See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/281742713

Determination of normal joint spacing from apparent joint spacing

measurements

Conference Paper · September 2013

DOI: 10.1201/b15794-99

CITATIONS READS
0 549

1 author:

Louis Ngai Yuen Wong

The University of Hong Kong
122 PUBLICATIONS   2,100 CITATIONS   

SEE PROFILE

All content following this page was uploaded by Louis Ngai Yuen Wong on 14 September 2015.

The user has requested enhancement of the downloaded file.

 Global View of Engineering Geology and the Environment – Wu & Qi (eds)
© 2013 Taylor & Francis Group, London, ISBN 978-1-138-00078-0

Determination of normal joint spacing from apparent joint spacing

measurements

Louis N.Y. Wong

School of Civil and Environmental Engineering, Nanyang Technological University, Singapore

ABSTRACT: Normal joint spacing is either measured directly from parallel joint members of the same
set or indirectly obtained from the apparent joint spacing measurements. The latter is achieved by con-
sidering the geometrical relationship between the joint orientation and the scanline/rock core orientation.
In common practice, one can compute either an individual weighting factor for each member in the joint
set, or only one weighting factor for the entire joint set based on the mean joint set orientation. These
methods yield the same normal joint spacing values if the joint members are perfectly parallel. However,
natural joints are rarely perfectly parallel, resulting in the discrepancy (error) of normal joint spacing. In
this paper, another means for determining the normal joint spacing from apparent joint spacing based
on the geometrical relationship between the scanline orientation and the averaged orientation of two
adjacent joint members of the same set is examined. A detailed analysis is conducted on different sizes of
randomly-generated joint set data with respect to differently oriented scanlines. The absolute error of the
normal joint spacing calculations obtained by proposed method, which is found to be always smaller than
that by the conventional practice, is thus recommended.

1 INTRODUCTION respectively. By defining the reciprocal of cosδ

as w, Eq. 1 can be expressed as
Joint set spacing (joint spacing for short) is com-
monly measured and presented in two ways. Sn = Sa/w (3)
Apparent joint spacings are readily obtained
during a scanline joint survey by recording the where w is the weighting factor originally developed
intersection distances of the joints on the scan- for balancing the orientation sampling bias intro-
line. The normal joint spacing is the perpendicular duced by linear sampling (Terzaghi, 1965, Priest,
distance between two adjacent joints belonging to 1993). In other words, the normal joint spacing is
the same joint set, which are parallel or almost obtained by dividing the apparent joint spacing by
parallel. the weighting factor.
If only apparent joint spacing measurements Although Eq.1 is expressed in a simple form,
are available, the normal joint spacing (Sn) can be complications arise due to its practical usage to
obtained from the apparent joint spacing (Sa) by
the following relation with reference to Figure 1
(ISRM, 1978, Giano, 1992). The rock joints are
assumed to be infinitely large. The joint size is not
taken into account.

Sn = Sαcosδ (1)

where δ is the acute inclination angle between the

joint normal and the scanline. The angle δ can be
found from the following expression

cosδ = |cos(αn – αs) cosβn cosβs + sinβn sinβs| (2)

In the above equation, αn and βn are the trend

and plunge of the downward directed end of the Figure 1. Relationship between apparent joint
normal to a given joint plane respectively, while spacing (Sa) and normal joint spacing (Sn) on a rock slope
αs and βs are the trend and plunge of the scanline exposure (modified from Giani, 1992).

615
determine the inclination angle δ, which is further
described below. Two approaches are commonly
used to compute the normal joint spacing from a
set of apparent joint spacing measurements.
In the first approach, the mean orientation of
the respective joint set is first identified from the
stereographic graphical method or by vectorial
calculation. The orientation (trend/plunge) of the
normal to the mean joint set orientation is then
substituted into Eq. 2 to obtain an inclination angle
δ. A normal joint spacing value is then computed
for each joint member according to Eq.1 based
on the same computed inclination angle δ and the
apparent spacing value associated with each joint
member. In the second approach, an inclination
angle δ is calculated for each joint member with
respect to the scanline, while without taking of
the remaining joint members in the same joint set
into consideration. A normal joint spacing value is
then computed for each joint member according
to Eq. 1 based on the individual inclination angle Figure 2. (a) Perfectly parallel joints belonging to the
δ and the apparent spacing value associated with same joint set, (b) Non-parallel joints belonging to the
each joint member. same joint set. Assume all joints are striking perpendicu-
These two methods yield the same normal joint lar to the paper.
spacing values if the joint members are perfectly
parallel. However, natural joints are rarely per-
fectly parallel, resulting in different normal joint Table 1. Apparent spacing and normal spacing
spacings obtained by these two methods. determination.
In this paper, another means for determining the Perfectly parallel Non-parallel joints
normal joint spacing from apparent joint spacing joints (Fig. 2a) (Fig. 2b)
measurement is proposed. This method is based
on the geometrical relationship between the scan- Between Between Between Between
line orientation and the averaged orientation of joint 1 joint 2 joint 1 joint 2
two adjacent joint members of the same set. This and and and and
approach closely resembles the practice of direct joint 2 joint 3 joint 2 joint 3
field measurement of joint normal spacing.
Apparent Sa11 = Sa12 Sa21 = Sa22 Sa11’ > Sa12’ Sa21’ < Sa22’
spacing
Normal Sn11 = Sn12 Sn21 = Sn22 Sn11’ > Sn12’ Sn21’ < Sn22’
2 PROPOSED METHOD FOR spacing
DETERMINING NORMAL JOINT
SPACING

As illustrated in Figure 2a, the normal joint spac- determined from Eqs. 1 and 2 with reference to
ings determined from the apparent joint spacing Figure 2b are thus not truly normal to the joint
measurements along scanline 1 and scanline 2 are planes.
equivalent, i.e. Sn11 = Sn12 and Sn21 = Sn22. It should also be noted that the apparent spac-
In nature, the members within the same joint set ing measurement and the subsequently determined
are unlikely to be perfectly parallel. As illustrated normal spacing values are dependent on the posi-
in Figure 2b, the orientation of joint 1, joint 3 and tion of the scanline (Table 1).
the scanlines are the same as those in Figure 2a, To alleviate the discrepancy of normal joint
expect for that of joint 2. Consider joint spacing spacing calculation due to the choice of the posi-
values along lines AB and CD. Due to the devi- tion and orientation of the survey scanline, instead
ated orientation of joint 2, even though lines AB of computing the normal joint spacing between
and CD are perpendicular to joint 1, they are not two neighboring joints based on the inclination
perpendicular to joint 2. In fact, it is impossible angle δ with respect to either one of the two joints,
to draw a line, which can be simultaneously per- computing the inclination angle δ with respect to
pendicular to both joint 1 and joint 2 (similarly the mean orientation of two neighboring joints is
for joint 2 and joint 3). The normal spacing values suggested. In other words, instead of substituting

616
the trend and plunge of the normal of one joint
member, the trend and plunge of the “averaged”
joint orientation are substituted in Eq. 2. Intui-
tively, this method resembles most closely the field
measurement of the normal spacing, during which
field engineers attempt to project a perpendicular
line to both neighboring joint planes. In this paper,
this proposed method is further investigated.

3 PARAMETRIC CASE STUDY

3.1 Methodology Figure 3. Automatic random joint set generation.

Different values of cone angle (5°, 10°, 15°) and joint
Two different methods (A and B) commonly used quantity (10, 50, 200) are examined in the present study.
for computing normal joint spacings based on the
field measurement of apparent joint spacings are
studied. These methods are described below. Table 2. Parameters used in the generation of
Method A—The inclination angle δ is com- random joints.
puted for a particular joint with reference to the
scanline orientation using Eq. 2, without taking Scanline number Trend (°) Plunge (°)
into account of the orientation of the neighboring
joints and the mean joint set orientation. The incli- 1 90 −45
nation angle δ hence obtained is used to compute 2 90 −30
the corresponding normal joint spacing associated 3 90 −15
with that particular joint based on the apparent 4 90 0
spacing. 5 75 0
Method B—A local mean joint orientation is 6 60 0
first obtained by averaging the orientation of two 7 45 0
8 30 0
neighboring joint members within the joint set. The
9 15 0
inclination angle δ is computed for this joint pair
using Eq. 2, which is the angle between the scanline
orientation and the local mean joint orientation.
The inclination angle δ hence obtained is used to model (Kent, 1982). Although the Fisher model is
compute the corresponding normal joint spacing very popular in modeling joint sets in geological
based on the apparent joint spacing associated and engineering studies (Priest, 1993, Kemeny &
with that particular joint pair. Post, 2003), Whitaker & Engelder (2005) pointed
To assess the influence of the computational out that any joint set with statistically greater
methods on normal joint spacing determination, variation in either the strike or dip direction
both methods A and B are used to compute the cannot be represented by the Fisher model.
normal joint spacings from the same set of joint The Kent model performs better in this case to
data for comparison. accommodate the elliptical dispersion of joint
To avoid sampling bias, randomly generated orientations about the mean (Peel et al., 2001,
joint data and the apparent joint spacing data are Whitaker & Engelder, 2005). However the Kent
used in the present study instead. Joint sets of vari- model requires computation of five parameters,
ous distribution properties and sample sizes are which are more than the two parameters required
generated by the JTdist function in the computer in the Fisher model. Since the objective of this
software DIPS (Rocscience, 2013) by specifying paper focuses on the methods of obtaining nor-
the following input parameters—Mean Dip, Mean mal joint spacing, only the data sets generated by
Dip Direction, Standard Deviation (Cone Angle), the JTdist are considered adequate. The conclu-
Joint Quantity, Dip Error and Dip Direction Error sions drawn from this study should however be
(Fig. 3). Due to the inherent randomness of the considered also applicable to the other types of
automatic joint set generation algorithm, the joint orientation distributions.
data generated from every run even with the same A probability distribution has to be assigned to
set of input parameters are not identical. provide a convenient means of describing the rela-
Various other types of orientation distribu- tive frequency of discontinuity spacings recorded
tion models are available in the literature, which along a scanline. Commonly used distributions
include the Fisher model (Fisher, 1953) and Kent include exponential (Priest & Hudson, 1976, 1981),

617
 Table 3. Calculation results of a particular set consisting of 10 joints (Scanline trend = 90 deg; Scanline plunge = −30 deg).

Orientation
Orientation of of normals
randomly to randomly Global mean method
generated generated (basis for comparison with
joints joints method A and method B) Method A Method B

Angle (δ) Angle (δ)

between between
normal to normal to
global Angle (δ) local mean
mean joint between Local Local joint
Apparent Dip orientation Weighting Normal joint normal Weighting Normal mean mean orientation Weighting Normal
spacing Dip direction Trend Plunge and scanline factor spacing and scanline factor spacing trend plunge and scanline factor spacing
Number (m) (deg) (deg) (deg) (deg) (deg) (1/cosδ) (m) (deg) (1/cosδ) (m) (deg) (deg) (deg) (1/cosδ) (m)

618
1 0.59 61 65 245 29 16.8 1.04 0.57 21.7 1.08 0.55 254.6 40.6 16.4 1.04 0.57
2 0.68 39 88 268 51 16.8 1.04 0.65 21.1 1.07 0.64 275.2 53.2 23.5 1.09 0.63
3 0.38 35 103 283 55 16.8 1.04 0.37 26.7 1.12 0.34 271.9 53.0 23.0 1.09 0.35
4 0.04 40 82 262 50 16.8 1.04 0.03 20.9 1.07 0.03 270.4 43.2 13.2 1.03 0.03
5 0.12 54 97 277 36 16.8 1.04 0.12 8.4 1.01 0.12 282.4 45.7 18.4 1.05 0.11
6 0.32 35 110 290 55 16.8 1.04 0.31 28.8 1.14 0.28 288.9 48.0 23.1 1.09 0.30
7 0.43 49 108 288 41 16.8 1.04 0.41 18.3 1.05 0.41 283.6 47.6 20.5 1.07 0.40
8 0.01 36 98 278 54 16.8 1.04 0.01 24.7 1.10 0.01 259.8 48.1 19.7 1.06 0.01
9 1.84 50 66 246 40 16.8 1.04 1.76 22.0 1.08 1.70 251.8 42.7 19.3 1.06 1.73
10 – 45 78 258 45 16.8 1.04 – 17.7 1.05 – – – – – –
Mean 0.49 88.3 43.3 268.3 46.7 0.469 0.454 0.460
Absolute error with respect to 3.30 2.07
basis (%)
log-normal (Sen & Kazi, 1984, Narr & Suppe, are repeated on the joint sets obtained from ten
1991), Weibull (Rouleau & Gale, 1985, Bardsley different runs of the JTdist function with the iden-
et al., 1990) and fractal type (Boadu & Long, 1994) tical set of input parameters. The rest of the input
distributions. In the present study, the apparent parameters for joint set generation are the same as
joint spacings are randomly generated according to those shown in Figure 3.
a negative exponential distribution, which is char- Two groups of results, which are obtained
acterized by the parameter λ. Although a mean based on the analysis of a Joint Quantity equal to
spacing value (1/λ = 0.5 m) is arbitrarily chosen in 10 and a Standard Deviation (Cone Angle) equal
the present study, the conclusions obtained from to 5°, are graphically shown in Figure 4a and 4b.
the present study are still valid irrespective of the
choice of λ since the normal joint spacings obtained
by different methods are eventually normalized for
comparison. Similarly, the conclusions drawn from
this study should also be considered applicable to
the other types of spacing distributions.
Three sets of random apparent joint spac-
ings consisting of 10, 50 and 200 joint data are
generated. Note that for a joint set consisting of
n joints, the number of apparent and normal joint
spacings are both n–1. In this paper, the effect
of the scanline orientation with respect to the
mean joint orientation on the normal joint spac-
ing determination is also studied. Nine scanlines
of different orientations (trend and plunge) are
assessed (Table 2).

4 RESULTS

In the analysis, a normal joint spacing is first com-

puted for each joint data by either method A or B
as described above. Within a particular joint set,
a mean normal joint spacing is then obtained by
averaging all the normal spacings for each compu-
tation method. Refer to Table 3 for the illustration
of the computations involving 10 joints (9 spacing
values). The mean normal joint spacings obtained
by methods A and B are then compared against
a common basis, which is obtained by the “global
mean method”.
The “global mean method” first involves com-
puting the mean joint set orientation, based on
which a single value of inclination angle δ is
computed with reference to the scanline orienta-
tion using Eq. 2. The common inclination angle
δ is then applied to the apparent spacing associ-
ated with each joint to compute the corresponding
normal joint spacing. A mean normal joint spac-
ing is later obtained by averaging all the normal
joint spacings within this joint set. The mean joint
spacing serves as a basis for comparison with those
obtained by the other two methods.
As discussed in Section 3.1, the automatic joint
generation in DIPS is associated with an inherent
randomness. The joint orientations generated from
different runs are not identical even with the same
set of input parameters. To assess this random Figure 4. Scattering of absolute error of normal joint
nature, analyses similar to that shown in Table 3 spacing values determined by (a) method A, (b) method B.

619
Figure 5. Influence of the computational methods A and B on the mean normal joint spacings.

The absolute error of the mean normal joint set contained in Figures 5 and 6, are summarized
spacings obtained by method A and method B below.
with respect to the basis are separately plotted
against the inclination angle δ. As illustrated in the • The absolute error with respect to the basis
plots, the degree of data scattering becomes more determined by method A is always larger than
significant as the inclination angle δ increases for that by method B for the same set of input data
both methods A and B. In order to minimize the (Fig. 5a–5i).
randomness effect of automatic joint set gen- • As the inclination angle δ between the joint nor-
eration on the analysis of normal joint spacings, mal and scanline increases, the absolute error
all the analyses by both methods A and B in the increases. The effect becomes particularly signif-
present studies are repeated individually ten times icant for δ larger than 60° for both methods A
similar to that shown in Figure 4a and 4b. The and B.
average absolute error value hence obtained is used • When the joint data become more scattered,
for comparison and assessment. i.e. the cone angle of the simulated joint set
The influences of calculation methods increases, the absolute error increases for both
(method A and method B), number of joints (10, methods A and B.
50, 200), cone angle of joint set (5, 10, 15) and • As the number of joints within a joint set
scanline orientation (Table 2) on the normal joint increases, while the cone angle of the joint
spacing calculation have been investigated. The set is maintained the same, the absolute error
results, which are graphically shown in the plots decreases (Fig. 6).

620
 study, however, show that for the same set of input
data, the absolute error of the normal joint spacing
calculations obtained by method B is always smaller
than that by method A. Calculating the normal
joint spacing based on the average orientation of
adjacent joint members (method B) is conceptually
similar to the normal joint spacing measurement
in the field, during which field engineers attempt
to project a perpendicular line to both neighboring
joint planes. Therefore, method B, which resembles
more closely the field practice of measurement
than method A does, is thus recommended to be
the preferred method for the normal joint spacing
calculation.
Assume that a joint set is characterized by a
specific degree of standard deviation (cone angle).
The larger the sampling population, the more reli-
able the normal joint spacing determination will
be. To ensure a large enough sample size, a suffi-
ciently long scanline is thus necessary, for exam-
ple ten times the estimated joint spacing (ISRM,
1978). Besides, the inclination angle δ between the
joint normal and the scanline should be as small as
possible as suggested from the present study.
If time and project budget permits, multiple
scanlines trending perpendicular to individual
joint sets are preferred to be surveyed. However,
if only one scanline is surveyed, it should be ori-
ented in such a way that its inclination angle δ with
the normals to multiple different joint sets is less
than 60°. Otherwise, highly erroneous normal joint
spacing values will result.
Figure 6. Influence of the number of simulated joints
(10, 50, 200) on the mean normal joint spacings.
REFERENCES

Bardsley, W.E., Major, T.J., Selby, M.J. 1990. Note on a

5 DISCUSSION AND CONCLUSIONS Weibull property for joint spacing analysis. International
Journal of Rock Mechanics and Mining Sciences &
In the present study, the mean normal joint Geomechanics Abstracts. Volume 11, No. 2: 133–134.
spacing of randomly-generated joint sets has Boadu, F.K., Long, L.T. 1994. The fractal character of
been determined by two different approaches fracture spacing and RQD. International Journal of
(methods A and B). The results are compared Rock Mechanics and Mining Sciences & Geomechan-
against a common basis, which is based on the ics Abstracts. Volume 31, No. 2: 127–134.
Fisher, R.A. 1953. Dispersion on a sphere. Proc. Roy.
global mean method. The normal joint spacings
Soc. London Ser. A. Volume 217: 295–305.
obtained by both methods are found to be close to Giani, G.P. 1992. Rock slope stability analysis, A.A
that obtained by the global mean method, when the Publishers.
cone angle of the randomly-generated joint set is ISRM. 1978. International society for rock mechanics
small, the number of joints is large, and the scanline commission on standardization of laboratory and field
is orientated close to the normals to the joints. tests: Suggested methods for the quantitative descrip-
Computing the normal joint spacing value from tion of discontinuities in rock masses. International
the apparent joint spacing value by method A is Journal of Rock Mechanics and Mining Sciences &
found to be more popular in practice due to is sim- Geomechanics Abstracts. Volume 15: 319–68.
ilarity of applying the weighting factor to a par- Kemeny, J. & Post, N. 2003. Estimating three-dimensional
rock discontinuity orientation from digital images of
ticular joint member. Since the latter has become fracture traces. Computers & Geosciences. Volume 29:
a standard practice in manipulating the joint map- 65–77.
ping data, method A is conveniently adopted for Kent, J.T. 1982. The Fisher–Bingham distribution on
calculating the inclination angle δ for each joint, the sphere. Journal of the Royal Statistical Society,
instead of a joint pair. The results in the present Series B. Volume 44: 71–80.

621
 Narr, W. & Suppe, J. 1991. Joint spacing in sedimentary Rocscience-DIPS v5.0—Features, Rocscience Inc. see
rocks. Journal of Structural Geology. Volume 13: website at http://download.rocscience.com/products/
1037–1048. (1991). dips/DipsUtilities.asp; (2013).
Peel, D., Whiten, W.J. & McLachlan, G.J. 2001. Fitting Rouleau, A. & Gale, J.E. 1985. Statistical characterization
mixtures of Kent distributions to aid in joint set of the fracture system in the stripa granite, Sweden.
identification. Journal of the American Statistical International Journal of Rock Mechanics and Min-
Association. Volume 96: 56–63. ing Sciences & Geomechanics Abstracts. Volume 22,
Priest, S.D. 1993. Discontinuity analysis for rock No 6: 353–367.
engineering. Champman & Hall. London. Sen, Z. & Kazi, A. 1984. Discontinuity spacing and RQD
Priest, S.D. & Hudson, J.A. 1976. Discontinuity spacings estimates from finite length scanlines. International
in rock. International Journal of Rock Mechanics and Journal of Rock Mechanics and Mining Sciences &
Mining Sciences & Geomechanics Abstracts. Volume Geomechanics Abstracts. Volume 21 (4): 203–212.
13: 135–48. Terzaghi R.D. 1965. Sources of error in joint surveys.
Priest S.D. & Hudson J.A. 1981. Estimation of disconti- Geotechnique. Volume 15: 287–304.
nuity spacing and trace length using scanline surveys. Whitaker, A.E. & Engelder, T. 2005. Characterizing stress
International Journal of Rock Mechanics and Mining fields in the upper crust using joint orientation distri-
Sciences & Geomechanics Abstracts. Volume 18: butions. Journal of Structural Geology. Volume 27:
183–197. 1778–1787.

622

View publication stats