Study of Vuggy Carbonates Using NMR
and X-Ray CT Scanning
I. Hidajat, SPE, and K.K. Mohanty, SPE, U. of Houston, and M. Flaum, SPE, and G. Hirasaki, SPE, Rice U.
Summary
Most existing nuclear magnetic resonance (NMR) permeability
correlations for carbonates assume that vugs do not contribute to
permeability. The objective of this work is to improve permeability
estimation from NMR responses for carbonates by including vug
connectivity. Six carbonate samples from a west Texas field were
studied. NMR T2 measurement, mercury porosimetry, thin-section
imaging [using optical microscopy and a scanning electron micro-
scope (SEM)], computerized tomography (CT) scanning, and elec-
trical measurements were conducted. The permeability of the
samples studied is controlled by the channels that connect vugs.
Capillary pressure and NMR T2 measurements show multimodal
pore-throat and pore-body-size distributions. A modified Chang
model, which includes the tortuosity factor, is proposed and proven to
yield a better permeability prediction than the original Chang model.
It is also shown that for the samples studied, the tortuosity can be
estimated from NMR T2 distribution, hence allowing permeability
prediction from the NMR T2 distribution alone.
Introduction
More than 50% of the world’s hydrocarbon reserves are in car-
bonate formations.1 However, estimating petrophysical properties
from NMR measurements in carbonate rocks has always been a
bigger challenge than in sandstone formations. Carbonates are
characterized by different types of porosity and complex pore-size
distributions. Because of their reactive nature, carbonates undergo
a more complicated post-depositional diagenesis compared to si-
liciclastic sandstones. The diagenesis process includes cementa-
tion, dissolution, dolomitization, recrystallization, and evaporite
mineralization.2 Carbonates are also sensitive to the microorgan-
ism activity in their depositional environment. Depending on
whether they are grain- or mud-supported, carbonates can be clas-
sified into grainstone (no mud), packstone, wackstone, and mud-
stone (mud-dominated).2,3
Porosity in carbonate rocks can be categorized into three dif-
ferent types of porosity: intragranular porosity, intergranular po-
rosity, and vugs.2 Intragranular porosity is the porosity inside the
grain, and intergranular porosity is the porosity between the grains.
A vug is defined as the pore space that is significantly larger than
grains or crystals (or that is within grains or crystals).2 In more
popular terms, a vug is a pore that is large enough to be visible to
the naked eye. Vuggy pore space can be subdivided into separate
vugs and touching vugs based on vug interconnections.2,4
A permeability correlation from NMR T2 measurement is given
by Chang et al.5:
k = 4.75 共␾750兲4 共T2lm,750兲2, . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (1)
where ␾750 is the porosity and T2lm,750 is the logarithmic mean of
T2 of the pore space consisting of pores with T2 less than 750 ms.
The correlation assumes that vugs do not contribute to the flow
and, hence, the contribution of the vugs in the T2 (which is above 750
ms) is not used. This may not always be the case; in some instances,
vugs may be connected and may contribute to the permeability.
Copyright © 2004 Society of Petroleum Engineers
This paper (SPE 88995) was revised for publication from paper SPE 77396, first presented
at the 2002 SPE Annual Technical Conference and Exhibition, San Antonio, Texas, 29
September–2 October. Original manuscript received for review 14 January 2003. Revised
manuscript received 7 June 2004. Paper peer approved 12 July 2004.
October 2004 SPE Reservoir Evaluation & Engineering
In this paper, we study six vuggy carbonate samples from a
west Texas field using NMR, X-ray CT scanning, and routine core
analysis (porosity, air permeability, mercury capillary pressure,
thin-section imaging, and formation factor). The paper is orga-
nized into four main sections: core-characterization procedure,
core-analysis results, permeability correlations from capillary pres-
sure and NMR, and conclusions.
Core-Characterization Procedure
Sample Selection. Six vuggy carbonate rock samples from a west
Texas field were selected. Before the samples were plugged and
cleaned, they were scanned by an X-ray CT scanner to identify
regions to be plugged. The samples were cleaned with toluene, chlo-
roform/methanol azeotrope, and methanol in a Dean Stark extraction
apparatus. The diameters of the plugs were 1 to 1.5 in. The porosity
and air permeability for the cleaned plugs were measured.
Mercury Capillary Pressure. Endpieces of the samples were
dried at approximately 100°C. They were evacuated to less than 50
microns vacuum, and mercury was injected over 117 pressure
steps ranging from 1.64 to 60,000 psia.
Thin Section. The thin section was viewed by using an optical
microscope. The picture was captured by using a CCD camera, and
the video output signal was sent to a PC computer by using a
frame-grabber card. The picture was then segmented into pore
space and solid by using Crabtree’s algorithm.6 Several contiguous
images were taken for the same sample and then composed to-
gether to form a larger image of approximately 1 to 2 cm coverage.
The resolution of each pixel is 10 ␮m. The endpieces or the stubs
of the thin sections were also studied by using SEM in the back-
scattered electron (BSE) mode.
NMR T2 Measurement. The NMR T2 measurement was conducted
in a Maran Ultra* low field NMR spectrometer using the Carr-
Purcell-Meiboom-Gill (CPMG) pulse sequence. The following pa-
rameters were used: number of scans⳱32, echo spacing⳱600
␮s, and number of echoes⳱10,240. The samples were saturated
with 1 wt% NaCl brine.
3D Porosity Distribution From CT Scanner. The dry and brine-
saturated samples were CT scanned at 2-mm intervals in the z-
direction. Then, the porosity distribution was calculated from the
CT-number difference between brine-saturated and dry core. The
CT scanner used was a Technicare Deltascan** 2060, and the
scanner was mounted vertically. The pixel resolution is 254×254
␮m in the xy plane and 2 mm in the z-direction. The parameters
used were 8-second scanning time, 120 kV, 75 mA, and 12.5 cm
scan diameter.
Tracer Experiment. A 12 wt% NaI in 1 wt% NaCl brine was
injected from the bottom into the brine-saturated core. CT scans
were taken during brine displacement. The NaI outlet concentra-
tion was measured by using an online microconductivity meter.
Swr Experiment. A 15 wt% iodo-decane in decane solution was
injected from the top into the brine-saturated core. CT scans were
done only at the end of the experiment. The brine and oil satura-
tions inside the core were calculated from the mass balance.
*Trademark of Resonance Instruments Ltd., Witney, U.K.
**Trademark of Johnson & Johnson USA.
365
Sor Experiment. Brine was injected from the bottom into the core
at residual water conditions. CT scans were taken only at the end
of the experiment. The final brine and oil saturations inside the
core were calculated from the mass balance.
Formation Factor. Electrical conductivity of brine-saturated samples
was measured, from which formation factors were calculated.
Core Analysis
Porosity and Permeability. Table 1 lists the measured perme-
ability, porosity, porosity from CT scans, rock descriptions, and
binarized thin-section images of the six cores. Samples 2a and 2b
366
are from the same well, as are samples 4a and 4b. The core
samples can be divided into two groups: low permeability and high
permeability. The porosity range for low-permeability samples is 6
to 7.4%, and the permeability is less than 0.1 md. The porosity
range for high-permeability samples is 15.2 to 21.8%, with per-
meability in the range of 56 md to more than 2,000 md. Fig. 1
shows the 3D porosity distribution from CT scanning. The cross-
sectional average porosity was calculated, and Fig. 2 shows the
average porosity variation along each core length. There is signifi-
cant porosity variation along the core length for each sample com-
pared to a Berea sample (Fig. 2). Hicks et al.7 also observed
similar porosity variation for different carbonate rocks.
October 2004 SPE Reservoir Evaluation & Engineering

Fig. 1—Cross-sectional porosity distribution from CT scans.
The spatial dependence of the porosity is characterized by
the semivariogram
兺共␾ − ␾
1 section images) leading to high permeability. The intergranular
SV共h兲 = i+h 兲 ,
2
. . . . . . . . . . . . . . . . . . . . . . . . . . . (2)
2N共h兲 i pore space controls the flow path in these two samples. Both
where SV is the semivariogram, h is the lag, N(h) is the number of
pairs of observed data points separated by a lag of h, and ␾ is the
porosity. The semivariograms were calculated from the x-z plane (the
middle slice in the flow direction) of the CT images and are plotted
in Fig. 3. The semivariograms are fitted by an exponential model:
冋
SV共h兲 = c 1 − exp − 冉 冊册 3h
a
, . . . . . . . . . . . . . . . . . . . . . . . . . . . (3)
where c is the sill and a is the range at which the semivariogram
reaches 95% of the sill.8 The fitting parameters c and a are given
in Table 2. Samples 3 and 4b have the highest sill or variance. The
heterogeneity of these samples is found to be in the range of 0.5 to
1.5 cm.
Rock Descriptions From the Thin Section. Sample 1 is a fine
peloidal wackstone/packstone dolomite. The matrix is filled with
anhydride, which reduces the permeability. From the SEM image
(Fig. 4), the matrix has fine and rather uniform crystals, with a
pore opening between the crystals. The vugs are small and mostly
less than 0.5 mm. A few vugs have a size of approximately 1 mm.
Vugs occur only in some layers, and the other layers are almost
free of vugs.
October 2004 SPE Reservoir Evaluation & Engineering
Fig. 2—Porosity variation along core length.
Samples 2a and 2b have very large crystals. The crystals are so
large that they grow very tight, leaving almost no intercrystalline
porosity. The SEM image (Fig. 4) shows that in between the large
crystals, there are sucrosic (or sugar-like) crystals, which have
significant pore openings. Both samples are very vuggy, and vugs
of the size between 1 and 5 mm are present in these two samples.
Vugs are interconnected through the intercrystalline porosity.
Thus, the microporosity in the matrix controls the permeability for
samples 1, 2a, and 2b.
Sample 3 is a peloid/fusulinid packstone/wackstone dolomite.
The matrix has very little intergranular porosity. The matrix is also
quite heterogeneous. As seen from the SEM image (Fig. 4), the
matrix also has very large crystals with sucrosic crystals in be-
tween. Also as seen in Fig. 1, there are patches of the lower-
porosity region in the matrix system. Some touching vugs and
fractures are observed from the thin section. Some vugs are con-
nected by microfractures. Anhydride occurs in the matrix and in
the fractures. This sample has several moldic vugs; the vug sizes
are approximately 1 to 3 mm. The vugs are distributed more or less
uniformly throughout the sample.
Samples 4a and 4b are very clumpy in appearance; the depo-
sitional texture cannot be seen. They look like breccia or an ag-
glomerate of different rocks. They may have been formed from
cave debris that recrystallized. From SEM images (Fig. 4), it is
found that the matrix of sample 4a is surprisingly tight, but both
samples have a large amount of intergranular porosity that forms
an interconnecting flow network (as observed from the thin-
samples are also quite vuggy, with vug sizes approximately 1 to 5
mm. Sample 4b seems to be more heterogeneous than 4a. In Fig.
1, sample 4b has patches of both high and low porosity.
Mercury Capillary Pressure. Fig. 5 shows the mercury capillary
pressure curves for all samples. Assuming the capillary bundle
model, the pore-throat size can be related to capillary pressure
injection by
2␴ cos ␪
Pc = , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (4)
rt
where Pc is the capillary pressure, ␴ is the surface tension (480
dynes/cm), ␪ is the contact angle (40°), and rt is the pore-throat
radius. This model neglects the issue of pore accessibility. The
pore-throat-size distributions calculated from Eq. 4 are shown in
Fig. 6.
Sample 4b has the lowest capillary pressure, and sample 1 has
the highest capillary pressure, which correspond to the highest and
lowest permeability samples, respectively (Fig. 5). The capillary
pressure curves for those two samples are also similar to those of
monomodal-porosity rocks. Fig. 6 shows that sample 1 actually
has a monomodal pore-throat-size distribution, and the throat sizes
are very small, in the range of 0.01 to 1 ␮m. Sample 4b has a broad
pore-throat-size distribution that also can be interpreted as a bi-
367
 Fig. 3—Porosity semivariograms.
modal size distribution, in which the two modes almost overlap
with each other (Fig. 6). The pore throats in sample 4b are large (in
the range of 1 to 100 ␮m).
The capillary pressure curves for samples 2a and 2b show the
existence of vugs that are connected to the surface (corresponding
to low capillary pressure), and then the curves steeply increase to
a high-capillary-pressure region, which corresponds to small throat
sizes in the matrix (Fig. 5). Both samples have very low perme-
368
ability; hence, the vugs in these two samples do not contribute to
the flow conductance. There are two distinct modes in the pore-
throat-size distribution; the smaller is on the order of 0.1 ␮m, and
the larger is on the order of 100 ␮m (Fig. 6).
The capillary pressure curve for sample 3 is similar to those of
samples 2a and 2b. Access to 50% of the pore volume is controlled
by large pore throats or pore throats located at the surface (Fig. 6).
NMR T2 Response. NMR T2 response provides the pore-body-
size distribution of brine-saturated samples (Fig. 7). Assuming the
fast diffusion regime and cylindrical pore shapes, the T2 response
is related to the pore body size by9
1 1 2
= + ␳ , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (5)
T2 T2B rb
where T2B is the brine bulk T2, ␳ is the surface relaxivity, and rb
is the pore-body radius.
The surface relaxivity (Table 3) is estimated by matching the
median of the cumulative distribution of rt computed from Pc and
rb computed from NMR T2. This method does not account for the
pore-body-to-throat ratio. The average value for samples 4a and 4b
is taken as the best estimate because these samples have the best
pore-throat/body correlation. This average surface relaxivity is
6.75 ␮m/s.
October 2004 SPE Reservoir Evaluation & Engineering
 Fig. 4—SEM images of matrix.
All the samples exhibit a multimodal T2 response, including the
bulk brine response, meaning that all the samples have large vugs
(Fig. 7). By comparing pore-body-size distributions with pore-
throat-size distributions (Fig. 8), it is observed that in higher-
permeability samples (samples 3, 4a, and 4b), the body and the
throat are better correlated than in lower-permeability samples
(samples 1, 2a, and 2b).
Tracer Experiment. Fig. 9 shows the in-situ tracer-concentration
profile for some of the samples after 1 PV tracer injection. Sample
2a has a preferential flow path at 45° to the horizontal in the
vertical plane. Sample 3 has a very distinct preferential flow path;
the tracer transport into the other areas is more of a diffusive
transport. Sample 4a has a more uniform profile at the bottom part
of the core; the flow then tends to go through the center at the top
part of the core. Sample 4b also has unswept areas at the bottom
left corner and top right corner. All these confirm the heterogeneity
of the carbonate rocks we studied at the centimeter scale.
The tracer effluent concentration profiles for the high-
permeability samples are plotted in Fig. 10. The very early break-
through followed by a long tail is characteristic of heterogeneous
carbonate systems. This behavior also seems to indicate the exis-
tence of a sample-spanning high-permeability streak in a tight
matrix.10 This also explains the early water breakthrough with high
oil saturation remaining in large unswept areas, which is usually
observed in many of the carbonate systems.10–13
The Coats and Smith14 dead-end pore model was used to ac-
count for early breakthrough and long tail on experimentally ob-
served effluent concentration profiles. The model divides the pore
space of the porous medium into flowing and nonflowing (or stag-
nant) fractions. The model is described as
⭸C ⭸C* ⭸2C ⭸C
fd + 共1 − fd兲 = Kl 2 − u . . . . . . . . . . . . . . . . . . . . . (6)
⭸t ⭸t ⭸x ⭸x
and
⭸C*
共1 − fd兲 = K⬘共C − C*兲, . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (7)
⭸t
where fd is the flowing fraction, C is the tracer concentration in the
flowing part, C* is the tracer concentration in the stagnant part, Kl
is the longitudinal dispersivity, u is the Darcy velocity, and K⬘ is
the mass-transfer coefficient. The analytical solution is given by
Brigham15 in an integral form. The experimental result is fitted by
three parameters: fd, Kl, and K⬘.16 The fitting parameters are listed
in Table 4.
Sample 3 has the largest stagnant fraction, which is consistent
with the CT images shown in Fig. 9. Sample 4b has the largest
longitudinal dispersivity coefficient, which also means that this
October 2004 SPE Reservoir Evaluation & Engineering
Fig. 5—Mercury capillary pressure curves.
sample has the longest mixing zone because of its heterogeneity.
The small mass-transfer coefficient for sample 3 seems to indicate
that this sample has long and narrow dead-end pores.
Swr and Sor. Fig. 11a shows the oil saturation (Soi) at Swr after the
drainage process, and Fig. 11b shows the residual oil saturation
(Sor) after the imbibition experiment for some of the samples.
Table 5 gives Soi , Sor, and the percentage of oil recovered by
waterflooding. For samples 2a, 4a, and 4b, oil fills the area outside
preferential flow paths seen in Fig. 9 at Soi because the oil injection
is from the top. In brine injection, oil is recovered mostly from the
preferential flow path. For sample 3, oil fills mostly the preferen-
tial flow path area at Soi. Sample 4b shows low oil recovery,
although it has the highest porosity and permeability. Both mass-
balance and oil-saturation calculations from the CT number give
approximately 20% oil recovery consistently. We suspect that this
sample is more oil-wet than the other samples.
Formation Factor. The formation factor is obtained by measuring
the resistivity of a brine-saturated rock and comparing it with the
resistivity of the brine itself. Assuming a bundle-of-tube model,
the tortuosity also can be estimated from formation factor
as follows:
Rrock+brine 1 ␶
F= = m = , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (8)
Rbrine ␾ ␾
where R is the resistivity, ␾ is the porosity, m is Archie’s cemen-
tation factor, and ␶ is the tortuosity.
Table 6 shows the measured formation factor, the cementation
factor, and the calculated tortuosity. The cementation factors are
larger than 2, which is the expected value for carbonates. It is also
observed that the formation factor and the tortuosity correlate well
with the permeability. The intergranular porosity and the vuggy
porosity can be estimated by using Myers’ pore-combination
model (PCM),17 which is given as
冉 冊 冉 冊
1 1.91 ␾int 1.11
F= , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (9)
␾int ␾
369
 Fig. 6—Pore-throat-size distribution from capillary pressure.
where ␾int⳱␾–␾vug is the intergranular porosity. The intragranu-
lar porosity is absent in these samples. The calculated results are
tabulated in Table 7. The results agree very well with the esti-
mated vuggy porosity calculated by using a 750-ms vugular cutoff
value in the NMR T2 distribution. Hence, it also means that the
formation factor of these samples can be estimated from the NMR
T2 distribution. Fig. 12 shows the relationship between the cemen-
tation factor and the ratio of vuggy porosity to the total porosity.
The cementation factor m increases with increasing vuggy poros-
ity. The relationship obtained from the linear regression is
m = 0.51 冉 冊
␾vug
+ 1.91. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (10)
␾
Permeability Correlation From
Capillary Pressure
The existing correlations to predict permeability by Winland-
Pittman,18 Swanson,19 and Katz-Thompson20 are tested for these
370
samples. The permeability models based on capillary pressure use
a characteristic throat size that governs the flow at the percolation
threshold of the porous medium, but each model uses a different
method to estimate that characteristic throat size. Winland devel-
oped a power-law model that relates permeability with porosity
and pore-throat radius:18
k = 17.6␾1.470r351.701, . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (11)
where r35 is the pore radius corresponding to 35% mercury satu-
ration. Swanson19 proposed the following correlation between the
permeability and mercury capillary pressure:
k = 399 冉冊Sb
Pc
1.691
max
, . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (12)
where Sb is the mercury saturation in percent bulk volume and Pc
is the mercury capillary pressure. The maximum is taken over all
possible values of Sb. Starting with the percolation concept, Katz
October 2004 SPE Reservoir Evaluation & Engineering
Fig. 7—NMR T2 distribution and trimodal Weibull distribution fit. The dots are NMR experimental data. The solid line is trimodal
Weibull distribution. The dashed line with the lowest T2 is the first mode, the dashed line with the medium T2 is the second mode,
and the dashed line with the highest T2 is the third mode of the trimodal Weibull distribution.

and Thompson20 have related the permeability of a porous medium where lc is the size of the smallest throat invaded by a nonwetting
to a length scale lc: phase at the percolation threshold; lc can be estimated from the
inflection point of the mercury-injection curve.
1 l c2 The predicted permeabilities from Eqs. 11 through 13 are tabu-
k= , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (13)
226 F lated in Table 8. The Winland-Pittman model18 gives the best
prediction, within a factor of 9 on average. The Katz-Thompson
model20 always underpredicts the permeability. The Swanson
model19 does not work for the west Texas samples because it is not
developed for rocks with multimodal porosity.
The Katz-Thompson model is modified by changing its con-
stant value from 1/226 to 32.336. The new constant is obtained by
plotting the measured permeability and lc2/F. The prediction is
improved (Table 9), but it is still not as good as the original
Winland-Pittman model. The Winland-Pittman model is modified
as follows:

r352
k = 102.36 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (14)
F

October 2004 SPE Reservoir Evaluation & Engineering 371

 Fig. 8—Pore-body and pore-throat sizes from NMR (assuming ␳=6.75 µm/s) and capillary pressure measurements.
However, on average, the original Winland-Pittman model gives a
slightly better prediction than the modified one.
Permeability Correlation From NMR
Trimodal Weibull Distribution Fit. The T2 response of most of
our samples exhibits multimodal behavior. To better estimate the
contribution of each mode, the T2 response was fitted with a tri-
modal Weibull distribution. Weibull distribution was selected be-
cause it is a versatile distribution that can take on the character-
istics of other types of distributions based on the value of the
shape parameter. The trimodal Weibull cumulative distribution is
given as21
兺␣ 再1 − exp冋−冉 m 冊 册冎
3 ␤i
x The vuggy porosity is estimated from the mode 3 fraction and
P共x兲 = i , . . . . . . . . . . . . . . . . . . . . (15)
i=1 i
where ␣i is the fraction of each mode, ␤i is the shape parameter, mi
is the scale parameter, and x is defined as
372
T2
x = ln , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (16)
T2,0
where T2,0 is the smallest T2 used in the T2 distribution (0.1 ms in
this case). There are nine parameters to fit, and this becomes a
constrained nonlinear least-squares problem.
The trimodal Weibull distribution can fit the T2 response very
well, as shown in Fig. 7. The fitting parameters are given in Table
10. Mode 1 is associated with the intragranular pores or the small-
est T2, mode 2 is associated with the intergranular pores or the
intermediate T2, and mode 3 is associated with the vugs at the
largest T2. The fitting provides a better picture of the contribution
of each mode and their overlap, especially for samples 4a and 4b.
agrees well with the calculated value from the PCM model and
from the vugular 750-ms cutoff (see Table 7). The vuggy porosity
estimation from the vugular 750-ms cutoff is slightly better than
the estimation from the Weibull distribution.
October 2004 SPE Reservoir Evaluation & Engineering

Fig. 9—Tracer-concentration profiles after 1 PV injection.
NMR Models. First, the existing NMR models (e.g., SDR,9
Chang,5 and Coates22) are tested against the measured air perme-
ability. The standard Schlumberger Doll Research (SDR) model is
given by
k = 4␾4T2lm2, . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (17)
where T2lm is the logarithmic mean of T2 distribution. The whole
T2 is used to calculate the permeability. The estimated permeabil-
ity using SDR correlation is given in Table 11. The correlation
works well for high-permeability samples but does not work for
low-permeability samples.
Chang’s correlation (Eq. 1) assumes that the intergranular po-
rosity controls the permeability, and its pore volume may be ap-
proximated by excluding components with T2 values above a cer-
tain cutoff value.5 Both porosity and logarithmic mean of T2 are
calculated from the T2 fraction below the cutoff value.5 The 750-
ms cutoff value was selected based on the comparison of the
correlation with the data of more than 24 samples.5 The estimated
permeabilities using this correlation for our samples are given in
Table 11. The correlation works well for low-permeability
samples. For higher-permeability samples, the correlation esti-
mates are too low; hence, the assumption that vugs do not con-
tribute to permeability is not always true. Most of the samples used
by Chang et al.5 in their study had low permeability.
Another approach to estimating permeability is using bound
volume of irreducible water (BVI) in a formation. In the absence
of centrifuge data, a value of 92 ms is used to estimate the T2 cutoff
for carbonate formations. The pore volume below the T2 cutoff is
October 2004 SPE Reservoir Evaluation & Engineering
Fig. 10—Experimental and model fit of tracer effluent concentrations.
assumed as the BVI. The Coates model to estimate the permeabil-
ity from BVI is given as22
k = 104␾4 冉 冊FFI
BVI
2
, . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (18)
where ␾ is the porosity and FFI is the free fluid volume
(FFI⳱␾–BVI). Table 11 shows that the permeability prediction
using Coates’ correlation works better for the high-permeability
samples and overpredicts the low-permeability samples by two
orders of magnitude.
It is observed from the above exercise that Chang’s model is
more suitable for low-permeability samples; on the other hand, the
SDR model is more suitable for high-permeability samples. Be-
cause the tortuosity is a good indicator to differentiate between the
low- and high-permeability samples, we propose to include the
tortuosity information in the NMR correlation as follows:
k = 4.75 冋冉 冊 册 冋冉
␾
␾750
a
␾750
4 T2lm
T2lm,750 冊 a
册 2
T2lm,750 , . . . . . . . . . . (19)
where a is defined as
␶
a=1− , 0 ⱕ a ⱕ 1, . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (20)
␶max
where ␶ is the tortuosity and ␶max is an adjustable parameter.
For the west Texas samples studied, ␶max is assumed to be 30.
The calculated permeability from this modified Chang’s model is
given in Table 12. The tortuosity is obtained from the resistivity
measurement. On average, the modified Chang’s model predicts
the permeability within a factor of 5.6. As shown in the earlier
section, the formation factor can be estimated from the NMR T2 by
using Myers’ PCM model; hence, the tortuosity also can be cal-
culated from NMR T2. Table 12 shows the predicted permeability
from the modified Chang’s model with tortuosity calculated from
NMR T2. The vuggy porosity is estimated by using the vugular
750-ms cutoff, and the intergranular porosity is calculated by sub-
tracting the vuggy porosity from the total porosity. This model (on
average) predicts the permeability within a factor of 4.8.
The Coates model (Eq. 18) was modified by changing its con-
stant from 104 to 504. The error in permeability estimation is
reduced only slightly from 36.2 to 32.3, and the predicted per-
meabilities are still not satisfactory. Other models, which in-
clude formation factor, were also tested. The models tested are
as follows:23–25
k = 81 冉 冊
T2lm,750
F
2
, . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (21)
k = 9.5 冉 冊
T2lm
F
2
, . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (22)
373
 Fig. 11—(a) Initial-oil-saturation distribution at Swr. (b) Residual-oil-saturation distributions.
T2lm,7502
k = 2.7 , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (23)
F
T2lm2
and k = 2.7 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (24)
F and 4b have a large intergranular pore size that interconnects the
The predicted values are tabulated in Table 12; the results are not
satisfactory because the error ratio is approximately two orders
of magnitude.
Conclusions
The rock fabric varies significantly between the samples studied
from a single field. Both pore-throat distribution and pore-body
distribution show multimodal behavior. The pore-throat and pore-
body sizes are better correlated in higher-permeability samples.
The tracer effluent concentration profiles and CT scans indicate
preferential flow paths and large porosity variation within the
cores. Different length scales of heterogeneity, from microscopic
to macroscopic level, make the property prediction for carbonate
formations very challenging.26
Sample 1 has a very fine grain size; hence, the intergranular
pore space is also very small. Although samples 2a and 2b have
very big crystal sizes, the intergranular pore space is very small
374
because the crystals are grown very close to each other. These
three samples have very low permeability, although they are vuggy
samples. Sample 3 is rather unusual; the intergranular pore space
is very small, but the permeability is rather high. The microfrac-
tures that connect the vugs play an important role here. Samples 4a
vugs. If the T2 signal above the 750-ms signal is neglected, then
the contribution of the intergranular porosity or microfractures to
flow is underpredicted in these three samples.
Deconvoluting the multimodal NMR T2 response into trimodal
Weibull distribution gives a better insight into the actual contri-
bution of each mode to porosity. It is found that the vuggy porosity
estimated from the Myers PCM model agrees very well with the
value calculated from NMR using the vugular 750-ms cutoff and
from the fraction of the largest mode of trimodal Weibull distri-
bution. The formation factor or the tortuosity for fully brine-
saturated core can be estimated from the NMR T2 response.
The existing permeability correlations from capillary pres-
sure curves were tested. The Winland-Pittman model gives the
best prediction.
A modification to Chang’s NMR permeability model is pro-
posed. The new model interpolates between the SDR model and
Chang’s model on the basis of tortuosity. The new model is based
on NMR T2 distribution alone and improves the permeability
predicition for the west Texas samples studied.
October 2004 SPE Reservoir Evaluation & Engineering
Nomenclature
a ⳱ semivariogram fitting parameter
c ⳱ semivariogram fitting parameter
C ⳱ flowing concentration
C* ⳱ stagnant concentration
fd ⳱ flowing fraction
fm ⳱ matrix volume fraction
fv ⳱ vug volume fraction
F ⳱ formation factor
k ⳱ permeability
keff ⳱ effective permeability
kmatrix ⳱ matrix permeability
kv ⳱ vug permeability
Kl ⳱ longitudinal dispersivity
K⬘ ⳱ mass-transfer coefficient
lc ⳱ throat size at percolation threshold
mi ⳱ scale factor in Weibull distribution
Pc ⳱ capillary pressure
P(x) ⳱ Weibull distribution
rb ⳱ pore-body radius
rt ⳱ pore-throat radius
R ⳱ resistivity
Sb ⳱ mercury saturation
Soi ⳱ initial oil saturation
Sor ⳱ residual oil saturation
Swr ⳱ residual water saturation
T2lm ⳱ logarithmic mean of T2 distribution
T2B ⳱ bulk fluid T2
T2,0 ⳱ smallest T2 in T2 distribution
u ⳱ superficial velocity
x ⳱ variable in Weibull distribution
冉 冊
T2
= ln
T2,0
␣i ⳱ fraction of mode i in Weibull distribution
␤i ⳱ shape parameter in Weibull distribution
␾ ⳱ porosity
October 2004 SPE Reservoir Evaluation & Engineering
Fig. 12—Relationship between Archie’s m factor and the vuggy
porosity/total porosity ratio.
␾CT ⳱ porosity by CT scanning
␾m ⳱ matrix porosity
␪ ⳱ contact angle
␴ ⳱ surface tension
␶ ⳱ tortuosity
Acknowledgments
This research is funded and supported by the U.S. Dept. of Energy
under Contract DE-AC26-99BC15201 and Rice U. Consortium
members: Baker Atlas, Chevron, Deepstar, Exxon Mobil, Norsk
Hydro, Halliburton-NUMAR, Kerr-McGee, Marathon, Ondeo
Nalco, Phillips, PTS, Schlumberger, and Shell. Special thanks to
Marathon, Shell, and PTS for providing core samples and core-
analysis measurements. The authors also would like to acknowl-
edge Charlotte Sullivan from the Geosciences Dept., U. of Hous-
ton, for assisting with the thin-section analysis.
References
1. Chilingar, G.V., Mannon, R.W., and Rieke, H.H. III: Oil and Gas
Production from Carbonate Rocks, Elsevier, New York City (1972).
2. Lucia, F.J.: Carbonate Reservoir Characterization, Springer-Verlag,
Berlin (1999).
3. Roehl, P.O. and Choquette, P.W.: Carbonate Petroleum Reservoirs,
Springer-Verlag, New York City (1985).
4. Jennings, J.W. Jr. and Lucia, F.J.: “Predicting Permeability From Well
Logs in Carbonates With a Link to Geology for Interwell Permeabil-
ity Mapping,” paper SPE 71336 presented at the 2001 SPE Annual
Technical Conference and Exhibition, New Orleans, 30 September–
3 October.
375

5. Chang, D. et al.: “Effective Porosity, Producible Fluid, and Permeabil-
ity in Carbonates from NMR Logging,” The Log Analyst (March–April
1997) 38, 60.
6. Crabtree, S.J. Jr., Erlich, R., and Prince, C.: “Evaluation Strategies for
Segmentation of Blue-Dyed Pores in Thin Sections of Reservoir
Rocks,” Computer Vision, Graphics, and Image Processing (January
1984) 28, 1.
7. Hicks, P.J. Jr., Deans, H.A., and Narayanan, K.R.: “Distribution of
Residual Oil in Heterogeneous Carbonate Cores Using X-Ray CT,”
SPEFE (September 1992) 235; Trans., AIME, 293.
8. Olea, R.A.: Geostatistics for Engineers and Earth Scientists, Kluwer
Academic Publisher, Boston (1999).
9. Kenyon, W.E.: “Petrophysical Principles of Applications of NMR Log-
ging,” The Log Analyst (March–April 1997) 38, 21.
10. Dauba, C. et al.: “Identification of Parallel Heterogeneities with Mis-
cible Displacement,” paper SCA-9933 presented at the 1999 SCA Sym-
posium, Golden, Colorado, 15–17 October.
11. Dauba, C. et al.: “Stochastic Description of Experimental 3D Perme-
ability Fields in Vuggy Reservoir Cores,” paper SCA-9828 presented at
the 1998 SCA Symposium, The Hague, 17–19 October.
12. Xu, B. et al.: “Use of Pore-Network Models to Simulate Laboratory
Corefloods in a Heterogeneous Carbonate Sample,” SPEJ (September
1999) 179.
13. Oshita, T. and Okabe, H.: “Water Breakthrough in Carbonate Core
Samples Visualized with X-Ray CT,” paper SCA 2000-18 presented at
the 2000 SCA Syposium, Abu Dhabi, UAE, 20–22 October.
14. Coats, K.H. and Smith, B.D.: “Dead-End Pore Volume and Dispersion
in Porous Media,” SPEJ (March 1964) 73; Trans., AIME, 231.
376
15. Brigham, W.E.: “Mixing Equations in Short Laboratory Core,” SPEJ
(February 1974) 91; Trans., AIME, 257.
16. Salter, S.J. and Mohanty, K.K.: “Multiphase Flow in Porous Media I:
Macroscopic Observations and Modeling,” paper SPE 11017 presented
at the 1982 SPE Annual Technical Conference and Exhibition, New
Orleans, 226–29 September.
17. Myers, M.T.: “Pore Combination Modeling: A Technique for Model-
ing the Permeability and Resistivity Properties of Complex Pore Sys-
tems,” paper SPE 22662 presented at the 1991 SPE Annual Technical
Conference and Exhibition, Dallas, 6–9 October.
18. Kolodzie, S. Jr.: “Analysis of Pore Throat Size and Use of the Wax-
man-Smits Equation to Determine OOIP in Spindle Field, Colorado,”
paper SPE 9382 presented at the 1980 SPE Annual Technical Confer-
ence and Exhibition, Dallas, 21–24 September.
19. Swanson, B.F.: “A Simple Correlation Between Permeabilities and
Mercury Capillary Pressures,” JPT (December 1981) 2498.
20. Katz, A.J. and Thompson, A.H.: “Quantitative Prediction of Perme-
ability in Porous Rock,” Phys. Rev. B (1986) 34, No. 11, 8179.
21. Weibull, W.: “A Statistical Distribution Function of Wide Applicabil-
ity,” J. of Applied Mechanics (1951) 18, 293.
22. Coates, G.R. et al.: “A New Characterization of Bulk-Volume Irreduc-
ible Using Magnetic Resonance,” paper QQ presented at the 1997
SPWLA Annual Logging Symposium, Houston, 6–9 June.
23. Coates, G.R. et al.: “The Magnetic Resonance Imaging Log Charac-
terized by Comparison With Petrophysical Properties and Laboratory
Core Data,” paper SPE 22723 presented at the 1991 SPE Annual Tech-
nical Conference and Exhibition, Dallas, 6–9 October.
24. Kenyon, W.E. et al.: “A Laboratory Study of Nuclear Magnetic Reso-
nance Relaxation and Its Relation to Depositional Texture and Petro-
physical Properties—Carbonate Thamama Group, Mubarraz Field,
Abu Dhabi,” paper SPE 29886 presented at the 1995 SPE Middle East
Oil Show, Bahrain, 11–14 March.
25. Dunn, K.J., LaTorraca, G.A., and Bergman, D.J.: “Permeability Rela-
tion with Other Petrophysical Parameters for Periodic Porous Media,”
Geophysics (1999) 64, No. 2, 470.
26. Ramakrishnan, T.S. et al.: “A Model-Based Interpretation Methodol-
ogy for Evaluating Carbonate Reservoirs,” paper SPE 71704 presented
October 2004 SPE Reservoir Evaluation & Engineering
 at the 2001 SPE Annual Technical Conference and Exhibition, New clude improved oil recovery, transport in porous media, and
Orleans, 30 September–3 October. formation evaluation. He holds a BTech degree from IIT, Kan-
pur, and a PhD degree from the U. of Minnesota, both in
chemical engineering. He was the Executive Editor of SPE Jour-
nal from 2002–2003. Mark Flaum is currently pursuing a doctor-
SI Metric Conversion Factors ate in chemical engineering at Rice U. e-mail: mflaum@
dyne × 1.0* E–02 ⳱ mN rice.edu. His research focuses on the use of NMR diffusion-
based measurements for characterization of porous media.
°F (°F – 32)/1.8 ⳱ °C He holds a BEng degree in chemical engineering from MacGill
in. × 2.54* E+00 ⳱ cm U. George J. Hirasaki had a 26-year career with Shell Devel-
psi × 6.894 757 E+00 ⳱ kPa opment and Shell Oil Cos. before joining the Chemical Engi-
neering faculty at Rice U. in 1993. e-mail: gjh@rice.edu. His
*Conversion factor is exact. research interests are in NMR well-logging, reservoir wettability,
enhanced oil recovery, gas hydrate recovery, asphaltene de-
position, emulsion coalescence, and surfactant/foam aquifer
Irwan Hidajat is a research engineer at GE Advanced Materi- remediation. He is a member of the Natl. Academy of Engi-
als. e-mail: Irawn.hidajat@ge.com. His research interests in- neering. Hirasaki holds a BS degree from Lamar U. and a PhD
clude advanced materials, flow through porous media, and degree from Rice U., both in chemical engineering. He was
NMR. He holds a PhD degree in chemical engineering from the named an Improved Oil Recovery Pioneer at the 1998 SPE/
U. of Houston. Kishore K. Mohanty is a professor at the U. of DOE IOR Symposium and was the 1989 recipient of the Lester
Houston. e-mail: mohanty@uh.edu. His research interests in- C. Uren Award.

October 2004 SPE Reservoir Evaluation & Engineering 377