Int. J. Rock Mech. Min. Sci. & Geomech. Abstr. Vol. 13, pp. 135-148. Pergamon Press 1976.

Printed in Great Britain

Discontinuity Spacings in Rock*

S. D. PRIEST, J. A. HUDSON

The possible distributions of discontinuity spacings along a straight line through a rock mass are considered. Unless there is a large predominance of evenly spaced discontinuities, any combination of evenly spaced, clustered and randomly positioned discontinuities leads to a negative exponential form of frequency vs spacing value curve. The negative exponential form was confirmed by field discontinuity scanline surveys in three tunnels. Utilizing the negative exponential distribution, a relation between the Rock Quality Designation (RQD*) and mean discontinuity frequency per metre (2) is established: R Q D * = 100e-'tx(0.12 + 1). The agreement with actual RQD was within 5% at 27 survey locations mainly in chalk and mudstone. Recommendations are made concerning the method of presenting discontinuity spacing data, the scanline length necessary for reasonable estimates of discontinuity frequency and the number of sample values required in a discontinuity survey.

INTRODUCTION

Rock discontinuities in engineering Faults, joints, bedding planes, fractures, fissures and microfissures are of widespread occurrence in rocks encountered in engineering practice. The purpose of this paper is to discuss the spacings between these inter- Indices of discontinuity intensity ruptions in the rock continuum. The general term 'disThe form of the index adopted to describe disconticontinuity' is used here, avoiding any genetic conno- nuity intensity is often influenced by the nature of the tations, to describe any mechanical break or interrup- exposure and the survey technique. Skempton et tion of the properties of a rock [1]. al. [10] described discontinuity intensity in the London Discontinuity characteristics play a major role in Clay in terms of the number of discontinuities per unit controlling the mechanical behaviour of a rock volume of material. The values were obtained by dismass [2-4]. The following characteristics were listed by secting blocks of clay. Using a similar technique on Piteau [5"] as having particular engineering significance: Cretaceous sediments in south-east England, Fo0kes (a) Nature of their occurrence--i.e, whether members and Denness [1] expressed intensity as discontinuity of sets or isolated; (b) Orientation and position in area per unit volume. space; (c) Continuity; (d) Intensity; (e) Surface geoPiteau [5] used a scanline survey technique (measurmetry; (f) Genetic type; (g) Nature and thickness of ing tape) on rock faces and expressed discontinuity inthe joint fill. tensity as the number of discontinuities per unit disThere are many engineering situations in which a tance normal to the strike of a set of sub-parallel disknowledge of the discontinuity characteristics of a rock continuities. mass is of importance and a variety of approaches can The results of a scanline survey or the quality of be adopted to analyse the stability and behaviour of rock cores recovered during a site investigation proa given rock mass utilizing discontinuity characteristics. gramme can be described using the method proposed A commonly used technique is that applied to slope by Deere [11]. The Rock Quality Designation (RQD) stability, based on considerations of discontinuity is the proportion of scanline or borehole core that conorientation viewed in terms of factors of safety against sists of intact lengths that are 0.1 m (4 in) or longer. failure along critical discontinuity surfaces [6,7]. To calculate the RQD value, these intact lengths are summed and expressed as a percentage of the total length: t Crown copyright 1975. Any views expressed in this paper are
not necessarily those of the Department of the Environment. Extracts from the text may be reproduced, except for Commercial purposes, provided the source is acknowledged. Higher Scientific Officer, Transport and Road Research Laboratory, Dept. of the Environment, Crowthorne, Berks, U.K. Principal Scientific Officer, Transport and Road Research Laboratory, Dept. of the Environment, Crowthorne, Berks, U.K. a.r~.M.s. 13 5 A

It is also important, however, to obtain data on discontinuity intensity in addition to discontinuity orientation. In the-case of tunnelling, for example, it is necessary to examine the ratio between discontinuity intensity and excavation dimensions [8, 9].

RQD = 100

i=1

~ x,/L,

(1)

where 3q is the length of the ith length 1>0.1 m, n is the number of intact lengths >/0.1 m, L is the length 135

136

S.D. Priest and J. A. Hudson

Otstonce from A to the ith discontinuity :d~

of scanline or boreh01e along which the RQD value is required. Thus, the RQD can be calculated for various sections or the complete length of the scanline or borehole. Deere et al. [12] suggested that a scanline, which can be used to measure discontinuity intensity at a face, may be regarded as directly analogous to a borehole core since the RQD can be found in both cases. This is important because the site investigation results can be compared with conditions actually experienced during excavation.

Spocing values (x) given

as x i :d i-dill for i = I ~ n

(o) Discontinuity intersection points along a straight line (A8)through the rock mOSS

(b) Sconiine(rr~sorinq tape)on exposed rock face

Af

I I I. -',, ,l'} If-B

till(

IJJll

}/Ill Ill((

Purpose and content of this paper

The purpose of this paper is to present a theoretical approach to discontinuity spacings and RQD based on the statistical distributions of spacing values that could occur along scanlines and to compare these results with experimental data obtained in field discontinuity surveys. In the theoretical section, the influence of spacing distributions on the RQD value is explained, and the sensitivity of the RQD value to mean discontinuity spacing is illustrated. In the experimental section, the scanline measurement technique for assessment of discontinuity spacing distributions in rock is explained. Experimental results are then presented for three different tunnels in the U.K. and compared with theoretical values. Finally, the usefulness of this interpretation of discontinuity spacings and the geotechnicai implications are discussed.
Discontinuity spacing values, x
(c}Fairly evenly sP~:ed distribution

Discontinuity spacing values, (d)Clustered distribuhon

/ -.\

, B

A_ I lIMIt \1 /I.',
I I~

I ~1(

I B_

I~/ I " \ I CIl~,(/ I

Clustered and random

L i\

n otiva e, -iol

f(x)=ke_X,

. f ~ - ' ~ distributions mutually---- 1

I ~ e i n f o ~ ~t~w spocms I
II I|

~.~ I
Discontinuity spacing values, (e) Random distribution

Evenly spaced ond rondom distritxJtions mutually interfere at

Discontinuity s0oeing values, x (f}Combirmtion of distributions

Fig. 1. Theoretical discontinuity spacing distributions.

THEORY

Possible spacing distributions

The discontinuity spacings will be considered with reference to the distances between points where discontinuities intersect a straight line through the rock mass--as shown in Fig. l(a). This is equivalent to the scanline or measuring tape that is used for field determinations of discontinuity spacing or to a thin borehole core. The spacing values, .xl for i = 1 to n, are shown in Fig. l(b). The discontinuity intersection points can be evenly spaced, clustered, random or some combination of these. If the discontinuities are fairly evenly spaced with a normal distribution of spacing values, the frequency distribution shown in Fig. l(c) will be appropriate. This type of distribution will tend to occur in, say, a uniform columnar jointed basalt or evenly bedded sandstone with the standard deviation reflecting the uniformity of jointing and bedding respectively. In the case of clustered discontinuities, a high frequency of low spacing values occurs within clusters and a low frequency of high spacing values occurs between clusters. This frequency distribution is shown in Fig. l(d). Clustering can develop as a result of spalling near a free face or around joints due to stress or weathering effects. In addition, cyclic variation in lithology such

as alternating layers of sandstone and highly fractured siltstone could produce this distribution. For randomly positioned discontinuities, the discontinuity-scanline intersection points are here defined as random if the presence of one intersection point does not affect the chance of another occurring in its neighbourhood. In other words, there is no interaction between discontinuities. This could occur in a homogeneous rock mass where the formation of discontinuities has not been dominated by lithological or structural variability. In Appendix A, from standard statistical theory it is explained that if each small segment of scanline has an equal but small chance of containing a discontinuity intersection point, the points are a Poisson process and the associated spacings follow a negative exponential distribution as shown in Fig. l(e). The negative exponential distribution is mathematically expressed as

f(x) = 2e -a~,

(2)

where f(x) is the frequency of a discontinuity spacing x, and ). is the average number of discontinuities per metre. This is a one parameter (2) distribution with the mean and S.D. both equal to 1/2. In a geologically complex rock mass with a varied mechanical history, it is likely that a combination of the evenly spaced, clustered and random distributions will be present. This will result in the type of distribution shown in Fig. l(f) which will be similar to the negative exponential distribution. If, however, the mean spacing of the superimposed random distribution is

Discontinuity Spacings in Rock large compared with that of the evenly spaced distribution, the latter will be largely unaffected and consequently dominate. In all other combinations, the clusters are largely unaffected whilst the even spacings are broken up by superimposed random discontinuity patterns. It is important to note that the orientation of discontinuities, relative to that of the scanline, in general controls the number of discontinuities intersected per unit length and not the overall pattern of the discontinuity spacing distribution.

137.

The total number of discontinuities in a scanline = 2L, where 2 = mean number of discontinuities/metre, L = scanline length. Thus, neglecting end effects, there are (2L - 1) spacing values just below 0.1 m and the rejected length of scanline is approximately 0.1 ( 2 L - 1) giving, from equation (1)

RQD,,In ~- 100[L - 0.1(2L - 1)]/L

"" 10011 - 0.12 + 0. l/L]. For a scanline which is long compared with the 0.I m threshold value, the term O.1/L can be ignored and

Influence of spacin9 distributions on the RQD The Rock Quality Designation (RQD) is the percentage of scanline or borehole core that consists of intact lengths over 0.1 m. Theoretical RQDs can therefore be determined directly from distribution curves of the type shown in Fig. 1 by considering the frequency of spacing values above the 0.1 m threshold value. This leads to the concept of maximum and minimum values of RQD for the clustered and evenly spaced distributions and a relation between discontinuity frequency and RQD for the random distribution. The maximum possible value of the RQD for any mean discontinuity spacing value is always close to 100% because: (a) if the mean discontinuity spacing is less than 0.1 m, an RQD value close to 100% is obtained when the discontinuities are very clustered with large intact lengths between the clusters; (b) if the mean discontinuity spacing is greater than 0.1 m, an RQD value close to 100% can be obtained either by clustering or even spacing. The minimum possible value for the RQD is 0 for an average spacing less than 0.1 m if the discontinuities are evenly spaced. When the average spacing is greater than 0.1 m, the minimum RQD occurs when all but one of the spacing values are just below 0.1 m with one high remaining spacing value. This leads to the linear relation derived below between the minimum possible RQD value (RQD.,i.) and the discontinuity fiequency.

RQDm;. " 10011

0.1'2) for

2 <~ 10/m.

(3)

This relation is plotted in Fig. 2 in terms of RQDmi, vs mean discontinuity spacing (1/2). To calculate the RQD value for a rock having random discontinuity positions, the percentage of scanline length containing intact lengths greater than 0.1 m is found by integrating the spacing values from 0.1m to L for the negative exponential distribution as shown in Fig. 2. The relation between RQD and discontinuity frequency for this case is derived below: The probability of discontinuity spacings occurring between x and x + dx is given by f(x)dx, where x is a spacing value, dx is an increment of spacing, f(x) is the probability density distribution of x. For a long total length of scanline, L, the total number of discontinuities is 2L. Thus, the number of intact lengths between x and x + dx is 2Lf(x)dx and the length of these is 2Lxf(x)dx. From the definition of RQD, equation (1), and in the continuous case, the theoretical RQD (RQD*) with an arbitrary threshold value, t, is given by: RQD,* = 100

and with a negative exponential distribution of spacing values, equation (2),

I:

, Lxf(x) rt' /L

RQD? = 10022

f:

x e -~x dx:

I001 90 ~ 8o
. 70

MaximumpossibleRQDvalue(all discontinuitiesclusteredin smallregionsalongscanline) I

I ~.~"~'~'~'~r~om discordir~uty positions ~CP =l OOe'~(O'l k+i) 7

60 5o

,
I / /
/ n

, 7
~ l
t

~ 40

RQI3P"I~ I-xflxldx/L 0.1 -x,

. )

IO 0

I / ! /
?'

\mmolnin(j WtQh spocing~,,OlUe /

, ,

I I

I""

i ^ ~ ~ _u, Im . . . . .
;

i
0

0.1 I RQD
Threshold volue

0.2 0.:5 Meondiscontinuityspocin~ I/Mrn)

0.4

0.,.5

Fig. 2. Maximum and minimum possible RQD values and RQD vs mean discontinuity spacing for randomly positioned discontinuities.

138
~oo

S.D. Priest and J. A. Hudson

95 % of the scanline consists of intact lengths I>0. I m "%

90

" o~

7O N 6O 5O

3O
20 tO
0 I I I I I | l i I I i i

1 6 0 of the sconlineconsists of intact lengths >~l.Om

0,1 0.2

03

0.4 0.5 0.6 0.7 0.8 0.9 1.0 I.I 1,2 1.3 Mean discontinuity spacing, m

1.4 1.5

1,6 1.7

1.8 1.9 2D

Fig. 3. Variation of

RQD with

mean discontinuity spacing for a range of

RQD threshold

values t.

For a long scanline, terms containing e -zL can be ignored and

Precision of the discontinuity frequency and RQD esti.mates

Assuming a negative exponential distribution of spacing values, it is possible to study the effect of scanFor the conventional RQD, with a threshold value of line length on the precision of the discontinuity fret = 0.1 m, the theoretical value is quency (2) estimate. A short scanline, say 10 times the RQD* = 100e-1~(0.12 + 1). (5) mean discontinuity spacing, might give a bad estimate The relation given in equation (5) is also plotted in of 2 and hence the RQD value will not be a good repreFig. 2. Note that equations (4) and (5) only apply if sentation of the rock mass. In fact, widely different the discontinuity spacing values follow a negative RQD values measured over metre or two metre lengths of a borehole can simply be a reflection of the random exponential distribution. positioning of the discontinuities. Variation of RQD with the threshold value The method for studying the precision of the 2 estimate is explained in Appendix B. Examination of Table In the conventional RQD, the percentage of scanline BI reveals that if, say, 50 discontinuities are measured consisting of intact lengths above 0.1 m gives the RQD along a scanline length of 5 m, the true 2 (i.e. 2L/5) value. However, it is possible to examine the magnitude is within 7.79/m and 12.66/m at the 90% confidence of the RQD value for a variety of threshold values, t. By the same arguments presented earlier, the maxi- level. The estimated 2 of 10/m (i.e. 50/5) gives the mum RQD value remains close to 100% and the mini- RQD* as 74% from equation (5) but, from the 2 varimum is found by substituting t for 0.1 in equation (3). ationS" at the 90% confidence level, the true RQD* The intermediate case is the suite of curves generated varies between 82% and 64%. An alternative approach by equation (4) for various values of t. These are shown is to determine the probability of estimating the disin Fig. 3 with RQD* plotted against the mean discon- continuity frequency (2) to within +20% or + 10% for different sample lengths; graphs of the resulting probatinuity spacing, 1/2. The curves in Fig. 3 show that, with the negative bilities are shown in Fig. B1. For example if the sample exponential assumption, the conventional RQD is sensi- length is 50 times the mean discontinuity spacing, the tive for mean discontinuity spacings between zero and estimated 2. will lie within + 20% of the true 2 86 times approximately 0.3 m; above 0.3 m, the RQD is always out of 100. In Fig. 4, the effect of the sample length on the RQD* above 95%. The RQD value can be made more sensitive for high mean discontinuity spacing values by increas- error band is shown. For the specific case of 5 discontinuities per metre encountered in sample lengths ing the threshold value. From scanline measurements made in a rock mass ranging from 1 m to 10 m, the RQD* error band varies with a mean discontinuity spacing of 0.3 m, the conven- from 98%-72% to 94%-87% respectively. tional RQD is 95%--i.e. 95% of the scanline contains intact lengths greater than 0.1 m. From equation (4), FIELD DISCONTINUITY SURVEYS however, the percentage of scanline containing intact lengths above any required theshold can be found. For Measurement technique example, from the vertical line in Fig. 3, 16% of the In the field surveys, discontinuity characteristics were scanline contains lengths greater than 1 m when the measured along scanlines (measuring tapes) set up on conventional RQD = 95~. exposed faces. Wherever possible, equal lengths of scanline were established in orthogonal directions to obtain f 2 is the true population parameter; ;. is the estimate obtained a true three-dimensional characterisation of the rock by sampling.

RQD* = 100 e-~'(2t + 1).

(4)

Discontinuity Spacings in Rock

I00 80 ,~ I m sample length 90% Confidence band I00 80
60

139
5m sample length

~ 60
4o 20 0 0

T', \
.Error band

ROD*=91%
Error bond

95%-84%

900/.-720

~' ~

- "~ . . . . . ~ . . . . . . ~ .
50

'~" ,~

2O 0 0

I I , -- ~ ~ i ~ I0 20 30 40 Obser~<l number of discontinuities

I0

20

30

40

50

Observed number of discontinuities IOta sample length

I00

2m sample length

o 60 80 I
40 20 0 0

;, t =o ,o r
I I I I I I I I I

I0

20

30

40

50

I0

20

:30.

40

50

Observed number of discontinuities

Observed number of discontinuities

Fig. 4. Effect of sample length on the

RQD* error band (90% confidence band).

mass [5,13]. In addition to other measurements (such as orientation and trace length) at points where discontinuities intersected the scanline, the distance along the scanline was recorded (Fig. i). Discontinuity spacing values were subsequently calculated by subtraction of consecutive intersection distances along the measuring tape.

Experimental results (a) Chinnor tunnel. Detailed assessment of chalk discontinuity characteristics formed an intrinsic part of the rock cutting experiments in the Lower Chalk carried out during the full scale tunnelling trials at Chinnor, Oxfordshire. A total of seven thousand discontinuity spacing values was processed for a total scanline length of approximately 750 m taken in the 23 m deep, 3 m dia shaft and in the 5 m dia tunnel. Figure 5 shows a discontinuity spacing histogram for all measurements taken in the shaft, and Fig. 6 a spacing histogram of measurements taken at selected faces during the first 85 m of tunnelling. The distribution of measured discontinuity spficing values is not strictly
Ol3
.c_

continuous because the spacing is measured to the nearest 0.01 m on the measuring tape; parallax and angle of discontinuity to the tape do not allow greater measurement precision. For this reason the histograms have been plotted as single line columns. Figures 5 and 6 include tabulated values of mean spacing and standard deviation together with total scanline length and number of observed values. In addition, the negative exponential probability density distribution is drawn on each histogram, using the parameter ~. equal to the reciprocal of the mean discontinuity spacing. The probability density distributions corresponding to Figs. 5 and 6 are replotted on a logarithmic scale in Figs. 7 and 8. The fitted negative exponential probability density distribution is a straight line (cf. equation 2):
lnf(x) = In ~. - 2x.

Examination of Figs. 5-8 shows that there is close agreement between the theoretical and measured form of the distribution curve. In addition, for each distribution, the value of the mean spacing is approximately

II
Total sconline length 253.81m Mean discontinuity spacing(~) O . I I 4 m Standard deviation O.107m Number of values 2221

~,o

Y9
8

"-- 4 "6 g3

Fitted negative exponential / probability density distribution, ~,=8.749/rn

~o

0.01

0.C~

0.10

0.15

020

]TITrFrrl-TT ,,
0.35 Discontinuity spacing, m
3m

>o5o .
0.40 0.45 0.50

0.25

0.30

Fig. 5. Discontinuity spacing histogram. Sum of all scanlines in the

dia shaft (Lower Chalk, Chinnor, Oxfordshire).

140
lltO

S. D. Priest and J. A. Hudson

Total scanline length 514.57m Mean discontinuityspacing(~) O,105m Standard deviation 0,I I 3m Number of values 4 8 8 4

_=
.,3 6 -

~5>.

"~. 4 c o 3-

~ d F i t t e d negative exponential

istribution, X: 9.488/m>0,50 ......b~

I-

"6

o;o,

0.05

0.10

0.15

O.aO

0.25

0.30

'i''l-:

0.35

0.40

. . . . . .

0.45

0.50

Discont"inui'fy spacing, rn
Fig. 6.

Discontinuity spacing histogram. Sum of all scanlines in the first 85 m of tunnel (Lower Chalk, Chinnor. Oxfordshire).

equal to the standard deviation; this is a theoretical characteristic of the negative exponential distribution and serves to confirm the conclusion based on visual comparison. (b) Kielder experimental tunnel. The Kielder experimental tunnel, Rogerley Quarry, Co. Durham, was excavated in rocks belonging to the Carboniferous Limestone sequence. The tunnel was driven at three levels: in sandstone, limestone and mudstone horizons. During tunnel excavation, a detailed assessment of rock quality at selected faces in the sandstone and mudstone horizons was undertaken by the authors. A part of this rock quality assessment included an examination of rock discontinuity characteristics using measurement techniques described earlier.
10 9 8 )o

Discontinuity spacing histograms for the sandstone and mudstone horizons are presented in Figs. 9 and 10 together with the corresponding negative exponential probability density distributions plotted for the appropriate value of mean discontinuity spacing. Figs. 9 and 10 show that there is good agreement between the theoretical and measured distributions. In addition, for each distribution there is close agreement between the mean spacing and standard deviation. This serves to verify the applicability of the negative exponential distribution in describing the real discontinuity spacing distribution for these Carboniferous rocks. It is important to note that Figs. 9 and 10 illustrate that the negative exponential distribution is equally applicable to the highly fractured mudstone as the less fractured sandstone.
~o Fitted negative exponential

7 ~oo Fitted negative exponential

6

~..^r~oabil~ density dimibution,

5 ~ i / m

~Al~obobiiity densify distribution,

.c

o> .E

21

o o
o

0.9 0.8 ~ 0.7

._ 0.7

0.6 0.5

~ 0.5

i . 0.5
0.3 o 02

e o

ooo

0.2

0.1

..i..,.i.,,,

I.,.,h.,,I,...I..,,I.,,,I..,,I.,..I.,,

0
Fig. 7.

O.lO 0.20 0.30 040 0.50 Discontinuity spacing, m

O. I _..~....,....,....,....~....,....,....,....,...o.,o.._ 0.10 0.20 0.30 0.40 0.50 Discontinuityspacing, m

Sample probability density distribution of discontinuity spacing values, plotted on a logarithmic scale. Sum of all scanlines in lirst 85m of tunnel. 4884 Values (Lower Chalk, Chinnor, Oxfordshire).
Fig. 8.

Sample probability density distribution of discontinuity spacing values, plotted on a logarithmic scale. Sum of all scanlines in the 3 m dia shaft. 2221 Values (Lower Chalk, Chinnor, Oxfordshire).

Discontinuity Spacings in Rock

141

tO .

~ 9 -

Total sconlinelength 37.22m Meandiscontinuityspacing(~) O. 129m Standarddeviation O. 116m Numberof values 249

.~ 7 =g64 3 2

Fitted negative exponential ^ ~robability density distribution, X=7.-F;'8/rn

>0.50--.-~I
OOI 0.05 O.10 0.15 0.20 0.25 0.30 Discontinuity spocincj, rn 0.35 .... 0.40 ,,..I 0.45

i
0.50

n e

Fig. 9. Discontinuity spacing histogram. Rogerley Quarry tunnel. Sum of all scanlines in the sandstone. In the section dealing with the theoretical approach construction programme provided an additional opporit was stated that a feature of the negative exponential tunity for examining in situ Lower Chalk. Assessment distribution is that the mean spacing is equal to the of discontinuity characteristics was carried out using standard deviation. Fig. 11 is a graph plotting the mean the scanline techniques described earlier. Figure 12 discontinuity spacing against the standard deviation of shows a histogram of discontinuity spacing values, for the spacing for separate survey locations at Chinnor a 0.5 m class interval. The fitted negative exponential and Rogerley Quarry. The data have been divided into probability density distribution indicates a reasonably two groups based on the sample size: <200 values and good agreement between actual and theoretical distri> 200 values. The straight line in Fig. 11 indicates the butions considering the small sample size. idealised relation for which the mean discontinuity spacing is equal to the standard deviation. It is clear General considerations from this figure that the experimental results conform In searches of published data only two papers, Mahwell to the theoretical line, particularly for the larger tab et al. [14] and McGown et al. [15], were found samples where sampling error is lower. Agreement is giving the results of measurements of discontinuity within 20~o for samples containing more than 200 spacing in histogram form. Although the class interval values. It is worth noting that in other distributions, was rather wide the indications are that the distribufor example the normal distribution, the mean and tions tended to follow the negative exponential form. standard deviation are totally independent. In view of the tunnel data presented in this section, (c) Channel tunnelt. Excavation of the machine erec- it is felt that the negative exponential probability dention chamber during Phase II of the Channel tunnel sity distribution provides a good approximation to the
Totalscanline length Mean discontinuity spacing (~) Standard deviation Number of values 50 60.75m O.O=,L3rn O.O~.rn 1828rn 0.22 0.20 0. 18 Number of values in sample <200 >200 values values Lower chalk Chionor O Sandstone Rooerley uorry - ML~stone Rogerley cluarry a o o

~28
-G 26

O /

24

O. 16 E 0.14 o O o O e4' oO

2-00

.~ 18
16

'-

;o

12

./.:

~ 1.4
E

\
~ r ~ Fitted negativeexponential A obability der

o,0
0 0 0.08

IO

p!!"
0.01 0.0,$ 0.05 0.07 0.09 0.11 Discontinuity spacing, m 0.13 0.15

~E
~di~:onti~i~

te: if the
~ing

0.04 0.02 /

=//

distribution follows the neojative eq)onentiol distribution the means0ac~,g = the standard deviation

0
0

r,
0.02

I
0.04

I
0.06

I
ODe

I
0.I0

I
0.12

I
0.14

I
0.16 0.18

Fig. 10. Discontinuityspacing histogram. RogerleyQuarry tunnel. Sum of all scanlinesin the mudstone.
t The Channel tunnel project was cancelled in January, 1975.

Standard deviation,

rn

Fig. l l. Mean discontinuity spacing and standard deviation measuredin the shaftand tunnelat Chinnor,and in the experimental tunnel, RogerleyQuarry.

142
2O
t9

S.D. Priest and J. A. Hudson

Total scanlinelength 50.65m Mean discontinuity spacing(i) 0.378m Standard deviation 0.585m Number of values 134

18

of rock types. It is hoped that, in the future, further work will shed light on the validity of wider application of the negative exponential distribution.
DISCUSSION

._~

Validity of the negative exponential discontinuity spacing distribution assumption

The validity of the negative exponential distribution, verified for the available experimental data, indicates that equation (5) can be used to calculate the RQD for a rock given the mean discontinuity spacing (1/2). Table I summarises values of RQD, calculated from measurements using equation (l), and RQD* calculated from theory using equation (5) for the measurement locations discussed earlier. This table, presented graphically in Fig. 13, demonstrates the close agreement between RQD and RQD*. The good agreement between the measured and theoretical RQD values is not a confirmation of equation (5) but a measure of how closely the discontinuity spacing values follow the negative exponential distribution in each location. The implication is that a value of RQD, accurate to within 5~, could have been obtained at each of the twenty-seven locations by simply counting the discontinuities and calculating the discontinuity frequency, 2.
RQD Theoretical Average number of discontinuities per m, /.

.-~ I0

:=
~

7 6

]
(Spacing >1.5 m) Fitted negative exponential / ~ o b a b i l i t y densitydistribution, Jl

5 ~

0 0.1 0.2 0.3 0.4 0.5 0,6 0.7 0.8 0.9 1.0 I.I

1.2 1.3 1.4 1.5

Discontinuity spocinq, rn
Fig. 12. Discontinuity spacing histogram. Channel tunnel. Sum of vertical and horizontal scanlines (Lower Chalk).

distribution of discontinuity spacings measured in the field. The applicability of the negative exponential has only been verified at a few locations for a limited range

TABLE 1. COMPARISON BETWEEN MEASURED AND THEORETICAL

Measured

RQD ",,
Rock formation, location Lower Chalk Chinnor shaft Lithology Chalk Scanline configuration Horizontal scanlines Vertical scanline I Vertical scanline II Chainage = 9.9m ll.5m 12.1m 14.0m 18.1m 20.4m 35.9m 55.4m 57.5m 60.5m 65.2m 70.2m 72.7m 73.9m 79.3m 85.0m (from equation 1)

RQD* %
(from equation 5)

79 67 7(/ 91 87 80 83 82 78 78 72 60 66 79 64 71 64 65 63 83 100 12 32 14 22 99 88

81 70 75 89 87 81 87 81 81 81 74 64 69 79 65 73 68 69 63 82 98 12 31 17 26 99 90

7.9 11.0 9.5 5.6 6.3 7.9 6.4 7.9 8.0 8.0 9.9 12.5 11.3 8.5 12.3 10.1 11.7 11.3 12.8 7.8 2.0 37.0 23.8 32.5 26.3 1.2 5.3

Lower Chalk Chinnor tunnel

Chalk

Horizontal plus vertical scanlines for each face

Sandstone Carboniferous Limestone Rogerley tunnel Limestone Mudstone

Vertical plus horizontal scanlines Vertical plus horizontal scanlines Vertical plus horizontal scanlines at selected faces Vertical plus horizontal scanlines

Lower Chalk Channel tunnel

Chalk

Discontinuity Spacings in Rock

Chinnor,Lowef Chalk

143

discontinuities already present in the rock mass and to results obtained from scanlines set up on rock faces. If, however, the discontinuity spacings are obtained LO0 from a borehole core, there is a possibility that additional discontinuities may have been introduced by 9O the sampling process---i.e, drilling the borehole [3]. Figure 15 summarises data presented by Deere et 80 al. [12] from cores and scanlines in a variety of rock 70 types. Figures 15(a--d) show the comparison between RQD* and the measured values of Deereet al., broken 6O down into the four rock types. These figures indicate 50 that for a given ~., equation (5) predicts an RQD* higher than the measured value. The effect causing RQD to be <RQD* (in particular for 2 >i 10/m) is a tendency 3O for the discontinuities to be more evenly spaced than predicted by the negative exponential distribution (cf. 20 J RQD=Theoreticol RQOlfrom equotion 51 Fig. 14). Most of the data in Fig. 15 relate to NX I0 / core; it is possible, therefore, that the process of rotary drilling added a more evenly spaced component of dis~ I I I 0 /~ I I I I 0 I0 20 30 40 50 60 70 80 90 I00 continuities to the recovered core, though this could RQD, % only be confirmed by detailed surveys on freshly Fig. 13. Comparison between measured and theoretical RQD. exposed rock faces. Experience at Chinnor has shown that lengths of inFigure 14 shows a graph of equation (5) relating RQD* to 2 together with the experimental data points tact core can snap near the centre during drilling; this, presented earlier. Between values of 2 = 6/m and together with the phenomenon of discing due to high 2 = 16/m the relation between ;t and RQD* is approxi- field stresses could lead not only to an artificially lowmately linear. The curve in Fig. 14 has an inflection ered RQD but also to an artificially imposed discontipoint P at 2 = 10/m; at this point the curve has a slope nuity spacing distribution. These effects must be borne equal to dRQD*/d2[~=lo = - 1 0 e -1 = -3.68 m. The in mind and allowed for when interpreting results constraight line drawn through P, tangential to the curve cerning rock quality obtained from boreholes referring, ideally, to the results of scanline surveys taken on defined by equation (5) is therefore described by freshly exposed faces. RQD* = -3.682 + 110.4. This line, shown in Fig. 14, gives a good approximation Geotechnical implications of RQD between values of 2 = 6/m and 2 = 16/m. The RQD, commonly adopted as a rock quality index Bieniawski [16], by reference to Deereet al. [12] sugis here regarded as having two major drawbacks when gested that the relation between 'fracture frequency' (i.e. viewed in the light of the present approach: 2) and RQD is linear; this statement is generally in (a) Calculating RQD is a time consuming process agreement with Fig. 14 within the stated range of 2. which, when applied to borehole cores, can give results Discontinuities imposed by sampling unrepresentative of the rock mass. In both the theoretical and experimental sections of (b) The conventional RQD is insensitive to variations this paper, the discussion has been with reference to in rock quality when the average discontinuity spacing
Ro~ey, sondstom |
# R o ~ I w , limestone} Corboniferous Limes~ae Rol~ey, mudstone J o Channel tunneI,LowefCholk I00 90 80 70 Linear approximation RODe=- 3.68), + I 10.4 # .~for6<X<16 Experimental doto points ChinnortLowerChalk

U'l"e,~

I'lk.
~"-~

Inflection point P

L~ 9w ~ " ~

. . . . . . I.)eVlOT IOn aue to clustered discontinuities .~ ~ / / ' ~ /

#RoQerley,limestone] CarboniferousUmestone a RogerlElYtmu~mne---J-L-I J 0 Chores, tunneltLower Chalk Theoreficol curve .-~1 RQ~-IOO. x (O. I X + I )

~g 6o c$5o

~ 40
:50 20 I0 0 0 I 2 I 4 I 8 I I I I I I I I I ~1 I0 12 14 16 18 20 22 24 26 28 30 Averoge number of discontinuities per m, X I 32 I 34 I 36 I 38

40

Fig. 14. R e l a t i o n b e t w e e n R.MM.S 13 5 a

RQD a n d

mean discontinuity frequency.

144
I O0

S.D. Priest and J. A. Hudson

r~~l
II B

50 O e. 25

:....-;,<
o r~ T t ~ o r e t i c a l curve

Theorei'icol curve

R O D .

100

75

~ ~

Theoreticol curve RQD*

50

25

J)NX core, Hockensock Siltstone 0 = I I = I I I I I 0 2 4 6 8 I0 12 14 16 18 20 AverSe number of discontinuities per m, I00

ff--u,,,~,~
.-

b) NX core, John Day Basalt

0 I 2 t 4 I 6 = I I I I I 8 10 17' 14 16 18 20

Avero, ~ number of discontinuities per m, ~, I00 ~ o Theoretical curve

- "~ ROD ~

75
50 0 w" 25
I I 0
I I I I D D I O D n O

d 50 o e~

25 ,:) NXcore, Dworshok Darn, granite gneiss

I I I I I I i

Tunnel woll, scanli ne o across discontinuities Tunnel wall,sconline parallel to discontinuities o a NX core
1~3.4~

t) Climax Stock Granite

I I

6 8 I0 12 14 16 18 20

2 4

6 8 I0 12 14 16 18 20

Average number of discontinuities per m, ~,

Fig. 15. C o m p a r i s o n between measured

/&verage number of discontinuities per m,

and theoretical

RQD (RQD

values from Deere

et al.,

1967).

exceeds 0.3 m. For example, for a large excavation, transition from 'favourable' to 'unfavourable' conditions may be represented by a variation in RQD of only 3%, from 98% to 95%. RQD has, however, an advantage over simple indices which express, say, an average discontinuity frequency per metre (2) in that it makes statements about the distribution of block sizes. If the negative exponential distribution is accepted as valid, from a simple knowledge of ~. additional data concerning RQD can be calculated if required. Moreover a two-tier RQD index can be produced by adopting a pair of threshold values at, say, 0.1 m and 1.0 m. The selection of the second threshold value, 1.0 m, ensures sensitivity of the two-tier RQD for mean spacing values up to 2.5 m for use when considering particularly large excavations. The suggested rock quality index would therefore be, for example: ).= 3.3/m, RQD&I=95%, RQD*.o= 16%, which gives information on discontinuity frequency, percentage of rock containing spacing values greater than 0.1 m and percentage of rock containing spacing values greater than 1.0 m. Acceptance of the negative exponential distribution allows calculations to be made concerning the precision of the discontinuity frequency estimates made from any given scanline. In this way, it is possible to determine the minimum scanline length, expressed as a multiple of the mean discontinuity spacing, required to produce a rock quality index at a specified precision. For example, if the sample length is 50 times the mean discontinuity spacing, the estimated 2 will lie within -t-20% of the true 2, 86 times out of 100. A knowledge that the distribution of discontinuity spacings follows a negative exponential form can be of use in rockbolt design. Rockbolt design criteria may embody requirements that the bolt should be of sufficient length to penetrate, say, at least the third joint block back. The probability of intersecting three joint

blocks for a given rockbolt length can be calculated directly from the statistical theory. The distribution of discontinuity spacings in a borehole is of critical importance when measuring mass permeability using packer test methods. In order to obtain a precise picture of mass permeability, minimising variations caused by sampling error, it is necessary to design the packer test length with regard to mean discontinuity spacing. Thus, a similar approach to that used when considering the effect of sample length on the precision of the ). estimate can be used to determine effective packer permeability test lengths. Snow[17,18] described a method for determining discontinuity frequency from the proportion of zero discharge packer permeability tests that occur. He made the basic assumption that the occurrence of open discontinuities in a given length of borehole obeys the Poisson distribution. A corollary of the Poisson process is that the discontinuities follow a negative exponential distribution. The assumptions in Snow's work, though dealing with the particular case of open, water bearing discontinuities, may therefore be valid for the rocks examined here. SUMMARY AND CONCLUSIONS The possible distributions of discontinuity spacings along a straight line through a rock mass were considered. The effects of evenly spaced, clustered and randomly positioned discontimJities were examined from a theoretical standpoint. It was found that, unless there is a large predominance of evenly spaced discontinuities, any combination of evenly spaced, clustered and randomly positioned discontinuities will lead to a negative exponential form of frequency vs spacing value curve. Analysis of field results obtained from scanline measurements carried out by the authors indicated that

Discontinuity Spacings in Rock a negative exponential distribution of discontinuity spacing values is a good approximation to the true distribution patterns. Deviations from negative exponential form in other rocks, though possibly a reflection of the true discontinuity characteristics of the rock mass, could in some cases result from data compiled from a small number of measurements (sampling error), or data obtained from borehole core (discontinuities superimposed by drilling). Using the negative exponential distribution of discontinuity spacing values, a relation was established between the theoretical Rock Quality Designation (RQD*) and the average number of discontinuities per metre:

145

water flow should be expected if the test lengths are only of the order of ten times the mean discontinuity spacing, implying that correlation of permeability values from borehole to borehole should not then be attempted. It is suggested that discontinuity surveys should be carried out in a variety of rock types to assess the wider applicability of the concepts presented in this paper. These surveys should be carried out on freshly exposed faces, with sample sizes consistent with the recommendations detailed above.
Acknowledgements--The work described in this paper forms part of the programme of the Transport and Road Research Laboratory and is published by permission of the Director. The authors wish to thank the following organisations for their help: Rugby Portland Cement Co; Mort Hay and Anderson; Sir Robert McAlpine and Sons Ltd; Soil Mechanics Ltd; Rio Tinto Zinc; Cross Channel Contractors and the Steering Committee for the Kidder Water Scheme Experimental Tunnel. The research was carried out under the general direction of Mr. M. P. O'Reilly, Head of Tunnels Division, Transport and Road Research Laboratory, Dr. P. B. Attewell and Dr. I. W. Farmer, both of the Engineering Geology Laboratories, University of Durham. The authors would like to thank Dr~ F. Garwood for helpful statistical advice and Dr. E. T. Brown, Imperial College of Science and Technology, Mr. J. B. Boden, Mr. C. McCaul and Mr. G. H. Alderman of TRRL for help in obtaining and recording discontinuity spacing data.

RQD*= 100e-'l;(0.12 + l).

With this formula it was possible to compare the theoretical RQD, obtained from the discontinuity frequency alone, with the actual RQD calculated from the spacing values. This was done for all measurement locations. The maximum error was 5%, indicating that the RQD could have been found to within 5% simply by counting the number of discontinuities along the scanline and using the formula. In addition, the formula shows that the conventional RQD with a threshold value of 0.1 m is insensitive when the mean discontinuity spacing is above 0.3 m. The authors recommend presenting discontinuity spacing data obtained from scanlines in the following form: (a))., the mean number of discontinuities.per metre, or ~ the mean discontinuity spacing (~ = 1/2). Since RQD is a commonly used method of expressing discontinuity spacing data in site investigation, a twotier RQD may be used: (b) RQDo.I, the conventional RQD. (c) RQDLo, the RQD with a threshold value of 1.0 m.t If the theory outlined earlier is applicable, (b) and (c) can be calculated directly from (a). It was also found that to estimate the number of discontinuities per metre to within a reasonable precision requires a scanline length at least fifty times the mean discontinuity spacing. Thus, if there are about five discontinuities every metre, the scanline must be at least ten metres long. Moreover, at least two hundred measurement values were generally required before the negative exponential distribution form became clear on a histogram. One of the features of this distribution is that theoretically the mean and standard deviation are equal: this was found to be the case to within 20% for the nineteen measurement locations in Lower Chalk and Carboniferous Limestone where greater than two hundred readings were taken. These sampling factors are relevant to packer permeability tests where water flow occurs essentially through the discontinuities. Very large variations of
t" A threshold value of 1.0 m is recommended for general use; however any value can be used to suit particular conditions.

Received 29 September 1975.

REFERENCES
1. Fookes P. G. & Denness B. Observational studies on fissure patterns in Cretaceous sediments of south-east England. Geotech. 19. 493-497 (1969). 2. Trollope D. H. The mechanics of discontinua or elastic mechanics in rock problems. In Rock Mechanics in Engineering Practice. (Edited by Stagg K. G. & Zienkiewicz O. C.), Chap. 9. pp. 275-320 (1968). 3. Jaeger J. C. & Cook N. G. W. Fundamentals of Rock Mechanics. p. 513. Methuen, London (1969). 4. Hock E. & Bray J. W. Rock Slope En~lineeriml. p. 309. Inst. Min. Mctall. (1974). 5. Piteau D. R. Geological factors significant to the stability of slopes cut in rock. S. Af Inst. Min. Met., Syrup. Planning Open Pit Mines, Johannesburg. 33-53 (1970). 6. AtteweU P. B. & Woodman J. P. Stability of discontinuous rock masses under polyaxial stress systems. In Stability of Rock Slopes. 13th Syrup. Rock Mech. (Edited by Cording E. J.) pp. 665-683. ASCE, NY. (1971). 7. Hock E., Bray J. W. & Boyd J. M. The stability of a rock slope containing a wedge resting on two intersecting discontinuities. Q. J. Engng Geol. 6, 1-55 (1973). 8. Watkins M. D. Terminology for describing the spacing of discontinuities of rock masses. Q. J. Engng Geol. 3, 193-195 (1971). 9. Barton N., Lien R. & Lunde J. Analysis of rock mass quality and support practice in tunneling, and a guide for estimating support requirements. Norwegian Geotechnical Institute, Internal report, 19 June (1974). 10. Skempton A. W., Schuster R. L. & Petley D. J. Joints and fissures in the London Clay at Wraysbury and Edgware. Geotech. 19, 205-217 (1969). 11. Dcere D. U. Technical description of rock cores for engineering purposes. Rock Mech. Eng. Geol. 1, 17-22 (1964). 12. Dcere D. U., Hendron A. J., Patton F. D. & Cording E. J. Design of surface and near surface construction in rock. In Failure and Breakage of Rock (Edited by Fairhurst C.), pp. 237-302. AIME, NY. (1967). 13. Terzaghi R. D. Sources of error in joint surveys. Geotech. 15, 287-304 (1965). 14. Mahtab M. A., Boistad D. D. & Kendorski F. S. Analysis of the Geometry of Fractures in San Manuel Copper Mine, Arizona. U.S. Department of the Interior, Bureau of Mines. No. RI 7715, 24 (1973).

146

S.D. Priest and J. A. Hudson In practice, the probability of a small unit length of scanline being intersected by a discontinuity need not be one half as for the unbiased coin. Thus, the analogy becomes direct for a biased coin1" with a different probability value for tails producing a random sequence such as HHHTHHHHTHTH HHH HTTH HT.

15. McGown A., Saldivar-Sali A. & Radwan M. Fissure patterns and slope failures in till at Hurlford, Ayrshire. Q. d. Engng Geol. 7, 1-26 (1974). 16. Bieniawski Z. T. 13ngineering classification of jointed rock masses. The Civil Engineer in South Africa 15, 335-343 (1973). 17. Snow D. T. Rock fracture spacings, openings, and porosities. J. Soil Mech. Found. Div. ASCE, 94, SMI Proc. Paper 5736, 73-91 (1968). 18. Snow D. T. The frequency and apertures oi' fractures in rock. lnt J. Rock Mech. Min. Sci. 7, 23-40 (1970). 19. Lindley D. V. Introduction to Probability and Statistics--l. Probability, p. 259. Cambridge University Press, London (1965). 20. Garwood F. Fiducial limits for the Poisson distribution. Biometrika, 28. Parts 3 and 4, 437-442 (1936). 21. Pearson E. S. & Hartley H. O. Biometrika Tables for Statisticians. Vol. I. Cambridge University Press, London (1954).

The probability of k tails occurring in n tosses, or k discontinuities in a length n units, is given.by the binomial distribution as: p~k tails io~ = (n!/(n -- k)!k!)qkp"-k, x n losses / (A1)

APPENDIX A The Poisson and negative exponential distributions

Using the scanline technique, discontinuity spacings are measured by setting up a measuring tape across the rock face and noting the tape positions where discontinuities are present. If it is assumed that each small section of scanline has an equal chance of containing a discontinuity, the discontinuity positions are random and the spacing distribution can be studied in two ways: (a) the discrete case--a unit length of rock (say 10 mm) is considered intact if no discontinuity is present, and fractured if a discontinuity is present; the total scanline length must then be an integral multiple of TABLE A 1 . SPACING SEQUENCES BETWEEN SUCCESSIVE TAILS the unit length; (b) the continuous case--the unit lengths become inNo. of heads Probability for finitely small and the total scanline length can have (spacing between unbiased coin Sequence tails) Probability (p = I/2) any value. In the discrete case, the binomial and geometric dis- T-T 0 q 1/2 tributions provide the probabilities associated with fre- T-HT 1 pq 1/4 2 ppq 1/8 quency of occurrence and spacing values respectively. T-HHT 3 pppq 1/16 In the continuous case, the equivalent distributions are T-HHHT T-HHHHT 4 ppppq 1/32 the Poisson and negative exponential functions. The discrete case is explained via a coin tossing analogy This probability density distribution is called the and the continuous case as a limit to the discrete case geometric distribution because the probability is when the number of units becomes infinite. The coin reduced by the constant factor p for each unit increase tossing analogy is useful because it provides an intuitive o f spacing. understanding of why the most likely spacing value is Thus zero when the discontinuities are randomly positioned. p / k heads between'~ ~ ~suessive tails ) = pkq pk(l p). (A2) (a) Discrete case. Consider, in analogy, an unbiased coin tossing sequence of heads and tails where a head This distribution is plotted in Fig. A1 for a specific represents a unit length of intact rock and a tail rep- value of p (19 = 0.91): for p = , the coin is unbiased; resents a unit length containing a discontinuity. For for other p values, the coin is biased but the coin tossing sequence is still random. Note that for all p values, example: the most frequently occurring spacing value is zero. The Coin sequence: THHTHHTHTTTHHTHT THT I(I I Ill I h i I/IIKI I I/I I / l l l I/1 mean of the geometric distribution is equal to p/(1 - p). Scanline: I ~ ~. ', ~, / t ~ / (b) Continuous case. In the coin tossing analogy, the There is an analogy between the number of tails that coin can only be thrown an integral number of times will occur in n tosses and the number of discontinuities and hence the scanline length can only be an integral that will occur along a scanline of length n units. In multiple of the unit length. The equivalent distribution addition, the spacings between successive tails and the that characterises the number of events occurring in spacings between successive discontinuities are analo- a continuous interval x is the Poisson distribution. The gous. probability of k events occurring in the interval x is t Although biased, the result is still random--similar to selecting given by a ball from a bag containing different numbers of black and red pek .... ts x i,~ (A3) ~interval ] ~ e - xx(2x)k/k!,
balls,

where P(x)= probability of x occurring, p = probability of a head (or an intact unit), q = 1 - p = probability of a tail (or a unit containing a discontinuity). The probabilities associated with encountering a certain number of discontinuities in a given length of scanline can be calculated from equation A I given q the probability of a discontinuity intersecting a unit length of scanline. The mean and standard deviation of the binomial distribution are np and x/npq respectively. However, we are also interested in the distribution of spacings between successive tails and, by analogy, the distribution of spacings between discontinuities. In this case, the probability of each sequence is considered. Assuming that a tail has just occurred, the sequences in Table A1 can follow:

Discontinuity Spacings in Rock

Geometric

147
Nc~jofi~ e~oon~t~l

distribution
O. lO
-

distribution

0.IO~
0.08 I

heads between~_^k q ~', pL~ __~' /s_ive toils K )"" ~ -v, ~ Where P(Head) : p P(Tail) = l-p=q Mean spacing = p/(l - p] Heads p= 0.91 i, Mean spacing value= I 0 heads 0.08

\ X

f(x)=Xe-x= whereh =nunt~ of events/unitlength Mean spacing=l/X units

~ 0.06

~0.06

X = O . I/unit length

~
~

value= IOunits

M e o n s p e c i n g

:~

004

"~
0.

0.02 I-

o.o~ IO/ 0

0 1

10

15

I I

',
25

20

25

Spacing value, K heads

t i 5 tO 15 20 Spacing value,x, arbitrary units

Fig. A1. Geometric and negative exponential distributions.

where 2 is the mean number of events occurring in a unit interval. In the Poisson distribution, the values of the mean and variance are both 2. Since we are interested in the discontinuity spacing probability density distribution, consider the distance / from one discontinuity to the next:

accidents follow a Poisson distribution. In the discontinuity context, an apparent clustering of discontinuities will occur as a natural consequence of the random positions. APPENDIX B

P(d <<,x ) = t - P(d > x)

Precision of the mean discontinuity frequency estimate

The parameter 2 in the negative exponential distribution (cf. equation 2) is estimated by counting the number of discontinuities intersecting a scanline and then dividing by the scanline length: The estimated number of discontinuities per metre (2) is given by

= 1-

P(no discontinuities in distance x)

and from equation (A3), with k = 0 P(t' ~ x) = 1 - e-xx. (A4)

This is a negative exponential cumulative probability distribution. The probability density distribution (f(x)) is the derivative of equation (A4) with respect to x:

n/L,

(B1)

f(x) = 2e -xx.

(A5)

This spacing distribution (equation A5) is plotted in Fig. A1 for a specific value of 2 (2 = 0.l/m). The mean and standard deviation are both equal to 1/2. In both the geometric and negative exponential distributions, the most likely spacing value is zero. Lindley [19] suggests that this is the reason for the phrase "it never rains but it pours" since rare events such as

where n is the number of discontinuities encountered; and L is the scanline length. We wish to know the probability of estimating 2 within a certain error band as a function of the scanline length. For example, if the true 2 = 10/m and a scanline length of one metre is taken, then 10 discontinuities must occur for 2 to be within 59/o of the value 10/m; if 9 or 11 discontinuities occurred, 2 would be 10% in error.

Curves ore upper envelopes of probability The actual curve has a saw tooth shape becausethe error bonds ore artificial (c. Appendix B) It is only passible to measure an integral number of discontinuities.Thus,if XL =3.4,it is not ,ossibleto estimate within IO%-i.e. measureo numberof discontinuitiesbetween 3.06-3.74

1.0

E'~ 0.8

=g
/ ""'~

"""'"""""*20/~r~--"~-~:to%etror bond

"1
-- -- --I I

~ OA

-- -- Ex~ple:To estimate X to within :1:10%, either 18,19,20,21 or 22 discontinuities must occur in the samplelength20 times the mean dsoontindty spaci~ Fromequation AS,the pmpabili~ is given as

x
0
I I I l I I

K=I8

~20~/Kt =0.42
I

I0

15

20

25

30

35

40

45

50

XL-Numl~r of times the sample length is91mtetfhanthe mean dsco~inuity spacing (for either sconline,borehole or packer permeability test length)

Fig. B1. Graph for determining the probability of estimating the discontinuity frequency (2) to within +20% or + 10% for different sample lengths.

148

S.D. Priest and J. A. Hudson 20, 21, n2 = 22. The probability of this occurring is 0.423---calculated from equation (B5). However, because of the arbitrary error band, it is preferable to use the upper and lower limits of 2L from the observed number of discontinuities at, say, the 98~o or 90~o confidence limit [20] as in Table B I below. A larger table is available in Pearson and Hartley [21].
TABLE B I . VARIATION OF 2 L AT 98% AND 900o CONFIDENCE LEVELS GIVEN THE OBSERVED NUMBER OF DISCONTINUITIES, FROM

Equation (A3) gives the probability of k events occurring in an interval x as

p t k ,nts in~
I,inlcr',al x }

e- ;x(~.x)k/k!

(B2)

Thus, for the example above pOO di. . . . tinuitics] = e-1o 101o/10! = 0.125. ~in one metre In this case, for a true ). of 10/m and taking a scanline length of 1 m, the probability of estimating ). within 5 ~ is only 0.125. Note that the use of 5 ~ is arbitrary here because the same probability applies to estimating 2 to within any percentage less than 10~--i.e. when ten discontinuities are measured, ). will be exactly equal to 2. In general for ~ to lie within a certain error band, + ~, 2(1 - E) ~< 2 ~< 2(1 + E) and, from equation (B1) 2L(1 - ~) ~< n ~< 2L(1 + e). is
k= a2

GARWOOD [20] Possible variation of 2L n 0 1 5 10 15 20 25 30 40 50 98% confidence 0-4.61 0.01-6.64 1.28-13.11 4.13-20.14 7.48-26.74 11.08-33.10 14.85-39.31 18.74-45.40 26.77-57.35 35.03-69.07 90% confidence 0-3 0.05-4.74 1.97-10.51 5.43-16.96 9.25-23.10 13.25-29.06 17.38-34.92 21.59-40.69 30.20-52.07 38.96--63.29

(B3)

(B4)

From equation (B2), the probability of this occurring

e-aL(2L)k/k!,
k=nt

(B5)

for integral values of nt to n2, satisfying inequality (B4). If, for example, the scanline length is 20 times the mean discontinuity spacing (2L = 20) and a 10~o error band is allowed (e = 0.1), n can take the values nt = 18, 19,

If 50 discontinuities are observed in a scanline length of 5 m, 2L varies from 38.96 to 63.29 at the 90~o confidence level (Table BI); i.e. 2 varies between 7.79/m and 12.66/m. From equation (5), the theoretical RQD* value then varies from 82~o to 64?/0. This illustrates the high sampling error in using short scanlines or borehole lengths for RQD determinations.