discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/269631174
CITATIONS READS
98 345
5 authors, including:
Some of the authors of this publication are also working on these related projects:
The Tropical Ostracod Record of The Maldives and its Paleoceanographic Implications View
project
1298 Geohorizons
proposed empirical equations to calculate separate- the advantages of quantitative geometrical param-
vug porosity from acoustic transit time. Anselmetti eters over qualitative pore-type classifications.
and Eberli (1993, 1997, 1999), however, showed
how in carbonates a variety of pore types produce
variable velocities in rocks with similar porosity. DATA SET
Other experiments documented that oomoldic car-
bonate samples with near-spherical pores show large One hundred twenty carbonate core-plug samples
scatter in velocities with up to 2500 m/s (8202 ft/s) (1-in. [25.4-mm] diameter by 1- to 2 in. [25.4–
difference at a given porosity (Baechle et al., 2007, 50.8 mm] long) were selected from cored wells at
2008a; Knackstedt et al., 2008). several locations in the Middle East, Southeast Asia,
In an attempt to quantify the influence of pore and Australia (Baechle et al., 2004). The Middle
structure on permeability, Anselmetti et al. (1998) East samples are from the Shu’aiba Formation and
defined the DIA parameter g that describes the are Aptian in age, the Southeast Asian samples are
roundness of pores and compared it with mea- from an isolated platform of Miocene age, and the
sured permeability values of plugs with character- Australian samples are from two drowned cool-
istic pore types. The parameter showed a strong subtropical platforms on the Marion Plateau and
correlation to permeability. Anselmetti and Eberli are also Miocene in age (Ehrenberg et al., 2006).
(1999) also quantified the pore-structure-induced Vertical plugs were drilled from reservoir and non-
scatter of velocities at any given porosity with the reservoir intervals to capture a wide range of total
velocity deviation, which is defined as the differ- porosity, rock types, and pore types. The set of se-
ence between measured velocities and velocities lected samples includes textures ranging from
estimated using Wyllie’s time-average equation. coarse-grained packstones with interparticle to
Intervals from Miocene–Pliocene cores from the vuggy porosity to fine-grained wackestone domi-
Great Bahama Bank with high positive velocity nated by interparticle to micromoldic porosity
deviation and oomoldic porosity show low perme- (mG). All samples are either limestone or dolomite
ability. This finding corroborated the general no- with less than 2% noncarbonate minerals.
tion that rocks with a high amount of separate-vug The samples have high-quality measurements
porosity have a high velocity and low permeabil- of velocity, porosity, and permeability. Thin sec-
ity. The application of the deviation log proved tions are impregnated with blue epoxy and cut
less successful in Cretaceous carbonates where from the end of the plug sample on which these
the separation between the medium and deep in- measurements were performed. Petrophysical mea-
duction curves better detected the high flow zones surements, geological description, and DIA param-
(Smith et al., 2003), indicating that the separation eter values are listed in the Appendix.
between interparticle or intercrystalline and intra-
grain or vuggy porosity is insufficient to capture
all pore type-velocity-permeability relationships. METHODS
These complications were the motivation behind
the study presented in this article. The goal was to Petrophysical Measurements
find a repeatable, independent measure of the pore
structure that is needed to quantitatively evaluate Sonic velocity was measured using an ultrasonic
the influence of pore geometry on acoustic velocity transmitter-receiver pair with piezoelectric trans-
and other petrophysical properties. ducers forming the core of the equipment. The
The here-described methodology of DIA pro- transducer arrangement measures one compres-
duces parameters that quantify the relationship sional and two independent orthogonally polarized
between pore geometry, acoustic velocities, and per- shear waves simultaneously using a pulse trans-
meability. The high correlation between the DIA mission technique developed by Birch (1960). Both
parameters and the petrophysical values illustrates transducers (compressional and shear) generate
1300 Geohorizons
Figure 1. (a) Image acquired using plane-polarized light shows a thin-section photomicrograph of a carbonate impregnated with blue
epoxy resin. Minerals and grains are beige, whereas pore space is blue except for an air bubble with color identical to the matrix. (b) The
intensity image of absolute cross-polarized-light (XPL) variation covers the same area and is derived using XPL images at different angles.
(c, d) The close ups and the distributions illustrated in panel (e) show that the red-green-blue (RGB) color bands of the subsection are not
capable of separating air bubbles from the matrix mineral, but the XPL variation of intensity is clearly different in those regions.
This difference is then combined with color values on computed tomography (CT) scans of core plugs
for image segmentation into pore space and rock at a resolution comparable to that of our OLM
(Weger, 2006). images that suggested that directionality has little
influence on geometrical parameter values.
Pore-Shape Parameters from Digital In our DIA, 37 parameters are measured on
Image Analysis each thin section. A principal component analysis
was performed to identify the most important and
Two different types of parameters exist for pore distinguishable parameters (Weger, 2006). Four
shape calculation (Russ, 1998): global parameters DIA parameters proved to best describe several
that describe the entire pore system on a photograph aspects of the pore system. Definitions and short
or thin section and local parameters that are ob- descriptions of the parameters’ characteristics are
tained from individual pores. All shape param- given below. More specific explanations on the deri-
eters used here are derived from two-dimensional vation and characteristics of these parameters are
(2-D) images. We are aware of the limitation of given by Weger (2006).
2-D-derived geometrical properties for correlation
to the physical property of the three-dimensional Perimeter Over Area
sample volume. However, any kind of thin-section Perimeter over area (PoA) is the ratio between the
analysis, quantitative or not, suffers from this limi- total pore-space area on a thin section and the total
tation. In addition, we performed a variety of tests perimeter that encloses the pore space. The PoA
Amount of Microporosity
In our methodology, macropores are defined by DVp ¼ Vp VpW ð2Þ
pores, which are vertically connected through the
thin section, resulting in a minimum pore diameter where Vp is the measured compressional velocity
of approximately 30 mm (the thickness of a thin of the sample.
1302 Geohorizons
Figure 2. Crossplot of perimeter over area (PoA) versus dominant pore size (DomSize) where the measured acoustic velocity is super-
imposed in color. (a–d) Thin-section images are shown to illustrate carbonate pore types corresponding to certain combinations of digital
image analysis (DIA) parameters and velocity. The samples shown as images are represented by enlarged dots, and exact parameter
values are listed below each thin-section photograph.
1304 Geohorizons
Figure 4. Comparison between (a) Choquette and Pray (1970) pore types, (b) microporosity fraction, and four digital-image-analysis
parameters: (c) dominant pore size, (d) gamma (g), (e) perimeter over area, and (f) aspect ratio. All parameters are superimposed in
color onto velocity-porosity crossplots. All show a gradient that differentiates samples with high velocity from samples with low velocity at
any given porosity.
clear trend in which at any given porosity, samples ual pores is captured by g, which shows generally
with a low value of PoA (simple pore geometry) low values in samples with relatively low velocities
have relatively high velocities, whereas samples at a given porosity and vice versa (Figure 4d). This
with high values of PoA (more complex pore ge- trend is similar as for the DomSize but not as well
ometry) have low velocities (Figure 4e). In other developed (Figure 4c, d). The AR only displays a
words, samples with simple pore geometries are very weak trend in velocity-porosity space where
faster than samples with a complicated pore struc- samples with low velocity for their given porosity
ture if porosities are the same. The DomSize also are generally those with high ARs (Figure 4f ). The
shows a clear trend of increasing velocity with in- parameters PoA and AR form trends with similar
creasing values of DomSize at a given porosity. orientation. Low values of PoA and AR correspond
This trend indicates that samples with larger pores to high velocities, and high values of PoA and AR
are faster than those with smaller pores at equal correspond to low velocities for a given porosity
porosities (Figure 4c). The roundness of individ- (Figure 4e, f). The parameters DomSize and g form
1306 Geohorizons
Table 1. Coefficients of Determination from the Correlation
trends in the opposite direction (Figure 4c, d), where
between Measured Velocity and Estimated Velocity from
low values correspond to low velocities and high
Regressions with the Following Digital Image Analysis Parameters
parameter values correspond to high velocities at
as Input Variables: Dominant Pore Size, Gamma, Perimeter
any given porosity. over Area, Aspect Ratio, and Percentage of Microporosity*
The trends formed by PoA and DomSize
(Figure 4c, e) are very strong (Figure 5), indicating Estimators Used for Velocity Prediction R2
that pore structure is a second independent param- Porosity 0.542
eter influencing velocity. In Figure 5, these quanti- Porosity and AR 0.549
tative DIA parameters are displayed together with Porosity and g 0.639
velocity and porosity in three dimensions. Many Porosity and DomSize 0.768
samples align closely with a simple linear best-fit Porosity and % microporosity 0.769
surface that is displayed for reference. This illus- Porosity and PoA 0.786
trates that what appears as a 2-D scatter (Figure 3) Porosity and PoA and AR 0.788
is mostly caused by the projection of this surface Porosity and PoA and DomSize 0.800
into a 2-D crossplot. Porosity and PoA and g 0.810
Porosity and PoA and % microporosity 0.820
A crossplot of PoA and DomSize with acoustic
Porosity and PoA and % microporosity and AR 0.822
velocity superimposed in color (Figure 2) illustrates
Porosity and PoA and % microporosity and g 0.832
the link between the parameters PoA, DomSize,
Porosity and PoA and % microporosity and DomSize 0.840
and acoustic velocity and rock texture. Four thin- Porosity and PoA and % microporosity and 0.841
section images are shown to illustrate the differ- DomSize and AR
ence in pore structure detected by high, medium, Porosity and PoA and % microporosity and 0.844
and low parameter values. Low-velocity samples are DomSize and g
characterized by DomSize below 200–300 mm and Porosity and PoA and % microporosity and 0.845
PoA above 50 mm−1. The corresponding thin-section DomSize and AR and g
images are dominated by small pores, a significant *The geometric parameters PoA and DomSize in addition to porosity significantly
amount of small particles, and/or abundant micro- improve the correlation, whereas the combination of several geometrical param-
porosity (Figure 2c, d). In contrast, high-velocity sam- eters does not produce significant improvement. DomSize = dominant pore size;
g = gamma; PoA = perimeter over area; AR = aspect ratio; % microporosity =
ples are characterized by DomSize above 300 mm percentage of microporosity.
and PoA below 50 mm−1. The corresponding thin-
section images show larger pores, larger particles,
and little to no mud (Figure 2a, b). In general, high As a first step, porosity alone is used as an esti-
velocities correspond to samples with simple and mator of compressional velocity. The correlation
large pores with smooth pore surfaces, low specific between the measured and the estimated velocity
surface, and a small amount of microporosity. resulted in a coefficient of determination (R2) of
0.542. Second, a linear combination of porosity
Quantitative Assessment of Different and a single DIA parameter (g, AR, PoA, DomSize,
Geometric Characteristics % microporosity) is used to estimate compres-
sional velocity. The parameter AR combined with
To explore the link between velocity, porosity, and porosity produces the least effective velocity esti-
pore-space geometry quantitatively, velocity is es- mate (R2 = 0.549, Table 1). The highest corre-
timated using multivariate linear regression from lation coefficient of all estimates (R2 = 0.845) is
combinations of porosity and the DIA parameters. obtained by combining porosity with all DIA pa-
The geometrical parameters g, PoA, DomSize, rameters (g, AR, PoA, DomSize, and % micro-
and AR, and the percentage of microporosity were porosity), but this correlation coefficient is only
used for multivariate linear regression. The correla- slightly better than the estimate from a combina-
tion coefficients between measured and estimated tion of porosity with PoA, DomSize, and % micro-
velocity are listed in Table 1. porosity (R2 = 0.840).
Crossplots between the velocity deviation and PERMEABILITY AND PORE SHAPE
the log10 of DIA parameters PoA and DomSize
(Figure 6) result in an R2 of 0.65 and an R2 of Pore size and specific surface influence permeabil-
0.62, respectively. This means that these two quan- ity. In our data, pore size and pore network com-
titative geometric parameters are able to explain plexity (PoA), which is the 2-D equivalent of a
62–65% of the deviation of acoustic velocity (DVp) specific surface, have a strong influence on perme-
from Wyllie’s time-average equation at a given ability (Figure 7). Samples with low permeability
porosity. for their given porosity have high values of PoA
Figure 7. Permeability-porosity (K-f) crossplots with perimeter over area (PoA) and dominant pore size (DomSize) superimposed in
gray scale. Both parameters exhibit trends in porosity-permeability space. Samples with low permeability despite relatively high porosity
have high values of PoA and low values of DomSize. Samples with high permeability have low values of PoA and high values of DomSize,
representing samples with a large and simple pore structure.
1308 Geohorizons
and low values of DomSize. In turn, samples with correlation between a measured and estimated
high permeability for their given porosity show low permeability of 0.419 (red dots in Figure 8).
values of PoA and high values of DomSize. Low val- Microporosity and pore-throat diameter are im-
ues of PoA represent a simple pore structure and portant to properly predict flow properties. Thin-
low specific surface, whereas high values of PoA section-based pore-structure analyses like the pre-
stem from a more complex pore structure and high sented method here do not capture pore geometries
specific surface (Figures 2, 7). below the 30-mm threshold, but the macro- and
mesopore system represents a large part of the flow
Quantitative Permeability Estimation capacity. This is reflected in the improvement of
the permeability estimates from R2 = 0.143 to R2 =
Bear (1972) refined Kozeny’s (1927) equation to 0.415 gained by incorporating the macroscale DIA
express permeability as a function of porosity, spe- parameters into the Kozceny equation (Figure 8).
cific surface, and tortuosity. Here we estimate per- Further improvements of permeability estimates
meability using pore geometry parameters and in- will be possible using micro-CT scans (Knackstedt
corporate them into Kozeny’s equation. et al., 2008) or by combining DIA analysis and mer-
cury injection capillary pressure.
k ¼ cf3 =S2 ð4Þ
of microporosity and (2) the size and complexity Baechle et al. (2008b) proposed that the frac-
of the macropore system are much more important tion of stiff macropores versus soft micropores is
factors for determining the stiffness and, thus, the responsible for the variation of velocity at any given
acoustic behavior of carbonates (Table 1). A recent porosity and develop a rock physics model that cap-
study of oomoldic rocks by Baechle et al. (2007) tures the presence of both macro- and microporosity
also documented that the percentage of spherical to better estimate velocity and permeability. The
pore shape is not the dominant factor in producing percentage of microporosity for this dual porosity
positive deviations from the Wyllie time-average DEM model is derived with the DIA methodology
equation. They attribute the variable acoustic re- described here (Baechle et al., 2008b).
sponse of up to 2000 m/s (6562 ft/s) at a given po- The assumption that rocks with mostly moldic
rosity to variations in intercrystalline porosity in and/or vuggy porosity will have a faster acoustic ve-
the rock frame; a conclusion that is corroborated locity than a formation with predominantly inter-
by ultra-high-resolution CT tomography and scan- crystalline and/or interparticle porosity has been
ning electron microscope (SEM) analysis on the used for quantitative estimates of separate-vug po-
same samples (Knackstedt et al., 2008). rosity from acoustic logs (e.g., Nurmi, 1984; Lucia
1310 Geohorizons
Figure 9. Velocity-porosity
crossplot of samples measured
at 20 MPa with annotation of
porosity types separated into two
groups. Open circles are sam-
ples with vuggy, moldic, intra-
frame, and intragrain porosity,
black and gray dots represent
samples with interparticle and
intercrystalline porosity. A large
overlap exists between these
two groups, indicating that rocks
with interparticle and/or inter-
crystalline porosity can in some
cases have a stiff framework
and high velocity.
and Conti, 1987; Wang and Lucia, 1993; Anselmetti 1996). In carbonates, cementation at grain contacts
and Eberli,1999). To test this assumption, we distrib- are meniscus cements derived from meteoric waters
ute the samples into two groups. The vuggy group (Harris, 1978; Longman, 1980) or micritic bridging
consists of samples whose primary pore types are cements in the marine realm (Hillgärtner et al.,
vuggy, moldic, intraframe, and intraparticle poros- 2001). In a Holocene grainstone, small amounts
ity. The interparticle group consists of samples with of bridging cement (15% of the total rock) produce
interparticle and intercrystalline porosity as the a Vp of 4500 m/s (14,764 ft/s) at 20 MPa (Eberli
primary pore type. Plotting the two groups in the et al., 2003). Some of the samples displayed in
velocity-porosity space reveals a considerable over- Figure 9 are dolomites; in this case, the extreme
lap (Figure 9). The samples of the vuggy group gen- stiffening of the frame is not caused by early cement
erally plot above the Wyllie time-average equa- but more likely by interlocking crystals. Anselmetti
tion, and a cluster of interparticle samples in the et al. (1997) documented this process on Neogene
low velocity area is observed. However, nearly an carbonates, in which the velocity of sucrosic dolo-
equal amount of samples from each group display mite increases dramatically as isolated rhombohe-
an exceptionally high velocity at a given porosity dra grow together to form a stiff framework.
(Figure 9).
High velocity at a given, sometimes high poros-
ity is possible if pore compressibility is low and CONCLUSIONS AND IMPLICATIONS
consequently if the stiffness of the rock is not sig-
nificantly decreased (Mavko and Mukerji, 1995). The DIA quantifies the influence of pore types on
Such a stiff frame is well known to occur in rocks velocity and permeability. A combination of porosity
with vugs or molds, but it also occurs in rocks with and (image-derived) microporosity is capable of
interparticle and intercrystalline porosity. A pro- estimating velocity with R2 = 0.77; a combina-
cess that can produce frame stiffening in these latter tion of porosity and digital image parameters is
rocks is contact cementation (Dvorkin and Nur, able to explain more than 85% of the variation
Appendix. Texture, Pore Type, DIM (Digital Image Analysis) Parameter Values, and Petrophysical
Measurements*
Dunham Dominant Minor DomSize PoA Micro
Sample Index** Pore Type† Pore Type† Gamma (mm) (mm−1) AR VP (m/s) Phi (%) Phi (%) K (md)
1312 Geohorizons
Appendix. Cont.
1314 Geohorizons
Appendix. Cont.
C5-M73 G-P MO IP 2.05 118 110 0.57 4092 23.4 20.1 8.6
C5-M76 W-P IP MO 2.81 96 116 0.57 3696 27.9 24.3 11.0
C5-M77 P IP MO 2.05 174 67 0.56 4450 21.7 20.2 3.9
C5-M78 P MO VUG, IP 1.91 148 67 0.57 4118 30.0 24.0 63.0
C5-M83 P MO 2.20 46 163 0.59 4117 24.1 21.6 5.5
*DomSize = dominant pore size; PoA = perimeter over area; AR = aspect ratio; VP = compressional acoustic velocity (values of water-saturated conditions with a confining
pressure of 20 MPa at a frequency of 1 kHz); Phi = porosity; K = permeability.
**G = grainstone; P = packstone; W = wackestone; M = mudstone; FL = floatstone; FR = framestone; RD = rudstone; B = boundstone; combinations are separated by a
hyphen; rDol = completely recrystallized rocks.
†
IP = interparticle; IX = intercrystalline; MO = moldic; VUG = vuggy; WPO = intraparticle; WF = intraframe; mG = micromoldic; FR = fracture. The dominant pore type listed
in the table is estimated to contain more than 50% of the visible pores. Minor pore types are listed if they are more than an estimated 5% of the total visible pores.
1316 Geohorizons
Shang, Q. Li, and T. F. Gao, eds., Theoretical and compu- Wang, Z., 1997, Seismic properties of carbonate rocks: Geo-
tational acoustics: Singapore, World Scientific, p. 335– physical Development Series, v. 6, p. 29–52.
347. Weger, R. J., 2006, Quantitative pore/rock type parameters
Van den Berg, E. H., A. G. C. A. Meesters, J. A. M. Kenter, in carbonates and their relationship to velocity devia-
and W. Schlager, 2002, Automated separation of touch- tions: Ph.D. dissertation thesis, University of Miami,
ing grains in digital images of thin sections: Computers Coral Gables, 232 p.
and Geosciences, v. 28, p. 179–190, doi:10.1016 Weger, R. J., G. T. Baechle, J. L. Masaferro, and G. P. Eberli,
/S0098-3004(01)00038-3. 2004, Effects of pore structure on sonic velocity in car-
Wang, R. F. P., and F. J. Lucia, 1993, Comparison of empiri- bonates: Society of Exploration Geophysicists, Ex-
cal models for calculating the vuggy porosity and cementa- panded Abstracts, v. 23, p. 1774.
tion exponent of carbonates from log responses: Bureau of Wyllie, M. R. J., A. R. Gregory, and L. W. Gardner, 1956, Elas-
Economic Geology, University of Texas at Austin, Geo- tic wave velocities in heterogeneous and porous media:
logical Circular 93-4, 27 p. Geophysics, v. 21, p. 41–70, doi:10.1190/1.1438217.