Volume 62 Number 2

of the South African Institution of Civil Engineering June 2020

QQ  echanical behaviour of semi-rigid connections

M
coupling two concrete segmental linings
QQ L aboratory and field investigations in granular soils to
correlate relative density, relative compaction and grain size
QQ  he liquefaction potential of the upper quaternary
T
sands of the Cape Flats, Western Cape, South Africa
QQ Warrants for roundabouts and other traffic control devices
QQ  he cost-effectiveness of water augmentation and
T
management: Assessing the Unit Reference Value
Volume 62 No 2 June 2020 ISSN 1021-2019

Publisher
South African Institution of Civil Engineering
Block 19, Thornhill Office Park,
Bekker Street, Vorna Valley, Midrand, South Africa
Private Bag X200, Halfway House, 1685, South Africa
Tel +27 11 805 5947, Fax +27 11 805 5971
http://www.saice.org.za
civilinfo@saice.org.za

Contents
Editor-in-chief
Dr Peter Day
Consultant, Jones & Wagener (Pty) Ltd
Adjunct Professor of Geotechnical Engineering,
University of Stellenbosch
Tel +27 11 519 0200
day@jaws.co.za
joint EditorS-in-chief 2 Mechanical behaviour of semi-rigid connections
Prof Chris Clayton
University of Southampton coupling two concrete segmental linings
c.clayton@soton.ac.uk
Prof Gerhard Heymann
P Zhai, P Xu DOI: 10.17159/2309-8775/2020/v62n2a1
University of Pretoria
gerhard.heymann@up.ac.za
Editor 12 Laboratory and field investigations in
Verelene de Koker
Tel +27 11 805 5947, Mobile +27 83 378 3996
granular soils to correlate relative density,
verelene@saice.org.za
relative compaction and grain size
journal editorial Panel
Prof Y Ballim – University of the Witwatersrand H Mujtaba, K Farooq, N Sivakugan, B M Das DOI: 10.17159/2309-8775/2020/v62n2a2
Prof W Burdzik – University of Pretoria
Prof CRI Clayton – University of Southampton
Dr P Day – Jones & Wagener (Pty) Ltd (chairman)
Prof J du Plessis – University of Stellenbosch 22 The liquefaction potential of the upper quaternary
Prof GC Fanourakis – University of Johannesburg
Prof PJ Gräbe – University of Pretoria sands of the Cape Flats, Western Cape, South Africa
Dr C Herold – Umfula Wempilo Consulting
Prof G Heymann – University of Pretoria N Fouché DOI: 10.17159/2309-8775/2020/v62n2a3
Prof A Ilemobade – University of the Witwatersrand
Prof SW Jacobsz – University of Pretoria
Prof EP Kearsley – University of Pretoria
Prof C Roth – University of Pretoria
31 Warrants for roundabouts and
Prof W Steyn – University of Pretoria
Mr M van Dijk – University of Pretoria
other traffic control devices
Prof C Venter – University of Pretoria
Prof A Visser – University of Pretoria
J D Sampson DOI: 10.17159/2309-8775/2020/v62n2a4
Dr E Vorster – Aurecon South Africa (Pty) Ltd
Prof J Wium – University of Stellenbosch
Prof A Zingoni – University of Cape Town
Prof M Zuidgeest – University of Cape Town
39 The cost-effectiveness of water augmentation and
Peer reviewing
management: Assessing the Unit Reference Value
The Journal of the South African Institution of
Civil Engineering is a peer-reviewed journal R Bester, J N Blignaut, P H van Niekerk DOI: 10.17159/2309-8775/2020/v62n2a5
that is distributed internationally
Design and reproduction
Marketing Support Services, Ashlea Gardens, Pretoria 45 Guidelines for the preparation of papers and
Printing technical notes
Fishwicks, Pretoria
Papers for consideration should be submitted
online at: http://journal.saice.org.za
The South African Institution of Civil Engineering accepts no
responsibility for any statement made or opinion expressed
in this publication. Consequently, nobody connected with
the publication of this journal, in particular the proprietor,
the publisher and the editors, will be liable for any loss or
damage sustained by any reader as a result of his or her action
upon any statement or opinion published in this journal.

© South African Institution of Civil Engineering 1

TECHNICAL PAPER Mechanical behaviour of
Journal of the South African
Institution of Civil Engineering semi-rigid connections
ISSN 1021-2019
Vol 62 No 2, June 2020, Pages 2–11, Paper 0596 coupling two concrete
PANPAN ZHAI received his MSc in Applied
Mathematics in 2014 from Wenzhou
segmental linings
University. Following a number of years as a
doctoral candidate, he changed his focus to
civil engineering. This paper is based on P Zhai, P Xu
research conducted for his PhD degree.

Contact details: This paper describes the use of segment joints in shield-driven tunnel linings. The opening of
PhD Candidate segment joints due to the application of a bending moment under different levels of axial stress
School of Water Science and Engineering is studied. The main focus is the mechanical behaviour of the joints, which are considered to be
Zhengzhou University
semi-rigid, and the moment-rotation behaviour of the joint. A physically based mathematical
Wenhua Road 97
Zhengzhou model is proposed based on the joint dimensions and material properties. Considering joints
China connected by two long curved bolts as an example, the model parameters are evaluated,
T: +86 188 3819 4945 and the moment-rotation curves for the joint are presented. The model distinguishes joint
E: panpanzhai2014@163.com
behaviour before and after opening, and agrees well with the real mechanical behaviour of
the joints. A comparison with previous experiments on segment joints and with Janssen’s and
PROF PING XU holds a PhD from the
Zhejiang University. He is currently a Gladwell’s models is provided, and the proposed model is found to be accurate and universal.
professor in Civil Engineering at the
Zhengzhou University.

Introduction Figure 1) connected by steel bolts are

The process of subway construction is commonly used in shield-driven tunnels.
undergoing rapid developments in China. These are one of the single most important
Contact details: Shield-driven tunnels are widely used in elements in the total cost of tunnelling,
School of Water Science and Engineering
Zhengzhou University
soft ground because of their flexibility, representing 15–40% of the total cost
Wenhua Road 97 cost-effectiveness, and minimal impact (Iftimie 1994).
Zhengzhou on ground traffic and surface structures. Segment joint opening is a common
China
Segmental precast concrete linings (see problem in shield tunnel operations, and is
T: +86 137 3318 9057
E: pingxu127@163.com

Bolt pocket

Segment joint

Segment

Key segment
Counter segment

Keywords: shield-driven tunnel, segment joint, semi-rigid connection,

M-θ behaviour Figure 1 Segment lining

Zhai P, Xu P. Mechanical behaviour of semi-rigid connections coupling two concrete segmental linings.
2 J. S. Afr. Inst. Civ. Eng. 2020:62(2), Art. #0596, 10 pages. http://dx.doi.org/10.17159/2309-8775/2020/v62n2a1
a major cause of water leakage. The width
of the segment joint opening is an indica-
tor of the deformation of the segmental
lining. The damage caused to segment
joints under large bending moments is the
most common structural problem with
tunnel linings.
Considerable research has focused on
the mechanical behaviour of the segmen-
tal linings. Originally the linings were
simplified as a uniformly rigid ring by
applying a reduction factor to the flexural
rigidity (ηEI) of the tunnel linings (Peck
et al 1972; Muir Wood 1975). Lee and Ge
(2001) derived an analytical formulation
of η for the flexural rigidity of segmental
tunnel linings.
The existence of joints between
segments is the main characteristic of
prefabricated linings, because they greatly
affect the bending moment, which is the
dominant factor in segment damage.
Thus, the influence of the segment joints
on the performance of tunnel linings has
attracted the attention of many research-
ers. Some consider the segment joints as
rotational springs, and assume that the
bending stiffness is constant (ITA 2000;
JSCE 2000; Ding et al 2004), which means
the moment-rotation (M-θ) relation is
linear. Subsequently, bilinear and trilinear
models for the structural analysis of lin-
ings (Atsushi 2012) have been proposed to
provide a better approximation.
Combining various linear models, the
following uniform model can be obtained
(see Figure 2):
⎫ ⎭ Nh
⎪
M = k1θ
⎪
⎪
where M ≤ |MT|
⎬(1)
⎪
M = k1θT + k 2(θ – θT)
M = k1θT + k 2(θY – θT) + k 3(θ – θY – θT) ⎪
where |MT| < M ≤ |MY|
⎪
⎭
where M > |MY|
where k 1 is the bending stiffness before
joint opening, k 2 is the bending stiffness
after joint opening, k 3 is the remnant
bending stiffness after plastic deformations
appear in the joint components, MT and
θT are the bending moment and rotation
angle just after joint opening, respectively,
MY and θY are the bending moment and
rotation angle, respectively, just as the joint
components (concrete or bolt) reach the
material strength limit.
In terms of M-θ curves, full-scale joint
tests suggest that the mechanical behaviour
is highly nonlinear (Ding et al 2013; Li et
al 2015; Blom 2000). This is because of
Journal of the South African Institution of Civil Engineering Volume 62 Number 2 June 2020
M
MY
k2
MT
k1
θT θY
Figure 2 Uniform linear model for segment joints
the nonlinearity of the components and
the discontinuity of the joint. The bending
stiffness of the joint decreases significantly
after the joint has opened.
A number of researchers have attempted
to carry out M-θ calculations of seg-
ment joints. This is a critical issue in the
structural analysis of the segment linings.
Janssen (1983) proposed a simple theoretical
model that describes the moment-rotation
behaviour of segment joints in linear materi-
als with full concrete-to-concrete surface
contact. The model represents the joint as
an equivalent concrete beam between two
segments. This concrete element simulates
the rotations in the joint and additional
curvature in the two adjoining segments
caused by a concentrated force introduced
to the segments. The Janssen model can be
written as follows:
M 2N
linear : θ = 12 ,θ< (2a)
Ec h2b Echb
8N
⎫ 2M ⎫
non-linear: θ = , θ ≥ 2N (2b)
9Echb⎪ – 1⎪
⎭ Nh ⎭
Echb
where Ec is the elasticity modulus of con-
crete, N is the axial force, b is the segment
width, and h is the segment thickness.
Under Janssen’s theory or the equivalent
theory developed by Blom (2000), a uni-
fied model involving linear and nonlinear
branches is established. The model can be
divided into three different stages:
k3
θ
1. Constant bending stiffness until
Nh bh2 Ec
M> . This gives cr =  (3)
6 12
 here Ec is the elasticity modulus of
w
concrete, N is the axial force, b is the
segment width, and h is the segment
thickness.
2. Bending stiffness is nonlinear, but the
ultimate compressive strains are in the
elastic branch, until ε = εc , where εc is
the initial plastic strain of concrete.
⎫ 2M ⎫
9bhEcM⎪ – 1⎪
2
⎭
This yields cr = (4)
8N
3. Bending stiffness is nonlinear and the
ultimate compressive strains are in the
plastic branch, until ε = εcu , where εcu is
the yield strain of concrete.
According to elasticity theory, the
compressive stress distribution in the
joint is nonlinear. In reality, the stresses
will reach infinity at the edge of the
contact area of the joint. Gladwell (1980)
developed a relation for the moment-
rotation between two flat surfaces.
The contact stresses are assumed to be
concentrated on the edges of the joint.
The initial stiffness of Gladwell’s model
is higher, and the joint stays closed
for longer than in Janssen’s model.
Gladwell’s model also contains linear and
nonlinear parts:
3
 1 πh2bEc 1
linear : M= , M< (5a)
32 1 – v2 4Nh

1 2b(1 – v 2)
non-linear: M = – ,
2Nh πhEc
1
M> (5b)
4Nh

where Ec is the elasticity modulus of con-

crete, N is the axial force, b is the segment
width, and h is the segment thickness.
In recent years, researchers have
concentrated on developing a better under-
standing of the joint behaviour in the hope
that this will lead to improved compre-
hension of test results and more realistic
analytical and numerical modelling. A pro-
gressive mechanical model for longitudinal
joints was derived by Li et al (2015), who Figure 3 Longitudinal joint sections
divided the duration of joint opening into
four stages. Dong and Xie (2013) proposed
an analytic three-stage model.
The objective of this paper is to inves-
tigate the opening of segment joints with
growing bending moments under different
levels of axial stress, and to propose an
analytical model for predicting the M-θ
behaviour of the joints.

Mechanical analysis of
concrete segment joints
This section considers a joint in a segment
lining connected by two long curved bolts
(see Figures 3 and 4) similar to that used
in Zhengzhou Metro Line 1 – concrete
strength grade C50, anti-seepage grade
P12, main reinforcement HRB335, outer Figure 4 Long curved bolts

d = 180

h = 300
R350

50
50

8 13
8
10

13 8
10
8

32
32

58

58

175 180 180 175

Figure 5 Dimensions of joint

4 Volume 62 Number 2 June 2020 Journal of the South African Institution of Civil Engineering
Table 1 Main physical and mechanical parameters
Parameter
Segment width b
Segment thickness h
Distance from bolt to external edge d
Yield stress of concrete σca
Yield stress of bolt σ ba
Elasticity modulus of concrete Ec
Axial force N
diameter of lining ring 6.0 m, inner
diameter 5.4 m and ring width 1.5 m. The
pre-tightening force of 6.8 grade M27 bolt
is 108 kN, and the distance between the
centre of the bolt hole at the joint and the
inner arc surface is 120 mm (see Figure 5).
The segment joint can be considered as a
semi-rigid connection, as in the design of
numerous frame structures (Xu 2002; Yee
& Melchers 1986; Dias 2014).
A concrete segment joint connected
by two long curved bolts is taken as an
example (Figure 5). In the case of this plane
joint, adjacent segments in a ring are most-
ly in concrete-to-concrete surface contact
without packing material. The contact area
will have a reduced height in comparison
with the segmental thickness (h).
The joint dimensions and material
properties are listed in Table 1.
For simplicity, several basic assump-
tions are needed to evaluate the model
parameters. These assumptions are as
follows:
1. The rotation angle and deformation of
the joint are very small compared with
the size of the segment.
2. The compressive and detached areas of
the joint remain in the plane.
3. The depth of influence of the
compression strain in the compres-
sive zone is twice the height of the
compressive area.
According to these assumptions, the
geometric model of the segment joint can
be simplified into the form illustrated in
Figure 6.
As the bolts are positioned only on
the inner side of the segments, the M-θ
behaviour should be investigated in two
rotational directions (use Figure 6 to define
+ and − moments). The arc structure of
the segment and the bolt position mean
that the joints respond differently under
positive and negative bending moments.
The moment-rotation curve can therefore
Journal of the South African Institution of Civil Engineering Volume 62 Number 2 June 2020
Value Unit
1.5 m the rate of decay of the M-θ curve.
0.3 m
0.18 m the joint, the nonlinear behaviour of the
5.0 × 104 kN/m2
4.8 × 105 kN/m2
3.45 × 107 kN/m2
1.5 × 103 kN/Ring
be divided into two separate branches. In
most analytical models of segment joints,
the arc structure is simplified as a straight
shape, so that only the location of the bolt
distinguishes the joints under positive and
negative bending moments. Therefore, the
parameter calculations in this model are
similar when the joints are under negative
bending moments. In this paper, we con-
sider the positive moment as an example,
and present a mathematical model describ-
ing the M-θ relation.
First, the physical meaning of some
parameters is discussed, as this is critical
to determining the M-θ behaviour of the
segment joint. The initial bending stiffness
of the joint, ki , describes the joint rotation
at the very early stage (i.e. M = θ = 0). The
joint then starts to rotate, but no opening
occurs. As the bending moment increases,
the one edge of the joint becomes divided
at a certain point in time. At this point,
M and θ reach values of Md and θd ,
respectively, and the bending stiffness is
kd . As the bending moment M continues
to increase, it reaches the plastic moment
capacity of the joint, Mp , at which point
the ultimate bending stiffness kp occurs.
These are the critical points of the M-θ
M+
d
h
x
N+
Figure 6 Joint under positive moment
relations, although the intervals between
these critical points are also important.
The parameter C is introduced to control
Regarding the relationship between the
bending moment and the rotation angle at
joint can be analysed using experimental
data. This means that the rotation stiffness
depends on the rotation angle of the joint.
According to the mechanical behaviour
reported by previous studies (Ding et
al 2013; Li et al 2015; Dong & Xie 2013;
Luttikholt 2007), the M-θ curve model
should satisfy the following requirements:
1. When no moment exists in a joint’s
section, no rotation angles will appear
(M = 0 or θ = 0), which means that the
M-θ curves pass through the origin.
2. Differentiating M and substituting for
θ = 0 gives the initial bending stiffness
ki .
3. The bending stiffness k then decreases,
and ki attains a value of kd at the point
at which the segment joint divides.
Thus, differentiating M and substituting
for θ = θd gives kd. This requires that
the M-θ curve is not only continuous,
but also differentiable at θd .
4. After the joint has opened, the bend-
ing stiffness k decreases rapidly. With
the rotation angle increasing, plastic
behaviour appears in one or more
components of the segment joint. The
moment M then increases very slowly
and the bending stiffness k remains
almost constant.
Besides the critical points mentioned
above, there are several desirable aspects
to the model. The slope of the math-
ematical expression of the model should
correspond to the bending stiffness of the
segment joint for any value of θ. Thus, the
Outer face
M+
dc σc
N+
Inner face
5
parameter C describing the decay rate of
the M-θ curve is introduced. Outside
From the experiments and numerical
simulations reported by Luttikholt (2007),
the M-θ curves exhibit exponential-like
N N
decay. As the rotation becomes large, the M M
slope of the M-θ curve approaches the
bending stiffness kp = 0, and the curve
tends asymptotically to the moment
capacity Mp of the joint. However, there Inside
are some differences before the joint opens
to after the joint has opened. These two
states should be considered separately. We
attempt to describe the M-θ behaviour
with exponential-like functions, while
considering that the model should possess
the properties mentioned above. Neutral axis
The mathematical model of the M-θ
curve is assumed to be as follows: Figure 7 Neutral axis is below the inner side

M = Φ1(θ), 0 ≤ θ ≤ θd(6a)
M = Φ2(θ), 0 ≥ θd(6b)
According to the requirements stated above,
nonlinear parts, an exponential-like func-
tion is introduced to reflect the exponential
decay of the M-θ curve.
According to the properties in
Equation 7, the following mathematical
model is derived:
unloaded state until the point of joint open-
ing. During this interval, the bending stiff-
ness of the joint is very large and decreases
only slightly. As the neutral axis is located
on the lower inside edge of the joint, i.e. the
moment when the joint opens is:

⎫
Model 6 should have the following properties:

⎪M =
x(3h – 2x)bEcθ Nh
⎪
, 0 ≤ θ ≤ θd Md =

⎫ ⎪
24 6 (9)

⎪ ⎪
⎬ M = (M – M ) ⎫⎬1 – exp
M=0 θ=0
⎪
⎪ ⎭
θ=0 ⎪
(8) it is difficult to determine the displacement

⎪ ⎪
p d

⎪ – (ki – kp + Cθ̂)θ̂ ⎫ + k θ̂ + M ,
dΦ1 ∆c of the concrete on the joint’s upper

⎪ ⎬
= ki
⎪ ⎭
⎪
dθ edge. According to Saint-Venant’s principle,
⎪
⎪
p d

⎭ θ ≥ θd
Mp – Md beyond a certain distance (lcc) from the
θ = θd ⎪
dΦ1 compression area, the stress becomes uni-
⎪
= kd

⎬
dθ form and uniaxial over the section profile

⎪
(7) where θ̂ = θ – θd , x is the height of the F

θ = θd ⎪
(Collins & Mitchell 1991), i.e., σz = c
dΦ2
⎪
= kd compressed region (see Figure 6), Md is bh

⎪
dθ the bending moment at the point at which and σ y = 0. Ding et al (2013) investigated

⎪
θ∞ ⎪
⎪
dΦ2
= kp
⎪
dθ
⎪
θ∞ ⎭
M  Mp
As shown in Figure 7, when the neutral axis
is below the inner side of the segment joint,
joint opening does not occur, and the whole
section of the joint bears the compressive
stress. Therefore, the bending stiffness of
the joint is extremely large in this situation,
and the moment mainly compensates the
axial forces. The rotation angle caused
by the compressive deformations is quite
small. Additionally, the critical issue of
the segment joint’s mechanical behaviour
needs only be considered after the joint has
opened. In conclusion, the first equation of
the mathematical model is assumed to be
linear. Thus, in Equation 7, ki = kd . For the
the joint opens, Mp is the plastic moment the value of lcc by means of finite element
capacity of the joint, C is a parameter analysis. In this study lcc is assumed to be
describing the decay in the bending stiff- twice the height of the compression area
ness, ki is the initial bending stiffness, and for simplicity (Atsushi 2012). Thus, the
kp is the ultimate bending stiffness. displacement ∆c of the upper outer edge
Model 8 satisfies the properties described has the following expression
at the beginning of this section. If the model
parameters can be evaluated exactly, the σc h
∆c = εc h = (10)
model will qualitatively agree with experi- Ec
mental results in terms of the M-θ relation.
In the next section, we concentrate on evalu- where εc and σc represent the strain and
ating the parameters in Model 8. stress of the upper edge, respectively. The
rotation angle can be calculated by
Evaluation of model θd =
∆c
=
σc
(11)
parameters h Ec
Evaluation of ki and so ki can be expressed as
The initial bending stiffness ki is considered
to be a uniform bending stiffness (secant Md NhEc
ki = = (12)
slope). Hence, ki will be calculated from the θd 6σc

6 Volume 62 Number 2 June 2020 Journal of the South African Institution of Civil Engineering
 1 N bh
Because N = hbσc , we have that = .
2 σc
Thus, the expression for ki can be rewritten as
bh2 Ec
ki = (13)
12
When M = Md , the rotation angle θd is
given by
2N
θd = (14)
bhEc
Therefore, we can distinguish two stages
in which the rotation angle θ of the joint
2N
increases: θ ≤ , when the joint is
bhEc
2N
always closed, and θ > , when the joint
bhEc
becomes divided along the inner edge.
Evaluation of Mp
As the bending moment increases, the
components of the joint generate plastic
deformations. The joint then develops into
a plastic hinge. It is the bending moment
M that transforms the joint into a plastic
hinge. Therefore, the moment capacity
of the joint, Mp , is a critical parameter in
Model 8. Mp is the maximum moment that
can be transmitted by the segment joint
without strain hardening, and is governed
squashing near the bolt pocket, and to the stress–strain relationship of the
2 Equation 16c represents concrete squash- segment concrete and the M-θ curves
ing of the compression zone. Calculations derived from previous full-scale tests,
indicate that concrete compression zone the M-θ curve of the joint is asymptotic
squashing will occur first, which is in with respect to Mp . In this case, a plastic
accordance with the results of full-scale hinge exists, i.e. plastic deformation of the
tests. Actually Mp can be calculated with concrete compression zone occurs, and
the help of the following equations: the bending stiffness of the joint tends
to k p .
Fbolt = (d – x)θk j(17a)
Calculation results of model
1 parameters
Fc = bxσca(17b)
2 Using the parameter values given at the
beginning of this section, the parameters
lσca of the M-θ model can be calculated. The
xθ = (17c)
Ec results are presented in Table 2.
Setting the rate of decay parameter
Fc – Fbolt = N(17d) C to values of 1 and 100, and using the
parameters in Table 2, the M-θ curve
where kj is the tension spring coefficient of can be plotted (see Figure 8). The tension
the segment joint. From Equation 17, the spring coefficient of the segment joint, kj , is
height of the compression area x can be approximately equal to the tension spring
obtained from: coefficient of the bolts, kb . Therefore, k j can
be calculated as (Atsushi 2012):
1 2σca
bxσca – (d – x) k j = N(18)
2 Ec Table 2 Calculation results of model parameters
Item Values Units
The distance x and moment capacity Mp
can then be calculated. The calculation Md 75 kN ∙ m
process is described in the Appendix at the θd 3.86 × 10 –4 rad
end of this article.
ki 1.94 × 105 kN ∙ m/rad

by the failure of the weakest component, Evaluation of kp x 5.0 × 10 –2 m

such as the bolts, bolt pockets, and the con- As strain-hardening of the joint is not
Mp 2.61 × 102 kN ∙ m
crete section of the joint. This parameter considered in this study, k p = 0. According
can be calculated as follows:

⎫ h⎫ ⎫h x ⎫
Mp = Fbolt⎪d – ⎪ + Fc⎪ – ⎪(15)
⎭ 2⎭ ⎭2 3⎭
M–q curves of segmental joint
350

The moment capacity Mp depends on the 300 c = 100

strength and size of the components of the
c=1
joint connected by two bolts. When one or
250
more of the following component failures
appear, the joint will lose its ability to sus-
Moment (kN*m)

tain further loads. 200

Fbolt = 2σba Ab(16a)

150
Fbolt = 2σca Ap(16b)
100
1
Fc = σcabx(16c) Joint open
2
50
where σba is the yield stress of bolt, σca is
the yield stress of concrete, Ab is the cross-
0
sectional area of the bolt, Ap is the bearing 0 1 2 3 4 5 6 7
area of the nut and washer on the concrete, Rotation angle (10 –3rad)
Equation 16a indicates bolt failure (in
tension), Equation 16b represents concrete Figure 8 Curves showing the M-θ relations

Journal of the South African Institution of Civil Engineering Volume 62 Number 2 June 2020 7
 2.5 1 500

2.0
ki (kN*m/rad) (× 106)

1 000
1.5

Mp (kN*m)
1.0
500

0.5

0 0
0 0.2 0.4 0.6 0.8 1.0 0 2 000 4 000 6 000 8 000 10 000
h (m) N (kN/ring)

Figure 9 Curve showing the ki – h relation Figure 10 Curve showing the N – Mp relation

E b Ab
kj = = 9.54 × 105 kN ∙ m(19)
Lb
Jack
where Eb is the elasticity modulus of the Jack

bolts, Ab is the useful cross-section of

the bolts, and Lb is the useful length of Pressure
the bolts. box Pressure
box

Model validation
The proposed model of this paper is
mainly dominated by the parameters of
critical points Md , Mp and the parameter
of decay rate C. The physical behaviour Concrete segment
of the segment joints with an increasing
bending moment, can also be reflected by
these parameters. The bending moment
Md describes the behaviour of the joint
opening, and Mp describes the situation
where the components of the joint perform
plasticity. The parameter C describes the
decay rate of the M-θ relation. Therefore,
when using this model, the parameters Md ,
Mp , and C should be calculated, so that the
model reflects the real physical behaviour
of the segment joints.
Concrete segment
Effects on model of parameters
Evaluations indicate that the model
parameters are closely related to the axial
force N and the intrinsic parameters of the
segment (such as the segment dimensions
b, h, d and the material properties Ec , σca).
Subsequently, Figure 8 was obtained, which
shows the M-θ relation. Figure 11 Schematic overview of the test set-up
The following paragraphs discuss the
main factors that influence the model increasing axial force N. This is because the deformation ∆c will enlarge θd . Additionally,
parameters. According to Equations 9, 13 bending moment mainly counteracts the Md will grow with the segment thickness
and 14, at the point at which the segment effects of N on the joint section. Thus, as h, and the rotation angle θd will decrease as
joint opens, the bending moment Md and N increases, a larger value of Md is needed the segmental size parameters (h, d) and the
the rotation angle θd both increase with to initiate joint opening, and the increased elasticity modulus Ec increase.

8 Volume 62 Number 2 June 2020 Journal of the South African Institution of Civil Engineering
 joint Mp will increase with the growth in
250 N. Figure 10 shows that Mp and N have an
approximately linear relationship, because
2Ec (the coefficient of N in Equation A4) is
small compared with bσca Ec + 4σca k j .
200

Comparison with other analytic

models and test results
150 The parameter C determines the rate of
Moment (kN*m)

decay of the M-θ curve (Maquoi & Jaspart

1987). In this study, we compare the effects
of C = 1 and C = 100.
100 Hordijk & Gijsbers (1996) conducted
experiments on segment joints without
packing materials. The specimens used
Test in the tests had a segmental thickness of
50 Model (closed) 350 mm and a segmental width of 500 mm.
Model (open) The contact height of the joint was
Janssen 158 mm. Two tunnel segments were loaded
Gladwell by increasing the bending moment under
0
0 0.001 0.002 0.003 0.004 0.005 0.006 0.007 0.008 0.009 0.01 various normal forces. The influence of the
Rotation angle (–) bolts on the rotation capacity of the joints
was also considered. A schematic overview
Figure 12 C
 omparison of proposed model with the test results and previous models (Janssen’s of the test set-up is shown in Figure 11.
and Gladwell’s) for axial force of N = 2.2 × 103 kN/m Under different axial forces, the model
proposed in this paper is compared with
the test results reported by Hordijk &
250 Gijsbers (1996), Janssen’s model, and
Gladwell’s model.
From Figures 12 – 16 it is clear that
the moment-rotation curves exhibit linear
200
and nonlinear branches. Gladwell’s model
always overestimates the bending moments
in the joint; Janssen’s model underestimates
150 the bending moments under higher axial
Moment (kN*m)

forces. Under lower axial forces, Janssen’s

100
50
0
0 0.001 0.002 0.003
Figure 13 C
 omparison of proposed model with the test results and previous models (Janssen’s
and Gladwell’s) for axial force of N = 1.5 × 103 kN/m
The initial bending stiffness k i will
become larger as the segmental size
parameters (h, d) and elasticity modulus
Ec increase. Consequently, the straight
part of the M-θ curve will extend as the
axial force N increases, i.e. k i is positively
related to bh2 Ec . Therefore, the segment
thickness h has a significant impact
model predicts larger bending moments
before the joint opens. For the Janssen’s
and Gladwell’s models, axial force N has
no effect on the initial bending stiffness ki ,
which is not in conformity with real behav-
Test iour of segment joints. From these com-
Model (closed) parisons, it appears that our model predicts
Model (open) the M-θ behaviour more accurately than
Janssen Janssen’s and Gladwell’s models.
Gladwell The axial force N has significant effects
0.004 0.005 0.006 0.007 0.008 0.009 0.01 on the M-θ relation of the joint. The initial
Rotation angle (–) bending stiffness ki increases with N, espe-
cially when N is relatively low. The plastic
moment capacity of the joint Mp grows
under larger values of N.
The N – Mp relation is almost linear
on the initial bending stiffness k i (see (see Figure 17), which is in accordance with
Figure 9). the proposed model.
Thus, the axial force N must exceed a
magnitude of 108 kN/ring for x to reach
the axis of symmetry (X = 11 250; see Conclusions
Equation A5 in the Appendix). Until the Bolted connections in segmental linings
axial force N reaches that order of magni- were regarded as semi-rigid connections in
tude, the plastic moment capacity of the establishing a physically based mathematical

Journal of the South African Institution of Civil Engineering Volume 62 Number 2 June 2020 9
 based parameters. With regard to critical
180 points of the M-θ curve, the parameters
Mp and Md accurately reflect the mechani-
160 cal behaviour of the joints. With the help of
a decay parameter C, the proposed model
140 was shown to be consistent with experi-
mental test data.
120
Moment (kN*m)

100 ACKNOWLEDGEMENTS
This work is supported by the National
80 Natural Science Foundation of China
(Grant Number 51278467), the National
60 Natural Science Foundation of China
(Grant Numbers 2015M582204 and
Test 2016T90681), the Program for Science
40
Model (closed) and Technology Innovation Talents at
Model (open) the University of Henan Province (Grant
20
Janssen Number 14HASTIT050), and the Special
Gladwell Scientific Research Foundation for Young
0
0 0.001 0.002 0.003 0.004 0.005 0.006 0.007 0.008 0.009 0.01 Teachers of Zhengzhou University (Grant
Rotation angle (–) Number 1421323078).

Figure 14 Comparison of proposed model with the test results and previous models (Janssen’s
and Gladwell’s) for axial force of N = 1.15 × 103 kN/m REFERENCES
Atsushi, K 2012. The segment design of shield
tunnelling. Beijing: China Architecture and Building
120 Press (Chinese Edition).
Blom, C 2000. Design philosophy of concrete linings for
tunnels in soft soils. PhD thesis. Delft, Netherlands:
100 Technical University of Delft.
Collins, M & Mitchell, D 1991. Prestressed Concrete
Structures. Upper Saddle River, NJ: Prentice Hall.
80 Dias, D 2014. A new numerical approach to the
hyperstatic reaction method for segmental tunnel
Moment (kN*m)

linings. International Journal for Numerical

60 and Analytical Methods in Geomechanics,

38: 1617–1632.
Ding, W, Peng, Y, Yan, Z et al 2013. Full-scale testing

40 and modeling of the mechanical behavior of shield

20
0
0 0.001 0.002
Figure 15 C
 omparison of proposed model with the test results and previous models (Janssen’s
and Gladwell’s) for axial force of N = 0.8 × 103 kN/m
model expressed by the joint dimensions
and material properties. This model allows
the M-θ behaviour of segment joints to be
predicted.
The model parameters were evaluated,
and M-θ curves for the joint were presented.
The calculation process allowed us to iden-
tify the key factors that influence the model
TBM tunnel joints. Structural and Mechanical
Test Engineering, 3: 337–354.
Model (closed) Ding, W, Yue, Z, Tham, L, Zhu, H, Lee, C, &
Model (open) Hashimoto, T 2004. Analysis of shield tunnel.
Janssen International Journal for Numerical and Analytical
Gladwell Methods in Geomechanics, 28: 57–91.
0.003 0.004 0.005 0.006 0.007 0.008 0.009 0.01 Dong, X & Xie F 2013. Analytical solution of segment
Rotation angle (–) joint model for segmented tunnel lining. Chinese
Journal of Geotechnical Engineering, 35: 1870–1975.
Gladwell, G M L 1980. Contact problems in the
classical theory of elasticity. Alphen aan den Rijn,
Netherlands: Springer Science & Business Media,
parameters. The segment thickness h has p 5.
a significant impact on the initial bending Hordijk, D & Gijsbers, F 1996. Laboratoriumproeven
stiffness ki . As h increases, the initial bend- tunnelsegmenten [Laboratory experiments with
ing stiffness ki also increases. The plastic tunnel lining segments]. CUR/COB. Rotterdam,
moment capacity of the joint increases Mp Netherlands: Projectbureau boortunnels.
approximately linearly with increases in N. Iftimie, T 1994. Prefabricated lining, conceptional
The critical mechanical behaviour of analysis and comparative studies for optimal
the model was revealed by the physically solution. Proceedings, ITA International Congress

10 Volume 62 Number 2 June 2020 Journal of the South African Institution of Civil Engineering
 Luttikholt, A 2007. Ultimate limit state analysis of
70 a segmented tunnel lining. Master’s thesis, Delft,
Netherlands: Technical University of Delft.
Maquoi, R & Jaspart, J 1987. Discussion: Moment
60
rotation curves for bolted connections. Journal of
Structural Engineering, 113: 2324–2327.

50 Muir Wood, A 1975. The circular tunnel in elastic

ground. Geotechnique, 25: 115–127.
Peck, R, Hendron, A & Mohraz, B 1972. State of the
Moment (kN*m)

40 art of soft round tunneling. Proceedings, 1st Rapid

Excavation and Tunneling Conference, Chicago, IL,
Vol 1, pp 259–286.
30
Xu, L 2002. Design and optimization of semi-rigid
framed structures. Chapter 7, in Burns, S A. Recent

20 Advances in Optimal Structural Design, Reston, VA:

Test American Society of Civil Engineers.
Model (closed) Yee, Y & Melchers, R 1986. Moment-rotation curves
10 Model (open) for bolted connections. Journal of Structural
Janssen Engineering, 112: 615–635.
Gladwell
0
0 0.001 0.002 0.003 0.004 0.005 0.006 0.007 0.008 0.009 0.01
Rotation angle (–) APPENDIX
From Equations 15, 17a, 17b and 18 we obtain:

⎫ h⎫ ⎫ h⎫
Figure 16 C
 omparison of proposed model with the test results and previous models (Janssen’s
Mp = d ⎪d – ⎪k jθ + bhσca – ⎪d – ⎪
1
⎭ 2⎭ ⎭ 2⎭
and Gladwell’s) for axial force of N = 0.45 × 103 kN/m
4
1 2
k jθ X – bhσca X (A1)
250 6

2NEc + 4σca k jd
X= (A2)
bσcaEcd + 4σcakj
200

Substituting Equation A1 into A2 yields:

⎫ h⎫
Mp = d ⎪d – ⎪k jθ – bσca
1
⎭ 2⎭
150
Mp (kN*m)

6
⎫ 2NEc + 4σca k jd ⎫
⎪ ⎪ +
2
1
⎭ ⎭
bhσca
100 bσ E + 4σ k 4
⎫ h⎫
ca c ca j

– ⎪d – ⎪k jθ
2NEc + 4σca k jd
⎭ 2⎭
(A3)
bσca Ec + 4σca kj
50
2σca
Here θ = , so:
Ec

⎫ h⎫ 2 1
Mp = d ⎪d – ⎪k j – b
0

⎭ 2 ⎭ Ec 6
0 0.5 1.0 1.5 2.0 2.5
N (*103 kN/m)

⎫ 2NEc + 4σca k jd ⎫
⎪ ⎪ +
2
1
⎭ ⎭
Figure 17 Influence of N on Mp bhσca
bσ E + 4σ k 4
⎫ h ⎫ 2 2NEc + 4σca k jd
ca c ca j

on Tunnelling and Ground Conditions, Cairo,
Egypt, pp 136A–137A.
ITA (International Tunnelling Association) 2000.
Guidelines for the design of shield tunnel lining.
Tunnelling and Underground Space Technology,
15: 303–331.
Janssen, P 1983. Tragverhalten von tunnelausbauten
mit gelenktubbings (in German). PhD thesis,
Braunschweig, Germany: Technischen Universitat
Carolo-Wilhelmina.
– ⎪d – ⎪k j
⎭ 2 ⎭ Ec bσca Ec + 4σca kj
JSCE (Japanese Society of Civil Engineers) 2000. The
(A4)
design and construction of underground structures
(in Japanese). Tokyo, Japan: JSCE.
Lee, K & Ge, X 2001. The equivalence of a jointed shield- Then:
driven tunnel lining to a continuous ring structure.
Canadian Geotechnical Journal, 38: 461–483. Mp = 2 .99 × 10 –4 + (5 625 – 1.66
Li, X, Yan, Z, Wang, Z & Zhu, H 2015. Experimental × 10 –3)X – 0.25X2(A5)
and analytical study on longitudinal joint opening
of concrete segmental lining. Tunnelling and
Underground Space Technology, 46: 52–63.

Journal of the South African Institution of Civil Engineering Volume 62 Number 2 June 2020 11
TECHNICAL PAPER Laboratory and field
Journal of the South African
Institution of Civil Engineering
investigations in granular
ISSN 1021-2019
Vol 62 No 2, June 2020, Pages 12–21, Paper 0516 soils to correlate relative
DR HASSAN MUJTABA is serving as Associate density, relative compaction
and grain size
Professor in the Civil Engineering Department at the
University of Engineering and Technology, Lahore,
Pakistan. Dr Mujtaba obtained his BS, MS and PhD in
Civil Engineering from the same university. He is a
registered Professional Engineer and has more than 15 years’ professional
experience in the field of geotechnical engineering. He has authored, presented
and published more than 30 technical papers in international/ national journals H Mujtaba, K Farooq, N Sivakugan, B M Das
and conferences. His research interests are correlations between cohesive and
non-cohesive soils, and problematic soils and their mitigation techniques.
Contact details
An attempt has been made to correlate relative density and relative compaction based on
Associate Professor, Department of Civil Engineering
University of Engineering and Technology laboratory and field test data. In this investigation, 185 sandy soil samples were tested in the
Lahore, Pakistan laboratory. The experimental investigations include classification tests, maximum and minimum
T: +92 32 19431303, E: hassanmujtaba@uet.edu.pk
density tests, and standard and modified Proctor and in-situ density tests. The values of the dry
DR KHALID FAROOQ is Professor in the Department of unit weight of these samples obtained by performing different tests fall between 14.7 kN/‌m3
Civil Engineering at the University of Engineering and and 20.8 kN/‌m3. Based on these results, linear and multivariate regression analyses were
Technology, Lahore, Pakistan, and also the
Chairperson of the Department. Dr Farooq completed
carried out to (a) relate relative compaction and relative density, (b) relate maximum (emax) and
his Bachelor and Master’s at the same university, and minimum void ratios (emin), and (c) express emax and emin in terms of median grain size (D50)
obtained his PhD in Civil Engineering from the University of Tokyo, Japan. He is and uniformity coefficient (Cu). Experimental and predicted values varied ± 5%, with a 95%
a registered Professional Engineer and has more than 25 years’ professional
experience in the field of geotechnical engineering. He has published more confidence interval for the relation between relative compaction and relative density, and for
than 65 technical papers in international/national journals and conferences. His other relations the variation was ± 10%. The proposed equations were validated using a new
research interests are: rain-induced landslides, problematic soils and mitigation data set which had not been used for the development of the correlations. Proposed equations
techniques, soil improvement and ground characterisation. He provides
substantial consultancy services for geotechnical engineering in Pakistan. were also compared with equations presented by various other researchers. Validation of the
Contact details proposed equations suggests that these equations may be used for future prediction of the
Professor and Chairperson, Department of Civil Engineering respective variables.
University of Engineering and Technology
Lahore, Pakistan
T: +92 42 99029202, E: kfch@uet.edu.pk
INTRODUCTION Relative density (Dr ) represents the degree
NAGARATNAM SIVAKUGAN is a Chartered
Field compaction is an integral part of of compaction of soil with respect to the
Professional Engineer, and a Fellow of both the
American Society of Civil Engineers and Engineers every civil engineering construction most dense and loosest states as measured
Australia. He graduated with First Class Honours in general and for the construction of in the laboratory. Dr as defined by the
from the University of Peradeniya, Sri Lanka, and
MSCE and PhD degrees from Purdue University, USA. His writings include
earth structures in particular. Relative American Society for Testing and Materials
eight books, 150 journal papers and 100 conference papers in conference compaction and relative density are two (ASTM 2007), is evaluated by Equation 2.

γdmax ⎫ γdf – γdmin ⎫

proceedings. He does substantial consulting work for geotechnical and parameters which can be used alternately

⎪ ⎪(2)
mining companies in Australia and internationally.

γdf ⎭ γdmax – γdmin ⎭

when writing technical specifications Dr =
Contact details
Associate Professor, College of Science and Engineering for field compaction control. In order to
James Cook University ascertain the degree of compaction in
Townsville, QLD 4811, Australia the field, usually relative compaction (Rc) Where: γdmax and γdmin are determined
T: +61 7 47814431, E: siva.sivakugan@jcu.edu.au
is determined. The concept of relative through laboratory tests (ASTM D-4253
DR BRAJA M DAS is Professor and Dean Emeritus density (Dr) has been found to be a more and ASTM D-4254 respectively) and γdf are
at California State University, Sacramento,
convenient basis for correlating many applicable to granular soils having fines up
California, USA. He is a Fellow and Life Member of
the American Society of Civil Engineers, Life granular soil properties, such as shear to 15%, and is the dry unit weight of the
Member of the American Society of Engineering strength, compressibility and liquefaction compacted soil in the field. However, Lee
Education, and Emeritus Member of the Committee on Stabilization of
potential of saturated sands (Lee & Singh and Singh (1971) defined maximum dry
Geomaterials and Recycled Materials of the Transportation Board,
Washington, DC. He has authored several text and reference books in the 1971). Relative compaction is defined as unit weight (γdmax) as the dry unit weight
field of geotechnical engineering, and has published over 300 research the ratio between the dry unit weight in of the material when arranged in the most
papers in journals and conference proceedings.
the field (γdf ) and the maximum dry unit compact state possible by any practical
Contact details
Professor and Dean Emeritus, California State University
weight (γdmax) that can be obtained in the engineering methods without significantly
2689 Chateau Clermont Street, Henderson, Nevada 89044 laboratory for a soil following a specified altering the grain size distribution, and
United States of America compaction procedure, and is mathemati- minimum dry unit weight (γdmin) as the
T: +1 702 616 2161, E: brajamdas@gmail.com
cally given by Equation 1. unit weight of the material when arranged
in the loosest state possible by any practical
Keywords: relative density, relative compaction, void ratio, uniformity
γdf engineering methods without the use of
coefficient, curvature coefficient, effective grain size, median Rc = (1)
grain size γdmax natural or artificial methods to increase

Mujtaba H, Farooq K, Sivakugan N, Das BM. Laboratory and field investigations in granular soils to correlate relative density, relative compaction
12 and grain size. J. S. Afr. Inst. Civ. Eng. 2020:62(2), Art. #0516, 10 pages. http://dx.doi.org/10.17159/2309-8775/2020/v62n2a2
the adhesion between particles and lead to
bulking. Based on these definitions, they
assumed that the values γdmax in Equations
1 and 2 are the same for a granular soil
irrespective of the test method being
adopted to determine γdmax . They also
proposed a correlation between relative
density and relative compaction for granu-
lar soils varying from silty sand to coarse
gravels as given in Equation 3.
Rc (%) = 0.2Dr (%) + 80
Both relative density and relative compac-
tion theoretically range from 0% to 100%.
Zero percent relative compaction refers
to zero dry density, which is physically
meaningless. Zero percent relative density
refers to a real and substantial value of
dry unit weight. Thus, the term relative
density covers a narrower range of possible
numerical values than relative compaction,
and is therefore considerably more sensitive
to small changes in actual dry densities.
Relative compaction is an indicator wheth-
er the required degree of field compaction
has been achieved or not, as per project
specifications. The concept of relative com-
paction is applicable to both cohesive and
non-cohesive soil, while relative density
(Dr ) is more appropriate for end-product
compaction specification for granular soils.
Tavenas and Rochelle (1972) suggested that
it is difficult to ensure that maximum and
minimum dry unit weights (determined by
experiments) correspond effectively to the
maximum and minimum possible state of
compactness of the material considered. If
Dr is negative, a collapsible structure may
be present, such as can occur with honey-
combed soils or loose cemented or calcare-
ous sand where in-situ dry unit weight
is smaller than the minimum dry unit
weight of the soil encountered (Kulhawy &
Mayne 1990).
The determination of in-situ void ratio
of sandy soils is very difficult, because to
obtain undisturbed samples in sandy soils
using conventional methods is not possible.
Equation 4 can be used to estimate the
in-situ void ratio from the results of relative
compaction.
⎫G γ ⎫
ef = ⎪ s w ⎪ – 1 (4)
⎭ γdf ⎭
Sand has particles varying in size between
4.75 mm to 0.075 mm (ASTM D 2487). The
behaviour of sandy soils depends largely
on grain size distribution, grain shape,
Journal of the South African Institution of Civil Engineering Volume 62 Number 2 June 2020
angularity, mineralogy, relative density,
mode of deposition and current state of the
deposit in which it is present. Cubrinovski
and Ishihara (2002) observed that the
generalised charts which are used to
estimate the response of sandy soils based
on material parameters use either fine
content (percentage finer than 0.075 mm,
ASTM D 2487) present in the sample and/
or median grain size (D 50). In these charts,
sand with the same amount of fines has
(3) been assumed to have similar response to
any loading condition, but practically it is
not true, as response depends mainly upon
grain size distribution, shape, mineralogi-
cal composition, etc. However, they showed
that void ratio range (emax – emin) may
be a better parameter than considering
fines, uniformity coefficient, curvature
coefficient, etc. The Japanese Geotechnical
Society (JGS 2000) procedure can be
employed to determine the void ratios
for clean sands having less than 5% fines.
However, Cubrinovski and Ishihara (2002)
revealed that the JGS procedure can be
applied to material having fines up to 30%.
Lade et al (1998) summarised the studies
of Graton and Fraser (1935), White and
Walton (1937) and McGeary (1961), and
presented five possible modes of packing
single-size spherical particles, along with
the void ratio for each type of packing.
They concluded that maximum and mini-
mum void ratios are independent of the
size of the sphere. Lade et al (1998) studied
the behaviour of poorly graded Cambria
sand mixed with non-plastic Nevada fines.
They concluded that both emax and emin
decrease as the fines increase from 0% to
20%. Within the range of 20% to 40% fines,
the relationships show a shift in pattern
indicating a transition from the filling of
voids to the replacement of solids. After
40%, the void ratio values tend to increase
steadily until they reach the highest value
at 100% fines. It should be mentioned that
the variation of emin with fines closely
resembles that of the packing of spheres.
The reason for this similarity is that a
sample consisting of a mixture of sand and
fines, and a sample consisting of spheres
having large and small diameters are both
gap-graded. Cubrinovski and Ishihara
(2002) studied the link between void ratio
range and fines content. As far as clean
sand, or sand with fines, is concerned,
there is an increase in void ratio range as
fines increase from 0% to 30%. However,
after 30% fines, there is a change of pattern
in the relationship between the void ratio
range (emax – emin) and the fines (Fc ). For
Fc < 30%,
emax – emin = 0.43 + 0.00867Fc (%)(5)
Equation 5 applies to situations where any
addition of fines fills the voids. If Fc < 30%,
addition of fines leads to replacement of
larger particles. Here,
emax – emin = 0.57 + 0.004 Fc (%)(6)
The reasons for the change in the behav-
iour of sand with 30% fines, compared
to sand having 0% to 20% fines, is that,
for sands having 30% fines, the fines are
controlling the grain fraction in the soil
structure, whereas for sands having 0%
to 20% fines, the packing is controlled by
sand grains. However, the role of fines in
the very dense packing state is different
from the role of fines in the loosest state
of packing.
Youd (1973) presented a relationship for
emax and emin as a function of roundness
(R), i.e. shape of the particle and uniformity
coefficient (Cu ). Roundness (R) is the
ratio of the average radius of the corners
and edges to the radius of the maximum
inscribed sphere. These relations are valid
for normal to moderately skewed grain size
distribution curves. Shimobe and Moreto
(1995) determined the variation of emax
with (R) for 40 uniform clean sands having
a uniformity coefficient less than or equal
to 2 (Cu ≤ 2). The experimental range of
emax with R is given by Equation 7.
emax = 0.642R–0.354(7)
Miura et al (1997) carried out research
on the physical characteristics of about
200 samples of granular materials, which
included clean sand, glass beads and light-
weight aggregates. Based on the data avail-
able, linear regression analysis was carried
out and the following equation (Equation 8)
was proposed:
emax = 1.62emin(8)
Cubrinovski and Ishihara (2002) collected
data consisting of over 300 sandy soil
samples including clean sand, clean sand
with fines, and silty sands, and divided
them based on the amount of fines (Fc ) and
clay size content (Pc ). Fc represents grain
sizes smaller than 0.075 mm, whereas Pc
is the percentage of grains smaller than
0.005 mm. To investigate the effect of
13
fines on emax and emin , regression analysis
was carried out on all four groups and
equations, i.e. Equations 9 to 12 (one for
each group). The strength of relationship
between the two variables is specified
statistically by the correlation coefficient
(r), and its value is .97, .94, .96 and .90 for
Equations 9 to 12, respectively.
a. Clean sand (Fc = 0–5%)
emax = 0.072 + 1.53emin(9)
b. Sand with fines (5 < Fc ≤ 15%)
emax = 0.25 + 1.37emin(10)
c. Sand with fines and clay (15 < Fc ≤ 30%,
Pc = 5–20%)
emax = 0.44 + 1.21emin(11)
d. Silty soils (30 < Fc ≤ 70%, Pc = 5–20%)
emax = 0.44 + 1.32emin(12)
Cubrinovski and Ishihara (2002; 1999),
based on a large database, developed a
unique relationship between void ratio
range (emax – emin) and mean grain
size (D 50 ). This relationship is given by
Equation 13 as:
0.06
emax – emin = 0.23 + (13)
D 50 (mm)
They also proposed the upper and lower
limits of (emax – emin) as a function of D 50
which can be approximated by Equations
14 and 15, respectively.
Upper limit:
0.045
emax – emin = 0.16 + (14)
D 50 (mm)
Lower limit:
0.079
emax – emin = 0.29 + (15)
D 50 (mm)
Patra et al (2010) conducted Proctor com-
paction tests (standard, modified Proctor
following ASTM D 698 and ASTM D 1557
along with reduced standard, reduced
modified compaction tests by altering
the number of blows per layer using the
relevant Proctor hammer) on 55 clean sand
samples – most of which were poorly grad-
ed. They also determined emax and emin
based on ASTM procedures. Compaction
tests including standard Proctor
(compaction energy = 600 kN‑m/‌m3),
reduced Proctor (compaction energy =
360 kN‑m/‌m3), modified Proctor (compac-
tion energy = 2 700 kN‑m/‌m3) and reduced
14
modified Proctor (compaction energy =
1 300 kN m/‌m3) were also carried out on
these samples. Based on the laboratory
tests, emax and emin and void ratio at maxi-
mum dry unit weight determined through
respective compaction tests, as mentioned
above, are expressed as a function of D 50 in
Equations 16 to 21. Correlation coefficients
of these equations are .87 ~ .92.
emax = 0.6042D 50 –0.304(16)
emin = 0.3346D 50 –0.491(17)
es = 0.4484D 50 –0.356(18)
em = 0.3825D 50 –0.04(19)
ers = 0.5039D 50 –0.327(20)
erm = 0.4087D 50 –0.389(21)
Where: D 50 is in mm, es = void ratio at
maximum dry unit weight from standard
Proctor, em = void ratio at maximum dry
unit weight from modified Proctor, ers =
void ratio at maximum dry unit weight
from reduced standard Proctor, erm = void
ratio at maximum dry unit weight from
reduced modified Proctor.
McCook (1996) developed the correla-
tion between a one-point standard Proctor
test performed on air-dry sand and a
relative density test for 29 filter sands. The
basic purpose was to relate the maximum
density test with an easy-to-perform field
test to determine the dry densities for
clean granular sands. He presented two
equations to estimate 50% and 70% relative
density from a one- point field test. These
equations are presented here by Equations
22 and 23. The correlation coefficient (r) is
.97, with a standard error of estimate (SEE),
i.e. the square root of average squared
deviation is 0.3 kN/‌m3 for Equation 22, and
0.28 kN/‌m3 for Equation 23.
γDr =50% = 1.07γd1pt – 1.96 (22)
γDr =70% = 1.073γd1pt – 1.484 (23)
Where: γDr =50% is the dry unit weight at
50% relative density, γDr =70% is the dry unit
weight at 70% relative density and γd1pt is
the field air-dry unit weight measured in
kN/‌m3 by compacting air-dry sand in a
standard Proctor mould having a volume
of 944 cm3, using three lifts, with a 2.5 kg
hammer dropped 305 mm for 25 blows
per lift.
Volume 62 Number 2 June 2020
Mujtaba and Farooq (2010) carried
out index density tests and compaction
tests (both standard Proctor and modified
Proctor) on sandy soils. Based on the com-
paction test results, they calculated relative
compaction separately for standard Proctor
and modified Proctor, and correlated
relative compaction with relative density.
The two equations proposed by them for
standard Proctor and modified Proctor,
respectively, are presented in Equations 24
and 25.
Rc = 0.13Dr (%) + 86.5 (24)
Rc = 0.13Dr (%) + 79.4 (25)
Based on the above preview, it can be
inferred that grading parameters includ-
ing uniformity coefficient, roundness,
grain size distribution and fines have a
significant influence on the maximum and
minimum void ratios of granular soils in
combination with mineralogical composi-
tion, mode of deposition and origin of the
respective soils. Also, there is a need to
correlate relative density and relative com-
paction, irrespective of the errors involved
in measuring maximum and minimum dry
unit weight for granular soils.
IN-SITU AND LABORATORY TESTS
The test results of the 185 samples, includ-
ing natural and reconstituted samples
belonging to classification groups SP,
SP-SM, SW and SW-SM, were used for
the development of correlations between
relative compaction and relative density,
maximum and minimum void ratios, and
gradational parameters versus void ratios.
The selected sand samples were subjected
to the following tests following ASTM
standards (ASTM 2007):
QQ Grain size analysis (ASTM D-422)
QQ Specific gravity test (ASTM D-854)
QQ Atterberg’s limit test (ASTM D-4318)
QQ Standard and modified Proctor compac-
tion test (ASTM D-698 and ASTM
D-1557)
QQ Index density test (ASTM D-4253 and
ASTM D-4254)
QQ In-situ density test (ASTM D- 1556)
The void ratios corresponding to maxi-
mum and minimum index unit weights
(γdmax and γdmin respectively) were calcu-
lated using Equations 26 and 27.
⎫G γ ⎫
emax = ⎪ s w ⎪ – 1
⎭ γdmin ⎭
(26)
Journal of the South African Institution of Civil Engineering
 100

90

80

70
Percent finer by weight (%)

60

50

40
.
30
Gravel = 0–20%
Sand = 80–100%
20
Silt & Clay = 0–12%
Cu = 1.4–14.0
10
Cc = 0.22–2.82
D50 = 0.15–2.5 mm
0
100.0 10.0 1.0 0.1 0.01 0.001
Particle size (mm)

Figure 1 Grain size distribution curves of soil samples

⎫G γ ⎫
emin = ⎪ s w ⎪ – 1
⎭ γdmax ⎭
(27)
24

Test results
The test data of 165 samples was used for 22

the development of correlations, whereas

20.8
data of 20 samples was used for validation 20.4
20
of the correlation. Figure 1 shows the grain 19.5
Dry unit weight (kN/m3)

size distribution (GSD) curves for all the

samples and Figure 2 shows the range of
18
γd obtained through index density tests,
standard Proctor and modified Proctor. 16.8
General findings listed below are made on 16
15.7 15.7
the basis of Figures 1, 2 and 3.
QQ All the selected samples are sandy in 14.7
nature, ranging from coarse to fine 14
sand. The median grain size (D 50) of
all samples is in the range of 2.5 mm to 12.5
0.15 mm, with an average of 0.6 mm, 12
and the standard deviation, which is the
amount of dispersion within the D 50
dataset, is 0.49. The effective grain size 10
Minimum index Standard Proctor Modified Proctor Maximum index
(D10) varies from 0.5 mm to 0.075 mm, density (γdmin) (γdstd) (γdmod) density (γdmax)
with an average of 0.13 mm and a stand- Test type
ard deviation of 0.05 mm.
QQ The sand content is between 80% and Figure 2 R
 ange of γd determined through maximum and minimum density tests, and standard
100% in the samples used in this study, and modified Proctor tests

whereas gravel (percent retained on
US # 4, R4) varies from 0% to 20% and
fines (percent passing US # 200, F200)
present in the samples are in the range of
0% to 12% according to ASTM D 2487.
QQ Based on the results of grain size distri-
bution curves, grain sizes corresponding
to 60%, 30% and 10% passing (D 60 , D30
and D10 respectively) were determined.
⎫ D ⎫
The uniformity coefficient ⎪ Cu = 60 ⎪
⎭ D10 ⎭
varies from 1.42 to 14.0. The average
value of Cu is 4.5 and its standard

Journal of the South African Institution of Civil Engineering Volume 62 Number 2 June 2020 15
 ⎫ D302 ⎫
deviation is 6.1%. The curvature coef-
ficient ⎪ Cc = ⎪ is in the range of
⎭‌ D 60 ∙ D10 ⎭
0.22 to 2.82, with a standard deviation
of 1.0 and an average of 1.1.
QQ Fines in all the samples are less than
12% and are non-plastic in nature.
QQ Specific gravity of the samples is in the
range of 2.50 ~ 2.76, with an average
value of 2.65.
QQ All the samples are classified accord-
ing to the Unified Soil Classification
System (USCS) as described in ASTM
D-2487. These samples are classified as
well-graded sand (SW), poorly graded
sand (SP), poorly graded sand with silt
(SP-SM), and well graded sand with silt
(SW-SM). Of the total samples tested
48% belong to the SP classification
group, 40% are SP-SM samples, while
representation of SW and SW-SM
groups is 6% each in the tested samples.
QQ The maximum dry unit weight (γdmax)
corresponding to modified compaction
tests is in the range of 15.7 ~ 20.4 kN/‌m3
as shown in Figure 2, whereas optimum
moisture content (OMC) ranges from
9% to 12%. Similarly, for standard
compaction tests, compaction param-
eters, i.e. γdmax, are in the range of
14.7 ~ 19.5 kN/‌m3, and OMC between
12 ~ 15.5%.
QQ The value of minimum dry unit weight
(γdmin) determined through minimum
density tests (ASTM D-4254) varied
between 12.5 ~ 16.5 kN/‌m3, and the
value of maximum dry unit weight
(γdmax) determined through maximum
density tests (ASTM D-4253) varied
between 15.7 ~ 20.8 kN/‌m3 as presented
in Figure 2. Maximum and minimum
density tests were repeated thrice on
each sample, and average values of γdmin
and γdmax were reported. However, vari-
ation in the values of γdmin and γdmax
are of the order of ± 0.5 kN/‌m3 and
± 0.3 kN/‌m3, respectively. The values
of maximum and minimum void ratios
(emax and emin) are determined through
Equations 26 and 27. The value of emax
varies from 0.47 to 1.08 and emin from
0.67 to 0.24.
QQ Field density tests (FDT) were performed
in the bed of River Ravi (Pakistan) using
the sand-cone method with a 10.16 cm
diameter cone. Uniformly graded sand,
passing from sieve # 25 and retained on
sieve # 52 and having a dry unit weight
of 13 kN/‌m3, was used in performing the
tests. Five test pits were excavated up to a
16
50
40
Number of samples
30
20
10
0
72 74 76 78 80 82 84 86
Relative compaction at zero relative density (%)
Figure 3 Values of relative compaction (Rc) at zero relative density
depth of 1.0 ~ 1.5 m. The water table was
encountered at a depth of 1.2 ~ 1.6 m
in these test pits. FDT was performed
at 0.3 m intervals in each test pit, the
depth of the test hole for each FDT
point was 10 cm, and the diameter of
the hole was 10.16 cm. The in-situ densi-
ties thus determined varied between
13.6 ~ 19.2 kN/‌m3 and the natural mois-
ture content varied between 5% and 22%.
QQ Values of relative compaction at zero
relative density, i.e. ratio of minimum dry
unit weight (γdmin) and maximum dry
unit weight (γdmax), determined through
index density tests, were computed for
each data set and are presented graphi-
cally in Figure 3. The mean value of rela-
tive compaction at zero relative density
for this data set is 83.
DEVELOPMENT OF
CORRELATIONS
Data analyses were carried out, and based
on this, correlations between relative densi-
ty and relative compaction, maximum and
minimum void ratios, material parameters
and void ratios have been proposed, which
are presented in the following section.
Correlation between relative
density and relative compaction
Relative compaction and relative density
were calculated by Equations 1 and 2
respectively, and correlation between
them is proposed. The value of γdmax in
Equations 1 and 2 was determined by
performing maximum index density tests
Volume 62 Number 2 June 2020
Mean value of Rc = 83
88 90 92 94 96 98
following the procedure given in ASTM
D 4253, whereas γdmin was determined by
minimum index density tests adopting
ASTM D 4254. The approach to determine
dry unit weight is based on the findings of
Tavenas and Rochelle (1972). They reported
that the limiting state of compactness
can be defined either by measuring the
real maximum and minimum densities of
a given material by experimenting with
every possible compaction technique, or
be selecting one particular compaction
procedure and declaring the state of com-
pactness as limiting even if they are not
necessarily limiting. The second approach
is adopted here to determine the dry unit
weight of the samples, and using these
values, correlation is proposed. Parameter
γdf in Equations 1 and 2 is in-situ dry unit
weight and is not measured practically
at this stage; its value is rather taken as
equal to dry unit weight, as determined
by performing the Proctor test (standard
and modified) in the laboratory on all
samples, and corresponding relative com-
paction and relative density are calculated.
Similarly, Rc and Dr are also calculated by
taking γdf equal to 95% of standard and
modified Proctor density for each test
sample. Also, it is considered that if the
soil is in its loosest state, γdf will be equal
to minimum dry unit weight (γdmin) and
its ef will be equal to maximum void ratio
(emax ). For this loosest state of the in-situ
soil, both relative density (Dr , i.e. zero at
loosest state) and relative compaction (Rc)
have been calculated. Linear regression
analysis was performed on this data set,
Journal of the South African Institution of Civil Engineering
 (SEE) and its value is 2.11 for Equation 28,
110 which is quite low, indicating the good
Rc = 0.17 Dr(%) + 83.0
prediction capability of the model. The
R2 = 0.88
model F value for Rc is greater than critical
100
Relative compaction (Rc) (%)

F, indicating that Equation 28 is significant.

Similarly, absolute t-statistics for input
90 parameters is greater than the t-significance
of the model. Equation 28 is presented in
graphical form in Figure 4. From Figure 4 it
80
Ratio of γdmin ∕ γdmax i.e values of relative compaction is clear that when relative density (Dr = 0)
when relative density is zero. The variation in values is zero, the relative compaction (Rc) varies
of Rc may be due to error in measuring the dry
70 between 74% ~ 88% with an average of
unit weights plus change in gradation and specific
gravity of the tested samples 83%. This larger spread in values of relative
compaction at zero relative density may be
60
0 20 40 60 80 100 120 due to errors, as explained by Tavenas and
Relative density (Dr) Rochelle (1972), in determining the values
of γdmin and γdmax , and their corresponding
Figure 4 Correlation between relative density and relative compaction effect on relative compaction, plus change in
the gradation and different specific gravity
values of the tested samples. However, it is
110 difficult to fix the real contribution of each
Rc = 0.17 Dr + 83.0
above-mentioned fact (i.e. dry unit weight,
change in gradation and specific gravity) in
100 the variation of Rc corresponding to zero
1:1 line Dr . This implies that when the soil is placed
in its loosest state, its relative density is
Rc (%) predicted

90 zero, whereas its average value of relative

80
±5%
70
60
60 70
Figure 5 Experimental versus predicted values of relative compaction (Rc)
and the equation between relative density
and relative compaction was formulated.
Regression analysis technique was used
to calibrate the correlation, and the val-
ues of coefficients for input and output
parameters were determined. Analysis of
variance (ANOVA) was also carried out to
determine F-statistics for output param-
eters and t-statistics for input parameters.
F-statistics represent the overall predictive
capability of the model by considering the
combined effect of all the variables present
within the model, and t-statistics is the
ratio between the difference between two
groups and the difference within the group,
and is accompanied by a p-value which
shows the probability that the results of the
dataset occurred by chance. Low p-values
indicate that the results do not occur by
chance. Coefficient of determination (R2) is
The marked portion indicates that experimental
values of Rc is different due to change in gradation
of samples. However, predicted values using
equation remain constant at an average value of 83.
80 90 100
Rc (%) experimental
equal to the percentage of variation in the
dependent variable that can be explained
by the independent variable. The final best
fit correlation is given by Equations 28.
Rc = 0.17Dr (%) + 83
The regression statistics for Equation 28
are: r (824) = .94; t (824) = 681; p < .001. A
significant regression equation was found:
(F (1), 824) = 5818, p < .001 with R2 = .88, as
well as true difference between the means,
with a 95% confidence interval = (16.85,
17.20) for relative density and (82.9, 83.45)
for intercept. The r value of .94 and R2 of
.88 for Equation 28 is rated as a reasonable
correlation coefficient in geotechnical
engineering. The variation between experi-
mental and predicted values of the output
parameter is a standard error of estimate
compaction is 83%. This finding is in good
agreement with Lee and Singh (1971) who
noted that when Dr is zero, then Rc is equal
to 80%. It is also inferred from Equation 28
that a one-point variation in relative com-
paction is equivalent to a six-point variation
in relative density, whereas a 1 percent
change in relative compaction value is
110 equivalent to a 5 percent change in relative
density in the Lee and Singh (1971) equa-
tion, indicating that Equation 28 is more
sensitive. The major advantage of writing a
technical specification in terms of Dr rather
than Rc is that it magnifies the values and
thus automatically emphasises the impor-
tance of small variations from the specified
values. A small variation in the value of
Dr may be significant when studying the
(28) liquefaction potential of granular soil during
earthquakes. The limitation of the proposed
Equation 28 is in line with the Tavenas and
Rochelle (1972) observation that an error
in computing Dr is a function of errors in
measuring γdmin, γdmax and γdf . Most of the
varying errors occurring during measure-
ment of the desired parameters (γdmin , γdmax
and γdf ) are due to the measuring technique
being used by different persons performing
the experiments. Although these errors may
be small, they have a significant effect on
the Dr values.
Figure 5 represents the extent of dif-
ference between experimental data and

Journal of the South African Institution of Civil Engineering Volume 62 Number 2 June 2020 17
predicted values, using Equation 28. These
plots show that the prediction accuracy of 110
the relation is within ± 5% for Rc (%) with
a 95% confidence interval. It can also be 105
observed from Figure 5 that, when soil is Loose to medium dense Dense to very dense
in the loosest state, the predicted value of ±5%
relative compaction value is 83.0, as cal- 100

Rc (%) predicted
1:1 line
culated using Equation 28, but practically
experimental values of relative compac- 95
tion at the loosest state are different due
to changes in grain size distribution and
90
errors in measuring the unit weights, indi-
cating the non-applicability of Equation 28
when the soil is in its loose state. However, 85
when the in-situ density of soil is more Lee and Singh (Equation 3)
than its loosest state, the variation between Present research (Equation 28)
80
experimental versus predicted values is 80 85 90 95 100 105 110
quite reasonable, justifying the use of this Rc (%) experimental
equation for practical purposes.
Figure 6 Comparison of experimental versus predicted values of relative compaction (Rc)
Validation of relation between
relative density and relative
compaction (Equation 28) 0.90
Development of any predictive model is
0.80
followed by its validation by some inde- ±10%
pendent data that had not been used for the 0.70
development of the model. In this study,
emin (predicted)

subsequent to the model formulation, five 0.60

test pits were excavated in the bed of River 1:1 line
0.50
Ravi (Pakistan) and field density tests were
performed using the sand replacement 0.40
technique. The in-situ density at various
depths in each test pit was determined. 0.30
Samples were also collected from each
test point and brought to the geotechnical 0.20

engineering laboratory, where index density

tests were conducted on these samples.
This test data was used for the validation
of Equation 28. The experimental values
of Rc of these samples were plotted against
the predicted values using Equation 28 and
are shown in Figure 6. A variation of ± 5%
was observed between the predicted and
experimental values of Rc . The empirical
relationship developed by the Lee and
Singh (1971) equation was also used to
predict relative compaction and plotted
against experimental values of Rc . It was
observed that both equations overesti-
mated relative compaction values, as they
were plotted above the 1:1 line in Figure 6.
However, all the predictions by the Lee
and Singh (1971) equation and Equation 28
fall within the limits of ± 5%. By careful
scrutiny of Figure 6, it can be observed that
when soil is between the loose to medium
dense state, i.e. relative compaction of the
soil is between 80% ~ 93%, the difference
in the predicted values of Rc by both the
equations is within 2% ~ 3%. However, this
0.10
0.10 0.20
Figure 7 Experimental versus predicted values of emin by Equation 29
difference diminishes when the subsoil
conditions are dense to very dense, or when
the relative compaction value is between
93% and 100%. The probable reasons for
this difference in predicted values of rela-
tive compaction are errors in measuring
dry unit weights plus the difference in
the intercept of these equations, which is
more dominating when the soil is between
loose to medium dense. This indicates that
Equation 28, as proposed in this research,
and the Lee and Singh (1971) relation can
be used effectively to correlate relative
density and relative compaction.
Correlation between maximum
and minimum void ratios
Regression analysis was carried out on
the test data to relate the maximum and
0.30 0.40 0.50 0.60 0.70 0.80 0.90
emax (experimental)
minimum void ratios. It was observed that
linear correlation exists between maximum
and minimum void ratios, and is given by
Equation 29. Experimental and predicted
values of emin by Equation 29 are plotted in
Figure 7 using the data set. It is observed
that variation in measured and predicted
values of void ratios is within ± 10%.
emax = 0.21 + 1.23emin(29)
The regression statistics for Equation 29
are: r (164) =.95; t (164) = 13.2; p < .001.
A significant regression equation was
found: (F (1), 164) = 1665, p < .001 with
R2 = .91. The lower and upper limit of
intercept at a 95% confidence interval is
.16 ~ .22 for intercept, and that of emin
is between 1.20 ~ 1.32. These regression

18 Volume 62 Number 2 June 2020 Journal of the South African Institution of Civil Engineering
 Cubrinovski and Ishihara (Equations 9 and
0.90 10), while predictions using the Miura et
Cubrinovski and Ishihara (Equation 9)
al equation falls within the ± 10% limit.
0.80 Present research (Equation 29)
±10%
The probable reason for this variation
Miura et al (Equation 8)
0.70 in prediction by the Cubrinovski and
Ishihara equation is that the test procedure
emin (predicted)

0.60 being employed in the present research

(Equation 29) in determining emin is based
0.50 1:1 line
on the ASTM standard, whereas the
0.40
Cubrinovski and Ishihara equation is based
on the Japanese Geotechnical Standard.
0.30 Therefore, caution is needed when apply-
ing their equations (Equations 9 to 12) to
0.20 estimate emin or emax .

Correlation between gradational

parameters and void ratios
0.10
0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 Multivariate regression analysis was
emin (experimental) carried out using Statistical Product and
Service Solution (SPSS) software on the
Figure 8 C
 omparison of experimental versus predicted values of emin by Equation 29, Cubrinovski data to develop a predictive model between
& Ishihara (2002) Equation 9 and Miura et al (1997) Equation 8 gradational parameters and voids. The
input parameters used for regression were
median grain size (D 50 ), effective grain
1.20 size (D10 ), uniformity coefficient (Cu),
Maximum void ratio (emax)
curvature coefficient (Cc) and specific
Minimum void ratio (emin) ±10%
gravity (Gs). Stepwise regression analysis
Maximum/minimum void ratio (predicted)

1.00
was carried out, and out of the five above-
mentioned parameters, median grain
0.80 size (D 50 ) and uniformity coefficient (Cu)
passed the F-test and t-test with a 95% con-
fidence interval. These relations are pre-
0.60
sented in Equations 30 and 31 respectively.
The experimental versus predicted values
0.40 of void ratios are presented in Figure 9.
The variation between experimental versus
predicted values falls within ± 10%.
0.20
0.033 0.370
emin = 0.24 + + (30)
1:1 line D50 Cu
0
0 0.20 0.40 0.60 0.80 1.00 1.20
Maximum/minimum void ratio (experimental)
0.072 0.306
emax = 0.48 + + (31)
Figure 9 E xperimental versus predicted values of maximum and minimum void ratios using D50 Cu
Equations 30 and 31
The R values of 0.90 and 0.92 for
statistics indicate that the correlation development, were tested in the labora- Equations 30 and 31 thus obtained as a
between maximum and minimum void tory for the validation of the correlation. result of regression analysis by SPSS are
ratio is significant. R2 for Equation 29 is .91, Experimental values of minimum void rated as reasonable correlation coefficients
indicating that only 9% of the variation in ratio (emin) are plotted against the values in geotechnical engineering. SEE values
the values of emax was not accounted for by predicted by Equation 29 and are shown in of 0.06 and 0.06 for Equations 30 and 31
the regression. Figure 8. It is observed that all the predict- are very low, representing good prediction
ed values fall within ± 10% of the measured capability of the models. Analysis of vari-
Validation of the relation values. The empirical relationship devel- ance (ANOVA) was also carried out by
between maximum and minimum oped by Cubrinovski and Ishihara (2002) SPSS software.
void ratios (Equation 29) and Miura et al (1997) are also used to pre- Regression statistics for Equation 30
In this study, after the formulation of cor- dict minimum void ratios of these twenty are (F (2 162) = 339.70; p < .000; with
relation between maximum and minimum samples, and are also presented in Figure 8. R2 = .81; t (162) = 13.05, 14.75; p < .000
void ratios, a new set of twenty soil samples For emin , six out of 20 observations 1 1
for and , respectively. Similarly,
which had not been used in the model fall outside the limit of ± 10% by using D 50 Cu

Journal of the South African Institution of Civil Engineering Volume 62 Number 2 June 2020 19
for Equation 31 regression statistics are
(F (2 162) = 438.03; p < .000 with R2 = .85;
t (162) = 20.64, 10.27; p < .000 for
1 1
and , respectively. Based on these
D 50 Cu
statistical results, it is evident that both
independent variables (Cu and D 50 ) are
significant predictors of emax and emin .
Furthermore, this suggests that a slight
change in Cu and D 50 values have a marked
effect on emax and emin . Also, it can be
inferred from Equations 30 and 31 that
emax and emin are inversely proportional to
median grain size and Cu . Void ratios pre-
dicted using Equations 30 and 31 decrease
as the median grain size increases and den-
sity increases. However, it is known that, as
median grain size increases, it leads to seg-
regation of the aggregates, resulting in loos-
er packing compared to a smaller median
size, which would lead to denser packing.
This is the limitation of these equations,
and it may be catered for by the fact that
these equations are valid for median grain
sizes ranging between 0.2 mm and 2.8 mm.
These equations would not be reliable to
estimate void ratios for samples compris-
ing bigger-sized particles (D 50 > 3 mm).
Secondly, fractions of the material bigger
than the diameter of the soil at 60% passing
(D 60) have no role in the prediction of emax
and emin using Equations 30 and 31, but in
practice bigger-sized particles may lead to
more voids if not properly packed, resulting
in lesser densities compared to medium
or smaller-sized particles which may have
uniform packing.
Validation of correlation between
grain size parameters and void
ratios (Equations 30 and 31)
After the formulation of correlations
between gradational parameter and void
ratios, a new set of 20 soil samples were
tested in the laboratory, and their results
were utilised for the validation of the cor-
relations. Experimental values of void ratios
were plotted against predicted values using
Equations 30 and 31. It was observed that
all the estimations fell within ± 10% limits
and are shown in Figure 10. The empiri-
cal relationship developed by Patra et al
(2010) was also used to predict void ratios
for these 20 samples. The comparison
of experimental versus predicted values
is also presented in Figure 10. It can be
inferred from Figure 10 that the Patra et
al (2010) equation overestimates emin by
about 15%. However, for emax five out of 20
predictions fall outside the ± 10% limits.
20
1.20
Maximum/minimum void ratio (predicted)
1.00
0.80
0.60
0.40
±10%
0.20
0
0 0.20
Maximum/minimum void ratio (experimental)
Figure 10 C
 omparison of experimental versus predicted values of maximum and minimum void
ratios using Equations 30, 31 and Patra et al (2010)
The probable reason for the estimation
variation by Patra et al (2010) may be that
it depends on only one parameter (median
grain size), and therefore does not reflect
the full representation of the grain size
distribution curve.
FIELD IMPLICATION OF THE
PROPOSED MODELS
Three models (relations between relative
density and relative compaction, maximum
versus minimum void ratio, and gradation-
al parameters versus maximum and mini-
mum void ratios) have been presented in
this research. These models are very useful
in estimating various properties based on
grain size or index density tests for granu-
lar soils, which are normally recommended
to be used in road construction or as a
backfill material. However, these equations
have a few limitations and cannot be used
for soil possessing more than 20% gravels,
or soils with fines greater than 15%. The
estimation in this case may be misleading.
Secondly, the uniformity coefficient is the
ratio of two grain sizes, and different soils
with different grain sizes may have the
same Cu values, but their corresponding
void ratios may be different, which could
be misleading. However, for prediction
purposes and for the sake of simplicity, the
proposed models may be very appealing.
CONCLUSIONS
A simple linear correlation between rela-
tive density and relative compaction has
Volume 62 Number 2 June 2020
1:1 line
Maximum void ratio by present research
Minimum void ratio by present research
Maximum void ratio by Patra et al
Minimum void ratio by Patra et al
0.40 0.60 0.80 1.00 1.20
been proposed, based on a large amount
of experimental data, followed by proper
validation. Linear and multiple regression
analyses were carried out on the same
data set to relate emax and emin and to
the parameters Cu and D 50 . The validity
of these correlations was verified using
an independent data set along with the
comparison of the proposed models with
similar correlations presented by other
researchers. The following conclusions
were made from the research.
QQ Based on extensive data of compaction
tests including both standard and modi-
fied Proctor and relative density tests,
correlation between relative density and
relative compaction has been developed
as given below:
Rc = 0.17Dr (%) + 83.
QQ It is inferred from the results of relative
density and relative compaction that,
when sandy soil is in its loosest state,
its relative density is zero, whereas its
relative compaction is 83%. Also, it can
be concluded that Rc predicted by the
proposed equation and Lee and Singh
(1971) fall within ± 5% with a 95% confi-
dence interval indicating the applicabil-
ity of both these equations for Pakistan’s
local sands having non-plastic fines up
to 10%.
QQ There is a linear relationship between
the maximum and minimum void ratio
of sandy soils. The maximum void ratio
can be estimated using the correla-
tion emax = 0.21 + 1.23emin with a 95%
Journal of the South African Institution of Civil Engineering
 confidence interval. The experimental Cubrinovski, M & Ishihara, K 2002. Maximum and Miura, K, Maeda, K, Furukawa, M & Toki, S 1997.
versus predicted values of emin fall with- minimum void ratio characteristics of sands. Soils Physical characteristics of sands with different
in ± 10% with a 95% confidence interval. and Foundations, 42(6): 65–78. doi.org/10.3208/ primary properties. Soils and Foundations,
QQ Based on the grain size data of the test sandf.42.6_65. 37(3): 53–64.
specimens and the index density test Graton, L C & Fraser, H J 1935. Systematic packing of Mujtaba, H & Farooq, K 2010. Experimental study on
data, maximum and minimum void spheres – with particular relation to porosity and relative compaction and relative density of sandy
ratios can be predicted using the follow- permeability. Journal of Geology, 43(8): 785–909. soils. Proceedings, International Conference on
ing relations: JGS (Japanese Geotechnical Society) 2000. JGS 0161. Geotechnical Engineering (ICGE 2010), Lahore,
Test Method for Minimum and Maximum Densities Pakistan, pp 67–74.
0.072 0.306 of Sand. Soil Testing Standard, pp 136–138. Patra, C R, Sivakugan, N, Das, B M & Rout, S K 2010.
emin = 0.48 + +
D50 Cu Kulhawy, F H & Mayne, P W 1990. Manual on Correlations for relative density of clean sand
estimating soil properties for foundation design. with median grain size and compaction energy.
0.033 0.370 Report EL-9800 to the Electronic Power Research International Journal of Geotechnical Engineering,
emin = 0.24 + +
D50 Cu Institute. Ithaca, NY: Cornell University. 4(2): 195–203. doi: 10.3328/IJGE.2010.04.02.195-203.
Lade, P V, Liggio, C D & Yamamuro, J A 1998. Effect of Shimobe, S & Moreto, N 1995. A new classification
The prediction accuracy of the above- non-plastic fines on minimum and maximum void for sand liquefaction. Proceedings, 1st International
mentioned correlations is within ± 10% ratios of sand. Geotechnical Testing Journal, ASTM, Conference on Earthquake Geotechnical
with a 95% confidence interval. 21(4): 336–347. Engineering, Tokyo, Japan, pp 315–320.
Lee, K L & Singh. A 1971. Relative density and relative Tavenas, F & Rochelle, P 1972. Accuracy of relative
compaction. Journal of the Soil Mechanics and density measurements. Geotechnique, 22(4): 549–562.
REFERENCES Foundations Division, ASCE, 97(7): 1049–1052. White, H E & Walton, S F 1937. Particle packing and
ASTM (American Society for Testing and Materials) McCook, D K 1996. Correlations between simple particle shape. Journal of the American Ceramic
2007. ASTM Book of Standards, Section 4, vol. 04.08, field test and relative density test values. Journal Society, 20(5): 155–166.
West Conshohocken, PA: ASTM International. of Geotechnical Engineering, 122(10): 860–862. Youd, T L 1973. Factors controlling maximum and
Cubrinovski, M & Ishihara, K 1999. Empirical doi: 10.1061/(ASCE)0733-9410(1996)122:10(860). minimum densities of sands. Evaluation of relative
correlation between SPT N-value and relative McGeary, R K 1961. Mechanical packing of spherical density and its role in geotechnical projects involving
density for sandy soils. Soils and Foundations, 39(3): particles. Journal of the American Ceramic Society, cohesionless soils. STP 523, West Conshohocken, PA:
6l–7l. doi.org/10.3208/sandf.39.5_61. 44(10): 513–522. ASTM, pp 98–122.

Journal of the South African Institution of Civil Engineering Volume 62 Number 2 June 2020 21
TECHNICAL PAPER The liquefaction potential
Journal of the South African
Institution of Civil Engineering of the upper quaternary
ISSN 1021-2019
Vol 62 No 2, June 2020, Pages 22–30, Paper 0684 sands of the Cape Flats,
Nanine Fouché graduated in 2008 with a
Bachelor’s degree in Earth Science from
Western Cape, South Africa
Stellenbosch University. Thereafter she
continued her studies and completed a
Post-Graduate Diploma and MSc Eng in N Fouché
Geotechnical Engineering. She gained
valuable experience as an engineering
geologist at Aurecon South Africa,
undertaking geotechnical and foundation investigations for a wide range Liquefaction of saturated cohesionless soils as a result of earthquake-induced ground shaking
of projects. Her passion for teaching and the academics led her to a career is a major concern for structures constructed within or on these soils. During an earthquake,
in teaching and research in the Civil Engineering Department at shear waves propagate through the ground, generating shear stresses and strains that are cyclic
Stellenbosch University. She currently teaches geotechnical engineering
in nature, causing the soil grains to rearrange into a denser state of packing. This process is too
courses and provides support to geotechnical research students, while
also completing her PhD. rapid for drainage to occur, which results in an increase in pore water pressure and a decrease
in effective stress and shear resistance of the soil. The occurrence of a seismic event associated
Contact details:
Department of Civil Engineering with the Milnerton Fault line extending below the Cape Flats, a low-lying sand-covered area
Stellenbosch University connecting the Cape Peninsula with the southwestern Cape mainland of South Africa, can cause
Private Bag X1 extensive damage to the largely inadequate infrastructure characterising the area. To evaluate
Matieland 7602
South Africa
the susceptibility of Cape Flats sand to undergo liquefaction during and after a seismic event,
T: +27 21 808 3816 the empirical SPT-based method put forth by Idriss & Boulanger (2004) was used to assess the
E: naninef@sun.ac.za cyclic liquefaction potential of dense, dilative sands, whereas the CPT-based method published
by Robertson (2016) was used to assess the volumetric response of the sands during undrained
shear. Resistance of the Cape Flats sands to liquefy at probable earthquake moment magnitude
and acceleration values of M = 6.0 and amax = 0.15 g respectively, was noted. Robertson’s
normalised Soil Behaviour Type (SBTn) chart revealed mostly dilative sands and silty sands
susceptible to cyclic liquefaction. Loose, contractive sands prone to flow liquefaction are
present, although limited in occurrence and extent.

INTRODUCTION saturated sand during undrained shear is

Keywords: Cape Flats sand, cyclic liquefaction, flow liquefaction,
standard penetration test, cone penetration test
Liquefaction of saturated cohesionless
soils as a result of earthquake-induced
ground shaking is a major concern for
structures constructed within or on these
soils. During an earthquake, shear waves
propagate through the ground, generating
shear stresses and strains that are cyclic in
nature, causing the soil grains to re­arrange
into a denser state of packing. This process
is too rapid for the drainage to occur,
which results in an increase in pore water
pressure and a decrease in effective stress
and shear resistance of the soil. The
various forms of unstable behaviour of
saturated sands during earthquakes are
typically all described by the general term
liquefaction. However, a distinction needs
to be made between the different forms
of liquefaction, namely flow liquefaction,
cyclic mobility and cyclic liquefaction.
Flow liquefaction describes the
un­drained flow of saturated, loose contrac-
tive sands when the static shear stress
surpasses the residual strength of the soil
(Rauch 1997). The response of a contractive
shown in Figure 1(a). During undrained
cyclic shear (with cycles of small shear
strains), dense, saturated sands have the
tendency to gradually soften as excess pore
water pressures are generated. However,
when these dense soils are subsequently
monotonically loaded without pore water
drainage, dilation of the soil occurs as the
particles move up and over each other,
resulting in a decrease in pore water pres-
sure and an increased shear resistance
(Rauch 1997). This behaviour of dense,
saturated sands is described as cyclic
mobility and is often referred to as limited
liquefaction. In addition, Rauch (1997) indi-
cates that a soil which normally tends to
dilate during undrained, monotonic shear
may temporarily lose its shear resistance
when the effective stress becomes zero,
thus leading to substantial deformations.
This occurrence, described by the term
cyclic liquefaction, ensues when cyclic
shear stresses become larger than the ini-
tial, static shear stress, producing a reversal
in the direction of shear stress (stress path

Fouché N. The liquefaction potential of the upper quaternary sands of the Cape Flats, Western Cape, South Africa.
22 J. S. Afr. Inst. Civ. Eng. 2020:62(2), Art. #0684, 9 pages. http://dx.doi.org/10.17159/2309-8775/2020/v62n2a3

Shear stress
Peak shear
strength
Static shear stress
Residual shear
resistance
Flow of
soil
(a)
Figure 1 B
 ehaviour of saturated sand in (a) loose (contractive) and (b) dense (dilative) states
during undrained shear
passes through a state of zero shear stress).
Deformations will, however, stabilise when
the cyclic loading ends. The tendency of a
dilative soil to produce an increase in shear
strength when monotonically sheared, is
shown in Figure 1b.
STEADY-STATE CONCEPT
When a soil deforms at a constant volume
or void ratio, constant effective stress,
constant shear stress or resistance, and
constant rate of shear strain, it is said to
be in the steady state (Rauch 1997). Steady-
state flow can be attained through drained
or undrained, and cyclic or monotonic
loading. For a specific soil, a plot of possible
conditions during steady-state flow gives
a single curve (the steady-state line) in
the three-dimensional space of void ratio,
effective stress and shear stress. When the
initial conditions of effective stress and
void ratio plot above the steady-state line,
the soil is considered contractive, whereas
dilative soils will plot below the steady-
state line.
SUSCEPTIBILITY OF THE CAPE
FLATS SANDS TO LIQUEFACTION
The Cape Flats is a low-lying sand-covered
area connecting the Cape Peninsula with
the southwestern Cape mainland of South
Africa (Hill & Theron 1981). This expan-
sive area of the Cape Town Metropole,
covering a surface area of about 630 km2,
is home to a substantial portion of the
population of the greater Cape Town
area. The terrain is flat, densely inhabited
and typically characterised by suburban
townships and informal settlements.
The occurrence of a seismic event, with
Journal of the South African Institution of Civil Engineering Volume 62 Number 2 June 2020
Shear stress
Dilation relieves excess pore
pressure at larger strains
Accumulation of excess pore
pressure and deformation
Axial strain (b) Axial strain
associated soil liquefaction, can cause
extensive damage to the largely inadequate
infrastructure, posing a significant risk to
residents. The evaluation of liquefaction in
the Cape Flats is considered necessary, and
an initial assessment of the susceptibility of
the upper quaternary sands to liquefaction
is given below.
The liquefaction characteristics of a
soil are not only affected by the initial
density and effective confining stress, but
also by factors such as particle cementa-
tion, soil fabric and ageing, which will
prevent movement and rearrangement
of the soil particles and thus increase
liquefaction resistance.
Liquefaction is mostly seen in loose,
saturated sand and silt deposits that
are subjected to strong ground motions
induced by earthquakes. Rounded soil par-
ticles of uniform gradation are especially
prone to liquefaction. It is the more stable
interlocking of well-graded sands with
angular particle shapes that increases its
resistance to liquefaction. Liquefaction is
additionally associated with recent deposits
in which particle movement is not hindered
by age-related cementation. According to
Rauch (1997), the presence of large propor-
tions of non-plastic fines in a soil, resulting
in lower permeability, will impede drainage
of excess pore water pressure, making the
soil more likely to liquefy.
The quaternary-aged windblown
sands and silty sands of the Witzand and
Springfontein Formations of the Cape Flats
are typically described as normally consoli-
dated, and fine and medium-grained. The
uniform gradation and loose particle pack-
ing, without cementation, make these soils
prone to liquefaction. In addition, shallow
groundwater often occurs widespread
across the Cape Flats, especially during
the wet winter months, making the sedi-
ments susceptible to the accumulation of
excess pore water pressures during seismic
shearing. Conversely, the consolidated sand
of the Langebaan Formation, comprising
calcretised layers, will provide significant
resistance to grain movement, and thus
liquefaction, during shearing.
Although the Western Cape Province
of South Africa, and specifically Cape
Town, is not located close to a plate bound-
ary, intraplate fault lines present local
weaknesses in the earth’s crust which are
vulnerable to slip in the case of accumula-
tion of sufficient regional tectonic strain.
An intraplate fault line known as the
Milnerton Fault, extends beneath the Cape
Flats in a northwesterly to southeasterly
direction. Seismic activity associated with
this intraplate fault is considered respon-
sible for the largest earthquake experienced
by the City of Cape Town in 1809. During
this event, observations of muddy water
squirting out of fountains were made, pos-
sibly providing evidence of soil liquefaction
induced by earthquake vibrations (De
la Harpe 2015). In 2003, the Council for
Geoscience produced a seismic hazard map
showing peak ground accelerations for the
whole of South Africa (De la Harpe 2015).
According to this map, a gravity accelera-
tion of 0.15 g (1.471 m/s2) – with a 10%
probability of exceedance in 50 years – is
applicable to the Cape Flats area. The
Council for Geoscience additionally pro-
duced a seismic intensities map of South
Africa showing probabilistic Modified
Mercalli Scale (MMS) intensities with
a 10% chance of exceedance in 50 years
(based on seismological data from 1620
to 1989) (Brandt 2011). Seismic intensity
Class VI was assigned to the study area.
EVALUATION OF LIQUEFACTION
When evaluating the susceptibility of a soil
to liquefaction, a distinction between cyclic
liquefaction, i.e. when deformations occur
only during cyclic loading, and liquefac-
tion resulting from strain-softening with a
subsequent loss of shear strength, needs to
be made.
The triggering of cyclic liquefaction
in a given soil deposit is often assessed
using empirical methods based on in-situ
penetration tests. To develop such an
empirical method, sites possibly subjected
to earthquake-induced liquefaction in
the past are investigated to determine
23
whether liquefaction had occurred, and
to measure the in-situ soil strength. The 0.6
Curves derived by:
shear stresses induced in the ground by the
Seed et al (1984) & NCEER/NSF Workshops (1997)
earthquake are also estimated. By separat-
Idriss & Doulanger (2004)
ing conditions where a soil liquefied from
Cetin et al (2004)
those where liquefaction did not occur, a 0.5
liquefaction assessment criterion is formu-
lated. To evaluate the cyclic liquefaction
potential of saturated cohesionless soils
during earthquakes, standard penetra-
0.4
tion test (SPT)-based correlations are the

Cyclic resistance ratio (CRR)

oldest and most widely used methods.
Notwithstanding this, the cone penetration
test has become increasingly popular in
estimating cyclic liquefaction potential.
To estimate the potential for cyclic
liquefaction for level ground sites due to
an earthquake, the cyclic stress ratio (CSR)
profile produced by the design earthquake,
demonstrating the seismic demand of a soil
layer, and the cyclic resistance ratio (CRR)
of the ground, which provides a measure
of the soils’ resistance to liquefaction,
should be known. The CSR is calculated
from the maximum horizontal surface
acceleration (amax) produced by the earth-
quake, whereas the CRR depends upon
penetration resistance data. When the CRR
exceeds the CSR induced by the earthquake
ground motions, the soil is considered
resistant to cyclic liquefaction. Most
empirical liquefaction assessment methods
are based on the cyclic resistance ratio and
a standardised SPT blow count.
One of the most widely used SPT-based
correlations is the deterministic relationship
presented by Seed et al (1984 & 1985 – cited
in Idriss & Boulanger 2004). This familiar
method has been the standard for many
years, with only minor modification at low
CSR suggested by the National Center for
Earthquake Engineering Research (NCEER)
Working Group (NCEER 1997). Since the
publication of this earlier method, SPT-
based correlations have been put forth
by numerous authors, including Liao et
al (1988; 1998), Youd and Noble (1997),
and Toprak et al (1999) – all cited in Seed
et al (2003). Seed et al (2003), however,
emphasised low data quality and overall
uncertainty related to these methods. More
recently, Cetin et al (2004), and Idriss and
Boulanger (2004) put forth similar relation-
ships between CRR and equivalent clean
sand corrected SPT blow count, (N1)60cs.
The three liquefaction triggering correla-
tions are compared in Figure 2.
A review of SPT-based liquefaction
triggering procedures for cohesionless soils
was undertaken by Idriss and Boulanger
0.3
0.2
0.1
0
0 10
Equivalent clean sand corrected standard penetration (N1)60cs
Figure 2 Comparison of SPT-based liquefaction triggering curves (Idriss & Boulanger 2010)
(2010) whereby the database of case histo-
ries was re-examined and updated. During
the review it was found that the lower
position of the Cetin et al (2004) curve was
mainly due to their interpretations of sev-
eral important case histories at a specific
range of effective stresses. In their report,
Idriss and Boulanger (2010) concluded that
the modified Seed et al (1984), and Idriss
and Boulanger (2004) procedures are real-
istic for depths less than 12 m, and that the
Idriss-Boulanger method is well supported
by existing data for extrapolation to even
greater depths. The Idriss-Boulanger (2004)
method is therefore considered superior
and was, for this reason, selected for the
current study.
To evaluate susceptibility to flow
liquefaction, i.e. liquefaction due to
strain-softening with a resulting loss of
shear strength, the potential for a soil to
strain-soften in undrained shear should
be evaluated. According to Robertson
(2010), experience has indicated that
loose sands are prone to abrupt strength
loss at small shear strains and, as such,
the identification of very loose and loose
coarse-grained soils is considered the key
M = 7.5; σ’γ = 1 atm
20 30 40
component to identify a soil susceptible to
flow liquefaction. Robertson (2016) created
a CPT-based normalised Soil Behaviour
Type (SBTn) chart (plotting normalised
cone resistance Qtn against normalised
friction ratio Fr) whereby an approximate
boundary between dilative and contrac-
tive soil response is given. This boundary,
presented by the solid line marked “CD”
in Figure 3, separates dense, dilative soils
in which deformations occur only during
cyclic loading (plotting above line CD)
from loose, contractive soils prone to flow
liquefaction (plotting below line CD).
Figure 3 serves as the latest CPT-based
SBT classification method following the
original SBT chart developed by Robertson
in 1990, and later updated by Robertson and
Wride (1998), Robertson (1999) and Zhang
et al (2002) – all cited in Robertson (2010)
– to enable flow liquefaction evaluation. To
assess the potential of a soil to strain-soften
in undrained shear, Robertson’s SBT chart
method to distinguish dilative and contrac-
tive soils, is considered state-of-the art and
was, for this reason, selected by the author
to assess the static liquefaction potential of
the Cape Flats sands.

24 Volume 62 Number 2 June 2020 Journal of the South African Institution of Civil Engineering
 SPT-based procedure proposed by Idriss and
1 000 Boulanger (2004) was applied to SPT blow
IB = 32 count data from 178 boreholes spread across
Soil behaviour type the Cape Flats. The Idriss and Boulanger
IB = 22 1. CCS Clay-like – Contractive – Sensitive
SD (2004) boundary curve was expressed using
100 2. CC Clay-like – Contractive
CD = 70 Equations 1 to 3.
3. CD Clay-like – Dilative
Qtn

4. TC Transitional – Contractive All the measured SPT N blow count

SC TD 5. TD Transitional – Dilative
CD values (considered N60 values in the cur-
TC 6. SC Sand-like – Contractive
rent study) were first corrected for over-
10 7. SD Sand-like – Dilative
burden stress at the test depths to give the
CD = (Qtn – 11)(1 + 0.06Fr)17
CCS CC (N1)60 values. The clean sand equivalent
IB = 100(Qtn + 10)/(70 + QtnFr)
values of (N1)60 were then calculated as fol-
lows (Idriss and Boulanger 2004):
1
0.1 1 10
Fr (%) (N1)60cs = (N1)60 + ∆(N1)60(1)

⎫ 15.7 ⎫
–⎪ ⎪ (2)
2
9.7
FC ⎭ FC ⎭
Figure 3 SBTn chart based on Qtn and Fr (Robertson 2016)
∆(N1)60 = exp 1.63 +

2.5 Where:
FC = f ines content (percentage smaller
than 0.075 mm)
2.0
Where grading results were available
at the SPT depths, the calculated fines
1.5 contents were inserted into Equation 2.
Where the actual fines content at an SPT
CRR

depth was unknown, an average fines con-

1.0
0.5
0
0 10
Figure 4: Liquefaction triggering curve for Cape Flats sands
According to Robertson (2016), the
CPT-based boundary between contractive
and dilative soils can be influenced by
tent value of 6.3%, calculated for the sands
of the Cape Flats, based on the results of
349 particle size analyses, was applied to
Equation 2.
The cyclic resistance ratio (CRR) was
subsequently calculated at each of the SPT
depths from the clean sand equivalent
20 30 40 50 (N1)60 values for a magnitude M = 7.5
(N1)60cs earthquake and an effective vertical
stress of 1 atmosphere (101.3 kPa), using
Equation 3. A total of 1 526 CRR (N1)60cs
data pairs were produced from the SPT
complete loss of shear strength (soil failure) data. These data pairs are shown graphi-
due to pore pressure increase during cyclic cally in Figure 4, revealing the Idriss and
loading, was estimated by the following Boulanger liquefaction triggering curve.

⎫ (N ) ⎫
several variables, such as in-situ stress state two approaches:
+ ⎪ 1 60cs ⎪
2
(N1)60cs
⎭ 126 ⎭
and, based on case histories by Robertson QQ Evaluation of the cyclic liquefaction
CRR = exp
(2010), this boundary is considered slightly potential of dense, dilative sands by 14.1
⎫ (N ) ⎫ ⎫ (N ) ⎫
– ⎪ 1 60cs ⎪ + ⎪ 1 60cs ⎪
conservative. As a result, a transitional means of an empirical SPT-based method 3 4

zone on the SBTn chart is recommended.
In this regard, Robertson (2016) suggests
an approximate lower boundary based on
these case histories, as shown in Figure 3.
The suggested boundary line marked “CD”
now represents the upper boundary.
METHODOLOGY
The susceptibility of Cape Flats soil to
put forth by Idriss & Boulanger (2004).
QQ Evaluating the volumetric response of
the Cape Flats sands during shear by
means of the CPT-based method pub-
lished by Robertson (2016), to determine
whether flow liquefaction can occur
(identification of loose, contractive soils).
SPT-based method for
cyclic liquefaction
⎭ 23.6 ⎭ ⎭ 25.4 ⎭
– 2.8
(3)
To finally separate the liquefiable and
non-­liquefiable soils, the cyclic stress ratios
induced by a design earthquake were calculat-
ed at the SPT depths as follows (Seed & Idriss
1971 – cited in Idriss & Boulanger 2004):

⎫σ a ⎫
undergo cyclic liquefaction, in which defor- The cyclic liquefaction potential of the Cape
CSR = 0.65⎪ vo max ⎪rd(4)
⎭ σ'vo ⎭
mations occur only during cyclic loading, Flats sands was studied in terms of the trig-
and flow liquefaction, in which there is a gering of liquefaction. The semi-empirical

Journal of the South African Institution of Civil Engineering Volume 62 Number 2 June 2020 25
Where:
QQ σ’vo and σ vo represent the effective and
the total vertical stresses respectively.
The depth of the groundwater table
was assumed at 3.0 m in cases where
the actual position of the phreatic
surface was unknown at an SPT loca-
tion. This is deemed realistic for the
study area.
QQ amax = maximum horizontal accelera-
tion at the ground surface. This value
was taken as 0.15 g (1.471 m/s2) for
the Cape Flats area, representing the
probabilistic peak horizontal ground
acceleration for the area with 10%
chance of exceedance in 50 years
(SANS 2010).
QQ rd = stress reduction coefficient
accounting for the flexibility of the soil
column, and obtained as follows:
Ln(rd ) = a(z) + β(z)M(5)
⎫ z
α(z) = –1.012 – 1.126sin⎪ + 5.133⎪(6)
⎭ 11.73
⎫ z
β(z) = 0 .106 + 0.118sin⎪ + 5.142⎪
⎭ 11.28
for depth z ≤ 34 m
Where: M = design earthquake moment
magnitude = 6.0
The earthquake moment magnitude and
the local Richter scale magnitude are, for
all practical purposes, considered equal.
The Richter scale magnitude was acquired
from the Modified Mercalli Scale (MMS)
intensity which was, in turn, obtained from
a seismic intensities map of South Africa
showing probabilistic MMS intensities with
a 10% chance of exceedance in 50 years
(Brandt 2011; De la Harpe 2015).
The CSR requires adjustment for the
equivalent number of stress cycles in dif-
ferent magnitude earthquakes. The CSR
induced by an earthquake with magnitude
M is routinely adjusted to an equivalent
CSR for an earthquake magnitude equal to
7.5 (Idriss & Boulanger 2004) as follows:
CSR
(CSR)M=7.5 = (8)
MSF
Where:
MSF = magnitude scaling factor, given by:
⎫ –M ⎫
MSF = 6 .9exp⎪ ⎪ – 0.058
⎭ 4 ⎭
for MSF ≤ 1.8
26
⎫
⎭
Figure 5 In-situ test locations (Source: Google Earth 2019)
⎫
M = earthquake moment magnitude =
⎭
6.0, obtained as described above.
(7) The cyclic stress ratios, calculated at each
of the SPT depths in the manner described
above, were subsequently plotted against
the clean sand equivalent (N1)60 values on
the liquefaction triggering curve (Figure 4).
A total of 1 526 CSR (N1)60cs data pairs
were included to reveal the liquefaction
potential of the Cape Flats sands.
To illustrate the occurrence and thick-
nesses of liquefiable soil layers, the CRR
and CSR values were plotted relative to
the SPT depths (with linear interpolation
between points), providing continuous
profiles of the factor of safety against lique-
faction. Liquefaction can only occur below
the water table and, as such, the indicated
liquefiable zones may vary based on the
depth of groundwater.
To illustrate the effects of a larger
magnitude earthquake producing higher
PGAs on the triggering of cyclic liquefac-
tion in the study area, a design earthquake
with moment magnitude M = 7.5 and
amax = 0.25 was applied in the calculation
of the cyclic stress ratio.
CPT-based method for
flow liquefaction
To evaluate the susceptibility of the Cape
Flats sands to flow liquefaction, the potential
for these soils to strain-soften in undrained
(9) shear was evaluated. The results from four
Volume 62 Number 2 June 2020
piezocone penetrometer tests undertaken at
two locations in the study area were applied
to the CPT-based method proposed by
Robertson (2016). The CPT parameters of
normalised cone resistance Qtn and normal-
ised friction ratio Fr were calculated from the
raw CPT parameters of cone tip resistance
qc, sleeve friction fs, and pore water pressure
u, continuously recorded from surface to
between 9.1 m and 10.98 m depth, using the
following equations (Robertson 2010):
⎫ fs ⎫
Fr = ⎪ ⎪100%(10)
⎭ qt – σvo ⎭
Where:
qt = c one tip resistance corrected for pore
pressure
⎫ q – σvo ⎫⎫ pa ⎫
Qtn = ⎪ t ⎪⎪ ⎪ (11)
n
⎭ pa ⎭⎭ σ'vo ⎭
Where:
pa = atmospheric pressure (101.3 kPa)
converted to same units as qt and σv
n = stress exponent varying with soil
type, defined as:
⎫ σ' ⎫
n = 0.381(Ic) + 0.05⎪ vo ⎪ – 0.15(12)
⎭ pa ⎭
Where:
Ic = soil behaviour type index, defined as:
Ic = [(3.47 – logQtn )2 + (logFr + 1.22)2]0.5(13)
Journal of the South African Institution of Civil Engineering
 showing all the test locations. A total of
0.20 182 in-situ test locations, including CPTu’s
0.18 and boreholes with SPTs, were plotted on
Liquefiable the Google Earth image. Final standard
0.16
Non-liquefiable penetration test depths in the boreholes
0.14
range between 1.95 m and 42.25 m below
0.12 ground level (average final test depth of
11.23 m). The four piezocone penetrometer
CRR

0.10
tests were advanced to between 9.5 m and
0.08
11 m below ground level. The depth to the
0.06 water table in the study area, as recorded in
0.04 the boreholes and from pore water pressure
0.02 dissipation tests, varied between ground
surface level and 6.0 m below ground level
0
0 5 10 15 20 25 30 35 40 45 50 (average water table depth of 2.3 m).
(N1)60cs
Cyclic liquefaction potential
Figure 6 C
 yclic stress ratio plotted against clean sand equivalent SPT blow count for design The liquefaction triggering curve with CSR-
earthquake with M = 6.0 and amax = 0.15 g (N1)60cs data points calculated for a design
earthquake with moment magnitude of 6.0
and a PGA of 0.15 g is shown in Figure 6.
Liquefaction profile (MDup5BR1) The liquefiable and non-liquefiable soils are
WT
0 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50 separated, illustrating that the CRR mostly
0
exceeds the CSR, thus providing a factor
1 of safety greater than 1. Of the 1 526 data
2 pairs obtained from 178 boreholes plotted in
the figure, only 19 pairs from 14 boreholes
3
showed the potential to liquefy during an
Depth (m)

4 earthquake with the above-mentioned mag-

5
6 and CSR values, the water table levels at
7 the time of each specific ground investiga-
8
9 Liquefaction can only occur in saturated
CRR
Figure 7 Liquefaction analysis in a Cape Flats borehole
To obtain the value of n an iterative process
involving Equations 11 to 13 was followed.
The depth of the water table at each of
the four CPT localities – required for the
calculation of the total vertical stress – was
calculated from pore water pressure dis-
sipation test results.
The normalised cone resistance and
friction ratio values obtained with depth at
the four piezocone penetrometer test posi-
tions in the manner described, were plotted
on the SBTn chart shown in Figure 3
using the CPeT-IT software developed by
GeoLogismiki for interpreting CPTu data.
By plotting the normalised parameters
on Figure 3, the depths and thicknesses of
dilative and contractive soil zones in the
profiles are not discernible. For this reason,
a contractive-dilative (CD) value was
calculated using Equation 14 and plotted
CSR
against depth to illustrate the volumetric
behaviour during shear. When the value of
CD exceeds 70, the soils are expected to be
dilative at large shear strains, whereas CD
values below 60 represent contractive soils.
A transitional zone exists between CD
values of 60 and 70, which was inserted on
the graphs.
CD = (Qtn – 11)(1 + 0.06 Fr)17(14)
RESULTS AND ANALYSIS
The in-situ test locations are shown on
the Google Earth image in Figure 5, show-
casing the distribution of the tests and
liquefiable soils. It should be noted that,
due to the large geographical area shown in
the image, test locations near one another
plot in an overlapping manner, not clearly
nitude and acceleration (refer to Figure 5). It
should be noted that, in calculating the CRR
tion were used (investigations undertaken
over a period of approximately 30 years).
soils and, as such, the outcome presented
in Figure 6 may vary depending upon the
depth of the phreatic surface.
The potential of the Cape Flats soils to
liquefy under static conditions at the SPT
locations shown in Figure 5 could not be
determined, due to the lack of an appropri-
ate SPT-based flow liquefaction method.
To illustrate the occurrence and thick-
nesses of liquefiable soil layers in the 14
boreholes which showed a likelihood to
liquefy, the CRR and CSR values were plot-
ted relative to the SPT depths (with linear
interpolation between points). A continu-
ous profile of the factor of safety against
liquefaction is shown in Figure 7 for one
such borehole. At this position, liquefiable
soil occurs between approximately 3 m and
5 m depths. In most cases, only a single
CSR value calculated at an individual SPT
depth in the soil profile (borehole location)
was higher than its corresponding CRR
value. Interpolation between SPT depths
hinders exact definition of liquefiable soil

Journal of the South African Institution of Civil Engineering Volume 62 Number 2 June 2020 27
layers. Liquefiable soils occur between
upper and lower depth limits of 3.2 m and 0.6
19 m respectively, with layer thicknesses
varying from 0.3 m to about 2.0 m. Rauch 0.5
(1997), however, notes that liquefaction is Liquefiable

unlikely at depths exceeding 15 m where 0.4

Non-liquefiable
confining stresses are large and frictional

CRR
0.3
resistance is likely to prevent liquefac-
tion. Liquefiable soils were identified as
0.2
fine to coarse-grained sands and silty
sands in boreholes in Athlone, Philippi,
0.1
Khayelitsha, Macassar, and in the vicinities
of the Cape Town International Airport in
0
Matroosfontein and the Tygerberg Hospital 0 5 10 15 20 25 30 35 40 45 50
in Bellville. Liquefiable layers were not (N1)60cs
intersected at comparable elevations across
different sites in the study area. Figure 8 C
 yclic stress ratio plotted against clean sand equivalent SPT blow count for design
To illustrate the effects of a larger mag- earthquake with M = 7.5 and amax = 0.25 g
nitude earthquake producing higher PGAs,
on the triggering of cyclic liquefaction in
the study area, the liquefaction triggering Modified Robertson (2016) SBTn Modified Robertson (2016) SBTn
1 000 1 000
curve with CSR-(N1)60cs data points was
Normalised cone resistance (Qtn)

Normalised cone resistance (Qtn)

lB = 32 lB = 32
generated for a design earthquake with
moment magnitude of 7.5 and a PGA of SD
lB = 22 lB = 22
100 100 SD
CD = 70 CD = 70
0.25 g, as shown in Figure 8.
A substantial decrease in liquefaction
SC TD TD
SC
resistance is associated with the above- TC
CD
TC
CD
mentioned design earthquake, with safety 10 10
factors at about 35% of SPT locations being CCS CC CCS CC
less than 1 (CSR > CRR). The possibility of
such an occurrence (large seismic event) 1 1
0.1 1 10 0.1 1 10
is, however, improbable considering the
(a) Normalised friction, F (%) (b) Normalised friction, F (%)
seismic hazard maps produced for southern
Africa (Brandt 2011; De la Harpe 2015).
Modified Robertson (2016) SBTn Modified Robertson (2016) SBTn
1 000 1 000
Normalised cone resistance (Qtn)

Normalised cone resistance (Qtn)

Flow liquefaction potential lB = 32 lB = 32

Robertson’s normalised Soil Behaviour
lB = 22 lB = 22
Type (SBTn) chart (including ­contractive- 100 CD = 70
SD
100
SD
CD = 70
dilative boundary) with Qtn‑Fr data points
calculated from the field data of four TD TD
SC SC
CD CD
piezocone penetrometer tests (CPTu) TC TC
10 10
undertaken at two locations in the Cape
CCS CC CCS CC
Flats, are shown in Figure 9 (also refer to
Figure 5). Figures 9(a) and 9(b) depict the
1 1
results from Site A: Southwest Cape Flats, 0.1 1 10 0.1 1 10
and Figures 9(c) and 9(d) from Site B: (c) Normalised friction, F (%) (d) Normalised friction, F (%)
Northeast Cape Flats. Both l­ocalities are
underlain by Witsand Formation ­deposits, Figure 9 S BTn charts showing dilative-contractive response of Cape Flats soils during shear
although Site B is situated near the
Springfontein Formation soil boundary. deformations can occur during cyclic load- enable proper interpretation of the results,
The soil behaviour types were determined ing when a brief loss in shear resistance revealing depths and thicknesses of dilative
as sand, silty sand and sandy silt with occurs when the effective stress becomes and contractive soil zones, the contractive-
irregular layers of clay and silty clay zero. Deformations will, however, stabilise dilative (CD) values were plotted against
underlying Site B. when the cyclic loading ends. depth for each of the CPTu’s. A typical
From the SBTn charts it is evident that Loose, contractive sands, in which graph for one CPTu position is shown
the sands and silty sands from the Cape Qtn‑Fr data pairs plot below the CD bound- in Figure 10. Note that the contractive-
Flats mostly dilate during undrained shear. ary, are also present at all four locations. A dilative boundary is transitional, the upper
The soils plotting above the contractive- limited number of data points plot within and lower boundaries of this zone being
dilative (CD) boundary may be susceptible the transitional or clay-like zones (some of represented by the grey and orange vertical
to cyclic liquefaction, thus substantial which display contractive behaviour). To lines respectively.

28 Volume 62 Number 2 June 2020 Journal of the South African Institution of Civil Engineering
 of data points plot within the transitional
CD value or clay-like zones.
0 100 200 300 400 500 600 700 800 These strain-softening soils (prone to
0
strength loss) can also experience cyclic
1 liquefaction depending on ground geom-
WT etry (Robertson & Cabal 2012). However,
2 the author was unable to assess and com-
pare liquefaction under static and dynamic
3
C D conditions for the same materials, as there
Depth (mm)

4 were no sites with both CPT and SPT

5 In conclusion, it is difficult to assess the
6
7 occurrence of such an event will certainly
8
9
C = Contractive   D = Dilative
Figure 10 Contractive-dilative chart for one CPTu (Site B)
Examination of the four graphs con-
firms the presence of contractive soils at
shallow depths (extending to a maximum
depth of 0.6 m), and at depths between
4 m and 5.5 m at the Site B CPTu loca-
tions. At between 4 m and 5.5 m depth,
liquefiable layers vary in thickness from
0.1 m to 0.5 m. At these depths in the
soil profiles, a complete loss of strength
leading to flow (under static shear stress
exceeding the residual shear strength) can
occur during a seismic event. The water
table was recorded between 2.45 m and
2.9 m depth in the four CPTu’s. It is the
build-up of pore water pressure during
dynamic loading that leads to strength
loss and, as such, flow liquefaction is only
likely to occur in the saturated sands
below the water table.
CONCLUSIONS
To evaluate the susceptibility of Cape
Flats soil to undergo liquefaction during
and after a seismic event, the empirical
SPT-based method put forth by Idriss and
Boulanger (2004) was used to assess the
cyclic liquefaction potential of dense, dila-
tive sands, while the CPT-based method
published by Robertson (2016) was used to
assess the volumetric response of the sands
during shear.
The liquefaction triggering curve with
CSR-(N1)60cs data points obtained for a
design earthquake with moment magnitude
of 6.0 and a PGA of 0.15 g, illustrated
the resistance of the Cape Flats sands to
liquefy at these probable magnitude and
acceleration values (obtained from the seis-
mic hazard map of southern Africa: Brandt
2011). Of the 1 526 data pairs obtained
from 178 boreholes, only 19 pairs from 14
boreholes showed the potential to liquefy.
Liquefied thicknesses in these boreholes
ranged from 0.3 m to about 2.0 m, occur-
ring between depths of 3.2 m and 19 m in
the soil profile.
A substantial decrease in liquefaction
resistance was observed when the liquefac-
tion triggering curve with CSR-(N1)60cs
data points was generated for a design
earthquake with moment magnitude of 7.5
and a PGA of 0.25 g. Approximately 35% of
the data points plotted above the trigger-
ing curve, illustrating the effects of a large
ground shaking event on the triggering of
liquefaction in the Cape Flats. The pos-
sibility of such an occurrence (large seismic
event) is deemed unlikely.
The water table levels encountered in
the boreholes at the time of each specific
ground investigation were used to calculate
the values of CSR and CRR. Liquefaction
can only occur in saturated soils and, as
such, the outcomes presented on the lique-
faction triggering curves may vary depend-
ing on the depth of the phreatic surface.
Robertson’s normalised Soil Behaviour
Type (SBTn) chart (including contractive-
dilative boundary) revealed mostly dilative
sands and silty sands possibly susceptible
to cyclic liquefaction at the four CPTu
locations. Loose, contractive sands are also
present at all four locations, occurring at
shallow depths, and between 4 m and 5 m
at two CPTu locations. A limited number
data available.
risk of earthquake occurrence and mag-
nitude in the study area. Nonetheless, the
result in cyclic liquefaction in the Cape
Flats area, the extent of which will depend
on the magnitude of the seismic event. The
limited assessment of static liquefaction in
the study area inhibits a true reflection of
the volumetric response of sand and clay-
like soils from the Cape Flats during static
or cyclic loading.
ACKNOWLEDGEMENTS
The piezocone penetrometer testing was
made possible by Prof Peter Day and
Jones & Wagener (Pty) Ltd. Thank you
for investing in this liquefaction study. I
also thank the following companies for
making historic SPT records available
for research: Kantey & Templer (Pty) Ltd,
Melis & Du Plessis Consulting Engineers
(Pty) Ltd, Core Geotechnical Consultants,
Fairbrother Geotechnical Engineering,
and the Council for Geoscience (Bellville
Office).
REFERENCES
Brandt, M 2011. Seismic hazard in South Africa.
Council for Geoscience Report 2011–0061. Pretoria:
Council for Geoscience.
Cetin, K O, Seed, R B, Der Kiureghian, A, et al 2004.
Standard penetration test-based probabilistic
and deterministic assessment of seismic soil
liquefaction potential. Journal of Geotechnical and
Geoenvironmental Engineering, ASCE, 130(12):
1314–340.
De la Harpe, C W H 2015. The development
of a seismic risk reduction procedure for the
prioritization of low-cost, load-bearing masonry
buildings. MSc Dissertation. Stellenbosch
University.
Hill, R S & Theron, J N 1981. Silica sand of the Cape
Flats. Bulletin Vol 69. Pretoria: Geological Survey of
South Africa.
Idriss, J M & Boulanger, R W 2004. State normalization
of penetration resistances and the effect of

Journal of the South African Institution of Civil Engineering Volume 62 Number 2 June 2020 29
 overburden stress on liquefaction resistance.
Proceedings, 11th International Conference on
Soil Dynamics and Earthquake Engineering and
3rd International Conference on Earthquake
Geotechnical Engineering, Berkeley, California,
pp 22–56.
Idriss, J M & Boulanger, RW 2010. SPT-based
liquefaction triggering procedures. University of
California, Department of Civil Construction
and Environmental Engineering, Center for
Geotechnical Modeling.
NCEER (National Center for Earthquake Engineering
Research) 1997. In Youd, T L & Idriss, I M (Eds).
Proceedings, NCEER Workshop on Evaluation of
30
Liquefaction Resistance of Soils.Technical Report
NCEER-97-02. New York: NCEER.
Rauch, A F 1997. EPOLLS. An empirical method for
predicting surface displacements due to liquefaction-
induced lateral spreading in earthquakes. PhD
Thesis. Blacksburg, VA: Virginia Polytechnic
Institute and State University.
Robertson, P K 2010. Evaluation of flow liquefaction
and liquefied strength using the cone penetration
test. Journal of Geotechnical and Geoenvironmental
Engineering, 136(6): 842–853.
Robertson, P K & Cabal, K L 2012. Guide to Cone
Penetration Testing for Geotechnical Engineering, 5th
ed. Signal Hill, CA: Gregg Drilling and Testing, Inc.
Volume 62 Number 2 June 2020
Robertson, P K 2016. Cone penetration test (CPT)-
based soil behaviour type (SBT) classification
system – An update. Canadian Geotechnical
Journal, 53: 1910–1927.
SANS (South African National Standard) 2010. SANS
10160-4: Basis of Structural Design and Actions for
Buildings and Industrial Structures. Part 4: Seismic
Actions and General Requirements for Buildings.
Pretoria: SABS Standards Division.
Seed, R B, Cetin, K O, Moss, R E S et al 2003.
Recent advances in soil liquefaction engineering:
A unified and consistent framework. Report
No. EERC 2003‑06. Oakland, CA: Earthquake
Engineering Research Center.
Journal of the South African Institution of Civil Engineering
Warrants for roundabouts and TECHNICAL PAPER
other traffic control devices Journal of the South African
Institution of Civil Engineering
ISSN 1021-2019
J D Sampson Vol 62 No 2, June 2020, Pages 31–38, Paper 0701

DR JOHN SAMPSON (Pr Eng, FSAAE, FSAICE)

There are many warrants for traffic signals, but no definitive warrant for roundabouts. Some has a BSc Civil Engineering (Wits), MSc
Transportation Engineering (California,
references (e.g. TRB 2000) give guidelines, and refer to the capacities of roundabouts and TWSC
Berkeley), BCom (UNISA) and a PhD (Wits).
(two-way stop control), but capacities address the upper limit of the control device, not the His experience includes Managing Director
lower values which would justify, or warrant, a change in control. Other references refer to the of TTT Africa, Partner and Director of
Jeffares & Green Consulting Engineers (now
need for a roundabout warrant but do not develop one. One paper by Sampson and Meijer
JG Afrika), Professor at the University of
(1999) discussed potential warrants, but no recommendation was made. Pretoria, Deputy City Engineer of City of Johannesburg, Chairperson of the
In this paper the role of roundabouts and mini-circles in filling the large gap between where Board of Directors of the Road Traffic Management Corporation, Director
stop control is no longer ideal, and where traffic signals are required, is examined. of the Road Traffic Infringement Agency, Expert on Gauteng Transport
Commission and Member of the Gauteng Integrated Transport Master
It is found that there are many considerations in addition to volume or delay that need to be
Plan Steering Committee.
taken into account when deciding on whether to implement roundabout type control, but as
Contact details
a rule of thumb, if the average queue at a stop control exceeds two (2Q) during the peak hour,
Consultant: Traffic and Transportation
roundabouts will improve the traffic flow, and if the queue totals around ten or more (10Q), TTT Africa
then the roundabout is likely to have reached its capacity. PO Box 1109
Sunninghill 2157
Sandton
South Africa
Introduction The 4Q / 6Q Warrant for T: +27 11 231 2298
As part of a doctoral thesis, the author Traffic Signals (COTO 2001) E: sampsonj@tttafrica.co.za
developed a queue-based traffic signal war- For background, the 4Q / 6Q warrant is
rant (Sampson 1992; Sampson 1999). That discussed. This is a conventional warrant
4Q / 6Q warrant was subsequently adopted which determines the point at which
in the South African Road Traffic Signs conversion from a stop to a signal becomes
Manual (COTO 2001) as the official war- justified.
rant for traffic signals in South Africa. When developing the queue length
At the time (1980s) traffic circles had warrant for traffic signals (Sampson 1992),
been phased out in South Africa (and in the it was found that to use traffic volumes as a
USA and Australia) and roundabouts had warrant was not accurate, and the require-
yet to find international acceptance. While ments for a signal depend on many more
modern roundabouts were developed in the considerations than a simple traffic count
1960s in the United Kingdom, they only would reveal.
appeared in the USA in the 1990s, and in Nor could “tolerable delay”, which refer-
South Africa they are still relatively rare. ences suggested was an average delay of 57
At present therefore, existing warrants seconds per vehicle at a stop street, be used
such as the MUTCD (US Department of as a warrant. This is because signals would
Transport 2009) and SARTSM (COTO be required long before they should be
2001) only consider the point at which the justified where side street traffic volumes
conversion from a stop-controlled intersec- are low, while at high side street traffic vol-
tion to a traffic signal is justified. Traffic umes traffic signals are needed long before
signal warrants do not provide for rounda- the average delay of 57 seconds is reached.
bouts as an option. It was therefore found that, for the
Because of their safety and capacity above and many other reasons (such as
benefits, however, roundabouts are becom- accuracy, politics and ease of observation),
ing important traffic control devices and, the best measure to use for a warrant
for good reason, are becoming more and would be queue length. Conveniently so,
more common throughout the world. because queue length is numerically equal
In this paper, a warrant for mini-circles to veh-hrs/hr (volume) of total delay, and
and roundabouts is developed. This is done total delay is directly aligned with cost.
by comparing the performance of stop Queue can also be used to determine
street, all-way stop, mini-circle, modern user costs.
roundabout and traffic signal control Further research then led to the con-
across a wide range of traffic volumes and clusion that, when a peak hour queue of Keywords: traffic signals, roundabouts, warrants, queue, delay, stop street,
geometric layouts. vehicles or pedestrians reaches or exceeds all-way (4-way) stop, mini-circle

Sampson J. Warrants for roundabouts and other traffic control devices.

J. S. Afr. Inst. Civ. Eng. 2020:62(2), Art. #0701, 8 pages. http://dx.doi.org/10.17159/2309-8775/2020/v62n2a4 31
four on average, this was the point at which
traffic signals could be justified. In addi-
tion, if the sum of all the queues of vehicles
and pedestrians on average throughout
a full hour exceeded six, this would also
justify signals. This became the 4Q / 6Q
warrant (Sampson 1992; COTO 2001).
Note that this is a peak hour warrant.
As was stated in the Introduction, the
4Q / 6Q warrant however only considers
the point at which a side road stop control
should be converted to an intersection sig-
nal control. This research is now extended
to include roundabouts.
GUIDELINES FOR ROUNDABOUTS
Before analysing the technical performance
of roundabouts compared to other con-
trol devices, conventional guidelines are
considered.
Roundabouts equalise the priority of
all approach roads. No matter how minor
the intersecting road may be, it is afforded
the same priority on entry as any of the
major routes. Furthermore, all vehicles
must slow and take gaps on approaching
the roundabout, and priority cannot be
given to any movement without violating
the roundabout operational principles (e.g.
once a traffic signal or stop street is intro-
duced at a roundabout, it ceases to operate
as a roundabout).
With these limitations in mind, the
following is a summary of the Guidelines
for the Location of Roundabouts, extracted
from local and international literature
and as contained in Roundabouts (Traffic
Circles) as Intersection Control Devices on
Provincial Roads (PWV Consortium 2001).
The best locations for roundabouts are
as follows:
QQ Intersections where safety would other-
wise be a problem
QQ Intersections where environmental
enhancement or landscaping is required
QQ Intersections where traffic signal main-
tenance or the availability of power is a
problem
QQ Where permanent traffic control with-
out enforcement is necessary
QQ Where there are four-way or all-way
stops
QQ Where the road standard or speed limit
changes (e.g. where an arterial road
changes to collector / local street status)
QQ Where urban and rural roads meet
(e.g. entrance to towns) or where com-
mercial/industrial and residential areas
meet
32
QQ In suburban areas where traffic calming
is required
QQ At intersections with high turning
movements
QQ Where U-turns are prevalent or
desirable
QQ Where three or more stages are
required at traffic signals
QQ At intersections with more than four
legs
QQ At Y-junctions or other junctions with
awkward geometry (e.g. sharp change in
direction)
QQ With other roundabouts in a network
where intersection spacing is too close
for signal coordination to be achieved.
The worst locations for roundabouts are
as follows:
QQ Where main road and side road traffic
flows differ greatly
QQ Where minor crossroads enter major
routes when a stop or yield street would
suffice
QQ In signalised coordinated networks
where they would break up the platoon
flow
QQ Where traffic signals will soon be
required.
As is the case with all intersection control
devices, roundabouts should be avoided on
roads with steep slopes or where the inter-
section is not visible. Longer “flat” areas are
required for roundabouts compared with
other intersection types, making them less
suitable on steep grades.
WARRANT DEVELOPMENT
Intersection control devices
The research that follows contrasts and
compares the full range of intersection
control devices, namely conventional stops,
all-way stops, mini-circles, modern rounda-
bouts and traffic signals.
To do this, an intersection simulation
software program called AutoJ (Sampson
2018) is used. AutoJ is based primarily
on the formulae in the Highway Capacity
Manual (TRB 2010).
For a given set of geometric and traffic
volume conditions, AutoJ compares the
V/C (volume to capacity ratio), delay, queue
and level of service performance of five
forms of priority control and 21 different
traffic signal staging and phasing options
simultaneously. It does this for vehicles,
pedestrians and bicycles for up to three
periods (typically AM, PM and off-peak
hours) at a time.
Volume 62 Number 2 June 2020
To compare controls, AutoJ summarises
the performance of each in a table (see
Appendix A on page 38).
Typical intersection geometry
The investigation is limited to a maxi-
mum of two lanes on each approach as
roundabouts with three approach lanes
are rare and require detailed site-specific
investigation.
The number of approach lanes tested
include 2 × 2, 2 × 1 and 1 × 1. A 2 × 1
cross-junction would be as follows:
QQ Main road: 2 lanes in each direction,
with 1 right-turning lane added if traffic
signals are considered, 2 approach and 2
circulatory lanes if a roundabout is con-
sidered, and 1 approach lane and 1 “circu-
latory” lane if a mini-circle is considered.
QQ Side road: 1 lane in each direction,
with 1 right-turning lane added if traf-
fic signals are considered, 1 approach
and 2 circulatory lanes (for main road
traffic) if a roundabout is considered,
and 1 approach lane if a mini-circle is
considered.
Geometrically this favours the traffic
signal, as the addition of a right-turn bay
means there are more approach lanes for
that option. Nevertheless, it will be shown
that even with that advantage, the traffic
signal only becomes the optimal control at
very high volumes.
Traffic volumes
Many volume combinations were consid-
ered, with the two-way main road ranging
from 0 to 3 600 vehicles per hour (with a
60:40 split), the “high” side street volume
ranging from 0 to 1 400 per hour, and the
other side street approach set at 100 vehi-
cles per hour.
For the 2 × 1 geometry, typically a sub-
urban collector side street entering a minor
arterial road, the turning volumes on the
main road were taken as 10% left and
right, and 80% straight, while the turning
volumes on the side road were taken as 45%
left and right and 10% straight.
For the 1 × 1 geometry, typically two
collector streets joining, the turning
volumes on the main road were taken as
20% left and right, and 60% straight, while
the turning volumes on the side road were
taken as 33% left, right and straight.
For the 2 × 2 geometry, typically two
minor arterials intersecting, the turning
volumes on both the main road and the
side road were taken as 20% left and right,
and 60% straight.
Journal of the South African Institution of Civil Engineering
RESULTS Table 1 Maximum delay to worst affected movement (2 × 1)
Tables 1 to 5 are a summary of the most
High side
important findings. The first three, i.e. 1 (veh/hr)
Maximum delay (sec/veh) (2 × 1)
Tables 1, 2 and 3, have the same geometric
configuration (2 × 1) but different measures stop 306 858 1 275
of performance (delay, queue and level of
mini 244 512 736
service based on V/C). Tables 2, 4 and 5
1 400
have the same measure of performance
r-about 236 477 683
(queue), but different geometric configura-
tions (2 × 1, 1×1 and 2 × 2). signal 27 33 297

Measures of performance stop 71 686 1 173 1 465

There are various performance measures
that can be considered when determining mini 37 286 548 767
the efficiency of a control device. Most 1 200
commonly these are: r-about 36 244 485 690

QQ Volume to capacity ratio (V/C)

signal 27 27 36 445
QQ Average delay (seconds per vehicle)
QQ Total delay (vehicle-hours per hour)
stop 25 446 1 030 1 381 1 590
QQ Queues
QQ Level of service. mini 13 36 283 547 780
When developing the warrant, all these 1 000
measures were considered across the r-about 14 30 208 454 665
full range of traffic volumes and control
devices, and all are presented in the Tables. signal 27 24 22 20 28
All-way stops, however, do not appear in
stop 16 86 815 1 254 1 516 1 671
the Tables, as they are never the preferred
control under any circumstances.
mini 9 12 25 272 617 847
800
Optimal performance r-about 10 12 20 100 364 590
To easily see where a control device is
optimal, the blocks in Tables 1, 2, 4 and 5 signal 27 22 19 17 20 123
are coloured as follows:
QQ Stop green stop 13 19 458 1 043 1 392 1 599 1 720
QQ Mini-circle yellow
QQ Roundabout pink
mini 7 8 11 54 431 695 883
600
QQ Traffic signal blue.
r-about 8 9 10 13 24 163 419
Table 3 is coloured differently as will be
described. signal 24 18 15 19 22 25 498

Measure 1: Maximum delay (2 × 1) stop 11 12 26 620 1 144 1 454 1 637 1 743

The first comparison is the maximum
hourly average delay experienced by the mini 6 7 9 18 219 521 737 898
worst affected movement in a 2 × 1 geo- 400
metric situation. r-about 7 7 8 8 9 11 17 50
The black line in Table 1 is the dividing
signal 20 14 18 21 24 25 243 774
line between stop control and traffic signal
control if mini-circles and roundabouts are
stop 10 10 11 17 400 1 020 1 385 1 695 1 879
not considered. It was previously indicated
that when developing the 4Q / 6Q warrant, mini 5 6 8 12 37 322 569 754 898
one consideration for warranting signals 200
was when the average delay at the stop r-about 7 7 7 8 8 10 12 18 45
exceeded 57 seconds (Sampson 1992). Even
though this was not used as a warrant, and signal 13 19 23 23 23 23 158 567 938
it is not easy to see in Table 1, the black line
400 800 1 200 1 600 2 000 2 400 2 800 3 200 3 600
appears approximately where the 57 second
cut-off would be.
Main 2-way (veh/hr)
For reasons stated earlier, maximum
delay is not the best criterion to be used Mini circle Traffic signal Roundabout
when warranting a control device. Were

Journal of the South African Institution of Civil Engineering Volume 62 Number 2 June 2020 33
it to be used, however, it is interesting to Table 2 Queue (2 × 1)
see that a stop control is never optimal (no
High side
block in Table 1 is coloured green). This 2 (veh/hr)
Total queue / delay (veh-hrs/hr) (2 × 1)
is because users of a mini-circle or round-
about will not need to stop in low flow stop 93.8 302.6 431.1
conditions, hence will always experience
mini 95.6 201.7 375.0
less delay than at a stop street.
1 400
r-about 92.7 187.1 268.5
Measure 2: Queue (2 × 1)
Table 2 is the result when queue length is signal 6.0 9.8 61.5
used as the criterion to choose a control
device. The numbers in Table 2 are the stop 17.4 197.8 326.3 406.0
total queue length (or total delay in vehicle-
hours per hour; whichever is preferred as a mini 13.0 97.3 228.3 433.1
measure, they are numerically equal). 1 200
As above, the black line is the split r-about 12.9 83.2 164.4 234.8

between a stop and traffic signal control

signal 5.2 8.7 13.6 116.4
if mini-circles or roundabouts are not
considered. It can be noted that when the stop 6.0 92.9 221.4 301.1 355.5
queue length exceeds four at the stop, traf-
fic signals are warranted, thus confirming mini 4.5 11.8 87.3 281.1 475.0
the 4Q warrant. 1 000
In Table 2, the inclusion of mini-circles r-about 4.9 9.9 60.4 130.3 191.9
and roundabouts changes the 4Q warrant
picture considerably. Clearly, mini-circles signal 4.6 7.8 10.5 13.2 16.1
and roundabouts provide the optimal
stop 3.4 12.7 116.5 196.3 249.2 300.2
solution over a wide range between the
stop and signal options. Neglecting the
mini 2.7 4.3 9.7 122.1 317.1 502.6
inclusion of these options, as the 4Q / 6Q 800
warrant does, will mean that the optimal r-about 3.1 4.5 7.1 26.1 86.7 142.0
control will not be considered. An impor-
tant conclusion of this research therefore is signal 4.2 6.9 9.1 11.2 13.4 29.3
that previous warrants need to be revised
to include the mini-circle and roundabouts stop 2.2 3.1 37.1 91.4 143.6 194.5 243.4
as options.
mini 1.9 2.9 4.8 18.3 154.8 338.0 520.5
Looking more closely at Table 2, a mini-
600
circle or roundabout appears to become
r-about 2.2 3.2 4.3 5.9 9.2 35.0 86.7
justified when the queue at the stop
exceeds two vehicles (2Q). signal 3.7 5.9 7.6 9.2 10.8 12.9 37.7
The upper limit of queue before a
signal is warranted is less obvious because, stop 1.4 1.6 2.5 32.7 59.9 94.3 138.1 204.4
as the main road volumes increase, the
length of queue required to warrant signals mini 1.4 2.2 3.5 7.0 76.0 213.4 356.7 587.1
also increases (from eight to twelve). For 400
warrant purposes, when the roundabout r-about 1.7 2.5 3.4 4.5 5.9 8.0 12.2 29.9
queues exceed approximately ten (10Q), a
signal 3.2 4.8 5.9 6.9 8.1 9.7 23.6 64.5
traffic signal control may perform better.
Note, however, that this is purely a
stop 0.8 0.9 1.0 1.4 11.5 36.0 73.3 131.5 180.2
peak-hour consideration. When volumes
drop off-peak, the roundabout will perform mini 1.0 1.8 2.8 4.8 14.8 132.7 275.3 479.7 716.9
better than the signal. 200
r-about 1.2 2.0 2.8 3.8 5.0 6.5 8.8 13.5 32.0
Measure 3: Level of service (2 × 1)
In Table 3, the level of service, based on signal 2.4 3.3 4.0 5.0 6.2 8.2 15.2 55.0 90.3
the volume to capacity ratio, is coloured as
400 800 1 200 1 600 2 000 2 400 2 800 3 200 3 600
follows:
QQ LoS A and B (V/C < 0.8) = green
Main 2-way (veh/hr)
QQ LoS C and D (V/C < 0.95) = blue
QQ LoS E (V/C < 1.0) = yellow Mini circle Traffic signal Roundabout Stop
QQ LoS F = red

34 Volume 62 Number 2 June 2020 Journal of the South African Institution of Civil Engineering
Table 3 Level of service (2 × 1) Table 3 is different from the other Tables
in that LoS is colour-coded and not the
high side
3 (veh/hr)
Level of service based on V/C (2 × 1) preferred control. Because traffic signals
have the greatest capacity, they will always
stop 1.13 1.75 2.94 have the lowest V/C regardless of volume. If
Table 3 were colour-coded in the same way
mini 1.11 1.33 1.59
as the others therefore, it would be all blue.
1 400
For that and other reasons, V/C is not
r-about 1.10 1.29 1.52
a suitable measure for a warrant (there is
signal 0.58 0.73 1.10 therefore no black line to indicate the stop
/ traffic signal boundary). It is, however,
stop 0.97 1.50 2.52 4.26 suitable to determine when the capacity of
the different control devices is reached.
mini 0.95 1.14 1.36 1.63 In Table 3 it can be noted that there is a
1 200 large area of overlap where the stop, mini-
r-about 0.94 1.11 1.30 1.53 circle, roundabout and traffic signal have
a similar level of service, indicating that
signal 0.50 0.54 0.84 1.21
there is a large range over which any of the
stop 0.81 1.25 2.10 3.55 6.05
controls could be appropriate from an LoS
viewpoint.
mini 0.79 0.95 1.14 1.36 1.65
1 000 Measure 4: Queue (1 × 1)
r-about 0.79 0.92 1.08 1.27 1.49 Table 2 provided the queue warrant with a
2 × 1 geometric configuration. Tables 4 and
signal 0.41 0.47 0.54 0.61 0.84 5 give the queue warrant with a 1 × 1 and
2 × 2 geometric layout.
stop 0.65 1.00 1.68 2.84 4.84 8.30
In this case of single lane approaches,
as expected, the mini-circle, being smaller
mini 0.63 0.76 0.91 1.13 1.44 1.76
800
and causing less delay, is generally slightly
r-about 0.63 0.74 0.87 1.02 1.20 1.41 better than the roundabout.
However, similar threshold values as
signal 0.34 0.41 0.47 0.56 0.71 1.00 in the 2 × 1 case can be observed, i.e. that
the mini-circle / roundabout is preferred
stop 0.49 0.75 1.26 2.13 3.63 6.23 10.74 when the queue at the stop reaches around
two (2Q), and the traffic signal becomes
mini 0.47 0.57 0.72 0.98 1.25 1.53 1.82
warranted when the total queue at the
600
mini-circle / roundabout reaches about ten
r-about 0.47 0.55 0.65 0.76 0.90 1.05 1.24
(10Q).
signal 0.27 0.34 0.40 0.52 0.69 0.87 1.27
Measure 5: Queue (2 × 2)
stop 0.32 0.50 0.84 1.42 2.42 4.15 7.16 12.44 The optimal control using queue as a crite-
rion with a 2 × 2 geometric configuration is
mini 0.32 0.41 0.63 0.86 1.09 1.34 1.59 1.85 shown in Table 5.
400 In a multi-lane situation, it would be
r-about 0.31 0.37 0.43 0.51 0.60 0.70 0.83 0.97 expected that a two-lane roundabout
would perform better than a single lane
signal 0.20 0.27 0.34 0.40 0.69 0.83 1.07 1.58
mini-circle, and this is indeed the case.
Again, it is found that the roundabout
stop 0.16 0.25 0.42 0.71 1.21 2.08 3.58 9.78 999.99
is preferred to a stop control at a queue
mini 0.18 0.36 0.55 0.75 0.95 1.16 1.38 1.61 1.85 length of about two (2Q), while the traffic
200 signal becomes the preferred device a little
r-about 0.16 0.19 0.29 0.39 0.50 0.61 0.72 0.84 0.96 later than in the single-lane cases, when the
queue at the roundabout exceeds between
signal 0.13 0.20 0.28 0.37 0.56 0.87 1.02 1.33 1.86 ten to fifteen (10Q).

400 800 1 200 1 600 2 000 2 400 2 800 3 200 3 600

LoS A and B (V/C < 0.8)
ConclusionS and
Main 2-way (veh/hr)
Recommendations
LoS C and D (V/C < 0.95) LoS E (V/C < 1.0) LoS F There are several recommendations
and conclusions that can be made

Journal of the South African Institution of Civil Engineering Volume 62 Number 2 June 2020 35
resulting from this research. The most Table 4 Queue (1 × 1)
important are:
High side
QQ Internationally, traffic signal warrants 4 (veh/hr)
Total queue / delay (veh-hrs/hr) (1 × 1)
need to be amended, because when a
stop street reaches capacity, the traffic stop 214.6 427.1 555.6
signal should not automatically be con-
mini 73.4 191.4 408.2
sidered as the next best option.
1 400
QQ When a stop control is not functioning
r-about 73.0 163.5 360.6
optimally, or when the queue exceeds
two on average at the stop throughout signal 6.7 12.3 26.0
the hour (2Q), this is an indication that
a mini-circle or roundabout should be stop 141.9 322.2 450.6
considered.
QQ When the total of queues waiting on mini 9.8 69.1 263.8
approaches to a roundabout reach ten 1 200
(10Q) in the peak hour, traffic signals r-about 10.3 57.4 216.3

may be considered.
signal 5.7 10.2 15.8
QQ Mini-circles and roundabouts have
therefore been found to provide the stop 71.0 217.2 345.7 417.7
optimum control in the wide range of
situations ranging between very low and mini 4.2 8.6 112.1 320.2
very high intersection volumes. 1 000
QQ Mini-circles are slightly better than r-about 4.7 8.3 65.6 268.3
roundabouts in single-lane approach,
low main road volume situations, while signal 5.0 8.7 12.6 241.6
roundabouts are better for multi-lane,
stop 9.3 112.8 240.7 312.6
higher main road volume intersections.
QQ There is significant overlap where more
mini 2.7 4.2 16.9 164.3
than one device will perform well. While 800
the colour coding in Tables 1, 2, 4 and r-about 3.1 4.6 10.6 131.8
5 tends to emphasise a clear division
between one control and another, exami- signal 4.4 7.4 10.3 16.6
nation of the numbers shows that different
controls perform similarly over a wide stop 2.8 37.9 135.8 207.6 254.8
range. This is most easily seen in Table 3.
QQ Queue length was found to be the best
mini 1.9 2.9 5.7 88.8 232.2
600
indicator of when a stop should be
r-about 2.3 3.4 5.8 66.0 207.8
converted to a traffic signal (Sampson
1992), and this research confirms that it signal 3.9 6.2 8.3 10.9 99.2
is also the best measure to use as a war-
rant when considering roundabouts. stop 1.5 2.4 32.6 102.7 147.9
QQ There are, however, additional consid-
erations for roundabouts, as indicated in mini 1.4 2.2 3.8 16.9 157.3
the Guidelines quoted. 400
QQ In the research for AutoJ (Sampson r-about 1.7 2.7 4.3 12.7 138.8
2018), to take into account that the con-
signal 3.2 4.9 6.2 7.8 23.1
trol device must operate over 24 hours
and not just the peak hour, a perfor-
stop 0.8 1.0 1.4 3.6 40.9 84.0
mance index was developed (Appendix
A). This is a more sophisticated approach mini 1.0 1.8 3.0 6.1 71.2 223.6
than using a simple single-number test 200
during one hour, as suggested by a war- r-about 1.3 2.2 3.4 6.4 60.3 209.1
rant. The warrant derived in this paper
can, however, be used as the basis for the signal 2.4 3.3 4.0 5.2 7.9 61.1
need for further investigation.
400 800 1 200 1 600 2 000 2 400 2 800 3 200 3 600
QQ Regardless of the technical calculations
above, engineering judgement remains
Main 2-way (veh/hr)
an important consideration, because
there are situations where roundabouts Mini circle Traffic signal Roundabout Stop
may not be preferred, e.g. when a traffic

36 Volume 62 Number 2 June 2020 Journal of the South African Institution of Civil Engineering
Table 5 Total delay / queue (2 × 2) signal already exists, or when there is
insufficient road reserve, or when any of
High side
5 (veh/hr)
Total queue / delay (veh-hrs/hr) (2 × 2) the worst locations for roundabouts list-
ed in the “Guidelines for Roundabouts”
stop 93.4 310.3 488.5 595.8 659.9 707.9 747.0 824.0 section above are present.
In conclusion, this research finds that:
mini 62.6 230.4 446.6 650.5 844.3 1 035.0 1306.4 1586.0
QQ An average of two vehicles in a queue (2Q)
1 400
at a stop street is the warrant to consider
r-about 4.4 6.0 13.3 140.8 293.5 519.8 740.4 950.2
conversion to roundabouts or mini-circles.
signal 6.1 8.7 12.0 16.0 25.6 34.8 80.3 210.8 QQ When the total queue on all approaches
to the roundabout reaches around ten
stop 18.6 224.7 383.5 489.9 553.5 602.1 641.6 718.7 (10Q), that is an indication that the
roundabout is nearing capacity and that
mini 8.9 85.8 303.6 509.1 704.4 896.5 1169.2 1450.1 traffic signals and possibly road widen-
1 200 ing should be considered.
r-about 3.6 4.8 7.3 64.4 207.1 357.9 570.7 782.8 QQ There is a wide range between the 2Q
and 10Q conditions where roundabouts
signal 5.5 8.0 10.9 14.0 22.4 28.8 42.6 134.8
would be the optimal form of control,
stop 6.2 140.0 278.4 384.8 446.8 496.2 536.1 613.4 725.8
and they could operate well even outside
this range.
mini 4.1 9.4 152.0 359.3 556.4 750.2 1024.6 1307.1 1582.0 QQ The total intersection volume range
1 000 over which roundabouts are optimal
r-about 3.0 4.0 5.6 11.3 120.1 264.5 412.3 603.1 810.0 is large, roughly from 1 000 to 2 400
vehicles per hour, indicating that, with
signal 5.0 7.3 9.7 12.2 19.5 24.4 32.6 65.4 186.7 average traffic growth, the roundabout
would be optimal for around 30 years
stop 3.5 55.7 178.9 279.7 339.7 390.0 430.5 508.1 620.6
without modification.
It is further recommended that all-way stops
mini 2.6 4.4 60.7 205.8 398.6 594.6 871.0 1155.4 1432.2
800
are never used in urban areas, and where
r-about 2.5 3.4 4.6 6.8 23.2 165.5 312.3 459.9 618.8 they exist they should be replaced with
mini-circles or, if need be, roundabouts.
signal 4.1 6.5 8.5 10.4 16.6 20.7 26.4 40.2 139.5 Over and above the unnecessary delay and
queues caused by all-way stops, motorists
stop 2.2 4.6 93.6 174.6 234.3 283.6 324.7 402.6 515.3 regularly ignore the sign and proceed
without stopping, thereby effectively turning
mini 1.9 3.0 8.3 128.4 270.2 427.5 706.4 993.2 1272.2
the control into a yield from all sides.
600
Formally recognising this typical behaviour
r-about 2.0 2.9 3.9 5.4 8.6 49.0 199.3 349.4 501.1
by converting the all-way stop into a mini-
signal 3.6 5.6 7.1 8.5 10.2 17.1 22.5 37.4 138.7 circle or roundabout will turn offenders into
law-abiding citizens.
stop 1.4 1.8 9.2 77.7 129.2 176.8 218.7 297.1 410.0

mini 1.4 2.3 4.1 43.9 187.8 337.3 575.0 828.5 1099.9 References
400 COTO (Committee of Transport Officials) 2001. South
r-about 1.6 2.4 3.3 4.5 6.3 10.9 65.7 220.2 375.9 African Road Traffic Signs Manual. Vol 3, Chapter 2.
Pretoria: National Department of Transport.
signal 3.1 4.6 5.6 6.5 7.9 14.5 19.9 33.1 138.7
PWV Consortium 2001. Roundabouts (traffic circles)
as intersection control devices on provincial
stop 0.8 0.9 1.1 2.3 31.1 69.9 112.5 191.4 304.6
roads. BL 99/5 Draft Report. Pretoria: Gauteng

mini 1.0 1.8 3.0 6.7 89.6 242.6 483.6 738.7 991.5 Department of Roads and Transport.

200 Sampson, J D 1992. Warrants for traffic signals.

r-about 1.2 2.0 2.9 3.9 5.3 7.5 12.9 67.0 227.8 PhD Thesis. Johannesburg: University of the
Witwatersrand.
signal 2.3 3.1 3.9 4.9 6.3 12.8 17.5 29.2 136.8 Sampson, J D 1999. Queue-based traffic-signal warrants:
The 4Q/6Q Warrant. Institute of Transportation
400 800 1 200 1 600 2 000 2 400 2 800 3 200 3 600
Engineers (ITE) Journal, 69, (4): 30–32.
Sampson, J D 2018. AutoJ Traffic Engineering
Main 2-way (veh/hr)
Software, Version &AutoJ 1808, Sandton, South

Mini circle Traffic signal Roundabout Stop Africa. Available from: sampsonjds@gmail.com;
www.‌autojtraffic.com.

Journal of the South African Institution of Civil Engineering Volume 62 Number 2 June 2020 37
Sampson, J D & Meijer M 1999. The use of modern
roundabouts on provincial roads. Proceedings,
18th South African Transport Conference,
Pretoria.
TRB (Transportation Research Board) 2000.
Roundabouts: An Informational Guide. National
Cooperative Highway Research Program (NCHRP)
Report 672. Washington, DC: National Research
Council.
TRB (Transportation Research Board) 2010. Highway
Capacity Manual. Washington, DC: National
Research Council.
US Department of Transport 2009. Manual of Uniform
Traffic Control Devices (MUTCD), Washington, DC:
Federal Highway Administration.
APPENDIX A: AUTOJ SUMMARY EXAMPLE
A typical AutoJ software summary (Sampson
2018) is given in Appendix A below. In the
summary the level of service using delay and
V/C numbers as performance measures is
colour-coded, with the optimal control(s)
shown in green-shaded blocks.
AutoJ uses a performance index to take
into account that the control device must
operate over 24 hours and not just the peak
hour. The performance index weights the
control’s efficiency as follows:
QQ maximum V/C of any movement (more
critical than delay in peak hours): 18%
AM and PM peaks, 4% off-peak
QQ average V/C of intersection combining
peak and off-peak: 10%
QQ maximum delay to any movement
(delay should be minimised if V/C
is not a concern, hence off-peak is
favoured): 4% AM and PM peaks,
7% off-peak
QQ overall intersection delay per vehicle
during worst period: 10%
QQ maximum queue on any approach:
6% AM and PM peaks, 3% off-peak
QQ total queue adding AM, PM and
off‑peak: 10%.

APPENDIX A Main Road (two lanes per direction) and Side Street (one lane per direction)
2 × 1 with added right turn lane if a signal

Summary

Volume / capacity (maximum) Delay / vehicle (maximum) Queue (maximum)

Weighting 100% 18% 4% 18% 10% 4% 7% 4% 10% 6% 3% 6% 10%
Perf index Peds AM Off PM i/s ave Peds AM Off PM i/s ave Peds AM Off PM i/s sum
Best ICD RR 86% 0.00 0.65 0.22 0.47 0.31 2 10 7 8 8 0.0 1.2 0.5 1.3 10.4
Best signal n3 63% 0.00 0.47 0.25 0.54 0.32 38 21 20 28 17 0.0 2.1 1.1 3.0 21.5
Control type PI 0.00 AM Off PM i/s ave Peds AM Off PM i/s ave Peds AM Off PM i/s sum
Xns 32% 0.00 0.97 0.37 1.19 0.48 85 68 11 374 65 0.0 10.8 0.9 28.0 52.0
Xwe 48% 0.00 1.27 0.34 0.77 0.29 100 465 10 22 71 0.0 34.9 0.3 0.4 40.5
Priority XX 37% 0.00 1.09 0.29 0.73 0.53 236 11 19 96 0.0 37.8 0.9 3.1 61.2
CC 57% 0.00 0.72 0.33 0.92 0.51 6 11 6 28 17 0.0 1.7 0.5 4.5 15.5
RR 86% 0.00 0.65 0.22 0.47 0.31 2 10 7 8 8 0.0 1.2 0.5 1.3 10.4
2 stage 2 52% 0.00 0.41 0.25 1.14 0.31 33 15 15 349 58 0.0 2.4 0.7 26.2 41.6
3ns 60% 0.00 0.46 0.26 0.56 0.34 38 20 20 28 19 0.0 3.2 1.1 3.1 24.6
3we 40% 0.00 0.46 0.35 1.78 0.37 38 20 20 904 140 0.0 3.2 1.1 67.8 88.5
n3 63% 0.00 0.47 0.25 0.54 0.32 38 21 20 28 17 0.0 2.1 1.1 3.0 21.5
3 stage
s4 39% 0.00 0.47 0.36 1.81 0.36 38 21 20 921 141 0.0 3.4 1.1 69.0 88.1
w3 39% 0.00 0.47 0.36 1.82 0.37 38 21 19 922 143 0.0 3.4 1.1 69.2 89.4
e3 39% 0.00 0.47 0.36 1.82 0.39 38 21 21 922 143 0.0 3.4 1.1 69.2 90.8
4nswe 51% 0.00 0.50 0.34 0.64 0.38 43 25 25 33 25 0.0 4.1 1.4 4.4 31.6
n4we 53% 0.00 0.54 0.28 0.63 0.36 43 27 25 33 21 0.0 2.8 1.5 4.3 28.0
s4we 32% 0.00 0.52 0.49 2.88 0.44 43 26 25 1 283 195 0.0 4.2 1.5 96.2 120.6
w4ns 50% 0.00 0.52 0.35 0.67 0.38 43 26 24 33 26 0.0 4.2 1.5 4.7 31.3
4 stage e4ns 49% 0.00 0.52 0.35 0.67 0.40 43 26 25 33 26 0.0 4.2 1.5 4.7 32.7
nw4 53% 0.00 0.54 0.29 0.66 0.35 43 27 25 33 22 0.0 2.9 1.6 4.6 27.5
ne4 52% 0.00 0.54 0.29 0.66 0.37 43 27 26 33 22 0.0 2.9 1.6 4.6 29.1
sw4 31% 0.00 0.54 0.50 2.89 0.44 43 27 25 1 287 196 0.0 4.4 1.5 96.5 120.2
se4 31% 0.00 0.54 0.50 2.89 0.45 43 27 26 1 287 196 0.0 4.4 1.5 96.5 121.8
n-s-3 49% 0.00 0.56 0.33 0.57 0.42 43 34 28 33 29 0.0 4.5 2.0 3.8 37.4
w-e-3 34% 0.00 0.53 0.39 1.89 0.43 43 32 27 965 154 0.0 4.3 1.5 72.4 101.7
Separate
n-s-4we 44% 0.00 0.61 0.36 0.62 0.45 48 40 33 38 34 0.0 5.4 2.4 4.6 44.2
(split) stage
w-e-4ns 43% 0.00 0.58 0.38 0.75 0.44 48 37 32 38 33 0.0 5.2 1.9 5.9 41.7
n-s-w-e-4 37% 0.00 0.73 0.44 0.78 0.55 48 45 36 42 40 0.0 6.4 2.7 6.2 52.2
Optimums 86% 0.41 0.22 0.47 0.29 10 6 8 8 1.2 0.3 0.4 10.4
Colour code Near best Best Within 10% of optimum LOS A-B C-D E F

38 Volume 62 Number 2 June 2020 Journal of the South African Institution of Civil Engineering
The cost-effectiveness of TECHNICAL PAPER
water augmentation and Journal of the South African
Institution of Civil Engineering

management: Assessing ISSN 1021-2019

Vol 62 No 2, June 2020, Pages 39–44, Paper 0840

the Unit Reference Value ROZANNE BESTER is a system dynamics

modeller and researcher affiliated with the
Institute of Water Research (Rhodes
University). She obtained a Master’s degree
R Bester, J N Blignaut, P H van Niekerk in Economics at the University of Pretoria in
2016 and has since worked in restoration
and development projects.

As societies develop, the problem of securing and mobilising adequate water resources Contact details
to support and sustain economic and social activities increases in complexity. As a cost- Postnet Suite 47
effectiveness measure the use of the Unit Reference Value (URV) has become standard practice Private Bag X16
Highveld Park 0169
in South African water resource management and development. It is informative and easy to
South Africa
understand. These two attributes inspire confidence and an ever-greater uptake of its use in T: +27 82 612 2471
evaluating suggested courses of action to be taken by society in ensuring their future. Deviations E: bester.roz@gmail.com
in how an URV is calculated emerge due to both the diversification of its user base and the
dynamic nature of the contexts in which it is applied. Such developments add to the pluralistic JAMES BLIGNAUT is a natural and
environmental resource economist attached
framings within which research and decision-makers operate. We provide three sub-types of
to both the School of Public Leadership,
URV application, and caution against comparing URVs across categories. To mitigate possible Stellenbosch, and the South African
confusion, we propose the development of standards/rules for the interpretation of URVs. Environmental Observation Network
(SAEON). He focuses on the economics of
the restoration of degraded landscapes
under the motto: The economics of
INTRODUCTION VRK-framework, is based on the work of restoration and the restoration of economics. He is especially focused on
The multifarious use, and even occasional Goddard et al (2016), and has been developed the advancement of regenerative agriculture and, within that context,
abuse, of research methods and/or indicators for the analysis of contested, complex and proper resource use and allocation inclusive of the economic benefits
of restoration.
are common features within the natural cross-scale “problems”. It is a very useful
development of science and scientific pro­ framework in the interpretation of the causes Contact details
SAEON Pretoria
gress. These multifarious uses often persist of the multifarious use of research methods and
due to a resistance to integrative and colla­ and indicators, as emerging from the interac- School of Public Leadership
borative learning approaches that emanate tion between a universal value for the use Stellenbosch University
Stellenbosch
from the structure of the various formal and of a particular method, and the evidence
South Africa
informal institutions involved in knowledge- and practice-based knowledge inequalities P: PO Box 144, Derdepark, 0035, South Africa
generating systems (Angelstam et al 2013). between user groups or individuals. T: +27 84 720 4127
The gradual entrenchment of indicator Here we apply the VRK-framework to E: jnblignaut@gmail.com

adaptations has been cited as the cause for
the increase in statistical errors in medical
research (Glantz 1980), the development of
expensive and ineffective educational prac­
tices (Alferink & Farmer-Dougan 2010), pri­
vate sector underinvestment (Drury & Tayles
1997), and the increase in the incidence of
physical injury caused to patients by medical
practitioners (Brisson & Brisson 2010). To
tunnel through this pervasive resistance
requires occasional reflection or, to borrow
jargon from systems thinking, feedback to
recalibrate. The question is, how do we con­
duct such mental feedback?
One proposed way is a systematic look
at the Values (or norms), system Rules (or
policies), and prevailing Knowledge at various
stages (i.e. over time) and how that shaped
the application, and thus change in and the
use of any particular research methods and/
or indicators. This model, herein called the
develop a deeper understanding of why, and
how, South Africa’s Unit Reference Value
(URV) has been used in the recent past,
and more specifically, how and why it has
undergone various indicator adaptations.
The URV is a cost-effectiveness measure
developed for use in the pre-planning stages
of water resource management and develop-
ment projects. It assists decision-makers
in allocating scarce resources in the most
effective way by comparing, in a consistent
manner, various water development schemes.
Over time, however, the URV has been
applied in various settings by an increas-
ingly diversified user base. This resulted in a
cumulative series of adaptations to the URV
from a diverse knowledge base following the
continuous application of it within varied new
organisational and environmental contexts.
While it could be argued that it would appear
that the multifarious use of the URV has
DR PETER VAN NIEKERK (Pr Eng, FSAICE) has
more than 40 years of experience in the
water resource planning and engineering
field. From 1987 to 2009 he was responsible
for all water resources planning in the
Department of Water Affairs, South Africa.
His extensive background includes
international water resource sharing and
management, and advanced academic studies in water resource
economic analysis. He continues to provide specialist input with respect
to the planning of hydrological, water supply and water
resources projects.
Contact details
Water Resource Appraisal Services
PO Box 47
Onrus River 7201
South Africa
T: +27 28 316 1968
E: waterappraisal@gmail.com
Keywords: Unit Reference Value (URV), Water Supply Management (WSM),
Integrated Water Resources Management (IWRM), Values-Rules-
Knowledge (VRK) framework, decision-making

Bester R, Blignaut JN, Van Niekerk PH. The cost-effectiveness of water augmentation and management: assessing the Unit Reference Value.
J. S. Afr. Inst. Civ. Eng. 2020:62(2), Art. #0840, 6 pages. http://dx.doi.org/10.17159/2309-8775/2020/v62n2a5 39
supported useful policy interventions and
necessary shifts in the narrative around the
management of water resources, there is a risk
that it might well be correlated with increased
ineffectiveness in water management. This
is due to the fact that the adaptations of the
URV have remained largely unchallenged.
This paper is an attempt to rectify this situa-
tion by conducting a critical review, based on
the VRK-framework, of the use of the URV
indicator in South Africa.
We commence with a discussion on the
natural progression of the context within
which the URV was developed, and which has
also given rise to its multifarious use based
on the VRK-framework. Thereafter we take a
critical look at the various ways the URV has
been calculated, and the implications thereof.
This is followed by recommendations with
respect to the future development thereof.
CONTEXTUAL PROGRESSION
AND THE GROWTH IN
URV THINKING
Introduction
The URV was developed in the 1980s by the
then Department of Water Affairs. It was
developed as a cost-effectiveness measure
that computes the unitary cost of supply-
ing a cubic metre of water at the required
assurance of supply, over the portion of the
water management or augmentation project’s
lifespan during which it produces economic
benefits for society. In its most basic form the
URV is calculated as the discounted present
value of the total (capital and operational)
lifecycle cost of a water augmentation or man-
agement scheme divided by the discounted
incremental water availability assured by the
expansion of the system and bounded by the
projected demand curve until full capacity
is reached (Van Niekerk & Du Plessis 2013a).
Discounting a natural resource stock such
as water is a contested issue (Costanza et al
1989). However, the URV has adopted this
approach in order to be able to compare, in a
consistent manner, the URV ratios of different
water development schemes, operating with
different time horizons, which have different
starting dates, thus avoiding an artificial
lowering of the cost of water development by
various schemes.
The URV, because of its simplicity
in calculation and the ensuing intuitive
understanding of the message it is carrying,
has become a very powerful indicator with
respect to water resource management and
development. Over time it has become not
40
Decision of the
context 1980s
Solve water scarcity by
V
bankrolling increasingly
complex and expensive
engineering solutions
WSM
Nationalism
Economic
growth
Ensure that
budget is spent
prudently
Engineering
and technical
water research
URV
K
emerges
Robust water
infrastructure for
industrial development
Notes:
QQ The Values were determined by Water Supply
Management (WSM) within the context of
the white nationalist apartheid state seeking
economic growth to enforce its ideology on
the people and to entrench its power.
QQ The Rules were determined by the 1956
Water Act within the context of increasing
Figure 1 The decision context of the 1980s
only standard practice to calculate the URV
for all water-related projects, but also manda-
tory within the context of any fiscal and
budgetary process to motivate for financial
or other resources. This natural progression
with respect to the use and popularity of the
URV coincided with a radical change within
the political-social-economic and ecological
context within which it has operated. We
analyse this change at the hand of the afore-
mentioned VRK-framework.
Cross-temporal analysis of the context
of water resource management
Goddard et al (2016) proposed a values,
rules and knowledge (VRK) framework with
which to understand how a societal system
of decision processes affects the manner in
which a particular problem is addressed. We
apply this framework to the South African
decision-making context for water resource
management, comparing a model of its state
in the 1980s when the URV was developed
(Figure 1) to a model of its current state
(Figure 2).
In the 1980s water resource management
was governed by Water Act No 54 of 1956.
This Act required government to con-
solidate existing water legislation, and thus
restructured the Department of Irrigation
to form the Department of Water Affairs
(DWA). The DWA had to ensure that there
Volume 62 Number 2 June 2020
DWS run by engineers
Apartheid state relied
on the white industrial
hydraulic mission to
placate their constituency
and entrench their power
Debt
crisis
Sanctions
R
Apartheid
Supply rapidly
Water Act growing urban
1956 areas
Take steps against
the increasing threat
of water pollution
fiscal illiquidity due to sanctions and the
need to take decisions in the most cost-
effective manner.
The Knowledge component was strongly
QQ
influenced by the search for engineering
and technical accuracy and precision.
was sufficient water supply to support South
Africa’s growing social, economic and indus-
trial development initiatives (Tempelhoff
2017). In addition to this drive towards the
heavy capitalisation of the country, it took
place within the context of a very unstable
internal and external political economy.
Thus, there was a need to manage water
resources more effectively and to secure
internal water security through water supply
management (WSM) (Lowenberg 1997). This
led to a very dominant supply-side policy
focus and water research. This produced
engineering and technically based solutions
(Siebrits et al 2014). With the DWA chiefly
the technocratic territory of engineers, plans
for the expansion of water supply to cities and
areas of industrial development were focused
on increasingly complex and costly transfer
schemes (Turton & Meissner 2002). The
URV was developed to assist in selecting the
most cost-effective option first. This context
is vastly different from that prevailing since
the 1994 democratic elections and the rapid
development of a natural resource manage-
ment agenda – elements that have shaped the
world of the current decade.
The prevailing context in the current
decade is illustrated in Figure 2. Supply-side
interventions were becoming costlier to
implement just as the government, having
been limited in their access to credit in
Journal of the South African Institution of Civil Engineering
 It is in this latter context that multi-
Decision of the
context 2010s Constitution
requires
water equity
Working
for Water
ple URVs emerged to assist and inform
decision-makers.
Protest culture disrupts
URV revered as an important water service delivery
science communication tool

Water research shifts

V IWRM
Redistribution of
hydro-political
ADVANCES WITH RESPECT
TO CALCULATING THE URV
privilege
towards topics of URV
sustainability, community, Protest culture National Social hydraulic
governance, integrative Redistribution Water Act mission Introduction
Constitution
catchment management, Equity Bill of Rights The URV is, essentially, a knowledge-­
socio-ecological systems

More disciplines
included in
mainstream
water research
Multiple K DWS dysfunctional
URVs emerge
Extend water and sanitation services to previously disadvantaged
Notes:
QQ The Values are determined by Integrated
Water Resources Management (IWRM)
principles that seek to protect culture,
promote redistribution with equity and
underlying value.
QQ The Rules are determined by the 1996
National Water Act, the Bill of Rights and
the Constitution, as well as the Employment
Figure 2 The decision context of the 2010s
response to their policies, was growing more
reliant on the expansion of the water supply
to support the needed growth of economic
activity. In such an environment, the applica-
tion of cost-effectiveness analysis to eliminate
the costlier supply interventions early in the
planning process was developed.
In post-apartheid South Africa, the policy
of government is not one of expanding the
water supply to support economic develop-
ment, but rather of extending water services
to the areas that were beyond the jurisdiction
of the DWA under apartheid (Schreiner 2013).
Integrative Water Resource Management
(IWRM) had replaced WSM as the dominant
paradigm, but the successful implementation
of its policies and interventions has been
questioned, and even failed (Ashton 2000;
Denby et al 2016). The most widely held belief
for its failure is that the persistent levels of
poverty and high-income inequalities, which
were carried over from the 1980s, limit the
successful implementation of IWRM (Movik
et al 2016). Water research is increasingly
dominated by ecologists and economists as
focus has shifted away from engineering and
technical solutions towards concepts such
as sustainability, governance, adaptation and
management (Siebrits et al 2014). The URV
is thus adapted and used as a science com-
munication tool (Blignaut et al 2010; Mander
et al 2017). A decision to fund the Working
Information age
Comprehensive
URV (Van Niekerk)
Employment
R DWS run by
bureaucrats;
technical skills
Engineering Equity Act
Economics shortage
Environmental movement
Governance Catchment The reserve
management
Ecology
Equity Act within a context where social
norms are challenged in public by protest
action and the need to redistribute the
hydro-political privileges.
QQ The Knowledge component has broadened
to include catchment management,
ecology, restoration, and the protection of
the environmental reserve.
for Water (WfW) programme, which was
made in 1995, had a positive impact on both
the livelihoods of local communities and the
availability of water, and has additionally pro-
moted advances in biological control, raised
conservation awareness, enacted legislation
and promoted research (Binns et al 2001; Van
Wilgen & Wannenburgh 2016). The expan-
sion of water delivery services is frustrated
by a growing protest culture (Tapela 2012)
and the Department of Water and Sanitation
(DWS) is evaluated as being in a state of
dysfunction, lacking technical personnel,
institutional capacity and funding to perform
maintenance on water infrastructure (Ruiters
& Matji 2015).
Comparing two epochs
The difference in social-political-ecological
and institutional context between the
1980s and the current decade can hardly be
greater. The 1980s were characterised by
technocratic, success-seeking precision with a
well-functioning and engineer-driven govern-
ment department focused on water supply
management to uphold a specific ideology of
exclusivity and privilege. The current decade,
however, is characterised by an inclusive,
pluralistic and eco-anthropological ideology
focused on integrated water resource man-
agement within the context of a hollow, and
often failed, state (Morokong et al 2017).
generating tool. However, due to the
contextual changes as described above, it
became widely distributed for use across a
broad number of organisational, sectoral
and disciplinary boundaries. This made new
applications possible that were not considered
at first, or even deemed possible. Originally,
in the 1980s, calculating a URV required one
to take into consideration discounted lifecycle
costs, projected water demand, the yield of
the existing system including water transfers,
and the incremental yield provided by the
project being evaluated. The boundary of the
problem is extended by the comprehensive
URV to include an analysis of how each
system’s water demand and supply vectors
vary over time and with respect to each other
in the case of evaluating transfer schemes
between two systems. In recent years, due
to the multifarious use of the URV, the
definition boundary has changed to exclude
some core components of the original URV
calculation, while expanding to include new
components. A process, referred to above as
indicator adaptation, occurred.
The URV’s many faces
All the applications of the URV use the same
definition with respect to costs, namely the
numerator (i.e. the total capital or establish-
ment costs plus the ongoing management and
operation cost over time, discounted over the
anticipated lifespan of the project). There are,
however, at least three different interpreta-
tions with respect to the calculation of the
denominator. We elaborate on these three
below, labelling them URV1, URV2 and URV3
(as illustrated in Figure 3).
1. URV1 (the original version, see Figure 3:
Panel A): The denominator is the addi-
tional water yield of a water development
scheme bounded by a water requirement
curve (Van Wilgen et al 1997; Larson et
al 2001; Hosking & Du Preez 2002; Van
Niekerk & Du Plessis 2013a, 2013b; DWS
2014; Blersch & Du Plessis 2017).
2. URV2 (URV1, but adapted for water
demand management, see Figure 3:
Panel B): The denominator is the addi-
tional structural water yield surplus, or

Journal of the South African Institution of Civil Engineering Volume 62 Number 2 June 2020 41
 Notes:
QQ Panel A: URV1 describes the case where the
Increase in yield at time t10 through new dam
Water requirement1 additional yield provided by a water project
is calculated as the hatched area, bounded
by the water requirement curve, which is
PV lifecycle costs effectively a reduction of future shortage.
Yield/water requirements (m3)

URV1 =
PV additional yield (area EBCD) QQ Panel B: URV2, which applies to water

demand interventions; the quantity in the

Shortage denominator (additional yield provided by
A E D the intervention) is taken as an increase in
Yield2
water surplus generated in the system.
QQ Panel C: URV3, the area taken as the additional

yield is not bounded by the water requirement

Yield1
B C curve and includes both an increase in yield
Surplus surplus and a decrease in shortage.

simply the “water saved” in response to

demand-side management interventions
(Joubert et al 2003; De Lange 2006;
tx tn Hoffman & Du Plessis 2008; Mckenzie &
Time
A Wegelin 2009; DWA 2011).
3. URV3 (URV1, but adapted to accom-
modate environmental management
such as rehabilitation and the clearing of
WDM shifts out water requirement curve invasive alien plants, see Figure 3: Panel
Water requirement1
C): The denominator is the ability to
Yield/water requirements (m3)

PV lifecycle costs of WDM

URV2 = Water requirement2 increase either water yield or base flow
PV additional yield (area FGH)
in a system, irrespective of direct human
consumption of that additional water for
water services (Marais & Wannenburgh
Shortage 2008; Blignaut et al 2010; Preston 2015;
F G
Yield1 Vundla et al 2016; Mander et al 2017;
Surplus Morokong et al 2017; Nkambule et al
2017). It is used particularly where, due
H to uncertainty regarding the extent to
which increased base flows will enhance
the capacity of water infrastructure to
t1 tx1 tx2
meet societal demands, or indeed in the
Time
B absence of such infrastructure for a river
system all together, it has become the
practice to forgo the consideration of a
Increase in yield at time t1 through clearing activities water requirement.
Yield curve shifts up
In literature, all three these methods, irre-
Water requirement1
spective of the way in which the denominator
PV lifecycle costs is being defined and calculated, are called
URV3 =
Yield/water requirements (m3)

PV additional yield (area IJK)

URV. This has the potential to lead to
misguided outcomes and erroneous, and
Shortage
J even inefficient, water resource management
Yield2
advice. The impacts of these three different
I
ways of calculating the URV will be discussed
at the hand of Figure 3, within the context of
Surplus
hypothetical examples.

Example URV1:
QQ Context: In Figure 3: Panel A, water
Yield1
K demand increases over time and the water
requirement of the system exceeds the
t1 tx1 tx2 tn water yield it can provide at time tx. A
Time water project that becomes operational at
C
time tx, and provides additional water yield
Figure 3 URV subtypes (shifts out yield curve 1) of size A-B for

42 Volume 62 Number 2 June 2020 Journal of the South African Institution of Civil Engineering
 the rest of the economic life of the system
until time tn.
QQ Denominator: The URV used to evaluate
the water project is concerned with the
cost-effectiveness of the water project
in enabling the system to meet its water
requirements, and it discounts the hatched
area (the area demarcated by EBCD) that
is bounded by the system’s projected water
requirement curve.
Example URV2:
QQ Context: In Figure 3: Panel B, the water
requirement of the system exceeds the
water yield it can provide at time tx1. Some
water demand management measure-
ments are implemented at time t1, lower-
ing the projected water requirement curve
(shifting it down) and ensuring the system
can meet its increasing demand for water
until time tx2. The system gains water
surpluses (unused capacity) and delays the
onset of future shortages.
QQ Denominator: The URV used to evaluate
the water project discounts the hatched
area (the area demarcated by FGH).
Example URV3:
QQ Context: In Figure 3: Panel C, the water
requirement of the system exceeds the
water yield it can provide at time tx1, due
to the spread of invasive alien plants (IAPs)
which reduce runoff. In anticipation of
future shortfall in the system’s capacity to
provide sufficient water, clearing of IAPs
commences at time t1 and continues for
the economic life of the system until tn.
Like other supply-side interventions, the
yield curve is shifted upwards improving
the system’s ability to meet its increasing
demand for water (by tx2–tx1 years).
QQ Denominator: In this case, the URV used
to evaluate the water project discounts the
hatched area (the area demarcated by KIJ).
Here the total additional yield provided
by the intervention is taken as measure
of effectiveness, irrespective of whether
or not this additional yield will be used
in the system. This is in contrast to the
example given in Panel A in which only
the additional water bounded by the water
requirement curve is used to evaluate the
cost-effectiveness of the intervention.
DISCUSSION
The process of calculating a URV is a
platform which supports the development
of synergies to create new knowledge on a
project by project basis (Siedlok & Hibbert
2014). Above the line sits Economics/Finance,
below the line sits Hydrology, Engineering
and Ecology – and the URV analysis requires
the inclusion of components from each.
The emergence of a multifarious use of
the URV, as seen in the previous section,
can be explained by applying Argote and
Miron-Spektor’s framework (Argote & Miron-
Spektor 2011) for organisational learning.
In each instance the URV tool was shaped
through use to accomplish the objectives of
the new user for the purpose of the adopting
organisation. For example, the URV was
adapted for use in evaluating IAPs and moti-
vating for their removal. The diffusion of the
URV tool across multiple organisations pre-
disposed users to have and acquire divergent
evidence and practise base knowledge endow-
ment. This made the emergence of several
distinct URV deviations a sure thing, as the
concepts of cost-effectiveness or accompany-
ing graphing techniques were lost during the
translation of the tool between organisations,
or were discarded over time.
By viewing the URV through the VRK and
organisational learning frameworks, we con-
clude that, since its emergence in the 1980s,
the use of the URV has succeeded to influ-
ence value systems and operating protocols
at a societal level. Politicians and researchers
are aware of, and wish to entrench, its
usefulness as a science-management com-
munication tool, hence the repeated use of
the URV. The URV is also a point in analysis
around which synergies occur, encouraging
those who engage with it to be exposed to a
more holistic view of a particular subsystem.
Owing to changes in the decision context of
South African water management, an influx
of researchers from a broadening set of aca-
demic disciplines expanded the experiential
and practical knowledge stock embedded
within the original definition and use of the
URV. Unfortunately, it is also potentially
misleading, as it is, essentially, not the same
indicator given the different definitions used
for the denominator.
Given the externality benefits produced by
the URV, any aim to address the multifarious
use of it must not inhibit any freedom in the
ease with which it can facilitate communica-
tion between scientists and managers, or
allow for innovation that comes from mov-
ing the problem boundary. Context is thus
important and valuable, and additional rules
imposed on its calculation must serve the plu-
ralistic operational context. The co-habitation
of the three URV formulae used, however,
provides sufficient space for such a pluralistic
environment.
CONCLUSION
The Unit Reference Value (URV) has been
used in South Africa for more than 30
years, initially by water engineers, and later
hydrologists, economists and ecologists alike
within a drastically changed social-political-
ecological and institutional context. As a cost-
effectiveness measure it is highly informative
and intuitively easy to understand – both
aspects that generate further support for its
use. Over time however, the multifarious
use thereof has led to various ways in which
researchers calculated its denominator, to
the extent that, effectively, the results across
the different types of URV measures cannot
be compared without some discussion and
distinction. This does not imply that the new
developments have been wrong. Rather, they
have added to the pluralistic context within
which research and decision-makers operate,
and is a good reflection of the current and
emerging debates.
To mitigate against possible confusion and
to avoid the misinterpretation of the results,
we recommend the following:
QQ The introduction of naming conventions
for different URV subtypes, namely URV1,
URV2 and URV3, as per the above catego-
ries thereof. This would force URV users
to make type-for-type comparisons and
reach conclusions that can be supported
from the theory, and which would make
type-for-type comparisons possible. This
will interact with the values and knowl-
edge components to self-regulate further
expansion.
QQ The redefining of the appropriate measure
for all the URVs as “the unit cost to reduce
future water shortage, or to enhance water
security, using measure 1, 2, or 3” – that
will be applicable in most interventions.
QQ That URV users provide, not only a
formula or description of the URV calcula-
tion according to the type they are report-
ing on, but also a diagram illustrating
what the effectiveness measure is that they
are calculating. This would imply that
users will engage with many of the prin-
ciples underlying the URV and potentially
be a point of learning between individuals
that could reduce knowledge inequalities.
Within a world of increasing complexity,
scarcity and dynamic change, the develop-
ment of simple and effective indicators to
assist in decision-making is of high value.
The URV is one such an indicator. Here we
propose three changes in the conventional
use of the URV metric that should facili-
tate dialogue within the water resource
management space, while improving the

Journal of the South African Institution of Civil Engineering Volume 62 Number 2 June 2020 43
nuances with respect to its use within vari-
ous different contexts.
REFERENCES
Alferink, L A & Farmer-Dougan, V 2010. Brain-
(not) based education: Dangers of misunder­
standing and misapplication of neuroscience
research. Exceptionality, 18(1): 42–52.
doi: 10.1080/09362830903462573.
Angelstam, P, Andersson, K, Annerstedt, M et al 2013.
Solving problems in social–ecological systems:
Definition, practice and barriers of transdisciplinary
research. Ambio: A Journal of the Human Environment,
42(2): 254–265. doi: 10.1007/s13280-012-0372-4.
Argote, L & Miron-Spektor, E 2011. Organizational
learning: From experience to knowledge. Organization
Science, 22(5): 1123–1137. doi: 10.1287/orsc.1100.0621.
Ashton, P 2000. Integrated catchment management:
Balancing resource utilization and conservation.
Occasional Paper No. 5, African Water Issues Research
Unit (AWIRU), University of Pretoria.
Binns, J A, Illgner, P M & Nel, E L 2001. Water shortage,
deforestation and development: South Africa’s Working
for Water programme. Land Degradation and
Development, 12(4): 341–355. doi: 10.1002/ldr.455.
Blersch, C L & Du Plessis, J A 2017. Planning for
desalination in the context of the Western Cape
water supply system. Journal of the South African
Institution of Civil Engineering, 59(1): 11–21.
doi: 10.17159/2309-8775/2017/v59n1a2.
Blignaut, J N, Mander, M, Schulze, R, et al 2010.
Restoring and managing natural capital towards
fostering economic development: Evidence from the
Drakensberg, South Africa. Ecological Economics,
69(6): 1313–1323.
Brisson, P & Brisson, M 2010. Variable application
and misapplication of cricoid pressure. Journal
of Trauma & Acute Care, 69(5): 1182–1184.
doi: 10.1097/‌TA.0b013e3181d2793e.
Costanza, R, Farber, S C & Maxwell, J 1989.
Valuation and management of wetland eco­
systems. Ecological Economics, 1(4): 335–361.
doi: 10.1016/0921-8009(89)90014-1.
De Lange, W 2006. Multi-criteria decision-making
for water resource management in the Berg Water
Management Area. PhD Thesis. Stellenbosch
University.
Denby, K, Movik, S, Mehta, L & Van Koppen, B 2016.
The ‘Trickle Down’ of IWRM: A case study of local-
level realities in the Inkomati Water Management
Area, South Africa. Water Alternatives, 9(3): 473–492.
DWA (Department of Water Affairs) 2011. Development
of a reconciliation strategy for the Olifants River Water
Supply System. Report No. WP10197. Pretoria: DWA.
DWS (Department of Water and Sanitation) 2014.
Feasibility study for the Mzimvubu Water Project:
Cost estimates and economic analysis. Report No.
P WMA 12/T30/00/5212/15. Pretoria: DWS.
44
Drury, C & Tayles, M 1997. The misapplication
of capital investment appraisal techniques.
Management Decision, 35(2): 86–93.
doi: 10.1108/00251749710160223.
Glantz, S A 1980. Biostatistics: How to detect, correct and
prevent errors in the medical literature. Circulation,
61(1): 1–7. doi: 10.1161/01.CIR.61.1.1.
Goddard, R, Colloff, M J, Wise, R M, Ware, D & Dunlop,
M 2016. Values, rules and knowledge: Adaptation as
change in the decision context. Environmental Science
Policy, 57: 60–69. doi: 10.1016/j.envsci.2015.12.004.
Hoffman, J J & Du Plessis, J A 2008. Water demand
management: An economic viable option. Stellenbosch
University. Available at: http://hdl.handle.
net/10019.1/43899.
Hosking, S & Du Preez, M 2002. Valuing water gains in
the Eastern Cape’s Working for Water Programme.
Water SA, 28(1): 23–28. doi: 10.4314/wsa.v28i1.4863.
Joubert, A, Stewart, T J & Eberhard, R 2003. Evaluation
of water supply augmentation and water demand
management options for the City of Cape Town.
Journal of Multi-Criteria Decision Analysis, 12(1):
17–25. doi: 10.1002/mcda.342.
Larson, E J, Marais, C & Görgens, A H M 2001. Water
resources planning with recognition of alien vegetation
eradication. Proceedings, 10th South African
Hydrological Symposium, Pietermaritzburg, pp 26–28.
Lowenberg, A D 1997. Why South Africa’s apartheid
economy failed. Contemporary Economic Policy, 15(3):
62–72. doi: 10.1111/j.1465-7287.1997.tb00478.x.
Mander, M, Jewitt, G, Dini, J et al 2017. Modelling
potential hydrological returns from investing in
ecological infrastructure: Case studies from the
Baviaanskloof-Tsitsikamma and Umgeni catchments,
South Africa. Ecosystems Services, 27: 261–271.
doi: 10.1016/j. ecoser.2017.03.003.
Marais, C & Wannenburgh, A M 2008. Restoration
of water resources (natural capital) through the
clearing of invasive alien plants from riparian
areas in South Africa: Costs and water benefits.
South African Journal of Botany, 74(3): 526–537.
doi: 10.1016/j.‌sajb.2008.01.175.
Mckenzie, R S & Wegelin, W 2009. Challenges facing
the implementation of water demand management
initiatives in Gauteng Province. Water SA, 35(2):
168–174. doi: 10.4314/wsa.v35i2.76735.
Morokong, T, Blignaut, J N & Bester, R 2017. Determining
the efficiency of contractors in clearing invasive alien
plant species: The case of Mopani District, Limpopo
Province, South Africa. African Journal of Agricultural
Research Economics, 12(4): 276–288.
Movik, S, Mehta, L, Van Koppen, B & Denby, K 2016.
Emergence, interpretations and translations of IWRM
in South Africa. Water Alternatives, 9(3): 456–472.
Nkambule, N P, Blignaut, J N, Vundla, T, Morokong,
T & Mudavanhu, S 2017. The benefits and costs of
clearing invasive alien plants in northern Zululand,
South Africa. Ecosystems Services, 27: 203–223.
doi: 10.1016/j.‌ecoser.2017.04.011.
Volume 62 Number 2 June 2020
Preston, I R 2015. Water supply development decision-
making in South Africa. MComm Dissertation.
Grahamstown: Rhodes University.
Ruiters, C & Matji, M P 2015. Water institutions and
governance models for the funding, financing and
management of water infrastructure in South Africa.
Water SA, 41(5): 660–676. doi: 10.4314/wsa.v41i5.09.
Schreiner, B 2013. Viewpoint – Why has the South
African National Water Act been so difficult to
implement? Water Alternatives, 6(2): 239–245.
Siebrits, R, Winter, K & Jacobs, I 2014. Water
research paradigm shifts in South Africa.
South African Journal of Science, 110(5–6): 1–9.
doi: 10.1590/‌sajs.2014/20130296.
Siedlok, F & Hibbert, P 2014. The organization of
interdisciplinary research: Modes, drivers and barriers.
International Journal of Management Reviews,
16(2): 194–210. doi: 10.1111/ijmr.12016.
Tapela, B N 2012. Social water scarcity and water use:
Report to the Water Research Commission. Pretoria:
Water Research Commission.
Tempelhoff, J 2017. The Water Act, No. 54 of 1956
and the first phase of apartheid in South Africa
(1948–1960). Water History, 9(2): 189–213.
doi: 10.1007/‌s12685-016-0181-y.
Turton, A & Meissner, R 2002. The hydrosocial
contract and its manifestation in society: A South
African case study. In Turton, A & Henwood, R
(Eds). Hydropolitics in the Developing World: A
Southern African Perspective. Pretoria: African
Water Issues Research Unit (AWIRU), University of
Pretoria, pp 37–60.
Van Niekerk, P & Du Plessis, J 2013a. Hydrologic-
economic appraisal of life-cycle costs of inter-basin
water transfer projects. Water SA, 39(4): 539–548.
doi: 10.4314/wsa.v39i4.13.
Van Niekerk, P & Du Plessis, J 2013b. Unit Reference
Value: Application in appraising inter-basin
water transfer projects. Water SA, 39(4): 549–54.
doi: 10.4314/‌wsa.v39i4.14.
Van Wilgen, B W & Wannenburgh, A 2016.
Co-facilitating invasive species control, water
conservation and poverty relief: Achievements and
challenges in South Africa’s Working for Water
programme. Current Opinion in Environmental
Sustainability, 19: 7–17.
Van Wilgen, B W, Little, P, Chapman, R, Görgens,
A, Willems, T & Marais, C 1997. The sustainable
development of water resources: History,
financial costs, and benefits of alien plant control
programmes. South African Journal of Science,
93(9): 404–411.
Vundla, T, Blignaut, J N, Nkambule, N, Morokong,
T & Mudavanhu S 2016. The opportunity cost of
not utilising the woody invasive alien plant species
in the Kouga, Krom and Baviaans catchments in
South Africa. South African Journal of Economic
and Management Sciences, 19(5): 814–830.
doi: 10.17159/2222-343/2016/v19n5a8.
Journal of the South African Institution of Civil Engineering
Guidelines for the preparation of papers and technical notes
Authors should comply with the following guidelines when preparing papers for publication in the journal
The Journal of the South African Institution of Civil
Engineering is published quarterly in March, June,
September and December. Articles submitted for
publication are reviewed by a panel of referees under
the guidance of the SAICE Journal Editorial Panel.
The journal publishes research papers covering all the
disciplines of civil engineering (structural, geotechnical,
railway, coastal/marine, water, construction, environ-
mental, municipal, transportation) and associated topics
that are relevant to the civil engineering profession, and
that preferably have particular relevance to civil engi-
neering in southern Africa and the African continent.
When preparing articles for publication, authors
should please take note of the following and comply
with the guidelines as set out:
CLASSIFICATION OF ARTICLES
CONSIDERED FOR PUBLICATION
QQ Technical papers are well-researched, in-depth,
fully referenced technical articles not exceeding
6 000 words in length (excluding tables, illustra-
tions and the list of references). Related papers that
deal with ‘softer sciences’ (e.g. education, social
upliftment, etc) are accepted if they are of a techni-
cal nature and of particular interest to the civil
engineering profession. The latter type of paper will
be subject not only to peer-review by civil engineers,
but also to review by non-engineering specialists in
the field covered by the paper.
QQ Technical notes are short, fully referenced technical
articles that do not exceed 2 000 words. A typical
technical note will have limited scope often dealing
with a single technical issue of particular importance
to civil engineering.
QQ Review papers are considered for publication as
either technical papers or technical notes on condi-
tion that they are the original work of the author
and will assist the reader with the understanding,
interpreting or applying of the subject under review.
A review paper must contain criteria by which the
work under review was evaluated, and contribute
by synthesising the information and drawing new
conclusions from the dissemination of the previously
published work.
QQ Discussion on published articles is welcomed up to
six months after publication. The length of discus-
sion contributions is limited to 1 500 words. Where
appropriate, discussion contributions will be subject
to the normal reviewing process and will be for-
warded to the authors of the original article for reply.
POLICY REGARDING LANGUAGE AND
ORIGINALITY OF SUBMITTED ARTICLES
QQ Language: Manuscripts should preferably be
presented in English, as the journal is distributed
internationally. Articles submitted in any of the
other official South African languages should be
accompanied by an expanded abstract in English.
QQ Original work: Papers and technical notes must
be original contributions. Authors must affirm
that submitted material has not been published
previously, is not under consideration for publica-
tion elsewhere and will not be submitted elsewhere
while under consideration by the SAICE Journal
Editorial Panel. It is the responsibility of the authors
to ensure that publication of any paper in the journal
will not constitute a breach of any agreement or
the transgression of any law. The corresponding
author should confirm that all co-authors have
read and approved the manuscript and accept these
conditions. Authors are responsible for obtaining
permission to publish experimental data and other
information that may be confidential or sensitive.
Authors are also responsible for obtaining permis-
sion from copyright owners when reproducing mate-
rial that had been published elsewhere. Proof of such
permission must be supplied.
SUBMISSION PROCEDURES
AND REQUIRED FORMAT
QQ Online submission: Manuscripts must be uploaded
as PDF files (http://journal.saice.org.za). Individual
file sizes may not exceed 10 MB. Should you experi-
ence problems uploading your paper, please contact
the editor (verelene@saice.org.za).
QQ Format: Manuscripts should be prepared in MS
Word and presented in double line spacing, single
column layout with 25 mm wide margins. Line
numbers must be applied to the whole document.
All pages should bear the authors’ names and be
numbered at the bottom of the page. With the excep-
tion of tables and figures (see below) the document
should be typed in Times New Roman 12 pt font.
Contributions should be accompanied by an abstract
of not more than 200 words.
QQ First page: The first page of the manuscript should
include the title of the paper, the number of words of
the main text (i.e. excluding figures, tables and the
list of references), the initials and surnames of the
authors, professional status (if applicable), SAICE
affiliation (Member, Fellow, Visitor, etc), telephone
numbers (landline and mobile), and e-mail and
postal addresses. The name of the corresponding
author should be underlined. Five keywords should
be suggested.
QQ Figures, tables, photos and illustrations: These
should preferably be submitted in colour, as the
journal is a full-colour publication.
QQ Their positions should be clearly marked in the
text as follows: [Insert Figure 1].
QQ Figures, tables, photos, illustrations and equa-
tions should be numbered consecutively and
should appear in the text directly after they have
been referred to for the first time.
QQ Illustrations must be accompanied by appropriate
captions. Captions for tables should appear above
the table. All other captions should appear below
the illustration (figures, graphs, photos).
QQ Only those figures and photographs essential to
the understanding of the text should be included.
All illustrations should be referred to in the text.
QQ Figures should be produced using com-
puter graphics. Hand-drafted figures will not be
­accepted. Lettering on figures should be equiva-
lent to a Times New Roman 9 pt font or slightly
larger (up to 12 pt) if desired. Lettering smaller
than 9 pt is not acceptable.
QQ Tables should be typed in Times New Roman
9 pt font. They should not duplicate information
already given in the text, nor contain material
that would be better presented graphically.
Tabular matter should be as simple as possible,
with brief column headings and a minimum
number of columns.
QQ Mathematical expressions and presentation
of symbols:
QQ Equations should be presented in a clear form
which can easily be read by non-mathematicians.
Each equation should appear on a separate line
and should be numbered consecutively.
QQ Symbols should preferably reflect those used in
Microsoft Word Equation Editor or Mathtype,
or should be typed using the Times New Roman
symbol set.
QQ Variables in equations (x, y, z, etc, as well as
lower case Greek letters) should be presented
in italics. Numbers (digits), upper case Greek
letters, symbols of metric measurement units
(m for metres, s for seconds, etc) and mathemati-
cal/trigonometrical functions (such as sin, cos
and tan) are not written in italics, but in upright
type (Roman). Variables and symbols used in the
body of the text should match the format used
in the equations, i.e. upright or italics, whichever
is applicable.
QQ Metric measurement abbreviations/units should
conform to international usage – the SI system of
units should be used.
QQ Decimal commas may be used, but decimal
points are preferred.
QQ Symbols should preferably be defined in the text,
but if this is not feasible, a list of notations may be
provided for inclusion at the end of the paper.
QQ Headings: Sections and paragraphs should not be
numbered. The following hierarchy of headings
should be followed:
HEADING OF MAIN SECTION
Heading of subsection
Heading of sub-subsection
QQ References: References should follow the Harvard
system. The format of text citations should be as
follows: “Jones (1999) discovered that …” or “recent
results (Brown & Carter 1985; Green et al 1999)
indicated that …”
 References cited in the text should be listed in
alphabetical order at the end of the paper. References
by the same author should be in chronological order.
The following are examples of a journal article, a
book and a conference paper:
QQ Donald, A W 1954. Pore pressure coefficients.
Géotechnique, 4(4):143–147.
QQ Terzaghi, K 1943. Practical soil mechanics. New
York: Wiley.
QQ Smith, A W, Black, C & White, A 1999. Factors
affecting the strength of soils. Proceedings, ASCE
Conference on Shear Strength of Soils, Colorado,
pp 503–532.
Papers published previously in the Journal of the
South African Institution of Civil Engineering should
be cited if applicable.
QQ Footnotes, trade names, acronyms, abbreviations:
These should be avoided. If acronyms are used, they
should be defined when they first appear in the text.
Do not use full stops after abbreviations or acronyms.
QQ Return of amended papers: Papers requiring
amendments will be accepted up to six months after
the referee reports had been sent to authors, after
which the paper will be withdrawn from the system.
FINAL ARTICLE
QQ Copyright: On acceptance of the paper or technical
note, copyright must be transferred by the author/s
to the South African Institution of Civil Engineering
on the form that will be provided by the Institution.
QQ Photos of authors: The final corrected version of
the paper should be accompanied by recent, high-
resolution head and shoulders colour photographs
and a profile not exceeding 100 words for each of
the authors.
QQ Proofs: First proofs of papers will be sent to authors
in PDF format for verification before publication.
No major re-writes will be allowed, only essential
minor corrections.