ortega et al 2006 a scale-independent approach to fracture
DESCRIPTION
Artigo que fala sobre a técnica Scan-lineTRANSCRIPT
AUTHORS
Orlando J. Ortega � Shell InternationalExploration and Production, 200 North DairyAshford, Houston, Texas 77079;[email protected]
Orlando J. Ortega received his M.S. degree andhis Ph.D. from the University of Texas at Austin.He is a structural geologist specializing in fracturesfor Shell Exploration and Production in Hous-ton, Texas.
Randall A. Marrett � Department of Geo-logical Sciences, John A. and Katherine G.Jackson School of Geosciences, University ofTexas at Austin, Austin, Texas 78712-1101
Randall A. Marrett is an associate professor inthe Department of Geological Sciences, JacksonSchool of Geosciences, University of Texas atAustin. Quantitative analysis of fracture systemsis a major area of his research. He receivedhis Ph.D. from Cornell University.
Stephen E. Laubach � Bureau of EconomicGeology, John A. and Katherine G. JacksonSchool of Geosciences, University of Texas atAustin, Austin, Texas 78713-8924;[email protected]
Stephen E. Laubach is a senior research scientistat the Bureau of Economic Geology, where heconducts research in structural diagenesis. HisPh.D. is from the University of Illinois.
ACKNOWLEDGEMENTS
This study was partially supported by the Na-tional Science Foundation Grant EAR-9614582,the Texas Advanced Research Program Grant003658-011, and the industrial associates of theFracture Research and Application Consortium:Anadarko, Bill Barrett Corp., BG Group, ChevronTexaco, Conoco Inc., Devon Energy Corpora-tion, Ecopetrol, EnCana, EOG Resources, Huber,Instituto Mexicano del Petroleo, Japan NationalOil Corp., Lariat Petroleum Inc., Marathon Oil,Petroleos Mexicanos Exploracion y Produccion,Petroleos de Venezuela, Petrobras, Repsol-YPF-Maxus, Saudi Aramco, Shell, Schlumberger,Tom Brown, TotalFinaElf, Williams Exploration& Production. We thank Faustino Monroy San-tiago, Julia Gale, Jon Olson, and Bob Gold-hammer for valuable discussion and Alvar Bra-athen, Ross Clark, Wayne Narr, and AmgadYounis for thoughtful reviews.
A scale-independent approachto fracture intensity and averagespacing measurementOrlando J. Ortega, Randall A. Marrett, andStephen E. Laubach
ABSTRACT
Fracture intensity, the number of fractures per unit length along a
sample line, is an important attribute of fracture systems that can
be problematic to establish in the subsurface. Lack of adequate con-
straints on fracture intensity may limit the economic exploitation
of fractured reservoirs because intensity describes the abundance of
fractures potentially available for fluid flow and the probability
of encountering fractures in a borehole. Traditional methods of
fracture-intensity measurement are inadequate because they ignore
the wide spectrum of fracture sizes found in many fracture systems
and the consequent scale dependence of fracture intensity. An al-
ternative approach makes use of fracture-size distributions, which
allow more meaningful comparisons between different locations
and allow microfractures in subsurface samples to be used for
fracture-intensity measurement. Comparisons are more meaning-
ful because sampling artifacts can be recognized and avoided, and
because common thresholds of fracture size can be enforced for
counting in different locations. Additionally, quantification of the
fracture-size distribution provides a mechanism for evaluation of
uncertainties. Estimates of fracture intensity using this approach for
two carbonate beds in the Sierra Madre Oriental, Mexico, illustrate
how size-cognizant measurements cast new light on widely ac-
cepted interpretation of geologic controls of fracture intensity.
INTRODUCTION
Sampling problems represent a fundamental challenge to subsur-
face fracture characterization because complete sampling of large
conductive fractures that dominate fluid flow in the subsurface is
GEOHORIZONS
AAPG Bulletin, v. 90, no. 2 (February 2006), pp. 193–208 193
Copyright #2006. The American Association of Petroleum Geologists. All rights reserved.
Manuscript received April 1, 2005; provisional acceptance August 5, 2005; revised manuscript receivedAugust 16, 2005; final acceptance August 25, 2005.
DOI:10.1306/08250505059
unfeasible. Large fractures in the subsurface typically
are widely spaced and have much larger dimensions
than the diameter of a borehole. Consequently, the
probability of encountering such fractures in a given
layer are small, and even where such fractures are in-
tersected, only fragmentary data can be collected (Narr,
1991). Large fractures commonly are more open than
smaller equivalents, so core recovery of these fractures
is typically incomplete (Laubach, 2003; Laubach et al.,
2004). Horizontal boreholes enhance the likelihood
of encountering large fractures but at most intersect
only a fraction of the population because of the finite
strike lengths of fractures.
One approach for overcoming limitations to sam-
pling fractures in the subsurface is to treat abundant
microfractures as proxies for related macrofractures in
the same rock volume. For example, orientations of
microfractures may provide estimates of macrofrac-
ture orientations (Laubach, 1997; Ortega and Marrett,
2000). Likewise, microfracture abundance is directly
related to macrofracture abundance in many cases
(Marrett et al., 1999; Ortega and Marrett, 2000). This
approach relies on systematic measurement of micro-
fracture attributes in samples retrieved from the sub-
surface. A more traditional approach for predicting
the abundance of subsurface fractures is to understand
how geologic parameters like mechanical-layer thick-
ness control variations in fracture intensity. For exam-
ple, measurable characteristics of sedimentary layers
might be used to predict the intensity of natural frac-
tures in the subsurface for identifying and ranking ex-
ploration and exploitation targets.
For the purposes of this article, fracture intensity
is the number of fractures encountered per unit length
along a sample line oriented perpendicular to the frac-
tures in a set. Nelson (1985) summarized some widely
accepted ideas of geologic controls on fracture inten-
sity: (1) composition, (2) texture (including grain size
and porosity), (3) structural position, and (4) stratig-
raphy (bed thickness). Compositional controls on frac-
ture intensity have been studied in laboratory experi-
ments (e.g., Handin et al., 1963) and with systematic
measurements of fracture spacings in outcrop (Das
Gupta, 1978; Sinclair, 1980). Core studies show that
progressive diagenesis causes mechanical properties to
evolve with marked effects on fracture-intensity pat-
terns (Marin et al., 1993). Price (1966) found an inverse
relation between rock strength and porosity, and
Nelson (1985) proposed that lower porosity rocks of
similar composition should have more abundant frac-
tures than higher porosity rocks. Perhaps the most
widespread paradigm of geologic controls on fracture
intensity is that of bed thickness (or mechanical-layer
thickness). Bogdonov (1947)was apparently the first to
document the intuitive result that fracture intensity
increases (i.e., fracture spacing decreases) with decreas-
ing bed thickness, but numerous other authors have
subsequently arrived at similar conclusions (Price, 1966;
Ladeira and Price, 1981; Huang and Angelier, 1989;
Narr and Suppe, 1991; Gross, 1993; Mandal et al.,
1994; Gross and Engelder, 1995;Wu and Pollard, 1995;
Becker and Gross, 1996; Narr, 1996; Pascal et al., 1997;
Ji and Saruwatari, 1998; Bai and Pollard, 2000).
However, a common flaw shared by all previous
empirical investigations quantifying controls on fracture
intensity is a lack of an explicit accounting for fracture
size. Althoughmany joints in outcrop apparently have a
narrow size range, opening-mode fractures formed in
the deep subsurface vary significantly in size, in some
cases ranging across at least five orders of magnitude,
with larger fractures typically being progressively less
abundant (Gillespie et al., 1993; Marrett et al., 1999).
Cumulative frequency is commonly related to fracture
size in the formof a power law (e.g.,Marrett et al., 1999).
Moreover, geomechanical models of fracture growth
indicate that many combinations of rock properties and
loading paths should result in fracture populations
having a wide spectrum of sizes (Olson, 1997, 2003).
As we demonstrate below, cumulative frequency
is a direct measure of fracture intensity, so there is an
implicit covariance of fracture intensity with fracture
size. Consequently, it is meaningless to quantify frac-
ture intensity without specifying what range of frac-
ture sizes are represented, and it is risky to compare
fracture-intensity measurements from different loca-
tions if the results have not been normalized to a com-
mon range of fracture sizes.
In this contribution, we present a new method
that explicitly accounts for variation of fracture inten-
sity according to fracture size. Applying this method
to several different locations permits meaningful com-
parison of resulting fracture-intensity measurements
and quantification of variation with independent geo-
logic parameters. Moreover, the approach is amenable
to systematic observations at any scale, so it can be
used for microfracture-based subsurface studies, thus
circumventing fracture-sampling challenges.
We use examples from the Sierra Madre Oriental,
northeastern Mexico, to illustrate the new approach.
The SierraMadreOriental is a northeast-directed, thin-
skinned fold-thrust belt of early Tertiary (Laramide)
age (Figure 1), which exposes a thick sequence of
194 Geohorizons
Mesozoic carbonate strata that, near Monterrey, have
been deformed into isoclinal anticlines (e.g., Marrett
and Aranda-Garcıa, 1999). Excellent outcrops in this
area provide an ideal opportunity to study sedimen-
tary and stratigraphic controls on fracture intensity.
This platform-to-basin carbonate system is the subject
of numerous sedimentologic and stratigraphic studies
(e.g., Goldhammer et al., 1991, and references there-
in). The fractures studied in the Lower Cretaceous
carbonate strata have high aperture-to-length ratios
and are mostly calcite and/or dolomite filled, so they
represent veins. The mineral-fill patterns and shapes of
these fractures closely match those found in core, and
except for late-stage calcite fill, these outcrop fractures
are good analogs to fractures in producing fields (Lau-
bach, 2003). Relative timing constraints suggest the
veins predominantly formed early in the burial history
of the rocks (Marrett and Laubach, 2001) and signifi-
cantly predate local Laramide-age folding.
FRACTURE DATA COLLECTION
The essential fracture documentation technique of this
study was to measure fracture attributes along linear
traverses (scan lines). Fracture orientation, morphol-
ogy, crosscutting relationships, composition and tex-
ture of fracture fill, and mechanical-layer thickness
were recorded for each fracture set of the beds studied.
Opening displacement (or kinematic aperture; here-
after, aperture) of each fracture, intercepted by a frac-
ture set–perpendicular scan line, was also recorded.
Beds were chosen so that they would represent con-
trasting sedimentary facies, various bed thicknesses,
and different degrees of dolomitization (Figure 2). Me-
chanical bed thickness ranges from less than 10 cm
(4 in.) to nearly 2 m (6.6 ft). Dolomite content ranges
from 0 to 100%, but most beds are either highly do-
lomitized or only weakly dolomitized. Dolomite con-
tent was measured in digital photomicrographs of
Figure 1. Location of the study area on a map of the geologic provinces of Mexico. Basins adjacent to the study area producehydrocarbons from units that are equivalent to those studied. Tectonic elements are conceptual. Modified from Benavides (1956).
Ortega et al. 195
thin sections stained with alizarine red, but we did not
distinguish dolomite resulting from different diagenet-
ic events. The beds studied were classified into six
different sedimentary facies on the basis of dominant
texture and allochem content, along with the inter-
pretation of the depositional environment. However,
fracture-intensity analyses were done independently
for two broadly defined sedimentary textures: mud-
supported and grain-supported facies. Here, we pres-
ent and illustrate the methodology; the complete
fracture analysis of the Sierra Madre Oriental carbon-
ate rocks will be presented elsewhere.
Fracture Set Determination
Patterns of fracture attributes, like fracture morphol-
ogy, timing, and orientation, can be used to establish
fracture sets. Although fracture orientation is com-
monly used alone for distinguishing fracture sets, in
many cases, fracture timing is mostly diagnostic. In-
terpretations of fracture timing are most based on
crosscutting relationships. Crosscutting principles sug-
gest that a fracture set that systematically crosscuts
another fracture set must postdate the other set. Al-
though crosscutting relationships do not constrain the
absolute time of fracture formation, they can be used to
separate groups of fractures that formed contempora-
neously and probably under similar remote stress con-
ditions. Fracture timing can also be constrained in rela-
tion to products of other geologic processes, such as
diagenesis and tectonism. Fractures that systematically
abut against other fractures can be interpreted as youn-
ger (e.g., Gross, 1993). Statistically based abutting re-
lationships between fracture sets have been used to
support fracture timing relationships based on orien-
tation and crosscutting relations.
Fracture Size Data
Fracture size was quantified by measuring apertures
along scan lines drawn perpendicular to each fracture
set in bed-perpendicular exposures (Figure 3). More
than 14,000 fracture apertures from 42 beds compose
the fracture-size database for the Sierra Madre Ori-
ental. Outcrop limitations generally prevented the col-
lection of sufficient fracture-aperture measurements
for a reliable determination of fracture intensity of
weakly developed fracture sets (Bonnet et al., 2001).
However, an effort was made to obtain fracture-
intensity data from low-fracture-intensity beds to doc-
ument approximate fracture intensity and to facilitate
comparison with other beds.
Accurate aperturemeasurement of these fractures is
possible because they are filled by secondary mineraliza-
tion. To collect fracture-aperture measurements, we
used a logarithmically graduated comparator (Figure 4).
Using this toolwith a hand lens allows documentation of
fracture apertures as small as about 0.05mm (0.002 in.)
on sufficiently exposed outcrops in the field. The com-
parator contains lines with increasing width starting
at 0.05 mm (0.002 in.) and ending at 5 mm (0.2 in.).
Increments in line width represent approximately uni-
formmultiples of each other and are thus evenly spaced
Figure 2. Variation of facies, bed thickness, and dolomitecontent among layers studied. The facies definition includestextural groups and depositional environment interpretations.Some histograms exclude multifacies beds and/or chert beds.Total dolomite content is based on thin-section point counting.
196 Geohorizons
whenplottedon a logarithmic axis. In thisway, apertures
were measured with consistent accuracy as viewed in
a log-log graph.
NORMALIZED FRACTURE INTENSITY
A complete inventory of apertures along a scan line
provides a more complete estimate of fracture inten-
sity than methods that do not account for size (e.g.,
Bogdonov, 1947; Ladeira and Price, 1981; Narr and
Suppe, 1991; Wu and Pollard, 1995; Narr, 1996). In
our data set, we find that fracture-size distributions
follow power-law scaling. However, the new method
is equally applicable to any type of cumulative size
distribution.
Fracture intensity, fracture density, average frac-
ture spacing, and fracture frequency all describe spatial
abundance of fractures. Fracture frequency is a general
term that subsumes fracture intensity and fracture den-
sity. Fracture intensity has units of inverse length and
is a ratio of the number of fractures, sum of fracture
lengths, or sum of fracture surface areas to the length,
area, or volume of observation, respectively (Mauldon
et al., 2001). The literature contains numerous exam-
ples of fracture-intensity determinations, mostly from
outcrops (e.g., Corbett et al., 1987; Huang and An-
gelier, 1989; Narr and Suppe, 1991; Gross, 1993) and
also from the subsurface (Corbett et al., 1987; Lorenz
and Hill, 1991; Narr, 1996) and laboratory tests (Rives
et al., 1992, 1994;Wu and Pollard, 1992, 1995;Mandal
Figure 4. Fracture-aperture comparator. Versions used in thefield have been calibrated by microscope. The version reproducedin this article is affected by printer limitations and paper qualityand should not be used without recalibration. Reproductionusing photocopying methods will further reduce the accuracyof the line widths.
Figure 3. Segment of the scan line foraperture-data collection in bed 3, La Es-calera Canyon. The mechanical bound-aries are indicated by dashed lines andcorrespond to the systematic terminationsof fractures. Two scan lines were con-structed for this bed, one for each ofthe two fracture sets present. The scanline shown (white traces) is approximatelyperpendicular to fracture set B (inclinedwith respect to bed boundaries).
Ortega et al. 197
et al., 1994). Fracture intensity (F) is most commonly
determined fromone-dimensional observation domains
(i.e., scan lines) by dividing the number of fractures (N)
by the total length (L) of the scan line:
F ¼ N
Lð1Þ
However, the most common indication of fracture
abundance reported in the literature is the average
spacing (S ) between fractures along a scan line (Huang
and Angelier, 1989; Narr, 1991; Gross, 1993; Ji and
Saruwatari, 1998). Average spacing is the inverse of the
fracture intensity:
S ¼ 1
N
XNi¼1
Si ¼L
N¼ 1
Fð2Þ
where Si is the spacing between the nearest neighbor
fractures along the scan line.
Some fracture spacing data sets follow a normal
distribution (Ji and Saruwatari, 1998), but most are
better modeled by negative exponential (Priest and
Hudson, 1976; Pineau, 1985), log-normal (Sen and
Kazi, 1984; Narr and Suppe, 1991; Rives et al., 1992;
Becker and Gross, 1996; Pascal et al., 1997), or gamma
distributions (e.g., Huang and Angelier, 1989; Gross,
1993; Castaing et al., 1996). Sampling bias might ex-
plain variations in the best mathematical model for
fracture spacing distributions, but Rives et al. (1992)
and Ji and Saruwatari (1998) have proposed alterna-
tive explanations.
Although numerous measurements of fracture in-
tensity or average spacing have been made previously,
these estimates typically do not explicitly account for
fracture size. Thus, data used to infer relations be-
tween bed thickness and fracture intensity or spacing
are commonly unclear about the size of fractures used
for spacing determinations. In some cases where nu-
merous smaller fractures are obvious, a rule is speci-
fied about the fracture size (bed-normal dimension) to
be measured relative to bed thickness; that is, only
fractures that span a given bed thickness are to be
measured (for example, Laubach and Tremain, 1991;
Laubach et al., 1998 for fractures in coal). However,
in most instances, fractures counted could be all that
were visible to the naked eye, perhaps only those that
reached the boundaries of the mechanical layer, or
merely those that are prominent (Huang and Angelier,
1989; Lorenz and Hill, 1991; Castaing et al., 1996).
Moreover, the size of fractures visible on an outcrop
varies according to exposure quality and fracture-fill
characteristics, which typically have not been accounted
for. Furthermore, if we changed the scale of observa-
tion, for example, from outcrop scale to microscopic
scale, estimates of fracture intensity or spacing would
typically be different. This lack of rigor potentially
introduces bias in the calculation of fracture intensity
or spacing and, consequently, may undercut conclu-
sions regarding relationships between fracture intensity
and geologic parameters.
In the subsurface, the bed thickness cannot be
generally measured, except under rare circumstances
(e.g., Marin et al., 1993). However, in most subsurface
cases, there will be insufficient fractures having the
size of interest (i.e., large conductive fractures) to de-
rive a statistically significant estimate of spacing.
Calculation of Normalized Fracture Intensity
Use of cumulative-frequency fracture-size distributions
addresses the potential scale bias inherent to average
fracture intensity or spacing determinations because
they yield a measure of fracture abundance that ex-
plicitly varies according to their size. This method
effectively quantifies fracture intensity for detection
thresholds corresponding to all fracture sizes considered.
It has the advantage of facilitating a comparison between
data sets collected at different scales of observation or
levels of resolution. The following example illustrates
the advantages of using cumulative-frequency fracture-
size distributions over the traditional method to mea-
sure and compare fracture abundance.
Figure 5 shows a schematic cross section of a frac-
tured layer in which fracture abundance is to be de-
termined along a 1-m (3.3-ft)-long scan line at three
different scales of observation. This example represents
real aperture data collected along a scan line in a do-
lomitized mudstone bed (bed 3) that is approximately
8 cm (3.1 in.) thick, studied in La Escalera Canyon,
northeasternMexico. The fractures have been random-
ly located in the cross section because, for the following
analysis, the actual location of the fractures is not im-
portant. First, only fractures with apertures greater
than 0.5 mm (0.02 in.) are counted (Figure 5A). Frac-
ture size spans an order of magnitude from 0.5 to 5mm
(0.02 to 0.2 in.) in aperture, measurable with a 0.5-mm
(0.02-in.) graduated ruler. At this scale of observation
(24 fractures in 1000 mm [39 in.]), we find that the
fracture intensity is approximately 24 fractures/m
198 Geohorizons
(7.3 fractures/ft). However, if we included all fractures
with apertures down to 0.05 mm (0.002 in.) along one
part of the scan line (28 fractures in 200 mm [8 in.])
measurable with a hand lens and the aperture compar-
ator (Figure 4), then the fracture intensity is 140 frac-
tures/m (42.7 fractures/ft) (Figure 5B) and nearly an
order of magnitude higher than the first estimate. Fur-
thermore, microfracture-aperture data from a thin sec-
tion of this bed (Figure 5C) indicate that the intensity
of fractures wider than 0.005 mm (0.0002 in.) (10 frac-
tures in 12 mm [0.47 in.]) is nearly another order of
magnitude higher, about 830 fractures/m (253 frac-
tures/ft). This example illustrates that fracture inten-
sity varies with changes in resolution (which typically
accompany changes in observation scale) because the
number of fractures observed in a domain will increase
as the threshold size for fracture detection decreases.
Consequently, fracture intensity (and average spacing)
is onlymeaningful if the fracture-detection threshold is
quantified.
Cumulative-frequency fracture-size distributions
provide a measure of fracture intensity (or average
spacing) that explicitly accounts for fracture size and
permits a comparison of data collected at different lo-
cations and/or observation scales. For purposes of com-
paring any two data sets, normalization of fracture
intensities is effectively achieved by choosing a com-
mon fracture-detection threshold. To estimate the
Figure 5. Schematic cross section of bed 3, La Escalera Canyon. Fractures of set A are solid white lines, and fracture aperture is thenumber next to each fracture. The scan line is represented by a dashed line that runs perpendicular to the fractures. (A) Fracturesshown all have apertures greater than 0.5 mm (0.02 in.), the minimum size of fractures that span the mechanical layer. The averagespacing between fractures of this size is 40 mm (1.57 in.). (B) The first 20% of the scan line (shaded area in A) showing the fractureswith apertures of 0.05 mm (0.002 in.) or greater. The average fracture spacing at this scale of observation is 7 mm (0.27 in.). (C) Scanline at the microscopic scale (black area in B) showing fractures having apertures of 0.005 mm (0.0002 in.) or larger in a thin section.The average fracture spacing at the thin-section scale is about 1.2 mm (0.04 in.). All fractures in the thin section are invisible atoutcrop scale, with the exception of one fracture 0.115 mm (0.0045 in.) wide.
Ortega et al. 199
cumulative-frequency fracture-size distribution for a
scan-line data set, we followed these steps:
1. List all measurements of fracture size (aperture is
used to quantify fracture size in the examples pre-
sented here), and sort them from the largest to the
smallest.
2. Number the sorted fracture sizes, such that the
largest is number 1 and successively smaller frac-
tures receive higher numbers as fracture size de-
creases (i.e., generate a list of cumulative numbers).
3. Simplify the list by eliminating all fractures having
the same size (given the resolution of measure-
ments), except for the one with the largest cumu-
lative number.
4. Normalize the cumulative numbers by the length of
the scan line. This will generate estimates of cumu-
lative fracture intensity (i.e., number of fractures of
a certain size or larger per unit length of scan line).
This parameter is a measure of the fracture inten-
sity as a function of detection threshold and can be
used to compare and contrast fracture intensities
from different beds and/or observation scales.
5. Plot cumulative fracture intensity versus fracture
size to provide a graphical display of the distribu-
tion. Several fracture-size distributions can be ade-
quately modeled using simple mathematical laws.
Power-lawdistributions plot as lines in a log-log plot.
Exponential fracture-size distributions plot as a line
in a log-linear graph.
6. Obtain the parameters of the bestmodel for the distri-
bution observed (e.g., coefficient and exponent of the
power law, as determined by least-squares regression).
The size distribution for the example discussed
above (Figure 5) is shown in Figure 6. A power law
adequately models these data across four orders of
magnitude variation in aperture. The power law relates
fracture size to the intensity of fractures having a certain
size or larger or, alternatively, to the average spacing of
fractures of a certain size or larger. The coefficient rep-
resents the value of the power law for a fracture size of
one (and, therefore, depends on the units chosen to
quantify fracture size). The exponent of the power law
represents the slope of the power-law line in a log-log
plot. This exponent is always negative, because by defi-
nition, the cumulative number of fractures must in-
crease or remain constant as fracture size decreases.
Obtaining the normalized fracture intensity or aver-
age spacing for a given fracture-detection threshold is
straightforward. From the cumulative-frequency frac-
ture-size distribution plot (Figure 6), we read the cor-
responding frequency (number of fractures per unit
length of scan line) for a given fracture size using the
power-law distribution model as a guide. We can also
obtain an average spacing estimate using the power law,
referring to an ordinate axis graduated with inverse fre-
quency values (i.e., average spacing). For the exercise
discussed previously, we can now obtain the fracture in-
tensity and average spacing for any scale of observation.
The arrows in Figure 6 indicate the scale depen-
dence of fracture abundance. Figure 7 illustrates the
range of possible fracture-intensity results based on
counting fractures without accounting for their sizes.
This approach generates a single estimate of average
spacing and intensity that is only valid for some specific
threshold of fracture aperture that remains unknown.
Figure 6. Normalized fracture intensityfrom cumulative fracture-size distributionfor fracture set A in bed 3, La EscaleraCanyon. Fracture abundance can be char-acterized as intensity (cumulative fre-quency) or average spacing (inverse ofcumulative frequency). Open and filledsquares are field data. The open squaresare data subject to truncation bias notused in regression. The dashed line isa power-law regression. In the formula,m�1 and mm�1 refer to the left-handand right-hand scales.
200 Geohorizons
Consequently, average spacing or intensity estimates
cannot be compared confidently with measurements
fromother scan lines because fracture-detection thresh-
olds might differ because of variable outcrop quality,
different fracture morphologies, or different sizes of
fractures that span a layer. Additionally, there is no
information about sampling bias possibly affecting the
data. If identical fracture sizes are chosen for fracture
spacing measurement at different locations, then con-
ventional approaches and this population-based ap-
proach will give the same results. However, this con-
dition is unlikely to occur without a conscious effort to
maintain a consistent fracture size threshold from loca-
tion to location. Figure 7 indicates ranges of fracture
intensity that might be inferred for different observa-
tion scales or resolution levels.
Sampling and Topologic Artifacts
Fault-scaling studies dominate the literature on frac-
ture scaling (Ackermann and Schlische, 1997; Marrett
et al., 1999; Bonnet et al., 2001, and references therein).
Much less abundant is the literature on opening-mode
fracture scaling. However, there is now agreement
about the power-law nature of many opening-mode
fracture-size distributions (Gudmundsson, 1987;Wong
et al., 1989; Heffer and Bevan, 1990; Barton and Zo-
back, 1992; Gillespie et al., 1993; Hatton et al., 1994;
Sanderson et al., 1994; Belfield and Sovich, 1995; Clark
et al., 1995; Gross and Engelder, 1995; Johnston and
McCaffrey, 1996; Marrett, 1997; Marrett et al., 1999;
Ortega and Marrett, 2000; Bonnet et al., 2001). Most
work on this aspect of fracture population statistics has
been done for fracture length. However, fracture-length
determinations are fraught with uncertainty, most of
which arise from the difficulties of clearly defining what
is a throughgoing fracture where fractures branch or
intersect (Ortega and Marrett, 2000). Studies of
fracture-aperture distributions are less common in the
literature, perhaps because of the lack of appropriate
tools for effective fracture-aperturemeasurement in the
field. However, considerable progress has been made in
recent years using new methods for accurate fracture-
aperture measurements, which have generated some
Figure 7. Representation of the fracture-counting method for fracture-intensity calculation in a log-log graph of fracture-sizedistribution. Thick lines represent the example explained in the text, and lower detection ranges are approximate for the CupidoFormation. The counting method is fraught with errors introduced by inconsistent limits of minimum fracture size counted for thecalculation. Patterned areas are determined by the approximate minimum size detectable at various observation scales. Note therange of possible fracture intensities and overlapping areas.
Ortega et al. 201
well-constrained fracture aperture-size distributions
(Ortega et al., 1998; Marrett et al., 1999).
Marrett et al. (1999) showed examples of opening-
mode fracture and fault-size distributions that follow
consistent power laws across three to five orders of
magnitude in size. Some of these distributions included
microscopic-scale aperture data from sandstones and
carbonate rocks. However, fracture-size distributions
commonly show deviations from an ideal power law in
the small and large fracture-size parts of the distribu-
tion. Deviations from power-law behavior can be ex-
plained by consideration of sampling and topologic and
statistical artifacts, which can be assessed from patterns
of fracture-size distributions. To use scaling methods,
these artifacts need to be understood and minimized.
Truncation Artifacts
Deviations of small fractures from the scaling shown
by larger fractures of a population have been long
recognized as truncation artifacts (Baecher and Lan-
ney, 1978; Barton and Zoback, 1992; Pickering et al.,
1995). Truncation artifacts arise from variable sam-
pling completeness for fracture sizes near the limits of
resolution of the observation tool (e.g., smallest frac-
ture size accurately measurable with the naked eye
varies from place to place in outcrop exposures). Al-
though a mathematical description of truncation arti-
facts has not yet been proposed, the part of a size dis-
tribution affected by truncation artifacts (commonly the
smallest fractures) typically shows a convex-upward
curve that decreases in slope with decreasing fracture
size. Truncation artifacts can be minimized by artifi-
cially imposing a threshold size for fracture measure-
ment that can be readily observed at all locations in a
sampling domain. Although smaller fractures might be
visible (at least locally), fractures of such size might not
be reliably measured everywhere in the sampling
domain and lead to an incomplete sample.
Censoring Artifacts
Censoring artifacts (Baecher and Lanney, 1978; Laslett,
1982; Barton and Zoback, 1992; Pickering et al., 1995)
have been best described for fracture-length distribu-
tions obtained from two-dimensional domains (i.e.,
maps). Censoring occurs where some or all of the
largest fractures in a population are incompletely sam-
pled, or only minimum estimates of their sizes can be
made. Censoring of fracture-length measurements can
result from fracture traces that continue beyond the
limits of an observation domain, so their true sizes
cannot be determined. Censoring of fracture-aperture
measurements can happen in core-based studies be-
cause the largest aperture fractures commonly result
in incomplete core recovery, so some of the largest
fracture apertures are preferentially missed. Enhanced
erosion at large fractures in outcrops can produce a
similar effect. Inclusion of minimum estimates of frac-
ture length in a size distribution acknowledges the
presence of such fractures but degrades accuracy of the
length distribution. Omission of large fractures in a
size distribution produces a more pronounced effect.
Censoring results in an apparent homogeneity of
large fracture sizes compared with the wider range of
sizes for uncensored fractures. Censoring dispropor-
tionately affects data from the largest fractures be-
cause the probability of censoring is proportional to
fracture size. Censoring effects on a fracture-size distri-
bution typically show a convex-upward curve that in-
creases in slope with increasing fracture size.
Topologic Artifacts
Topologic artifacts are best illustrated by referring to a
specific example. Assume that several layer-perpen-
dicular fractures are randomly distributed in a layer of
rock, and that these fractures follow a power-law dis-
tribution of sizes. The volume of rock will contain
some fractures that span the thickness of the layer and
other fractures whose tip lines are partly or totally
embedded within the layer. In a layer-parallel two-
dimensional slice of this volume, the probability of
sampling a fracture that does not reach both layer
boundaries depends on the fracture height, but a com-
plete sampling of fractures that span the layer thickness
will be obtained because they will intersect any layer-
parallel two-dimensional slice. Thismeans that a three-
dimensional sampling of spanning fractures is effectively
obtained from a two-dimensional observation domain.
However, fractures that do not span the layer thickness
will show lower apparent abundance than their true
abundance in three dimensions. The cumulative size
distribution will reflect this change in apparent abun-
dance as a discrete decrease in the slope of the power-
law distribution for nonspanning fractures, which typi-
cally are most abundant (Marrett, 1996).
A similar reasoning can be extended to one-
dimensional sampling domains. A scan line perpen-
dicular to the fractures in the layer will sample a sub-
set of all fractures in the volume. The probability that
a scan line will sample a fracture is proportional to the
fracture surface area. In this case, not only fracture
height but also fracture length and the geometry of the
fracture surfacewill determine the sampling probability.
202 Geohorizons
Fractures that span the layer and the limits of the study
volumewill reflect three-dimensional sampling. Fractures
that do not reach the bed boundaries or the lateral bound-
aries of the study region will reflect one-dimensional
sampling.
Mechanical Boundary Effect
The effect of mechanical-layer thickness on fracture-
height distribution is trivial and can be explained using
topologic arguments. However, the effects of mechan-
ical-layer boundaries on fracture-length and aperture
distributions are poorly known. Ortega and Marrett
(2000) showed an example in which uncensored
fractures in fracture-length distributions deviate from
ideal power-law distributions with exponents that can-
not be explained by topologic artifacts. It is possible
that limits to fracture propagation affect the relation-
ship between aperture and length; however, no system-
atic study of fracture-aperture– and fracture-length–
scaling variations below and above the length scale of
the mechanical layer has been done. If the length and/
or aperture growth rates are perturbed when fractures
reach the bed boundaries, then nontopologic effects
might result, and these might be difficult to distinguish
from topologic effects.
Undersampling of Large Fractures
Power-law scaling predicts that the largest fractures
are least common in a fractured volume of rock. Con-
sequently,most size distributions are poorly constrained
at the large-scale end of the distribution. In addition,
the spatial distribution of fractures will affect sam-
pling probabilities. For example, if fractures are highly
clustered and a short scan line intersects a fracture
swarm, then the size distribution may show an anom-
alously high number of fractures for the size of the
observation domain. Fracture clusters commonly in-
clude the largest fractures of a population (Gomez,
2004). Mechanical modeling of en echelon fracture
arrangements suggests that the enhanced growth of
clustered large fractures is favored in this type of spatial
distribution (Olson and Pollard, 1991; Olson, 2004).
More typically, scan lines shorter than the half-spacing
between clusters will undersample large fractures. Con-
sequently, spatial distribution of fractures affects esti-
mates of a large-fracture intensity. The calculation of
intensity or average spacing of small fractures is more
reliable, particularly if long scan lines and large amounts
of data are collected.
Quantification of Uncertainties
Measuring fracture intensity for a volume of rock is
challenging, but estimating uncertainties is even more
difficult. In the previous sections, we discussed how
the scales of observation and fracture size affect esti-
mates of fracture intensity. The use of fracture-size
distributions promises a solution to the influence of
fracture size on fracture intensity by choosing a com-
mon fracture size as the basis of fracture-intensity
comparisons. However, sampling, topology, number
of observations, and spatial architecture of fracture
arrays can limit the accurate attainment of the under-
lying size distribution for a given fracture system. For
example, we expect a progressively more reliable ap-
proximation of the fracture intensity to be obtained as
we collect progressively more fracture data by ex-
tending an observation scan line. Extending the scan
line reduces possible spatial distribution effects on
fracture-size distributions and improves the likelihood
of adequately sampling the largest fractures.
One estimate of the uncertainty in fracture-intensity
determinations is the variance of consecutive fracture-
intensity estimates as data collection progresses along
a scan line. To evaluate this uncertainty, the normal-
ized fracture intensity can be calculated for progres-
sively larger parts of a complete scan line, choosing a
minimum fracture size to be considered for fracture
intensity (this does not have to be the threshold size
for fracture measurement). The standard deviation of
the normalized fracture-intensity estimates can then
be calculated for every new fracture encountered along
the scan line, using all of the estimates obtained up to
that point along the scan line. The normalized fracture-
intensity estimate and its uncertainty can be plotted
in a graph of the normalized fracture frequency ver-
sus the fracture number along the scan line (Figure 8).
Such graphs show that when a fracture of the mini-
mum size or larger is found along an observation scan
line, normalized abundance increases if the last spac-
ing is smaller than the average spacing up to that point.
However, if the last spacing is greater than the average
spacing up to that point, the normalized abundance de-
creases.Where only fractures smaller than theminimum
size are found, fracture-intensity estimates decrease be-
cause of the increase in scan-line length. The standard
deviation of successive normalized fracture intensities
provides a measure of the variability of fracture in-
tensity. One standard deviation around the expected
fracture-intensity estimate encompasses approximately
68% of the normalized intensity estimates determined
Ortega et al. 203
up to a given spot along the scan line (Koch and Link,
1971).
For the estimates of intensity uncertainties in
Figure 8, the first 20 fracture-intensity estimates were
ignored because the normalized fracture intensity is
unstable when a small number of fractures are used.
We assumed that the normalized fracture intensity up
to a given point in the scan line is independent of pre-
vious determinations, and thus, the expected value is the
last normalized intensity obtained. We justify the use
of this method because we expect a progressively more
representative fracture-intensity determination as the
scan line progresses, whereas a mean fracture-intensity
determination is a worse approximation. Estimation of
uncertainty (i.e., dispersion) using one standard devi-
ation around the expected value is conservative.
The determination of total fracture intensity in a
bed requires accounting for all fracture sets present in
the bed. We do this by simply adding the scan-line–
based estimates of fracture intensity for each set. The
propagation of errors for the addition of normalized
fracture intensities must be estimated. If Q is the sum
of the fracture intensities for all sets (e.g., sum of in-
tensities for sets A, B, C, etc.), we can obtain the ap-
proximate uncertainty sQ as
sQ ¼ffiffiffiffiffiffiffiffiffiffiffiffiffiXni¼1
s2i
sð3Þ
where s is the standard deviation, and i represents eachof the n fracture sets present in the bed (Young, 1962).
Anotherway of estimating the combined uncertainty of
the sum of fracture intensities (sall) is by weighting the
intensity of each fracture set (Fi) by the inverse of the
uncertainty associated with each of the estimate (1/si).This method of weighting aims to decrease the impact
of the least confident estimates commonly associated
with problematic size distributions or limited data and
provides a better approximation to the uncertainty of
fracture intensity in the bed. The formula used to cal-
culate this dispersion is
sall ¼
Pni¼1
Fi
Pni¼1
Fisi
ð4Þ
Uncertainty estimates using equations 3 and 4 prove
to be similar in most cases, but consistently higher es-
timates result from equation 3. The uncertainties re-
ported below and in Figure 8 were obtained using
equation 4.
COMPARISON OF FRACTURE INTENSITY INTWO LAYERS
For a comparison of the results shown in Figures 6
and 8, we provide analogous results from another
layer studied in the Sierra Madre Oriental. Bed 7 is a
carbonate mudstone approximately 60 cm (24 in.)
thick and only slightly dolomitized (10% of the total
volume of rock) studied near Iturbide, northeastern
Mexico. This bed is thicker than bed 3. Thicker beds
Figure 8. Uncertainty (68% confidenceinterval) for fracture-intensity estimateof set A in bed 3, La Escalera Canyon. Onlyfractures having an aperture of 0.2 mm(0.008 in.) or larger were included in thefracture-intensity estimate. The graphsuggests that the fracture-intensity aver-age does not change much after collectingthe first 100 fractures. The uncertaintyin the intensity determination diminishesprogressively toward the end of the scanline, suggesting that there are no majorchanges in the spatial distribution of frac-tures in the scan line measured. Largechanges in fracture intensity at the begin-ning of the scan line are expected forsparse data.
204 Geohorizons
are expected to have a wider average spacing than
thinner beds (Bogdonov, 1947; Price, 1966; Ladeira
and Price, 1981; Huang and Angelier, 1989; Narr
and Suppe, 1991; Gross and Engelder, 1995; Wu and
Pollard, 1995; Becker and Gross, 1996; Pascal et al.,
1997; Ji and Saruwatari, 1998; Bai and Pollard, 2000),
and consequently, thicker beds have lower fracture
intensity than thinner beds. Bed 7 is also less dolomi-
tized than bed 3. Sinclair (1980) found that dolostones
have higher fracture intensity than limestones. Both of
these considerations indicate that bed 7 is expected to
have lower fracture intensity than bed 3. The fracture-
aperture distribution obtained frommeasurements along
a scan line perpendicular to set C in bed 7 produces a
fracture-size distribution that can be adequately mod-
eled by a power law across more than one order of
magnitude variation in aperture (Figure 9). Set C is
the most abundant fracture set in bed 7, and the
fracture set studied in bed 3 is also the most abundant
in bed 7.
For comparing fracture intensity in the two layers,
we use 0.2 mm (0.008 in.) as the minimum aperture
considered for fracture-intensity estimation. Aper-
tures as small as 0.05 mm (0.002 in.) were measured
in both layers, but apertures of 0.2 mm (0.008 in.)
were well constrained for all layers in our Sierra Madre
Oriental database. Fracture intensity in bed 3 is nearly
five times higher than the fracture intensity in bed 7
(Figure 10). Confidence intervals for fracture intensi-
ties in the two beds do not overlap after 50 or more
fractures have been observed, suggesting that fracture
intensity in bed 7 is significantly lower than in bed 3.
The uncertainty of fracture intensity is slightly higher
for bed 3 than for bed 7, suggesting that fractures
having apertures of 0.2 mm (0.008 in.) or larger are
slightly more clustered in bed 3 than in bed 7. The
variation of fracture intensity and uncertainty in bed 7
suggest that several apertures 0.2 mm (0.008 in.) or
larger were encountered near the beginning of the scan
line. After encountering 50 fractures, the fracture in-
tensity progressively decreases, and the fracture inten-
sity changes little after sampling 100 fractures, al-
though it continues decreasing very slightly until the
end of the scan line is reached.
The difference in the fracture intensity obtained
from beds 3 and 7, where 229 and 197 fractures were
measured, respectively, implies considerably different
scan-line lengths. The scan line for bed 3 was 1.7 m
(5.5 ft) long, but a scan line 6.2 m (20.3 ft) long was
needed to encounter a comparable number of frac-
tures in bed 7.
CONCLUSIONS
Fracture-intensity estimates for a single rock volume
vary according to the observation scale because of varia-
tion of the smallest fracture size included. A compari-
son of fracture intensities among different volumes of
rock requires fracture-intensity estimates that share a
common minimum size of fracture. The method de-
scribed here provides a scale-independent approach
to fracture intensity that uses fracture-size distributions
Figure 9. Aperture distribution offracture set C in bed 7, Iturbide. Theaperture distribution can be ade-quately modeled by a power law formore than an order of magnitudevariation in aperture. The coefficientof this power law is approximatelya factor of 8 smaller than the coef-ficient in the power-law regressionfor bed 3, suggesting that this bed haslower fracture intensity than thefirst bed. Filled squares are fielddata used for regression. In the for-mula, m� 1 and mm�1 refer to theleft-hand and right-hand scales.
Ortega et al. 205
and permits choosing the minimum fracture size such
that sampling artifacts are minimized. The technique
can be used to estimate the fracture intensity and po-
rosity in the subsurface.
Uncertainty of fracture-intensity estimates is ob-
tained by calculating the standard deviation of fracture
abundance as collection of data progresses. A graph of
the variation of fracture abundance and uncertainty as
data collection progresses provides information on the
clustering of fractures larger than the normalization
size. This graph also illustrates whether the number of
data collected are sufficient for a stable estimate of
fracture intensity.
The use of scale-independent fracture-intensity es-
timates allows a reassessment of the geologic controls
on fracture intensity that have hitherto been broadly
accepted. Because previous methods do not account
for fracture size, it is likely that many fracture-intensity
estimates reflect arbitrary (or at least unexamined)
choices or assumptions about which fractures to mea-
sure. Moreover, this variation might have been sys-
tematically related to the very geologic controls under
examination. For example, if smaller fractures aremore
readily visible in one lithology than another, then biased
fracture-intensity estimates can falsely indicate a pat-
tern of geologic control. Lithologic controls on fracture
intensity need to be reexaminedwith appropriate scale-
independent estimators.
REFERENCES CITED
Ackermann, R. V., and R. W. Schlische, 1997, Anticlustering ofsmall normal faults around larger faults: Geology, v. 25, no. 12,p. 1127–1130.
Baecher, G. B., and N. Lanney, 1978, Trace length biases in jointsurveys, in Proceedings of the 19th Symposium on Rock Me-chanics: Rotterdam, Balkema, v. 1, p. 56–65.
Bai, T., and D. D. Pollard., 2000, Fracture spacing in layered rocks:A new explanation based on the stress transition: Journal ofStructural Geology, v. 22, p. 43–57.
Barton, C. A., and M. D. Zoback, 1992, Self-similar distribution andproperties of macroscopic fractures at depth in crystalline rockin the Cajon Pass Scientific Drill Hole: Journal of GeophysicalResearch, B, v. 97, no. 4, p. 5181–5200.
Becker, A., and M. R. Gross, 1996, Mechanisms for joint saturationin mechanically layered rocks— An example from southernIsrael: Tectonophysics, v. 257, p. 223–237.
Belfield, W. C., and J. P. Sovich, 1995, Fracture statistics fromhorizontal wellbores: Journal of Canadian PetroleumTechnolo-gy, v. 34, no. 6, p. 47–50.
Benavides, L., 1956, Notas sobre la geologia petrolera de Mexico, inE. J. Guzman, ed., Symposium sobre yacimientos de petroleo ygas: XX Congreso Geologico Internacional, Tomo III, p. 350–562.
Figure 10. Comparison of fracture intensity and uncertainty for beds 3 and 7, La Escalera Canyon and Iturbide, respectively. Onlyfractures having aperture of 0.2 mm (0.008 in.) or larger were included in fracture-intensity estimates. In bed 7, fracture intensityremains approximately constant after the first 100 fractures measured. The uncertainties for fracture intensity in the two beds studieddo not overlap at the 68% confidence level after 50% or more fractures have been observed, suggesting that the fracture intensity inbed 7 is statistically lower than in bed 3. The scan-line length used to estimate the fracture intensity in bed 3 (229 fracturesmeasured) was 1.7 m (5.5 ft) long. The scan-line length used to estimate the fracture intensity in bed 7 (197 fractures measured) was6.2 m (20.3 ft) long. Normalized fracture intensities and their associated uncertainties allow the direct comparison of fractureabundance in these beds, independent of observation scale and resolution levels.
206 Geohorizons
Bogdonov, A. A., 1947, The intensity of cleavage as related to thethickness of beds: Soviet Geology, v. 16, p. 102–104.
Bonnet, E., O. Bour, N. E. Odling, P. Davy, I. Main, P. Cowie, andB. Berkovitz, 2001, Scaling of fracture systems in geologicalmedia: Reviews of Geophysics, v. 39, no. 3, p. 347–383.
Castaing, C., et al., 1996, Scaling relationships in intraplate fracturesystems related to Red Sea rifting: Tectonophysics, v. 261,p. 291–314.
Clark, M. B., S. L. Brantley, and D. M. Fisher, 1995, Power-law veinthickness distributions and positive feedback in vein growth:Geology, v. 23, p. 975–978.
Corbett, K., M. Friedman, and J. Spang, 1987, Fracture develop-ment and mechanical stratigraphy of Austin Chalk, Texas:AAPG Bulletin, v. 71, no. 1, p. 17–28.
Das Gupta, U., 1978, A study of fractured reservoir rocks, withspecial reference to Mississippian carbonate rocks of southwestAlberta: Ph.D. thesis, University of Toronto, Toronto, 261 p.
Gillespie, P. A., C. B. Howard, J. J. Walsh, and J. Watterson, 1993,Measurement and characterization of spatial distributions offractures: Tectonophysics, v. 226, p. 113–141.
Goldhammer, R. G., P. J. Lehman, R. G. Todd, J. L. Wilson, W. C.Ward, and C. R. Johnson, 1991, Sequence stratigraphy andcyclostratigraphy of the Mesozoic of the Sierra Oriental,northeast Mexico: Gulf Coast Section SEPM, 85 p.
Gomez, L., 2004, Predicting macrofracture spatial arrangementfrom small rock samples: Testing new analytical techniquesusing microfracture spacing (abs.): AAPG Annual MeetingProgram, v. 13, p. A53.
Gross, M. R., 1993, The origin and spacing of cross joints: Exam-ples from the Monterrey Formation, Santa Barbara coastline,California: Journal of Structural Geology, v. 15, no. 6, p. 737–751.
Gross, M. R., and T. Engelder, 1995, Strain accommodated bybrittle failure in adjacent units of the Monterey Formation,U.S.A.: Scale effects and evidence for uniform displacementboundary conditions: Journal of Structural Geology, v. 17,p. 1303–1318.
Gudmundsson, A., 1987, Geometry, formation, and developmentof tectonic fractures on the Reykjanes Peninsula, southwestIceland: Tectonophysics, v. 139, p. 295–308.
Handin, J., R. V. Hager Jr., M. Friedman, and J. N. Feather, 1963,Experimental deformation of sedimentary rocks under confin-ing pressure; pore pressure tests: AAPG Bulletin, v. 47, no. 5,p. 717–755.
Hatton, C. G., I. G. Main, and P. G. Meredith, 1994, Non universalscaling of fracture length and opening displacement: Nature,v. 367, p. 160–162.
Heffer, K. J., and T. G. Bevan, 1990, Scaling relationships andnatural fractures: Data, theory and applications: Society ofPetroleum Engineers, Europec 90, The Hague, October 22–24, SPE Paper 20981, p. 367–376.
Huang, Q., and J. Angelier, 1989, Fracture spacing and its relation tobed thickness: Geological Magazine, v. 126, no. 4, p. 355–362.
Ji, S., and K. Saruwatari, 1998, A revised model for the relationshipbetween joint spacing and layer thickness: Journal of StructuralGeology, v. 20, no. 11, p. 1495–1508.
Johnston, J. D., and K. J. W. McCaffrey, 1996, Fractal geometries ofvein systems and the variation of scaling relationships with mech-anism: Journal of Structural Geology, v. 18, no. 2–3, p. 349–358.
Koch Jr., G. S., and R. F. Link, 1971, Statistical analysis of geo-logical data: New York, Dover Publications, Inc., 438 p.
Ladeira, F. L., and N. J. Price, 1981, Relationship between fracturespacing and bed thickness: Journal of Structural Geology, v. 3,no. 2, p. 179–183.
Laslett, G. M., 1982, Censoring and edge effects in areal and line
transects sampling of rock joint traces: Mathematical Geology,v. 14, no. 2, p. 125–140.
Laubach, S. E., 1997, A method to detect natural fracture strike insandstones: AAPG Bulletin, v. 81, no. 4, p. 604–623.
Laubach, S. E., 2003, Practical approaches to identifying sealed andopen fractures: AAPG Bulletin, v. 87, no. 4, p. 561–579.
Laubach, S. E., and C. M. Tremain, 1991, Regional coal fracturepatterns and coalbed methane development, in J. C. Rogiers,ed., Rock mechanics as multidisciplinary science: Proceedingsof the 32nd U.S. Symposium: Rotterdam, A. A. Balkema,p. 851–859.
Laubach, S. E., R. A. Marrett, J. E. Olson, and A. R. Scott, 1998,Characteristics and origins of coal cleat: A review: Interna-tional Journal of Coal Geology, v. 35, p. 175–207.
Laubach, S. E., R. M. Reed, J. E. Olson, R. H. Lander, and L. M.Bonnell, 2004, Coevolution of crack-seal texture and fractureporosity in sedimentary rocks: Cathodoluminescence observa-tions of regional fractures: Journal of Structural Geology, v. 26,no. 5, p. 967–982.
Lorenz, J. C., and R. E. Hill, 1991, Subsurface fracture spacing:Comparison of inferences from slant/horizontal core andvertical core in Mesaverde reservoirs: Society of PetroleumEngineers, Rocky Mountain Regional Meeting and Low-Permeability Reservoirs Symposium, Denver, April 15–17,SPE Paper 21877, p. 705–716.
Mandal, N., S. K. Deb, and D. Khan, 1994, Evidence for a non-linear relationship between fracture spacing and layer thickness:Journal of Structural Geology, v. 16, no. 9, p. 1275–1281.
Marin, B. A., S. J. Clift, H. S. Hamlin, and S. E. Laubach, 1993,Natural fractures in Sonora Canyon sandstones, Sonora andSawyer fields, Sutton County, Texas, in Rocky Mountain Re-gional Meeting/Low Permeability Reservoirs Symposium: So-ciety of Petroleum Engineers, SPE Paper 25895, p. 523–531.
Marrett, R., 1996, Aggregate properties of fracture populations, inP. A. Cowie, R. J. Knipe, I. G. Main, and S. F. Wojtal, eds.,Special issue: Scaling laws for fault and fracture populations;analyses and applications: Journal of Structural Geology, v. 18,no. 2–3, p. 169–178.
Marrett, R., 1997, Permeability, porosity, and shear-wave anisotro-py from scaling of open fracture populations, in T. E. Hoak,A. L. Klawitter, and P. K. Blomquist, eds., Fractured reser-voirs: Characterization and modeling guidebook: Rocky Moun-tain Association of Geologists, p. 217–226.
Marrett, R., and M. Aranda-Garcıa, 1999, Structure and kinematicdevelopment of the Sierra Madre Oriental fold-thrust belt,Mexico, in J. L. Wilson, W. C. Ward, and R. A. Marrett, com-pilers and eds., Stratigraphy and structure of the Jurassic andCretaceous platform and basin systems of the Sierra MadreOriental, Monterrey and Saltillo areas, northeastern Mexico, afield book and related papers: South Texas Geological Society,for the Annual Meeting of the AAPG and SEPM, p. 69–98.
Marrett, R., and S. E. Laubach, 2001, Fracturing during burialdiagenesis, in R. Marrett, ed., Genesis and controls of reservoir-scale carbonate deformation, Monterrey salient, Mexico: Uni-versity of Texas at Austin Bureau of Economic Geology Guide-book, v. 28, p. 109–120.
Marrett, R., O. Ortega, and C. Kelsey, 1999, Extent of power-lawscaling for natural fractures in rock: Geology, v. 27, no. 9,p. 799–802.
Mauldon, M., W. M. Dunne, and M. B. Rohrbaugh Jr., 2001, Cir-cular scanlines and circular windows: New tools for charac-terizing the geometry of fracture traces: Journal of StructuralGeology, v. 23, p. 247–258.
Narr, W., 1991, Fracture density in the deep subsurface: Techniqueswith application to Point Arguello oil field: AAPG Bulletin,v. 75, no. 8, p. 1300–1323.
Ortega et al. 207
Narr, W., 1996, Estimating average fracture spacing in subsurfacerock: AAPG Bulletin, v. 80, no. 10, p. 1565–1586.
Narr, W., and J. Suppe, 1991, Joint spacing in sedimentary rocks:Journal of Structural Geology, v. 13, p. 1037–1048.
Nelson, R. A., 1985, Geologic analysis of naturally fracturedreservoirs: Houston, Gulf Publishing, 320 p.
Olson, J. E., 1997, Natural fracture pattern characterization using amechanically-based model constrained by geologic data—Moving closer to a predictive tool: International Journal ofRock Mechanics and Mining Sciences, v. 34, no. 3–4, paperno. 237, p. 403 and CD-ROM.
Olson, J. E., 2003, Sublinear scaling of fracture aperture versuslength: An exception or the rule?: Journal of GeophysicalResearch, B9, v. 108, no. 2413, doi:10.1029/2001JB000419.
Olson, J. E., 2004, Predicting fracture swarms— The influence ofsubcritical crack growth and the crack-tip process zone on jointspacing in rock, in J. W. Cosgrove and T. Engelder, eds., Theinitiation, propagation, and arrest of joints and other fractures:Geological Society (London) Special Publication 231, p. 73–87.
Olson, J. E., and D. D. Pollard, 1991, The initiation and growth ofen echelon veins: Journal of Structural Geology, v. 13, no. 5,p. 595–608.
Ortega, O., and R. Marrett, 2000, Prediction of macrofractureproperties using microfracture information, Mesaverde Groupsandstones, San Juan basin, New Mexico: Journal of StructuralGeology, v. 22, no. 5, p. 571–588.
Ortega, O., R. Marrett, S. Hamlin, S. Clift, and R. Reed, 1998,Quantitative macrofracture prediction using microfractureobservations: A successful case study in the Ozona Sandstone,west Texas (abs.): AAPG Annual Meeting Program, v. 7,p. A503.
Pascal, C., J. Angelier, M.-C. Cacas, and P. L. Hancock, 1997,Distribution of joints: Probabilistic modeling and case studynear Cardiff (Wales, U.K.): Journal of Structural Geology,v. 19, p. 1273–1284.
Pickering, G., J. M. Bull, and D. J. Sanderson, 1995, Samplingpower-law distributions: Tectonophysics, v. 248, p. 1–20.
Pineau, A., 1985, Echantillonnage des espacements entre fractures:
Une distribution exponentielle negative troquee: ComptesRendus de l’Academie des Sciences, II, v. 301, p. 1043–1046.
Price, N. J., 1966, Fault and joint development in brittle andsemibrittle rocks: Oxford, Pergamon Press, 176 p.
Priest, S. D., and J. A. Hudson, 1976, Discontinuity spacing in rock:International Journal of Rock Mechanics, Mining Science, andGeomechanics Abstracts, v. 13, p. 135–148.
Rives, T., M. Razack, J.-P. Petit, and K. D. Rawnsley, 1992, Jointspacing: Analog and numerical simulations: Journal of Struc-tural Geology, v. 14, p. 925–937.
Rives, T., K. D. Rawnsley, and J.-P. Petit, 1994, Analoguesimulation of natural orthogonal joint set formation in brittlevarnish: Journal of Structural Geology, v. 16, no. 3, p. 419–429.
Sanderson, D. J., S. Roberts, and P. Gumiel, 1994, A fractalrelationship between vein thickness and gold grade in drill corefrom La Codosera, Spain: Economic Geology, v. 89, p. 168–173.
Sen, Z., and A. Kazi, 1984, Discontinuity spacings and RQD es-timates from finite-length scanlines: International Journal ofRock Mechanics, Mining Sciences, and Geomechanics Ab-stracts, v. 21, p. 203–212.
Sinclair, S. W., 1980, Analysis of macroscopic fractures on Tetonanticline, northwestern Montana: M.S. thesis, Texas A&MUniversity, College Station, Texas, 102 p.
Wong, T. F., J. T. Fredrich, and G. D. Gwanmesia, 1989, Crackaperture statistics and pore space fractal geometry of WesterlyGranite and Rutland Quartzite: Implications for an elasticcontact model of rock compressibility: Journal of GeophysicalResearch, v. 94, p. 10,267–10,278.
Wu, H., and D. D. Pollard, 1992, Propagation of a set of opening-mode fractures in layered brittle materials under uniaxial straincycling: Journal of Geophysical Research, v. 97, p. 3381–3396.
Wu, H., and D. D. Pollard, 1995, An experimental study of therelationship between joint spacing and layer thickness: Journalof Structural Geology, v. 17, p. 887–905.
Young, D. H., 1962, Statistical treatment of experimental data:New York, McGraw-Hill Book Company, Inc., 172 p.
208 Geohorizons