR), Eqs. (2)-(6). Note that experimental data are from
basin of the world may not be applicable to other forma- Salazar et al.’s paper20 (Figure 1).
tions with a different diagenetic history.
Figure 6 compares permeability values rendered by
lattice-Boltzmann flow simulations to those estimated with To further verify and compare empirical models dis-
different empirical models, Eqs. (2)-(6), for all computer- cussed previously, we have chosen core data from a tight-
generated rock samples with different clay morphologies. If gas reservoir consisting of sand units of turbiditic ori-
we assume that the calculated permeabilities with lattice- gin.20, 21 Core samples were acquired from a shallow inter-
Boltzmann simulations are ‘true’ values of these numerical val of A1-A5 m and a deep interval B1-B5 m. Laboratory
samples, empirical models, such as Tixier’s, Timur’s, and measurements on these samples, including porosity, irre-
Coates’ (Eqs. (2)-(4)) generally underestimate the actual ducible water saturation, and permeability, were well con-
values of permeability. ducted and documented. Figure 7 displays the measured
One also observes that the spatial distribution of clays permeability of core samples, together with the calculated
in the pore space (or irreducible water) has a significant permeability of computer-generated rock samples F with
effect on the accuracy of permeability estimates obtained PT clay morphology. These calculated/measured values
with different empirical equations. Empirical models give a are compared to the estimated values determined from the
better estimate of permeability for samples with a PT clay developed correlation in this paper, Eq. (6). Our model
growth pattern than for those with a PU growth pattern. overestimates the permeability of core samples. This dis-
6 T. TORSKAYA, G. JIN, AND C. TORRES-VERDÍN SPE 109878
A2
Depth, m
A3
A4
A5
Figure 9— Measured permeability of core samples compared to their estimated values from different empirical models: Tixier:
Eq. (2), Timur: Eq. (3), Coates: Eq. (4), and this study: Eq. (6). Core samples were acquired in the shallow interval of A1-A5 m.
The length of each tick mark in depth is 5 m.
crepancy may be caused by different grain size distribu- shallow interval A1-A5.
tions and different types of clays present in the core sam- The above discrepancy is due to different values of free
ples and computer-generated samples. It is well known parameters included in Eqs. (2)-(6). Table 2 lists all
that different clay morphologies significantly affect rock the free parameters together for comparison. Compared
porosity/permeability, capillary pressure curve, and asso- to values in Timur’s (or Coates’) model, we observe that
ciated pore-size distributions in various ways.14 Different the value of b in Tixier’s model is larger (the difference
values of these parameters in Eqs. (5) and (6) further con- is about 2), whereas the coefficient a is also one order of
firm the effect of the spatial distribution of irreducible wa- magnitude larger. Average values of porosity and irre-
ter on specific parameters included in the calculations. ducible water saturation of core samples studied in this
Figure 8 shows the overall comparison of permeabil- paper are about 10% and 50%, respectively. Therefore,
ity between laboratory measurements on core samples and Tixier’s model would underestimate the permeability by
those estimated with empirical models, Eqs. (2)-(6). We approximately one order of magnitude, as reflected by the
observe that Timur’s model gives the best estimate of per- plots of Figures 9 and 10. Similarly, our equation (PT pat-
meability, while Tixier’s and Coates’ equations underesti- tern) yields larger values of a (about 2 order of magnitude
mate the permeability, and our correlation, Eq. (6), over- larger) and b (the difference is about 1). It also yields a
estimates the measured permeabilities. smaller value of c (the difference is about 1). These differ-
In Figures 9 and 10, we observe more clearly the differ- ences of parameters in our equation contribute to overes-
ence between permeability estimates for different models, timate the permeability of core samples by approximately
where permeability is plotted as a function of depth. The one order of magnitude.
length of each tick mark in depth is 5 m. Timur’s pre- However, we note that the fact that Timur’s model
dictions of permeability are in excellent agreement with yielded a better estimate of permeability for core sam-
laboratory measurements for both depth intervals, while ples does not indicate that it is superior to other empirical
our model overestimates the permeability approximately models and that this model gives the best estimates of
by one order of magnitude, and Tixier’s model underesti- permeability for all types of rock formations. Moreover,
mates it by one order of magnitude. Coates’ predictions we emphasize that our equations (5) and (6) were derived
nearly match experimental data in the deep depth inter- from computer-generated samples, which exhibit the same
val B1-B5, but underestimate slightly their values in the uniform grain size distribution. Our approach to assign the
SPE 109878 PORE-LEVEL ANALYSIS OF THE RELATIONSHIP BETWEEN POROSITY, IRREDUCIBLE WATER . . . 7
B2
Depth, m
B3
B4
B5
Figure 10— Measured permeability of core samples compared to their estimated values from different empirical models: Tixier:
Eq. (2), Timur: Eq. (3), Coates: Eq. (4), and this study: Eq. (6). Core samples were acquired in the deep interval of B1-B5 m.
The length of each tick mark in depth is 5 m.
Our correlations were derived from computer-generated liburton Energy Services, Hydro, Marathon Oil Cor-
rocks, in which we used the same uniform grain-size distri- poration, Mexican Institute for Petroleum, Occidental
bution and spherical grains. A simple but flexible cement- Petroleum Corporation, Petrobras, Schlumberger, Shell
growth algorithm was used to geometrically distribute ir- Inter national E&P, Statoil, TOTAL, and Weatherford.
reducible water on grain surfaces. This approach ignores The authors acknowledge the Texas Advanced Comput-
many complications, and of course, may not represent the ing Center (TACC) at The University of Texas at Austin
real distribution of irreducible water in the pore space. We for providing high performance computing resources that
expect that all these factors will play a significant role in have contributed to the research results reported in this
establishing a correlation between permeability, porosity, paper.
and irreducible water saturation. Therefore, we emphasize
that the correlation developed in this paper is only valid References
for our numerical rocks and is not universal. [1] S. Ameri, D. Molnar, S. Mohaghegh, and K. Aminian.
It is well known that permeability is dominated by the Permeabilitty evaluation in heterogenous formations us-
smallest portions (pore throats) of flow channels in the rock ing geophysical well logs and geological interpretations. In
pore space, while porosity and water saturation are con- SPE Western Regional Meeting, Anchorage, Alaska, 1993.
trolled by the volume of large pores, not by pore throats.
[2] G. E. Archie. The electrical resistivity log as an aid in
Irreducible water saturation may be closely related to the determining some reservoir characteristics. Trans. AIME,
solid surface area of a rock. Hence, correlations for per- 146:54–62, 1942.
meability may be inherently limited in their accuracy and
reliability when assuming specific correlations with poros- [3] M. P. Tixier. Evaluation of permeability from electric-log
resistivity gradients. Oil and Gas Journal, June 16, 1949.
ity and water saturation, or with any other rock property,
that is not strongly affected by the pore throats of porous [4] M. R. J. Wyllie and W. D. Rose. Some theoretical con-
media. These factors explain why empirical permeabil- siderations related to the quantitative evaluation of the
ity models are not universal; different predictive equations physical charateristics of reservoir rock from electrical log
must be established for different types of rocks. Therefore, data. Petroleum Transactions, AIME, 189:105, 1950.
it is necessary to address these factors suitably when es- [5] A. Timur. An investigation of permeability, porosity, and
tablishing a correlation among rock properties to calculate residual water saturation relationships for sandstone reser-
permeability. voirs. The Log Analyst, 9(4):8–17, 1968.
In addition, empirical permeability models, such as [6] George R. Coates and J. L. Dumanoir. A new approach
Tixier’s, Timur’s and Coates’, relates the permeability to improved log derived permeability. The Log Analyst,
of a porous medium to its porosity with only one con- XV(1):17–29, 1974.
stant parameter for all values of porosity. Laboratory
[7] Schlumberger. Log Interpretation Principles/Applications.
measurements performed on Fontainebleau sandstones22
Houston, 1989.
have shown that the correlation between permeability and
porosity has two different families (low and high porosities) [8] Gray Mavko and Amos Nur. The effect of a percola-
for which the correlation law has the same mathematical tion threshold in the Kozeny-Carman relation. Geophysics,
shape, but not the same characteristic parameters. There- 62(5):1,480–1,482, 1997.
fore, we also do not expect that a simple correlation be- [9] Guodong Jin, Carlos Torres-Verdı́n, Sarath Devarajan,
tween permeability, porosity, and irreducible water satura- Emmanuel Toumelin, and E. C. Thoms. Pore-scale analy-
tion would work well for all rock formations with different sis of the Waxman-Smits shaly-sand conductivity model.
values of porosity. Petrophysics, 48(2):104–120, 2007.
Our fully-explicit pore-scale geometrical approach en- [10] Guodong Jin, Carlos Torres-Verdin, F. Radaelli, and
ables one to study the sensitivity of permeability to amount E. Rossi. Experimental validation of pore-level calcula-
and spatial distribution of hydrated clay minerals and their tions of static and dynamic petrophysical properties of
exchange cations, brine salinity, and irreducible water sat- clastic rocks. In SPE Annual Technical Conference and
uration for clastic rocks with variable grain-size distribu- Exhibition, Anaheim, California, USA, 2007.
tions. Compared to laboratory measurements, this numer- [11] Guodong Jin, Tad W. Patzek, and Dmitriy B. Silin.
ical approach is particularly attractive, due to its low cost Physics-based reconstruction of sedimentary rocks. In SPE
and high speed of calculation. Our study is but a first step Western Regional/AAPG Pacific Section Joint Meeting,
toward applying pore-scale models to well-log interpreta- Long Beach, California, USA, 2003.
tion. [12] Guodong Jin, Tad W. Patzek, and Dmitriy B. Silin. Direct
prediction of the absolute permeability of unconsolidated
Acknowledgements and consolidated reservoir rock. In SPE Annual Technical
The work reported in this paper was funded by the Conference and Exhibition, Houston, Texas, USA, 2004.
University of Texas at Austin’s Research Consortium on [13] Michael D. Wilson and Edward D. Pittman. Authigenic
Formation Evaluation, jointly sponsored by Anadarko, clays in sandstones: recognition and influence on reservior
Aramco, Baker Atlas, BHP Billiton, BP, British Gas, properties and paleoenvironmental analysis. Journal of
ConocoPhillips, Chevron, ENI E&P, ExxonMobil, Hal- Sedimentary Petrology, 47(1):3–31, 1977.
SPE 109878 PORE-LEVEL ANALYSIS OF THE RELATIONSHIP BETWEEN POROSITY, IRREDUCIBLE WATER . . . 9