You are on page 1of 123

Stereological Interpretation of Rock Fracture Traces on

Borehole Walls and Other Cylindrical Surfaces


Xiaohai Wang
Dissertation submitted to the Faculty of the
Virginia Polytechnic Institute and State University
in partial fulfillment of the requirements for the degree of
Doctor of Philosophy
in
Civil Engineering
Mauldon, Matthew, Chair
Dove, Joseph E.
Dunne, William M.
Gutierrez, Marte S.
Westman, Erik C.
September 16, 2005
Blacksburg, Virginia

Keywords: fractures, cylindrical sampling, borehole, stereology, Monte Carlo method,


intensity measures, conversion factors, mean fracture length and width

Copyright 2005, Xiaohai Wang

Stereological Interpretation of Rock Fracture Traces on Borehole


Walls and Other Cylindrical Surfaces
Xiaohai Wang

Abstract
Fracture systems or networks always control the stability, deformability, fluid and gas
storage capacity and permeability, and other mechanical and hydraulic behavior of rock
masses. The characterization of fracture systems is of great significance for
understanding and analyzing the impact of fractures to rock mass behavior. Fracture
trace data have long been used by engineers and geologists to character fracture system.
For subsurface fractures, however, boreholes, wells, tunnels and other cylindrical
samplings of fractures often provide high quality fracture trace data and have not been
sufficiently utilized. The research work presented herein is intended to interpret fracture
traces on borehole walls and other cylindrical surfaces by using stereology. The
relationships between the three-dimension fracture intensity measure, P32, and the lower
dimension fracture intensity measures are studied. The analytical results show that the
conversion factor between the three-dimension fracture intensity measure and the twodimension intensity measure on borehole surface is not dependent on fracture size, shape
or circular cylinder radius, but is related to the orientation of the cylinder and the
orientation distribution of fractures weight by area. The conversion factor between the
two intensity measures is determined to be in the range of [1.0, /2]. The conversion
factors are also discussed when sampling in constant sized or unbounded fractures with
orientation of Fisher distribution. At last, the author proposed estimators for mean
fracture size (length and width) with borehole/shaft samplings in sedimentary rocks based
on a probabilistic model. The estimators and the intensity conversion factors are tested
and have got satisfactory results by Monte Carlo simulations.

Acknowledgments
I am indebted to the assistance of my dissertation committee: Dr. Matthew Mauldon, Dr.
Joseph E. Dove, Dr. William M. Dunne, Dr. Marte S. Gutierrez, and Dr. Erik C.
Westman. From my proposal to the final form of this dissertation, they have given great
amount of valuable suggestions and made the study in this Ph.D. program priceless
experience to me.
My advisor, Matthew Mauldon, whom I met two weeks after I arrived at this country,
generously provided the support for me to enroll as a Ph.D. student. In the passed four
years, he and his insights had showed me many times the lights of the way and lead me
out of the darkness of confusion and uncertainty. Though, what I have learned from him
is far beyond what I can put in words. I Thank Matthew, his wife Amy and their
daughters for their kindness and support.
Special thanks to Dr. Dunne and his student Chris Heiny in the University of Tennessee.
The collaborations with them on fracture size estimators pushed the dissertation to a new
level. Their work and suggestions as geologists have made the estimators more practical
and useful.
I also owed thanks to Jeremy Decker of Virginia Tech, who helped me testing my
program and carrying out numerous simulations. I always regret that I can not include in
my dissertation the great figures he worked out in Matlab.
I am grateful to have my friends around me in the years in Ozawa library and Rm19,
Patton Hall. My colleagues consideration and thoughtfulness makes the days and nights
in the office wonderful memory.
Last, but not least, I am beholden to my wife Hui Cheng, her family and my family in
China. Without their great love, this dissertation is impossible.

iii

Contents
Acknowledgments .........................................................................................iii
1 Introduction.................................................................................................. 1
2 Multi-dimensional intensity measures for Fisher-distributed fractures ...... 3
2.1 Introduction ............................................................................................................3
2.2 General form of conversions ..................................................................................4
2.3 Linear and planar sampling of fisher-distributed fractures.....................................7
2.4 Sampling on a cylindrical surface ........................................................................14
2.5 Example: 3-d fracture intensity inferred from scanline data ................................19
2.6 Discussion and Conclusions .................................................................................22
Acknowledgments ......................................................................................................22
Appendix 2.A Probability density function (pdf) f ( ) of angle ........................23
Appendix 2.B Numerical approach for obtaining F() and F() ...........................27
References...................................................................................................................28

3 Estimating fracture intensity from traces on cylindrical exposures .......... 31


3.1 Introduction ..........................................................................................................32
3.2 Basic assumptions.................................................................................................35
3.3 General form of the relationship between areal intensity P21,C and volumetric
intensity P32 for right circular cylinders ..................................................................36
3.4 General case of cylindrical sampling....................................................................38
3.5 Special case: Sampling fractures of constant orientation .....................................42
3.6 Special case: fractures with uniform orientation distribution...............................45
3.7 Cycloidal Scanline Technique ..............................................................................47
3.7.1 Unbiased sampling criterion........................................................................47
3.7.2 Cycloidal scanlines......................................................................................50
3.8 Monte Carlo Simulations......................................................................................50

iv

3.9 Discussion & Conclusions....................................................................................54


Acknowledgements.....................................................................................................55
Appendix 3.A Determine |cos | .................................................................................56
References...................................................................................................................59

4 Estimating length and width of rectangular fractures from traces on


cylindrical exposures ............................................................................ 62
4.1 Introduction ..........................................................................................................63
4.2 Assumptions .........................................................................................................67
4.4 Probabilistic model for occurrence of intersection types .....................................71
4.4.1 w > D ..........................................................................................................77
4.4.2 w D ..........................................................................................................79
4.4.3 Summary of fracture length and width estimators ......................................84
4.5 Examples ..............................................................................................................86
4.6 Monte Carlo simulations ......................................................................................90
4.7 Discussion & Conclusions....................................................................................96
References...................................................................................................................98

5 Conclusions and discussions....................................................................103


6 Appendix: Programs used in the dissertation ..........................................106
A. FISHER - Simulate the Fisher distribution.........................................................106
B. TRACE - Simulate fracture population sampled by a borehole ..........................107

7 Vita...........................................................................................................111

List of Figures

Fig. 2.1. Geometry of linear and planar sampling of a fracture.......................................... 5


Fig. 2.2. For an isotropic fracture orientation distribution, the distributions of and are
proportional to the sin and sin , respectively. ................................................. 7
Fig. 2.3. Spherical triangle formed by n, m, and s, where n is fracture normal, m is
Fisher mean pole, and s is the sampling line....................................................... 9
Fig. 2.4. Fisher-distributed fracture normals in relation to sampling line (+). ................. 10
Fig. 2.5. pdfs of Fisher distribution with = 20, 40, and 100......................................... 10
Fig. 2.6. Coefficients a, b and c for conversion factor [1/C13 ] as functions of Fisher
constant . ......................................................................................................... 12
Fig. 2.7. Coefficients a, b and c for conversion factor [1/C23 ] as functions of Fisher
constant . ......................................................................................................... 13
Fig. 2.8. Cylindrical sampling of Fisher-distributed fractures with mean pole m. The
shaded area is a slice of the cylinder surface with normal c. ............................ 15
Fig. 2.9. Cylinder axis (z), Fisher mean pole (m), and normal (c) of a slice on the
cylinder surface.................................................................................................. 16
Fig. 2.10. Coefficients a, b and c for conversion factor [1/C23,C] as functions of Fisher
constant . ......................................................................................................... 18
Fig. 2.11. Fracture normals () and mean pole () in lower hemisphere projection....... 21
Fig. 2.A-1. Coordinate system for spherical triangle formed by m, s and n. .................. 24
Fig. 2.A-2 The figure shows the range, R , of , as a function of , and . Angle
(between m and s) is a constant. ....................................................................... 26
Fig. 3.1. Borehole or shaft sampling of fractures in a rock mass. .................................... 33
Fig. 3.2. Fracture traces on a cylindrical shaft. Intersections between fractures and the
shaft are traces (curved line segments) on the shaft surface.............................. 38
Fig. 3.3. A thin slice of the shell sampling in fractures. The total trace length on its
surface is dl. ....................................................................................................... 39

vi

Fig. 3.4. A cylindrical shell (axis Z, height = H) intersects a set of fractures with constant
orientation (normal n)........................................................................................ 43
Fig. 3.5. For cylindrical sampling in fractures with constant orientation, the correction
factor C23,C between areal intensity P21,C and volumetric intensity P32 is a
function of angle 0 between the cylinder axis and fracture normal. ................ 45
Fig. 3.6. Illustration of linear (vector) IUR sampling in 3-d space................................... 49
Fig. 3.7. The cycloid (heavy curve) is the path of a point on the circle of radius r0 as the
circle rolls from left to right along the x-axis. .................................................. 51
Fig. 3.8. The computer program is used to generate rectangular fractures intersecting with
a borehole........................................................................................................... 52
Fig. 3.9. Illustration (to the scale) of the five cases studied. Shaded rectangles are
simulated fractures, and circles are sampling cylinders..................................... 53
Fig. 3.10. Simulation results of the conversion factor 1/ C23,C, compared with the
calculated curve by Eq.(3.19). ........................................................................... 54
Fig. 3.A-1. Unit vectors S, T, n, and nr in Cartesian coordinate system, where Z is
parallel to the borehole axis. The coordinates of unit vectors S and n are given
based on the geometry. ...................................................................................... 57
Fig. 4.1. Joints on limestone bed at Llantwit Major, Wales (photo provided by Matthew
Mauldon). Cross joints terminate at primary systematic joints. ........................ 65
Fig. 4.2. Schematic drawing of dipping sedimentary beds, with primary joints either
terminating on bedding planes or cutting across several layers......................... 65
Fig. 4.3. Borehole/shaft and rectangular fractures and their projections on the axis-normal
plane. Note true width w and apparent width w. .............................................. 66
Fig. 4.4. A vertical borehole of diameter D intersects rectangular fractures in six ways.
The unrolled trace map is developed from the borehole wall by cutting along
fracture dip direction. Intersection types are marked beside the corresponding
traces. ................................................................................................................. 70
Fig. 4.5 Six types of intersection between projected fractures (shaded) and
boreholes/shafts (dashed circles) are shown on the axis-normal plane. ............ 71

vii

Fig. 4.6. The locus for borehole/shaft-projected fracture intersection on the axis-normal
plane is the region inside by the dashed line. ................................................... 73
Fig. 4.7. Each intersection type has a corresponding locus on the projected fracture (bold
rectangle) for the center of the borehole. In this case, w > D. .......................... 73
Fig. 4.8. Each intersection type has a corresponding locus on the projected fracture (bold
rectangle) for the center of the borehole. In this case, D/2 < w D. ................ 74
Fig. 4.9. The corresponding locus for the center of the borehole/shaft for each intersection
type around the projected fracture (bold rectangle) on the axis-normal plane for
case w D/2...................................................................................................... 74
Fig. 4.10. Flowchart of choosing estimators to estimate mean fracture length and width.
........................................................................................................................................... 85
Fig. 4.11. A computer program was developed to generate a population of rectangular
fractures intersected by a borehole/shaft............................................................ 92
Fig. 4.12. Comparison of computed fracture length and width vs. actual fracture length
and width for scenario 1..................................................................................... 94
Fig. 4.13. Percent error and coefficient of variation of estimators for (a) fracture length
and (b) fracture width, in comparison with observed counts of B3-type
borehole/shaft-fracture intersections.................................................................. 95
Fig. App-1. The geometry of fracture, sampling cylinder, and three different shapes of
generation region. ............................................................................................ 108

viii

List of Tables

Table 2.1. Factor 1/C13 vs. and ................................................................................... 12


Table 2.2. Factor 1/C23 with different values of and . ................................................. 13
Table 2.3. 1/C23,C, the conversion factor between P21 and P32 when sampling with
cylinder surface............................................................................................... 18
Table 2.4. Orientation data for a set of fractures on the Huckleberry Trail...................... 20
Table 3.1. Simulation parameters and results. .................................................................. 53
Table 4.1. Six borehole/shaft-fracture intersection types ................................................. 69
Table 4.2. Defined symbols .............................................................................................. 72
Table 4.3. Areas of regions corresponding to each borehole/shaft-fracture intersection
type from geometry......................................................................................... 76
Table 4.4. Borehole-fracture intersection counts from a borehole sampling.................... 86
Table 4.5. Borehole-fracture intersection counts from borehole sampling ...................... 89
Table 4.6. Parameters and results of Monte Carlo simulations ........................................ 93
Table App-1. Inputs for generating the Fisher distribution. ........................................... 106
Table App-2. Parameters for simulating fractures sampled by a borehole..................... 107
Table App-3. Minimum dimension of different generation regions............................... 109

ix

This page intentionally left blank.

Chapter 1
1 Introduction

Characterization of rock fractures is essential in engineering geology, civil engineering,


mining engineering and oil and gas industry. Geometric and mechanical parameters of
fractures are widely used for estimating fractured rock mass strength, deformability,
permeability, and fluid storage capacity. Currently geological investigations have
provided a great amount of fracture data from boreholes, tunnels, shafts as well as other
cylindrical sampling surfaces. Therefore, the study of fracture characterization based on
cylindrical sampling of fractured rock mass is of great significance. In this dissertation,
the author intends to study the stereological relationships in cylindrical samplings,
unbiased scanline techniques and their applications, and estimation of fracture size in
sedimentary rocks. These studies are demonstrated in the following three chapters.
Chapter 2 discusses the conversions (linear fracture intensity measure P10, planar fracture
intensity measure P21 and cylindrical fracture intensity measure P21,C, to the volumetric
fracture intensity measure P32) appropriate for constant size or unbounded fractures with
a Fisher distribution of orientation. The corresponding paper is submitted to
Mathematical Geology. Chapter 3 discusses the estimating of fracture intensity, more
specifically, fracture volumetric intensity P32, from fracture trace data in cylindrical
(borehole, tunnel or shaft) samplings. The conversion factor between the cylindrical
fracture intensity measure P21,C and the fracture volumetric intensity P32, is presented in a
general form and some special cases are also discussed. The corresponding paper is for
submission to International Journal of Rock Mechanics & Mining Sciences. In Chapter 4
the author intends to develop a general model for estimating mean rectangular fracture
length and width from traces on cylinder walls. The corresponding paper is for
submission to Rock Mechanics & Rock Engineering.

This page intentionally left blank.

Chapter 2
2 Multi-dimensional intensity measures for Fisher-distributed
fractures
Abstract: Fracture intensity is fundamentally a three dimensional concept, relating the
total area (m2) or volume (m3) of fractures to the volume of the rock mass studied.
However, field measurements of fracture intensity in rock masses are usually either one
dimensional - along sampling lines or boreholes, or two dimensional - on tunnel walls or
trace planes. In this paper, conversions between these one and two dimensional intensity
measures, and the three dimensional intensity measure P32, are developed for constant
size or unbounded Fisher-distributed fractures, for three types of sampling domain: lines,
planes and cylindrical surfaces. Conversion factors for each of these sampling domains
are derived semi-analytically, and then computed, graphed and tabulated for a wide range
of cases. The practical significance of this work is that it enables rock engineers and
geologists to deduce 3-d fracture intensity from 1-d or 2-d field measurements.

2.1 Introduction
The Fisher distribution (Fisher, 1953) is the most commonly assumed distribution for
natural fracture orientations (Cheeney, 1983). This is largely due to its relatively simple
form, as compared to other distributions for spherical data (N. Fisher et al., 1987). The
Fisher distribution also has the advantage that it is the theoretical analogue of the normal
distribution, for spherical data. Because of these advantages, the Fisher distribution is
widely used for hydrological and geomechanical modeling in fractured rock (Cheeney,
1983; Priest, 1993).
One dimensional (1-d) and two dimensional (2-d) fracture intensity measures P10 and P21
are defined, respectively, as the number of fractures per unit length and the number of
fractures per unit area (Dershowitz & Herda, 1992; Mauldon, 1994) in the rock mass.

These measures are directionally dependent and are strongly affected by the relative
orientation of the fractures and the sampling domain, e.g., scanline, or planar surface. In
contrast, the three dimensional (3-d), or volumetric, fracture intensity measure P32,
defined as area of fractures per unit volume, is not directionally dependent (Dershowitz &
Herda, 1992; Mauldon, 1994). Measures P10 and P21 are easy to measure in the field, but
they cannot be used as general parameters to characterize fracture intensity because of
their directional dependence. For these reasons, the ability to convert linear intensity P10
or areal intensity P21 to the volumetric intensity P32, which is difficult to measure in the
field but directional independent, will be very useful.
Previous work (Dershowitz & Herda, 1992; Mauldon, 1994, Mauldon & Mauldon, 1997)
has developed some of the theoretical background for fracture intensity measures. In the
present paper, the authors derive conversions between field measures of fracture
intensity, P10 and P21, and the three dimensional volumetric fracture intensity measure P32
for fracture sets with the Fisher orientation distribution. This study focuses on fracture
orientation instead of fracture size; we assume fractures are either of constant size or are
unbounded. Based on this assumption, factors to convert measured 1-d or 2-d fracture
intensity for Fisher-distributed fractures to volumetric intensity are obtained semianalytically for sampling domains on lines, planes and cylinders.

2.2 General form of conversions


Conversions between 1-d intensity measure P10 and 3-d intensity measure P32, or between
2-d intensity measure P21 and 3-d intensity measure P32, require consideration of the
sampling bias that arises from the relative orientation of the sampling domain and the
fracture. This bias was first described by R. Terzaghi (1964), and later explored by Yow
(1987), Priest (1993), Martel (1999), and Mauldon and Mauldon (1997), among others.
In the general case, for linear or planar sampling of constant size or unbounded fractures,
P10 and P21 are related to P32 in the following ways (Dershowitz, 1992; Mauldon, 1994):

P10 = P32 | cos | f ( )d and

(2.1)

P21 = P32 sin f ( )d ,

(2.2)

where is the angle between the sampling line and the fracture normal (Fig. 2.1a); is
the angle between the sampling plane normal and fracture normal (Fig. 2.1b); and
f() and f() are the probability density functions (pdfs) of and , respectively. In
the following, we assume a statistically homogeneous sampling domain, and it is to be
understood that the given relationships refer to expected values of the intensity measures.
Here, for simplicity, the integrals in Eqs. (2.1) and (2.2) are each functions of a single
variable. The angles or are themselves functions of conventional geologic fracture
orientation parameters such as dip and dip-direction, and orientation of the sampling line
or sampling plane, and can be calculated from orientations of sampling line or plane and
fracture normal.

Fracture
normal

Sampling
line

Sampling
plane normal

Fracture
normal

Fracture

Fracture

Sampling
plane

(a)

(b)

Fig. 2.1. Geometry of linear and planar sampling of a fracture.

Define conversion factors C13 and C23 by

C13 = | cos | f ( )d
0

(2.3)

and

C 23

= sin f ( )d ,
0

(2.4)

so that
C13 P10 = P32 and

(2.5)

C 23 P21 = P32 .

(2.6)

The integrals in Eqs. (2.3) and (2.4) are on [0, 1], so the ranges of the conversion factors
C13 and C23 are from 1 to .
As an example, for the isotropic case of a uniform fracture orientation distribution,

f ( ) = 12 sin and f ( ) = 12 sin (Fig. 2.2), for and in the range [0, ].
Introducing these pdfs into (3) and (4), respectively, we have
1

C13 (isotropic )

= 2 sin | cos | d
0

C23 ( isotropic )

= 12 sin 2 d = (1 cos 2 )d ,
0

4 0

and

(2.7)
1

(2.8)

which, combining with Eqs. (2.5) and (2.6), yield (Dershowitz, 1985)
1
P10 ( isotropic ) = P32 and
2

(2.9)


P21 ( isotropic ) = P32 .
4

(2.10)

Fracture
normal

Sampling
line

Fracture
normal

Sampling
plane normal

length sin

length

Fracture
Fracture

Sampling

plane
(a)

(b)

Fig. 2.2. For an isotropic fracture orientation distribution, the distributions of and are
proportional to the sin and sin , respectively.

Eqs. (2.9) and (2.10) imply that for uniformly distributed fractures (the isotropic case),
P32 is twice the average scanline frequency and 1.27 times the mean areal trace length
intensity. In the following, we determine the conversion factors C13 and C23 for the case
of Fisher-distributed fractures.

2.3 Linear and planar sampling of fisher-distributed fractures


The probability density function of the Fisher distribution is given as (N. Fisher et al,
1987)

f ( ) =

eCos Sin
e e

(0 ) ,

(2.11)

where is the angle between a fracture normal and the Fisher mean pole (Fig. 2.3); f()
is the probability density function of ; and is the Fisher constant related to the amount
of dispersion ( has high values for low dispersion and low values for high dispersion).
Because of the radial symmetry of the Fisher distribution about its mean pole, we express
its probability density function as a function only of for a given dispersion constant.
The local azimuth of the Fisher mean pole is uniform on [0, 2] and is independent of .
Fig. 2.4 shows a set of fracture normals following the Fisher distribution, in upper
hemisphere projection. Fisher mean pole m corresponds to a plane with dip 80 and dipdirection 45. The Fisher dispersion constant in this case is equal to 60.
The theoretical range of is from 0 to , with low values indicating a high degree of
dispersion. As approaches 0, the fractures approach a uniform orientation distribution.
Typical graphs of the pdf of the Fisher distribution are shown in Fig. 2.5.
In order to obtain the conversion factor between 1-d intensity measure P10 and 3-d
intensity measure P32, we need to know f() , the probability density function of angle
between the sampling line and the fracture normal.
Based on the geometry of the spherical triangle formed by the fracture normal n, the
Fisher mean pole m and the sampling line s (Fig. 2.3), the theoretical probability density
function f() is given by (see Appendix 2.A):

f ( ) =

eCos Sin

sin
sin 2 sin 2 (cos cos cos ) 2

e e

(2.12)

for in the range | | + , where the range of integration Rd is given by:


R = [ - , + ] , if , or

R = [0, 2 ] , if > .

(2.13)

The integral in Eq. (2.12) cannot, however, be expressed in closed form. We use
numerical simulation to find the set of values of the conversion factor, following the
procedure described in Appendix 2.B.

Fig. 2.3. Spherical triangle formed by n, m, and s, where n is fracture normal, m is


Fisher mean pole, and s is the sampling line. The spherical angles , , and are,
respectively, the angles between n & s, m & n, and m & s.

North

Fisher Mean Pole m


(dip 80, dip
-direction 45)

Sampling line s (trend


225, plunge 45)
20

Fracture normal ni
Small circle
with = 20

Upper Hemisphere
Equal Area

Fig. 2.4. Fisher-distributed fracture normals in relation to sampling line (+).

= 100

6
5

= 40

f() 3
2

= 20

1
0
0

10
20 30 40
50 60 70
80
Angular deviation (deg.) from Fisher mean pole

Fig. 2.5. pdfs of Fisher distribution with = 20, 40, and 100.

10

90

Tabulated values of the factor 1/C13 (= P10/P32) are shown in Table 2.1 as a function of
the Fisher constant and angle . The reciprocal of C13, rather than C13, is tabulated in
order that values range between 0 and 1. When is relatively small ( < 1), indicating
that fracture orientations have close to a uniform distribution, the factor 1/C13 is close to
0.50, which agrees with Eq. (2.9). The factor 1/C13 can be fitted to the family of curves
given by 1 / C13 = a cos(b ) + c (Fig. 2.6). Regression coefficients a, b and c can be
computed for 1, according to the logarithmic expression given in Fig. 2.6. For < 1,
it is recommended to treat the distribution as uniform and to use the conversion factor
given by Eq. (2.9).
Following the procedure described in Appendix B, the conversion factor 1/C23 (= P21/P32)
is also computed numerically. The values of the factor 1/C23 are tabulated in Table 2.2 as
a function of and . As with the case of linear sampling, for a given value of , the
conversion factor 1/C23 is relatively insensitive to changes in for > 50. When is
relatively small ( < 1), the factor is close to 0.79, which agrees with Eq. (2.10). The
conversion factor 1/C23 can be fitted to the family of curves given by
1 / C23 = a sin(b d / 2) + c (Fig. 2.7). Regression coefficients a, b, c and d can be

computed for 1, according to the logarithmic expression given in Fig. 2.7. For < 1,
it is recommended to treat the distribution as uniform and to use the conversion factor
given by Eq. (2.10).

11

Table 2.1. Factor 1/C13 vs. and

0
5
10
20
30
40
50
60
70
80
90

0.1

10

50

100

200

500

0.50
0.50
0.50
0.50
0.50
0.50
0.50
0.50
0.50
0.50
0.50

0.53
0.53
0.53
0.53
0.52
0.51
0.51
0.49
0.48
0.48
0.48

0.62
0.62
0.61
0.59
0.56
0.54
0.51
0.48
0.45
0.44
0.44

0.79
0.79
0.78
0.75
0.68
0.62
0.54
0.47
0.39
0.34
0.33

0.90
0.89
0.88
0.84
0.77
0.67
0.57
0.45
0.34
0.26
0.24

0.98
0.97
0.96
0.91
0.84
0.74
0.62
0.47
0.32
0.18
0.11

0.99
0.98
0.97
0.92
0.85
0.75
0.62
0.48
0.32
0.16
0.08

0.99
0.99
0.98
0.93
0.85
0.75
0.63
0.48
0.32
0.16
0.06

1.00
0.99
0.98
0.93
0.85
0.75
0.63
0.48
0.32
0.16
0.04

Cos

3.0

1 / C13 = a cos(b ) + c

2.5

b = -0
.1655
ln(

Coefficient

2.0

)+ 2

1.5

.0951

1.0

7
) + 0.050 a

(
n
l
7
4
= 0.12

a
c = - 0.05

0.5
0.0
0.1

10

51 ln( )

+ 0.5988

100

c
1000

Fig. 2.6. Coefficients a, b and c for conversion factor [1/C13 ] as functions of Fisher
constant . The equations for a, b and c shown in the figure are for > 1.

12

Table 2.2. Factor 1/C23 with different values of and .

0
5
10
20
30
40
50
60
70
80
90

0.1

10

50

100

200

500

0.79
0.79
0.79
0.79
0.79
0.79
0.79
0.79
0.79
0.79
0.79

0.77
0.77
0.77
0.77
0.78
0.78
0.78
0.79
0.80
0.80
0.80

0.70
0.70
0.70
0.72
0.74
0.76
0.78
0.79
0.82
0.83
0.83

0.53
0.54
0.54
0.58
0.64
0.71
0.77
0.82
0.87
0.90
0.91

0.39
0.40
0.43
0.49
0.58
0.68
0.77
0.85
0.90
0.94
0.95

0.19
0.21
0.25
0.38
0.53
0.66
0.77
0.87
0.94
0.98
0.99

0.14
0.17
0.22
0.37
0.52
0.66
0.77
0.87
0.94
0.98
0.99

0.11
0.14
0.20
0.36
0.52
0.66
0.78
0.87
0.94
0.99
1.00

0.07
0.12
0.20
0.36
0.52
0.66
0.78
0.87
0.94
0.99
1.00

Sin

2.5

1 / C 23 = a sin(b d / 2) + c

2.0

Coefficient

b=-

0.13

1.5

d = -0

1.0

c = -0.07

0.5

.1297

ln( )

71 ln( )
+

76 ln
(

0.1

10

2.17
45

+ 1.07
83

0.8112

)+
4 ln(
6
0
1
.
a=0

0.0

)+

b
a
c
d

1
0.035

100

1000

Fig. 2.7. Coefficients a, b and c for conversion factor [1/C23 ] as functions of Fisher
constant . The equations for a, b, c and d shown in the figure are for > 1.

13

2.4 Sampling on a cylindrical surface


In this section we discuss the conversion factor between the areal intensity measure
obtained by sampling rock fractures on the surface of a cylinder, and the volumetric
intensity measure P32. Again, constant size or unbounded Fisher-distributed fractures are
assumed. The practical significance of this case arises, on one hand, from the availability
of fracture trace data obtained from borehole image or FMI and FMS logs (Dershowitz et
al., 2000), and on the other hand, from fracture trace maps obtained from circular tunnel
walls (Mauldon and Wang, 2003).
Let P21,C denote fracture trace length per unit area on the cylinder surface. For constant
size or unbounded Fisher-distributed fractures, the relationship between P21,C and
volumetric measure P32 is a function only of the angle between the Fisher mean pole m
and the cylinder axis z (Fig. 2.8) because of the circular symmetry of the sampling
surface with respect to the cylinder axis. Define C23,C as the conversion factor between
P21,C and P32 for cylinder sampling, with

C 23,C P21,C = P32 .

(2.14)

For a slice of the cylinder surface, such as the shaded area in Fig. 2.8, the normal c of the
surface element makes an angle with the mean fracture pole m. Let C23 = P32 / P21
denote the conversion factor between P21 and P32 for a sampling plane which has the
same normal as the slice (e.g., vector c in Fig. 2.8), then

1 / C23,C =

max

(1/ C ) f
23

( )d ,

(2.15)

min

where f() is the pdf of and the integration is carried out over the full range of . It
should be noted that C23 in Eq. (2.15) is a function of and refers to a specific slice (such
as the shaded strip in Fig. 2.8).

14

We adopt the Cartesian coordinate system shown in Fig. 2.9, where the xy plane is
perpendicular to the cylinder axis z, and for convenience, the x axis is selected to be
perpendicular to the zm plane. If m and c are unit vectors, then

cos = m c = sin cos ,

(2.16)

from which

cos
sin

= cos 1

(2.17)

where is the angle between the y-axis and c. Note that is uniformly distributed on [0,
2] because of the radial symmetry of the cylinder (Fig. 2.9). The pdf f ( ) of is
given by
f ( ) =

1
.
2

(2.18)

z
m

Fig. 2.8. Cylindrical sampling of Fisher-distributed fractures with mean pole m. The
shaded area is a slice of the cylinder surface with normal c.

15

z
m

(0, sin , cos )

y
c

(sin , cos , 0)

Fig. 2.9. Cylinder axis (z), Fisher mean pole (m), and normal (c) of a slice on the
cylinder surface. The xyz cylinder coordinate system is also shown.

And the cdf F ( ) of is


F ( ) = Prob ( < ) =

f ( )d ,

(2.19)

where R is the range of corresponding to < , with limits determined by Eq. (2.17)
for specified and . Noting the symmetry of the range of with respect to y-axis, and
utilizing Eq. (2.17),

F ( ) = 2

cos 1 cos
sin

1
d ,
2

(2.20)

from which
F ( ) =

cos
cos 1

sin
1

, 0 2 , ( / 2 ) ( / 2 + ) .

The pdf f ( ) of is found by differentiation as

16

(2.21)

f ( ) = F ( ) =

sin

sin (1 cos 2 / sin 2 )1 / 2

, ( / 2 ) ( / 2 + ) .

(2.22)

Substituting Eq. (2.22) into Eq. (2.15), we obtain

1 / C 23,C =

max

sin d
(1 / C 23 ) sin (1 cos 2 / sin 2 )1/ 2

min

(2.23)

where min = ( / 2 ) max = ( / 2 + ) and where (1/C23) is already given


numerically in Table 2.2.
With Table 2.2 and Eq. (2.24), the conversion factor 1/C23,C is computed numerically, and
is tabulated in Table 2.3 as a function of and . It is interesting to note that when angle
(between the Fisher mean pole and the cylinder axis) is around 60, the conversion
factor 1/C23,C is relatively insensitive to and has the value 0.79, which is also the
value obtained (/4) on any sampling surface for a uniform fracture orientation
distribution, as is implied by Eq. (2.10).
The conversion factor 1/C23,C increases to 1.0 if the cylinder axis is close to the Fisher
mean pole, and decreases to 0.64 (2/, the theoretical solution for ) if the cylinder
axis is perpendicular to the mean pole. The factor 1/C23,C can be fitted to a family of
curves given by 1 / C23,C = a cos(b ) + c (Fig. 2.10). Regression coefficients a, b and c can
be computed for 1 according to the logarithmic expression given in Fig. 2.10. For <
1, it is recommended to treat the distribution as uniform and the conversion factor1/C23,C
0.79.

17

Table 2.3. 1/C23,C, the conversion factor between P21 and P32 when sampling with
cylinder surface.

5
10
20
30
40
50
70
90

0.1
0.79
0.79
0.79
0.79
0.79
0.79
0.79
0.79

1
0.81
0.81
0.81
0.80
0.80
0.79
0.79
0.79

2
0.83
0.83
0.83
0.82
0.81
0.80
0.78
0.77

2.5

10
0.95
0.95
0.93
0.91
0.88
0.84
0.77
0.73

50
0.99
0.99
0.97
0.95
0.91
0.86
0.77
0.70

100
0.99
0.99
0.97
0.95
0.91
0.86
0.76
0.70

500
1.00
1.00
0.98
0.95
0.92
0.87
0.77
0.70

1 / C 23,C = a cos(b ) + c

2.0

Coefficients

5
0.91
0.91
0.90
0.88
0.85
0.83
0.77
0.75

b = -0.0944 ln(

1.5
1.0

) + 2.1971

c = 0.0026 ln( ) + 0.8045

0.5
a = 0.0274 ln( ) + 0.0359

0.0
0.1

10

100

a
1000

Fig. 2.10. Coefficients a, b and c for conversion factor [1/C23,C] as functions of Fisher
constant . The equations for a, b and c shown in the figure are for > 1.

18

2.5 Example: 3-d fracture intensity inferred from scanline data


Table 2.4 contains the orientation data for a set for 19 subparallel fractures. Fracture
orientation data were collected along a 400 ft straight scanline (trend 180, plunge 0) on
a rock slope along a former railroad alignment, on the Huckleberry Trail near
Blacksburg, Virginia. Fig. 2.11 shows the fracture normals in lower hemisphere
projection. The fractures are thought to be of approximately the same size, and
orientations to follow the Fisher distribution. Since all fractures belong to a well-defined
set, the Terzaghi bias associated with sampling along a straight scanline (Priest, 1993)
is approximately the same for all fractures, and is therefore neglected here.
Let the y-axis be directed horizontally to the north, the x-axis horizontally to the east, and
the z-axis vertically upward. The Fisher mean pole and dispersion constant can be
estimated as follows (Cheeney, 1983).
(1) The arithmetic means of the direction cosines are calculated from

cl = li N

cm = mi N

cn = ni N ,

(2.24)

where here the total number of fractures N = 19.


(2) The length of the mean vector R is calculated from:

R = cl2 + c m2 + c n2 ,

(2.25)

(3) Direction cosines of the estimated Fisher mean pole are computed from:

l = cl R

m = cm R

n = cn R ,

19

(2.26)

Table 2.4. Orientation data for a set of fractures on the Huckleberry Trail.
DipDirection
Number
Dip
1
80
44
2
84
76
3
80
44
4
80
270
5
88
260
6
82
48
7
88
227
8
86
244
9
86
238
10
89
75
11
86
256
12
76
43
13
75
58
14
74
50
15
70
42
16
90
248
17
90
66
18
86
252
19
84
240
totals
arithmetic means

Direction cosines of
normals
l
m
n
0.684
0.708
0.174
0.965
0.241
0.105
0.684
0.708
0.174
0.985
0.000
-0.174
0.984
0.174
-0.035
0.736
0.663
0.139
0.731
0.682
-0.035
0.897
0.437
-0.070
0.846
0.529
-0.070
0.966
0.259
0.017
0.968
0.241
-0.070
0.662
0.710
0.242
0.819
0.512
0.259
0.736
0.618
0.276
0.629
0.698
0.342
0.927
0.375
0.000
0.914
0.407
0.000
0.949
0.308
-0.070
0.861
0.497
-0.105
15.942
8.766
1.100
0.839
0.461
0.058

(4) If R has magnitude greater than about 0.65, the Fisher constant can be
approximated by:

= 1 (1 R ) ,

(2.27)

In this example, the direction cosines of the mean pole are estimated to be:

l = 0.875

m = 0.481

n = 0.060 ,

(2.28)

which gives a mean plane with dip-direction 61.2 and dip 86.5. The angle between the
Fisher mean pole and the scanline is calculated to be about 61.0. The mean resultant
length R is 0.96, from Eq. (2.25), and the Fisher constant is estimated by Eq. (2.27),
which gives = 24.6. A similar procedure for computing the Fisher parameters is given
by Goodman (1989), who takes R to be the resultant vector rather than the mean.

20

Mean

Scanline
lower hemisphere

equal area

Fig. 2.11. Fracture normals () and mean pole () in lower hemisphere projection.

We can calculate the (direction dependent) fracture frequency P10 along the scanline by
dividing the total number of fractures N by scanline length L:
P10 = N L = 19 400 = 0.048 ft 1 .

(2.29)

By using Table 2.1, Fig. 2.6, or the curves defined by the coefficients in Fig. 2.6, we can
interpolate the value of 1/C13. In this example, 61.0 and 24.6, so 1/C13 0.46
and C13 2.17. The volumetric intensity measure P32 (fracture area per unit rock mass
volume) can be determined for this fracture set by multiplying C13 and P10, giving

P32 = C13 P10 (2.17)(0.048 ft 1 ) = 0.10 ft 1

21

(2.30)

2.6 Discussion and Conclusions


Fracture intensity is a key input for computer models that deal with flow through a
fractured rock mass. Fracture intensity (P32) is inherently three-dimensional, but is
usually approximated via measurements on 1-d or 2-d sampling domains. Conversions
from 1-d or 2-d intensity, however, necessarily depend on the orientation (or orientation
distribution) of the sampling domain with respect to the fracture orientation distribution
of the rock mass. In this paper, conversion factors between 1-d and 2-d fracture intensity
measures (P10 and P21) and the 3-d intensity measure (P32) are discussed for the cases of
constant size or unbounded Fisher-distributed fractures. The needed conversion factors
for linear, planar and cylindrical sampling domains are computed semi-analytically, with
the aid of Monte Carlo simulation.

For linear sampling and planar sampling, the

conversion factors C13 and C23 are determined to be in the range of [1.0, ]. For
cylindrical surface sampling of constant size or unbounded Fisher-distributed fractures,
the conversion factor C23,C is determined to be in the range of [1.0, /2]. These
conversion factors are graphed and tabulated for a wide variety of cases.
In practice, straight scanlines run on a rock mass exposure, as well as straight smalldiameter boreholes, can be considered linear sampling. Rock exposures such as rock
slopes, or mine drift walls, are typical examples of planar sampling of fractures. Tunnel,
shaft or borehole walls give rise to cylindrical surface sampling of fractures. After
collecting fracture data on a sampling domain, e.g., a scanline, a planar rock slope, or a
borehole, engineers and geologists can estimate the volumetric intensity measure P32 by
using the conversion factors presented in this paper.

Acknowledgments
Support from the National Science Foundation, Grant Number CMS-0085093, is
gratefully acknowledged.

22

Appendix 2.A Probability density function (pdf) f ( ) of angle


Angles , , and between Fisher mean pole m and sampling line s, sampling line s and
fracture normal n, and fracture normal n and Fisher mean pole m, respectively, are
shown in Fig. 2.3 and Fig. 2.A-1.
To simplify determination of the pdf of , we define a coordinate system as shown in Fig.
2.A-1, in which m is perpendicular to the xy plane, and s is in the zy plane. Vectors s'
and n' are the projections of s and n on xy plane, respectively. The angle between s' and
n' is defined as . For the spherical triangle formed by m, n, and s, the following

relationship holds (Ayres, 1954):

cos = cos sin sin + cos cos ,

(2.A-1)

so that,
cos cos cos

sin sin

= cos 1

(2.A-2)
.

The probability density function of angle depends on and ( being kept constant in
the derivation). For the Fisher distribution, the joint pdf of angle and is given by (N.
Fisher et al., 1987)
f , ( , ) =

eCos Sin
(0 ) .
2 (e e )

23

(2.A-3)

z
m
n

Fig. 2.A-1. Coordinate system for spherical triangle formed by m, s and n.

Angle is uniformly distributed in the range [0, 2] and is independent of angle .


Therefore the pdf of is
f ( ) = 1/ 2 ,

(2.A-4)

from which,

f ( ) =

eCos Sin
e e

(0 ) .

(2.A-5)

The pdf of angle can be derived through its cumulative distribution function (cdf).
Given , the cdf of is

24

F| ( | ) = Prob ( | ) = R f ( ) d ,

(2.A-6)

where R is the range of when . Given , increases from the minimum of |-|
to the maximum of + when increases from 0 to . Fig. 2.A-2 shows the relationship
among angles , , , and also the range of . Note that R is symmetric about the y-axis.
Below is the determination of R with different range of , , and .

[0, max ]

R =
0

[0, max ]

R = 2

[ - , + ], if ( + ) /2
[ - , /2], if ( + ) /2 ,
else

[ , + ], if ( + ) < /2
[ - , /2], if ( + ) /2

[0, - ]
[ + ), /2]

(2.A-7)

where max is the upper limit of , which is determined by Eq.(2.A-2), from which

cos cos cos


.
sin sin

max = cos 1

(2.A-8)

Then
cos cos cos

cos1
sin sin

F| ( | ) = 2

1
d , | | + ,
2

(2.A-9)

from which,
F| ( | ) =

cos cos cos | | +


,
cos 1
, and

sin sin

f | ( | ) = F| ( | ) =

sin

sin sin (cos cos cos ) 2


2

| | + .

25

(2.A-10)

(2.A-11)

Then, the pdf of is


f ( ) =

f | ( | ) f ( )d =

eCos Sin

sin
sin 2 sin 2 (cos cos cos ) 2

e e

| | + ,

(2.A-12)

where Rd is given by
R = [ - , + ] , if , or
R = [0, 2 ] , if

>.

(2.A-13)

Fig. 2.A-2 The figure shows the range, R , of , as a function of , and . Angle
(between m and s) is a constant. Angles and are the semi-apical angles of small
circles about z and y, respectively. R delimits the intersection of the above-mentioned
small circles, projected into the xy plane.

26

Appendix 2.B Numerical approach for obtaining F() and F()

Rewrite Eqs. (2.3) and (2.4) as


1

C13 = cos dF ( ) , and


0

(2.B-1)

C23 = sin dF ( ) ,
0

(2.B-2)

where F() and F() are the cdfs of and , respectively. Numerical evaluation of the
finite integrals in the equations gives the conversion factors C13 and C23, using the
procedure described below for F(). F() can be obtained through a similar procedure.
1. Set the mean pole and the Fisher constant for the Fisher distribution.
2. Generate a set of Fisher-distributed fracture normals by using the cdf of the Fisher
distribution, given by (Dershowitz, 1985)

F ( ) =

eCos
e 1

(0 ) ,

(2.B-3)

Fig. 2.4 shows a simulated population of 3000 fracture normals with the Fisher
mean pole corresponding to a plane with dip 80 and dip-direction 45, and = 60.
for a detailed description of the simulation procedure, see Priest (1993).
3. For a given sampling orientation, draw small circles (Fig. 2.4) with values of at
fixed increments.
4. The cdf F() of , is calculated empirically by the number of fracture normals
falling inside small circles divided by the total number of fracture normals. For
instance, in Fig. 2.4, 104 out of 3000 fracture normals are inside the small circle
with = 20. Therefore the cdf F() of , evaluated at = 20, is
F ( = 20) =

104
= 0.035 .
3000

(2.B-4)

27

References
Ayres, F. Jr. (1954) Schaums Outline Series of Theory and Problems of Plane &
Spherical Trigonometry. McGraw-Hill
Cheeney, R. F. (1983) Statistical methods in geology for field and lab decisions, Allen
& Unwin Ltd. London. UK
Dershowitz, W.S. (1985) Rock Joint System Ph.D. Dissertation, MIT, Cambridge, Mass.
Dershowitz, W.S. and H.H. Einstein (1988) Characterizing rock joint geometry with
joint system models Rock Mechanics and Rock Engineering 21: 2151
Dershowitz, W. S. and Herda, H. H. (1992) Interpretation of fracture spacing and
intensity Proceedings of the 33rd U.S. Symposium on Rock Mechanics, eds. Tillerson, J.
R., and Wawersik, W. R., Rotterdam, Balkema. 757-766.
Dershowitz, W., J. Hermanson & S. Follin, M. Mauldon (2000) Fracture intensity
measures in 1-D, 2-D, and 3-D at Aspo, Sweden, Proceedings of Pacific Rocks 2000,
eds. Girard, Liebman, Breeds & Doe
Einstein, H. H. and Baecher, G. B. (1983) Probabilistic and statistical methods in
engineering geology Rock Mechanics and Rock Engineering 16: 39-72.
Fisher, N. I., T., Lewis, B.J.J. Embleton (1987) Statistical analysis of spherical data.
Cambridge University Press, Cambirdge UK
Fisher, R. A. (1953) Dispersion on a sphere Proc. Roy. Soc. London, Ser. A, 217: 295305
Goodman, R. E. (1989) Introduction to Rock Mechanics. John Wiley & Sons, New
York.
Martel, S.J. (1999) Analysis of fracture orientation data from boreholes. Environmental
and Engineering Geoscience. 5: 213-233.
Mauldon, M. (1994) Intersection probabilities of impersistent joints, International
Journal of Rock Mechanics and Mining Science & Geomechanics Abstracts, 31(2): 107115.
Mauldon, M., J. G. Mauldon. (1997) Fracture sampling on a cylinder: from scanlines to
boreholes and tunnels. Rock Mechanics and Rock Engineering. 30: 129-144.
Mauldon, M., M.B. Rohrbaugh, W.M. Dunne, W. Lawdermilk (1999) Fracture intensity
estimates using circular scanlines. In Proceedings of the 37th US Rock Mechanics
Symposium, eds. R.L. Krantz, G.A. Scott, P.H. Smeallie, Balkema, Rotterdam. 777-784.

28

Mauldon M., W. M. Dunne and M. B. Rohrbaugh, Jr. (2001) Circular scanlines and
circular windows: new tools for characterizing the geometry of fracture traces. Journal
of Structural Geology, 23(3): 247-258
Mauldon M. and X. Wang (2003) Measuring Fracture Intensity in Tunnels Using
Cycloidal Scanlines Proceedings of the 12th Panamerican Conference on Soil
Mechanics and Geotechnical Engineering and the 39th U.S. Rock Mechanics Symposium.
Owens, J.K., Miller, S.M., and DeHoff, R.T. (1994) Stereological Sampling and
Analysis for Characterizing Discontinuous Rock Masses. Proceedings of 13th
Conference on Ground Control in Mining. 269-276.
Priest, S.D. (1993) Discontinuity Analysis for Rock Engineering. Chapman and Hall,
London.
Russ, J. C., DeHoff, R. T. (2000) Practical Stereology Kluwer Academic/Plenum
Publishers, New York
Terzaghi, R.D. (1965) Sources of errors in joint surveys. Geotechnique. 15: 287-304.
Yow, J.L. (1987) Blind zones in the acquisition of discontinuity orientation data.
International Journal of Rock Mechanics and Mining Sciences and Geomechanics
Abstracts. Technical Note. 24: 5, 317-318.

29

This page intentionally left blank.

30

Chapter 3
3 Estimating fracture intensity from traces on cylindrical
exposures
Abstract
Fracture intensity is a fundamental parameter when characterizing fractures. In the field,
a great amount of fracture data is collected along boreholes, circular tunnel or shaft walls.
The data reveal some characteristics of fractures in rock masses; however, it has not been
sufficiently interpreted. In this paper, we discuss estimating of fracture intensity, more
specifically, fracture volumetric intensity P32, from fracture trace data in cylindrical
(borehole, tunnel or shaft) samplings. We built up the relationships between the 2-d
fracture intensity measure and the 3-d fracture intensity measure theoretically.
Stereological analyses show that the conversion factor between the two intensity
measures is not dependent on fracture size, shape or circular cylinder radius, but is related
to the orientation of the cylinder and the orientation distribution of fracture area. It is also
found that the fracture volumetric intensity measure P32 is always 1.0 to 1.57 times of
fracture trace length per unit borehole surface area (P21,C). The technique of using
cycloidal scanlines to estimate the fracture volumetric intensity is also discussed. A
computer program is developed to generate synthetic fractures sampled by a circular
cylinder and the derived conversion factor between the two intensity measures is tested
by Monte Carlo simulations.
Key words: cylindrical sampling, fracture networks, stereology, rock mass, intensity

measures, conversion factors

31

3.1 Introduction
Natural rock masses are commonly dissected by discontinuities such as fractures, faults
and bedding planes, which influence or even control the behavior of rock masses
(Goodman, 1989; Priest, 1993). Therefore, characterization of the fracture system in a
rock mass, including properties such as fracture orientation, shape, size, aperture, and
intensity (ISRM, 1978), is necessary for many engineering applications. Examples of
such applications include hydrocarbon extraction, control of contaminants in landfills,
tunneling, and rock slope engineering.
Fracture intensity, which represents the amount of fractures in the rock mass, is one of
the fundamental parameters for characterizing fracture systems. Fracture intensity can be
interpreted in several ways, corresponding to a set of fracture abundance measures,
depending on the dimension of the sampling domain. (Dershowitz, 1984, 1992; Mauldon
1994). The most commonly used measure is the frequency of fractures, defined as
number of fractures per unit length. Frequency, which is also referred to as the onedimensional (1-d, linear) intensity, P10, is often measured along a scanline (Fig. 3.1(a)) of
fixed orientation on a planar exposure, or along the length of a borehole. The sampling
bias (R. Terzaghi 1965) induced by scanline or borehole measurements of fracture
frequency, or P10, remains a problem with scanline measurements. The major difficulty
with implementing frequency data as a fracture intensity measure has to do with the socalled blind zone (Terzaghi, 1965; Yow, 1987), which refers to fracture orientations
that are not seen or under-sampled by a borehole or scanline. The geometric
(Terzaghi) correction factor for fractures in the blind zone can lead to gross distortion
of the data (Yow, 1987). A review of scanline sampling is presented by Priest (1993,
2004).
On cylindrical exposures such as borehole walls, circular tunnel or shaft walls, the
fracture system is revealed in a two-dimensional (2-d) form. Besides features of fractures
such as orientation, aperture, or infilling that can be measured directly on cylindrical

32

exposures, the intensity, pattern, and termination relationships of fracture traces on the
cylindrical exposure surfaces provide much more information about fracture networks
than a one-dimensional exposure (scanline) does.

Borehole

Rock mass

Fractures

Fracture traces
total length = l

Scanline
Fracture trace
on the slope

Unrolled (developed) trace map


(total area A)

(a)

(b)

Fig. 3.1. Borehole or shaft sampling of fractures in a rock mass. (a) Vertical shaft
intersects several fractures, which yield traces on the cylinder surface and on the face of
the rock mass; horizontal scanline on the rock face intersects three fracture traces. (b)
Unrolled trace map developed from the borehole or shaft wall.

33

To explore the relationships between fracture traces on a cylindrical surface and the 3-d
fracture system, we introduce the following notation. Let P21,C denote the twodimensional (2-d, areal) fracture intensity on the circular sampling cylinder surface,
defined as trace length per unit sampling surface area. The subscript C denotes the
cylindrical sampling domain. P21,C is determined as the sum of trace length on tunnel or
borehole walls divided by the total surface area of tunnel or borehole walls. In Fig. 3.1(b),
for instance, assume the total trace length on the unrolled trace map is l and the total area
of the unrolled trace map is A. Then the areal fracture intensity is simply P21,C = l / A.
For a fractured rock mass, this measure is a function of tunnel or borehole size and
orientation, as well as the fracture orientation distribution (weighted by fracture size).
Therefore it is also a directionally biased measure, as is as the linear intensity measure
P10.
Let P32 denote the three-dimensional (3-d, or volumetric) fracture intensity, defined as
fracture area per unit volume of rock mass. P32 is independent of the sampling process
and is an unbiased measure of fracture intensity (Dershowitz, 1992; Mauldon 1994).
Interpreted as an expected value, P32 is also scale independent. P32 is a crucial parameter
for numerical analyses in models such as the discrete fracture flow and transport model
(Dershowitz et al., 1998). However, P32 is impossible to measure directly in an opaque
rock mass.
This paper proposes approaches to utilize fracture trace data collected on the cylindrical
exposures of rock mass, such as borehole walls, tunnel or shaft walls, to estimate
volumetric fracture intensity of the rock mass. This determination is based on the derived
relationship (conversion factor) between the fracture areal intensity on a cylindrical
surface (P21,C) and the fracture volumetric intensity measure (P32).
Following stereological principles (Russ and DeHoff, 2000) we first discuss the general
form of the conversion factor between the areal intensity P21,C on circular cylinder
surface and fracture volumetric intensity measures P32. Theoretical solutions for the
conversion factor between the two measures are derived in the case of cylindrical

34

sampling of constant orientated fractures, and also sampling of fractures with a uniform
distribution. The conversion factor can be calculated analytically if the fracture
orientation distribution with respect to its area is known. Secondly, another approach to
estimate fracture volumetric intensity, based on the cycloidal scanline technique, is also
discussed. By counting the intersections between cycloidal scanlines and fracture traces
on the circular cylinder surface, the fracture volumetric intensity can be estimated
without knowing the orientation of fractures. Finally Monte Carlo simulations are carried
out to verify the derived correction factors.

3.2 Basic assumptions


In this paper, we study a fractured rock mass sampled by a borehole or tunnel/shaft by
using stereology. For convenience, we make the following assumptions with respect to
the geometry of the sampling domain, e.g., the surface of the tunnel/shaft or borehole;
and of fractures in the rock mass.
a) The surface of the sampling domain is a right circular cylinder, long in relation to
its diameter. Borehole, tunnel or shaft ends are not included in the sampling
domain.
b) Fractures are planar features with negligible thickness. No assumptions are made
regarding the spatial distribution of fractures, or fracture shape. In particular, it is
not necessary that fracture centers follow a Poisson process, or that fractures have
the shape of circular or elliptical discs.
c) No prior assumptions are made about fracture size, or orientation distribution;
however, for the first method discussed below, the fracture orientation distribution
in terms of area must be known.
d) The sampling domain is independent of the rock mass fracture network to be
characterized. What this means in practical terms is the borehole/shaft or tunnel is
emplaced without consideration of fracture locations.

35

The above assumptions are fairly standard in engineering analysis of fractured rock
masses (Priest & Hudson, 1976; Warburton, 1980; Cheeney, 1983; Dershowitz, 1984;
Priest 1993; Mauldon & Mauldon, 1997). Furthermore, these assumptions are applicable
in most rock engineering situations either because of the lack of knowledge of
underground fracture networks before boreholes are excavated or, because the location of
a tunnel or shaft is predetermined, based on external factors.
In accordance with principles of stereology, the 1-d, 2-d and 3-d fracture intensities
discussed in this paper refer to expected values, if not specified otherwise. The acronym
IUR - isotropic, uniform, random denotes, in general, desirable properties of
stereological samples (Russ and DeHoff, 2000; Mauton 2002). In the present situation,
isotropy is ensured in the plane perpendicular to the borehole/shaft/tunnel axis by the
circular symmetry of the cylinder; the directional relationship between the cylinder axis
and the fracture system, however, is not in general, one of isotropy, except in the special
case of a uniform fracture orientation distribution. One of the primary tasks of this paper
is to account for the directional relationship between cylinder and fractures, with respect
to the determination of fracture intensity.

3.3 General form of the relationship between areal intensity P21,C and
volumetric intensity P32 for right circular cylinders
In this section, we relate the volumetric fracture intensity measure P32 (fracture area per
unit rock mass volume) to the areal fracture intensity measure P21,C as measured on a
cylinder (fracture trace length per unit sampling surface area). The relationship is
presented here in a general form.
We define a geometric correction factor, C23,C by

P32 = C23,C P21,C ,

(3.1)

36

where the subscript 23 denotes conversion from a two-dimensional to a threedimensional measure, and the subscript C denotes a cylindrical surface sampling domain.
The conversion factor C23,C is a function of cylinder orientation and the fracture
orientation distribution; it does not depend on cylinder radius, as demonstrated in next
section.
The geometric meaning of this conversion factor can be illustrated using a simple model
of a cylindrical surface sample (Fig. 3.2), in which five fractures are sampled by a
vertical shaft of radius r and height H. Let l denote the total summed trace length on the
shaft surface. Given a population of fractures, l is a function of cylinder orientation,
radius r, and height H; and the area-weighted fracture orientation distribution.
Consider a thin cylindrical shell (Fig. 3.2) with radius r. The shell thickness dr is taken to
be infinitesimal, so that the area of fractures contained inside the shell, Afractures, can be
approximated as

A fractures = C 23,C l dr .

(3.2)

where C23,C is the geometric correction factor. If the fractures are perpendicular to the
circular cylinder surface at the intersections, this correction factor is 1.0 (and the
expression is exact). Otherwise, it is greater than 1.0.
The volumetric fracture intensity measure P32, fracture area per unit volume, for the shell
can be expressed as

P32 =

A fractures

(3.3)

2r H dr

Substituting Eq. (3.2) into (3.3), we obtain

P32 = C 23,C

l
= C 23,C P21,C .
2r H

(3.4)

37

Shaft with radius r,


shell thickness dr

Fracture
traces, total

Fig. 3.2. Fracture traces on a cylindrical shaft. Intersections between fractures and the
shaft are traces (curved line segments) on the shaft surface

This is the general form of the relationship between the 2-d intensity measure for trace
length and the 3-d intensity measure for fracture area in a rock mass. In the following we
derive the correction factor C23,C for the general case of fractures that are distributed
according to a known probability density function for fracture orientation with respect to
fracture area. Then we discuss two special cases: fractures of constant orientation and
fracture orientations uniformly distributed in the rock mass.

3.4 General case of cylindrical sampling


As discussed in section 2, the sampling cylinders radius, orientation and location is
assumed independent of the rock mass and fracture geometry. Let f(,) denote the
probability density function (pdf) of fracture orientation weighted by area, where is the
acute angle between the sampling cylinder axis (Z-axis in Fig. 3.3(a)) and the normal n to
a fracture; and is the angle between the Y-axis and the projection of the fracture normal
n onto the XY plane (Fig. 3.3(a)).

38

Ti

Z
Y

ni

npi
LH
Z

dr

S
npi
1

ni

Ti

X
(a)

rd
(b)

Fig. 3.3. A thin slice of the shell sampling in fractures. The total trace length on its
surface is dl. (a) A cylindrical shell (axis Z) intersects a set of fractures with orientation
distributed as f(,). For a fracture with unit normal n, is the angle between Y-axis and
the projection of n on the XY plane; is the angle between n and Z. (b) A portion (unit
height) of a slice from the shell is taken out for study. The ith fracture intersected with the
portion has a unit normal ni and the trace of this fracture on the circular cylinder surface
is represented by a unit vector Ti. The figure above shows the vectors in a lower
hemisphere projection.

39

Consider a thin, narrow slice of unit length, width = rd and thickness = dr, taken out
from the shell (Fig. 3.3(b)). Let dli denote the length of the trace of fracture i on the
outside surface of the slice; let Ti be the unit vector representing the direction of the
corresponding fracture trace on the slice surface; and let npi denote the unit normal to a
plane passing through the trace, and perpendicular to the slice surface (npi is the
normalized vector of the cross product STi). Finally, let i be the angle between npi
and the normal ni of fracture i. Then the infinitesimal area dAi of fracture i inside the
slice is

dAi =

dl i dr
cos i

(3.5)

Notice that i varies for different fractures intersecting the same slice, and for the same
fracture intersecting by different slices from the cylindrical shell.
The expected area dAi of fracture i inside the unit length slice can also be expressed in
terms of P32 and the probability density function f(,),

dAi = P32 f ( i , i ) 1 rd dr ,

(3.6)

where i and i are the angles representing the orientation of the normal to fracture i in
the coordinate system shown in Fig. 3.3(a). Equating Eqs. (3.5) and (3.6), the expected
trace length dli of fracture i on the unit slice surface is found to be,

dli = P32 f ( i , i )rd cos i .

(3.7)

The expected total length dl of fracture trace segments on the outer cylindrical surface
contained within the slice of height H is the integration of trace lengths of all fractures
intersecting the slice, with respect to fracture orientation:

dl = P32 Hrd

f ( , ) cos dd .

40

(3.8)

where is the angle between the normal n to a fracture and the normal np to the plane
passing through the trace of the fracture and perpendicular to the slice surface. Note that

is a function of , and (Appendix 3.A) and that, in this context, and d are
constant.

Denote the integral in Eq. (3.8) as

Io =

f ( , ) cos dd ,

(3.9)

where Io is a function of f(, ) and the orientation of the cylinder axis. For this general
case, cos is determined in Appendix 3.A as

cos = cos 2 + sin 2 sin 2 ( ) ,

(3.10)

so that

dl = P32 Hrd I o .

(3.11)

The expected total trace length l on the cylindrical sampling surface is obtained by
integrating dl over all values of ,
2

l = dl =P32 Hr I o d .

(3.12)

The fracture areal intensity on the cylinder surface can be expressed as

P21,C

P
P32 Hr 2
l
=
=
I o d = 32

2rH 2rH 0
2

I o d
0

Then the conversion factor C23,C relating areal intensity on a cylinder to volumetric
intensity (c.f. Eq. (3.1)) is given by

41

(3.13)

C 23,C

= 2 I o d .
0

(3.14)

For this general case, Eq. (3.14) shows that the conversion factor C23,C is dependent
neither on the size of the circular cylinder surface, nor on fracture shape. It is a function
of the orientation of cylinder axis and the area-weighted fracture orientation pdf f(,).
The range of the conversion factor will be discussed in the next section.

3.5 Special case: Sampling fractures of constant orientation


When a cylindrical surface samples a set of fractures with constant orientation, we can
always choose the Y-direction so that cylinder axis Z, fracture normal n, and the Y-axis
are coplanar (Fig. 3.4). In this coordinate system, angle between Y and the projection
of n on the XY plane is 0. Let 0 denote the acute angle between n and Z (Fig. 3.4). It is
a constant in this context.
For fractures with constant orientation, cos is determined to be (Appendix 3.A)

cos = 1 sin 2 0 cos 2 .

(3.15)

Note that cos is not a function of either or . Then the integral in Eq.(3.10) is

Io =

f ( , ) cos dd = cos f ( , )dd

= cos = 1 sin 2 0 cos 2 .

(3.16)

42

Fig. 3.4. A cylindrical shell (axis Z, height = H) intersects a set of fractures with constant
orientation (normal n). 0 is the angle between the fracture normal and the cylinder axis.

From Eq. (3.14),

C 23,C

= 2 I o d
0

= 2 1 sin 2 0 cos 2 d
0

(3.17)

Evaluating the above integral using Mathematica (Wolfram Research, Inc, 2004), we
obtain

43

2
1
4

1 sin 2 0 cos 2 d = EllipticE (sin 2 0 ) ,

(3.18)

2
where EllipticE (sin 0 ) is a complete elliptic integral of the second kind.

Combining Eqs. (3.18) and (3.14), the conversion factor C23,C relating areal intensity on
a cylinder to volumetric intensity is

C 23,C =

[EllipticE (sin
2

0 )

(3.19)

The conversion factor C23,C takes on values ranging from 1 to /2 for 0 ranging from 0
to 90, respectively (Fig. 3.5). Note in particular that fracture volumetric intensity P32 is
equal to fracture areal intensity P21,C on the cylinder surface if fractures are perpendicular
to the sampling cylinder(C23,C = 1); and P32 is 1.57 times fracture areal intensity P21,C on
the cylinder surface if fractures are parallel to the cylinder axis (C23,C = /2).
It should be noted that the case above of constant fracture orientation is the least isotropic
of all orientation distributions and that the above orientations of the fractures relevant to
the cylinder i.e. parallel and perpendicular to the cylinder axis, also represent extreme
cases. Therefore, for a general case of fracture orientation distribution, the conversion
factor C23,C is in the range [1, /2] as well. This result is very important to rock
engineering practitioners, especially when there is not much information about the
fracture orientation distribution with respect to area. Since the range of the conversion
factor C23,C is fairly small (1.0 to 1.57), it will be convenient and will not cause major
errors to approximate the fracture volumetric intensity P32 by using Eq. (3.19) or Fig. 3.5,
where 0 is estimated as the average acute angle between fractures and the sampling
cylinder axis.

Finally, for the special case of constant fracture orientation, Eq. (3.19) shows clearly that
the conversion factor C23,C is only a function of the angle between the cylinder axis and

44

the fracture normal. It is independent of the radius of the sampling cylinder, as well as of
fracture shape and size.

Correction factor C23,C

1.6

/2

1.5
1.4
1.3
1.2
1.1
1.

0 (degree)

Fig. 3.5. For cylindrical sampling in fractures with constant orientation, the correction
factor C23,C between areal intensity P21,C and volumetric intensity P32 is a function of
angle 0 between the cylinder axis and fracture normal. The elliptic integral required to
obtain the curve was evaluated using Mathematica.

3.6 Special case: fractures with uniform orientation distribution


We apply the general result of the conversion factor C23,C to the isotropic case, in which
fracture orientations are uniformly distributed with respect to area.
In this case,

45

f ( , ) =

sin
, and from Appendix 3.A
2

(3.20)

cos = cos 2 + sin 2 sin 2 ( ) .

(3.21)

Therefore, the integral in Eq.(3.9) is

Io =

f ( , ) cos dd

= / 2 =2

=0

sin

=0 2

cos 2 + sin 2 sin 2 ( )d d

(3.22)

And from Eq. (3.14),

C 23,C

= 2 I o d
0

=2 = 2 =2 sin
= 2
=0 =0 =0 2

cos + sin sin ( )d d d

(3.23)

The definite integral in Eq. (3.23) was evaluated in Mathematica (Wolfram Research, Inc,
2004), which gives
= 2 = 2 = 2

=0

=0

=0

sin
2
2
2
2
cos + sin sin ( )d d d = 4.9348
2
2

So,

46

(3.24)

C23,C

= 2 = 2 = 2 sin
cos 2 + sin 2 sin 2 ( )d d d
= 2

= 0 = 0 = 0 2

2 2 / 2

(3.25)

This result can be compared with the results for plane sampling of isotropically
distributed fractures (Dershowitz, 1984), namely


P21 ( isotropic ) = P32 ,
4

(3.26)

where P21(isotropic) is the trace length per unit area of sampling plane.

3.7 Cycloidal Scanline Technique

In this section we discuss a sampling technique that uses a special curved scanline based
on a cycloid, which automatically takes care of the directional bias described by Terzaghi
(1965). By correctly deploying cycloidal scanlines on the cylindrical surface, we can
make an unbiased estimate of fracture volumetric intensity with no need to know the
orientation of fractures (either ahead of time or at the time of sampling).

3.7.1 Unbiased sampling criterion

As mentioned earlier, a basic strategy in stereology involves the use of IUR (IsotropicUniform-Random) sampling (Russ and Dehoff, 2000). A perfectly isotropic 2-d
sampling surface is a sphere, on which the surface area is distributed uniformly with
respect to direction. Similarly, on a plane, a circular scanline is a perfectly isotropic 1-D

47

sampling domain, with length segments uniformly distributed in every direction


(Mauldon et al., 2001). IUR scanlines produce unbiased samples automatically, and thus
obviate the need for any bias correction.

Fig. 3.6(a) shows uniformly distributed unit vectors (directed line segments) on a
hemisphere. Let denote the angle (colatitude) between a unit vector and axis Z. If the
unit vectors have a uniform orientation distribution, the probability p( ) of choosing a
line segment of unit length and along a vector with colatitude must be proportional to
l() = 2 sin ( ) (Fig. 3.6(a)). Choosing a normalizing constant such that p( ) d
has the value unity when integrated over all values of (0 to ) for vectors uniformly
distributed in all orientations), we have

p ( ) = 12 sin

(3.27)

As an alternative to selecting scanline orientations from a probability distribution, it is


possible to specify a curved scanline (Fig. 3.6(b)) that utilizes all values of (0 )
with differential scanline arc lengths dL() proportional in all cases to sin , or

dL( ) sin .

(3.28)

One form of scanline that has this property is the cycloid, which we discuss in the next
section.

48

l = 2sin

sin

Cycloidal
curve

(b)

(a)

Fig. 3.6. Illustration of linear (vector) IUR sampling in 3-d space. (a) Uniformly oriented
unit vectors on a hemisphere. (b) Length-scaled vectors on the cylinder surface.

From stereological principles, (Russ and DeHoff, 2000; Dershowitz, 1984; Mauldon and
Wang 2003), linear fracture intensity P10(unbiased) measured on the cylindrical surface by
such unbiased sampling probes (scanlines) has the following relationship with the
volumetric fracture intensity P32.

P10(unbiased ) = 12 P32 .

(3.29)

49

3.7.2 Cycloidal scanlines


Mathematically, a cycloid is the locus of a point on the rim of a circle rolling along a
straight line, as shown in Fig. 3.7. For a generating circle of radius r0, the coordinates of a
point on the cycloid are given by:

x' = r0 ( sin )
z = r (1 cos )
0

(3.30)

where is the angle of rotation of the circle.


One of the properties of a cycloid is that for any point on the cycloid with angle
between Z and the tangent to the cycloid, the incremental arc length dl is proportional to
sin ( dl = 4 r0 sin d ). Therefore the cycloid as a sampling probe satisfies Eq. (3.28)

and can be used as a directionally unbiased (IUR) sampling probe for measuring fracture
intensity on the walls of a borehole or tunnel/shaft. In other words, cycloids can be used
as scanlines on (right-circular) cylindrical surfaces without the need to correct for
directional sampling bias, and without the need to know fracture orientation. In practice,
cycloidal scanlines can be modified in various ways for more efficient deployment (Russ
and DeHoff, 2000; Mauldon and Wang 2003), as long as the correct relationship between
arc length and orientation is maintained. The fracture volumetric intensity P32 can then
be estimated by Eq. (3.29) - which in terms of expected values is an exact expression.

3.8 Monte Carlo Simulations


A computer program was developed in Visual C++ and used to generate a population of
synthetic fractures, of rectangular shape, intersecting a cylindrical surface such as the
wall of a borehole, tunnel or shaft (Fig. 3.8). The fracture traces are computed and shown
on the unwrapped cylindrical surface (right-hand window in Fig. 3.8). For each
simulation, the total number of generated fractures, the area of each fracture, as well as
the size of the generation region were recorded, in order to calculate the volumetric

50

fracture intensity P32. Total fracture trace length on the circular cylinder surface was also
recorded to calculate P21,C, the 2-d intensity on the sampling circular cylinder, by
dividing by total cylinder surface area.

dl = 4 r0 sin d

Generating
circle

dl

d
Cycloid

r0

X
Fig. 3.7. The cycloid (heavy curve) is the path of a point on the circle of radius r0 as the
circle rolls from left to right along the x-axis.

Five cases are chosen, to represent different fracture sizes and shapes intersecting a
cylinder of constant size (Fig. 3.9). In each case, the angle 0 between fracture normal
and cylinder axis, is set to be 0, 30, 60, and 90, respectively. Ten simulations were run
for each fracture orientation. The parameters for each case and the results of the
simulations are listed in Table 3.1. For comparison, the conversion factor C23,C calculated
by Eq. (3.19) for each 0 is also listed in Table 3.1. In all the simulations, fracture
volumetric intensity P32 was set constant, P32 = 1.0.

51

The simulation results are plotted in Fig. 3.10, where they are compared with the curve of
C23,C computed by Eq. (3.19). The simulations show that for fractures with constant
orientation, the areal fracture intensity measure on a cylindrical surface P21,C (trace length
per unit cylinder surface area), is related to the volumetric fracture intensity P32 (fracture
area per unit volume), only by angle 0 between the fracture normal and the cylinder axis.
The conversion factor is independent of the cylinder radius, as well as of the size or shape
of fractures. The derived conversion factor, expressed by Eq. (3.19), is also verified from
the simulations.

Fractures

Trace map
Borehole

Fig. 3.8. The computer program is used to generate rectangular fractures intersecting with
a borehole. Fracture orientation can be set to constant or vary according to given
parameters.

52

Case #

Cylinder
radius

10

10

10

10

10

Fracture
length l

100

10

20

100

20

Fracture
width w

100

10

20

20

Aspect
ratio
l/w

Fig. 3.9. Illustration (to the scale) of the five cases studied. Shaded rectangles are
simulated fractures, and circles are sampling cylinders (radius is constant 10 for all
simulations).
Table 3.1. Simulation parameters and results.
Average C23,C for each case

Fracture

Fracture

Aspect

length l

width w

ratio l / w

0 = 0

0 = 30

0 = 60

0 = 90

Case 1

100

100

1.0

1.00

1.04

1.31

1.53

Case 2

10

10

1.0

1.01

1.07

1.29

1.56

Case 3

20

20

1.0

1.01

1.06

1.28

1.62

Case 4

100

20

5.0

1.04

1.09

1.32

1.56

Case 5

20

5.0

0.99

1.06

1.30

1.57

Average C23,C for each angle 0

1.01

1.07

1.30

1.57

C23,C calculated by Eq. (3.19)

1.00

1.07

1.30

1.57

53

Conversion factor C23,C

1.6
1.5

C23,C calculated by
Eq.(3.19)

Case
1
Case
2
Case
3
Case
4
Case
5

1.4
1.3
1.2

C23,C from
simulations

1.1
1.0

0 (degree)
Fig. 3.10. Simulation results of the conversion factor 1/ C23,C, compared with the
calculated curve by Eq.(3.19).

3.9 Discussion & Conclusions


In this paper, we used stereological principles to study the conversion factor between the
2-d fracture intensity measure on a cylinder surface and the 3-d fracture volumetric
intensity measure. The derived conversion factor between the two intensity measures is
not dependent on fracture size, shape or circular cylinder radius, but is related to the
orientation of the cylinder and the distribution of fracture area with respect to its
orientation.

54

By studying a special case of cylindrical sampling of fractures with constant orientation,


it is found that the fracture volumetric intensity measure P32 is always 1.0 to 1.57 times of
fracture trace length per unit borehole surface area (P21,C). The two values are also the
minimum and maximum limit of the conversion factor between the two measures in a
general case of cylindrical sampling of fractures, which provides a very practical means
in the field to estimate fracture volumetric intensity.
Based on Isotropic-Uniform-Random principle of stereology, cycloidal scanlines, as
directional unbiased probes to estimate the fracture volumetric intensity, is also
introduced in this paper.
A computer program simulating synthetic fractures sampled by a circular cylinder was
developed and the derived conversion factor between the two intensity measures is
confirmed by Monte Carlo simulations.

Acknowledgements
Partial support from the National Science Foundation, Grant Number CMS-0085093, is
gratefully acknowledged.

55

Appendix 3.A Determine |cos |


Fig. 3.A-1 shows the unit vectors S, T, n, nr in a Cartesian coordinate system.
In this coordinate system, Z represents the borehole or sampling circular cylinder axis. n
is the normal to a fracture and it makes an acute angle with Z-axis. is the angle
between Y-axis and the projection of n on XY. S is a unit normal to a small slice of the
circular cylinder surface, which is parallel to the circular cylinder axis, and it makes an
angle with Y-axis. T is a unit vector parallel to the intersection of two planes whose
normals are S and n respectively (Fig. 3.A-1). nr is the unit normal to a plane containing
both T and S.
Let vector T be the cross product of n and S.

T = n S
x
y
z
= sin sin sin cos cos
sin
cos
0
= [ cos cosx + cos sin y sin sin ( )z ]

(3.A-1)

Then unit vector T will be the normalized T.

T = [ cos cosx + cos sin y sin sin ( )z ] T

(3.A-2)

where

T = cos 2 cos 2 + cos 2 sin 2 + sin 2 sin 2 ( )


= cos 2 + sin 2 sin 2 ( )

56

(3.A-3)

Z
nr
n

(sin sin,
sin cos,
cos)

X S

(sin, cos, 0)

Fig. 3.A-1. Unit vectors S, T, n, and nr in Cartesian coordinate system, where Z is


parallel to the borehole axis. The coordinates of unit vectors S and n are given based on
the geometry.

Let vector nr be the cross product of S and T, which gives a unit vector.

nr = S T
x
y
z
=
sin
cos
0
cos cos cos sin sin sin ( )
T
T
T
sin sin ( ) cos
sin sin ( )sin
cos
=
x +
y +
z
T
T
T
Then nr is the same as nr.

57

(3.A-4)

|cos | is given by the dot product of nr and n.

cos = nr n
sin sin ( ) cos
=

=
=

sin sin ( )sin


T

cos
T

cos 2 + sin 2 sin ( )sin


T

sin

cos

cos 2 + sin 2 sin ( )sin


cos 2 + sin 2 sin 2 ( )

(3.A-5)

In the special case that fractures are of constant orientation, we can always rotate the
coordinate system around Z-axis and make n inside ZY plane. Then angle , the angle
between Y-axis and the projection of n on XY, turns to be zero. Let 0 denote the acute
angle between n and Z, which is a constant in this special case.
Therefore, |cos | given by Eq. (3.A-5) will be simplified as follows.

cos =

cos 2 0 + sin 2 0 sin 2


cos 2 0 + sin 2 0 sin 2

= cos 2 0 + sin 2 0 sin 2


= 1 sin 2 0 cos 2

(3.A-6)

58

References

Cheeney, R. F., (1983) Statistical methods in geology for field and lab decisions, Allen
& Unwin Ltd. London. UK
Dershowitz, W.S. (1984) Rock joint systems. Ph.D. Thesis, Massachusetts Institute of
Technology, Cambridge, Massachusetts.
Dershowitz, W.S. and H.H. Einstein, (1988) Characterizing rock joint geometry with
joint system models Rock Mechanics and Rock Engineering 21: 2151
Dershowitz, W. S. and Herda, H. H. (1992) Interpretation of fracture spacing and
intensity Proceedings of the 33rd U.S. Symposium on Rock Mechanics, eds. Tillerson, J.
R., and Wawersik, W. R., Rotterdam, Balkema. 757-766.
Dershowitz, W.S., Lee, G., Geier, J., Foxford, T., LaPointe, P., and Thomas, A. (1998)
FracMan, Interactive discrete feature data analysis, geometric modeling, and
exploration simulation, User documentation, version 2.6, Seattle, Washington: Golder
Associates Inc.
Einstein, H. H. and Baecher, G. B. (1983) Probabilistic and statistical methods in
engineering geology Rock Mechanics and Rock Engineering 16: 39-72.
Goodman, R. E. (1989) Introduction to Rock Mechanics. John Wiley & Sons, New
York.
ISRM, Commission on Standardization of Laboratory and Field Tests. (1978) Suggested
methods for the quantitative description of discontinuities in rock masses. International
Journal of Rock Mechanics and Mining Science, 15: 319-368
Martel, S.J. (1999) Analysis of fracture orientation data from boreholes. Environmental
and Engineering Geoscience. 5: 213-233.
Mauldon, M. (1994) Intersection probabilities of impersistent joints, International
Journal of Rock Mechanics and Mining Science & Geomechanics Abstracts, 31(2): 107115.
Mauldon, M., J. G. Mauldon. (1997) Fracture sampling on a cylinder: from scanlines to
boreholes and tunnels. Rock Mechanics and Rock Engineering. 30: 129-144.
Mauldon M., W. M. Dunne and M. B. Rohrbaugh, Jr. (2001) Circular scanlines and
circular windows: new tools for characterizing the geometry of fracture traces. Journal
of Structural Geology, 23(3): 247-258

59

Mauldon M. and X. Wang (2003) Measuring Fracture Intensity in Tunnels Using


Cycloidal Scanlines Proceedings of the 12th Panamerican Conference on Soil
Mechanics and Geotechnical Engineering and the 39th U.S. Rock Mechanics Symposium.
Mauton, Peter R. (2002) Principles and practices of unbiased stereology: an
introduction for bioscientists. Johns Hopkins University Press.
Owens, J.K., Miller, S.M., and DeHoff, R.T. (1994) Stereological Sampling and
Analysis for Characterizing Discontinuous Rock Masses. Proceedings of 13th
Conference on Ground Control in Mining. 269-276.
Priest, S. D. & Hudson, J. (1976) Discontinuity spacing in rock. Int. J. Rock Mech. Min.
Sci. Geomech. Abstr. 13: 135-148
Priest, S.D. (1993). Discontinuity Analysis for Rock Engineering. Chapman and Hall,
London.
Russ, J. C., DeHoff, R. T. (2000) Practical Stereology Kluwer Academic/Plenum
Publishers, New York
Terzaghi, R.D. (1965) Sources of errors in joint surveys. Geotechnique. 15: 287-304.
Warburton, P. M. (1980) A stereological interpretation of joint trace data. Int. J. Rock
Mech. Min. Sci. Geomech. Abstr., 17: 181-190
Wolfram Research, Inc. (2004). Mathematica, Version 5.1, Champaign, IL.
Yow, J.L. (1987) Blind zones in the acquisition of discontinuity orientation data.
International Journal of Rock Mechanics and Mining Sciences and Geomechanics
Abstracts. Technical Note. 24: 5, 317-318.

60

This page intentionally left blank.

61

Chapter 4
4 Estimating length and width of rectangular fractures from
traces on cylindrical exposures
Abstract

This study focuses on estimating length and width of subsurface fractures in sedimentary
rocks. Fractures in sedimentary rock are typically elongated along their strikes and their
shapes can be considered rectangles. The study shows how information about length and
width of rectangular fractures can be discerned from study of borehole/shaft-fracture (or
core-fracture) intersections. Based on the possible geometric relations between a fracture
and a sampling cylinder, six types of intersection: transection, long-edge, short-edge,
corner, single piercing, and double piercing, are defined. The probabilities of occurrence
of these intersection types are related to the length and width of the fractures and
borehole/shaft diameter. The mean length and width of the fractures are estimated
directly from the observed counts of different types of intersection in a borehole/shaft or
rock core. A computer program is developed to generate synthetic fractures sampled by a
circular cylinder and the derived estimators are tested by Monte Carlo simulations, which
show satisfactory results.
Key words: cylindrical rock exposures, fracture networks, fracture length and width,

rectangular fractures

62

4.1 Introduction

Rock engineers, geologists, and hydrologists have long made use of fracture trace data
from planar rock exposures to extract characterization of rock fractures and fracture
systems, and procedures for inferring the three-dimensional (3-d) fracture geometry from
traces have been the subject of considerable research (Priest & Hudson, 1976; Cruden,
1977; Baecher et al., 1977; ISRM, 1978; Warburton 1980; Cheeney, 1983; Einstein &
Baecher, 1983; Kulatilake & Wu, 1984; LaPointe & Hudson 1985; Dershowitz &
Einstein, 1988; Dershowitz & Herda, 1992; Priest 1993, 2004; Mauldon et al. 1994;
Zhang & Einstein, 1998; Mauldon et al. 2001; Zhang et al., 2002). For subsurface rock
masses, large planar exposures are, however, rare. Direct measures from cylindrical
exposures, such as circular tunnel and shaft walls, borehole images (Dershowitz et al.,
2000), rock bores, as well as geophysical surveys, often provide the main sources of
subsurface fracture data, for characterization of fracture systems.
Common practice for borehole sampling of fractures (here used as a generic term for
discontinuities of all types) is to treat the borehole as a one-dimensional (1-d) sampling
domain, equivalent to a scanline. Fracture frequency is then taken to be inversely
proportional to the probability of the observed fractures being intersected by the 1-d
sampling line (Terzaghi, 1965; Priest & Hudson, 1976; Dershowitz & Einstein, 1988;
Priest, 1993). If the ratio of sampling cylinder (i.e. borehole) diameter to the average size
of fractures is, however, greater than about 20%, the sampling domain effectively takes
on a higher dimension either 2-d or 3-d depending on whether fractures are sampled
using the cylinder surface only, or using the cylinder volume (Mauldon & Mauldon,
1997). Yet another form of 2-d sampling has been described, making use of virtual 2-d
boreholes applied to a subsurface cross-section. (Narr, 1993; Pascal et al., 1997; Fouch
& Diebolt, 2004). Methods have been proposed to use borehole data to determine
fracture orientation distribution (Martel, 1999), average spacing (Narr 1996), 3-d fracture
intensity (Owens et al., 1994; Mauldon & Wang, 2003), fracture surface roughness
(Thapa et al., 1996) and fracture size (Stone, 1984; Mauldon, 2000; Zhang & Einstein,

63

2000; zkaya, 2003; Wang et al., 2004, 2005). The present paper addresses inference of
fracture size, and also shape.
A wide variety of fracture geometry (or fracture system) models have been proposed in
the literature (e.g. Ruhland, 1973; LaPointe & Hudson, 1985; Dershowitz et al., 1998),
and most of these make assumptions about fracture shape and size distribution. The
Baecher model (Baecher et al. 1977; Dershowitz & Einstein, 1988), for example, assumes
circular disks with lognormally distributed radii. Orthogonal fracture models may
comprise either unbounded (e.g. Snow, 1965) or bounded (e.g. Mller, 1963; Gross, 1993)
joints. Field observations and mechanical consideration lend support to fracture models
for layered sedimentary rocks in which termination and propagation relationships (and
hence fracture size) are governed by elastic properties of the layers, boundary conditions
during fracturing and other mechanical and geometric factors (Engelder, 1993; Gross et
al., 1995). In particular, late forming fractures are likely to be normal to and terminate at
the primary fractures (or mechanical layer boundaries, Fig. 4.1), which may be either
bedding planes or systematic joint sets (Price, 1966; Helgeson & Aydin 1991; Gross,
1993; Engelder & Gross, 1993; Gross et al., 1995; Engelder & Fischer, 1996; Ruf et al.,
1998; Bai & Pollard 2000; Cooke & Underwood, 2001). One of the commonly observed
fracture patterns is that in which the late-forming cross joints propagate between and
orthogonal to preexisting primary joints (e.g. Fig. 4.1) in a ladder pattern (Gross, 1993;
Engelder & Gross, 1993). Field observations have also confirmed that fractures in
sedimentary rock are commonly perpendicular to bedding and elongated in one direction
(typically along strike, as shown schematically in Fig. 4.2) (Price, 1966; Suppe, 1985;
Priest, 1993). In all such cases, the shapes of fractures can be approximated as rectangles.
Estimates of (and models for) fracture size are usually predicated on assumed fracture
shape, such as circular disks (Baecher et al. 1977; Mauldon, 2000; zkaya, 2003),
elliptical disks (Zhang & Einstein, 2002), or rectangles (Narr 1996; Wang et al. 2004,
2005).

64

Fig. 4.1. Joints on limestone bed at Llantwit Major, Wales (photo provided by Matthew
Mauldon). Cross joints terminate at primary systematic joints.

Bedding
planes

Fractures (joints)

Rock mass

Fig. 4.2. Schematic drawing of dipping sedimentary beds, with primary joints either
terminating on bedding planes or cutting across several layers.

65

In this paper we focus on fractures in sedimentary rocks, in which fracture shape is


assumed to be rectangular; and we introduce methods to estimate the mean length and
width of fractures by using borehole/shaft-fracture intersection (trace) data. For
convenience, the derivations are based on the model of a vertical borehole/shaft sampling
a layered rock mass that contains strike-elongated fractures. The results can also be
applied, however, to a general orientation of the borehole/shaft, as long as the
assumptions discussed in Section 2 are applicable.
A simple model for borehole/shaft sampling of fractures in sedimentary rock is shown in
Fig. 4.3, in which fracture long axes align in the direction perpendicular to the borehole
axis. Consider a fracture of length l and width w (Fig. 4.3). The apparent width w is
defined as the width of the fracture when projected onto a plane normal to borehole axis
(the axis-normal plane in Fig. 4.3), and is related to fracture true width by
Ground surface

Ground surface

Borehole

Axis-normal
plane
(a)

Borehole

Axis-normal
plane
(b)

Fig. 4.3. Borehole/shaft and rectangular fractures and their projections on the axis-normal
plane. Note true width w and apparent width w. (a) vertical borehole/shaft; (b) general
case of a skew borehole/shaft

66

w = w cos .

(4.1)

where angle is the minimum angle between the fracture and the axis-normal plane (Fig.
4.3), i.e. the true dip in the case of a vertical borehole/shaft.

4.2 Assumptions

We make the following assumptions regarding fracture geometry, the borehole/shaft


sampling domain and the interrelationship between sampling domain and fracture system.
a) Fractures are planar rectangular objects with negligible thickness.
b) A shaft or a borehole is considered to be a right circular cylinder of diameter D,
oriented normal to the fracture elongation direction (or to strike when the
borehole/shaft is vertical).
c) The sampling domain refers to the cylindrical surface of the borehole/shaft.
d) The shaft/borehole is assumed to be long compared to its diameter. The end (or
ends) is not included in the cylindrical surface sampling domain.
e) Fracture length is greater than borehole/shaft diameter and also greater than the
apparent width wof the fracture. Note that if the latter condition is not the case,
the length and width can be interchanged.
f) The location of the borehole/shaft is independent of the locations of fractures in
the rock mass to be explored. This is the case when we have little knowledge of
fracture networks before the excavation of a borehole or a shaft. Statistically, this
assumption ensures that the portion of the rock mass intersected by the cylindrical
surface of the borehole/shaft corresponds to a uniformly distributed, random

67

sample. Isotropy is achieved automatically with respect to directions


perpendicular to the borehole/shaft axis.

4.3 Borehole/shaft-fracture intersection types

Six types of intersection: transection, long-edge, short-edge, corner, single-piercing, and


double-piercing, are defined, based on the possible geometric relations between a
rectangular fracture and a borehole/shaft (Table 4.1). Corresponding to each intersection
type are characteristic types of fracture trace on the unrolled borehole surface (Fig. 4.4).
It may be observed that type A intersections can occur only if fracture apparent width w
is greater than borehole/shaft diameter D. Type C1 and C2 intersections can occur only if
w is less than borehole/shaft diameter D. Note that Wang et al. (2004) used a slightly
different terminology, referring to transactions as complete intersections. A simple
illustration of all six intersection types is shown in Fig. 4.5, in which rectangular fractures
and boreholes/shafts are projected onto the axis-normal plane, on which boreholes/shafts
project as circles.
For the model discussed in this paper rectangular fracture elongated along strike we
can identify each of the six intersection types from fracture traces on the borehole/shaft
surface or unrolled trace map (Table 4.1, Fig. 4.4). Note that in Table 4.1, the
characteristics of each intersection type are based on knowing fracture dip-direction (cut
line along fracture dip-direction). If fractures are perpendicular to the borehole/shaft, this
direction can not be determined. In this case, only A-type (Transection) and C2-type
(double piercing) intersections can be easily identified; and it will be difficult to apply the
estimators discussed in this paper.

68

Table 4.1. Six borehole/shaft-fracture intersection types


Intersection

Symbol

type

Example

Trace (or traces) on

in Fig. 4.4 borehole/shaft surface

Trace (or traces) on the


unrolled trace map with cut
line along dip-direction

Transection

Full ellipse

Full sine curve

Long-edge

B1

Partial ellipse,

Partial segments of sine curve,

symmetric with respect

symmetric with respect to cut

to dip-direction or anti-

line or anti-dip-direction

dip-direction
Short-edge

Corner

Single

B2

B3

C1

piercing

Double
piercing

C2

Partial ellipse, centered

Partial segments of sine curve,

with respect to strike

centered along strike

Partial ellipse, not

Partial segments of sine curve,

symmetric with respect

not symmetric with respect to

to any direction

any direction

Single partial ellipse,

Single partial segments of sine

similar to one of the

curve, similar to one of the

paired C2 traces

paired C2 traces

Paired partial ellipse,

Paired partial segments of sine

symmetric with respect

curve, symmetric with respect

to dip-direction or anti-

to dip-direction or anti-dip-

dip-direction

direction

69

-/2

/2

3/2

Cut line
1

1
A

Dip direction

B2

B1

B1

3
Borehole
axis direction

B3 4

C1

C2

C2

D
Cut line
(a)

Cut line
(b)

Fig. 4.4. A vertical borehole of diameter D intersects rectangular fractures in six ways.
The unrolled trace map is developed from the borehole wall by cutting along fracture dip
direction. Intersection types are marked beside the corresponding traces. (a) borehole and
fractures; (b) Unrolled trace map. Coordinate axis is defined with = -/2 at the cut
line. If the cut line were taken along strike, the angular coordinate would be from 0 to
2.

70

Boreholes (diameter D < w)


B3
B2

(a)

B1

Projected
Fracture

Boreholes (diameter D w)
(b)
w

C1

C2

l
Fig. 4.5 Six types of intersection between projected fractures (shaded) and
boreholes/shafts (dashed circles) are shown on the axis-normal plane. (a) D < w; (b) D
w

4.4 Probabilistic model for occurrence of intersection types


We define symbols in Table 4.2.
In this section, we discuss the probabilities of occurrence of each borehole/shaft-fracture
intersection type. The key to the probabilistic model is that, on the axis-normal plane, the
center of a borehole/shaft must be inside a specific region around the projected fracture in
order for an intersection to occur (Fig. 4.6). Each intersection type, therefore, has a

71

corresponding locus with respect to the projected fracture on the axis-normal plane (Figs.
4.74.9).
Table 4.2. Defined symbols
Symbol

Definition

Aspect ratio of a fracture: = l / w .

~
N

~
Number of occurrences of borehole/shaft-fracture intersections. N with a
subscript (e.g. B1, B2) indicates the number of occurrences for a specific
intersection type (or several types).

Length of the borehole/shaft.

~
Expected frequency of borehole/shaft-fracture intersections: = N / H .
with a subscript (e.g. B1, B2) indicates the expected frequency of
intersections for a specific intersection type (or several intersection types).

(l, w)

Expected frequency of borehole/shaft-fracture intersections when sampling in


fractures of constant size (the projected fracture has the dimension of l w on
the axis-normal plane).

Area of a region on the axis-normal plane corresponding to an intersection


type (or several intersection types). with a subscript (e.g. B1, B2)
indicates a specific intersection type (or several intersection types).

fL,W(l,w)

Joint probability density function (pdf) of fracture length and fracture


apparent width.

l and w

Expected values of fracture length and apparent width, respectively. For


constant fracture orientation, w = w sec , where w is the mean fracture
width.

72

Borehole/shaft
location

D/2

D/2

w
Projected Fracture

Region of
intersection

Fig. 4.6. The locus for borehole/shaft-projected fracture intersection on the axis-normal
plane is the region inside by the dashed line. If the center of borehole/shaft is inside the
region, an intersection occurs.

Projected
fracture

D/2

D/2

B3

B2

B3

B1

B1

B3

B2

B3

l
Fig. 4.7. Each intersection type has a corresponding locus on the projected fracture (bold
rectangle) for the center of the borehole. In this case, w > D.

73

D/2

D/2

D/2

B1

C1

C1

C2

B2, B3

B1

B2, B3

Projected

fracture

Fig. 4.8. Each intersection type has a corresponding locus on the projected fracture (bold
rectangle) for the center of the borehole. In this case, D/2 < w D.

D/2

D/2
B1

D/2

C2

w
C1
B2, B3

C2
B1

C1
B2, B3

Projected
fracture

Fig. 4.9. The corresponding locus for the center of the borehole/shaft for each intersection
type around the projected fracture (bold rectangle) on the axis-normal plane for case w
D/2.

74

The regions, separated by dashed lines in Figs. 4.74.9, each define the possible locus of
the center of a borehole/shaft, corresponding to each intersection type (e.g. A-type, B1type, B2-type, B3-type, C1-type and C2-type). For instance, when the center of the
borehole/shaft falls into the shaded region marked as A, a transection intersection will
occur and a full cosine trace will be induced on the trace map. Consider a fracture of
dimension l w projected on the axis-normal plane so that its projection has size of l w.
Then the area of the region for each intersection type as well as the area of all intersection
regions can be determined from simple geometry (Table 4.3).
If the last assumption in Section 4.2 holds, the location of a borehole/shaft is independent
of the location of fractures. If we were to introduce a Cartesian coordinate system on the
axis-normal plane, with origin at the borehole/shaft center, the locations of projected
fractures would be uniform on that plane (this holds even when projected fractures
overlay). In other words, projected fractures on the axis-normal plane have the same
probability to be at any point on that plane. Therefore, for a rectangular fracture, the
frequency of any type of intersection is proportional to the area of the corresponding
region on the axis-normal plane (Figs. 4.7-4.9). For fractures of constant orientation and
size, this can be expressed as the equations below.

(4.2)

where is a frequency (Table 4.2), is an area (Tables 4.2 and 4.3), and is identical to
the 3-d fracture density P30 (Dershowitz, 1992), i.e., number of fractures per unit volume
of the rock mass. P30 is assumed to be constant.

There are three cases to consider for the probabilistic model, depending on the relative
magnitudes of borehole/shaft diameter D and fracture apparent width w. In each case,
fractures are assumed to be of constant orientation.

75

Table 4.3. Areas of regions corresponding to each borehole/shaft-fracture intersection


type from geometry

Area

Case 4.4.1: w > D

Case 4.4.2.1: D w > D/2

Case 4.4.2.2: w D/2

w l Dl D w + D 2

NA

NA

B1

2Dl D 2

B2

D 2

2Dw D 2

2 w l 2 D w + D 2

Dw

* B1+C 2 = wl + Dl Dw D

D 2
4

D 2

w
+ D 2 cos 1
4
D

w D 2 w2

* B 2 + B 3 = 2 D w +

B3

D2 +

3D
4

Dw +

D 2
2

w
D cos + w D 2 w 2
D
2

3D 2
w
2 D 2 cos 1
4
D

+ 2 w D 2 w2

C1

NA

w
D 2 cos 1 w D 2 w 2
D

w
D 2 cos 1 w D 2 w 2
D

C2

NA

wl + Dl + wD D 2

* B1+C 2 = wl + Dl Dw D

total

wl + Dl + Dw +

* Area of combined regions

76

D 2
4

4.4.1 w > D
In this case, the apparent widths of all fractures are greater than the borehole/ shaft
diameter D. When these relatively large fractures intersect a borehole/shaft, a transection
intersection (type A) may occur. Piercing intersection (C types), on the other hand, are
impossible.
For fractures of constant orientation and size with projected size of l w on the axisnormal plane, if the fracture density is , from Eq. (4.2) and Table 4.3, the frequencies,
here interpreted as expected values (Owens et al., 1994), of B1, B2 and B3 intersection are
given by,

B1 (l , w) = B1 = 2Dl D 2 (1 + / 4 )

(4.3)

B 2 (l , w) = B 2 = 2Dw D 2 (1 + / 4 )

(4.4)

B 3 (l , w) = B 3 = D 2 (1 + 3 / 4 ) .

(4.5)

and

For a set of fractures with constant orientation but varied size, let fL,W(l, w) denote the
joint pdf of fracture length and apparent width. The expected value of frequency B3 is
obtained by integrating the right side of Eq. (4.5) over all values of l and w (in this case,
w l < ; D < w < ), noticing that B3(l, w) in Eq. (4.5) is expressed as a function
independent of fracture size.

B 3 = D 2 (1 + 3 / 4 ) f L ,W (l , w)dldw
l , w

= D (1 + 3 / 4 )
2

The constant can then be expressed as

77

(4.6)

B 3
.
D (1 + 3 / 4)

(4.7)

Similarly, the expected value of frequencies B1 and B2 can be obtained by integrating


the right side of Eqs. (4.3) and (4.4), respectively, over all values of l and w (in this case,
w l < ; D < w < ). Noticing that the right side of Eq. (4.3) is not a function of w
and the right side of Eq. (4.4) is not a function of l.

B1 = 2Dlf L,W (l , w)dldw


l , w

D 2 (1 + / 4 ) f L,W (l , w)dldw
l , w

(4.8)

= 2Dl D 2 (1 + / 4 )
and

B 2 = 2Dwf L ,W (l , w)dldw
l , w

D 2 (1 + / 4 ) f L ,W (l , w)dldw

(4.9)

l , w

= 2D w D 2 (1 + / 4 )
where l and w are the mean fracture length and the mean fracture apparent width,
respectively.

Substituting Eq.(4.7) into Eqs. (4.8) and (4.9), the mean fracture length l and the mean
fracture apparent width w can be determined as:

l =

D B1
(4 + 3 ) + 4 + ,

8 B 3

(4.10)

and

78

w =

D B 2
(4 + 3 ) + 4 + .

8 B 3

(4.11)

Note that the expected values of fracture length and apparent width are proportional to
borehole diameter D, and are linear functions of the ratios of expected frequency of B1type and B2-type intersections over B3-type intersections, respectively.

4.4.2 w D
When fractures are narrow, or sufficiently steep that their apparent widths are smaller
than borehole/shaft diameter, piercing intersections (C types) may occur, whereas a
transection intersection (type A) is impossible. These narrow fractures are called piercing
fractures. Piercing fractures can pierce a borehole/shaft in either of two ways: singly or
doubly, as fracture #5 (doubly piercing) and fracture #6 (singly piercing) show in Fig.
4.4(a). Both single piercing and double piercing fractures intersect the borehole with two
long edges and leave similar traces on borehole/shaft walls, except that double piercing
fractures have paired traces (Fig. 4.4(b)). Double piercing fractures are easily identified
on unrolled trace maps derived from shaft surface or borehole imagery; and the amplitude
h of the traces (Fig. 4.4(b)) can be used to determine the apparent width of fractures by

w = h / tan .

(4.12)

79

4.4.2.1 D/2 < w D


The procedure to determine the length of piercing fractures with apparent widths greater
than the radius and less than the diameter of the sampling borehole/shaft is as follows.
For fractures of constant orientation and size with projected size of l w on the axisnormal plane, if the fracture density is , from Eq. (4.2) and Table 4.3, consider the
following combinations of frequencies, as functions of fracture size.

total (l , w) B1 (l , w) C 2 (l , w)
= [ total B1 C 2 ]
= 2Dw + 2D 2 / 4

(4.13)

and

B1 (l , w) + 2 C 2 (l , w) = [ B1 + 2 C 2 ]
= 2Dl D 2 (1 + / 4 )

(4.14)

For a set of fractures with constant orientation but varied size, the expected value of the
linear frequency combination (total - B1 - C2) can be obtained by integrating right side
of Eq. (4.13) over all values of l and w(in this case, w l < ; D/2 < w D), noticing
that the combination, given fracture size, is not a function of l.

total B1 C 2 = 2D wf L,W (l , w)dldw


l ,w

+ 2D 2 / 4

) f
l , w

L ,W

(l , w)dldw

= 2D w + D 2 / 2

(4.15)

Note that the mean apparent width w can be estimated directly by averaging all values
of w determined by Eq.(4.12). Then the constant can be estimated by

80

total B1 C 2
.
2 D w + D 2 / 2

(4.16)

The expected value of the linear frequency combination (B1 + 2C2) can be obtained by
integrating right side of Eq. (4.14) over all values of l and w(w l < ; D/2 < w D),
noticing that the combination, given fracture size, is not a function of w.

B1 + 2C 2 = 2D lf L,W (l , w)dldw
l , w

D 2 (1 + / 4) f L,W (l , w)dldw
l , w

= 2Dl D 2 (1 + / 4)

(4.17)

Substituting Eq. (4.16) into Eq. (4.17) yields,

B1 + 2C 2 =

2 l D (1 + / 4)
(total B1 C 2 ) .
2 w + D / 2

(4.18)

Finally, the mean fracture length l can be determined as

l =

B1 + 2C 2
( w + D / 4) + D (4 + ) .
total B1 C 2
8

(4.19)

4.4.2.2 w D/2
The fractures in this case are very narrow piercing fractures or very steep fractures (angle

close to 90) whose apparent widths are smaller than borehole/shaft radius. This
scenario is rare in borehole samplings, but may occur for shafts or tunnels. The regions
separated by the dashed lines in Fig. 4.9 show the locus of the center of a borehole/shaft

81

corresponding to each intersection type (e.g. C1-type, C2-type, B1-type, B2 and B3-type)
for this case.

For fractures of constant orientation and size with projected size of l w on the axisnormal plane, if the fracture density is , consider the total borehole/shaft-fracture
intersection frequency and the following linear combination of frequencies from Eq. (4.2)
and substitute for areas from Table 4.3, we have

total (l , w) = total = wl + Dl + Dw + D 2 / 4

(4.20)

total (l , w) B1+ C 2 (l , w) = ( total B1+ C 2 )


= 2Dw + D 2 / 2

(4.21)

For a set of fractures with constant orientation but varied size, the expected value of the
frequency combination (total - B1+C2) can be obtained by integrating right side of Eq.
(4.21) over all values of l and w(in this case, w l < ; 0 w D/2), noticing that the
combination, given fracture size, is not a function of l.

total B1+C 2 = 2D wf L,W (l , w)dldw


l ,w

+ D 2 / 2

) f
l ,w

L ,W

= 2D w + D 2 / 2

(l , w)dldw .
(4.22)

Again, the mean apparent width w can be estimated by averaging all w determined by
Eq.(4.12). Then constant can be estimated by

total B1+ C 2
.
2 w D + D 2 / 2

(4.23)

82

The expected value of the total frequency total is expressed as

total = wl + Dl + Dw + D 2 / 4 f L,W (l , w)dldw .


l , w

(4.24)

The integral in Eq.(4.24) is difficult to evaluate unless we know or could assume the joint
pdf of fracture length and apparent width. For instance, if we assume:
a) Fracture length is constant (and equal to l*). Then the integral in Eq.(4.24) is
simplified as follows.

total = wl + Dl + Dw + D 2 / 4 f L,W (l , w)dldw


l , w

= l* + D

) wf
l , w

L ,W

(l , w)dldw

+ D l * + D / 4

) wf
l , w

L ,W

= l * + D w + D l * + D / 4

(l , w)dldw

(4.25)

Substituting Eq. (4.23) into Eq. (4.25), l* can be determined.

w D + D 2 / 4
2total
l =
1
.
D + w
total B1+ C 2
*

b) Fracture aspect ratio is constant (* = l / w = l / w).


Then Eq.(4.24) is expressed as

83

(4.26)

total = * w 2 + * Dw + Dw + D 2 / 4 f L,W (l , w)dldw


l ,w

= * w 2 f L,W (l , w)dldw
l , w

+ D 1+*

+ D 2 / 4

( )

) wf
l ,w

) f
l , w

L ,W

L ,W

(l , w)dldw

(l , w)dldw

= * E w 2 + D 1 + * w + D 2 / 4

(4.27)

where E(w2) is the second moment of fracture apparent width probability density
function, which can be estimated by averaging all w2 determined by Eq.(4.12).

Substituting Eq. (4.23) into Eq. (4.27), * can be solved.

D + D 2 / 4
2total
1 w 2
total B1+ C 2 E w + D w

* =

( )

(4.28)

Finally, the mean fracture length can be estimated as:

2 D + D 2 w / 4
2total
l = * w =
1 w 2
total B1+ C 2 E ( w ) + D w

(4.29)

4.4.3 Summary of fracture length and width estimators


The estimators discussed in this section (Eqs. (4.10, 4.11, 4.19, 4.26, 4.29)) are
categorized in three cases, depending on the relative magnitudes of borehole/shaft
diameter D and fracture apparent width w. Judgment should be made to determine
which estimator(s) will be applied to estimate fracture length and apparent width. The
flowchart in Fig. 4.10 shows how this procedure is carried out. Note that if there are no
C-type intersections, it is much likely that fracture apparent width is greater than

84

borehole/shaft diameter. Estimators for the case 4.4.1 (Eqs. (4.10, 4.11)), therefore, can
be used to estimate fracture length and apparent width. If there are no A-type
intersections, on the other hand, it is much likely that fracture apparent width is less than
borehole/shaft diameter. Estimators for the case 4.4.2 (Eqs. (4.19, 4.26, 4.29)), can be
used, depending on whether the measured fracture apparent width is greater than
borehole/shaft radius or not.

In practice, when using the estimators, the ratios of expected frequencies can be replaced
by the ratios of the corresponding observed intersection counts. This is demonstrated in
the following example.

Unrolled borehole/shaft trace map


Identify intersection types and their counts
No C-type intersections

Both A and C intersections

No A-type intersections

Case 4.4.1: w > D

Use judgment

Case 4.4.2: w D

Determine mean fracture apparent width w


by averaging all w calculated by Eq. (4.12)

Mean fracture length: Eq. (4.10)


Mean fracture width: Eq. (4.11)

Case 4.4.2.1

Yes

w > D/2 ?
No

Mean fracture length: Eq. (4.19)

Case 4.4.2.2
Mean fracture length:
Eq. (4.26) or Eq. (4.29)

Fig. 4.10. Flowchart of choosing estimators to estimate mean fracture length and width.

85

4.5 Examples

Example 1: suppose a 600.0 feet long, 6.0 inch in diameter vertical borehole was drilled

in a sedimentary rock mass. From borehole image, three sets of fractures were observed,
with an average dip of 35.0, 60.0, and 80.0, respectively. The borehole-fracture
intersection types were identified by unrolled borehole images, and the counts of
intersections for each intersection type are listed in Table 4.4.

Table 4.4. Borehole-fracture intersection counts from a borehole sampling


Intersection counts

Intersection type

Set 1

Set 2

Set 3

18

B1

105

233

58

B2

38

78

34

B3

57

63

62

C1

C2

55

21

Total number of intersections

218

429

175

Average dip

35.0

60.0

80.0

Borehole length (ft.)

600.0

Borehole diameter (in.)

6.0

Estimated mean fracture length l (ft)

2.0

2.4

1.0

Estimated mean fracture apparent width w (ft)

1.0

0.4

0.1

Estimated mean fracture width w (ft)

1.2

0.8

0.6

Estimated mean squared fracture width w2 (ft2)


Estimated Aspect Ratio (l/w)

0.015
1.6

86

3.0

1.7

For fracture set 1, no C-type intersections were found. Therefore, we use estimators Eqs.
(4.10) and (4.11) for the case 4.4.1 (w > D) to estimate mean fracture length and width.
The ratios of expected intersection frequencies are replaced by the ratios of intersection
counts.

l =

D B1
(4 + 3 ) + 4 +

8 B3

D N B1
~ (4 + 3 ) + 4 +
8 N B3

6.0 70
(4 + 3 ) + 4 +

8 25

= 23.9in. = 2.0 ft

(4.30)

and

w =

D B 2
(4 + 3 ) + 4 +

8 B3

D N B2
~ (4 + 3 ) + 4 +
8 N B3

6.0 18
(4 + 3 ) + 4 +

8 25

= 12.1in. = 1.0 ft

(4.31)

Assume that fractures in set 1 are of constant orientation (dip = 35), fracture true width
is estimated by

w = w / cos = 1.0 ft / cos(35) = 1.2 ft .

For fracture set 2, no A-type intersections were identified and the average fracture
apparent width is w = 0.4 ft by averaging all values of fracture apparent width

87

(4.32)

determined by Eq. (4.12). We use estimator Eq. (4.19) for case 4.4.2.1 (D/2 < w D) to
estimate the mean fracture length.

l =

B1 + 2C 2
( w + D / 4) + D (4 + )
total B1 C 2
8

~
~
N B1 + 2 N C 2
D
~
~
~ ( w + D / 4) + (4 + )
8
N total N B1 N C 2
=

(4.33)

110 + 2 55
(0.8 + 6.0 / 4) + 6.0 (4 + )
204 110 55
8

= 28.5in. = 2.4 ft

Assume that fractures in set 2 are of constant orientation (dip = 60), fracture true width
is estimated by

w = w / cos = 0.4 ft / cos(60) = 0.8 ft .

(4.34)

For fracture set 3, no A-type intersections were identified and the average fracture
apparent width is w = 0.1 ft by averaging all values of fracture apparent width
determined by Eq. (4.12) and E(w2) = 0.015 ft2. By assuming that fractures are of
constant aspect ratio, we use estimator Eq. (4.29) for case 4.4.2.2 (w D/2) to estimate
the mean fracture length.

2 D + D 2 w / 4
2total
1 w 2
l =
total B1+C 2 E ( w ) + D w

w2 D + D 2 w / 4
2 N total
~
1
~
2
N total N B1+C 2 E ( w ) + D w

2
2
2 175
0.1 0.5 + 0.5 0.1 / 4
=
1
0.015 + 0.5 0.1
175 58 21

= 1.0 ft

88

(4.35)

Assume that fractures in set 3 are of constant orientation (dip = 80), fracture true width
is estimated by

w = w / cos = 0.1 ft / cos(80) = 0.6 ft .

(4.36)

Example 2: a 10.4 inch in diameter borehole was drilled in a sedimentary rock mass.

From borehole FMI image, a set of fractures was observed, with an average dip of 82.3.
The borehole-fracture intersection types were identified and the counts of intersections
for each intersection type are listed in Table 4.5.

Table 4.5. Borehole-fracture intersection counts from borehole sampling


Intersection type

Intersection counts

17

B1

103

B2

21

B3

23

C1

C2

98

Total number of intersections

262

Average dip

82.3

Borehole diameter (in)

10.4

Estimated mean fracture length l (in)

59.0 ~ 65.9

Estimated mean fracture apparent width w (in)

3.1

Estimated mean fracture width w (in)

23.1

Estimated mean squared fracture width w2 (in2)

14.5

Estimated Aspect Ratio (l/w)

2.5 ~ 2.8

89

Although both A-type and C-type intersections were identified for this fracture set, Ctype intersection counts (98) are much higher than A-type intersection counts (17).
Therefore, w D (case 4.4.2) is assumed to be more suitable for this case. The average
fracture apparent width is w = 3.1 in by averaging all values of fracture apparent width
determined by Eq. (4.12) and E(w2) = 14.5 in2. Since w is less than borehole radius (5.2
in), assuming that fractures are of constant aspect ratio, we use estimators Eq. (4.29) for
case 4.4.2.2 (w D/2) to estimate the mean fracture length.

( 2 D + D 2 w / 4 )
2total
1 w 2
l =
total B1+C 2 E ( w ) + D w

w2 D + D 2 w / 4
2 N total
~
1
~
2
N total N B1+C 2 E ( w ) + D w
2
2
2 262

3.1 10.4 + 10.4 3.1 / 4


1
=
14.5 + 10.4 3.1
262 103 98
= 59.0in

(4.37)

Assume that fractures are of constant orientation (dip = 82.3), fracture true width is
estimated by

w = w / cos = 3.1in / cos(82.3) = 23.1in .

(4.38)

4.6 Monte Carlo simulations


Monte Carlo simulations were carried out using a computer program developed in Visual
C++. The program generates a population of synthetic rectangular fractures intersected
by a borehole/shaft (Fig. 4.11). Fracture traces are computed (right-hand side window in
Fig. 4.11) and used to determine the occurrence of each borehole/shaft-fracture

90

intersection type. For each simulation, the size (length and width) of fractures and the
dimensions of borehole (length and diameter) are constant; and the total number of
generated fractures, the area of each fracture, as well as the size of the generation region,
were recorded. The observed counts of intersection types are used to estimate fracture
length and width by using Eqs. (4.10) and (4.11), or Eqs. (4.12) and (4.19), depending on
the relative size of borehole/shaft diameter D and fracture apparent width w.
Twenty-seven scenarios (Table 4.6) were simulated by systematically changing fracture
dip (0, 30, and 60 ), length l (2, 10, and 20) and aspect ratio (1, 2, and 10). For all
cases, the cylinder (borehole/shaft) diameter was held constant (D = 0.2), and the fracture
intensity P32 was held constant at 1.0 [L-1]. Depending on the mean size of the fracture,
the number of fractures varied from one scenario to the next. For each scenario, fifty
simulations were run; and the average estimated fracture length and width, and average
percent error, variance and coefficient of variation are listed in Table 4.6.
An example of simulation results for scenario 1 is shown in Fig. 4.12. In this scenario,
the length and width of the generated fractures are both set to be 2.0, i.e., the fractures are
square. In 50 simulations, there are an average of 354 B1-type intersections, 353 B2-type
intersections and 65 B3-type intersections observed. The estimated fracture length and
width are 2.04 and 2.02, by using Eqs. (4.10) and (4.11), respectively.
Overall, the simulation results (Table 4.6) show that the derived equations produce
reasonable, good estimates of fracture length and width (absolute percent error is less
than 15% of the actual fracture size). The largest errors occur for scenarios 4, 7, 13, 16,
22, and 25, in which the data are distorted by very low or even no occurrence of B3-type
intersections observed in some simulations. The comparison of percent error and
coefficient of variation versus observed B3-type intersection counts (Fig. 4.13) shows a
big increase of both error and variation of the estimators when the observed B3-type
intersection counts drops from 26 to 7. This suggests that cares should be taken when

91

applying these estimators in practice, especially when fractures are much larger than
borehole/shaft diameter and the counts of B3-type intersections is very low.

1
2
3
4

Fractures

5
6
7

8
9

Borehole

10
11

Trace map

Fig. 4.11. A computer program was developed to generate a population of rectangular


fractures intersected by a borehole/shaft. In the simulation shown above, number of
fractures generated = 4796; fracture length = 2; aspect ratio = 1; dip/dip direction = 30
deg/north; borehole length = 10; borehole diameter = 0.2; fracture area per unit rock mass
volume = 5.0. The trace analysis produced 36 A-type intersections; 6 B1-type
intersections; 6 B2-type intersections; 1 B3-type intersections; 0 C1-type intersections; 0
C2-type intersections.

92

Dip

Fracture length [L]

Fracture width [L]

Aspect Ratio

Apparent width [L]

Number of fractures
generated

Fracture density (P30) [L-3]

B1

B2

B3

C2

Total

Average

Absolute percent
error

Variance

Coefficient of
variation

Average

Absolute percent
error

Variance

Coefficient of
variation

Estimated fracture
width [L]

Scenario

Estimated fracture
length [L]

Averaged number of
intersections

Table 4.6. Parameters and results of Monte Carlo simulations

3673

0.25

1569

354

353

65

2341

2.0

2%

0.08

14%

2.0

1%

0.08

14%

4754

0.50

1443

735

329

134

2640

2.0

2%

0.03

9%

1.0

1%

0.01

8%

0.2

10

0.2

19562

2.50

3660

45

665

4369

2.0

1%

0.01

5%

0.2

1%

0.00

2%

10

10

10

3295

0.01

1917

77

78

2075

12.2

22%

55.56

61%

12.3

23%

56.94

62%

10

4150

0.02

1884

157

79

2125

11.7

17%

60.40

66%

6.0

21%

21.01

76%

10

10

16832

0.10

1566

794

66

26

2451

10.9

9%

4.76

20%

1.1

6%

0.04

19%

20

20

20

3249

0.00

1955

40

41

2036

11.6

42%

12.00

30%

11.5

43%

10.77

29%

20

10

10

4077

0.01

1939

79

40

2058

20.3

2%

62.12

39%

10.3

3%

16.46

39%

20

10

16505

0.03

1783

400

37

2227

24.4

22%

137.81

48%

2.4

21%

1.18

45%

10

30

1.73

3673

0.25

1374

366

316

66

2121

2.1

4%

0.07

13%

2.1

5%

0.07

12%

11

30

0.87

4754

0.50

1198

730

276

134

2338

2.0

1%

0.03

9%

1.0

1%

0.01

8%

12

30

10

0.17

19562

2.50

3162

27

610

239

4038

2.0

0%

0.01

4%

0.2

0%

0.00

0%

13

30

10

10

8.66

3295

1654

81

70

1807

13.4

34%

59.92

58%

13.4

34%

61.76

58%

14

30

10

4.33

4150

0.02

1623

156

65

1850

10.5

5%

44.11

63%

5.1

2%

12.34

69%

15

30

10

10

0.87

16832

0.10

1308

793

55

28

2184

10.1

1%

3.67

19%

1.0

1%

0.03

18%

16

30

20

20

17.3

3249

0.00

1692

39

34

1766

10.5

47%

14.20

36%

10.8

46%

16.40

38%

17

30

20

10

8.66

4077

0.01

1677

81

34

1794

19.3

3%

70.41

43%

9.7

3%

21.64

48%

18

30

20

10

1.73

16505

0.03

1513

396

32

1947

22.4

12%

101.61

45%

2.3

13%

1.09

46%

19

60

3673

0.25

717

365

163

64

1309

2.1

5%

0.06

12%

2.1

4%

0.06

12%

20

60

0.5

4754

0.50

539

725

128

133

1525

2.0

1%

0.02

7%

1.0

1%

0.00

7%

21

60

0.2

10

0.1

19562

2.50

1847

381

896

3130

2.0

1%

0.01

4%

0.2

0%

0.00

0%

22

60

10

10

3295

0.01

943

79

37

1062

14.1

41%

63.84

57%

13.2

32%

55.09

56%

23

60

10

2.5

4150

0.02

901

156

40

1103

11.3

13%

64.42

71%

5.9

19%

13.07

61%

24

60

10

10

0.5

16832

0.10

590

788

26

27

1432

10.2

2%

4.14

20%

1.0

2%

0.05

21%

25

60

20

20

10

3249

0.00

971

41

20

1033

11.8

41%

17.73

36%

11.5

43%

20.33

39%

26

60

20

10

4077

0.01

949

81

20

1051

21.7

9%

51.16

33%

11.3

13%

25.64

45%

27

60

20

10

16505

0.03

800

398

17

1223

22.9

14%

123.09

48%

2.3

14%

0.92

42%

Diameter of borehole/shaft [L]

0.2

Fracture area per unit rock mass volume (P32) [L-1]

Simulations per scenario

50

93

Length

Estimated Length
Average Estimated Length
Actual Length
0
0

10

20

30

40

50

Simulations
3

Width

Computed Width
Average Estimated Width
Actual Width
0
0

10

20

30

40

50

Simulations

Fig. 4.12. Comparison of computed fracture length and width vs. actual fracture length
and width for scenario 1. Fracture dip = 0; length = 2.0; aspect ratio = 1.0.

94

80%

B3 intersection counts

700

# of B3 intersections
Absolute percent error
Coefficient of variation

600

70%

60%

500

7 B3
intersection

400

50%

40%
300

26 B3
intersection

200

30%

20%

100

Percent error or
coefficient of variation

(a)

10%

0%

Scenarios

B3 intersection counts

80%

# of B3 intersections
Absolute percent error
Coefficient of variation

600

70%

60%

500

7 B3
intersection

400

50%

40%
300

26 B3
intersection

200

30%

20%

100

Percent error or
coefficient of variation

700

(b)

10%

0%

Scenarios
Fig. 4.13. Percent error and coefficient of variation of estimators for (a) fracture length
and (b) fracture width, in comparison with observed counts of B3-type borehole/shaftfracture intersections.

95

The fracture apparent widths in Scenario 12 and 21 are 0.17 and 0.10, respectively, which
are less than borehole diameter (0.2). The estimators (Eqs. (4.12) and (4.19)) for case D
w > D/2 give very good estimates for fracture width and length (Table 4.6). In
addition, estimator (Eq. (4.10)) for case w > D were also used to estimate fracture length
and resulted an estimated length of 1.9 and 1.8 for scenario 12 and 21, respectively. The
percent errors are 4% and 10%, respectively, which implies that it will not cause major
errors by using Eq. (4.10) to estimate fracture length even if fracture apparent width is
smaller than borehole/shaft diameter.

4.7 Discussion & Conclusions


The goal of this study was to develop a general model for estimating mean rectangular
fracture length and width from traces on cylinder walls. Fractures in sedimentary rocks
are commonly elongated along strike and terminated on bedding planes or primary joint
sets, therefore assumed rectangular in shapes. From the geometric relations between a
fracture and a borehole/shaft, six types of intersection are defined. The features of each
intersection type described in the paper can be used to identify the six intersection types
from unrolled borehole/shaft trace maps. The occurrences of the intersection types are
related to fracture size and borehole/shaft diameter, assuming independence between
locations of borehole/shaft and the fractures.
Three cases regarding relations between fracture apparent width and the borehole/shaft
diameter are discussed. For each case, estimators are derived to estimate mean fracture
length and width based on probabilistic models. The estimators are confirmed by Monte
Carlo simulations, which gave satisfactory results. It is also pointed out in the paper that
caution should be used when applying the estimator to the cases that the size of the
sampled fractures is much larger than the diameter of boreholes/shafts.

96

Acknowledgements

Partial support from the National Science Foundation, Grant Number CMS-0085093, is
gratefully acknowledged. Also should be acknowledged are Chris Heiny from University
of Tennessee and Jeramy Decker from Virginia Tech, who helped carrying out
simulations.

97

References

Baecher, G.B., Einstein, H.H., and Lanney, N.A. (1977). Statistical description of rock
properties and sampling, In: Proc. 18th U.S. Symp. Rock Mech., Colorado School of
Mines. 5C: 1-8.
Bai, Taixu, Pollard, D.D. (2000). Closely spaced fractures in layered rocks: initiation
mechanism and propagation kinematics, Journal of Structural Geology, 22: 1409-1425
Cheeney, R.F. (1983). Statistical methods in geology for field and lab decisions, Allen
& Unwin Ltd. London. UK
Cooke, M.L. and Underwood, C.A. (2001). Fracture termination and step-over at
bedding interfaces due to frictional slip and interface opening, Journal of Structural
Geology, 23: 223-238.
Cruden, D.M. (1977). Describing the size of discontinuities, International Journal of
Rock Mechanics and Mining Science & Geomechanics Abstracts, 14: 133-137
Dershowitz, W.S. (1984). Rock joint systems, Ph.D. Thesis, Massachusetts Institute of
Technology, Cambridge, Massachusetts.
Dershowitz, W.S. and Einstein, H.H. (1988). Characterizing rock joint geometry with
joint system models, Rock Mechanics and Rock Engineering, 21: 2151
Dershowitz, W.S. and Herda, H.H. (1992). Interpretation of fracture spacing and
intensity, Proceedings of the 33rd U.S. Symposium on Rock Mechanics, eds. Tillerson,
J. R., and Wawersik, W. R., Rotterdam, Balkema. pp. 757-766.
Dershowitz, W.S., Lee, G., Geier, J., Foxford, T., LaPointe, P., and Thomas, A. (1998).
FracMan, Interactive discrete feature data analysis, geometric modeling, and
exploration simulation, User documentation, version 2.6, Seattle, Washington: Golder
Associates Inc.
Dershowitz, W., Hermanson, J., Follin, S., and Mauldon, M. (2000). Fracture intensity
measures in 1-D, 2-D, and 3-D at Aspo, Sweden, Proceedings of the North American
Rock Mechanics Symposium (Pacific rocks 2000), v.4: 849-853.
Einstein, H.H. and Baecher, G.B. (1983). Probabilistic and statistical methods in
engineering geology, Rock Mechanics and Rock Engineering, 16: 39-72.
Engelder, T. (1993). Stress regimes in the lithosphere, Princeton University Press,
Princeton, NJ.

98

Engelder, T. and Gross, M.R. (1993). Curving cross joints and the lithospheric stress
field in eastern North America, Geology, 21: 817-820.
Engelder, T. and Fischer, M.P. (1996). Loading configurations and driving mechanisms
for joints based on the Griffith energy-balance concept, Tectonophysics, 256: 253-277.
Fouch, O. and Diebolt, J. (2004). Describing the Geometry of 3D Fracture Systems by
Correcting for Linear Sampling Bias, Mathematical Geology, 36(1): 33-63.
Gross, M.R. (1993). The origin and spacing of cross joints: examples from the Monterey
Formation, Santa Barbara Coastline, California, Journal of Structural Geology, 15(6):
737-751
Gross, M.R., Fischer, M.P., Engelder, T., and Greenfield, R.J. (1995). Factors
controlling joint spacing in interbedded sedimentary rocks: integrating numerical models
with field observations from the Monterey Formation, USA, In: Ameen, M.S. (ed.)
Fractography: Fracture topography as a tool in fracture mechanics and strain analysis.
Geological Society, London, Special Publications, 92: 215-233.
Helgeson, D.E. & Aydin, A. (1991). Characteristics of joint propagation across layer
interfaces in sedimentary rocks, Journal of Structural Geology, 13(8): 897-911.
ISRM, Commission on Standardization of Laboratory and Field Tests. (1978).
Suggested methods for the quantitative description of discontinuities in rock masses,
International Journal of Rock Mechanics and Mining Science, 15: 319-368
Kulatilake, P.H.S.W., Wu, T.H. (1984). The density of discontinuity traces in sampling
widows, International Journal of Rock Mechanics and Mining Science & Geomechanics
Abstracts, 21(6): 345-347.
LaPointe, P.R. and Hudson, J.A. (1985). Characterization and Interpretation of Rock
Mass Joint Patterns, Special Paper 199, Geological Society of America Book Series.
Martel, S.J. (1999). Analysis of fracture orientation data from boreholes,
Environmental and Engineering Geoscience, 5: 213-233.
Mauldon, M. (1994). Intersection probabilities of impersistent joints, International
Journal of Rock Mechanics and Mining Science & Geomechanics Abstracts, 31(2): 107115.
Mauldon, M. and Mauldon, J.G. (1997). Fracture sampling on a cylinder: from
scanlines to boreholes and tunnels, Rock Mechanics and Rock Engineering, 30: 129144.
Mauldon, M. (2000). Borehole estimates of fracture size, In Proceedings of the 4th
North America Rock Mechanics Symposium. Seattle, pp. 715-721

99

Mauldon, M., Dunne, W.M. and Rohrbaugh, M.B.Jr. (2001). Circular scanlines and
circular windows: new tools for characterizing the geometry of fracture traces, Journal
of Structural Geology, 23(3): 247-258
Mauldon M. and Wang, X. (2003). Measuring Fracture Intensity in Tunnels Using
Cycloidal Scanlines Proceedings of the 12th Panamerican Conference on Soil
Mechanics and Geotechnical Engineering and the 39th U.S. Rock Mechanics Symposium.
Mller-Salzburg, L. (1963). Der Felsbau, Bd.I, Theoretischer Teil, Felsbau ber Tage,
1. Teil. Enke, Stuttgart 1-624
Narr, W. (1996). Estimating average fracture spacing in subsurface rock, AAPG
Bulletin, 80(10): 1565-1586.
Ozkaya, S.I. (2003). Fracture length estimation from borehole image logs,
Mathematical Geology, 35(6): 737-753.
Owens, J.K., Miller, S.M., and DeHoff, R.T. (1994). Stereological Sampling and
Analysis for Characterizing Discontinuous Rock Masses, Proceedings of 13th
Conference on Ground Control in Mining. pp. 269-276.
Pascal, C., Angelier, J., Cacas, M.-C., and Hancock P.L. (1997). Distribution of joints:
probabilistic modelling and case study near Cardiff (Wales, U.K.), Journal of Structural
Geology, 19(10): 1273-1284
Price, N.J. (1966). Fault and joint development in brittle and semi-brittle rock,
Pergamon Press, New York.
Priest, S.D. and Hudson, J. (1976). Discontinuity spacing in rock, International Journal
of Rock Mechanics and Mining Science & Geomechanics Abstracts, 13: 135-148
Priest, S.D. (1993). Discontinuity Analysis for Rock Engineering, Chapman and Hall,
London.
Priest, S.D. (2004). Determination of Discontinuity Size Distributions from Scanline
Data, Rock Mechanics and Rock Engineering, 37(5): 347-368
Ruf, J.C., Rust, K.A., Engelder, T. (1998). Investigating the effect of mechanical
discontinuities on joint spacing, Tectonophysics 295: 245-257
Ruhland, M. (1973). Mthode d'tude de la fracturation naturelle des roches associe a
divers modeles structuraux, Sciences Gologiques Bulletin, 26(2-3): 91-113
Snow D.T. (1969). Anisotropic permeability of fractured media, Water Resources
Research, 5(6): 1273-1289.
Suppe, J. (1985). Principles of structural geology, Prentice-Hall, Inc., Englewood
Cliffs, New jersey.

100

Terzaghi, R.D. (1965). Sources of errors in joint surveys. Geotechnique. 15: 287-304.
Thapa, B.B., Ke, T.C., Goodman, R.E., Tanimoto, C., and Kishida, K. (1996).
Numerically simulated direct shear testing of in situ joint roughness profiles,
International Journal of Rock Mechanics and Mining Science & Geomechanics Abstracts,
33(1): 75-82
Wang, X., Mauldon, M., Dunne, W., Heiny C. (2004). Using Borehole Data to Estimate
Size and Aspect Ratio of Subsurface Fractures, In: Proceedings of the 6th North
American Rock Mechanics Symposium (NARMS), Houston, Texas
Wang, X., Mauldon, M., Dunne, W., Heiny C. (2005). Extracting fracture
characteristics from piercing-type intersections on borehole walls, In: Proceedings of
the 40th U.S. Symposium on Rock Mechanics (USRMS), Anchorage, Alaska
Warburton, P.M. (1980). A stereological interpretation of joint trace data,
International Journal of Rock Mechanics and Mining Science & Geomechanics Abstracts,
17(4): 181-190
Zhang, L. and Einstein, H.H. (1998). Estimating the Mean Trace Length of Rock
Discontinuities, Rock Mechanics and Rock Engineering, 31(4): 217-235
Zhang, L. and Einstein, H.H. (2000). Estimating the intensity of rock discontinuities,
International Journal of Rock Mechanics and Mining Sciences, 37(5): 819-837
Zhang, L., Einstein, H.H. and Dershowitz, W.S. (2002). Stereological relationship
between trace length and size distribution of elliptical discontinuities, Geotechnique,
52(6): 419433

101

This page intentionally left blank.

102

Chapter 5
5 Conclusions and discussions

Based on the stereological analyses and numerical simulation results of sampling


fractures by a line, a plane, and most important, cylindrical surfaces, the following
conclusions are drawn:
1. For linear sampling in constant sized or unbounded fractures with orientation
given by the Fisher distribution, the conversion factor C13 [1.0, ] between the
fracture linear intensity and the volumetric intensity is a function of the angle
between the sampling line and the Fisher mean pole, and the Fisher constant .
2. For planar sampling in constant sized or unbounded fractures with orientation
given by the Fisher distribution, the conversion factor C23 [1.0, ] between the
fracture areal (planar) intensity and the volumetric intensity is a function of the
angle between the normal of the sampling plane and the fracture Fisher mean pole,
and the Fisher constant .
3. For cylindrical surface sampling in constant sized or unbounded fractures with
orientation given by the Fisher distribution, the conversion factor C23,C [1.0, /2]
between the fracture areal (cylindrical surface) intensity and the volumetric
intensity is a function of the angle between the axis of the sampling cylinder
(borehole) and the fracture Fisher mean pole, and the Fisher constant .
4. For a general case of cylindrical surface sampling of fractures, the conversion
factor C23,C [1.0, /2] between the fracture areal (cylindrical surface) intensity and
the volumetric intensity is only a function of orientation of cylinder axis relative
to the fracture system and the pdf of fracture orientation weighted by area. It is
independent of fracture size or shape, or the sampling cylinder size.

103

5. Cycloidal scanlines, when deployed on the cylinder surface in a certain pattern,


give directional unbiased estimates of fracture volumetric intensity. Fracture
orientation information is not required by using this technique.
6. Fractures in sedimentary rocks can be approximated rectangular in shapes and the
estimators for their mean length and width are derived for three cases. The
estimators are independent of fracture size distributions.
The author also recommends the following work to be done in the future.
1. Interpretation of fracture traces to estimate other fracture properties, such as
roughness, connectivity, and so on, by means of stereology.
2. Study on fracture trace length distribution on borehole walls. It may be another
way to make estimates of fracture size, based on the assumptions of fracture
shapes.
3. Apply the conversion factors of fracture intensities and the estimators of fracture
size to real fracture trace data. Find ways to verify the obtained results.

104

This page intentionally left blank.

105

6 Appendix: Programs used in the dissertation

Two programs were used in this dissertation to carry out Monte Carlo simulations.

A. FISHER - Simulate the Fisher distribution


This program was developed by using Microsoft Excel. The inputs for this program are
listed in the table below.
Table App-1. Inputs for generating the Fisher distribution.
Input

Range

Fisher constant,

[0.1, 700]

Number of fracture poles, N

[0, 3000]

Fisher mean pole dip

[0, 90]

Fisher mean pole dip-direction

[0, 360]

To simulate a fracture normal given by the Fisher distribution, first we rotate the Fisher
mean pole to be upward (Fig. 2.A-1). A random number (between 0 and 1) is generated,
and by using the cdf of the Fisher distribution (Eq. (2.B-3)), angle , the angle between a
fracture normal and the Fisher mean pole, is calculated. This angle and another generated
random number between 0 and 360 define a unique orientation in the coordinate system
shown in Fig. 2.A-1. The dip and dip-direction of the simulated fracture are then
calculated from the paired angles by rotating the upward axis back to the Fisher mean
pole. An example of the simulated Fisher distributed fracture normals is shown in Fig.
2.4. This program was used to study linear and planar samplings of the Fisher distributed
fractures.

106

B. TRACE - Simulate fracture population sampled by a borehole


The program was developed in Visual C++. OpenGL was used to visualize the simulated
fracture population and the borehole in three-dimensional graphics (Figs. 3.8 and 4.11).
In this program, fractures are rectangular in shape and borehole is considered as a
cylinder. The parameters user may change are listed in Table App-2.
Table App-2. Parameters for simulating fractures sampled by a borehole.
Parameter

Range

Fracture length, f_l

>0

Fracture aspect ratio,

> 1.0

Fracture width, f_w

f_w = f_l

Fracture Dip, or dip of the Fisher mean pole

[0, 90]

Fracture dip-direction, or dip-direction of the Fisher


mean pole

[0, 360]

Fisher constant,

>0

Generation region shape

Box, Cylinder, Ball

Fracture volumetric intensity, P32

>0

Number of fractures, N

>0

Borehole (sampling cylinder) length, c_l

>0

Borehole (sampling cylinder) radius, c_r

>0

Borehole plunge

[0, 90]

Borehole trend

[0, 360]

User may fix the number of fractures to be generated or fix fracture volumetric intensity
and let the program calculate the number of fractures. The generation region, i.e., a box,

107

a cylinder, or a ball (Fig. App-1), is a region in which the centers of generated fractures
are located. The sampling region refers to the region that the sampling cylinder (borehole)
is within. The relationship between sampling region and generation region is showed in
Fig. App-1. Note that the maximum fracture dimension is: max_f_l =
the maximum sampling cylinder dimension is max_c_l =

max_c_l

f _ w2 + f _ l 2

c _ l 2 + 4c _ r 2 .

max_f_l
c_l

f_w

f_l
2 c_r
Largest fracture

Sampling cylinder

R_Cylinder

R_Ball

l_Cylinder

l_Box

Generation Box

Generation Cylinder

Generation Ball

Fig. App-1. The geometry of fracture, sampling cylinder, and three different shapes of
generation region.

108

Given dimensions of the largest fracture and sampling cylinder, the minimum size of
generation region varies with the shape of the region. The dimensions for different
shaped generation regions are listed in Table App-3.
Table App-3. Minimum dimension of different generation regions
Region shape

Length

Radius

Box

max_ c _ l + max_ f _ l

Cylinder

c _ l (max_ c _ l + max_ f _ l ) / max_ c _ l

c _ R + max_ f _ l

Ball

(max_ c _ l + max_ f _ l ) / 2

In three shapes of generation region, generation box is the simplest. The algorithm of
calculating fractures truncated by the region boundaries is also simple. Generation
cylinder is for the case that no rotation of the sampling cylinder is involved, while
generation ball allows rotation of the sampling cylinder.
After generating a set of synthetic fractures, the program computes the intersections
between the sampling cylinder and the rectangular fractures. Fracture traces are shown
on an unrolled trace map (Figs. 3.8 and 4.11). The outputs (in a text file) of the program
includes: fracture volumetric intensity (either set by the user, or calculated by the
program), fracture areal intensity on the borehole surface (calculated by the program by
dividing the total trace length by the cylinder area), and count of each intersection type.

109

This page intentionally left blank.

110

7 Xiaohai Wang
EDUCATION

Ph.D., Civil Engineering, Virginia Polytechnic Institute and State University, December
2005
Ph.D., Rock Mechanics & Rock Engineering, Institute of Soil & Rock Mechanics, Chinese
Academy of Sciences, September 1999
M.S., Mining Engineering, Taiyuan University of Technology, 1996
B.S., Mining Engineering, Taiyuan University of Technology, 1993
RESEARCH AND WORK EXPERIENCE
Research

Research assistant, Department of Civil & Environmental Engineering, Virginia Tech,


Blacksburg, VA
August 2001 September 2005
Characterization of rock fractures based on cylindrical samples, supported by National
Science Foundation
Scanline bias estimate and techniques to minimize the directional bias in cylindrical
sampling
Techniques to estimate fracture size and aspect ratio in sedimentary rocks
Computer simulation of rock mass fractures with three-dimensional visualization
Computer program to deploy unbiased scanlines on fracture trace maps and characterize
fractures
Research Fellow, Institute of Soil & Rock Mechanics, Chinese Academy of Sciences,
Wuhan, China
August 1998 May 2001
Integration of Three-dimensional Strata Information System (3DSIS), supported by
Chinese Academy of Sciences
Strata geological structure analysis and modeling
Research Assistant, Department of Mining Engineering, Shanxi Mining Institute, Taiyuan,
Shanxi, China
August 1994 July 1996
Fractals of distribution features of rock mass fissures, supported by Shanxi Natural
Science Foundation
Teach

Teaching Assistant, Department of Civil & Environmental Engineering, Virginia Tech,


Blacksburg, VA
May 2004 July 2004
Assist teaching the Intensive Summer Course in Geology Engineering & Rock
Mechanics for the US Army Corps of Engineers
Work

111

Programmer, Institute of Soil & Rock Mechanics, Chinese Academy of Sciences, Wuhan,
China
August 1999 May 2001
Develop the visualization module for program Three-dimensional Limit Equilibrium
Method in Slope Stability Analysis
Design and develop the program Supporting System in Deep Foundation Excavation
PUBLICATIONS

Wang, Xiaohai, M. Mauldon, W. Dunne. Estimating size and aspect ratio of rectangular
fractures from traces on cylindrical rock exposures. (for submission to Rock Mechanics &
Rock Engineering)
Mauldon, Matthew, X. Wang. Estimating fracture intensity from traces on cylindrical
exposures. (for submission to International Journal of Rock Mechanics & Mining Sciences)
Wang, Xiaohai, M. Mauldon, and W. S. Dershowitz. Multi-dimensional intensity measures
for Fisher-distributed fractures. (submitted to Mathematical Geology, May 2005)
Wang, Xiaohai, M. Mauldon, W. Dunne, C. Heiny. 2005. Extracting fracture characteristics
from piercing-type intersections on borehole walls. In: Proceedings of the 40th U.S.
Symposium on Rock Mechanics (USRMS) (2005), Anchorage, Alaska.
Wang, X., M. Mauldon, W. Dunne, C. Heiny. 2004. Using Borehole Data to Estimate Size
and Aspect Ratio of Subsurface Fractures. In: Proceedings of the 6th North American Rock
Mechanics Symposium (NARMS), Houston, Texas
Mauldon, M. and X. Wang. 2003. Measuring Fracture Intensity in Tunnels Using Cycloidal
Scanlines. In: Proceedings of the 12th Pan-Am. Conf on Soil Mech & Geotech Eng. & 39th
U.S. Rock Mech Symp. Soil & Rock America 2003, Culligan, Einstein & Whittle, eds.,
Boston: Vol. 1, 123-128
Jiang, Q., X. Wang, D. Feng, S. Feng. 2003. Three Dimensional Limit Equilibrium Analysis
System Software 3D_SLOPE for Slope Stability and its Application. Chinese Journal of
Rock Mechanics and Engineering. 22 (7): 1121-1125
Mauldon, M., X. Wang, D. Peacock. 2002. Fracture frequency predictions using doublecorrected data. In: Proc. of the 5th North American Rock Mechanics Symp. And the 17th
Tunnelling Association of Canada Conference: NARMS-TAC 2002, Hammah, R. et al. ed.,
Toronto, Canada: 27-34
Jiang, Q., M.R. Yeung, X. Wang, D. Feng. 2002. Development of the interactive
visualization system for three dimensional slope stability analysis. In: Proc. of the 9th
Congress of the International Association of Engineering Geology and the Environment,
Durban, September 16-20, 244-252
Zhao, Y., X. Wang, K. Duan, D. Yang. 2002. Unsymmetry of scale transformation of rock
mass anisotropy, Chinese Journal of Rock Mechanics and Engineering, Vol. 21. No. 11:
1594-1597
Zhang, Y., X. Wang, J. Chen, S. Bai. 2000. Application of 3D Volume Visualization in
Geology, Journal of Rock Mechanics & Engineering (in Chinese), Vol. 20. No. 5.
Wang, X., S. Bai. 1999. 3D Topological Grid Data Structure for Modeling Subsurface In:
Proc. of International Symposium on Spatial Data Quality (ISSDQ 1999), Hong Kong.

112

Wang, X., S. Bai, Z. Gu. 1998. The Problems in the Applications of GIS in Rock and Soil
Projects. Research and Practice in Rock and Soil Mechanics. Zhengzhou.
Wang, X., S. Bai. 1998. An Easily Integrated Three-dimensional Data Structure in Strata
Modeling. In: Proceedings of International Conference on Modeling Geographical and
Environmental Systems with Geographic Information Systems. Hong Kong
Zhao, Yangsheng, X. Wang, K. Duan. 1997. The Scale-invariability of the Distribution of
Rock Mass Fissures. Modern Mechanics and Technology Progressing. Beijing.
AWARDS / AFFILIATION

Graduate Research Development Project Grant, Virginia Polytechnic Institute & State
University, 2003-2004
Outstanding poster presenter, 20th Annual Graduate Student Assembly Research Symposium
& Exposition, Virginia Polytechnic Institute & State University, 2004
Best poster (tied), 6th North American Rock Mechanics Symposium (NARMS), June 6-10,
Houston, TX, 2004
Member of American Rock Mechanics Association, 2004, 2005

113

You might also like