8

The Collection and Analysis of

Discontinuity Orientation Data for
Engineering Design, with Examples
STEPHEN D. PRIEST
University of South Australia, Pooraka, Australia; Formerly University of
Adelaide, South Australia, Australia
t. I NTRODUCTI ON
i~ DATA COL L E CTI ON
83.1 The Scanline Technique
822 Preliminary Data Processing
t_3 GRAPHI CAL RE PRE S E NTATI ON OF ORIE NTATION DATA
Example 1
14 V E CTORI AL RE PRE S E NTATI ON OF ORIE NTATION DATA
Example 2
\i ORIE NTATION SAMPL ING BIAS DUE TO A L INE AR SURVE Y
Example 3
Example 4
ti . I DE NTI FYI NG AND DE L I MI TI NG SETS
Example 5
S." RE PRE S E NTATI V E ORI E NTATI ON FOR A S E T
Example 6
| THE FI S HE R DISTRIBUTION
Example 7
19' SUMMARY AND CONCL US I ONS
0 RE FE RE NCE S
16$
169
172
173
174
176
1.77
177
178
180
181
182
184
185
187
189
190
191
1 I NTRODUCTI ON
Rock masses usually contain such features as fractures, faults, joints, bedding planes, cleavageand
r mechanical defects which are herecollectively referred to as discontinuities. Thesediscontinuities
i haveamajor influenceon thedeformability, strength, stability and permeability of therock mass
thereby play a major part in dictating the design of foundations, slopes, tunnels and other
ctures associated with discontinuous rock masses. Oneof theaims of this chapter is to present alist
lily accessiblereferences [1-62] which providesupplementary reading and further material for
concerned with the collection and analysis of discontinuity orientation data. The greater
rtion of thesereferences comefromthemost activeperiod of research in this area: 19701985.
167
Discontinuities
The discontinuity properties that have the greatest influence at the design stage have been 1
by Piteau [ 1, 2] as follows: (i) orientation; (ii) size; (iii) frequency; (iv) surface geometry; (v) ge
type; and (vi) infill material.
Of thesePiteau concluded that discontinuity orientation with respect to the engineering stru
is the most important. This conclusion is reflected in the fact that all the currently accepted dcsfa
methods for foundations, slopes and underground excavations in discontinuous rocks reeaaa
information on discontinuity orientation in oneformor another [3-9]. An unfavourably oriental*1
discontinuity, or group of discontinuities, adjacent to a rock face subject to low stress levels
causerigid block failures involving sliding, toppling or falling mechanisms, or a combination of
[10-12]. Discontinuities in zones of high stress adjacent to an underground excavation :
orientated in an unfavorable way, provide planes for shear failure and displacement [3. _
Discontinuity networks can, depending on the orientation of individual open fractures, prow*
paths of high permeability through otherwise relatively impermeable rock materia! [14, 151. 1
chapter is primarily concerned with the measurement, analysis and presentation of discon
orientation data in aformsuitable for input to design calculations to tackle the problems listed
A discontinuity is here regarded as afinite nonplanar surface, which can be represented by a
of points in three-dimensional space. Although discontinuities often have an irregular or c
geometry, there is usually a scale at which the whole surface, or a portion of it, is sufficiently ptaac
to be represented by a single orientation value. Field measurements of orientation are usually takes
using a simple compass-clinometer device[16]. Such a device is designed lo take angular measure-
ments of the orientation of a line or plane in three-dimensional spaceby reference to magnetic Dona
and the horizontal plane. Before proceeding with a discussion of discontinuity orientatioi
necessary to set out some fundamental definitions concerning the orientation of lines in three-
dimensional space.
Plunge, p ( - 90 <p <90). This is the acute angle measured in a vertical plane between a pva
line and the horizontal plane. A line directed beiow the horizontal is here described as ban*]
a downward senseand is taken to have a positive plunge; a line directed upwards is described m
having an upward senseand is taken to have a negative plunge. The downward-directed i.posiswj
value of plunge will always be taken for lines, such as the normal to a plane, that require aa
specification of direction.
Trend, a (Qe <a <360e). This is the geographical azimuth, mea.41j.red in clockwise rotation bem
north (0), of the vertical plane containing the given line. Any vertical plane possesses ma
geographical azimuth directions, 180 apart; trend is the azimuth that corresponds to the direcooa
of plunge of the line.
The orientation of any line can be recorded unambiguously in terms of its trend a and plunget
theformof a threedigit and a two digit number separated by a slash. For example 268/31 refers
a line plunging downwards at 31' towards 268". A plane can be regarded as an infinite number d
coplanar lines radiatingfroman arbitrary point. The line of maximumdip of a nonhorizontai piaae
is the imaginary line whose plunge exceeds that of all other lines in the plane. The trend ad and tie
downward plunge /3d of the line of maximumdip of a plane are here referred to as the dip directie
and dip angle of the plane.
Discontinuity orientation data can be presented and utilized in design by two different methods. W
the rock face at the design site is readily accessible, the locations and orientations of acraal
discontinuities can be measured and used explicitly in the design calculations. If, however, the rock,
face is not accessible it is necessary to measure discontinuity orientation at other rock faces, or bam
boreholes, and build up a statistical model that represents the discontinuity orientation charac**-
istics of the rock mass. I n most cases the statistical model is based on the fact that geological
processes usually generate one or more clusters (or sets) of nearly parallel discontinuities in a grvea
rock mass.
This chapter is primarily concerned with the latter of the above methods. The first pan of tie
chapter deals with dtscotvtmutt^samipVmj. methods, a.tvd the ^TOphkaX Mvd Nectot\a.\.
of orientation data. A new method for eliminating sampling bias will be presented and incorporate!
into an algorithmfor identifying clusters of preferred orientation. A vectorial method for detenaiB-
ing representative orientations for setswill be introduced and linked lo a statistical analysis of set
data based on the Fisher distribution. Finally a new Turbo Pascal computer programCANDO.
based on the methods explained in this chapter, for rapid analysis of large volumes of discontinue?
orientation datawill be presented. One of the main features of this chapter is the use of examples tc
illustrate the data processing methods. The aimof theseexamples is to illustrate principles rather
than to simulate an actual data processing exercise. For this reason, and to save space, data volumes
have been kept to an absolute minimum. Numerical results in theseexamples are presented to more
The Collection and Analysis oj Discontinuity Orientation Data for Engineering Design 169
significant figures than are warranted by the nature of the data and the precision of the measurement
techniques. These extra significant figures are provided to allow the reader to work through the
calculations by hand and to check intermediate results without being troubled by accumulated
round-off errors.
8.2 DATA C O L L E C T I O N
I n order to understand and quantify the influence that discontinuities have on rock mass behavior
it is first necessary to measure and to represent quantitatively the relevant characteristics of the
discontinuities that form the complex three-dimensional fabric of mechanical breaks, termed the
rock structure, in the mass. Discontinuity orientation, and other properties such as size, frequency
and surface geometry are essentially geometrical properties that must be measured physically in
some way. There are two general sampling regimes that can be adopted.
(i) Borehole sampling. The recovery, examination and testing of high quality drill core has been
used for many years to probe rock conditions at depth [13]. Borehole sampling can provide
a relatively undisturbed sample of core from deep within the rock close to the area of interest. I n
addition mechanical and fluid flow tests can be carried out on the recovered core and on the walls of
the borehole [14]. The borehole walls can also be inspected by remote visual techniques, and the
surrounding rock can be probed by geophysical devices. One significant problemwith borehole
sampling is tiiat the core can rotate during extraction, so special sampling and analysis techniques
are needed to determine the true orientation of the sampled discontinuities within the rock mass
[13]. A second difficulty is that the core is usually of small diameter (<100 mm) so it is impossible to
measure discontinuity size. High equipment hire charges, the costs of skilled operatives and (he
desirability of professional supervision can make borehole sampling a relatively expensive opera-
tion. I nteresting papers by Andersson [17] and by Karzulovic and Goodman [18] have addressed
some of the problems of sampling discontinuity characteristics from boreholes.
(ii) Measurement at exposed rock faces. Measurement at exposed rock faces, either above or
below ground, has the advantage of utilizing a relatively large area of rock, which enables the direct
measurement of discontinuity orientation, size and other large-scale geometrical features. The
geological relations between the various discontinuity groups can also be observed. One disadvan-
tage of this approach is that the rock face is often remotefrom the zone of interest and may suffer
from blasting damage or degradation by weathering and vegetation cover. Although skilled
personnel are needed to take the measurements, equipment and labor costs are negligible compared
with borehole sampling costs. Unti l recently, measurements at exposed faces were taken in an
arbitrary, subjective way derived fromgeological mapping techniques; this yielded little quantitative
data of value for engineering design. Recently more rigorous statistical sampling methods have been
adopted. The most widely used of these methods is the scanline technique. This technique has been
described and discussed by a number of authors [16, 19-21, 25], The relative simplicity of the
measurement process at exposed faces, and the statistical rigor of the scanline technique, make this
method ideally suited to the determination of discontinuity orientation, and other large-scale
geometrical properties of the rock structure. Thefollowing section contains a brief summary of the
principal features of the scanline technique.
&2.1 The Scanline Technique
I t is important to recognize that there is no universally accepted standard for scanline sampling.
I ndeed it is desirable to modify the details of the technique to suit local rock conditions and to
provide specific data that may be required for a particular design exercise. The methods summarized
below are based on the author's experience of some 20 years of discontinuity sampling, involving the
measurement of many thousands of discontinuities [9, 11, 22-26],
A clean, approximately planar rock face is selected that is large relative to the size and spacing of
the discontinuities exposed. Such exposures can be found on beach cliffs, in gorges, road cuttings,
quarries, open pit mines and unsupported adits. I ntersections between discontinuities and the rock
iace will produce linear traces which provide an essentially two-dimensional sample of the rock
structure. The scanlines themselves are simply measuring tapes, between 2 and 30 mlong, pinned to
the rock face in both horizontal and vertical directions. I deally, further scanlines should be set up on
a -second rock face, at right angles to the first, to provide a three-dimensional picture of the rock
;structure. The aimhere is to impose a rigorous linear sampling regime that is similar to that of
170 Discontinuities
a borehole but which does not suffer from the small sample width. The location, orientation and
condition of the rock face are recorded on thefirst page of the logging sheet together with the tread
and plunge of each scanline. Color photographs of the face, including a scale and appropriate iabeL
may also be taken before commencing the "sampling process [27]. Each scanline tape is scanned,
startingfrom the zero end. until a discontinuity trace is intersected. Thefollowing properties of ooK
those discontinuities intersected by the tape are measured and recorded in a systematic tabular forat
on a logging sheet such as the one illustrated in Figure 1.
(i) Intersection distance d. This is the distance along the scanline to the intersection point nsk
the discontinuity. I f the face is not perfectly planar it may be necessary to extrapolate some of Ike
discontinuities to obtain this point. The intersection distance on a given scanline provides a sircrie
and unambiguous method for identifying any particular discontinuity in the sample. Ksconti ci aw
spacings can be obtained by subtracting consecutive intersection distances for a specified group of
discontinuities [24, 28], Scanline discontinuity frequency, which is the average number of interjec-
tions per unit length of scanline, is usually calculated separately for each identifiable set and the*
processed to obtain the expected frequency along the normal to each set [16. 18. 26, 29].
(iij Orientation. This is thefive digit dip direction/dip angle of the discontinuity measured a: tae
point of intersection with the scanline. If the discontinuity is poorly exposed at this point it may be
necessary to measure the orientation at an exposed surface on the discontinuity some distance free
the scanline. The orientation is usually measured by means of a magnetic compass and clinomeasr
device fitted with a spirit level [16, 30]. Ewan and West [31] concluded that different operauxs
measuring the orientation of the same feature have a maximum error 10 for dip direction and
5C for dip angle. Care must be taken to correct the readings of geographical azimuth for k ni
differences between true and magnetic north. This correction can often be achieved during Ac
measurement processitself by means of a minor adjustment to the compass. If this adjustment is aet
possible it will be necessary to correct the azimuth readings during data processing and plorcaaj.
Compass needles balanced for magnetic inclination in the northern hemispherewi ll be severely oar
of balance in the southern hemisphere. It should be noted that the scanline will tend to intersect
University ct Adelaide - sconline logging form
Page
Date
Face location Rock type
Scanline label
Foce, dip direction /
dip ongie /
Excavation met nod
Measurements by Non -overhanging /overhanging Condition of exposure
Scanline, trend/plunge /
Face height Comments
Semi-traces left/right/above/
below
Figure 1 Example of a scanline survey logging form
The Collection and Analysis of Discontinuity Orientation Data for Engineering Design 171
preferentially those discontinuity traces that make a large angle with it, and in thelimit, the scanline
will fail to intersect traces that lie parallel toit. This effect, which is discussed in detail in Section 8.5,
can be allowed for by applying simple trigonometrical correction factors to the data before plotting
and analysis.
(Hi) Semi-trace length I. This is the distance fromthe intersection point on the scanline to the end
of the discontinuity trace. This distance can cither be measured directly if the face is accessible,
estimated by eye or scaled fromthe photograph of the rock face. Therewill be two semi-trace lengths
associated with each discontinuity: one above and one below a horizontal scanline; one to the left
and one to the right of a vertical scanline. In many cases the scanline will have been set up close to
the edge of the exposure so it will only be possible to measure semi-trace lengths on one side of it. This
sampling systempresents no problembecause the intersection point will be at a randomlocation
along each discontinuity trace, permitting the use of simple statistical techniques 10 estimate the
properties of the complete trace length distribution [20, 32-37], Some operators choose to ignore
semitraces that are smaller than a given threshold value lt. If no record is kept of the number of
traces that are ignored, this practice has the effect of truncating the semi-trace length sample for
values less than k- If, on the other hand, a record is kept of the number of traces that are ignored, the
semi-trace length sample is said to be censored for values less than lt [38-41]. The different effects of
truncation and censoring must be allowed for during statistical processing. During this statistical
analysis allowance can also be made for the fact that the scanline will tend to have intersected
preferentially discontinuities with longer traces [20,25]. Although it can be argued that these longer
traces are more significant from an engineering point of view and should, therefore, feature more
prominently in the sample, it is not satisfactory to replace engineering judgement by this arbitrary
sampling bias. In the author's opinion it is preferable to remove all sampling bias and then to
examine the true rock structure data, taking into account not only discontinuity size but other
properties such as orientation, frequency and strength. A number of papers have been written
describing ways of removing sampling bias, and ways of determining the distributional formand the
distribution parameters of discontinuity size measurements taken at planar rock faces of limited
extent [20, 25, 32, 35, 37, 42-46]. Discussion of these techniques, which comprise a variety of
ingenious algorithms for processing truncated and censored samples, is beyond the scope of this
chapter.
(iv) Termination. It can be helpful to keep a record of the nature of the termination of each
semi-trace. The following recording scheme has proved to be adequate: A =discontinuity trace
terminates at another discontinuity; I =terminates mintact rock material; and O termination is
obscured. The larger the proportion of discontinuities that terminate at other discontinuities the
greater the tendency for the rock mass to be blocky, relatively weak, of low stiffness and highly
permeable. A trace can be obscured by scree, vegetation or by extending beyond the limits of the
exposure. Obscuring in this way has the effect of censoring the sample of semi-trace lengths for values
larger than a variable length that depends upon the geometry of the rock face, the location of the
discontinuity within the face and the orientation of the discontinuity. This complex censoring
problemhas been examined by Baecher [32], Kulatilake and Wu [37, 42, 47], Laslett [43], Pahl
[44], Priest and Hudson [25], and Warburton [45, 46] who present a number of techniques for
determining the distribution formand distribution parameters of the complete trace lengths. One of
the approaches adopted by these authors has been to adopt an areal sampling system, rather than
the traditional line sampling'techniques.
(v) Roughness. Surface irregularities with a wavelength less than about 100 mmare here referred
to as roughness. Roughness can be measured by taking an impression of the surface, then digitizing
and quantifying representative profiles. For most practical purposes, however, it is sufficient to
assess roughness visually on a five point scale ranging from1 =smooth to 5 =very rough, based on
experience at a given site. The results can then be correlated with shear strength parameters for each
discontinuity type on the site, or alternatively analyzed following the methods recommended by the
ISRM [16],
(vi) Curvature. Surface irregularities with a wavelength greater than about 100 mmare here
referred to as curvature. Curvature can be determined by measuring offsets at 100 mmintervals
along a straight base line then digitizing and quantifying the resulting profile. As with roughness,
however, it is often sufficient to assess curvature visually on a five point scale ranging from
1 =planar to 5 =very curved.
(mi) Comments. The comments column is used to provide additional qualitative information
about each discontinuity in abbreviated form. The following information may be recorded.
(a) Type of discontinuity, i.e. whether it is a joint, bedding plane, fault or blasting crack [48]. I t
may be possible to recognize and ignore blasting cracks since they are usually small, rough, clean.
172
Discontinuities
randomly orientated and associated with blast holes. If there is any doubt it is best to take
measurements and record any appropriate comments.
(b) Nature of infill, i.e. whether the discontinuity is clean or is infilled with clay or
deposits.
(c) Aperture. The aperture, or opening, of a discontinuity can be estimated with feeler gauge*
measured by other more sophisticated techniques [15], It should be noted, however, that disco"
ity apertures are highly sensitive to blast vibrations, block movements and local weathering
(dj Water flow. I t can be helpful to note those discontinuities that show signs of water fio*
seepage.
(e) Slickensides. These scratch marks on the surface of discontinuities indicate the directkx:
previous shear displacement. I f they are present their orientation should be recorded.
(f) Broken rock. There is little point in trying to measure discontinuities in broken or hea
fractured rock. It is better simply to record the extent of the broken zone in terms of distance
the scanline.
(g) Foliation. Closely spaced parallel discontinuities with a spacing of less than about 5 mm at
be produced by bedding, slaty cleavage or schistosity. It is rarely necessary, or feasible, to rotH I T
each feature. I t is better to take representative measurements and then to estimate the frequency a
the features at 0.5 to 1 mintervals along the scanline.
(h) Uncertainty. I n some cases it may be difficult to decide whether a given feature is a a w
discontinuity or not. For most applications only those features that form true mechanical breaks of
geological origin should be measured. Any uncertainty about the origin of a particular feature
should be recorded in the comments column. I n some applications, for example blasting studies, i
may be desirable to record separately those fractures induced by the blasting process.
Before proceeding with a detailed description of the presentation and analysis of orientation cua
it is worth considering thefollowing two conflicting principles: there is little point in collecting and
processing data that cannot be used as input for design calculations; conversely it is much easie- lo
ignore superfluous data than to return to a site and collect additional data. Experience has shows
that between 1000 and 2000 discontinuity measurements of the type listed in this section are
generally sufficient to provide a preliminary characterization of a typical site [24]. Addition*!
measurements are required where a site exhibits highly variable discontinuity characteristics or
where a higher degree of confidence in the engineering design is required.
8.2.2 Preliminary Data Processing
Field measurements of discontinuity orientation are usually recorded in terms of the dip directsoc
and dip angle of a plane that lies parallel to a portion of the discontinuity. For the purpose of
discontinuity orientation analysis, however, it is generally preferable to refer instead to the trend s.
and the plunge/? of the normal to the plane. Unlike vectors such as force and displacement, the lines
referred to above do not necessarily require a prescribed direction. For example, an upward directed
normal with a trend/plunge 308/ - 35 refers to the same plane as a downward directed normal of
trend/plunge 128/35. I n view of this, it is convenient to work consistently with only the downward
directed end of a given orientated line. This approach ensures that all angles of plunge are positne
and makes it possible to write down thefollowing expressions linking ad / / ?d and a//? for a give
plane
If a <180 then a = a +180
If a >180" then a = a - 180
fi. = 90 -ft at
An important exception to this approach occurs when determining the representative orientation for
a set, as explained in Section 8.7. The data collected during a scanline survey require considerable
further processing before they can be used in engineering design. Much of this processing requires
relatively advanced statistical techniques that are dealt with later in this chapter and in the next three
chapters of this volume. Software is now available to allow the processing and graphical presema'jce
of the data to be carried out on mainframe and personal computers. Most of the available programs,
such as the programCANDO referred to in this chapter, require orientation values and other
discontinuity data to be stored in a separate datafile for access by the main program. I n view of this,
it is desirable to type the relevant data into an appropriate file by means of a convenient word
The Collection and Analysis of Discontinuity Orientation Data for Engineering Design 173
processor systemas soon as the preliminary processing referred to abovehas been complefed. The
particular programCANDO has been designed to accept up to 3000 orientation values sampled
along up to 100 separatescanlines. Thesedata are read froma file specified by the user at the start of
the analysis session.
Z3 G R A P H I C A L R E P R E S E N T A T I O N O F O R I E N T A T I O N DATA
Theaimof collecting discontinuity orientation data is to delect consistent patterns (or the absence
of them) and to predict their influenceon theengineering performanceof therock mass. A graphical
representation of theorientation data can help i n achieving this aim, not only by presenting the raw
data but also by displaying the results of any statistical analyses.
One of thesimplest forms of graphical representation is the rosediagram[19, 49]. This approach
is particularly suited to cases wheremost of thediscontinuities haveangles of dip in excess of about
60. I n such casesthe dip direction data are of primary significance and can beplotted on a simple
circular histogram. The0 to 360 circleis divided into convenient class intervals, usually of 5, 10 or
15 depending upon the samplesize, and thenumber of dip direction values in each class interval is
counted. The results are plotted as wedges which havea radial extent that is proportional to the
frequency in each class interval. This frequency can be expressed either as a raw number, as
a proportion or as a percentageof the total samplesize. The rosediagramin Figure 2 based upon
a sampleof 226 discontinuities obtained by Priest [22], shows how theorientation and the relative
significance of clusters of preferred dip direction areclearly visible. In the author's opinion thereis
little point in plotting rosediagrams based upon strikedirection sincethis produces a duplication of
information across the rosediagramand an associated loss of dip direction data.
The disadvantageof rosediagrams is that they contain no information on dip angle. This can be
overcometo someextent by selecting data fromthemoresignificant class intervals and then plotting
a histogramof dip angles. Thehistogramin Figure3, based upon 52 valuesfromFigure2 with a dip
direction in the range320 to 020, shows that 33 of thediscontinuities in this rangehavean angleof
dip greater than 60; only one discontinuity has an angleof dip less than 30. I n this case, then, the
rosediagram provides a reasonablerepresentation of discontinuity orientation.
Thedifficulties of representing three-dimensional orientation data in two dimensions on a sheet of
paper can be overcome by adopting the technique known as stereographic, or hemispherical,
projection [4, 5,23, 50-52]. The basic principleof hemispherical projection is that theorientation of
a linein three-dimensional spaceis uniquely represented by the position of a point within a circular
projection area of radius R. The perimeter of this projection area is associated with the circleof
geographical azimuth directions so that the radius on which the point plots is given directly by the
trend direction, a, of the linethat it represents. I t is usual to mark thenorth (azimuth 0) direction
with a small linelabeled 'N' to providea referencedirection on theprojection. This convention will
t
ieo*
Figure 2 Rose diagram for discontinuity dip direction, Chinnor, Location 2, 226 values
174
Discontinuities
Dip direction 320 - 020
52 values
10 -
i
0 10 20 30 40
60
70
ec
90
Dip angle (i
Figure 3 Histogram for discontinuity dip angle, Chinnoi. Location 2, dip direction 320-020t
be adopled here. The radial distance, r, of the point from the center of the projection circle u
afunction of the plunge, /S, of the line that it represents, subject to the boundary conditions that r =I
when P = 90 and r = R whenfi =0. Theform of the functional relation between r and fi depends
upon the projection method adopted. Two of the most common projection methods utilize as
imaginary hemisphere of radius R positioned below the plane of projection so that its circular face
forms the projection circle. The first of these methods, termed the lower hemisphere equal arigk
projection, gives rise to thefollowing relation between r and ji
The second method, termed the lower hemisphere equal area projection, gives the following
relation
A full discussion of the properties and the relative merits of these two methods of projection s
given by Priest [23].
Example 1 (Figure 4)
Plot on a lower hemisphere equal angle projection of 100 mm diameter, points repres-
enting the line of maximum dip and the normal to a plane of dip direction/dip angle
The trend and plunge of the line of maximum dip a*, (jd are given directly by the dip direction and
dip angle of the plane. Hence ad =146 and /Jd =57. The trend and plunge of the normal a, fl, are,
by equations (1) and (2), respectively 326 and 33. The point D, representing the line of maximum
dip, and the point N, representing the normal, plot along radii of azimuths 146 and 326*
respectively on the 100 mm diameter projection circle in Figure 4. SettingR =50 mm in equation in-
puts the point D 14.8 mm from the center, and the point N 27.1 mm from the center of the
projection.
I f a large number of data points are to be plotted it is usually convenient to adopt computer
graphics in preference to manual plotting methods. Figure 5(a) shows a circular projection area of
radius R related to anx, y Cartesian coordinate system in which po'sitive .x is horizontal to the north
(trend 000) and positivey is horizontal to the east (trend 090). The x, y Cartesian coordinates of
a point on the projection representing a line of trend/plungea/B are given in Table1 for equal angle
and equal area projections. The x, y coordinates for the points D and N in Example 1, determined
from the top row of Table 1, are given in Figure 4. I t is a relatively straightforward matter to wine
a computer program that utilizes the expressions in Table 1 to enable automatic plotting of dau.
Large numbers of discontinuity orientation values can be plotted on a lower hemisphere
projection to give a graphical representation of the orientation properties of the rock structure. Toe
146/157.
The Collection and Analysis of Discontinuity Orientation Data for Engineering Design
175
/ N
/r-S2.5mm\.27.l mm
/ y -15.2 mm\
Figure 4 Lower hemisphere equal angle projection of the line of maximum dip, D, and the normal. N, of a plane of dip
direction/dip angle 146/57 (Example 1)
y Eos*
x Horfrontal north
y Horizontal east
7 Vertical down
Figure 5 (a) Two-dimensional Cartesian coordinate system, (b) Three-dimensional Cartesian coordinate system
Table 1 The x, y Cartesian Coordinates of a Point Representing a Line of Trend/Plunge
a//? on a Lower Hemisphere Projection of Radius R
x coordinate {north 000] y coordinate {east 090)
Equal angle R cos a tan
Equal area R^/lcosac
/90
/90 - fi\
R sin a tan I 1
IT . fW + 0
R J 2 sin a cos I
best approach is to plot only thepoints representing thenormals to each discontinuity plane. It is
often feasible to usesymbols of different shapes to represent different types of discontinuities (joints.,
bedding planes, faults, etc.) and to usesymbols of different sizes to represent a rangeof discontinuity
sizes. An exampleof such a plot, from Priest [23] is reproduced in Figure6. Plots of this typepermit
the identification of groups of subparallel discontinuities, or 'sets', that havebeen produced by
a singlegeological process [48]. The orientation of thesesets, and the degreeof clustering within
each set, can havea major influence on the engineering performance of the rock mass. It can be
beneficial, therefore, to apply quantitativemethods in theanalysis of discontinuity orientation data,
to providean objective assessment of any preferred orientation.
176 Discontinuities
Trace length
< 2 nn 2-4 m >4m
joint
bedding
O
/
Sampling line I
O O QC O
3 c. Qoo0 o
Samplingline I
0=
Figure 6 Lower hemisphere, equa! angle plot of discontinuity normals (reproduced from ref. 23 with permission c4 \
Hyman)
8.4 VECTORI AL REPRESENTATI ON OF ORI ENTATI ON DATA
Figure 5(b) shows aright-handed Cartesian coordinate systemi n which positivex is horizontal at
thenorth (trend000), positivey is horizontal to the east (trend090) and positivez is vertical ckmm.
This coordinate systemhas been selected to maintain compatibility with the two-dimenssotaa'
systemin Figure 5(a) while ensuring that positivez i s associated with positive angles of plunge, ha
worth noting that this coordinate systemis slightly different fromthe left-handed systemadoptee*
Priest [23] inwhichx is east and y is north. Conversionfromthis left-handed systemto the currcat
right-handed systemcan be achieved simply by swoppingx and y on the diagrams and m nc
associated equations. Any line or vector u in three-dimensional spacecan be represented mike
Cartesian systemof Figure 5(b) by putting the start point of theline or vector at theorigin of tar
systemand then noting the Cartesian coordinates u, u us of its end point. Thesecoordinates aae
usually referred to as the Cartesian components of the vector. The length of thel i D e , or magrmuoe of
the vector, is given by
ll =Jul +<4 +u\
The trenda and plungeB of alinewith Cartesian components u, uy anduz in the systemof aiesai
Figure 5(b) are given by
The parameter Q is an angle, in degrees that ensures that alies in the correct quadrant and m3e
range 0 to 360. This parameter, which depends upon the signs of ux and u,. as listed in Table2. at
required becausethe arc tangent function of most computers returns avalue in the range - 9C
to +90. Care must be taken when the denominators in equations (6) and (7) are zero I n equaaoa
(6), if ux =0 and u, >0, then a=90 whileif ux =0 andu, < 0 thena = 270. I n equation(7). a zsaa
denominator means that if u2 > 0 then B = 90' while if u. < 0 then B = 90.
The inverse forms of equations (6) and (7) are
ux = \m\s acos/ ?, u, , = | i / | si nacos/J , uz = | | si n^
The Collection and Analysis of Discontinuity Orientation Data for Engineering Design 111
Table 2 The Quadrant Parameter Q in
Equation (6)
Q
20 20 0
<0 20 180"
<0 <0 180
20 <0 360
Equations (8) allow discontinuity orientation data to be converted to vectorial form. For this
purpose it is convenient to consider the downward directed unit vector normal to the discontinuity
plane. Equations (6) and (7) give the trend and plunge of a vector and thereby permit its graphical
representation on a hemispherical projection. Only downward directed vectors, associated with
positive values of plunge, B, and positive values of the z components, u., can be plotted on a lower
hemisphere projection. This point is illustrated in the next example.
Example2
Utilizing the coordinate axes in Figure 5(b): (i) calculate the Cartesian components of
the downward directed unit vector normal to a plane of dip direction/dip angle 146/57; and
(ii) calculate the trend, plunge and magnitude of a vector with x, y, z components 1.36,
-2.85, - 1.93.
(i) This is the plane previously considered in Example 1. By equations (1) and (2) the trend and
plunge of the downward normal are respectively 326 and 33. Substituting these values into
equations (8), and taking a vector of unit magnitude, gives x, y and z components of 0.695, 0.469
and 0.545 respectively.
(ii) Before applying equation (6) it is necessary to evaluate the parameter Q by examining the signs
of the x and y components. I n this case, since the x component is positive and the y component is
negative, Q = 360. Equation (6) gives a =295.5; equation (7) gives B = 31.4". By equation (5) the
vector has a magnitude of 3.70. I t is important to note that the vector has a negative plunge; this
ans that it has an upward sense and cannot be plotted directly on to a lower hemisphere
irojection. The line of action of the vector can, however, be represented on a lower hemisphere
irojection by taking its reverse or downward directed end. The Cartesian components of this reverse
vector, found by multiplying the original components by - 1.0, are 1.36, 2.85 and 1.93. The
ameter Q in equation (6) for this reverse vector is 180 giving a =115.5. Equation (7) gives
fi = 31.4 for this reverse vector. When this reverse vector is plotted as a point on a lower hemisphere
projection it is necessary to annotate it to record its magnitude and also the fact that the original
vector has an upward sense.
85 OR I EN TA TI ON S A MPL I N G BI AS DUE TO A L I N EA R S UR V EY
Although the linear sampling regime of the scanline survey produces an objective sampling
strategy it does introduce an orientation sampling bias. One of the first to analyze this bias was
Terzaghi [53] who advocated the application of a geometrical correction factor based on the
observed angle between the sampling line and the normal to a particular discontinuity. Others who
have considered this problem include Baecher [33], Kulatilake and Wu [47], and Priest [23]. Before
. proceeding with the use of a correction factor it is first necessary to understand the exact nature of
the orientation sampling bias introduced by a linear survey.
Consider a single planar discontinuity with a normal of trend/plunge a, '/} and surface area A.
The probability, P that a randomly located long scanline of trend/plunge xs/B, will intersect, and
thereby sample, this discontinuity is directly proportional to the area, A, of the discontinuity
:.projected onto a plane normal to the scanline direction. Hence
P, cc A,
A, = A cosh
(9)
176 Discontinuities
Trace length
<2rr 2-4 m >4m
joint o o O
bedding o D /
N
1
Sampling line I
/Sampling line &80t
?o %c *
Sampling :me ;
Figure 6 Lower hemisphere, equal angle plot of discontinuity normais (reproduced from ref. 23 with permission of Uc-s^e
Hyman)
3
8.4 VECTORI AL REPRESEN TATI ON OF ORI EN TATI ON DATA
Figure 5(b) shows a right-handed Cartesian coordinate systemin which positivex is horizontal to
the north (trend 000), positivey is horizontal to the east (trend 090) and positive z is vertical down.
This coordinate systemhas been selected to maintain compatibility with the two-dimensional
systemin Figure 5(a) while ensuring that positive z is associated with positive angles of plunge. I t is
worth noting that this coordinate systemis slightly different fromthe left-handed systemadopted by
Priest [23] in which x is east andy is north. Conversion fromthis left-handed systemto the current
right-handed systemcan be achieved simply by swopping x and y on the diagrams and in the
associated equations. Any line or vector u in three-dimensional space can be represented in the
Cartesian systemof Figure 5(b) by putting the start point of the line or vector at the origin of the
systemand then noting the Cartesian coordinates ux, uf, us of its end point. These coordinates are
usually referred to as the Cartesian components of the vector. The length of thel i D e , or magnitude of
the vector, is given by
|*| = Jul + u, + u\)
The trend a and plunge 6 of a line with Cartesian components ux, uv anduz in the systemof axes in
Figure 5(b) are given by
a =arctan^J +Q {(
IS =arctan( , = _^| f7l
The parameter Q is an angle, in degrees that ensures that a lies in the correct quadrant and in the
range 0 to 360. This parameter, which depends upon the signs of ux and u,. as listed in Table 2, is
required because the arc tangent function of most computers returns a value in the range - 90L
to +90. Care must be taken when the denominators in equations (6) and (7) are zero I n equation
(6), if ux =0 anduy > 0, then a =90 whileif ux =0 and u,. <0 then a =270. I n equation (7), a zero
denominator means that if u2 > 0 then B =90 while if u. <0 then /) = 90.
The inverse forms of equations (6) and (7) are
ux = |u|cosacosft uy = ||sinacos/!, uz = |u|sin/3 (8)
The Collection and Analysis of Discontinuity Orientation Data for Engineering Design 177
Table 2 The Quadrant Parameter Q in
Equation (6)
u Q
2 0
<0
>o
20
<0
<0
0
M
180
360
Equations (8) allow discontinuity orientation data to be converted to vectorial form. For this
purpose it is convenient to consider the downward directed unit vector normal to the discontinuity
plane. Equations (6) and (7) give the trend and plunge of a vector and thereby permit its graphical
representation on a hemispherical projection. Only downward directed vectors, associated with
positive values of plunge, B, and positive values of the z components, u., can be plotted on a lower
hemisphere projection. This point is illustrated in the next example.
Utilizing the coordinate axes in Figure 5(b): (i) calculate the Cartesian components of
the downward directed unit vector normal to a plane of dip direction/dip angle 146/57; and
(ii) calculate the trend, plunge and magnitude of a vector with x, y, z components 1.36,
-2.85, - 1.93.
(i) This is the plane previously considered in Example 1. By equations (1) and (2) the trend and
plunge of the downward normal are respectively 326 and 33. Substituting these values into
equations (8), and taking a vector of unit magnitude, gives x, y and z components of 0.695, 0.469
and 0.545 respectively.
(ii) Before applying equation (6) it is necessary to evaluate the parameter Q by examining the signs
of the x and y components. I n this case, since the x component is positive and the y component is
negative, Q - 360. Equation (6) gives a =295.5; equation (7) gives fi =31.4. By equation (5) the
vector has a magnitude of 3.70. I t is important to note that the vector has a negative plunge; this
means that it has an upward sense and cannot be plotted directly on to a lower hemisphere
projection. The line of action of the vector can, however, be represented on a lower hemisphere
projection by taking its reverse or downward directed end. The Cartesian components of this reverse
vector, found by multiplying the original components by - 1.0, are - 1.36, 2.85 and 1.93. The
parameter Q in equation (6) for this reverse vector is 180 giving a =115.5. Equation (7) gives
IB = 31.4 for this reverse vector. When this reverse vector is plotted as a point on a lower hemisphere
projection it is necessary to annotate it to record its magnitude and also the fact that the original
vector has an upward sense.
85 OR I EN TA TI ON S A MPL I N G BI AS DUE TO A L I N EA R S UR V EY
Although the linear sampling regime of the scanline survey produces an objective sampling
strategy it does introduce an orientation sampling bias. One of the first to analyze this bias was
Terzaghi [53] who advocated the application of a geometrical correction factor based on the
observed angle between the sampling line and the normal to a particular discontinuity. Others who
have considered this problem include Baecher [33], Kulatilake and Wu [47], and Priest [23]. Before
proceeding with the use of a correction factor it is first necessary to understand the exact nature of
the orientation sampling bias introduced by a linear survey.
Consider a single planar discontinuity with a normal of trend/plunge //} and surface area A.
The probability, Ps, that a randomly located long scanline of trend/plunge a,//?, will intersect, and
! thereby sample, this discontinuity is directly proportional to the area, A, of the discontinuity
^projected onto a plane normal to the scanline direction. Hence
Example2
But
P. x A,
A, = A cosS
(9)
178
Discontinuities
where<5 is the acuteangle between the discontinuity normal and the sampling line. So
F , cc ,4cos< 5
The angle S can be determined either graphically from a hemispherical projection, by
algebramethods or fromthefollowing expression
cosd =.|cos(a - a,) cos ft cos ft +sin ftsinftj
The absolute valuefunction in equation (11) ensures that the angle6 is always acute: this pro]
is important in subsequent analysis. The highest probability of intersection, P,m, occurs whenI
andA. = A, The lowest probability of intersection is zero; this occurs whent5 =90 andA, - 0.
scanline samplewill, therefore, be biased to contain alower proportion of thosediscontinuities
happen to have ahigher value of 6. This reduced proportion can be expressed by the ratio
P,
= 005(5
P
This reduced sample size at the higher values of 6 can be compensated for by assigning abj
weighting to thosediscontinuities that are sampled [23, 33]. A discontinuity with anorm
trend/plungea//S sampled by ascanline of trend/plungea.JBs is assigned aweightingw give
1
w S < 90"
cos r)
where cosS is given by equation(11). For alarge sample size this weightingwill serveto balance1
orientation sampling bias introduced by linear sampling. When<5 approaches 90, w becomes \
large, to the extent that asingle datapoint could dominate thedistribution pattern, as shownm I
next example. To avoid this it may be desirable to set amaximumallowableweighting. A maxirm
value of 10 has been adopted in the programCANDO presented later in this chapter, corresponds
to <5 =84.3. This means that all discontinuities whose normals make an angle of more than 8*J*
with the sampling linewill have their maximumweighting arbitrarily limited to 10. Although* m
desirable to keep analyses of size and orientation separate, it would be arelatively simple maf^r v
provide an additional weighting that is proportional to the area(or trace length) of each disconrauv
ity in the sample, thereby giving greater emphasis to the larger features. I f such aweighting were -aec
it should also take account of the size sampling bias imposed by alinear survey, as discussed fc;
Priest and Hudson [25] and referred to briefly in Section 8.2.1. This bias is caused by the fact :ssr
a linear sampling line will already have tended to sample preferentially those discontinuities sT
alarger area, as shown by equation (10).
I f each discontinuity normal is represented by adownward directed vector, theweighting car. he
regarded as the magnitude of this vector. Utilizing the Cartesian system in Figure 5fB. l i e
components of this weighted normal vector are, fromequation (8), given by
n, w cos. cos ft
n, = w sin a cos ft,
nz =w sin ft,
Example3
Part of asample obtained fromascanline of trend/plunge 128/15 contains discontinui-
ties of dip direction/dip angle: (i) 297/73; (ii) 162/49; and (iii) 133/18. Calculate the
weighting factor for each discontinuity required to compensate for the orientation samp-
ling bias, and hencedetermine the components of the respective weighted normal vectors.
I t is first necessary to apply equations(1) and (2) to find the trend and plunge of the normal tocad
discontinuity. Theweighting factor for each is found by applying equations (11) and (12) noting tser.
in this caseocs =128 and /), =15. The results arelisted in Table 3. The Cartesian comporieras a
the weighted normal vectors, found fromequations (13), arelisted in Table 4.
I t is interesting to note that in the above example the theoretical weighting factor of 1 <? :
discontinuity (iii) is considerably higher than thosefor the other two. This is becausethediscocaa-
The Collection and Analysis of Discontinuity Orientation Data for Engineering Design 179
Table 3 Weighting Factors for Example 3
Discontinuity a n Weighting, w
(i)
297/73 117/17 10.8 1.018
(ii)
162/49 342/41 64.2 2.301
(iii) 133/18 313/72 87.1 19.531
Table 4 Cartesian Components for Weighted Normal Vector;;
in Example 3
Discontinuity Cartesian components of the
weighted normal vector
n. ny n,
(i)
-0.442 0.867 0.298
(ii)
1.652 -0.537 1.510
(iii) 4.116 -4.414 18.575
ity planelies almost parallel to thescanlineand so 5 is closeto 90. Theassumpti on is that this single
di sconti nui ty represents 18 or so other di sconti nui ti es of si mi lar ori entati on that werenot intersec-
ted by thescanline dueto thei r lyi ng almost parallel to thesampli ng line. Wi thout further sampli ng i t
is impossible to tell whether theseother di sconti nui ti es actually exist or not. Great caremust be
exercised when i nterpreti ng data that contai n hi gh wei ghti ng values such as thi s. The approach
adopted here, that of li mi ti ng themaxi mumwei ghti ng to 10, provi des a reasonablecompromi se for
most appli cati ons.
The bias compensati on methods described abovewi ll result i n ali st of ori entati on values, each of
which is associated wi th a wei ghti ng factor. I f the dataare presented graphi cally on a hemi spheri cal
projecti on, each datapoi nt must be annotated to record the associated wei ghti ng factor. Sincevi sual
i nterpretati on of the resulti ng matri x of numbers could be di ffi cult, i t is recommended that
alternative approaches to ori entati on data analysis are adopted. One of thesealternatives, whi ch
involves contouri ng thewei ghted datapoi nts, is described i n detai l by Priest [23] and also by Ragan
[30], and wi ll be menti oned only briefly here. Computeri zed plotti ng and contouri ng methods
have been discussed by Attewell and Woodman [54], by Beasley [55], and bv Kalkani and von
Frese[51].
Essentially the process of contouri ng i nvolves setting up a small movi ng sample-wi ndow on the
lower hemisphere projecti on. Thi s sample-wi ndow is usually ci rcular wi th a radius r that is
one-tenth of that of the projecti on, thereby gi vi ng a wi ndow area that is 1 % of the area of the
projection ci rcle. Thewi ndow, drawn on a transparent plasti c fi lm, is moved systematically so that
its center is posi ti oned at thei ntersecti on poi nts of a squaregri d that has agri d di mensi on rw and is
positioned beneath theprojecti on. Theweighted sample sizewi thi n theci rcular wi ndow is counted
for each gri d i ntersecti on poi nt and then recorded on a blank transparent sheet superimposed on the
projection. As theci rcular wi ndow is moved from gri d poi nt to gri d poi nt a matri x of numbers is
generated whi ch provi des a movi ng averageof weighted sample sizeper 1 % areaof the projecti on.
Thesevalues can beconverted to a percentageof thetotal wei ghted sample sizefor all of the dataon
theprojecti on and then contoured by hand. A typi cal result, presented i n Fi gure 7, i llustrates how
a contoured projecti on provi des an easily i nterpreted pi ctureof themajor clusters of di sconti nui ty
ori entati on i n a given rock structure. Themajor disadvantages of contouri ng are that the process is
largely subjective and provi des no quanti tati vei nformati on about the mean ori entati on and degree
of clusteri ng wi thi n each set. An alternati ve approach to the analysis of di sconti nui ty ori entati on
data, whi ch i nvolves the stati sti cal analysis of the ori entati on data expressed i n vectori al form, is
presented i n Sections 8.6 to 8.8.
Before proceedi ng wi th the vectori al analysis of weighted ori entati on data i t is worth noti ng
a simplepracti cal step that can betaken to mi ni mi zetheori entati on sampli ng bias. I f the geometry
of the rock facepermi ts, i t may bepossible to conduct two, or i deally three, mutually orthogonal
- scanline surveys of approxi mately the samelength at a gi ven locati on. Thi s wi ll ensurethat any
^discontinuities whi ch, by vi rtueof a large angleS, tend to bei gnored by onescanline wi ll be sampled
preferentially along another. The aggregated values from the orthogonal scanlines wi ll, therefore,