of the South African Institution of Civil Engineering Volume 59 Number 1 March 2017

of the South African Institution of Civil Engineering

Volume 59 No 1 March 2017 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 (0)11 805 5947/48, Fax +27 (0)11 805 5971
http://www.saice.org.za
civilinfo@saice.org.za

Contents
Editor-in-chief
Prof Gerhard Heymann
University of Pretoria
Tel +27 (0)12 420 3627
gerhard.heymann@up.ac.za
joint Editor-in-chief
Prof Chris Clayton
University of Southampton
2 Vorticity filaments beneath regular turbulent flow
c.clayton@soton.ac.uk
R Mukaro DOI: 10.17159/2309-8775/2017/v59n1a1
MANAGING Editor
Verelene de Koker
Tel +27 (0)11 805 5947, Cell +27 (0)83 378 3996 11 Planning for desalination in the context of
verelene@saice.org.za
the Western Cape water supply system
journal editorial Panel
Prof G Heymann – University of Pretoria C L Blersch, J A du Plessis DOI: 10.17159/2309-8775/2017/v59n1a2
Prof CRI Clayton – University of Southampton
Prof Y Ballim – University of the Witwatersrand
Prof W Burdzik – University of Pretoria 22 Numerical modelling of flexible pavement
Dr P Day – Jones & Wagener (Pty) Ltd
Prof J du Plessis – University of Stellenbosch incorporating cross‑anisotropic material properties
Prof GC Fanourakis – University of Johannesburg
Prof M Gohnert – University of the Witwatersrand
Part I: Surface circular loading
Prof PJ Gräbe – University of Pretoria
Dr C Herold – Umfula Wempilo Consulting J W Maina, F Kawana, K Matsui DOI: 10.17159/2309-8775/2017/v59n1a3
Prof A Ilemobade – University of the Witwatersrand
Prof SW Jacobsz – University of Pretoria
Prof EP Kearsley – University of Pretoria
28 Numerical modelling of flexible pavement
Prof C Roth – University of Pretoria incorporating cross‑anisotropic material properties
Prof W Steyn – University of Pretoria
Mr M van Dijk – University of Pretoria Part II: Surface rectangular loading
Prof JE van Zyl – University of Cape Town
Prof C Venter – University of Pretoria J W Maina, F Kawana, K Matsui DOI: 10.17159/2309-8775/2017/v59n1a4
Prof A Visser – University of Pretoria
Dr E Vorster – Aurecon South Africa (Pty) Ltd
Prof J Wium – University of Stellenbosch 35 Potential solution to pollution of groundwater
Prof A Zingoni – University of Cape Town
Prof M Zuidgeest – University of Cape Town by diffusion of volatile organic compounds
Peer reviewing through the primary HDPE geomembrane
The Journal of the South African Institution of
Civil Engineering is a peer-reviewed journal in composite lining systems of landfills
that is distributed internationally
R A Pienaar, W Fourie DOI: 10.17159/2309-8775/2017/v59n1a5
Design and reproduction
Marketing Support Services, Ashlea Gardens, Pretoria
Printing
45 Updated provisions of SANS 10160‑4
Fishwicks, Pretoria for steel structures
Papers for consideration should be e-mailed to
the Managing Editor at: verelene@saice.org.za C P Roth, A Gebremeskel DOI: 10.17159/2309-8775/2017/v59n1a6

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

Journal of the South African Institution of Civil Engineering • Volume 59 Number 1 March 2017 1
TECHNICAL PAPER
Vorticity filaments beneath
regular turbulent flow
Journal of the South African
Institution of Civil Engineering
ISSN 1021-2019
Vol 59 No 1, March 2017, Pages 2–10, Paper 1445
R Mukaro

DR RAPHAEL MUKARO is a senior lecturer in the

Department of Physics and Electronics at
Two-dimensional vorticity flow fields created in the wake of a plunging breaker were
North-West University, Mafikeng Campus,
where he has worked since 2015. From 2008 to investigated for regular turbulent flow at a Reynolds number of 30 000. Velocity flow fields
2014 he lectured Applied Physics at the obtained from an earlier study that had employed digital particle image velocimetry, were
University of KwaZulu-Natal (UKZN), where he analysed to determine vorticity shedding patterns and the interactions between the vorticity
taught third-year modules (mainly Electronics, filaments as flow progressed. Central difference approximations were applied to the velocity
Signal Processing Techniques and Measurement
fields to determine vorticity at each point in the field. Most of the strong instantaneous vorticity
Systems). He was also a senior lecturer in physics at Bindura University of
Science from 1999 to 2007. He received his PhD degree from UKZN in 2014. observed in the flow field was in the form of filaments. A hierarchy of filaments of different
His current research interests are in the digital, statistical and wavelet study of lengths were observed, with the longest being as long as the height of the wave used. During
turbulence generated by laboratory-plunging water waves. the early phases of the flow, instantaneous vorticity tended to organise into thin filaments of
Contact details: counter-rotating pairs. Eventually, the co-rotating vorticity filaments coalesced and ultimately
North West University merged in the turbulent flow as flow progressed, while counter-rotating vorticity filaments were
Mafikeng Campus cancelled by viscous dissipation. The results suggested that filaments travel more slowly than
School of Mathematical and Physical Sciences the wave velocity and drifted towards the bed as they became elongated, and the number of
Department of Physics and Electronics
filaments remaining in the flow were observed to decrease as flow progressed. Whereas phase-
Private Bag X2046
Mmabatho resolved instantaneous vorticity results showed pairs of counter-rotating vorticity filaments near
2735 the crest, the phase-averaged vorticity description of flow fields showed a dominant primary
South Africa positive vorticity filament around the shear boundary layer.
T: +27 18 389 2151
E: raphael.mukaro@nwu.ac.za

INTRODUCTION turbulence and extreme waves and their

Wave breaking is one of the most important
problems for coastal engineers. The breaking
process leads to either surging, spilling or
plunging breaker types. A plunging breaker
is the most violent breaker, initiated when
a steepening wave crest curls over to form
an overturning jet. Steepening of the wave
naturally induces high curvature and conse-
quently strong vorticity. The jet plunges into
the water ahead to form a plunger vortex,
and also produces a splash-up leading to
strong turbulence in the flow. The generated
turbulence is a collection of weakly cor-
related vortical motions, which, despite their
intermittent and chaotic distribution over a
wide range of space and time scales, actually
consist of local characteristic eddy patterns
that persist as they move around under the
influences of their own and other eddies’
vorticity fields (Hunt & Vassilicos 2000).
Energy in the wave is transferred from the
average flow to vortex structures at different
scales. Due to the non-linear interactions
between different scales, cascade processes
of energy are very complex (Lin et al 2002).
Wave-breaking turbulence accounts for
most of the sediment suspension through
the introduction of eddies associated with
breaking waves, with increasing levels of tur-
bulence coming towards the boundary layer
from the water surface.
Keywords: turbulent flow, velocity flow field, vorticity filament, Ocean and coastal engineers have been
instantaneous vorticity, mean vorticity interested in wave loading generated by
interaction with marine structures. Ryu
et al (2007) investigated potential damage
to coastal structures caused by significant
impacts of breaking waves and associated
overtopping greenwater. The formation
of bed topography, which, as a result of a
complicated interaction between flow and
sediment particles along the bed, has also
received considerable attention. Lajeunesse
et al (2010) used a high-speed video-imaging
system to record the trajectories of the mov-
ing particles over a flat bed, and observed
that entrained particles exhibit intermittent
motion, composed of the succession of
periods of flight and rest. Keshavarzi and
Ball (2011) suggested that the structure of
turbulent flow over the ripples in the bottom
of an open channel is important for under-
standing of sediment particle entrainment
and its transport. Two important issues
which they suggested need to be understood
in sediment movement are the stochastic
nature of instantaneous shear stresses
over a ripple bed, and how it influences on
sediment entrainment and transport. They
employed an Acoustic Doppler Velocity
Meter (Micro-ADV) with a sampling rate
of 50 Hz to measure three-dimensional
velocities, a charge-coupled device camera
to record images of particle motion, and
image-processing techniques to provide an
accurate measurement of flow structure
and particle entrainment from the bed. In

2
Mukaro R. Vorticity filaments beneath regular turbulent flow. J. S. Afr. Inst. Civ. Eng. 2017:59(1),
Art. #1445, 9 pages. http://dx.doi.org/10.17159/2309-8775/2017/v59n1a1
a study of sediment entrainment from the
bed, Williams (1990) and Nelson et al (1995)
investigated and found a high correlation
between the stream-wise velocity component
and the sediment flux. Additionally, Nelson
et al (1995) found that the transport rate
tends to be higher when the vertical velocity
and Reynolds momentum flux are angled
towards the bed. They found the best cor-
relation between the sediment flux and the
stream-wise velocity component to occur
with a lag of 0.1 second, and consequently
they suggested that a measuring frequency
of 10 Hz would give the best results in terms
of time scales. Nielsen (1984) showed experi-
mentally that the process of sediment sus-
pension is convective rather than diffusive,
and that it is dominated by such features as
travelling vortices, jets and turbulent bursts.
In addition, the effect of wave-breaking on
the concentration profile in terms of turbu-
lence associated with each breaker type was
discussed. Experimental results near the bed
showed that sediment concentration profiles
are very similar, suggesting that the near-bed
reference concentration is independent of
the external turbulence (breaking waves)
and is controlled mainly by the bottom
boundary layer. Voulgaris and Collins (2000)
observed that an efficient numerical model
of time-averaged suspended sediment con-
centration should be related to the breaking
characteristics of the waves. A commonly
used parameter to define the onset of wave-
breaking is the ratio of wave height to water
depth. Galvin (1972) defined another simple
wave-breaking parameter that is a function
of breaking wave height (Hb), beach slope
(tan β) and wave period (T):
Hb
= Bb(1)
gT 2 tan β
For Bb > 0.068 the breaking waves are spill-
ing, for 0.003 < Bb < 0.068 the waves are
plunging, while surging breaking waves are
present when Bb < 0.003. Galvin’s parameter
has been used extensively in the study of
the effect of the breaking-wave-induced
vortices in sediment resuspension (Zhang &
Sunamura 1990; Zhang 1994). Zhang (1994)
identified four types (A–D) of oblique rolling
vortices. The conditions for the occurrence
of the observed types of vortices were
defined as a function of Galvin’s breaker
parameter (Equation (1)), and the breaking
wave Reynolds number Re = HbLb/νT where
Lb and ν are the wavelength at the breaking
position and the seawater kinematic viscos-
ity, respectively. From laboratory experi-
ments, Zhang (1994) found that oblique
vortices whirl the bed material and lift the
sediment up into suspension, acting as a
Journal of the South African Institution of Civil Engineering • Volume 59 Number 1 March 2017
– – –
Suspended load
transport +
+
Flow separation + Irrotational
+ + + + + +
– – –
+ + +
Figure 1 S chematic representation of the vorticity behaviour and corresponding flow structure in
the surf and swash zones (Sou & Yeh 2011)
tornado. They observed that types A and
B vortices last longer than types C and D.
Type A horizontal vortices act in a similar
manner as the oblique vortices, but are capa-
ble of re-suspending more sediment. Type B
horizontal vortices dig the bed material like a
cultivator, so that a larger amount of sand is
lifted up into suspension and is entrapped in
the vortex. Types C and D do not reach the
bottom and thus do not contribute to sedi-
ment re-suspension. Considering the dura-
tion that the vortex is in contact with the
bed, the strength of the vortex and thus its
capability to re-suspend sediment was found
to be highest in the B-type horizontal vortex,
decreasing in the A-type horizontal vortex
and the oblique vortex, in that order.
The creation of vorticity due to dissipat-
ing or breaking waves is a basic fluid dyna­
mics phenomenon that is central to many
geophysical and engineering applications
(Buhler 2000). As stated by Long and Özkan-
Haller (2009), vorticity motions in the near
shore contribute to mixing and dissipation.
The presence of vorticity induces shear at the
water surface which changes the properties
of the free surface waves when meeting an
underlying current. Turbulent flows display
intermittent pressure drops associated with
vorticity concentrations, so pressure mea-
surements can also be used to locate regions
with high vorticity or dissipation in turbu-
lent flows (Abry et al 1994). Bracket et al
(1983) were the first to show, via simulations,
that there is a direct relationship of the small
pressure filaments with vorticity filaments.
There have also been many experimental
attempts to visualise coherent vorticity
structures in boundary layers or in rotating
flows, but only a few in the bulk of fairly
homogeneous isotropic flows (Dernoncourt
et al 1998). Sou and Yeh (2011) employed
particle image velocimetry to investigate
the fundamental characteristics of the flow
structure in the vertical cross-shore plane
as the wave evolves from the outer surf
zone to the swash zone. They suggested
that vorticity generated near the bed has a
simple relation to the pressure gradient for
boundary layer flows. The no-slip boundary
+
+
vortex
– – + +
Flow attachment
Fast velocity = deposition
= bed load transport
condition, i.e. vanishing inertial force at the
boundary, means that the phase-averaged
momentum equation on a plane surface of a
two-dimensional flow can be expressed (Sou
& Yeh 2011) as:
д 2u дωy дP'
μ =μ = (2)
дz 2 дz дx
where μ is the dynamic viscosity and P’ is
the excess pressure at the bed surface, given
by P' = P + ρgz, where P is the atmospheric
pressure, ρ is the water density, g is accelera-
tion due to gravity and z is the water depth.
Equation (2) implies that the flux of vorticity
from the bed is created by the pressure gra-
dient. This indicates that when the pressure
gradient in the x-direction is negative, the
vorticity gradient in the z-direction ∂ωy /∂z
is also negative, suggesting that the positive
or clockwise (CW) vorticity is generated at
the bed during uprush (Sou & Yeh, 2011).
Figure 1 shows a schematic sketch by Sou
& Yeh (2011) that summarises the observed
flow structure and flow characteristics of
vorticity observed in the surf and swash
zones, which show the resulting load
deposition, transport and suspension. They
observed that in the surf zone a thin layer
of counter-clockwise (CCW) vorticity on
top of the CW vorticity is generated along
the water surface, due to strong induction
of the CW vorticity by the collision of the
uprush and downwash flows. The surface
flow pattern can support the presence of
CCW vorticity on the surface behind the
surface roller. They also pointed out that the
presence of CCW vorticity on the surface
is necessary, based on the Stokes theorem.
Sou & Yeh (2011) also observed that flow
separation occurs at the bore front, so the
detached sediments from the bed by flow
separation must remain suspended by the
wave-breaking-generated turbulence and
the shear-layer-generated turbulence. They
also noted that the vorticity intensity at
the shear layer decreases in magnitude in
the offshore locations. In addition, sus-
pended sediments should deposit on the
bed with the flow attachment process as the
3
wave-breaking-generated turbulence and
the shear-layer-generated turbulence levels
become relatively low at the flow reversal.
Dabiri and Gharib (1997) studied the
vorticity generation within a spilling water
wave using the particle image velocimetry
technique. Douady et al (1991) studied
turbulent Von Karman swirling flows using
water seeded with air bubbles which dis-
played intermittent formation of filaments of
bubbles, ascribed to vorticity concentrated
on tube-like structures. Pressure measure-
ments (Fauve et al 1993; Abry et al 1994) and
pressure-velocity correlations (Cadot et al
1995) were used in the detection of vorticity
filaments, and in the study of their dynamic
and statistical properties. Dernoncourt et
al (1998) used an ultrasound scattering
technique that allowed measurement of
the vorticity in the bulk of the flow. They
used short-time Fourier transforming of the
incoming sound frequency and filtering of
the scattered pressure signal, allowing the
detection of large vorticity concentrations on
small spatial scales, but not the study of their
detailed internal dynamics. Their essential
findings were that the vorticity at small
scales has a clear tendency to exhibit intense
rare events that are associated with localised
structures compatible with the visualisa-
tion of low-pressure structures in the same
flow. Lien and Sanford (2000) observed that
turbulence vorticity flux plays an important
dynamic role in the turbulence boundary
layer. They used similarity scaling of the
turbulence vorticity flux to estimate the bed
stress, and observed that turbulence vorticity
flux was related to the divergence of turbu-
lence momentum flux, which represents the
turbulence force on the mean flow. They
argued that understanding the spectral prop-
erties of the turbulence vorticity flux would
help improve turbulence parameterisation
schemes in numerical models.
Kevlahan and Farge (1997) stated that,
in order to successfully model a two-
dimensional turbulent flow, it is essential to
understand the role of vorticity filaments.
They used the term ‘filament’ to refer exclu-
sively to the thread-like structures observed
in two-dimensional turbulence. Siggia (1981)
was the first to show in a simulation that, in a
turbulent flow, the regions in which vorti­city
had the largest modulus were in the shape
of elongated filaments. Numerical studies
by Jimenez et al (1993) have also shown that
intense vorticity levels in turbulent flows
are organised in Burger-type filaments. The
same objects have been identified by locating
local minima of pressure fields in numerical
experiments (Belin et al 1996). Dritschel et
al (1991) noted that a common feature of
strongly non-linear, high Reynolds number,
4 Journal of the South African Institution of Civil Engineering • Volume 59 Number 1 March 2017
two-dimensional flows is the presence of
thin filaments of vorticity, while Jimenez
and Wray (1998) also observed that most of
the strong vorticity in the flow field is in the
form of filaments. According to Elhmaidi et
al (2004), vorticity filaments are easily gener-
ated during vortex–vortex interactions (such
as vortex merger), and when vortices are
exposed to strong shearing. Kimmoun and
Branger (2007) used calculations of the spa-
tial derivatives of the velocity field measured
through particle image velocimetry, in order
to obtain relevant information on vorticity.
Dritschel (1989) observed that, once large
vortices are formed in a turbulent flow, like-
signed vortices merge together in an inelastic
process. They observed that, during the
merger process, filamentation occurs, leading
to the formation of thin strips of vorticity.
While some filaments remain attached to the
large coherent vortices, others are expelled
during the merger and remain detached
from the large vortices. In a related study
on vorticity filaments in two-dimensional
turbulence Kevlahan and Farge (1997) sug-
gested that vorticity filaments form in two-
dimensional turbulence during the merger
of coherent vortices, due to the strong shear
forces that develop during the interaction.
These forces pull some of the vorticity out
into long, narrow strips of intense vorticity
which are usually spiral in shape. Petitjeans
(2003) also weighed in by pointing out that
vorticity filaments are generated by roll-up
of a fluid sheet around the vortex core, and
that it is now well known that local stretching
of vorticity in turbulent flows produces very
intense vortices called vorticity filaments.
Meunier et al (2005) studied the physics of
vortex merging. They observed that at the
beginning of the merging, two strong fila-
ments of vorticity are ejected, and roll up
around the final vortex due to the differential
rotation, leading to an axisymmetric vortex
at late stages. Just as in previous work by
Kevlahan and Farge (1997), Elhmaidi et al
(2004), Dritschel (1989), Jimenez et al (1993),
Jimenez and Wray (1998), Meunier et al
(2005) and Petitjeans (2003), narrow strips
of intense vorticity are what is meant when
referring to ‘vorticity filaments’ in this work.
However, as pointed out by (Petitjeans 2003),
the presence of vorticity does not imply the
existence of a vortex, e.g. in shear boundary
layers, where we may have vorticity but no
vortices. Where a vortex is present, the term
‘vorticity’ describes its rate of rotation about
some axis, locating and giving the vortex its
strength or intensity.
Results presented herein are related to
an earlier study in which experiments were
performed in a laboratory wave flume to
study the turbulence generated by 0.4 Hz
breaking water waves (with a wave height of
12 cm) as they propagate along a glass-walled
laboratory flume and break on a 1:20 plane
slope. Characterisation of the flow structure,
description of the experimental setup and
methods used have been previously reported
in Mukaro and Govender (2011; 2013). Here it
will suffice to just mention that the temporal
and spatial evolution of instantaneous and
phase-ensemble-averaged velocity flow fields
induced were examined using flow visualisa-
tion and digital correlation image velocimetry.
A computer-driven electronic measurement
system was designed, developed and employed
to capture images of breaking waves. An
eight-bit, monochromatic, progressive-scan
digital camera was connected to the com-
puter and mounted on the side of the flume
to capture images of these breaking waves.
Through the use of a trigger pulse from
the wave generator, the computer synchro-
nises image acquisition by the camera at the
instance the computer drives the strobe lights
to illuminate the field of view. The wave cycle
was subdivided into 20 overlapping phases
or fields of view. For a particular phase,
100 sequential image pairs were captured
and the information saved on a computer.
Instantaneous velocities were later deter-
mined by measuring the combined displace-
ment of polystyrene particles and air bubbles
entrained in the flow. A digital correlation
image velocimetry technique was employed to
calculate the spatial cross-correlation of the
grey-scale image data by means of computing
the cross-power spectrum of the Fourier-
transformed image samples. This yielded 100
instantaneous velocity flow fields for each
phase. Velocity measurements obtained from
this turbulent flow enabled the computation
of instantaneous and mean flow vorticity.
A single component of vorticity is obtained
from velocity fields by measuring the rotation
of particles seeded in a turbulent flow. The
velocity data were differentiated to yield vor-
ticity field data that characterises the wake of
a plunging breaker. The spatial and temporal
evolution of both instantaneous vorticity and
vorticity of the mean flow are presented only
for phases where turbulence is predominant.
VORTICITY CALCULATION
Vorticity of a flow field is defined as the curl
of the velocity field. It is a vector field that
gives a microscopic measure of the rotation
at any point in the fluid, and plays a vital role
in the dynamics of turbulent fluid flows. The
vorticity vector plays an outstanding role in
turbulence kinematics and dynamics of the
flow (Wang 2012), and has been described as
the principal quantity that defines the flow
structure (Kim et al 1995; Wu et al 2006). If
resolved temporally, the vorticity field can
be much more useful in the study of flow
phenomena than the velocity field (Raffel et
al 1998). At any instant eddies are present in
the flow, ranging in size from the largest geo-
metric scales of the flow down to small scales
where molecular diffusion dominates. These
eddies are continuously evolving in time, and
the superposition of their induced motions
leads to the fluctuating time records normally
observed. By calculating the vorticity fields
from the velocity field it is possible to follow
the motion of coherent structures in the flow.
For a three-dimensional fluid flow,
vorticity can be calculated in Cartesian
coordinates, from the partial derivatives of
the velocity components as:
→ дw дv дu дw
ω = × u→ = x̂ – + ŷ –
дy дz дz дx
дv дu
+ ẑ – (3)
дx дy
where x̂, ŷ and ẑ are the unit basis vectors for
the three-dimensional Euclidean space and u,
v and w are velocities in the x, y, and z direc-
tions, respectively. The turbulent flow studied
here has velocity fields confined only in the
two-dimensional x-z plane. This means that
only one component of vorticity is present.
This component, ωy, which is perpendicular
to the flow and points in the orthogonal
y-direction is given (Sou & Yeh 2011) as:
→ дu дw
ω y = ŷ – (4)
дz дx
Figure 2 shows data points on a computa-
tional grid of a velocity mesh used for calcu-
lating vorticity at a particular point (i, j). The
four yellow-shaded grid points were used to
numerically estimate vorticity at this point.
Instantaneous vorticity at a given grid
point (i, j) was computed from derivatives of
the instantaneous horizontal, ui, and vertical,
wi, velocity components using second-order
central difference approximation (Sou & Yeh
2011; Lee & Lee 2001) as :
u (i, j + 1) – ui(i, j –1)
ωy(i, j) =  i
2∆z
w (i + 1, j) – wi(i –1, j)
– i (5)
2∆x
where (i, j) represent the spatial grid point
in the x-z plane, and ∆x and ∆z are the x
and z-grid spacing, with values 4.1 mm and
5.0 mm respectively.
Vorticity of the mean flow is the trans-
verse vorticity component calculated from
the phase-ensemble averaged velocity com-
ponents < u > and < w >, given by Sou and
Yeh (2011) as:
Journal of the South African Institution of Civil Engineering • Volume 59 Number 1 March 2017
Grid point
j+1
j
j–1
i–1 i i+1
Figure 2 V
 elocity mesh grid used for
calculating vorticity at point (i, j)
using second-order central difference
approximation
→ д〈u〉 д〈w〉
〈ω y〉 = ŷ – (6)
дz дx
This means that vorticity was estimated
from the derivatives of the phase-ensemble-
averaged velocity components, using the
central difference method as:
〈u(i, j + 1)〉 – 〈u(i, j –1)〉
〈ωy(i, j)〉 = 
2∆z
〈w(i + 1, j)〉 – 〈w(i –1, j)〉
– (7)
2∆x
RESULTS AND DISCUSSIONS
Spatially and temporally resolved velocity vec-
tor fields previously measured for a plunging
breaker were used to investigate the phase
evolution of the instantaneous and mean flow
vorticity. For easier visualisation, the vorticity
components are presented in the form of con-
tour plots, and show evolution of the vorticity
as flow progressed. In order to describe the
two-dimensional configuration of the wake
of this turbulent flow, instantaneous vorticity
and vorticity of the mean flow were computed
from the instantaneous velocity fields. It
should, however, be emphasised that most
flows of practical interest are three-dimen-
sional, so vorticity behaves very differently in
three dimensions, compared to two.
Wave images
Figure 3 shows a series of typical grey-scale
images of a 0.4 Hz plunging breaking wave
captured by a monochrome digital camera
at six different phases at one station, where
turbulence was observed to be predominant.
These photographs were captured at one
position and were intended to illustrate
the qualitative features of the breaking
waves that were analysed and presented in
this paper. The temporal evolution of the
breaking is shown in two space coordinates
Figure 3 T ypical grey-scale level images of the
0.4 Hz breaking wave showing six
consecutive phases that include the
crest; these images are for phases t/T =
0.00, 0.05, 0.10, 0.15, 0.20 and 0.25; the
wave is propagating from left to right
in a series of photographs where the time
separation between two consecutive frames
is 0.125 s. The magnitude of the intensity of
each pixel in the images is represented by
levels of grey from black to white. Images
from the 20 wave phases are labelled
relative to the wave period, with t/T = 0.00
representing the first phase; 0.05 for the
second, through to phase t/T = 0.95 for
the twentieth phase. These correspond to
phases: t/T = 0.00, 0.05, 0.10, 0.15, 0.20 and
0.25. The waves propagate from left to right,
and the images evolve from left to right, top
to bottom, as shown in Figure 3. The photo
for phase 0.00 (top left) shows an image of
the breaking wave at the phase when the
crest is just coming into the field of view of
the camera. The last photo in Figure 3 shows
the image at a later phase when the crest has
just passed. White polystyrene beads that
were used as tracer particles together with
air bubbles to determine the velocity flow
fields are clearly visible in the images. The
white cap in the images shows aerated water.
The breaking process is observed to create a
high-speed roller that rides on the front face
of the wave crest. The water mass in the roll-
er is rotating about the horizontal axis that is
parallel to the front face of the wave, creating
a huge amount of turbulence and shear stress
at the front face of the wave. Images at other
phases show the trough part of the wave
which is not as dynamic, so vorticity results
only for the six phases presented.
Instantaneous vorticity (ωy)
Characteristics of vorticity structures in
the turbulent flow, and their evolution as
phase progresses, are presented in Figure 4,
which shows the sequence of contour plots
of instantaneous vorticity generated from
the images. These reveal the evolution of
the vorticity fields as flow progressed. These
5
plots were obtained from the instantaneous
velocity fields of the images presented in t/T =0.00 t/T =0.05

Elevation, z, relative to

Elevation, z, relative to
still water mark (cm)

still water mark (cm)

10 10
Figure 3. These six snapshots of the vorticity
5 5
field are presented here to best illustrate
spatio-temporal evolution of near-surface 0 0
eddies during the passage of the wave crest. –5 –5
The colour bar shows the magnitude and
–10 –10
direction of the vorticity. Colours are coded
according to the intensity of the vorticity –250 –240 –230 –220 –250 –240 –230 –220
field from minimum vorticity (green) to Distance from still water mark (cm) Distance from still water mark (cm)
the maxima (either red or blue). Positive
vorticity indicates motion in clockwise rota- t/T =0.10 t/T =0.15

z, to

z, to
SWL (cm)

SWL (cm)
10 10

z, relative

z, relative
tion, and the direction is into the plane of

to mark

to mark
5 5
the figure, while negative vorticity indicates

Elevation,

Elevation,
anticlockwise rotation, with direction out of 0 0

water

water
Elevation,

Elevation,
the plane of the figure. Positive vorticity has
relative

relative
–5 –5
been conveniently taken to point in the posi- still

still
–10 –10
tive y-direction. Red-filled contours indicate
clockwise or positive vorticity, whereas the –250 –240 –230 –220 –250 –240 –230 –220
blue-filled contours represent counter- Distance from still water mark (cm) Distance from still water mark (cm)
clockwise or negative vorticity. The abscissa
shows distance from the intersection of the t/T =0.20 t/T =0.25
Elevation, z, relative to

Elevation, z, relative to
still water mark (cm)

still water mark (cm)

still water line with the beach slope, and is
taken as negative away from the shore. The
still water line is at elevation z = 0 cm on the
contour plots.
The spatial variation of the instantaneous
vorticity shows a series of strong counter-
rotating filaments observed behind the lead-
ing edge of the breaker. The presence of vor-
ticity patches near the water surface shows
that wave-breaking which occurs at the free
surface is the major source of vorticity in the
flow. The underlying physical mechanism for
this vorticity near the surface was explained
by Sou and Yeh (2011), who suggested that
water on the front face must move faster
than the propagation speed, hence the for-
mation of a surface roller. On the other hand,
the water behind the front face must move
more slowly than the propagation speed, cre-
ating a divergence on the surface. The fila-
ments are observed to diffuse to the bottom
of the flume, reaching the bed after the crest
has passed (phases 0.20 and 0.25). As pointed
out by Longo (2008), deepening is quite fast
immediately after breaking and is slower at
subsequent phases. Pairs of counter-rotating
filaments have peak vorticity of magnitude
100 s–1 which is observed to decrease as flow
progresses. These patches have only been
shown to exist, but the mechanism by which
they are set up is not clear. However, they
are understood to dissipate the remaining
wave energy. It is the interactions between
these adjacent counter-rotating filaments
that produce a wake of complex vortex
distribution behind the wave crest. These
filaments, firstly observed at elevations above
z = –5 cm at phase 0.00, originate from
the free surface and obliquely propagate in
the direction of the wave. Their presence
is an indication of strong mixing taking
10 10
5 5
0 0
–5 –5
–10 –10
–250 –240 –230 –220 –250 –240 –230 –220
Distance from still water mark (cm) Distance from still water mark (cm)
–100 –50 0 50 100
1/s
Figure 4 C
 ontour plots showing the evolution of instantaneous vorticity filaments with phase, for
breaking wave images shown in Figure 3; the still water line (SWL) is at z = 0 cm
place there. Near the trough level there are saddle points, as seen in the last two panels
occasional regions devoid of strong vorticity. of Figure 4. This cancellation is the key
Early phases also show large expanses of mechanism for the decay of two-dimensional
the fluid with nearly zero vorticity. Bakewell vorticity distributions. Leweke et al (2016)
and Lumley (1967) and Aubry et al (1988) illustrated this mechanism to show how, in a
used the proper orthogonal decomposition viscous flow, two identical co-rotating vorti-
in the near-wall region, and observed that a ces ultimately merge, while counter-rotating
pair of counter-rotating stream-wise vortices ones cancel. Coalescing of co-rotating vor-
contain the largest amount of energy. Ting ticity filaments is evidenced by the eventual
(2006; 2008) used particle image velocimetry formation of a positive vorticity filament ring
(PIV) to study instantaneous turbulent veloc- observed to be centred around x = –240 cm
ity fields associated with a broken solitary in the last panel of Figure 4. Negative vorti-
wave on a plane slope, and also observed city (anticlockwise) filaments are observed
that large eddies were composed of two to disappear as a result of mixing, leaving
counter-rotating vortices. the fluid volume filled mainly with positive,
Just as parallel electric currents in the clockwise filaments. The filaments deepen
same direction attract one another because and diffuse towards the bottom, as shown
of their magnetic interaction, co-rotating in phases 0.10–0.25. Saddle points near the
filaments coalesce and ultimately merge crest are characterised by strong filaments
in viscous flow as flow progresses, while of obliquely descending positive and negative
counter-rotating vorticity filaments inter- vorticity below them.
fere with one another by viscous vorticity Characteristics of instantaneous vor-
cancellation. Vorticity cancellation ulti- ticity structures in the flow, such as the
mately leaves predominantly weaker positive vorticity filament width (core diameter in
vorticity throughout most of the wake and three-dimensional vorticity), length and
negative vorticity confined to just below the translational speed of the filaments between

6 Journal of the South African Institution of Civil Engineering • Volume 59 Number 1 March 2017
 Wray (1998) have shown that the filaments
100 (cm/s) can simultaneously have lengths of the order
of the injection scale and a core diameter
10 as small as the Kolmogorov scale. A pair of
Elevation, z, relative to still water mark (cm)

counter-rotating filaments has an estimated

centre separation of about 2.0 cm. Vorticity
5 analysis not only identifies any shearing
motion, but also vortex cores present in the
flow (Adrian et al 2000). Kevlahan and Farge
0 (1997) noted that most of the time filaments
formed during a vortex merger tend to be
distributed as a sequence of approximately
–5 circular rings around the coherent vortex.
In the contour plot of phase 0.20, a large
positive vorticity structure of about 60 s–1
–10 vorticity magnitude is observed, which
is centred around (x, z) = (–250, –7) cm.
Results show this structure to have devel-
–250 –240 –230 –220 oped into a vortex ring that has evolved and
Distance x, from still water mark (cm) moved to position (x, z) = (–240, –7) cm in
the plot of phase 0.25. Figure 5 shows the
Figure 5 C
 orresponding instantaneous velocity field for the vorticity field for phase t/T = 0.25 instantaneous velocity field that gave rise to
shown in Figure 4 vorticity contours presented in the last panel
of Figure 4. The vortex ring is clearly visible
in the velocity field.
(a) (b) The wave was divided into 20 phases and
Elevation, z, relative to

Elevation, z, relative to
still water mark (cm)

still water mark (cm)

10 10
the period of wave used is 2.5 s. The time
5 5
interval between two consecutive phases is
0 0 0.125 s. Thus the ring structure has moved
–5 –5 about 10 cm in 0.125 s. This implies that the
ring propagates at about 80 cm/s towards
–10 –10
the shore, which is about 0.74 c, where
–300 –290 –280 –270 –250 –240 –230 –220 c = 108 cm/s is the wave phase velocity.
Distance from still water mark (cm) Distance from still water mark (cm) This is in agreement with Kimmoun and
Branger (2007), who observed that vorticity
10 (c) (d) filaments propagated obliquely towards
z, to

z, to
SWL (cm)

SWL (cm)
z, relative

z, relative

5 5 the bottom, moving more slowly than the

to mark

to mark

wave crest. Miller (1976) traced vortices and

Elevation,

Elevation,

0 0 noted that they travelled more slowly than

water

water
Elevation,

Elevation,

the wave velocity, and drifted downwards

relative

relative

–5 –5 while expanding.
still

still

–10 Figure 6 shows contours of instantaneous

–210 –200 –190 –180 –160 –155 –150 –145 –140 vorticity under the wave crests, measured
Distance from still water mark (cm) Distance from still water mark (cm) at five different cross-shore positions along
the flume. The Figure 6(e) measurements
(e) were captured at a point closest to the shore,
Elevation, z, relative to
still water mark (cm)

5
with its centre only about 1.0 m from the
shore, while the panel in Figure 6(a) is in the
0 deepest water and furthest from the shore,
just shoreward of the break point. The flume
–5 –100 –50 0 50 100 bed is located at different positions in the
(s–1) panels. While the size of vorticity filaments
–110 –105 –100 –95 –90 is observed to decrease towards the shore,
Distance from still water mark (cm) there is a general increase in the intensity.
Stretching a filament along its axis will make
Figure 6 C
 ontour plots of instantaneous vorticity under the wave crests, measured at five it rotate faster and decrease its width (or
different positions along the flume – (a) is furthest from the shore, while (e) is closest; diameter in three-dimensions) in order to
note that there is no time relationship between these plots maintain its angular momentum constant.
A well-known example in fluid mechanics is
frames as the structures travel downstream They are estimated to have individual length the bath tub vortex that becomes smaller and
in the flow, have been estimated. It can be scales that stretch up to 12 cm (height of rotates faster as it goes from the fluid surface
observed that instantaneous vorticity fila- wave used) and a typical mean width of to the exit (Petitjeans 2003). Both negative
ments are created in a large range of sizes. about 2.0 cm. Simulations by Jimenez and and positive vorticity filaments of magnitude

Journal of the South African Institution of Civil Engineering • Volume 59 Number 1 March 2017 7
up to 100 s–1 can be observed impinging
the flume bed as flow progresses from deep

Elevation, z, relative to

Elevation, z, relative to
still water mark (cm)

still water mark (cm)

10 10
water to shallow water. Such high-vorticity
5 5
filaments are responsible for lifting sedi-
ments from the bed and transporting them. 0 0

–5 –5
Mean flow vorticity 〈w y〉
–10 –10
Figure 7 shows contour plots of the mean
flow vorticity for the six phases of interest. –250 –240 –230 –220 –250 –240 –230 –220
The plots show that the mean flow is charac- Distance from still water mark (cm) Distance from still water mark (cm)
terised by a single, strong, positive vorticity
filament located below the crest, at the shear

z, to

z, to
SWL (cm)

SWL (cm)
10 10

z, relative

z, relative
boundary layer. The shear layer arises from

to mark

to mark
5 5
the fast-moving mixture of crest water/

Elevation,

Elevation,
bubbles and the opposing slow-moving water 0 0

water

water
Elevation,

Elevation,
near the trough. Mean flow vorticity was
relative

relative
–5 –5
obtained by averaging over 100 instantane- still

still
–10 –10
ous vorticity fields at a particular phase to
give the ensemble-averaged vorticity. Mean –250 –240 –230 –220 –250 –240 –230 –220
vorticity distribution shows moderate spatial Distance from still water mark (cm) Distance from still water mark (cm)
concentration gradients, in contrast to the
patchy instantaneous vorticity. During the
Elevation, z, relative to

Elevation, z, relative to
still water mark (cm)

still water mark (cm)

early phases of the flow, there is a concen-
trated region of positive vorticity, about 5 cm
thick near the breaker shear layer. This posi-
tive vorticity was due to the shear generated
by the high-speed uprush flow overlaying
a slower receding downwash layer of fluid
(Misra et al 2005). Co-rotating instantane-
ous filaments observed earlier coalesce to
become a larger and stronger single filament.
Kevlahan and Farge (1997) observed that
strong shearing associated with the vortex
mergers produces intense filaments of vorti-
city. Thus mean flow organises with the for-
mation of a large coherent positive vorticity
filament near the shear boundary layer.
Even though the instantaneous flow
between the shear layer and the free surface
is very turbulent (Figure 4), the observed
mean vorticity field is rather weak, except
within the shear layer. This is consistent
with observations by Lin and Rockwell
(1994). Averaging instantaneous vorticity
makes small-scale vorticity filaments van-
ish, so contours of the mean flow vorticity
only show the development of large-scale
structures within and around the shear layer,
which develops and moves downstream. The
fluid beneath the elevation of z = –5 cm is
relatively vorticity-free for the first two phas-
es, but as flow progresses, it is observed that
positive vorticity diffuses towards the flume
bed. As observed by Sou and Yeh (2011) and
presented in Figure 7, the fluid immediately
behind the crest (for phases 0.10 to 0.25) is
characterised by negative vorticity that rides
above positive vorticity at elevation centred
around z = –5 cm. As pointed out by Dabiri
and Gharib (1997), the negative vorticity
seen on the wave, below the surface, and
above the positive shear layer above z = 0 cm,
indicates the existence of a stagnation point
10
5 5
0 0
–5
–10
–250 –240 –230 –220
Distance from still water mark (cm)
–40 –20 0
Figure 7 C
 ontour plots showing evolution of the vorticity of the mean flow as flow progressed
at that location. In agreement with observa-
tions by Sou and Yeh (2011), the maximum
intensity of mean vorticity occurs around
the shear layer, and the intensity decreases
in magnitude as flow progresses. Thus
maximum positive vorticity will remain in
the shear layer region between the upper,
faster-moving part of the breaking wave and
the quiescent region below. Below the trough
level mean flow vorticity is about the same
order of magnitude as the wave phase speed
divided by the local water depth, c/h = √‌gh/h.
For the same region, Chang and Liu, (1998;
1999) also found that the vorticity generated
by wave breaking was of the same order of
magnitude, as the phase speed divided by the
local water depth.
Mean vorticity contour plots also reveal
the phenomenon of vorticity-shedding
at later phases of the flow, in which the
tail of the initially strong boundary layer
vorticity peels off, weakens and diffuses
from the shear boundary layer, reaching
the flume bed after the crest has passed.
This results in the decay of the vorticity
from the neighbourhood of the initially
stronger region (Jimenez 2004). The original
10
–5
–10
–250 –240 –230 –220
Distance from still water mark (cm)
20 40 60 80
1/s
layer breaks in this way into nearly circular
cores. The strong negative vorticity patches
near the free surface directly above regions
–255 cm < x < –242 cm for phase 0.10, and
–242 cm < x < –232 cm for phase 0.15, may
be responsible for the shedding observed.
The explanation for the observed decay is
that, at later phases, the effects of small-scale
mixing due to the strong turbulence in
the flow have greatly reduced the presence
of patches of vorticity in the flow. It was
shown by other researchers (e.g. Stanly et al
2002) that, while the large scales in the flow
field adjust slowly to variations in the local
mean velocity gradients, the small scales
adjust rapidly.
CONCLUSIONS
Results have been presented of instantaneous
and mean vorticity, aimed at determining
their evolution as flow progressed. These
were estimated from the measured velocity
fields using a central difference method.
Results have shown that the spatial
distribution of instantaneous vorticity is
extremely patchy near the crest, with isolated

8 Journal of the South African Institution of Civil Engineering • Volume 59 Number 1 March 2017
filaments of high positive and negative vorti-
city that have steep gradients. In-between the
patches and below the still water line are large
expanses with nearly zero vorticity. Vorticity
measurements at different phases at the same
point showed that, as flow progressed, there
is two-dimensional merging of co-rotating
filaments and cancellation of the counter-
rotating. It was observed that instantaneous
vorticity filaments are created in a large
range of sizes, peaking at the height of the
wave used. Mean vorticity was observed to
have completely different characteristics.
The principal feature of the phase-averaged
vorticity is a strong positive vorticity structure
that appears in the flow near the shear bound-
ary layer. A possible reason for this is that
vorticity filaments have sizes distributed over
several scales present in the flow. Averaging
instantaneous vorticity makes small-scale
vorticity filaments vanish, so the contours
of the mean flow vorticity only show the
development of large-scale structures within
and around the shear layer, which develops
and moves downstream. This occurs near the
shear layer, around elevation z ≈ 0, where the
uprush opposes the undertow. At this point
a large clockwise-rotating vorticity is formed
underneath the wave crest. As the wave crest
propagates along the flume, the width of the
positive vorticity broadens as it sequentially
interacts with downstream vorticity which are
remnants of previous breaking waves. This
main structure of the wave diffuses rapidly
into the interior of the wave after breaking,
where molecular diffusion plays a part in
stretching the filaments and redistributing the
vorticity. Below the trough level, mean flow
vorticity is about the same order of magnitude
as the wave phase speed divided by the local
water depth c/h = √‌gh/h. Peak magnitudes of
both instantaneous and mean flow vorticity,
generated by wave breaking, are one order
of magnitude greater than the phase speed
divided by local depth c/h.
The results presented here may be useful
for modelling two-dimensional turbulent
flows. Identification of scaling laws for fila-
ments may also bring some new insights.
Statistical properties of the filaments, includ-
ing filament life time, number of filaments
per unit volume, number of high vorticity
filaments impinging the bottom and their
magnitudes could also be determined from
this data, as this information is considered
vital in sediment transport studies.
ACKNOWLEDGEMENTS
The experiments reported here were
conducted in the Coastal and Hydraulics
Laboratory at the CSIR in Stellenbosch,
South Africa. The author would like to
Journal of the South African Institution of Civil Engineering • Volume 59 Number 1 March 2017
acknowledge the technical support received
during the experimental runs.
REFERENCES
Abry, P, Fauve, S, Flandrin, P & Laroche, C 1994.
Analysis of pressure fluctuations in swirling turbulent
flows. Journal de Physique II, 4(5): 725–733.
Adrian, R J, Christensen, K T & Liu Z-C 2000. Analysis
and interpretation of turbulent velocity fields.
Experiments in Fluids, 29: 275–290.
Aubry, N, Holmes, P, Lumley, J L & Stone, E 1988. The
dynamics of coherent structures in the wall region
of a turbulent boundary layer. Journal of Fluid
Mechanics, 192: 115–173.
Bakewell, H P, Lumley, J L 1967. Viscous sublayer and
adjacent wall region in turbulent pipe flow. Physics of
Fluids, 10: 1880–1889.
Belin, F, Maurer, J, Tabeling, P & Williame, H 1996.
Observation of intense filaments in fully developed
turbulence. Journal de Physique II (France), 6: 573–583.
Brachet, M, Meiron, D, Orszag, S A, Nickel, B, Morf, R &
Frisch, U 1983. Small-scale structure of a Taylor-Green
vortex. Journal of Fluid Mechanics, 130: 411–452.
Buhler, O 2000. On the vorticity transport due to
dissipating or breaking waves in shallow-water flow.
Journal of Fluid Mechanics, 407: 235–263.
Cadot, O, Douady, S & Couder, Y 1995.
Characterization of the low-pressure filaments in a
three-dimensional turbulent shear flow. Physics of
Fluids, 7: 630–646.
Chang, K A & Liu, P L-F 1998. Velocity, acceleration
and vorticity under a breaking wave. Physics of
Fluids, 10(1): 327–329.
Chang, K A & Liu, P L-F 1999. Experimental
investigation of turbulence generated by breaking
waves in water of intermediate depth. Physics of
Fluids, 11(11): 3390–3400.
Dabiri, D & Gharib, M 1997. Experimental investigation
of the vorticity generation within a spilling water
wave. Journal of Fluid Mechanics, 330: 113–139.
Douady, S, Couder, Y & Brachet, M-E 1991. Direct
observation of the intermittency of intense vorticity
filaments in turbulence. Physical Review Letters,
67: 983–986.
Dritschel, D G 1989. The survivability of filaments in
two-dimensional fluids at extremely high Reynolds
numbers. In: Moffatt, H K & Tsinober, A (Eds.),
Topological Fluid Mechanics, Proceedings of the
IUTAM Symposium, Cambridge, UK: Cambridge
University Press, p 286.
Dritschel, D G, Haynes, P H, Juckes, M N & Shepherd,
T G 1991. The stability of two-dimensional vorticity
filaments under uniform strain. Journal of Fluid
Mechanics, 230(1): 647–665.
Elhmaidi, D, Provenzale, A, Lili, T & Babiano, A 2004.
Stability of two-dimensional vorticity filaments.
Physics Letters A, 333: 85–90.
Fauve, S, Laroche, C & Castaing, B 1993. Pressure
fluctuations in swirling turbulent flows. Journal de
Physique II (France), 3: 271–278.
Galvin, C J 1972. Wave breaking in shallow water. In:
Meyer, R E (Ed.), Waves on Beaches and Resultant
Sediment Transport. New York: Academic Press,
413–456.
Hunt, J C R & Vassilicos, J C (Eds.) 2000. Turbulence
Structure and Vortex Dynamics. Cambridge, UK:
Cambridge University Press.
Jimenez, J, Wray, A A, Saffman, P G & Rogallo, R S
1993. The structure of intense vorticity in isotropic
turbulence. Journal of Fluid Mechanics, 255: 65–90.
Jimenez, J & Wray, A A 1998. On the characteristics of
vortex filaments in isotropic turbulence. Journal of
Fluid Mechanics, 373: 255–285.
Jimenez, J T 2004. Turbulence and vortex dynamics.
Available at: https://www.scribd.com/doc/80320496/
Javier-Jimenez-Turbulence-and-vortex-dynamics
Keshavarzi, A & Ball, J 2011. Stochastic nature of flow
turbulence and sediment particle entrainment over
the ripples at the bed of open channel using image
processing technique. In: Bhuiyan, F (Ed.), Sediment
Transport – Flow Processes and Morphology,
Chapter 3. Rijeka, Croatia: InTech, 69.
Kevlahan, N K-R & Farge, M 1997. Vorticity filaments
in two-dimensional turbulence: Creation, stability
and effect. Journal of Fluid Mechanics, 346: 49–76.
Kim, W Y, Walker, P G & Pedersen, E M 1995. Left
ventricular blood flow patterns in normal subjects: A
quantitative analysis by three-dimensional magnetic
resonance velocity mapping. Journal of the American
College of Cardiologists, 26, 224–238.
Kimmoun, O & Branger, H 2007. A particle image
velocimetry investigation on laboratory surf-zone
breaking waves over a sloping beach. Journal of Fluid
Mechanics, 588: 353–397.
Lajeunesse, E, Malverti, L & Charru, F 2010. Bed
load transport in turbulent flow at the grain scale.
Experiments and modeling. Journal of Geophysical
Research, 115: F04001.
Lee, S J & Lee, S H 2001. Flow field analysis of a
turbulent boundary layer over a riblet surface.
Experiments in Fluids, 30: 153–160.
Leweke, T, Dizes, S L, Williamson, I & Charles, H K
2016. Dynamics and instabilities of vortex pairs.
Annual Review of Fluid Mechanics, 48: 1–35.
Lien R-C & Sanford, T B 2000. Spectral characteristics
of velocity and vorticity fluxes in an unstratified
turbulent boundary layer. Journal of Geophysical
Research, 105(C4): 8659–8672.
Lin, J C & Rockwell, D 1994. Instantaneous structure of
a breaking wave. Physics of Fluids, 6: 2877–2879.
Lin, J, Shao, X & Ni, L 2002. Wavelet analysis of
coherent structures in a three-dimensional mixing
layer. Acta Mechanica Sinica, 18(1): 42–52.
Long, J W & Özkan-Haller, H T 2009. Low-frequency
characteristics of wave group-forced vortices.
Journal of Geophysical Research, 114: C08004.
Longo, S 2008. Vorticity and intermittency within the
pre-breaking region of spilling breakers. Coastal
Engineering, 56(3): 285–296.
Meunier, P, Le Dizès, S & Leweke, T 2005. Physics
of vortex merging. Comptes Rendus Physique,
6(4–5): 431–450.
Miller, R L 1976. Role of vortices in surf zone
prediction: Sedimentation and wave forces. In: Davis,
R A & Ethington, R L (Eds.), Beach and Nearshore
9
 Sedimentation, Tulsa, OK: Society of Economic
Paleontologists and Mineralogists, 92–114.
Misra, S K, Kirby, J T & Brocchini, M 2005. The
turbulent dynamics of quasi-steady spilling breakers
– Theory and experiments. Research Report No.
CACR-05-08, Newark, NJ: Center for Applied
Coastal Research.
Mukaro, R, Govender, K & Gledhill, H 2011. Velocity
flow fields and water level measurements in shoaling
and breaking water waves. Proceedings, 7th South
African Conference on Computational and Applied
Mechanics, 10–13 January, Pretoria.
Mukaro, R & Govender, K 2013. Experimental study
of turbulence and water levels in shoaling and
breaking water waves using digital image processing
techniques. Journal of the South African Institution
of Civil Engineering, 55(2): 59–74.
Nelson, J M, Shreve, R L, McLean, S R & Drake, T G
1995. Role of near-bed turbulence structure in bed
load transport and bed form mechanics. Water
Resources Research, 31: 2071–2086.
10 Journal of the South African Institution of Civil Engineering • Volume 59 Number 1 March 2017
Nielsen, P 1984. Field measurements of time-averaged
suspended sediment concentration under waves.
Coastal Engineering, 8: 51–72.
Petitjeans, P 2003. Stretching of a vortical structure:
Filaments of vorticity. Europhysics News, 34(1): 20–23.
Raffel, M, Willert, C E, Wereley, S & Kompenhans, J
1998. Particle Image Velocimetry. A Practical Guide,
2nd ed. Berlin: Springer, 189.
Ryu, Y, Chang, K A & Mercier, R 2007. Runup and green
water velocities due to breaking wave impinging and
overtopping. Experiments in Fluids, 43(4): 555–567.
Siggia, E D 1981. Numerical study of small-scale
intermittency in three-dimensional turbulence.
Journal of Fluid Mechanics, 107: 375–406.
Sou, I M & Yeh, H 2011. Laboratory study of the cross-
shore flow structure in the surf and swash zones.
Journal of Geophysical Research, 116: C03002.
Stanly, A S, Sarkar, S & Mellado, J P 2002. A study
of the flow-field evolution and mixing in a planar
turbulent jet using direct numerical simulation.
Journal of Fluid Mechanics, 450: 377–407.
Ting, F C K 2006. Large-scale turbulence under a
solitary wave. Coastal Engineering, 53: 441–462.
Ting, F C K 2008. Large-scale turbulence under a
solitary wave. Part 2. Forms and evolution of coherent
structures. Coastal Engineering, 55: 522–536.
Voulgaris, G & Collins, M B 2000. Sediment
resuspension on beaches: Response to breaking
waves. Marine Geology, 167: 167–187.
Wang, L 2012. Structures of the vorticity tube segment
in turbulence. Physics of Fluids, 24: 045101.
Williams, J J 1990. Video observations of marine gravel
transport. Geo-Marine Letters, 10: 157–164.
Wu, J Z, Ma, H Y & Zhou, M D 2006. Vorticity and Vortex
Dynamics. Berlin-Heidelberg, Germany: Springer.
Zhang, D & Sunamura, T 1990. Conditions for
the occurrence of vortices induced by breaking
waves. Coastal Engineering, 33: 145–155.
Zhang, D P 1994. Wave Flume Experiments on the
Formation of Longshore Bars Produced by Breaking
Waves. Science Report 15, Institute of Geoscience,
University Tsukuba, Japan, 47–105.
Planning for desalination in TECHNICAL PAPER
the context of the Western Journal of the South African
Institution of Civil Engineering

Cape water supply system ISSN 1021-2019

Vol 59 No 1, March 2017, Pages 11–21, Paper 1200

C L Blersch, J A du Plessis CATHERINE BLERSCH (Pr Eng, MSAICE) is an

Associate in Aurecon’s Cape Town office, where
she has been involved in water engineering for
the past six and a half years, with a focus on
As water demands continue to grow, South Africa is starting to consider seawater desalination wastewater treatment works, and water
as a potential future supply source, and it is currently being investigated at a feasibility level resource management and planning. She
in a number of coastal cities, including Cape Town. Desalination is different to conventional obtained a BEng (Civil) from the University of
Pretoria, and an MEng from Stellenbosch
surface and groundwater supply sources in that it is climate-resilient, having an assurance of
University. Currently she is busy with an MBA at the University of Cape Town’s
supply of essentially 100 percent. However, the increased reliability comes at a great cost. This Graduate School of Business.
paper presents a methodology developed for modelling a proposed desalination plant as an
Contact details:
integrated component of the Western Cape Water Supply System, in order to optimise system
Department of Civil Engineering
operating rules and cost. The modelling entailed short-term and long-term system analyses in Stellenbosch University
the Water Resources Yield Model and Water Resources Planning Model, and estimation of first- Private Bag X1
order capital and operating costs in order to calculate and compare Unit Reference Values. The Matieland 7602
maximum increase in yield was found to occur when the seawater desalination plant is used South Africa
T: +27 21 526 9400
as a base supply, operational all the time. There was little benefit, in terms of system yield, in
E: catherine.blersch@aurecongroup.com
using the desalination plant as an emergency supply source only. Unit reference values for the
desalination plant decrease as the percentage supply from the plant increases, meaning that the
PROF KOBUS DU PLESSIS (Pr Eng, MSAICE,
lowest possible cost per cubic metre of water supplied is when the desalination plant is used FIMESA) is Head of the Department of Civil
as a base supply. It was also apparent that the unit reference values decrease with an increase Engineering at Stellenbosch University, and has
in desalination plant capacity, suggesting that, from an economic perspective, the optimal been responsible for Hydrology and
solution would be to have one large desalination plant operational immediately. Environmental Engineering at the university for
the past 13 years. He has a special interest in the
integrated management of water resources in
South Africa as applied by local authorities, as
INTRODUCTION traction as a viable alternative. Seawater well as in flood hydrology. He obtained his PhD (Water Governance), MEng
Desalination can take the form of seawater desalination is currently being investigated (Water Resource Management) and BEng (Civil) from the University of
desalination, desalination of brackish or pol- at a feasibility level in Saldanha Bay, Cape Stellenbosch. During his more than 29 years of experience in the water
sector, he also worked for the Department of Water Affairs, the City of Cape
luted groundwater and the use of membrane Town, Port Elizabeth and Durban, and
Town and the West Coast District Municipality. He presently serves on the
processes in advanced water treatment pro- is mentioned in all of the national water Executive Committee of IMESA (Institute of Municipal Engineering of
cesses for water reuse. All of these forms of planning documents (Department of Water Southern Africa) and on the Education and Training Panel of SAICE (South
desalination are relevant in South Africa and Affairs and Forestry 2008:26; Department of African Institution of Civil Engineering).
are currently being considered as possible Water Affairs 2010:7; Department of Water Contact details:
supply sources for many major cities, includ- Affairs 2013b:31). Department of Civil Engineering
ing Cape Town. However, the focus of this Considering the existing plants, there Stellenbosch University
Private Bag X1
paper is seawater desalination only, firstly have been some limitations in terms of
Matieland 7602
because it is likely to be a large-scale supply integrated planning for seawater desalination South Africa
source (as opposed to brackish groundwater in South Africa. In Mossel Bay, for example, T: +27 21 808 4358
desalination which will be on a much smaller a 15 Mℓ/day seawater desalination plant was E: jadup@sun.ac.za
scale), and secondly because the practicality constructed in 2011 in response to the severe
and acceptance of reuse of treated effluent drought in the area. Since the commission-
has not yet been fully researched in South ing of the plant, it has remained virtually
Africa. Further limitations are discussed in untapped, with the municipality preferring
the conclusion of the paper. to make use of its less expensive surface
Historically, less costly surface and water resources (Mallory et al 2013). Similar
groundwater resources have been available to problems have been experienced in Australia,
meet South Africa’s water demands, mean- where large-scale seawater desalination was
ing that seawater desalination has not been implemented, at great cost, in response to a
considered a viable water source. There are decade-long drought (Onishi 2010).
currently only six small desalination plants A feasibility study for large-scale seawater
in operation in the country, half of which desalination for Cape Town is currently
were constructed as emergency schemes under way. The decision regarding when
in response to the severe Southern Cape to proceed with seawater desalination will
drought of 2009/2010. However, as growing most likely be driven by the feasibility study
water demands place pressure on water and Western Cape Water Supply System
resources, desalination has begun to gain (WCWSS) planning mechanisms. However, Keywords: seawater desalination, water resource planning, system analysis

11
Blersch CL, Du Plessis JA. Planning for desalination in the context of the Western Cape water supply. J. S. Afr. Inst. Civ. Eng. 2017:59(1),
Art. #1200, 11 pages. http://dx.doi.org/10.17159/2309-8775/2017/v59n1a2
seawater desalination differs from the
surface and groundwater resources of the
WCWSS, firstly in its higher capital and
operating costs, and secondly in its resilience
to climate variability, having an assurance
of supply of essentially 100 percent. Factors
such as integration into the WCWSS, supply
risks and cost implications may therefore be
more relevant in evaluating desalination than
a 98 percent assurance of supply from tradi-
tional water supply options, as is currently
the case. The higher costs and assurance of
supply of desalination will also impact on the
overall operational philosophy.
The aim of this research was to deter-
mine what operating rules and modified
planning criteria are required to optimise the
implementation of large-scale desalination as
part of the WCWSS, in order to ensure cost-
effective provision of water at an appropriate
assurance of supply.
BACKGROUND AND
LITERATURE REVIEW
Existing Western Cape Water
Supply System
The Western Cape area is characterised by a
Mediterranean climate, with rainfall occur-
ring in the winter months (May to October)
when demands are low, and minimal rainfall
occurring in the summer months when
demands are high. Approximately 50 per-
cent of the total system storage is available
for storing winter flows to meet summer
demand, with the remaining 50 percent
required for long-term drought storage
(Shand & Sparks 2004:3). The main dams
of the WCWSS are operated as an integral
system in order to reduce the probability of
spillage, hence increasing the overall system
yield (Department of Water Affairs 2007:31).
The WCWSS supplies water to various
towns and irrigators in the Western Cape,
including the City of Cape Town. The latest
WCWSS planning documents give the total
water usage from the system as 503 mil-
lion m3/a in 2013 (Department of Water
Affairs 2013a:21) and the total integrated
system yield at a 98 percent assurance of
supply as 596 million m3/a (Department of
Water Affairs 2011:7). To manage demands,
yield and planning models of the WCWSS
have been set up in the Water Resources
Yield Model (WRYM) and Water Resources
Planning Model (WRPM) respectively in
order to simulate the inflows and demands
on the system and plot storage trajectories
of the coming year. These models (WRYM
and WRPM) are run annually in November
in order to determine whether the dams are
full enough to meet the projected summer
12 Journal of the South African Institution of Civil Engineering • Volume 59 Number 1 March 2017
demands at the required assurance of sup-
ply, or whether water restrictions should be
implemented (Shand & Sparks 2004:6).
Seawater desalination as possible future
supply source for the WCWSS was first
assessed at a pre-feasibility level in 2003
as part of the Bulk Water Supply Study. A
feasibility study was commissioned in 2011,
and was set for completion at the end of
2013, although the findings have not yet been
released. The anticipated capacity based on
the feasibility study would be 150 Mℓ/day,
upgradeable to 450 Mℓ/day (Department of
Water Affairs 2013a:12).
Local research into desalination
and integrated system modelling
Despite the various feasibility studies which
are currently under way, and the ambitious
outlook for seawater desalination develop-
ment in South Africa, there has been little
research into the conjunctive use of seawater
desalination with surface water supply
sources, and how to integrate seawater
desalination plants into existing water supply
systems. The only South African authors
who have presented any material on the topic
are Mallory et al (2013) who undertook a
study into the optimal operation of the sea-
water desalination plant in Mossel Bay using
the Water Resources Modelling Platform.
The work of Mallory et al (2013) began to
tackle some of the key questions associated
with the value of water to different users, and
setting of operating rules for a desalination
plant.
Although not specifically related to sea-
water desalination, the work of Van Niekerk
and Du Plessis (2013a:543) into the conjunc-
tive evaluation of the costs and benefits of
inter-basin water transfer schemes illustrates
a number of useful principles which could be
easily applied to integrate system modelling
for seawater desalination. Van Niekerk and
Du Plessis (2013a:543) found that the so-
called incremental approach has historically
been followed in evaluating the viability of
inter-basin transfer schemes, which assumes
that volume transferred per annum is equal
to the difference between the projected
future water demands and the current sys-
tem yield. The authors found that the origi-
nal projections vastly exceeded the actual
transfers, mainly because the incremental
approach ignores the stochastic nature of
the conditions of the receiving system. To
address these concerns, Van Niekerk and Du
Plessis (2013a:543) propose a comprehensive
approach in which the receiving water
basin and the transfer scheme are modelled
stochastically as an integrated system in the
WRPM. Taking this concept further, Van
Niekerk and Du Plessis (2013b:551) postulate
that the results of such an analysis should
be used when calculating the Unit Reference
Value (URV) of any newly proposed scheme
(Van Niekerk & Du Plessis 2013b:552).
Estimating desalination costs
South African approach to
desalination costing
Du Plessis et al (2006) present a step-by-step
guide to determining first-order capital and
operating costs for desalination plants of
different sizes in South Africa. The process
centres on selecting a plant capacity, deter-
mining the required membrane area and
feed water pressures, and calculating capital
and operating costs with these parameters as
a base. In terms of capital costs, the desalina-
tion cost component can be estimated as
a function of the membrane area, the pre-
treatment costs can be calculated as a func-
tion of the feed flow rate, and the combined
cost of the intake, outlet and post-treatment
systems can be estimated as a percentage of
the capital cost of the desalination plant and
pre-treatment system (15 to 30 percent).
Du Plessis et al (2006) suggest that the
operating costs comprise energy costs, which
can be estimated based on the total energy
requirement of the plant (which is primarily
a function of feed-water pressure), chemical
costs of approximately R0.50/m3, membrane
replacement costs based on a unit cost of
R165/m2 of membrane area and a membrane
lifespan of six years, and annual labour
and maintenance costs of about 5 percent
of the total capital cost. All costs and rates
provided by Du Plessis et al (2006) are for a
base date of 2006 and were escalated to 2014
at 6 percent per annum.
This methodology was applied to the pro-
posed Cape Town desalination plant, situ-
ated on the west coast, based on a capacity of
100 Mℓ/day and a two-pass system, giving a
first-order (2014) estimate of the capital cost
of R1 054 million and the total specific cost
of R16.44/m3.
International seawater desalination
literature
In terms of international literature on seawa-
ter desalination cost estimation, Karagiannis
and Soldatos (2008) studied almost 100 dif-
ferent seawater desalination plants and pre-
sented a summary of first-order desalination
costs derived from literature. Based on their
costs, a 100 Mℓ/day seawater desalination
plant would cost approximately R9.84/‌m3.
Similarly, Wittholz et al (2008) collated a
database of over 300 seawater desalination
plants and derived an equation for estimat-
ing capital costs for seawater reverse osmosis
desalination plants based on the plant

Capital costs for existing large seawater RO desalination plants
100 000
90 000
80 000
Capital cost (R/m3/day)
70 000
60 000
50 000
40 000
30 000
20 000
10 000
0
0 50
Figure 1 S eawater desalination plant capital costs in Australia, Israel, Algeria, Spain and South Africa
capacity. Using their equation, a 100 Mℓ/day
seawater desalination plant would have an
equivalent 2014 cost of R1 862 million and
unit production cost of R9.95/m3. In a more
recent study, Ghaffour et al (2013) provided
a review of the economics of desalination
worldwide, noting that unit water costs for
seawater reverse osmosis desalination have
decreased rapidly over the past decade,
meaning that a 100 Mℓ/day plant could cost
as little as R960 million.1
Analysis of known costs of seawater
desalination plants
Figure 1 shows costs in Rand per m3/day of
desalination plant capacity plotted against
plant capacity for the recently constructed
desalination plants in Australia, and plants
in Israel, Algeria and Spain, extracted from
Ghaffour et al (2013), and the Mossel Bay
desalination plant, being the only large
desalination plant in South Africa.
Attempts to derive any kind of relation-
ship from the values presented in Figure 1
were unsuccessful, given the large scatter.
One of the problems with desalination costs
quoted in literature is that it is often unclear
whether the costs are all-inclusive or for
the desalination portion of the plant only
(i.e. excluding intake and outlet structures
or general civil works). This is particularly
relevant for larger plants. It is clear that the
Australian desalination plants were signifi-
cantly more expensive than recent plants in
other parts of the world, most likely a result
of the emergency nature of the schemes,
which invariably pushes up costs. However,
for Israel, Algeria and Spain the costs are
fairly similar across the range of capacities,
varying from R10 700 to R14 600/m3/day of
desalination plant capacity. At these rates,
100 150 200 250 300 350 400
Plant capacity (’000 m3/day)
Desalination plants in Australia
Desalination plants in Israel, Algeria and Spain
Desalination plants in South Africa
a 100 Mℓ/day seawater desalination plant
would cost in the order of R1 290 million,
which is comparable to the values derived
from the methodologies found in the
literature.
METHODOLOGY
Integration of seawater desalination
into the WCWSS models
In order to model the integration of sea-
water desalination into the WCWSS for
the purpose of this research, an approach
was required in which the desalination
plant could be modelled as an integral
component of the existing system. The
WRYM and WRPM were selected as the
most appropriate tools to achieve this. Given
the long history involved in the develop-
ment of the existing WCWSS models in
WRYM and WRPM, it seemed logical to
use them as the base for the modelling in
this research. The latest versions of the
WRYM and WRPM models of the WCWSS
were obtained from Aurecon with the
permission of the Department of Water and
Sanitation (DWS)2.
Operating scenarios
After studying the existing operational
philosophy of the WCWSS, four possible
scenarios were developed for the implemen­
tation and integration of a seawater
­desalination plant:
■■ Scenario A: Base scenario with current
system and no desalination plant, rep-
resenting the status quo or base case for
comparison purposes.
■■ Scenario B: Desalination plant used as
a base load supply, always operational
Journal of the South African Institution of Civil Engineering • Volume 59 Number 1 March 2017
regardless of the conditions in the rest of
the supply system.
■■ Scenario C: Desalination plant used as
a base load supply, but only if the major
dams are not spilling.
■■ Scenario D: Seawater desalination
plant used as an emergency supply, only
operational when the dams reach a cer-
tain threshold. Considering the penalty
structure zones of the main reservoirs,
four threshold or “trigger” levels were
selected to give a representative spread.
Based on the level-capacity curves of
the major reservoirs, the four trigger
450 levels were estimated to be equivalent to
90 percent, 70 percent, 45 percent and
15 percent of the capacity of the main
system reservoirs.
For each of the desalination operating sce-
narios B to D, three different desalination
plant capacities were tested, based on infor-
mation extracted from the latest WCWSS
SSC Progress Report (Department of Water
Affairs 2013a:13), i.e. 150 Mℓ/day (54.8 mil-
lion m3/a), 300 Mℓ/day (109.6 million m3/a)
and 450 Mℓ/day (164.4 million m3/a).
Water Resources Yield Model
(WRYM) setup
A number of options were considered and
tested for modelling a seawater desalination
plant in the WRYM and WRPM, including
the use of “artificial natural runoff” files or
a specified “inflow channel”, as defined by
the different models. These options would
entail the creation of an inflow sequence to
represent the seawater desalination plant,
with uniform flows in each month equal
to the seawater desalination plant capacity.
Creating the requisite summary of statisti-
cal parameters for the inflow file for the
purposes of stochastic analyses, however,
proved challenging and, as a result, a simpler
approach was adopted in which a “multi-
purpose min-max channel” was used as
an inflow channel from a zero node. The
capacity of the channel was constrained
to the desalination plant capacity and an
appropriate penalty was assigned to the
channel in order to model all the selected
operating scenarios.
Historic analyses were run for a total of
77 years (1928 to 2004) for a range of target
drafts in order to determine the historic firm
yield, and long-term stochastic analyses were
run for a total of 51 sequences to calculate
the long-term assurance of supply. The
number of sequences were selected based on
the suggestion of Basson et al (1994:34) that
reasonable results in a long-term analysis
can be obtained with at least 40 stochastic
sequences, as well as consideration of other
WCWSS planning studies.
13
 Short-term stochastic analyses were
Growth in water demands in the WRPM
also performed in the WRYM in order to 1 000
create short-term yield curves for use in the
WRPM. For short-term analyses, Basson et 900
al (1994:34) suggest that at least five times
800

Water demands (million m3/annum)

the number of sequences used in a long-term
analysis are required, hence at least 255 in 700
this case. The most recent WRPM studies
600
carried out by Aurecon use 401 sequences.
Given that the computational time is 500
much shorter for a short-term analysis,
401 sequences seemed reasonable, and were 400
therefore used.
300

Water Resources Planning Model 200

(WRPM) setup
The WRPM setup was modified in the same 100

way as for the WRYM. The base model setup 0

had only ten years of growth information, 2012 2014 2017 2020 2023 2025 2028 2031 2034
starting in 2013, for the 12 master control
channels. The growth factors were extended UBergFarmDam only demand UBergIB demand
to cover a period of 20 years, based on the UBergFarmDam only return UBergIB return
growth calculation spreadsheets prepared by RSI IB demand UBergFarmDam plus riperian demand
Aurecon as part of a recent DWS study into Total demand UBergFarmDam plus riperian return
operating rules for the WCWSS. The growth MetroPlus demand
scenario assumes an increase in urban
demands of 3 percent per annum, and that Figure 2 P
 rojected growth in water demands for the WCWSS as used in the WRPM
only 80 percent of the anticipated water con-
servation and demand management savings
Increase in historic firm yield for all scenarios and capacities
would be achieved. The projected demands 180
Increase in historic firm yield (million m3/annum)

for each master control channel are shown

in Figure 2. 160
The WRPM was run for a period of 140
20 years starting in 2013. A total of 401
stochastic sequences were analysed in order 120
to match the number of sequences used in
100
deriving the short-term yield curves.
80
Cost estimation and calculation
60
of URVs
In addition to comparing the yields and sup- 40
ply from the WCWSS with the introduction
of seawater desalination, it was considered 20

worthwhile to compare costs of the selected 0

operating scenarios. Based on a review of the
available costing methodologies, the South
African approach followed by Du Plessis
et al (2006) was selected as most fitting,
given that it is a local methodology, uses
membrane area rather than plant capacity
alone to estimate capital costs, is simple to
understand and apply, and for the test plant
of 100 Mℓ/day compares well in terms of
calculated costs to the methods considered.
The methodology of Du Plessis et al
(2006) can be used for estimating capital and
operating costs, but the methodology as is
only gives unit costs of water based on aver-
age plant capacity. In order to provide costs
which can be used to compare the operating
scenarios, it was considered more appropri-
ate to calculate URVs. The URV approach, as
presented by Hoffman and Du Plessis (2008),
40 60 80 100 120 140 160 180
Desalination plant capacity (million m3/a)
Scenario B Scenario D – 90% Scenario D – 45%
Scenario C Scenario D – 70% Scenario D – 15%
Figure 3 I ncrease in historic firm yield for increasing seawater desalination capacities
was adapted by Van Niekerk and Du Plessis ■■ Capital costs were determined based
(2013a) by using the actual volume of water on the desalination plant capacity and
supplied based on a stochastic analysis in the escalated to the start date of the analysis.
WRPM to calculate the URVs of inter-basin For the same desalination plant capacity,
transfers. This approach was applied to the the capital costs for all scenarios were
WCWSS by using the modelled annual vol- the same.
umes of water supplied from the desalination ■■ The total energy consumption of the plant
plant, as extracted from the WRPM analyses, was calculated based on its design capa­
to calculate costs. The following adaptations city, and then these costs were factored to
were made to the costing approach of Du calculate the annual energy costs based on
Plessis et al (2006) in order to provide URVs the annual volume of water supplied, as
for the modelled scenarios: derived from the WRPM analyses.

14 Journal of the South African Institution of Civil Engineering • Volume 59 Number 1 March 2017
 consumption, chemical costs, maintenance
Increase in historic firm yield for different storage trigger levels
180 costs and membrane lifespan.
Increase in historic firm yield (million m3/annum)

Note that all URVs calculated as part of

160 this research were for the desalination plant
140
and its associated infrastructure only, and
y = 63.806ln(x) + 159.35 not for the WCWSS as a whole.
120 R 2 = 0.996

100
PRESENTATION AND
y = 47.038ln(x) + 106.17
80 R 2 = 0.9981
DISCUSSION OF RESULTS
60
System yield
40 y = 25.023ln(x) + 52.378
R 2 = 0.9789 Base scenario
20 For Scenario A (the base scenario with
0
no desalination plant) the historic firm
0 10 20 30 40 50 60 70 80 90 100 yield of the system was calculated as
Desalination trigger level (percentage of total system storage) 530 ­million m3/a. Based on the stochastic
450 Mℓ/day capacity 300 Mℓ/day capacity 150 Mℓ/day capacity analyses, this corresponds to an assurance
of supply of approximately 70 percent, or
Figure 4 I ncrease in historic firm yield versus reservoir trigger level 1 in 215 years. The 1 in 50, 1 in 100 and 1 in
200 year yields of the system were calculated
as 580 million m3/a, 553 million m3/a and
Increase in historic firm yield for different storage trigger levels
180 532 million m3/a respectively. The results for
Increase in 1 in 50 year yield (million m3/annum)

Scenarios B to D were compared to the base

160 scenario results in order to determine the
140
increase in yield resulting from the addition
y = 63.767ln(x) + 163.83
R 2 = 0.9833 of a desalination plant to the system.
120
Historic firm yield
100
y = 41.439ln(x) + 108.38 Figure 3 shows the increase in the historic
80 R 2 = 0.9875 firm yield compared to the base scenario for
all the seawater desalination plant operating
60
scenarios that were analysed.
40 y = 22.665ln(x) + 56.133 For Scenario B, with the desalination
R 2 = 0.944
plant operational 100 percent of the time, the
20 increase in historic firm yield is very close to
0
the desalination plant capacity (as expected)
0 10 20 30 40 50 60 70 80 90 100 for all three capacities. For Scenario C, with
Desalination plant trigger level (percentage of total system storage) the desalination plant operational only when
450 Mℓ/day capacity 300 Mℓ/day capacity 150 Mℓ/day capacity the dams are not spilling, the results are
almost identical, which suggests that reduc-
Figure 5 I ncrease in 1 in 50 year yield versus reservoir trigger level ing the desalination plant output when the
dams are spilling has little impact on the
■■ It was assumed that the membranes (r) of 8 percent. Similarly, the NPV of the system yield. Considering Scenario D, as the
would be replaced every six years for water supplied from the desalination plant reservoir trigger level at which the seawater
a plant operating at full capacity. The (Wn) was determined, and hence, the URV desalination plant kicks in is lowered, the
membrane life was increased to up to 12 (in R/‌m3) calculated for each scenario based increase in historic firm yield decreases sub-
years, depending on the actual annual on Equation 1. stantially. This trend is more clearly viewed
output from the plant as a percentage of in a plot of the increase in historic firm yield
its capacity. ∑(Cn) of the system against the desalination plant
■■ Maintenance and labour costs were cal- trigger level, as presented in Figure 4.
(1 + r)n
culated as a function of the capital costs, URV = (1) The results show a clear logarithmic
and were adjusted depending on the ∑(Wn) pattern. Logarithmic trend lines plotted for
annual output of the plant. (1 + r)n each desalination plant capacity showed a
■■ Chemical costs were calculated by mul- good correlation with the modelled data.
tiplying a specific cost of chemicals by As in the approach of Van Niekerk and Du The coefficients of the logarithmic equations
the actual desalination plant supply per Plessis (2013a), cognisance was taken of the for each curve appeared to be related to the
annum. stochastic variation in the supply and the desalination plant capacity, and were there-
The capital, operating and maintenance resulting stochastic variation of the URVs. fore normalised based on the plant capacity
costs were summed per annum (Cn), and The cost components which are a function and averaged to provide the generic equation
the NPV determined over the analysis of the actual desalination plant output, and as shown in Equation 2. This equation could
period (n) of 20 years for a discount rate hence vary stochastically, include the energy be used to estimate the increase in historic

Journal of the South African Institution of Civil Engineering • Volume 59 Number 1 March 2017 15
firm yield in million m3/a (∆HFY) of the
Annual supply from a 150 Mℓ/day desalination plant as a percentage of capacity
WCWSS for any desalination plant capac- 100

Annual supply from desalination plant as a percentage of capacity

ity and reservoir trigger level. The average
annual desalination plant capacity (Qa) is 90
expressed in million m3/a, and the desalina-
tion plant trigger level (Tr) as a percentage of 80
the system storage in the main reservoirs.
70

∆HFY = 0.425QalnTr + 0.965Qa(2)

60

The logarithmic shape suggests that 50

increasing the dam trigger level from
say 20 percent to 30 percent, will have a 40
significantly greater impact on the historic
firm yield than increasing the trigger level 30
from say 80 percent to 90 percent. It also
shows that there is no “turning point” or 20

optimal trigger level. In other words, the

10
maximum increase in historic firm yield
is achieved when the desalination plant is 0
always operational. 2013 2015 2017 2019 2021 2023 2025 2027 2029 2031 2033
Year
Stochastic results Scenario B Scenario D (90%) Scenario D (45%) Water balance
Figure 5 shows the increase in 1 in 50 year Scenario C Scenario D (70%) Scenario D (15%)
yield from the base scenario, based on the
results from the stochastic analyses. Figure 6 A
 nnual supply from a 150 Mℓ/day seawater desalination plant as a percentage of capacity
The trends across the trigger levels and
capacities are very similar to the historic
Annual supply from a 150 Mℓ/day desalination plant as a percentage of capacity
firm yield results, showing a clear logarith- 100
Annual supply from desalination plant as a percentage of capacity

mic pattern, and logarithmic trend lines

fitted to the data show a good correlation. As 90
for the historic firm yield results, the loga-
rithmic equation coefficients were normal- 80
ised based on the desalination plant capacity,
70
and averaged, giving the generic Equation 3
for calculating the 1 in 50 year yield in mil-
60
lion m3/a (ΔSY) for any desalination plant
capacity and trigger level. 50

∆SY50 = 0.393QalnTr + 1.003Qa(3) 40

Comparable plots prepared for the 1 in 100 30

year and 1 in 200 year yields showed similar
trends, suggesting that the introduction 20

of the desalination plant merely shifts the

10
yield-reliability curve up without chang-
ing its shape. A summary of the equations 0
derived for estimating the increase in yield 2013 2015 2017 2019 2021 2023 2025 2027 2029 2031 2033
for the three return periods is presented in Year
Table 1. Scenario C 95th percentile Scenario D (70%) 95th percentile
Scenario C 50 th percentile Scenario D (70%) 50 th percentile
Table 1 E quations derived for estimating the Scenario 5th percentile Scenario D (70%) 5th percentile
increase in yield of the WCWSS for
any desalination plant capacity and Figure 7 S tochastic variation in annual supply from a 150 Mℓ/day seawater desalination plant
reservoir trigger level
Return
Seawater desalination plant supply analyses and are based on the median annual
Equation for estimating supply of the 401 stochastic sequences that
period of
increase in yield
yield (years) Median supply based on WRPM analyses were analysed.
1 in 50 ∆SY50 = 0.393Qa lnTr + 1.003Qa Figure 6 illustrates the supply from the sea- For Scenario B (as expected) the sea-
water desalination plant as a percentage of its water desalination plant would always be
1 in 100 ∆SY100 = 0.384Qa lnTr + 0.996Qa capacity in each year (2013 to 2032), for all 100 percent operational. For Scenario C
scenarios, based on a 150 Mℓ/day plant. The the seawater desalination plant would start
1 in 200 ∆SY200 = 0.387Qa lnTr + 0948Qa
values shown were derived from the WRPM out at 90 percent of its capacity, increasing

16 Journal of the South African Institution of Civil Engineering • Volume 59 Number 1 March 2017
 median, decreasing to zero by the end of the
Annual supply from desalination plant as a percentage of capacity based on traditional approach
100 analysis period (2028), when even the wetter
Annual supply from desalination plant as a percentage of capacity

sequences required the desalination plant to

90 be fully operational. For Scenario D, with a
70 percent trigger level, the stochastic range
80 is much greater, varying between -25% and
+50% from the median, and hence making
70
the choice of what values to use as input into
the URV calculations more critical.
60

50 Comparison with traditional approach

The research of Van Niekerk and Du Plessis
40 (2013b:543) showed that the annual supply
from a new water supply source (in their case
30 an inter-basin transfer scheme) is generally
estimated by calculating the deficit between
20 the projected annual demands of the system
and the existing system capacity, taking any
10
limitations in the capacity of the new water
0
supply source into consideration. Applying
2013 2015 2017 2019 2021 2023 2025 2027 2029 2031 2033 this principle, using the demand projections
Year as presented in Figure 2 and the 1 in 50 year
150 Mℓ/day plant 300 Mℓ/day plant 450 Mℓ/day plant yield of 580 million m3/a, the annual system
deficit was calculated. Assuming that a new
Figure 8 A
 nnual supply from a seawater desalination plant based on the traditional approach desalination plant would have to meet this
deficit, for the selected capacities of 150 Mℓ/
day, 300 Mℓ/day or 450 Mℓ/day, the annual
Water balance of WCWSS based on Scenario B
1 100 supply as a percentage of the desalination
WCWSS demand and supply (million m3/annum)

plant capacity was calculated, as shown in

Figure 8.
1 000
Based on the traditional approach, a
150 Mℓ/day seawater desalination plant
900
would operate at 18 percent of its capacity
in 2013, increasing to 100 percent by 2018.
800
Comparing these values to the WRPM
results in Figure 6, the traditional approach
700 and the WRPM approach provide vastly
different values in terms of supply from the
600 desalination plant, particularly for the lower
reservoir trigger levels. This will impact on
500 the calculated URVs.

400
2012 2014 2016 2018 2020 2022 2024 2026 2028 2030 2032
Year
1 in 50 year yield with 450 Mℓ/day desalination plant Projected demands
1 in 50 year yield with 300 Mℓ/day desalination plant 1 in 50 year current base yield
1 in 50 year yield with 150 Mℓ/day desalination plant
Figure 9 P
 rojected demand and supply of WCWSS with and without a seawater desalination plant
to 100 percent by 2015. Lowering the trig- Stochastic variability in supply
ger level (Scenario D), lowers the output The results presented in Figure 6 are based on
from the seawater desalination plant as a median values of the 401 stochastic sequences
percentage of its capacity. For the lowest analysed, and were selected as representa-
trigger levels of 45 percent and 15 percent, tive values for calculating URVs. In order to
the seawater desalination plant would not illustrate the possible stochastic variability in
be operational for the entire analysis period. supply from the desalination plant, the 5th and
Similar plots prepared for 300 Mℓ/day and 95th percentile annual values were plotted for
450 Mℓ/day plants show that, as the seawater Scenario C and for Scenario D with a 70 per-
desalination plant capacity increases, the cent trigger level, as shown in Figure 7.
supply from the plant as a percentage of its For Scenario C the maximum stochastic
capacity decreases. range varies between -28% and +9% from the
Planning for future interventions
2034
Traditional approach
The typical approach that would be followed
in determining when a future intervention is
required in a water supply system would be a
water balance of system demands and supply.
The water demands that were projected as
part of this study, based on the recent work
by Aurecon (presented in Figure 2), were
plotted along with the calculated 1 in 50 year
base yield of 580 million m3/a (Scenario A)
and the 1 in 50 year yields with a seawater
desalination plant as derived from the
WRYM analysis (Scenario B), as shown in
Figure 9. Based on this approach, a 150 Mℓ/
day seawater desalination plant would meet
the system demands until 2017, a 300 Mℓ/
day plant would meet the demands until
2021 and a 450 Mℓ/day plant would meet the
requirements until 2024.

Journal of the South African Institution of Civil Engineering • Volume 59 Number 1 March 2017 17
 Sub-system curtailment – Ex+BRD+Sup – Decision month (2)
5.0
4.8
4.6
4.4
4.2
4.0 1 in 200 year curtailment level
3.8 1 in 100 year curtailment level
3.6
3.4
3.2
Curtailment level

3.0
2.8
2.6
2.4
2.2
2.0
1.8
1.6
1.4
1.2
1.0
0.8
0.6 1 in 20 year curtailment level
0.4
0.2
0
2014 2026
2013/12

2014/12

2015/12

2016/12

2017/12

2018/12

2019/12

2020/11

2021/11

2022/11

2023/11

2024/11

2025/11

2026/11

2027/11

2028/11

2029/11

2030/11

2031/11

2032/11
Year

Figure 10 S ystem curtailment for base scenario with no seawater desalination plant (Scenario A)

Table 2 Level of restrictions used in WRPM analyses Figure 10 shows a box-and-whisker plot
of the subsystem curtailment for the base
Restricted water demand as a
Acceptable frequency percentage of normal demand scenario (Scenario A) from 2013 to 2032 as
Level of curtailment
of restrictions derived from the WRPM analysis. The criti-
Domestic Agricultural
cal lines or whiskers which correspond to 1
0 1 in 10 years 100% 100% in 20 year, 1 in 100 year and 1 in 200 year
curtailment levels are labelled.
1 1 in 20 years 93% 75%
Curtailment Level 1 would be reached
2 1 in 100 years 85% 50% (approximately) at a frequency of 1 in 20
years by 2018. Curtailment Level 2 would
3 1 in 200 years 71% 20%
be reached at a frequency of 1 in 100 years
by 2016, and curtailment Level 3 would be
Table 3 Year at which curtailment level would be reached for all scenarios reached at an assurance of 1 in 200 years by
2016. These critical points are marked with
1 in 20 1 in 100 1 in 200 1 in 20 1 in 100 1 in 200 1 in 20 1 in 100 1 in 200
year year year year year year year year year circles. Considering these three trigger years,
Scenario
the earliest that a new water supply interven-
150 Mℓ/day plant 300 Mℓ/day plant 450 Mℓ/day plant
tion would be required for the WCWSS in
B 2021 2019 2019 2024 2021 2021 2027 2025 2023 order to ensure that the frequency of curtail-
C 2021 2019 2019 2024 2021 2021 2027 2025 2023 ments stays within acceptable limits would
be 2016.
D (90%) 2021 2019 2019 2027 2021 2022 2027 2024 2023

D (70%) 2021 2018 2019 2023 2021 2019 2027 2023 2023 Seawater desalination plant
D (45%) 2020 2017 2016 2020 2018 2018 2021 2018 2018
scenarios in the WRPM
Plots similar to Figure 10, showing curtail-
D (15%) 2019 2017 2016 2020 2018 2018 2020 2018 2018 ment levels, were extracted from the WRPM
No desalination plant for the various seawater desalination operat-
ing scenarios and capacities. A summary of
A 2018 2016 2016 2018 2016 2016 2018 2016 2016
the critical points at which the acceptable
frequency of curtailments are exceeded are
Base scenario using the WRPM WRPM analyses. As shown in Table 2, Level provided in Table 3. The earliest dates at
In the WCWSS analysis in the WRPM and 1 curtailments cannot be applied more than which an intervention is required in each
part of the recent study on the development once in every 20 years, Level 2 curtailments case are highlighted in bold.
of integrated annual and real time operating no more than once in every 100 years and Considering Scenario B, the addition
rules for the WCWSS, revised water restric- Level 3 curtailments no more than once of a 150 Mℓ/day plant to the system would
tion levels were developed. The scenario in every 200 years. Stricter curtailments delay the date at which a new intervention is
which Aurecon found to be the most realistic are applied to agricultural demands than required to 2019. A plant capacity of 300 Mℓ/
is provided in Table 2, and was used in the ­domestic demands. day would provide a further two years to

18 Journal of the South African Institution of Civil Engineering • Volume 59 Number 1 March 2017
Table 4 U
 RVs for all scenarios and seawater desalination plant capacities derived from WRPM analyses
150 Mℓ/day plant
Scenario
5th 50 th
percentile percentile
B R10.96 R10.96
C R10.96 R10.96
D (90%) R10.96 R12.21
D (70%) R11.16 R15.49
D (45%) R37.06 R99.00*
D (15%) R54.68 R99.00*
* URVs in excess of R99/m3 were capped at R99/m3 as a representative upper limit.
2021, and a capacity of 450 Mℓ/day, an addi-
tional two years to 2023. For Scenario C and
Scenario D with a 90 percent trigger level,
the results are identical to those of Scenario
B, suggesting that the decision regarding
whether to operate the seawater desalination
plant 100 percent of the time or only when
the dams are not spilling would not be influ-
enced by curtailment requirements.
As the trigger levels decrease, the benefit
of the seawater desalination plant in terms of
delaying the requirement for a new scheme
is reduced. For the lowest trigger levels the
seawater desalination plant provides almost
no benefit in terms of reducing curtailments,
suggesting that there is very little benefit in
using the desalination plant as an emergency
supply source.
Unlimited desalination plant capacity
The results are derived from assumed sea-
water desalination plant capacities of 150,
300 and 450 Mℓ/day. The reverse of this
approach would be to ask: “What desalina-
tion plant capacity is required in order to
meet future demands for the next 10 or 20
years?” Based on the traditional approach,
a 400 Mℓ/day seawater desalination plant
would meet the system requirements up to
2023, and a 1 000 Mℓ/day plant would meet
the requirements up to 2033. The WRPM
results confirm that, with a 450 Mℓ/day
desalination plant, curtailments would be
kept within acceptable limits until 2023.
However, from the WRPM results it is not
possible to confirm what capacity is required
to meet the demands beyond this point.
The WRPM analyses were repeated for
Scenarios C and D (with a 70 percent trigger
level only) with a seawater desalination plant
of unlimited capacity. The monthly supply
from the seawater desalination plant over
the analysis period was determined, based
on the median values of the 401 stochastic
sequences. The maximum monthly supply
for Scenario C would be 2 100 Mℓ/day in
2023, increasing to 2 900 Mℓ/day in 2033.
Journal of the South African Institution of Civil Engineering • Volume 59 Number 1 March 2017
300 Mℓ/day plant
95th 5th 50 th
percentile percentile percentile
R10.96 R10.38 R10.38
R11.75 R10.38 R10.53
R14.56 R10.38 R11.71
R22.77 R10.69 R14.79
R99.00* R28.43 R99.00*
R99.00* R36.71 R99.00*
For Scenario D the plant would have to sup-
ply up to 1 500 Mℓ/day in 2023, increasing to
2 100 Mℓ/day by 2033.
The problem with adopting this approach
is that it does not allow for any failures over
the analysis period, i.e. for Scenario D with a
70 percent trigger level, a 1 500 Mℓ/day plant
would meet the demands until 2023 with a
100 percent assurance of supply. This differs
from the normal risk allowance that would
be applied in planning for a water supply
system in South Africa.
Comparison of costs
Traditional approach to calculating URVs
The research of Van Niekerk and Du Plessis
(2013a:551) showed that the traditional
approach to calculating URVs uses the water
supply from a scheme calculated based on
a water balance between annual demands
and a 1 in 50 year system yield. As a start-
ing point, this traditional approach was
applied for calculating the URV of the
seawater desalination plant. For a discount
rate of 8 percent, the traditional approach
yielded URVs of R11.35/m3 for a 150 Mℓ/
day seawater desalination plant, R11.46/m3
for a 300 Mℓ/day seawater desalination plant
and R11.48/m3 for a 450 Mℓ/day seawater
desalination plant.
WRPM-based approach to
calculating URVs
In the WRPM-based approach, as applied
for this research, capital and operating costs
were calculated for each scenario using the
annual supply from the seawater desalination
plant as derived from the WRPM analyses.
Using these costs and the associated sea-
water desalination plant supply, URVs were
calculated for each scenario and seawater
desalination plant capacity as presented in
Table 4. The values highlighted in bold are
the median values (50th percentile) which
were considered to be the most reasonable
for the purposes of comparison between the
450 Mℓ/day plant
95th 5th 50 th 95th
percentile percentile percentile percentile
R10.38 R9.81 R9.81 R9.81
R11.38 R9.81 R10.08 R10.92
R13.48 R9.83 R11.11 R12.47
R19.90 R10.33 R13.71 R17.71
R99.00* R24.22 R99.00* R99.00*
R99.00* R31.35 R99.00* R99.00*
scenarios and with other possible interven-
tions. The 5th and 95th percentile URVs are
also shown in order to provide an indication
of the possible stochastic range of the costs.
Considering the median (50th percentile)
values in Table 4, it would appear that, for
all capacities, the URV increases as the trig-
ger level decreases. It therefore seems that
operating the seawater desalination plant as
an emergency type scheme, as exemplified by
the 45 percent and 15 percent trigger levels,
would be expensive. Although the annual
operating costs would be much lower for an
emergency scheme, the initial capital cost
would be the same regardless of the actual
seawater desalination plant output. The
results suggest that this initial capital cost
outweighs any reduction in operating costs
when the seawater desalination plant output
is reduced.
For a trigger level of 70 percent, the URV
for a 150 Mℓ/day plant is much more reason-
able at R15.49/m3, decreasing to R12.21/m3
for a 90 percent trigger level. For Scenario B
and Scenario C the desalination plant would
be fully operational once constructed, giving
a URV of R10.96/m3. Based on these values
it would appear that the lowest possible
URV occurs when the desalination plant is
100 percent operational.
Comparing the three capacities that were
analysed, the URVs decrease slightly with
an increase in seawater desalination plant
capacity. This suggests that it would be more
beneficial from an economic perspective to
construct a larger seawater desalination plant
now, rather than phasing it in smaller lanes.
Comparing the results from the tra-
ditional approach to the WRPM-based
approach, the results for a 150 Mℓ/day plant
are actually similar to Scenario B/C. This
suggests that, although the annual sup-
ply from the seawater desalination plant
differs vastly between the traditional and
WRPM approaches in the initial years,
both approaches suggest that the seawater
desalination plant would be 100 percent
19
operational in the later years, meaning
that the calculated URVs are similar. The
traditional approach, however, suggests an
increase in the URV with an increase in
capacity, which differs from the WRPM
approach. For the lower trigger level sce-
narios, the WRPM-based approach has much
higher URVs than the traditional approach.
Considering the 5th and 95th percentile
values, it appears that the possible stochastic
range increases with a decrease in the trigger
level and with an increase in capacity. Hence,
for a fairly predictable scenario in terms of
seawater desalination plant supply (when
it will be operational most of the time), the
stochastic variability in stream flows in
the existing system would have little effect
on the URVs and would not be too far off
from URVs calculated using the traditional
approach. However, for a less predictable
operating scenario, using the seawater desali-
nation plant as an emergency type scheme,
the stochastic variability in the URV would
be more noticeable, and hence the selection
of an appropriate percentile for reporting on
the costs would become more important.
CONCLUSIONS
Through the current WCWSS planning
processes, seawater desalination has been
earmarked as a potential future supply
source for the area. In order to maximise the
benefits and minimise the costs of a seawater
desalination plant, it needs to be considered
as an integral part of the current system. In
order to undertake integrated system analy-
sis, the WRYM and WRPM were selected
as the most appropriate tools. Short-term
and long-term analyses were completed for
a variety of possible seawater desalination
plant operating scenarios and capacities in
order to determine the increase in system
yield and the annual supply from the sea-
water desalination plant. First-order capital
and operating costs were estimated using the
South African methodology of Du Plessis et
al (2006). Combined with the annual supply
values from the WRPM, these costs were
used to calculate and compare the URVs of
the desalinated water.
The following conclusions can be drawn
from the results of the modelling and cost
analyses:
■■ It is possible to model a desalination plant
as an integral part of a surface water sup-
ply system. With the appropriate costing
tools, a similar analysis approach could
be applied to other non-conventional
resources, such as water reuse.
■■ The increase in system yield as a result
of the desalination plant appears to be
logarithmically related to the reservoir
20 Journal of the South African Institution of Civil Engineering • Volume 59 Number 1 March 2017
trigger level. Generic equations were
developed which could be used to esti-
mate the increase in historic firm yield
and 1 in 50 year yield of the WCWSS for
any seawater desalination plant capacity
and reservoir trigger level.
■■ The maximum increase in yield occurs
when the seawater desalination plant is
used as a base supply, operational all the
time. The increase in yield decreases with
a decrease in the reservoir trigger level,
meaning that there is little benefit in
using the seawater desalination plant as
an emergency supply source.
■■ With a 150 Mℓ/day desalination plant,
curtailment requirements would be
kept within acceptable levels until 2019,
with a 300 Mℓ/day plant until 2021 and
a with a 450 Mℓ/day plant until 2023,
for a desalination plant which is 100
percent operational. Using the plant as an
emergency supply source would add little
benefit in terms of limiting curtailment
requirements.
■■ URVs for the seawater desalination plant
decrease as the percentage of time for
which the seawater desalination plant is
operational increases. This means that
the lowest possible cost per cubic metre
of water supplied is when the seawater
desalination plant is operational 100 per-
cent of the time (R10.96/m3 for a 150 Mℓ/
day plant).
■■ It would also appear that the URVs
decrease with an increase in seawater
desalination plant capacity, meaning
that constructing a larger seawater
desalination plant upfront may be more
cost-effective. However, it would have to
be constructed in small enough lanes to
enable restriction of the output when the
total supply is not required, and cogni-
sance would have to be taken of other rel-
evant factors, such as availability of other
schemes already in place and growth in
water demands.
■■ The lower the reservoir trigger level, the
larger the stochastic variation in the sup-
ply from the seawater desalination plant,
and hence the larger the variation in the
URVs. Hence, for a seawater desalination
plant operational as a base supply source,
undertaking stochastic modelling to
calculate URVs is less important than for
an emergency supply source.
LIMITATIONS AND
FURTHER RESEARCH
In concluding, the following limitations and
areas for further research should be noted:
■■ Of the available non-conventional supply
sources, only seawater desalination was
considered, and the model was applied to
the specific case of the WCWSS, hence
the results are mostly case-specific. As
part of further research, it would be
worthwhile to repeat the methodology
and analysis procedure for other large
water supply schemes in South Africa,
and for other sources such as water reuse.
■■ The main limitation of this research is
that only the cost of the desalination
scheme was considered in the analysis.
As part of future research, it may be
worthwhile to consider developing a cost-
ing model for the WCWSS as a whole and
attempting to optimise the overall system
cost with the implementation of a desali-
nation plant. Under such an approach,
reducing the desalination plant output
may prove to be more cost-effective.
Desalination technology has advanced
rapidly in the last few years, and as a
result capital and operating costs today
are likely to be lower than those estimat-
ed using the methodology of Du Plessis et
al (2006) used in this research. Updating
of the costing methodology could form
part of further research; however, it is
expected that adjustments to the costing
would not affect the applicability of the
analysis methodology and relevance of
the conclusions.
NOTES
1. The following exchange rates were issued in the
costing analyses and comparison of known costs
(South African Reserve Bank 2014):
■ 1 Australian Dollar (AUS$) is equivalent to R9.80
■ 1 United States Dollar (US$) is equivalent to
R10.70
■ 1 Euro (€) is equivalent to R14.70
2. DWS (Department of Water and Sanitation) was
previously the Department of Water Affairs and
Forestry (prior to 2009) and subsequently the
Department of Water Affairs (from 2009 to 2014).
In this paper, the department is referred to by
its current designation (DWS), apart from in the
references, where the name at the time of publication
of the referenced document is used.
REFERENCES
Basson, M S, Allen, R B, Pegram, G G S & Van Rooyen,
G A 1994. Probabilistic Management of Water
Resource and Hydropower Systems. Littleton, CO:
Water Resources Publications.
Department of Water Affairs 2007. Western Cape Water
Supply System Reconciliation Strategy. Cape Town:
Department of Water Affairs, South Africa.
Department of Water Affairs 2010. Integrated Water
Resource Planning for South Africa: A Situation
Analysis 2010. Pretoria: Department of Water
Affairs, South Africa.
Department of Water Affairs 2011. Western Cape Water
Supply System Reconciliation Strategy, Progress
Report of the Strategy Steering Committee. Cape
Town: Department of Water Affairs, South Africa.
Department of Water Affairs 2013a. Support to the
Continuation of the Water Reconciliation Strategy
for the Western Cape Water Supply System: Status
Report October 2013. Cape Town: Department of
Water Affairs, South Africa.
Department of Water Affairs 2013b. National Water
Resource Strategy: Water for an Equitable and
Sustainable Future. Pretoria: Department of Water
Affairs, South Africa.
Department of Water Affairs and Forestry 2008.
Water for Growth and Development in South Africa.
Pretoria: Department of Water Affairs and Forestry,
South Africa.
Du Plessis, J A, Burger, A J, Swartz, C D & Musee,
N 2006. A Desalination Guide for South African
Municipal Engineers. Pretoria: Department of Water
Affairs, South Africa.
Journal of the South African Institution of Civil Engineering • Volume 59 Number 1 March 2017
Ghaffour, N, Missimer, T M & Amy, G I 2013.
Technical review of the economics of water
desalination: Current and future challenges for
better water supply sustainability. Desalination, 309:
197–207.
Hoffman, J J & Du Plessis, J A 2008. Water demand
management: An economic viable option. WISA,
147–163.
Karagiannis, I C & Soldatos, P G 2008. Water
desalination cost literature: Review and assessment.
Desalination, 223: 448–456.
Mallory, S J L, Ballim, F, Pashkin, J & Ntuli, C 2013.
Introducing desalination plants into system
operating rules for bulk water supply systems.
Unpublished paper delivered at the 16th South
African National Hydrology Symposium, 1 October,
Pretoria.
Onishi, N 2010. Arid Australia sips seawater, but
at a cost. Available at: http://www.nytimes.
com/2010/07/11/world/asia/11water.html?_r=0
[accessed on 23 March 2012].
Shand, M & Sparks, A 2004. Optimisation of system
operating rules for the Western Cape Water Supply
System. s.l.:s.n.
South African Reserve Bank 2014. Selected
historical rates. Available at: https://www.
resbank.co.za/REsearch/Rates/Pages/
SelectedHistoricalExchangeAndInterestRates.aspx
[accessed on 11 September 2014].
Van Niekerk, P H & Du Plessis, J A 2013a. Hydrologic-
economic appraisal of lifecycle costs of inter-basin
water transfer projects. Water SA, 39(4): 539–548.
Van Niekerk, P H & Du Plessis, J A 2013b. Unit
Reference Value: Application in appraising inter-
basin water transfer projects. Water SA, 39(4):
549–554.
Wittholz, M K, O’Neill, B K, Colby, C B & Lewis, D
2008. Estimating the cost of desalination plants
using a cost database. Desalination, 229(2008):
10–20.
21
TECHNICAL PAPER Numerical modelling
Journal of the South African
Institution of Civil Engineering of flexible pavement
incorporating cross‑anisotropic
ISSN 1021-2019
Vol 59 No 1, March 2017, Pages 22–27, Paper 1366

PROF JAMES MAINA (Pr Eng, MSAICE, FSAAE) is a

professional pavement engineer, full-time
material properties
professor of civil engineering at the University of
Pretoria, and currently on secondment as a
technical director of a QA/QC Project for roads
Part I: Surface circular loading
in the State of Qatar. He obtained his PhD from
Miyazaki University in Japan. His professional J W Maina, F Kawana, K Matsui
activities include QA/QC in road projects,
pavement materials, and the development of advanced numerical analysis
(modelling) tools for pavement engineering application. He also teaches
both under- and post-graduate classes at the University of Pretoria. Accurate numerical modelling of the behaviour of road pavement layers is an important
Contact details:
requirement for the design and evaluation of road pavements. This modelling includes the
Department of Civil Engineering prediction of pavement performance under the action of traffic loading and environmental
University of Pretoria factors. Depending on the complexity of the models, properties of pavement layers that may
Private Bag X20 be considered are wide-ranging – from linear or nonlinear elastic to cross-anisotropic through
Hatfield 0028
to linear visco-elasto-plastic. Some properties, such as cross-anisotropic, are not only related
South Africa
T: +27 12 420 6608
to placement and compaction of the pavement layers, but are also inherent to the materials
E: james.maina@up.ac.za used. Other properties, such as linear visco-elasto-plastic, are specific to asphalt concrete
and depend on the speed and magnitude of traffic loading, as well as the environment
PROF FUTOSHI KAWANA is an associate (temperature) in which the road is located. This paper presents basic theoretical derivation of
professor at the Tokyo University of Agriculture numerical modelling of a flexible pavement considering cross-anisotropic material properties
in Japan. His specialises in structural mechanics. (with isotropic properties as a special case). The solutions derived in this paper are based on
He received his doctorate in engineering from Hankel transformation of Navier’s equations. The accuracy and validity of the solutions are
the Science University of Tokyo in 2004. He is a
member of the Japan Society of Civil Engineers,
verified through comparisons with a proprietary finite element method (FEM) package. For
the Pavement Diagnosis Researchers Group this purpose, a pavement structure composed of five main layers constituted by isotropic and
(NPO) and the Japanese Society of Irrigation, cross-anisotropic (also known as transversely isotropic) material properties is analysed. In order
Drainage and Rural Engineering. to vary some of the layer properties with depth, the main layers were sub-layered, resulting in a
Contact details: 17-layer pavement system.
Department of Bioproduction and Environment Engineering
Faculty of Regional Environment Science
Tokyo University of Agriculture
background information a better understanding of the properties of
1-1-1 Sakuragaoka
Setagaya-ku
To support the movement of people and materials used in road constructions and
Tokyo, 156-8505 goods, as well as access to education and modelling of structural behaviour of pave-
Japan training, employment and health care, ments was deemed necessary.
T: +81 3 5477 2342 a sound road network – which in South The SAPDM, which uses multilayer linear
E: fk205262@nodai.ac.jp
Africa is most often designed and evaluated elastic (MLLE) theory to determine pave-
using the South African Pavement Design ment structural responses, was introduced
PROF KUNIHITO MATSUI is emeritus professor of
Method (SAPDM) – plays a key role in the in the late 1960s and early 1970s. Since its
civil and environmental engineering at Tokyo
Denki University in Japan. He obtained his PhD
socio-economic development of a country. introduction, developments in material char-
from the Department of Mechanics and South Africa, in the recent past, has expe- acterisations and numerical modelling have
Hydraulics at the University of Iowa, USA, in rienced considerable growth in both pas- taken place. In addition, advanced computer
1974. His areas of interest include structural senger and freight traffic volumes because software and hardware make it possible now
analysis in pavement structures, static and
of increased economic activities. The high to determine the stress/strain/displacement
dynamic back-calculation of pavement systems,
non-destructive testing, thermal analysis of pavement systems, parameter level of truck traffic is causing premature distributions of pavements under surface
identification, sensitivity analysis and structural optimisation. failures, mostly on provincial and municipal loading in a matter of seconds.
road networks. This paper presents the outcome of
Contact details:
Department of Civil and Environmental Engineering The South African National Roads the efforts to improve the standard pave-
Tokyo Denki University Agency Ltd (SANRAL) teamed up with the ment analysis for circular surface loading
Hatoyama roads industry, the Council for Scientific and by introducing capability to consider
Hiki Industrial Research (CSIR) and academia cross-anisotropic material properties (with
Satama, 350-0394
in an effort to improve the SAPDM and isotropic as a special case). This capability is
Japan
T: +81 492 96 5703/2549 the design of roads in South Africa. The not possible in the current SAPDM software.
E: matsui@g.dendai.ac.jp partnership called for skills development and After rigorous validation and verification,
the building of capacity to optimise design, software resulting from this development
Keywords: pavement, linear-elastic analysis, transversely isotropic, maintenance and repair strategies against has now become the analysis engine of the
cross‑anisotropy, isotropic, circular loading road (premature) failures. To achieve this, new SAPDM, and it will be used when the

22
Maina JW, Kawana F, Matsui K. Numerical modelling of flexible pavement incorporating cross-anisotropic material properties – Part I: Surface circular loading.
J. S. Afr. Inst. Civ. Eng. 2017:59(1), Art. #1366, 6 pages. http://dx.doi.org/10.17159/2309-8775/2017/v59n1a3
loading on the surface of the road pavement
is circular in shape.
motivation for incorporating
cross-anisotropic properties
Cross-anisotropy is the most prevalent but
less considered type of materials properties
in pavement layers. In this regard, the elastic
properties in the lateral and vertical direc-
tions are considered to be different. The road
construction process involves directional
rolling compaction of materials, which
invariably results in different mechanistic
properties in vertical and horizontal direc-
tions (Masad et al 2006). A cross-anisotropic
material has a vertical axis of symmetry
of rotation, i.e. the elastic properties are
equivalent in all directions perpendicular to
the axis of symmetry (horizontal or radial
direction). In general, these properties are
different from those in the direction parallel
to the axis (vertical direction).
In engineering mechanics, cross-­
anisotropy of an elastic material is defined
by five independent elastic constants – two
elastic moduli in vertical and horizontal
directions (Ev and Eh), two Poisson’s ratios
in vertical and horizontal directions (vvh and
vhh) and one shear modulus (Ghv) as pre-
sented by Love (1944). Several experiments
have also proven the existence of cross-
anisotropic properties in pavement layers,
and these properties have been measured in
the laboratory (Correia 1999; Adu-Osei et
al 2001; Tutumluer et al 2003; Masad et al
2006). Typically, the level of cross-anisotropy
is characterised by the ratio of the horizontal
to vertical modulus (Eh/Ev). In contrast,
isotropic materials have the same elastic
properties in both the vertical and horizontal
directions, which are described by three
independent elastic constants: E, v and G.
The importance of incorporating cross-
anisotropic material property in the pavement
analysis was demonstrated by Masad et al
(2006), who showed how the accuracy of calcu-
lated pavement displacements improved when
the aggregate base layer was considered to be
cross-anisotropic. In their study, Masad et al
(2006) used a Finite Element (FE) program to
calculate pavement surface displacements, and
compared their results with measured values
from 246 different sections in the Association
of American State Highway Officials (AASHO)
road test. The study revealed that the displace-
ments calculated using isotropic material
properties tended to be smaller than measured
displacements in the field. Further analysis
showed that the errors between measured and
calculated responses were minimised when
the elastic modulus in the horizontal direction
was assumed to be 30% of the elastic modulus
Journal of the South African Institution of Civil Engineering • Volume 59 Number 1 March 2017
P (0, 0, 0)
X
x 2a
θ
r
σz α
y
θ
τ rz r x
τ rz
y Point of interest
σθ σr
(X0 + rCos(θ + α), Y0 + rSin (θ + α), 0)
Z Y
Figure 1 Cylindrical coordinate referencing for pavement analysis
in the vertical direction (Eh/Ev = 0.3). Also, r or z face in the respective z or r direction,
the findings seem to be in agreement with as shown on Figure 1.
results reported in another study on the level of In this regard, the differential equilibrium
anisotropy (Eh/Ev) of unbound and chemically equations in cylindrical coordinates may be
stabilised aggregate systems (Salehi-Ashtiani expressed using modified Navier’s Equations
et al 2008). 1 and 2:
Other researchers such as Emeriault and
Chang (1997), and Masad et al (2004) have ∂σ r ∂τzr σr – σθ
+ + = 0(1)
also used the micromechanics theory to ∂r ∂z r
determine cross-anisotropic properties of
granular base materials. Adu-Osei et al (2001) ∂τzr ∂σz τzr
+ + = 0(2)
used the linear elastic theory coupled with ∂r ∂z r
system identification to develop a testing sys-
tem, in a triaxial test setup, that could identify Corresponding axisymmetric strains may be
the five parameters that represent cross- expressed in terms of axisymmetric displace-
anisotropic properties of unbound materials. ments as follows:
This test was developed in a collaborative
research between the universities of Illinois ∂u u ∂w ∂u ∂w
εr = , εθ = , εz = , γzr = + (3)
and Texas, whereas the field validation data ∂r r ∂z ∂z ∂r
was collected from a full-scale pavement test
study conducted at Georgia Tech (Georgia Generalised Hooke’s law
Institute of Technology). The validation of the In linear elasticity, if the stress is sufficiently
anisotropic modelling approach was accom- small, Hooke’s law is used to represent
plished by analysing conventional flexible the material behaviour and to relate the
pavement test sections, where the GT-PAVE unknown stresses and strains. The general
finite element program was used to predict equation for Hooke’s law is given below:
responses to loading in an unbound aggregate
εij = sijklσkl = ∑k=1∑l=1 sijklσkl(4)
3 3
base layer, and compared these predicted
responses to the measured values (Tutumluer
et al 2003). Wang and Liao (1999) listed and Where: i, j = 1, 2, 3
meticulously summarised the numerous stud-
ies for cross-anisotropic half-space subjected In this case sijkl is a fourth-rank tensor called
to different types of loading, such as line elastic compliance of the material. Each
loads, circular loads, parabolic loads, ring subscript of sijkl takes on the values from 1 to
loads, and varying rectangular loads. 3, giving a total of 34 = 81 independent com-
ponents in s. However, due to the symmetry
of both εij and σkl, the elastic compliance s
THEORETICAL DEVELOPMENT must satisfy the relation:
Linear elasticity sijkl = sjikl = sijlk = sjilk(5)
Because of azimuthal symmetry for circular
surface loading, responses present in axisym- It follows from Equation 5, therefore, that
metric elastic problems are displacements the generalised Hooke’s law presented in
u = u(r,z),w = w(r,z) in horizontal, r– and Equation 4 can be simplified to become:
vertical, z–directions respectively, normal
stresses σ r, σz and σ θ in the horizontal, r–and εi = sijσ j(6)
vertical, z– and circular, θ– directions
respectively, as well as shear stress τrz on the Where: i, j = 1,..,4
23
This relationship reduces the number of s Where: cij is a fourth-rank tensor called Substituting Equations 19–22 into Equations
components to 16, as seen in the following elastic stiffness of the material. 1 and 2, with some rearrangement, will result
linear relation between the pseudovector The stiffness and compliance matrices in in Equations 23 and 24:
forms of the strain and stress: Equation 10 are related in the following form:
∂2 1 ∂ 1 c44 ∂ 2
+ – + u
 ε1   s11
s12 s13 s14   σ 1  c= s–1(11) ∂r 2 r ∂r r2 c11 ∂z 2
  
ε2
 =
s21 s22 s23 s24 σ 2
 
  +
(c13 + c44) ∂2
w = 0 (23)
ε3 s31 s32 s33 s34 σ 3 Three components (c11, c12, c44) exist for c11 ∂rdz
  
 ε4   s41 s42 s43 s44   σ4    isotropic materials, whereas, in cross-ani-
(c13 + c44) 1 ∂ ∂u
 εr   s11 s12 s13 s14   σ r  sotropic materials, five unique components +
 ε  
s s s s
⇒  θ  =  21 22 23 24   θ (7)
σ   (c11, c12, c13, c33, c44) exist. c44 r
∂r ∂z
εz s31 s32 s33 s34 σz It should be noted that responses for a ∂2 1 c ∂2
∂
+ 33
  
 γ rz   s41 s42 s43 s44   τ rz   special cross-anisotropy case where Ev = Eh and
+ +
∂r 2 r ∂r c44 ∂z 2
w = 0 (24)

vvh = υhh exist and will be similar to the isotro-

The s matrix in this form is also symmetric. pic case presented by Maina and Matsui (2004). Derivation of displacement
It, therefore, contains only ten independent Equation 12 below shows the five compo- and stress solutions using
elements. The number of independent ele- nents of the elastic stiffness c and Equations Hankel transformation
ments is obtained by counting elements in the 13–17 define each of the five components: One convenient way of deriving displacement
upper right triangle of the matrix, including and stress solutions is to use Hankel trans-
the diagonal elements (i.e. 1 + 2 + 3 + 4 = 10).  c11 c12 c13  formation (Miyamoto 1977; Sneddon 1951).
Furthermore, if the material exhibits sym-  c12 c11 c13  In this regard, Hankel transformation of
c= (12)
metry in its elastic response, the number of c c c Equations 23 and 24 would yield, respectively:
 13 13 33 
independent components in the compliance s  c44 
matrix will be reduced even further. ∂2 ∂w̃
–ξ 2 + b ũ + aξ = 0 (25)
For example, in the simplest case of isotrop- Where: ∂z 2 ∂z
ic materials whose elastic moduli are the same
2
in all directions, only three unique components Eh(–Ev + Ehvhv ) ∂ũ ∂2
c11 = (13) cξ +b + –ξ 2 + d w̃ = 0 (26)
2
(s11, s12, s44) exist, as shown in Equation 8. In (1 + vhh)(–Ev + Evvhh + 2Ehvhv) ∂z ∂z 2
the elastic range, these three coefficients of
2
the ordinary model require three parameters Eh(Evvhh + Ehvhv ) Where:
c12 = – (14)
2
(elastic modulus E, Poisson’s ratio v and shear (1 + vhh)(–Ev + Evvhh + 2Ehvhv)
modulus G) for their definitions as follows: (c13 + c44) c44 (c13 + c44)
a= ,b= ,c= ,
EhEvvhv c11 c11 c44
c13 = – (15)
2
–Ev + Evvhh + 2Ehvhv c33
 s11 s12 s12  d= (27)
s s s  c44
s =  12 11 12 (8) Ev2(–1 + vhh)
s s s
 12 12 11  c33 = – (16) ∞
 s44  2
–Ev + Evvhh + 2Ehvhv ũ = ũ(ξ, z) = ∫ ru(r, z)J1(ξr)dr(28)
0
∞
Where: s 11 = 1/E, s12 = –v/E, s44 = 1/G, Ev w̃ = w̃(ξ, z) = ∫ rw(r, z)J0(ξr)dr(29)
c44 = G = (17) 0
G = E/[2(1 + v)] 2(1 + vhv)
ũ and w̃ are Hankel transformations of the
In cross-anisotropic materials, however, Substituting Equation 12 into 10 yields: displacements for u and w, whereas J0(ξr) and
five unique components (s11, s12, s13, s33, J0(ξr) are Bessel functions of first kind and
s44) exist, as shown in Equation 9. In the σr   c11 c12 c13 0   εr  order 0 and 1, respectively. ξ is the parameter
elastic range, these five components of the σθ  =  c12 c11 c13 0  εθ  
  ε (18)
of Hankel transformation corresponding
ordinary cross-anisotropic model require σz c c13 c33 0 to r. Equations 25 and 26 can be simplified
 τ rz   0  
 13  z
five parameters (Ev, Eh, vhv, υhh, Ghv) for their 0 0 c44   γ rz  by representing a differential function with
definitions (Love 1944). respect to z as λ. The next step is to elimi-
Expanding Equation 18 by expressing strains nate ũ from the modified equations, and
 s11
s12 s13  in terms of elastic displacement based on simplifying, resulting in:
 s12 s11 s13  Equation 3 gives:
s= (9)
s s s (λ4 – t 1ξ 2 λ2 + t 2ξ 4)w̃ = 0 (30)
 13 13 33  ∂u u ∂w
 s44  σ r = c11 + c12 + c13 (19)
∂r r ∂z b – ac + d 1
Where: t 1 = , t2 =
Where: s 11 = 1/Eh, s12 = –vhh/Eh, s13 = –vhv/‌Ev, bd bd
s33 = 1/Ev, s44 = 1/Ghv, ∂u u ∂w
σ θ = c12 + c11 + c13 (20)
Ghv = Ev/[2(1 + vhv)] ∂r r ∂z With the four roots, λ derived as:

Writing the stresses in terms of the strains ∂u u ∂w t 1 ± t12 – 4t 2

σz = c13 + c13 + c33 (21) λ = ±ξ (31)
would require Equation 7 to be inverted, ∂r r ∂z 2
yielding:
∂u ∂w Putting back the differential function in
τrz = c44 + (22)
σ i = cijεj (10) ∂z ∂r Equation 30 yields:

24 Journal of the South African Institution of Civil Engineering • Volume 59 Number 1 March 2017
 p Applied load/stresses
P/σ

Contact
stresses

a
R
Stress
Surfacing: h1, E1v, E1h , v1v, v1h
rotation

z
Base layer:
Figure 2 Surface circular load Pavement h2 , E2v, E2h , v 2v, v 2h
response

∂4 ∂2
– t 1ξ 2 + t 2ξ 4 w̃ = 0 (32)
∂z4 ∂z 2 Stresses
Sub-base layer:
and
h3, E3v, E3h , v 3v, v 3h
strains
With reference to the roots in Equation 31,
the solution form of Equation 32 depends on
the sign of the coefficient: t12 – 4t 2.
Furthermore, and this is very important, Selected layer:
h4, E 4v, E 4h , v4v, v4h
depending on the numerical integration
methods used, it may be necessary to modify
the solution for w̃ to obtain stable and accu-
rate responses of the pavement structure. The Sub-grade:
h5, E5v, E5h , v5v, v5h
solutions presented hereunder were based on
the numerical methods used in this research.

1. Solution 1: t12 – 4t 2 > 0, where Eh > Ev Figure 3 Hypothetical pavement structure

 1(ξ)e λ1z + C2(ξ)e –λ1z

w̃ = w̃ (ξ, z) = C 3. Solution 3: t12 – 4t 2 < 0, where. Eh < Ev. external and internal vertical and shear
+ C3(ξ)e λ2z + C4(ξ)e –λ2z(33) stresses. For the case shown in Figure 2,
 1(ξ)e λ1z + C2(ξ)e –λ1z
w̃ = w̃ (ξ, z) = C where there is only uniformly distributed
Where: + C3(ξ)e λ2z + C4(ξ)e –λ2z(36) vertical circular load P, whose radius is a on
the road surface, the Hankel transformation
t 1 + t12 – 4t 2 t 1 – t12 – 4t 2 Where: of the equilibrium of internal and external
λ1 = ξ , λ2 = ξ ,
2 2 stresses is given as:
and C1(ξ), C2(ξ), C3(ξ), and C4(ξ) are coef- t 1 + t12 – 4t 2
λ1 = ξ , and
ficients of integration. 2  σ̃ 1z (0, ξ)   –p̃ (ξ) 
 1  = 
 τ̃ rz (0, ξ)   0 (38)
In order to obtain stable and accurate results t 1 – t12 – 4t 2
λ2 = ξ
of pavement responses, Equation 33 was 2 Where:
modified to:
∞ pa
For stable and accurate results of pave- p̃(ξ) = ∫ rpJ0(ξr)dr = J1(ξa)
0 ξ
 1(ξ) cosh(r 2z) e r1z
w̃ = w̃ (ξ, z) = C ment responses, it was necessary to modify
+ C2(ξ) cosh(r 2z) e –r1z Equation 36 to:
– C3(ξ) sinh(r 2z) e r1z Pavement responses
– C4(ξ) sinh(r 2z) e –r1z(34)  1(ξ) cos(r 2z) e r1z
w̃ = w̃ (ξ, z) = C By applying Hankel inverse transformation
+ C2(ξ) cos(r 2z) e –r1z of all the Hankel transformed solutions, it is
Where: – C3(ξ) sin(r 2z) e r1z possible to determine solutions for pavement
– C4(ξ) sin(r 2z) e –r1z(37) responses at any point in a pavement struc-
t 1 – t12 – 4t 2 + t 1 + t12 – 4t 2
r1 = ξ , and ture. However, when r = 0, the computation
2 2 Where: of σ r and σ θ using Equations 19 and 20 is not
t1 t 12 – 4t 2 straightforward because of the term 1/r. The
t 1 + t12 – 4t 2 + t 1 – t12 – 4t 2 r1 = ξ , and r 2 = ξ
r2 = ξ 2 2 L’Hospital rule (Taylor 1952) is used to derive
2 the solutions.
After Hankel transforms have been deter-
2. Solution 2: t12 – 4t 2 = 0, where Eh = Ev Boundary condition – vertical mined, it is possible to determine responses
circular surface loading at any point (r, z) in the pavement structure
 1(ξ)eξz + C2(ξ)e–ξz
w̃ = w̃ (ξ, z) = C The boundary condition at the surface of the through Hankel inverse transformation using
+ C3(ξ)eξz + C4(ξ)e–ξz(35) pavement considers an equilibrium between the following equations:

Journal of the South African Institution of Civil Engineering • Volume 59 Number 1 March 2017 25
 Stress, Szz (MPa) Stress, Sxx (MPa)
–100 100 300 500 –2 000 0 2 000 4 000
0 0

100 100

200 200

300 300
Depth (mm)

Depth (mm)
400 400

500 500

600 600

700 700
This study Szz FEM Szz This study Sxx FEM Sxx

Figure 4 Stress distribution under the centre of the load with depth between this study and proprietary FEM package
∞
u(r, z) = ∫ ξũ(ξ, z)J1(ξr)dξ(39) Table 1 Profile of the hypothetical pavement structure
0
∞ Layer Layer E-Modulus E-Modulus P-Ratio P-Ratio Thickness Slip
w(r, z) = ∫ ξw̃(ξ, z)J0(ξr)dξ(40) type number vert (MPa) horiz (MPa) vert horiz (mm) rate
0
∞ AC 1 20 281 18 618 0.14 0.13 10 0
σz(r, z) = ∫ ξσ̃ z(ξ, z)J0(ξr)dξ(41)
0
AC 2 19 781 19 767 0.15 0.15 10 0
∞
τrz(r, z) = ∫ ξτ̃z(ξ, z)J1(ξr)dξ(42)
0 AC 3 19 303 19 282 0.15 0.15 10 0
c11 – c12 ∞
σθ – u = ∫ ξH1(ξ, z)J0(ξr)dξ(43) AC 4 18 846 15 583 0.12 0.11 10 0
r 0
G1 5 450 450 0.35 0.35 37.5 0
∞
σ r + σ θ = ∫ ξH2(ξ, z)J0(ξr)dξ(44)
0 G1 6 450 450 0.35 0.35 37.5 0

Where: G1 7 450 450 0.35 0.35 37.5 0

∞ c11 – c12 G1 8 450 450 0.35 0.35 37.5 0

H1(ξ, z) = ∫ r σ θ – u J0(ξr)dr(45)
0 r C3 9 250 250 0.35 0.35 75 0
∞
H2(ξ, z) = ∫ r (σ r – σ θ) J0(ξr)dr(46) C3 10 250 250 0.35 0.35 75 0
0
C3 11 250 250 0.35 0.35 75 0

C3 12 250 250 0.35 0.35 75 0

worked examples
Solutions developed in this study were used to G7 13 50 50 0.45 0.45 37.5 0
compute responses at different positions with-
G7 14 50 50 0.45 0.45 37.5 0
in a pavement structure. The pavement struc-
ture for which the simulation results are pre- G7 15 50 50 0.45 0.45 37.5 0
sented here are shown in Figure 3 on page 25. G7 16 50 50 0.45 0.45 37.5 0
A vertical load of 21.5 kN and a diameter of
SG 17 239 239 0.4 0.4 Semi-infinite
238 mm resulting in a 483 kPa contact stress
was used in the analysis. All layers except the
asphalt layer were modelled with isotropic, layers into four sub-layers resulted in a total of specific time of the day used in the analysis.
linear-elastic properties. The asphalt layer was 17 layers, as shown in Table 1. The top and bottom asphalt sub-layers
modelled with cross-anisotropic, linear elastic The vertical stiffness in the asphalt show significant reduction in the effective
properties. Information on all the layers, sub-layers shows a marginal increase close horizontal stiffness resulting from cracking
including sub-layers, is shown in Table 1. The to the surface that is a function of the effect initiating from both the top and bottom of
sub-layering of each of the upper four main of temperature variations with depth for a the main asphalt layer.

26 Journal of the South African Institution of Civil Engineering • Volume 59 Number 1 March 2017
 Strain, Ezz (MPa) Strain, Exx (MPa)

–5.0E-05
–2.0E-04

–1.5E-04
–1.0E-04

–1.0E-04
0.0E+00

0.0E+00
2.0E-04

4.0E-04

6.0E-04
3.0E-04

1.5E-04
5.0E-04
1.0E-04

5.0E-05

1.0E-04
0 0

100 100

200 200

300 300
Depth (mm)

Depth (mm)
400 400

500 500

600 600

700 700
This study Ezz FEM Ezz This study Exx FEM Exx

Figure 5 Strain distribution under the centre of the load with depth between this study and proprietary FEM package

Discussion of results
Figures 4 and 5 show comparisons between
the numerical method developed in this study
and a proprietary FEM package for stresses
(Szz) and (Sxx) as well as strains (Ezz) and
(Exx) distributions with depth, z-direction,
under the centre of a circular surface load
of magnitude 21.5 kN load (483 kPa contact
stress). From these results it is clear that the
closed-form solutions developed in this study
have achieved a very good level of accuracy,
as their results compare very well with results
from a proprietary FEM package.
It is also important to mention here
that the software developed from this work
has become the analysis engine for the
new SAPDM.
CONCLUSIONs
1. A numerical tool for the analysis of an
elastic multilayer system under the action
of surface circular load, considering both
cross-anisotropy and isotropy material
properties, has successfully been developed.
2. The numerical tool developed in this study
is capable of performing analyses for an
unlimited number of points in an elastic
multi-layered pavement system with an
unlimited number of layers, and on the
surface where an unlimited number of
uniformly distributed circular loads act.
3. The results shown in this paper confirm
the accuracy and reliability of the closed-
form theoretical solutions developed.
A very good match of results was obtained
between the numerical tool developed in
this study and a proprietary FEM package.
4. The numerical tools developed can be
used to improve the design, evaluation and
analysis of road/runway pavement systems.
acknowledgEmentS
This research study was part of the revision
of the SAPDM, a project sponsored by the
South African National Roads Agency SOC
Ltd (SANRAL) and the Council for Scientific
and Industrial Research (CSIR). The incentive
funding for rated researchers received from
the National Research Foundation (NRF) is
also gratefully acknowledged, as well as the
support from Dr Hechter Theyse for setting
up the numerical problem together with Prof
Rudi Du Preez to run the proprietary FEM
analyses for comparative purposes.
References
Adu-Osei, A, Little, D N & Lytton R L 2001. Cross-
anisotropic characterization of unbound granular
materials. Transportation Research Record No. 1757,
82–91.
Correia, A G (Ed.) 1999. Unbound granular materials.
Laboratory testing, in-situ testing and modelling.
Proceedings, International Workshop on Modelling
and Advanced Testing for Unbound Granular
Materials, Lisbon, Portugal, 21–22.
Emeriault, F & Chang, C S 1997. Anisotropic
elastic moduli of granular materials from a
micromechanical approach. Journal of Engineering
Mechanics, ASCE, 123(12): 1289–1293.
Love, A 1944. A Treatise on the Mathematical Theory of
Elasticity. New York: Dover Publications.
Maina, J W & Matsui, K 2004. Developing software
for elastic analysis of pavement structure responses
to vertical and horizontal surface loadings.
Transportation Research Record No. 1896, 107–118.
Masad, S, Little, D & Masad, E 2006. Analysis of flexible
pavement response and performance using isotropic
and anisotropic material properties. Journal of
Transportation Engineering, ASCE, 132(4): 342–349.
Masad, S, Little, D & Lytton, R 2004. Modeling nonlinear
anisotropic elastic properties of unbound granular bases
using microstructure distributed tensor. International
Journal of Geomechanics, ASCE, 4: 254–263.
Miyamoto, H 1977. Three-Dimensional Problems in the
Theory of Elasticity (in Japanese). Tokyo: Syokabo
Publishers.
Salehi-Ashtiani, R, Little, D N & Masad, E 2008.
Anisotropy level prediction model of unbound
aggregate systems. Proceedings, First International
Conference on Transportation Geotechnics (ICTG‑1),
25–27 August, Nottingham, UK, pp 253–262.
Sneddon, I N 1951. Fourier Transforms. New York:
McGraw-Hill.
Taylor, A E 1952. L’Hospital’s rule. American
Mathematical Monthly, 59(1): 20–24.
Tutumluer, E, Little, D N & Kim, S H 2003. Validated
model for predicting field performance of aggregate
base courses. Transportation Research Record
No. 1837, 41–49.
Wang, C D & Liao, J J 1999: Computing in transversely
isotropic rocks using influence charts. Rock
Mechanics and Rock Engineering, 32(1): 51–70.

Journal of the South African Institution of Civil Engineering • Volume 59 Number 1 March 2017 27
TECHNICAL PAPER Numerical modelling
Journal of the South African
Institution of Civil Engineering of flexible pavement
incorporating cross‑anisotropic
ISSN 1021-2019
Vol 59 No 1, March 2017, Pages 28–34, Paper 1385

PROF JAMES MAINA (Pr Eng, MSAICE, FSAAE) is a

professional pavement engineer, full-time
material properties
professor of civil engineering at the University of
Pretoria, and currently on secondment as a
technical director of a QA/QC Project for roads
Part II: Surface rectangular loading
in the State of Qatar. He obtained his PhD from
Miyazaki University in Japan. His professional J W Maina, F Kawana, K Matsui
activities include QA/QC in road projects,
pavement materials, and the development of advanced numerical analysis
(modelling) tools for pavement engineering application. He also teaches
both under- and post-graduate classes at the University of Pretoria. In order to better understand the impact of increased loading on roads, studies on tyre-road
interaction have gained prominence in recent years. Tyres form an essential interface between
Contact details:
Department of Civil Engineering vehicles and road pavement surfaces. These are the only parts of the vehicle that are in contact
University of Pretoria with the road and transmit the vehicle loading to the road surface. The use of the Cartesian
Private Bag X20 coordinate system is convenient in dealing with a uniform/non-uniform tyre load acting over
Hatfield 0028 a rectangular area, but few research reports are available that provide any form of theoretical
South Africa
solutions for pavement responses. This paper presents analytical solutions of responses due to
T: +27 12 420 6608
E: james.maina@up.ac.za rectangular loading acting on the surface of a multi-layered pavement system. The solutions
developed incorporate both isotropic and cross-anisotropic material properties. The method
PROF FUTOSHI KAWANA is an associate followed is based on classical trigonometric integral and Fourier transformation of Navier’s
professor at the Tokyo University of Agriculture equations. Accuracy and validity of the solutions are verified through comparisons with a
in Japan. His specialises in structural mechanics. proprietary finite element method (FEM) package. For this purpose, a pavement structure
He received his doctorate in engineering from composed of five main layers constituted by isotropic and cross-anisotropic (also known
the Science University of Tokyo in 2004. He is a
as transversely isotropic) material properties is analysed. In order to vary some of the layer
member of the Japan Society of Civil Engineers,
the Pavement Diagnosis Researchers Group properties with depth, the main layers were sub-layered, resulting in a 17-layer pavement system.
(NPO) and the Japanese Society of Irrigation,
Drainage and Rural Engineering.

Contact details: INTRODUCTION to carry the same load as the dual-tyres, the
Department of Bioproduction and Environment Engineering Motivation for this work is discussed in the wide-base tyre may have a much greater tyre
Faculty of Regional Environment Science companion paper by Maina et al (2017), (see inflation pressure and a larger individual
Tokyo University of Agriculture
pages 22–27 in this edition). Although mod- “footprint” (but could also be smaller than
1-1-1 Sakuragaoka
Setagaya-ku
ern era trucks transport heavier cargo, they the two combined “footprints” from stand-
Tokyo, 156-8505 are using relatively fewer tyres than their pre- ard dual-tyres). Research done by De Beer
Japan decessors, and as a result they are purported (2008) on tyre-pavement contact stresses has
T: +81 3 5477 2342 to be exerting much higher contact stresses also shown tyre-pavement contact stresses
E: fk205262@nodai.ac.jp
on the road surface. A good understanding of to be, although dependent on the loading
tyre-road contact stresses, and the ability to magnitude and inflation pressure, mostly rec-
PROF KUNIHITO MATSUI is emeritus professor of
model the macroscopic behaviour of materi- tangular and occasionally circular in shape.
civil and environmental engineering at Tokyo
Denki University in Japan. He obtained his PhD
als when subjected to varying traffic loading Development of solutions for circular surface
from the Department of Mechanics and and environmental conditions, is therefore loading has already been presented in the
Hydraulics at the University of Iowa, USA, in important for better road pavement designs companion paper published in this edition
1974. His areas of interest include structural and improved pavement performance. Tyres (Maina et al 2017).
analysis in pavement structures, static and
are the only part of the vehicle that are in In order to develop closed-form solutions
dynamic back-calculation of pavement systems,
non-destructive testing, thermal analysis of pavement systems, parameter
contact with the road, and transmit the for resilient responses of a pavement struc-
identification, sensitivity analysis and structural optimisation. vehicle loading to the road surface through ture under the rectangular tyre loading, the
a very small contact area, called the “contact Cartesian coordinate system may be conve-
Contact details:
Department of Civil and Environmental Engineering patch” or “tyre footprint”. Generally, there are nient to use. Bufler (1971) derived the theoret-
Tokyo Denki University two main types of truck tyres widely used on ical solution for multi-layered systems using
Hatoyama our roads – the single (or so-called wide-base the Cartesian coordinate system for isotropic
Hiki tyre) and the conventional dual-truck tyres. materials, but did not provide any worked
Satama, 350-0394
A single wide-base tyre is a proportionately examples. Similarly, Ernian (1989) used both
Japan
T: +81 492 96 5703/2549 larger and more robust tyre that is now being the cylindrical and Cartesian coordinate sys-
E: matsui@g.dendai.ac.jp used on trucks for heavy cargo. This type of tems to derive solutions for both circular and
tyre is expected to replace dual-tyres in the rectangular uniformly distributed loads acting
Keywords: pavement, linear-elastic analysis, transversely isotropic, future, on condition of minimal damage to on the surface of a multi-layered system with
cross‑anisotropy, isotropic, rectangular loading the existing road infrastructure. To be able isotropic material properties.

28
Maina JW, Kawana F, Matsui K. Numerical modelling of flexible pavement incorporating cross-anisotropic material properties – Part II: Surface rectangular loading.
J. S. Afr. Inst. Civ. Eng. 2017:59(1), Art. #1385, 7 pages. http://dx.doi.org/10.17159/2309-8775/2017/v59n1a4
 pz (x, y, 0)
2a
2b
x x
y y
z z
(a) Vertical pz (x, y, 0)
Figure 1 Uniformly distributed rectangular loads
This paper presents the development of
closed-form solutions for a multi-layered
pavement system under static rectangular
loadings, considering both isotropic and
cross-anisotropic material properties.
Isotropic materials have the same elastic
properties in both the vertical and horizontal
directions, and can be described by three
independent elastic constants – elastic
modulus E, Poisson’s ratio v and shear
modulus G. In contrast, cross-anisotropy of
an elastic material is defined by five indepen-
dent elastic constants – two elastic moduli
in vertical and horizontal directions (Ev, Eh),
two Poisson’s ratios in vertical and horizontal
directions (vvh, υhh) and one shear modulus
(Ghv), as presented by Love (1944).
In this research study, two classical math-
ematical methods, i.e. classical trigonometric
integral and classical potential function,
were investigated for flexibility and effi-
ciency. The former was adopted in this study
and its use in the determination of pavement
responses is presented in this paper. This
work is an extension of work where solutions
of responses due to circular loading were
presented (Maina et al 2017). This method is
flexible enough and can easily be extended
to dynamic and wave propagation problems,
although this is not the focus of the paper.
Accuracy and validity of the solutions
are verified through comparisons of the
computed responses to results obtained
using a proprietary finite element package
for a five-layer pavement structure, where
the four upper main layers were sub-layered,
resulting in a pavement system with 17
layers. The pavement layers were composed
of materials with isotropic and cross-aniso-
tropic (also known as transversely isotropic)
properties.
THEORETICAL DEVELOPMENT
Three-dimensional linear elasticity
For 3D problems shown in Figure 1, the
differential equilibrium equations in
Journal of the South African Institution of Civil Engineering • Volume 59 Number 1 March 2017
px (x, y, 0) py (x, y, 0)
2a 2a
2b 2b
x
y
z
(b) Horizontal px (x, y, 0) (c) Horizontal py (x, y, 0)
Cartesian coordinates may be expressed
using modified Navier’s equations. However,
y X
the equations are cumbersome to deal x
with because of the need to solve three
coupled partial differential equations for Y z
the three displacement components. The σzz
difficulty with finding particular solutions Z
of the system of equations in terms of the
displacements arises because each of the τyz τxz
sought-after deflection functions in the τxy τxy
σ xx
Cartesian coordinates (x, y and z) appear in σ yy
all three equilibrium equations.
The solutions may be simplified by
representing displacements in terms of har- Figure 2 S tress distributions in a Cartesian
monic potentials. It is because this approach coordinate system
decouples the equations in various different
ways. The most common approach is to use In the second approach the stress tensor is
the so-called Papkovich-Neuber potentials determined first, and then this tensor is used
to represent the solution (Ozawa et al 2009; to determine the strain tensor and displace-
Borodachev & Astanin 2008). This approach ment vector (known as problem in stresses).
enables the use of a well-known catalogue of In the third approach the strain tensor is
particular solutions of the Laplace equation, determined first, and then stress and dis-
and sometimes even reduces the problem, if placement tensors are determined (known
not completely, to one of the classical prob- as problem in strains). The work presented
lems of the theory of harmonic functions in this paper followed the first approach,
(theory of potential). Despite the simplifica- namely problem in displacements.
tion, it is difficult to extend this approach to A system of rectangular Cartesian
problems of dynamic or moving load analysis coordinates (x, y, z) is used. By assuming the
(Ozawa et al 2010). body forces to be zero, equilibrium equations
This paper aims at presenting closed- for an infinitesimal element (Figure 2) can be
form solutions of pavement responses due expressed using Navier’s equations as follows
to static rectangular loading in the verti- (Filonenko-Borodich 1963):
cal direction. The solutions presented in
this paper were derived based on a more ∂σx ∂τxy ∂τxz
+ + = 0 (1)
flexible and efficient classical transform ∂x ∂y ∂z
integral method. A similar approach can
be followed to derive solutions for rectan- ∂τxy ∂σ y ∂τzy
+ + = 0 (2)
gular loads acting in the longitudinal and ∂x ∂y ∂z
transverse directions.
∂τxz ∂τyz ∂σz
+ + = 0 (3)
Theoretical development ∂x ∂y ∂z
Three different approaches may be used
to solve problems of the theory of elastic- Where: σx, σ y, σz are normal stress and τxz,
ity (Borodachev 1995; 2001). In the first τyz and τxy are shear stresses acting on an
approach, the displacement vector is infinitesimal element. The two subscripts
determined first, and this vector is then on the symbols for shear stresses represent,
used to determine the stress and strain ten- respectively, the face and direction on which
sors (known as problem in displacements). the shear stress is acting.
29
 The strain-displacement relationship may elements. The number of independent ele- Three components (c11, c12, c44) exist
be represented as follows: ments is obtained by counting elements for isotropic materials, whereas, in
in the upper right triangle of the matrix, cross-­anisotropic materials, five unique
∂u ∂v ∂w ∂v ∂u including the diagonal elements (i.e. components (c11, c12, c13, c33, c44, c 66) exist.
εx = , εy = , εz =
, , γxy = +
∂x ∂y ∂z ∂x ∂y 1 + 2 + 3 + 4 + 5 + 6 = 21). Furthermore, It should be noted that a special cross-
∂u ∂w ∂w ∂v if the material exhibits symmetry in its anisotropy solution exists, where Ev = Eh and
γxz = + , γ yz = + (4)
∂z ∂x ∂y ∂z elastic response, the number of independent vvh = υhh. Equation 13 shows the six compo-
components in the s matrix will be reduced nents of the elastic stiffness matrix, whereas
Where: u = u(x, y, z), v = v(x, y, z) and even further. Equations 14–19 define each of these six
w = w(x, y, z) are displacements in the direc- For example, in the simplest case of components:
tions of x, y and z axes. Furthermore, ε x, εy isotropic materials whose elastic moduli are
and εz are normal strains corresponding to the same in all directions, only three unique  c11 c12 c13 0 0 0 
normal stresses σx, σ y and σz, whereas γxz, γ yz
and γxy are shear strains corresponding to
components (s11, s12, s44) exist, as shown in
Equation 9. In the elastic range, these three 
c12
c
c =  13
c11
c13
c13
c33
0
0
0
0
0
0 
(13)

 
shear stresses τxz, τyz and τxy. coefficients of the ordinary isotropic model 0 0 0 c44 0 0
require three parameters (E, v, G) for their 0 0 0 0 c44 0
Generalised Hooke’s law definitions as follows:  0 0 0 0 0 c 66 
In linear elasticity, if the stress is sufficiently
small, Hooke’s law is used to represent  s11 s12 s12 0 0 0  Where:
the material behaviour and to relate the
unknown stresses and strains. The general s12
s
s =  12
s11
s12
s12
s11
0
0
0
0
0
0 
(9) c11 =
2
Eh(–Ev + Ehvhv)
(14)
2

 
equation for Hooke’s law is: 0 0 0 s44 0 0 (1 + vhh)(–Ev + Evvhh + 2Ehvhv)
0 0 0 0 s44 0
2
εij = sijkl σkl = ∑3 Eh(Evvhh + Ehvhv )
k=1 ∑3l=1 sijkl σkl(5)  0 0 0 0 0 s44  c12 = – (15)
2
(1 + vhh)(–Ev + Evvhh + 2Ehvhv)
where i, j = 1, 2, 3 1 v 1
Where: s 11 = , s12 = – , s44 = ,
E E G EhEvvhv
c13 = – (16)
In this case sijkl is a fourth-rank tensor called E 2
–Ev + Evvhh + 2Ehvhv
G=
elastic compliance of the material. Each 2(1 + v)
subscript of sijkl takes on the values from 1 to Ev2 (–1 + vhh)
c33 = (17)
2
3, giving a total of 34 = 81 independent com- In cross-anisotropic materials, however, –Ev + Evvhh + 2Ehvhv
ponents in s. However, due to the symmetry six unique components (s11, s12 , s13, s 33,
of both εij and σkl, the elastic compliance s s44) and s 66 exist, as shown in Equation Ev
c44 = G = (18)
must satisfy the relation: 10. These six components of the ordinary 2(1 + vhv)
cross-anisotropic model require five param-
sijkl = sjikl = sijlk = sjilk(6) eters (Ev, Eh, vhv, υhh, Ghv) for their defini- Ev (c11 – c12)
c 66 = = (19)
tions (Love 1944). 2(1 + vhv) 2
It follows from Equation 5, therefore, that the
generalised Hooke’s law in Equation 5 can be  s11 s12 s13 0 0 0  Substituting Equation 13 in 11 yields:
simplified to become:
s12
s
s11
s13
s13
s33
0
0
0
0
0
0   σx   c11
c12 c13 0 0 0  εx 

   
s =  13 (10)

 
εi = sijσ j, where i, j = 1 , .., 6 (7) 0 0 0 s44 0 0 σy c12 c11 c13 0 0 0 εy
0 0 0 0 s44 0 σz c13 c13 c33 0 0 0 εz
 =  (20)

   
This relationship reduces the number of s  0 0 0 0 0 s66  τyz 0 0 0 c44 0 0 γ yz
components to 36, as seen in the following τxz 0 0 0 0 c44 0 γxz
linear relation between the pseudovector 1 vhh vhv  τxy   0 0 0 0 0 c 66   γxy 
Where: s 11 = , s12 = – , s13 = –
,
forms of the strain and stress: Eh Eh Ev
1 1 Ev Derivation of the solutions
s33 = , s44 = , Ghv = ,
 ε1   s11
s12 s13 s14 s15 s16   σ 1  Ev Ghv 2(1 + vhv) It is convenient to use the three-dimen-


ε2
ε3
 =
s21 s22 s23 s24 s25 s26 σ 2
s31 s32 s33 s34 s35 s36 σ 3
 ⇒
  s66 = 2(s11 – s12) sional Cartesian coordinate system and
make an assumption that displacement


ε4 s41 s42 s43 s44 s45 s46 σ4
ε5 s51 s52 s53 s54 s55 s56 σ 5
 ε 6   s61 s62 s63 s64 s65 s66   σ 6 
 ε x   s11 s12 s13 s14 s15 s16   σ x 
 
Writing the stresses in terms of the strains functions u(x, y, z), v(x, y, z) and w(x, y, z) in
would require Equation 8 to be inverted, the x, y and z directions, respectively, may
yielding: be represented using double trigonometric
functions, as detailed below. With this

 

εy
εz
=
s21 s22 s23 s24 s25 s26 σ y
s31 s32 s33 s34 s35 s36 σz
 (8)
  σ i = cij εj(11) approach, the x and y dependencies of the
displacement functions u(x, y, z), v(x, y, z)

 
γ yz s41 s42 s43 s44 s45 s46 τyz
γxz s51 s52 s53 s54 s55 s56 τxz
 γxy   s61 s62 s63 s64 s65 s66   τxy 
The s matrix in this form is also symmetric.
It therefore contains only 21 independent
 
Where: cij is a fourth-rank tensor called
elastic stiffness of the material. The stiffness
and compliance matrices in Equation 11 are
related in the following form:
c = s–1(12)
and w(x, y, z) are accommodated by means
of analytical double integral Fourier trans-
form, with the z dependence approximated
by using closed-form solutions.
The Fourier transform of the displace-
ments employs transform pairs that are

30 Journal of the South African Institution of Civil Engineering • Volume 59 Number 1 March 2017
defined in terms of the Fourier parameters in Using Equations 27 and 28 it becomes con- 2. Solution 2: t12 – 4t 2 = 0, where Eh = Ev
the x and y directions (ξx and ξy) as follows: venient to eliminate ũ(ξx, ξy, z) and ṽ(ξx, ξy, z)
from Equations 24–26, then simplify ξy ξ
∞ ∞ w̃ (ξx, ξy, z) = e ξz –C1(ξ) – C3(ξ)
ũ(ξx, ξy, z) = ∫
 ∫ u(z) sin(ξx x) to obtain: ξ ξx
–∞ –∞
cos(ξy y)dξx dξy(21) 3c11 – c12 zξ
+ C5(ξ) –
(λ4 – t 1ξ 2 λ2 + t 2ξ4)w̃ = 0 (29) ξx(c11 + c12) ξx
∞ ∞ ξy ξ
ṽ (ξx, ξy, z) = ∫
 ∫ v(z) cos(ξx x) The roots of Equation 29 are determined as: + e–ξz C2(ξ) + C4(ξ)
–∞ –∞ ξ ξx
sin(ξy y)dξx dξy(22)
t 1 ± t12 – 4t 2 3c11 – c12 zξ
λ = ±ξ (30) + C6(ξ) +  (34)
∞ ∞
2 ξx(c11 + c12) ξx
w̃(ξx, ξy, z) = ∫
 ∫ w(z) cos(ξx x)
–∞ –∞
cos(ξy y)dξx dξy(23) Where: t 1 = (b + (a + d) – c f ), t 2 = a d + b d 3. Solution 3: t12 – 4t 2 < 0, where Eh < Ev
from Equation 27.
Putting back the differential function in  1(ξ)e λ1z + C2(ξ)e –λ1z
w̃ (ξx, ξy, z) = C
Where: ũ(ξx, ξy, z), ṽ (ξx, ξy, z) and w̃(ξx, ξy, z) Equation 29 yields: + C3(ξ)e λ2z + C4(ξ)e –λ2z(35)
are the Fourier transforms of displacement
functions about the coordinate z (Sneddon ∂4 ∂2 Where:
– t 1ξ 2 + t 2ξ 4 w̃ = 0 (31)
1951). The procedure that follows is to ∂z4 ∂z 2
expand Equation 20 and to express strains t 1 + t12 – 4t 2
λ1 = ξ , and
in terms of elastic displacements based on With reference to the roots in Equation 30, 2
Equation 4. After that the Fourier trans- the solution form of Equation 31 depends on
forms of the displacements using Equations t 1 – t12 – 4t 2
the sign of the coefficient t12 – 4t 2. λ2 = ξ
21, 22 and 23 are applied. Then the result- 2
Furthermore, and this is very important,
ing functions are substituted into Equations
depending on the numerical integration
1–3 to yield their Fourier transforms For stable and accurate results of pavement
methods used, it may be necessary to modify
as follows: responses, Equation 35 was modified to:
the solution for w̃ to obtain stable and accu-
rate responses of the pavement structure.
– ũ(ξx, ξy, z)(ξy2(c11 – c12) + 2c11ξx2) + The solutions presented hereunder were  1(ξ) cos(r 2z) e r1z
w̃ (ξx, ξy, z) = C
based on the numerical methods used in this + C2(ξ) cos(r 2z) e –r1z
ξxξy(c11 + c12) ṽ (ξx, ξy, z) + 2ξx(c13 + c44) research. – C3(ξ) sin(r 2z) e r1z
∂w̃(ξx, ξy, z) ∂ 2ũ(ξx, ξy, z) – C4(ξ) sin(r 2z) e –r1z(36)
– 2c44 = 0 (24) 1. Solution 1: t12 – 4t 2 > 0, where Eh > Ev
∂z ∂z 2
Where:
 1(ξ)e λ1z + C2(ξ)e –λ1z
w̃ = w̃ (ξx, ξy, z) = C
– ξx, ξy(c11 + c12)ũ(ξx, ξy, z) + ṽ (ξx, ξy, z) + C3(ξ)e λ2z + C4(ξ)e –λ2z(32) t1 t12 – 4t 2
r1 = ξ and r 2 = ξ
2 2
(ξx2(c11 – c12) 2c11ξy2) + 2ξy(c13 + c44)
Where:
∂w̃(ξx, ξy, z) ∂ 2ṽ (ξx, ξy, z) Solutions for the remaining Fourier
– 2c44 = 0 (25) t 1 + t12 – 4t 2 t 1 – t12 – 4t 2 transformed displacements
∂z ∂z 2 λ1 = ξ , λ2 = ξ ,
2 2 Substituting Equations 33, 34 and 36 into
∂ũ(ξx, ξy, z) ∂ṽ(ξx, ξy, z) and C1(ξ), C2(ξ), C3(ξ), and C4(ξ) are Equations 24 and 25, the solutions for
(c13 + c44) ξx + ξy
∂z ∂z coefficients of integration determined, as ũ(ξx, ξy, z) and ṽ(ξx, ξy, z) are derived.
∂ 2w̃(ξx, ξy, z) described later, by using boundary loading As an example, solutions for ũ(ξx, ξy, z)
+ c33 – c44w̃(ξx, ξy, z)
∂z 2 conditions. and ṽ(ξx, ξy, z) for the case where Eh = Ev
(solution 2) are obtained as follows:
(ξx2 + ξy2 = 0 (26)
In order to obtain stable and accurate results
of pavement responses, Equation 32 was
Equations 24–26 can be simplified further modified to:  2(ξ)e ξz + C4(ξ)e –ξz
ũ(ξx, ξy, z) = C
by representing a differential function with +C 5(ξ)zeξz + C6(ξ)ze–ξz(37)
∂  1(ξ) cosh(r 2z) e r1z
w̃ (ξx, ξy, z) = C
respect to z as λ = as well as making the ξy ξy
∂z + C2(ξ) cosh(r 2z) e –r1z ṽ(ξx, ξy, z) = e ξz C1(ξ) + C3(ξ) + C5(ξ) z
following substitutions: – C3(ξ) sinh(r 2z) e r1z ξx ξx
– C4(ξ) sinh(r 2z) e –r1z(33) ξy
+ e–ξz C2(ξ) + C4(ξ)
c11 – c12 c11 + c12 c13 + c44 ξx
a= ,b= ,c= ,
2c44 2c44 c44 Where: ξy
+ C6(ξ) z (38)
c44 c + c44 ξx
d= , f = 13 (27) t 1 – t12 – 4t 2 + t 1 + t12 – 4t 2
c33 c33 r1 = ξ , and
2 2 Solutions for ũ(ξx, ξy, z), ṽ(ξx, ξy, z) and
and w̃(ξx, ξy, z) are then substituted into
t 1 + t12 – 4t 2 – t 1 – t12 – 4t 2
r2 = ξ Equations 4 and 20 to determine the
ξ = ξx2 + ξy2(28) 2 2 Fourier transforms of normal and shear

Journal of the South African Institution of Civil Engineering • Volume 59 Number 1 March 2017 31
 pz (x, y, 0) Applied load/stresses
P/σ

Contact
stresses

2a
2b x
y

Stress
z Surfacing: h1, E1v, E1h , v1v, v1h
rotation

Figure 3 S urface rectangular vertical load

(pz(x, y, z) = 0) Base layer:
Pavement h2 , E2v, E2h , v 2v, v 2h
response
stresses σ̃ x(ξx, ξy, z), σ̃ y(ξx, ξy, z), σ̃ z(ξx, ξy, z),
τ̃xz(ξx, ξy, z), τ̃ yz(ξx, ξy, z), and τ̃xy(ξx, ξy, z).
Stresses
Sub-base layer:
Boundary condition – and
h3, E3v, E3h , v 3v, v 3h
strains
surface rectangular vertical loading
For the loading case shown in Figure 3, there is
only a single uniformly distributed surface rec-
tangular vertical load, P, whose sides are 2 × a Selected layer:
h4, E 4v, E 4h , v4v, v4h
and 2 × b in dimensions. Boundary conditions
for rectangular loads acting on a surface of a
semi-infinite medium shown in in Figure 3 may
be represented by taking into consideration Sub-grade:
h5, E5v, E5h , v5v, v5h
the equilibrium between external and internal
vertical and shear stresses, as shown below:
When x ≤ |a| and y ≤ |b| then:
Figure 4 Hypothetical pavement structure
1
 σ̃ z(ξx, ξy, z = 0)   p̃z(ξx, ξy, z = 0) 
   
τ̃ 1
 xz x y(ξ , ξ , z = 0)  = –  0 (39) Table 1 Profile of the hypothetical pavement structure
 1   
 τ̃ yz(ξx, ξy, z = 0)   0 
Layer Layer E-Modulus E-Modulus P-Ratio P-Ratio Thickness Slip
type number vert (MPa) horiz (MPa) vert horiz (mm) rate
Where:
AC 1 20 281 18 618 0.14 0.13 10 0
a b pz cos(ξx x) cos(ξy y)
p̃z(ξx, ξy) = ∫ ∫ – d(ξy)d(ξx)
–a –b AC 2 19 781 19 767 0.15 0.15 10 0
2π
2pz sin(ξx x) sin(ξy y)
=– and AC 3 19 303 19 282 0.15 0.15 10 0
ξ x ξy π
P AC 4 18 846 15 583 0.12 0.11 10 0
pz =
(2 × a)(2 × b) G1 5 450 450 0.35 0.35 37.5 0

In addition, when x > |a| and y > |b| then: G1 6 450 450 0.35 0.35 37.5 0

G1 7 450 450 0.35 0.35 37.5 0

1
 σ̃ z(ξx, ξy, z = 0)   0 
   
 τ̃ 1xz(ξx, ξy, z = 0)  =  0 (40) G1 8 450 450 0.35 0.35 37.5 0
 1   
 τ̃ yz(ξx, ξy, z = 0)   0 
C3 9 250 250 0.35 0.35 75 0

Pavement responses C3 10 250 250 0.35 0.35 75 0

By applying Fourier inverse transformation
of all the Fourier transformed solutions, it is C3 11 250 250 0.35 0.35 75 0
possible to determine solutions for pavement
C3 12 250 250 0.35 0.35 75 0
responses at any point in a pavement structure.
G7 13 50 50 0.45 0.45 37.5 0
1 ∞ ∞
u(x, y, z) =  ∫ ∫ ũ(kx, ky, z) sin(ξx x)
2π –∞ –∞ G7 14 50 50 0.45 0.45 37.5 0

cos(ξy y)dξxdξy(41) G7 15 50 50 0.45 0.45 37.5 0

1 ∞ ∞
v(x, y, z) =  ∫ ∫ ṽ(ξx, ξy, z) cos(ξx x) G7 16 50 50 0.45 0.45 37.5 0
2π –∞ –∞
SG 17 239 239 0.4 0.4 Semi-infinite
sin(ξy y)dξxdξy(42)

32 Journal of the South African Institution of Civil Engineering • Volume 59 Number 1 March 2017
 Stress, Szz (MPa) Stress, Sxx (MPa)
0 200 400 600 –2 000 0 2 000 4 000
0 0

100 100

200 200

300 300
Depth (mm)

Depth (mm)
400 400

500 500

600 600

700 700
This study Szz FEM Szz This study Sxx FEM Sxx

Figure 5 Stress distribution under the centre of the load with depth between this study and proprietary FEM package

1 ∞ ∞
w(x, y, z) =  ∫ ∫ w̃(ξx, ξy, z) cos(ξx x) results are presented here is shown in
2π –∞ –∞ Figure 4. Based on historical tyre loading on
sin(ξy y)dξxdξy(43) South African roads, a tyre vertical load of
21.5 kN on a rectangular patch of 231 mm
1 ∞ ∞
σx(x, y, z) =  ∫ ∫ σ̃ x(ξx, ξy, z) cos(ξx x) by 238 mm resulting in a 390 kPa contact
2π –∞ –∞ stress was used in the analysis (De Beer
cos(ξy y)dξxdξy(44) 2008). All l­ayers, except the asphalt layer,
were m ­ odelled with isotropic, linear-elastic
1 ∞ ∞
σ y(x, y, z) =  ∫ ∫ σ̃ y(ξx, ξy, z) cos(ξx x) ­properties. The asphalt layer was modelled
2π –∞ –∞ with cross-anisotropic, linear elastic
cos(ξy y)dξxdξy(45) ­properties. Information on all the layers,
including sub-layers, is shown in Table 1.
1 ∞ ∞
σz(x, y, z) =  ∫ ∫ σ̃ z(ξx, ξy, z) cos(ξx x) The sub-layering of the upper four main
2π –∞ –∞ ­layers into four layers each resulted in a total
cos(ξy y)dξxdξy(46) of 17 layers, as shown in Table 1.
The vertical stiffness in the asphalt sub-
1 ∞ ∞
τxz(x, y, z) =  ∫ ∫ τ̃xz(ξx, ξy, z) sin(ξx x) layers shows a marginal increase close to the
2π –∞ –∞ surface that may be attributed to binder age-
cos(ξy y)dξxdξy(47) ing, but more probably to a slight reduction
in the temperature conditions. The top and
1 ∞ ∞
τyz(x, y, z) =  ∫ ∫ τ̃ yz(ξx, ξy, z) cos(ξx x) bottom asphalt sub-layers show significant
2π –∞ –∞ reduction in the effective horizontal stiffness
sin(ξy y)dξxdξy(48) resulting from cracking initiating from both
the top and bottom of the main asphalt layer.
1 ∞ ∞
τxy(x, y, z) =  ∫ ∫ τ̃xy(ξx, ξy, z) sin(ξx x) Discussion of results
2π –∞ –∞ Figures 5 and 6 show comparisons between
sin(ξy y)dξxdξy(49) the numerical method developed in this
study and a proprietary FEM package for
stresses (Szz) and (Sxx) as well as strains
worked examples (Ezz) and (Exx) distributions with depth,
Solutions developed in this study were used z-direction, under the centre of a rectangular
to compute responses at different positions surface load of magnitude 21.5 kN (390 kPa
within a pavement structure. The pave- contact stress). From these results it is clear
ment structure for which the simulation that the closed-form solutions developed in
this study have achieved a very good level of
accuracy, as their results compare well with
results from a proprietary FEM package.
It is also important to mention here
that the software developed from this
work has become the analysis engine for
the new SAPDM (South African Pavement
Design Method).
What is also evident in the strains plots
is that, as the FEM increases the size of
the elements at points far from where the
load is acting, the accuracy is reduced a
little bit and the results start moving away
from the results of this study. However, all
in all, the agreement of the two methods,
i.e. approximation by FEM and closed-
form solution developed in this paper, is
very good.
CONCLUSIONs
1. A numerical tool for analysis of an elastic
multilayer system under the action of
a surface rectangular load, considering
both cross-anisotropic and isotropic
material properties, has successfully
been developed.
2. The numerical tool developed in this
study is capable of performing analyses
for an unlimited number of points in an
elastic multi-layered pavement system
with an unlimited number of layers,
and on the surface where an unlim-
ited number of uniformly distributed
­rectangular loads act.

Journal of the South African Institution of Civil Engineering • Volume 59 Number 1 March 2017 33
 Strain, Ezz (MPa)
–1.0E-04
0.0E+00
1.0E-04
0 0
100
200
300
Depth (mm)
400
500
600
700
This study Ezz
Figure 6 Strain distribution under the centre of the load with depth between this study (closed-form) and proprietary FEM package
3. The results shown in this paper confirm
the accuracy and reliability of the closed-
form theoretical solutions developed.
4. A very good match of the stress results
was obtained between the numerical tool
developed in this study and a proprietary
FEM package.
5. Differences in the strain results at points
far from where the load acts, were
observed. The differences seem to be
emanating from the FEM, which is a pro-
prietary package. In conventional FEM,
stress computations come after strain
computations, and the trends should
have been similar. The reason for the
differences is not clear, but the results, as
measured, are nevertheless reported here.
6. The numerical tools developed can be
used to improve the design, evaluation
and analysis of multi-layered road/runway
pavement systems.
acknowledgEmentS
This research study was part of the revision
of the SAPDM, a project sponsored by the
South African National Roads Agency SOC
34
–1.5E-04
–1.0E-04
2.0E-04
4.0E-04
3.0E-04
5.0E-04
100
200
300
Depth (mm)
400
500
600
700
FEM Ezz
Ltd (SANRAL) and the Council for Scientific
and Industrial Research (CSIR). The incen-
tive funding for rated researchers received
from the National Research Foundation
(NRF) is also gratefully acknowledged,
as well as the support from Dr Hechter
Theyse for setting up the numerical
problem together with Prof Rudi Du Preez
to run the proprietary FEM analyses for
comparative purposes.
References
Borodachev, N M 1995. Three-dimensional problem
of the theory of elasticity in strains. Strength of
Materials, 27(5–6): 296.
Borodachev, N M 2001. Construction of exact solutions
to three-dimensional elastic problems in stresses.
International Applied Mechanics, 39(6): 438.
Borodachev, N M & Astanin, V V 2008. Solution of a
three-dimensional problem of the elasticity theory in
terms of displacements for an isotropic elastic layer.
Strength of Materials, 40(3): 308–315.
Bufler, H 1971. Theory of elasticity of a multilayered
medium. Journal of Elasticity, 1(2): 125–143.
De Beer, M 2008. Stress-In-Motion (SIM) – A new
tool for road infrastructure protection? Proceedings,
Journal of the South African Institution of Civil Engineering • Volume 59 Number 1 March 2017
Strain, Exx (MPa)
–5.0E-05
0.0E+00
1.5E-04
5.0E-05
1.0E-04
This study Exx FEM Exx
International Conference on Heavy Vehicles, WIM
Session 7, 19–22 May, Paris/Marne-la-Vallée.
Ernian, P 1989. Static response of a transversely
isotropic and layered half-space to general surface
loads. Physics of the Earth and Planetary Interiors,
54: 353–363.
Filonenko-Borodich, M 1963. Theory of Elasticity.
Moscow: Peace Publishers.
Love, A 1944. A Treatise on the Mathematical Theory of
Elasticity. New York: Dover Publications.
Maina, J W, Kawana F & Matsui, K 2017. Numerical
modelling of flexible pavement incorporating cross-
anisotropic material properties – Part I: Surface
circular loading. Journal of South African Institution
of Civil Engineering, 59(1): 22–27.
Ozawa, Y, Maina, J W & Matsui, K 2009. Linear
elastic analysis of pavement structure loaded over
rectangular area. Paper presented at the 88th
Transportation Research Board Annual Meeting,
11–15 January, Washington, DC. (CD-ROM.)
Ozawa, Y, Maina, J W & Matsui, K 2010. Analysis of
multilayered half space due to rectangular moving
load. Paper presented at the 89th Transportation
Research Board Annual Meeting, 10–14 January,
Washington, DC. (CD-ROM.)
Sneddon, I N 1951. Fourier Transforms. New York:
McGraw-Hill.
Potential solution to pollution TECHNICAL PAPER
of groundwater by diffusion Journal of the South African
Institution of Civil Engineering
of volatile organic compounds ISSN 1021-2019
Vol 59 No 1, March 2017, Pages 35–44, Paper 1435

through the primary HDPE REON PIENAAR (Pr Eng) holds a BEng (Civil),

geomembrane in composite BEng Honours and an MEng in Environmental

Engineering from the University of Pretoria. He
has twelve years’ experience in waste

lining systems of landfills management, dam engineering and alternative

waste treatment technologies. He is an
Associate Member of the South African
Institution of Civil Engineering (SAICE), and
vice-chairman of the Central Branch Committee of the Institute of Waste
R A Pienaar, W Fourie Management of South Africa (IWMSA). He works for AECOM South Africa
where he is an Associate Engineer and the Practice Area Lead for Waste
Services (Africa).
Waste in a landfill is exposed to the chemicals and heat generated over time, producing Contact details:
harmful fluids in the form of leachate or landfill gas that migrate from the landfill towards the AECOM
liner or capping, and include organic contaminants. The high-density polyethylene (HDPE) PO Box 3173
geomembrane (GM) component of the landfill liner is often believed to be the primary barrier to Pretoria, 0001
South Africa
contaminant transport, but volatile organic compounds (VOCs) diffuse through geomembranes
T: +27 12 421 3893
at appreciable rates. The aim of this study was to obtain reliable data on the reduction in E: reon.pienaar@aecom.com
diffusion of VOCs through the HDPE GM component in the composite liner systems of landfills
by extracting air through the leakage detection layer or drainage layer of the composite liner. It DR WALTER FOURIE holds a Bachelor’s Degree in
was shown that introducing a flow of air through a pervious zone adjacent to the GM layer in a Chemical Engineering from the University of
landfill liner would significantly reduce the concentrations of VOCs in the groundwater beneath Pretoria, and a Master’s and PhD in
landfills and waste containment facilities. Environmental Engineering from the University
of Alaska. For the last 15 years he has consulted
on engineering and environmental projects
from Antarctica to Alaska. He is currently the
INTRODUCTION AND BACKGROUND liner is often believed to be the primary bar- Director of EnviroEng (Pty) Ltd in Nelspruit, in
Waste and the adverse effects of waste stor- rier to contaminant transport, but the clay the Mpumalanga Province of South Africa.
age and disposal is not a modern concept. component in the composite liner usually Contact details:
It has always been a byproduct of human controls the rate of transport of VOCs, since EnviroEng (Pty) Ltd
beings’ use of the earth’s natural resources researchers like Edil (2003) have shown that PO Box 12282
Steiltes
for survival. VOCs diffuse through geomembranes at
Nelspruit, 1213
The safe and reliable long-term disposal appreciable rates. Therefore, the effectiveness South Africa
of solid waste residues is an important com- of modern landfill liner systems in minimis- T: +27 72 966 5197
ponent of integrated waste management. ing the migration of VOCs merits scrutiny. E: walter@enviroeng.co.za
Solid waste residues are waste components Landfill and landfill liner technology has
that are not recycled, that remain after gone through significant developments in
processing at a material recovery facility, recent years. Waste disposal landfills (gener-
or that remain after the recovery of energy. al and hazardous) have evolved from uncon-
Historically, solid waste was placed in trolled dumps to highly engineered facilities
depressions in the soil of the earth’s crust designed to protect the environment and
through a process called landfilling. promote environmental sustainability. Liner
When waste materials in a landfill or any technology and the relevant regulations that
other waste body are exposed to the air and govern them have also evolved from rudi-
water infiltration over time, heat is generated mentary compacted clay liners to complex
and harmful fluids in the form of leachate composite engineered lining systems com-
and/or landfill gas are generated. Leachate prising a range of layers such as compacted
and landfill gas migrate from the landfill clay, geosynthetic clay liners, geomembranes,
towards the basal liner or capping, and geocomposite drains and geotextiles.
most often include organic contaminants. Early concerns regarding composite liners
These organic contaminants include a group typically focused on their hydraulic conduc-
commonly referred to as volatile organic tivity and their ability to limit advective trans-
compounds (VOCs) that have been known port (Edil 2003), but evidence has been pre-
to migrate to and pollute the underlying sented subsequently that highlights diffusive
groundwater (Prosser & Janechek 1995). transport (i.e. contaminant migration driven
The high-density polyethylene (HDPE) by the difference in concentration between
geomembrane (GM) as part of a landfill the upper and lower sides of the liner) as a Keywords: waste, liner, VOC, diffusion, groundwater

35
Pienaar RA, Fourie W. Potential solution to pollution of groundwater by diffusion of volatile organic compounds through the primary HDPE geomembrane in
composite lining systems of landfills. J. S. Afr. Inst. Civ. Eng. 2017:59(1), Art. #1435, 10 pages. http://dx.doi.org/10.17159/2309-8775/2017/v59n1a5
dominant mode of transport in well-built
liner systems (McWatters & Rowe 2009).
Although HDPE GMs are used for a vari-
ety of applications as barriers for contaminant
transport, for the purpose of this study, the
only application of HDPE investigated will be
that of waste disposal landfill liner.
STUDY OBJECTIVES
The aim of this study was to obtain reliable
data on the reduction in diffusion of VOCs
through the HDPE GM component in com-
posite liner systems of landfills by extracting
air through the leakage detection layer or
drainage layer of the composite liner adja-
cent to the HDPE component of the liner.
With the extraction of air through the liner
system, some of the VOCs that could dif-
fuse through the HDPE GM layer would be
extracted with the air and thus reduce VOCs
in the underlying soil and/or groundwater.
The objective was to undertake tests in
the following three phases:
■■ Phase 1 aimed to prove that the chosen
VOCs diffuse from source to receptor
through an HDPE GM layer, and to
compare the results obtained with those
in the literature.
■■ Phase 2 aimed to prove that, even if
the separation between the source and
receptor consisted of two HDPE GMs
separated by an air-filled pervious zone,
diffusion of the VOCs would still occur
from the source to the receptor volumes.
■■ Phase 3 aimed to prove that, by intro-
ducing airflow into the pervious zone
between the two HDPE GMs, the concen-
tration of VOCs in the receptor volume
(due to diffusion through the HDPE GM)
could be reduced significantly. The test-
ing in this phase also aimed to determine
if the rate of air removal would play a role
in the diffusive process and the resultant
VOC concentrations in the receptor.
For the purposes of this study, the onsite
conditions were replicated in a laboratory
using diffusion test cells similar to those
used by Sangam and Rowe (2001). All of the
HDPE GMs used in this study were supplied
by Aquatan (Pty) Ltd in South Africa, and
the diffusion test cells were built by Interlock
Systems (Pty) Ltd in Pretoria, South Africa.
Literature review
Waste disposal and
containment barriers
In South Africa, waste disposal landfills are
grouped into four classes according to the
waste types earmarked for disposal (Classes
A, B, C and D).
36 Journal of the South African Institution of Civil Engineering • Volume 59 Number 1 March 2017
Geotextile
filter layer
2 mm HDPE
geomembrane
Geotextile
filter layer
1.5 mm HDPE
geomembrane
150 mm base
preparation
layer
Figure 1 Class A containment barrier design prescription (Government Gazette 36784 p 37)
Class A landfills require a minimum of
a double-composite containment barrier
system and are meant for the disposal of
hazardous wastes. The Class A landfill liner
prescribed in the Waste Classifications and
Management Regulations of the Waste Act
(DWAF 1998) are presented in Figure 1.
The liner detail in Figure 1 is a general
minimum standard, and every containment
facility needs to have its own fit-for-purpose
engineered lining system that conforms
to the Class of facility and waste type for
which it is intended. The layers can be
replaced by other layers of equal or improved
performance, and the compacted clay layers
are often replaced with a geosynthetic clay
liner (GCL). The leakage detection system
normally made up of granular material can
also be replaced by an approved geosynthetic
equivalent alternative such as a cuspated
HDPE drainage sheet or geocomposite drain.
A fluid could pass through this layer in order
to remove VOCs from the system and pos-
sibly cool the liner. HPDE GMs have been
widely used in landfill and waste contain-
ment barriers due to their high resistance to
advective flow of leachate and resistance to
chemical attack (Islam and Rowe 2009).
Although the focus of this study was the
HDPE GM, in most cases the HDPE GM
is used as a primary barrier in conjunction
with other engineered layers to form a
composite lining system that is designed and
engineered to be fit for purpose. The HDPE
Waste body
200 mm stone leachate collection
system
100 mm protection layer of silty
sand or a geotextile of equivalent
performance
600 mm compacted clay liner
(in 4 × 150 mm layers)
150 mm leakage detection
system of granular material
or geosynthetic equivalent
100 mm protection layer of
silty sand or a geotextile of
equivalent performance
200 mm compacted clay liner
In situ soil
GM is an integral part of a composite land-
fill lining system, and in South Africa the
Department of Environmental Affairs has
included the HDPE GM in the liner require-
ments for the successful application of any
waste licence to own or operate a waste
disposal facility.
HDPE GMs remain fit for purpose
in landfill liner applications for up to
1 000 years (Rowe 2005), depending on a
range of factors, such as the period of expo-
sure to active leachate, the height of waste
on the liner, the chemical composition of the
waste being contained and the temperature
of the waste body. The temperature that the
HDPE GM is exposed to has a significant
impact on the service life of the HDPE and
Rowe (2005) has shown that HDPE GM ser-
vice life (or half-life), based on 50% reduction
in tensile strength at break, can be between
565 and 900 years when exposed to tempera-
tures not exceeding 20°C, but can reduce
to as little as 15–20 years when exposed to
temperatures of more than 60°C.
Thus, for an HDPE GM to be a success-
ful barrier in the landfill liner, it needs to
be manufactured, installed and monitored
according to the specifications given by
design engineers and manufacturers.
Volatile organic compounds
Volatile organic compounds (VOCs) are
organic chemical compounds that have high
enough vapour pressures under normal
conditions to significantly evaporate and
enter the atmosphere (Dikshith 2011). Many
carbon-based molecules, such as aldehydes,
ketones and other light hydrocarbons are
VOCs. The term may refer both to well-
characterised organic compounds and to
mixtures of variable composition.
Sometimes VOCs are released into the
environment accidentally, where they can
contaminate soil and groundwater, for exam-
ple the deposition of waste and waste-related
products into engineered waste disposal
landfills equipped with an engineered lining
system. VOCs have been shown to diffuse
through the lining systems of landfills,
resulting in, among other things, ground­
water pollution (Touze-Foltz et al 2011).
Methane is the most commonly known
VOC and, as a greenhouse gas, is a signifi-
cant contributor to environmental pollution.
Methane generally migrates to the surface
of landfills, while other VOCs, such as the
aromatic hydrocarbons (benzene, toluene,
ethylbenzene, xylene (BTEX)) commonly
found in petroleum products, migrate to the
base of the landfill and contribute to ground-
water contamination (Prosser et al 1995).
In recent years, the BTEX hydrocarbons
have attracted much attention because they
constitute one of the most common and
serious threats to groundwater reservoirs
in close proximity to contaminated sites.
The VOCs used for this study are BTEX
(the xylene being p-xylene) and chloroform.
These contaminants were selected because
they are commonly found in landfill lea-
chates (Rowe 2005) and are also significant
contributors to groundwater contamination.
Exposure to VOCs in the short and long
term can be detrimental to human health in
various ways.
Diffusion of VOCs
Diffusion is often referred to as molecular
diffusion and is the net transport of mol-
ecules from a region of higher concentration
to one of lower concentration by random
molecular motion. Due to the nature of
the material and how it is manufactured,
intact HDPE GMs prevent advective flow
of contaminants through their structure,
hence their widespread use in the contain-
ment of water and other liquids. HDPE GMs
do, however, allow movement of certain
contaminants through their structure by
means of molecular diffusion (Rowe 1998).
The diffusive movement of contaminants
through an intact GM with no faults or holes
involves a cooperative rearrangement of the
penetrant molecule and the surrounding
polymer chain segments. For the penetrant
molecule to move into the polymer structure
of the HDPE GM, the process requires the
Journal of the South African Institution of Civil Engineering • Volume 59 Number 1 March 2017
Fluid 1 (Source)
cg1
Geomembrane
c g = Sg f c f
cg2
cf 2
z
Fluid 2 (Receptor)
Figure 2 S chematic diagram of diffusion
through HDPE GM
(adapted from Rowe 1998)
localisation of energy to be available (Rowe
1998). Thus, the diffusive motion requires
energy and depends on the relative mobility
of the penetrant molecules in the contami-
nant or leachate and polymer chains in the
HDPE. In turn, this will depend on tempera-
ture, concentration, the size and shape of
the penetrant and the nature of the polymer
itself (Rowe 1998).
Sangam and Rowe (2001) describe the
molecular diffusion of penetrants such as
BTEX and chloroform through an intact
HDPE GM as a molecular-activated process
that occurs in a series of steps following the
path of least resistance. For diluted aqueous
solutions, as is the case in this study, this
involves three steps (Park & Nibras 1993):
1. Adsorption (partitioning of contaminant
between the inner surface of the HDPE
GM and the medium containing the
contaminant).
2. Diffusion of the permeant through the
HDPE GM.
3. Desorption (partitioning of the contami-
nant between the outer surface of the
HDPE GM and the outer medium).
For water or water-based solutions like those
used in this experiment, the adsorption and
desorption processes can be seen as similar
and inverted (Sangam & Rowe 2001). In the
Keynote Lecture for the 6th International
Conference on Geosynthetics held in Atlanta
in 1998, Professor Kerry Rowe presented a
schematic drawing similar to the one shown
in Figure 2 to illustrate the diffusive trans-
port of contaminants through an HDPE GM.
The figure shows partitioning between the
concentration in solution and the concentra-
tion dissolved in the GM.
Figure 2 illustrates that the process starts
with the removal of the molecule from the
solution fluid onto the surface of the HDPE
GM (Step 1: Adsorption). The sorption of
the permeant onto the polymer (the HDPE
GM) depends on a variety of factors and thus
the extent to which permeant molecules
are sorbed, and their mode of sorption in a
polymer depends upon the activity of the
permeant within the polymer at equilibrium
(Mueller et al 1998). For the simplest case
where the permeant does not interact with
the polymer (as is the case for a HDPE GM) or
tGM
at low concentrations (as in landfill leachates),
the relationship between the concentration in
the fluid and the HDPE GM (solid) is given by
the Nernst distribution function which takes
the linear form shown in Equation 1 (Henry’s
Law) (Sangam & Rowe 2001):
Cg = Sg f Cf(1)
Where: Sg f is called a partitioning or sorption
coefficient [–] and, in principle, is a constant
for the given molecule, fluid, HDPE GM and
temperature of interest, and Cg and Cf are
the concentrations of the permeant in the
HDPE GM and the fluid respectively.
For organic compounds in aqueous
solution, like BTEX and/or chloroform in
water, the value of Sg f is strongly related to
the solubility of the compound of interest in
water (Rowe 1998). The lower the solubility
in water, the higher the affinity of the HDPE
GM to attract the compound, and thus the
higher Sg f will be when in aqueous solution
(Rowe 1998). Compounds with high solubil-
ity thus generally give lower Sg f values.
The process ends with desorption (Step
3: Desorption), which is similar to adsorp-
tion and, for an aqueous solution in contact
with a HDPE GM, it can be assumed that
Equation 1 also holds true, meaning the par-
titioning coefficient for adsorption and des-
orption of BTEX and chloroform in aqueous
solutions are equal (Sangam & Rowe 2001).
The diffusion process in the HDPE GM
happens between the adsorption and desorp-
tion processes and can be explained by Fick’s
First Law:
∂cg
f = –Dg (2)
∂z
Where: f is the rate of transfer per unit area
[ML–2T–1] (typically mg per m2 per second),
Dg is the diffusion coefficient in the HDPE
GM [L2T–1] (typically m2 per second), Cg is
the concentration of the substance that is
diffusing, and z is the direction parallel to
the direction of the diffusion (typically the
thickness of the HDPE GM).
∂c / ∂z is thus the concentration gradient,
and in transient state, allowing for the con-
servation of mass, the governing differential
equation is given by Fick’s Second Law (Rowe
1998):
∂cg ∂ 2 cg
= Dg (3)
∂t ∂z 2
37
This equation needs to be solved for the
appropriate boundary and initial conditions
to obtain the diffusion coefficient of the
solution / HDPE GM system at equilibrium.
To measure the concentration change in
the HDPE GM when doing diffusion tests is
difficult, so it is useful to express the diffu-
sion equations in terms of the concentration
in adjacent solutions (Sangam & Rowe 2001).
Equation 1 gives the relationship between
the concentrations in the GM and the
adjacent fluid. Equation 3 gives the flux (dif-
fusion) within the HDPE GM, so substituting
Equation 1 into Equation 2 gives the flux
on one side of the HDPE GM to a similar
fluid on the other side of a HDPE GM (Rowe
1998), i.e. Equation 4.
∂cg ∂cf ∂cf
f = –Dg = –Sg f Dg = –Pg
∂z ∂z ∂z
Where: Pg = Sg f Dg(5)
Pg gives the relationship between the diffu-
sion coefficient and the sorption coefficient,
and is referred to in polymer literature as
the permeability coefficient (Sangam &
Rowe 2001). This permeability coefficient
(Pg) should not be confused with the soil
mechanics term coefficient of permeability,
which more often is called the hydraulic
conductivity, or the intrinsic permeability of
a porous medium. It has nothing to do with
Darcy’s Law or the flow through the open
voids within porous media, but accounts for
the effects of both diffusion and partitioning.
Based on Equation 5, the mass flux
across an HDPE GM of thickness tGM is thus
given by:
∆cf
f = Sg f D g (6)
tGM
Where: Sg f Dg can be replaced by Pg
(Equation 5) and where Δcf is the difference
in concentration in the fluid on either side of
the GM (cf1 and cf2 in Figure 2).
The purpose is thus to determine the Sg f
and Dg (and thus Pg) values of the system in
question, and compare them to the values
found in the literature, before trying to prove
that VOCs can be extracted successfully
through a pervious zone in the liner system.
For the purpose of this study, it was the
intention to draw air through a pervious
zone in the liner system, thereby removing
VOCs that would have contaminated the
groundwater. This is simulated by separat-
ing the source and receptor volumes using
two HDPE GMs with an air void between
them. The two GMs would be identical and
the theory applied for the calculation of
38
the sorption and diffusion coefficients will
remain. A study by McWatters and Rowe
(2009) investigated the transport of VOCs
through GMs from both aqueous and vapour
phases and found that: “… diffusive transport
of VOC contaminants through geomem-
branes in a simulated landfill environment
is identical despite the phase they originated
from, simplifying the analysis of contami-
nant transport”.
This principle would be adopted
in the work for this dissertation study,
indicating that diffusive transport should
still occur across a system similar to the
description above.
Calculating coefficients
in diffusion process
When undertaking diffusion tests the sorp-
(4)
tion (Sg f ) and diffusion (Dg) coefficients need
to be calculated in order to understand and
comment on the diffusion process.
Sg f is the Henry’s coefficient, and is also
called a solubility, sorption or partitioning
coefficient (Rowe 1998). It is the ratio of the
concentration of the chemical in the HDPE
GM at equilibrium to the concentration
of the chemical in the solution in contact
with the HDPE GM (Park & Nibras 1993).
Sg f is most often unitless and, when doing
diffusion tests where the concentration of
contaminants in the source and receptor is
monitored over time, can be calculated using
Equation 7, and can then be used to infer the
diffusion coefficient (Dg) using the computer
software program POLLUTE® which was
first developed by R K Rowe and J R Booker
in 2004 (POLLUTE 2004). This program
implements a one-and-a-half dimensional
solution to the advection-dispersion equation
(Equation 3).
The diffusion coefficient (Dg), or so-
called diffusivity, has the dimensions of
[length2 time–1], which result from the
underlying kinetic theory. It is a proportional
constant between the molar flux due to
molecular diffusion and the gradient in the
concentration of the species (or the driving
force for diffusion). Generally, it is prescribed
for a given pair of species, but for a multi-
component system, it is prescribed for each
pair of species in the system.
The higher the diffusivity (of one
substance with respect to another), the
faster they diffuse into each other; thus the
higher the diffusion coefficient of the VOC
in question, given a certain GM and con-
centration profile, the faster diffusion will
occur through the GM into the underlying
groundwater.
When the flux across one HDPE GM is
investigated, the Sg f value can be determined
in various ways, as described by Rowe (1998),
Journal of the South African Institution of Civil Engineering • Volume 59 Number 1 March 2017
and Sangam and Rowe (2001). For VOCs,
however, the best method is the diffusion
test method, which is the diffusion from
solution on one side of the GM to solution
on the other side, and monitoring the change
in concentration in the source and receptor
over time until equilibrium is reached (no
significant change in the concentrations in
the source and receptor volumes). The value
for Sg f is then calculated using Equation 7
(Rowe 1998).
CfoVs – CfF(Vs + Vr ) – ∑Vi Ci
Sg f = (7)
AtGMCfF
Where:
cfo is
 the initial concentration of fluid in
the source reservoir [ML–3]
Vs ,Vr are the volumes of the source and
receptor reservoirs [L3]
CfF is the final equilibrium concentration
in the source and receptor reservoirs
[ML–3]
ΣViCi is the mass removed by sampling
events [M] (Vi and Ci being the vol-
ume and concentration removed at
each sampling event)
A is the area of the GM through which
diffusion occurs [L2]
tGM is the thickness of the GM [L].
The diffusion coefficient (Dg) is then
inferred by using Equation 3 and the varia-
tion in source and receptor concentrations
with time (Fick’s Second Law) at the given
boundary conditions. This is done using
POLLUTE v7®, which solves the one-dimen-
sional contaminant migration equation sub-
ject to boundary conditions at the top and
bottom of the GM being modelled (Sangam
& Rowe 2005).
experimental METHODOLOGY
Laboratory tests were carried out at the
University of Pretoria in South Africa. The
tests undertaken were based on the methods
used by Sangam and Rowe (2001).
Sorption tests
Sorption tests were done to determine the
sorption coefficient (Sgf ) for the HDPE GM
and permeant in question. The coefficient is
defined as the ratio of the concentration of
the chemical in the HDPE GM at equilibrium
to the concentration of the chemical in the
solution in contact with the HDPE GM (Park
& Nibras 1993). Sorption tests were done
using glass vials with sampling caps, about
80 mm high and 50 mm diameter, as shown
in Figure 3. The glass sorption cells are
similar in shape and size to the glass sorption
cells used by Sangam and Rowe (2001).
 Sampling port
Fill to mark on neck at
first filling (120 ml)
Artificial leachate solution
surrounding GM
80 mm
Immersed geomembrane
50 mm
Figure 3 Sorption test vials used
Filling port
Source cell, 100 mm high,
70 mm diameter
Sampling port with septum
Lock screws with wingnuts
HDPE geomembrane
Sampling port with septum
Receptor cell, 70 mm high,
70 mm diameter
Figure 4 Diffusion test cell
Tests were done on both the 1 mm
and 2 mm GM using an aqueous solution
containing BTEX and chloroform. The
tests were done in duplicate for both the
1 mm and the 2 mm GMs, and two control
cells were included to measure and assess
losses. The control vials were identical to
the test vials, except that they contained no
GM. Experiments were performed at room
temperature in the laboratory (24 ± 2°C)
and 1 ml samples were taken at days 0, 2, 5,
12 and 21. Samples were placed into glass
gas chromatograph (GC) sampling vials for
testing in the GC (GC-MS). The concentra-
tions of contaminants were monitored until
the equilibrium concentration was reached.
The system was deemed to be in equilibrium
if no significant attributable change in
concentrations occurred for consecutive
sampling events. The sorption coefficient
of each contaminant was then calculated
using Equation 7. The synthetic leachates
used in the sorption vials were made up in
the laboratory using laboratory-grade BTEX
and chloroform made by SAAR Chem-Trade
(Pty) Ltd.
Diffusion tests
Diffusion tests were done following the
examples given in the work by McWatters
and Rowe (2009) to determine the rate of
diffusion of the VOCs through the HDPE
GM by measuring the change in concentra-
tions of solutions on either side of the HDPE
GM. Diffusion tests were carried out in
three phases.
■■ Phase 1 – Diffusion tests using one
HDPE GM: This test would replicate
work already done by others in order to
Journal of the South African Institution of Civil Engineering • Volume 59 Number 1 March 2017
prove that, by using the equipment and
laboratory setup for this project, diffusion
would take place across a 2 mm intact
HDPE GM separating a source and recep-
tor volume (see Figure 2) of a diffusion
test cell.
■■ Phase 2 – Diffusion tests using two
HDPE GMs: The second phase of testing
replicated the first, with the 2 mm HDPE
GM being replaced by 2 x 1 mm HDPE
GMs separated by an 8 mm gap filled
with air to replicate the top and bottom
of an HDPE cuspated leakage detection
system. The purpose was to prove that
diffusion would take place across both
HDPE GMs (separated by air in the pervi-
ous zone) and still reach groundwater
beneath the liner system. The two HDPE
GMs were assumed identical since they
were cut from the same roll and when
there is no flow through the system the
VOC concentration in the water above
and below the HDPE GMs will reach
equilibrium with the concentration of the
VOC in the air layer, resulting in the Sg f
and Dg values being the same for the two
HDPE GMs.
■■ Phase 3 – Extraction of air: The third
phase replicated Phase 2, with air being
extracted through the gap between the
two HDPE GMs to represent the flow of
a fluid through the leakage detection sys-
tem of a landfill liner. The purpose was to
prove that, by removing the air between
the two HDPE GMs at regular intervals,
the VOCs would be removed from the
system and would not reach the ground-
water. The two HDPE GMs were assumed
identical, but since there would now be a
flow of air through the gap between the
two GMs, the concentration profile would
change, resulting in a change of flux,
which could result in a change in the Dg
values of the two HDPE GMs.
Stainless steel, which has been used by
several investigators to examine the dif-
fusion of BTEX compounds (Sangam &
Rowe 2001) was used to manufacture the
diffusion cells, and the cells were designed
to replicate the diffusion test cells used by
Professor Rowe at the Queens University in
Kingston, Canada. The test cells were made
in South Africa by Interlock Systems and
had the dimensions and properties shown
in Figure 4.
Five cells were made so that tests could be
done in triplicate (for each phase of testing),
with one control cell and one blank cell to
measure losses and outside influences. The
receptor reservoir represented the ground-
water beneath lined landfill facilities and
was filled with deionised water at the start of
testing. The source reservoir represented the
leachate in a lined landfill and was filled with
a prepared synthetic leachate solution using
the filling port.
For the Phase 1 diffusion tests, the
source and receptor cells were separated
only by the HDPE GM, and during Phases 2
and 3 a centre-piece was added to introduce
the pervious zone into the liner system. The
centre-piece was separated from the source
and receptor reservoirs by HDPE GMs so
that the configuration was Source-GM-
Pervious Zone-GM-Receptor. During Phase
3 testing the holes in the centre-piece were
used to introduce air flow to the system.
Using four test cells (A to D), different air
flow rates were introduced in three of the
cells, leaving one cell permanently closed
as for Phase 2 testing in order to provide
a control. The air flows were achieved
by replacing the air in the pervious zone
between the GMs with clean ambient air
once every 24 hours, 72 hours and 7 days.
In addition to the sorption test described
earlier, the Sg f value was also obtained
from the diffusion test using Equation 7
after the completion of diffusion tests at
equilibrium. The diffusion coefficient (Dg)
was then inferred using Equation 3, and
the variation in source and receptor con-
centrations with time (Fick’s Second Law)
at the given boundary conditions using the
software POLLUTE®, which solves the one-
dimensional contaminant migration equa-
tion subject to boundary conditions at the
top and bottom of the GM being modelled
(Sangam & Rowe 2005).
results
Sorption
Table 1 summarises the averaged and cor-
rected Sg f values obtained.
Table 1 S ummary of Sgf values obtained using
different test methods
Aqua sorption Diffusion test
VOC
1 mm 2 mm 1 mm 2 mm
Benzene 27.8 14.1 – 116
Toluene 61.9 198 – 183
Ethylbenzene 87.2 326 – 391
p-Xylene 80.2 102 – 188
Chloroform 25.1 14.2 – –
Diffusion
Phase 1 – Diffusion tests using
one HDPE GM
Concentrations in the source and receptor
cells were measured, averaged and plotted
39
 1.20 0.35

1.00 0.30

0.25
0.80
0.20
ct/ci

ct/ci
0.60
0.15
0.40
0.10

0.20 0.10

0 0
Day 0 Day 3 Day 8 Day 15 Day 22 Day 0 Day 3 Day 8 Day 15 Day 22
p-Xylene Toluene Benzene Ethelbenzene p-Xylene Toluene Benzene Ethelbenzene

Figure 5 Phase 1 – VOC concentration in source volume Figure 6 Phase 1 – VOC concentration in receptor volume

200 200
180 180
160 160
Concentration (mg/l)

Concentration (mg/l)
140 140
120 120
100 100
80 80
60 60
40 40
20 20
0 0
Day 0 Day 3 Day 8 Day 15 Day 22 Day 36 Day 0 Day 3 Day 8 Day 15 Day 22 Day 36
Pollute receptor Pollute source Receptor Source Pollute receptor Pollute source Receptor Source

Figure 7 Phase 1 – Combined output graph for benzene Figure 8 Phase 1 – Combined output graph for toluene

160 160

140 140

120 120
Concentration (mg/l)

Concentration (mg/l)

100 100

80 80

60 60

40 40

20 20

0 0
Day 0 Day 3 Day 8 Day 15 Day 22 Day 36 Day 0 Day 3 Day 8 Day 15 Day 22 Day 36
Pollute receptor Pollute source Receptor Source Pollute receptor Pollute source Receptor Source

Figure 9 Phase 1 – Combined output graph for ethylbenzene Figure 10 Phase 1 – Combined output graph for p-xylene

(initial concentration over measured concen- The Phase 1 diffusion coefficients Table 2 C
 alculated diffusion coefficients (Dg)
tration) against time for the source Figure 5 obtained are then given in Table 2 (chloro- for Phase 1 testing
and receptor Figure 6 volumes. form was not tested during Phase 1). VOC Diffusion coefficient in m2/s
The concentration versus time output
Benzene 9.26 × 10 –13
graphs that POLLUTE® produces if the Phase 2 – Diffusion tests using
methodologies described earlier were cor- two HDPE GMs Toluene 8.68 × 10 –13
rectly followed, were combined with the Concentrations in the source and receptor Ethylbenzene 1.39 × 10 –12
actual laboratory test results and are shown cells were measured, averaged and plot-
p-Xylene 2.32 × 10 –12
in Figures 7 to 10 for Phase 1 testing. ted (initial concentration over measured

40 Journal of the South African Institution of Civil Engineering • Volume 59 Number 1 March 2017
 1.40 concentration) against time for the source
Figure 11 and receptor Figure 12 volumes.
As for Phase 1 testing, the diffusion coef-
1.20
ficient Dg was determined using POLLUTE®.
The concentration versus time output graphs
1.00
that POLLUTE® produced were combined
with the actual laboratory test results and
0.80 then plotted as shown in the example for
ct/ci

benzene in Figure 13.

0.60 The Phase 2 diffusion coefficients
obtained are then given in Table 3.
0.40
Table 3 C
 alculated diffusion coefficients (Dg)
for Phase 2 testing
0.20
VOC Diffusion coefficient in m2/s
0 Benzene 8.10 × 10 –13
Day 0 Day 5 Day 12 Day 19 Day 28 Day 41 Day 53 Day 86
p-Xylene Toluene Benzene Ethylbenzene Chloroform Toluene 8.10 × 10 –13

Ethylbenzene 5.79 × 10 –13

Figure 11 Phase 2 – VOC concentration in source volume
p-Xylene 8.10 × 10 –13

0.35 Chloroform 5.79 × 10 –13

0.30 Phase 3 – Extraction of air

Concentrations in the source and receptor
0.25 cells were measured and plotted (initial
concentration over measured concentration)
0.20 against time for the source and receptor
volumes of each cell. The results showed that
ct/ci

0.15
the VOC concentrations in the source gradu-
ally decreased while concentrations in the
receptor gradually increased over the 48-day
0.10
testing period. When looking at the graphs
of concentrations (initial concentration over
0.05 measured concentration) against time per
individual VOC in the various cells, another
0 trend becomes clear. Figures 14 and 15 show
Day 0 Day 5 Day 12 Day 19 Day 28 Day 41 Day 53 Day 86
this trend for benzene, but the trend was also
p-Xylene Toluene Benzene Ethylbenzene Chloroform
the same for the other VOCs (not shown due
to article space constraints).
Figure 12 Phase 2 – VOC concentration in receptor volume These figures show that concentrations
of the VOCs increase more significantly in
12 the receptor volume of Cell A where no air
extraction took place, compared to the other
cells’ charts that represent various air extrac-
10
tion rates. The concentration versus time
output graphs that POLLUTE® produced
were combined with the actual laboratory
Concentration (mg/l)

8
test results and are shown in Figures 16 to 19
for Phase 3 testing, again using only benzene
6
as an example.
For Phase 3 testing the diffusion coef-
4 ficients obtained through POLLUTE® are
given in Table 4 on p 44.

DISCUSSION
0 Phase 1 results show that over the 22-day
Day 0 Day 5 Day 12 Day 19 Day 28 Day 41 Day 53 Day 86 diffusion test period, the VOC concentra-
Source Receptor Pollute source Pollute recepter tions in the source decreased and the VOC
concentrations in the receptor increased.
Figure 13 Phase 2 – Combined output graph for benzene The detection of VOC concentrations in the

Journal of the South African Institution of Civil Engineering • Volume 59 Number 1 March 2017 41
receptor started on day 8 and increased to 1.20
between 10% and 30% of the original source
concentration at day 22. The VOC concen-
trations in the source immediately decreased 1.00
as the VOC sorbed onto the HDPE GM, and
gradually decreased over the 22-day testing 0.80
period to between 5% and 30% of the original
source concentration. Measured losses can

ct/ci
be attributable to sorption of the VOCs onto 0.60
items such as the stainless steel cell, the
septa, the gaskets or the screw in the filling
0.40
port, but since great care was taken to limit
losses due to sorption to these areas, the
most plausible reason for the losses would be 0.20
due to the sampling process. Phase 1 testing
met its objective of proving that the VOCs
0
in question diffuse from the source, through Day 0 Day 5 Day 11 Day 18 Day 25 Day 32 Day 39 Day 48
the 2 mm HDPE GM, into the receptor that Cell A Cell B Cell C Cell D
represents the groundwater, and the diffu-
sion coefficients obtained compare well with Figure 14 Phase 3 – Concentrations of benzene in source volume
those from literature.
Phase 2 results show that over the 86-day
0.12
diffusion test period, the VOC concentra-
tions in the source decreased and the VOC
concentrations in the receptor increased. 0.10
The detection of VOC concentrations in
the receptor were evident from the samples
0.08
taken on day 5 already, and increased to
between 15% and 32% of the original source
concentration at day 86. The trend in the
ct/ci

0.06
data shows an increase in the receptor and a
decrease in the source concentrations over
time, indicating that diffusion took place 0.04
across the divide between the source and
receptor. The VOC concentrations in the 0.02
source immediately decreased as the VOC
sorbed onto the HDPE GM, and continued
to decrease gradually over the 86-day testing 0
Day 0 Day 5 Day 11 Day 18 Day 25 Day 32 Day 39 Day 48
period to between 80% and 20% of the origi-
Cell A Cell B Cell C Cell D
nal source concentration. It is also evident
that the concentrations of chloroform in
the source reduced at a slower rate than the Figure 15 P
 hase 3 – Concentrations of benzene in receptor volume
other VOCs, indicating that it would take
longer for the chloroform in the system to 16
reach equilibrium. It took longer to reach
equilibrium in the system than for Phase 1 14
testing, since the sorption and diffusion
process had to take place over two HDPE 12

GMs and the 8 mm air-filled pervious zone.

Concentration (mg/l)

10
For diffusion to occur through HDPE GM
separating the receptor from the pervious
8
zone, the concentration of the VOCs in
the pervious zone had to be higher than in
6
the receptor to drive the diffusive process.
Phase 2 testing proved that the diffusion
4
of BTEX and chloroform takes place from
source to receptor across a divide consisting 2
of two 1 mm HDPE GMs separated by an
air-filled pervious zone. 0
Phase 3 results show that concentrations Day 0 Day 5 Day 11 Day 18 Day 25 Day 32 Day 39 Day 48
in the source volumes decreased over the Source Receptor Pollute source Pollute receptor
testing time to about 20% of the original
source concentration. Chloroform is the Figure 16 P
 hase 3 – Combined output graph: Cell A benzene (no extraction)

42 Journal of the South African Institution of Civil Engineering • Volume 59 Number 1 March 2017
 8 exception and its concentration reduced
to about 50% of the initial concentration.
7 This is very similar to the data shown on
the source graph of Cell A (no air flow),
6 indicating that the reduction in source
concentrations are comparable, regardless
Concentration (mg/l)

5 of air flow through the pervious zone, and

that the assumption to use the same sorption
4 coefficient in the modelling of Phase 2 and
Phase 3 work was sound. Results also show
3 that VOC concentrations in the receptor
volumes of Cells B, C and D increased over
2 the testing period, indicating that, even with
airflow through the system, concentrations
1
of BTEX and chloroform were observed in
the receptor. However, the concentrations
0
Day 0 Day 5 Day 11 Day 18 Day 25 Day 32 Day 39 Day 48 of the VOCs in the receptor volumes were
Source Receptor Pollute source Pollute receptor at most 0.9% (average 0.4%) of the original
source concentrations compared to 20%
Figure 17 P
 hase 3 – Combined output graph: Cell B benzene (extraction every 24 hours) found during Phase 2 tests. This indicates
that airflow resulted in diffusion taking place
significantly slower. Due to the very low
7
VOC concentrations measured in the recep-
tor volumes of Cells B, C and D, the graphs
6 look slightly distorted and trend identifica-
tion is difficult. The receptor graphs showing
5 the concentration profile per cell for each
Concentration (mg/l)

individual VOC against time indicate that

4 the concentrations measured in the receptor
volumes of Cell A, where airflow was not
3
introduced, is much higher than the concen-
trations measured in the receptor volumes of
Cells B, C and D, again indicating that dif-
2
fusion took place significantly slower in the
cells where airflow was introduced. The aim
1 of Phase 3 was to prove that, by introducing
airflow into the pervious zone between the
0 two 1 mm HDPE GMs, the concentration
Day 0 Day 5 Day 11 Day 18 Day 25 Day 32 Day 39 Day 48
of VOCs in the receptor volume (due to
Source Receptor Pollute source Pollute receptor
diffusion through the HDPE GM) could be
reduced significantly, and the results indicate
Figure 18 P
 hase 3 – Combined output graph: Cell C benzene (extraction every 72 hours) that this aim was comfortably achieved.

8
CoNCLUSIONS AND
7 RECOMMeNDATIONS

6 Conclusions
Various studies have shown that VOCs can
Concentration (mg/l)

5
penetrate even the most well designed liners
of waste containment facilities to pollute the
4
groundwater. The method of penetration is
advection and/or diffusion, with the main
3
contributor to pollution of groundwater
beneath landfills being diffusion.
2
Phase 1 of the tests undertaken for this
1
study showed that BTEX diffuses through a
2 mm HDPE GM over time, with significant
0 concentrations found in the receptor volumes
Day 0 Day 5 Day 11 Day 18 Day 25 Day 32 Day 39 Day 48 of diffusion test cells specially made for this
Source Receptor Pollute source Pollute receptor project. This confirmed studies undertaken
by many researchers in the past. Phase 2
Figure 19 P
 hase 3 – Combined output graph: Cell D benzene (extraction every week) testing proved that diffusion of BTEX and

Journal of the South African Institution of Civil Engineering • Volume 59 Number 1 March 2017 43
Table 4 Calculated diffusion coefficients (Dg) for Phase 3 testing
VOC Cell A
Dg(GM1=GM2) DgGM1
Benzene 1.04 × 10 –12 1.16 × 10 –10
Toluene 1.04 × 10 –12 1.16 × 10 –10
Ethylbenzene 1.15 × 10 –12 1.16 × 10 –10
p-Xylene 1.15 × 10 –12 1.16 × 10 –10
Chloroform 9.26 × 10 –13 1.16 × 10 –13
chloroform takes place from a source to a
receptor reservoir separated by two 1 mm
GMs with an air-filled pervious zone between
them. Phase 3 proved that by extracting air
through a pervious zone beneath the GM
component of a landfill liner, the concentra-
tion of VOCs present in the underlying
groundwater can be reduced, since the air
removed from the system also removes the
majority of VOCs. This phase of testing also
confirmed that more frequent removal of air
further reduces the VOC concentrations in
the receptor, thus implying that a constant
airflow through a pervious zone in a landfill
liner can significantly reduce concentrations
of VOCs in the groundwater beneath landfills
and waste containment facilities.
Recommendations
To add to the work done for this study it is
recommended that extraction fluids other
than air be used at more frequent rates. This
could include the use of a GCL in testing to
understand whether continuous hydration
of the bentonite in the GCL will benefit
the reduction in contaminant transport.
Also, the VOCs that were extracted from
the system in the tests undertaken for this
study were not captured or measured. If the
VOCs are removed from beneath the liner,
they need to be routed somewhere (releasing
44 Journal of the South African Institution of Civil Engineering • Volume 59 Number 1 March 2017
Diffusion coefficient in m2/s
Cell B Cell C
DgGM2 DgGM1
2.31 × 10 –14 1.04 × 10 –10
2.31 × 10 –14 1.04 × 10 –10
2.31 × 10 –14 1.04 × 10 –10
2.31 × 10 –14 1.04 × 10 –10
5.79 × 10 –14 1.16 × 10 –13
them into the atmosphere does not protect
the environment). It needs to be further
investigated how to trap the VOCs and treat
them after removal.
References
Dikshith, T S S 2011. Handbook of Chemicals and
Safety. Boca Raton, FL: Taylor and Francis Group
(CRC Press).
Edil, T B 2003. A review of aqueous-phase VOC
transport in modern landfill liners. Waste
Management Journal, 23: 561–571.
Islam, M Z & Rowe, R K 2009. Permeation of BTEX
through unaged and aged HDPE geomembranes.
Journal of Geotechnical and Geoenvironmental
Engineering, 135(8): 1130–1140.
McWatters, R S & Rowe, R K 2009. Transport of volatile
organic compounds through PVC and LLDPE
geomembranes from both aqueous and vapour
phases. Geosynthetics International, 16(6): 468–481.
Mueller, W, Jakob, I, Tatzky, G R & August, H 1998.
Solubilities, diffusion and partitioning coefficients
of organic pollutants in HDPE GMs: Experimental
results and calculations. Proceedings, 6th
International Conference on Geosynthetics, 25–29
March, Atlanta, GA, 239–248.
DWAF (Department of Water Affairs and Forestry,
South Africa) 1998. Minimum requirements for the
handling, classification and disposal of hazardous
waste. Pretoria: DWAF.
Cell D
DgGM2 DgGM1 DgGM2
2.31 × 10 –14 9.84 × 10 –11 2.31 × 10 –14
2.31 × 10 –14 9.84 × 10 –11 2.31 × 10 –14
2.31 × 10 –14 9.84 × 10 –11 2.31 × 10 –14
2.31 × 10 –14 9.84 × 10 –11 2.31 × 10 –14
5.79 × 10 –14 1.16 × 10 –13 5.79 × 10 –14
Park, J K & Nibras, M 1993. Mass flux of organic
chemicals through polyethylene geomembranes.
Water and Environmental Research, 65(3), 227–237.
Prosser, R & Janechek, A 1995. Landfill gas and
groundwater contamination. Paper presented at the
October 1995 American Society of Civil Engineers
(ASCE) Convention, 23–27 October, San Diego,
California.
POLLUTE® v7 2004 Reference Guide. Ontario,
Canada: GAEA Technologies Ltd, R.K Rowe and
J.R Booker.
Rowe, R K 1998. Geosynthetics and the minimization
of contaminant migration through barrier
systems beneath solid waste. Keynote lecture,
6th International Conference on Geosynthetics,
25–29 March, Atlanta, GA.
Rowe, R K 2005. Long-term performance of
containment barrier systems: 45th Rankine Lecture.
Geotechnique, 55(9), 631–678.
Sangam, H P & Rowe, R K 2001. Migration of dilute
aqueous organic pollutants through HDPE
geomembranes. Geotextiles and Geomembranes,
19(6), 329–357.
Sangam, H & Rowe, R K 2005. Effect of surface
fluorination on diffusion through a high density
polyethylene geomembrane. Geotechnical and
Geoenvironmental Engineering, 19, 329–357.
Touze-Foltz, N, Rosin-Paumier, S & Mazeas, L 2011.
Diffusion of volatile organic compounds through an
HDPE geomembrane. Paper presented at the Geo-
Frontiers Congress, ASCE 2011, 1121–1130.

Updated provisions of SANS
10160-4 for steel structures
C P Roth, A Gebremeskel
A forthcoming revision of SANS 10160-4 (Seismic actions and general requirements for
buildings) addresses the omission of structural steel design provisions from the standard, as well
as contradictions between SANS 10160 and SANS 10162, the steel design standard. This note
discusses the background to the proposed provisions.
Introduction
Basic provisions for the design of buildings
and other structures to withstand earth-
quake loads have been available in the SANS
codes since 1989. However, compliance with
the requirements has not been vigorously
enforced by the authorities and owners over
the past two decades. This is largely due to
lack of awareness by design engineers and
academics. In some cases it has also been
caused by skepticism about the level of seis-
mic risk that exists in South Africa.
Despite having relatively low seismicity
by global standards, South Africa has one
of the most up to date codes in Africa when
it comes to design for seismic resistance.
Whereas the code addresses the design of
concrete and masonry structures adequately,
a gap exists when it comes to the design of
structures that are framed in structural steel.
Moreover, the fact that South African codes
are derived from different code cultures,
namely European and North American, at
times causes confusion when attempting to
design structures for seismic resistance.
Efforts have been made by the Southern
African Institute of Steel Construction
(SAISC), in collaboration with the South
African Bureau of Standards TC98 to address
the gap described above. The proposed
provisions for the design of steel structures
to resist seismic loading are expected to
be published in a forthcoming revision of
SANS 10160‑4 (Seismic actions and general
requirements for buildings). This note dis-
cusses the background to these provisions,
for the benefit of practising engineers and
future code developers.
CURRENT CODE PROVISIONS
Current seismic design practice allows struc-
tures to dissipate energy by nonlinear mate-
rial behaviour during major seismic events.
In order to avoid undertaking nonlinear
dynamic analysis, the codes allow a designer
to do a linear elastic analysis and reduce
Roth CP, Gebremeskel A. Updated provisions of SANS 10160-4 for steel structures.
J. S. Afr. Inst. Civ. Eng. 2017:59(1), Art. #TN75, 3 pages. http://dx.doi.org/10.17159/2309-8775/2017/v59n1a6
the resulting forces by a behaviour factor to
reflect the energy dissipated by nonlinear
behaviour. The value of the behaviour factor
depends on the material type and structural
system used. Structures with more inherent
ductility and energy dissipation capacity have
larger behaviour factors, which allow greater
reduction in design forces.
The original SANS 10160‑4:2010
(SANS 2010) and its first amendment
SANS 10160‑4:2011 (SANS 2011a) both
allow a behaviour factor q of 5.0 for ordinary
braced steel frames, and 4.5 for steel frames
in a moment-resisting frame system. No
specific detailing requirements are specified
other than “Detailing rules of SANS 10162‑1
and SANS 10162‑2 shall apply” and “Also
refer to EN 1998‑1 for detailing rules of
structural steelwork.” This is different to
the situation for masonry and reinforced
concrete structures, where higher behaviour
factors are only allowed for structures com-
plying with certain detailing requirements
which are given in Annexes A and B. For
example, the behaviour factor for a building
frame system with reinforced concrete shear
walls increases from 2.0 to 5.0 when specific
detailing rules are applied. It is therefore
inconsistent for the code to allow relatively
high behaviour factors for steel structures
without providing specific detailing rules for
seismic design in SANS 10160‑4.
This inconsistency was recognised when
the original SANS 10160‑4 was written.
Wium (2009) identified quantification and
confirmation of the behaviour factor for
structural steel as a research need in his
discussion of the background to the develop-
ment of the original SANS 10160‑4.
The Eurocode EN 1998‑1 (EN 1998) allows
behaviour factors in the range of 4.5–5.0 only
for steel structures properly analysed and
detailed in accordance with the requirements
for “Dissipative structural behaviour”. For
steel structures where no specific detailing
rules are applied, a value of 1.5 is recom-
mended in Section 6 of EN 1998‑1.
TECHNICAL NOTE
Journal of the South African
Institution of Civil Engineering
ISSN 1021-2019
Vol 59 No 1, March 2017, Pages 45–47, Note 75
PrOF CHRIS ROTH (Pr Eng, FSAICE) is Associate
Professor in Civil Engineering at the University of
Pretoria, working in the discipline of structural
engineering. He started his career in consulting
engineering before joining the University of
Pretoria, and is the current chairman of SANS
Technical Committee 98 Sub-committee 1 on
“Basis of Design and Actions (including
Earthquake Design)”. He obtained a BEng degree in civil engineering at
Stellenbosch University, and an MS and PhD at Cornell University. His interests
are in structural reliability and structural analysis.
Contact details:
Department of Civil Engineering
University of Pretoria
Pretoria
0002
South Africa
T: +27 12 420 2185
E: chris.roth@up.ac.za
AMANUEL GEBREMESKEL (PE, SE) is registered as
a Structural Engineer in the United States. He
received degrees in Engineering and Business
Administration from the University of
Minnesota. He presently directs technical
publications and research for the Southern
African Institute of Steel Construction and chairs
the SANS 10160-4 Seismic Sub-committee. He
has been designing buildings and writing national codes in the United States
and South Africa for the past 15 years.
Contact details:
Southern African Institute of Steel Construction
PO Box 291724
Melville
Johannesburg
2109
South Africa
T: +27 11 726 6111
E: amanuel@saisc.co.za
Keywords: seismic design, code development, SANS 10160-4,
support structures, steel design
45
 SANS 10160-4 also contains no provi-
sions for seismic steel support frames for
heavy rigid objects, such as tanks, bins and
similar containers storing liquids, gases
and granular materials. These common
structures are included in the scope of
SANS 10160‑4, which covers “industrial
structures utilising structural systems simi-
lar to those of building structures”. It was
felt that provisions for seismic design of such
structures would be beneficial.
harmony between current
code provisions
In addition to the omission of structural steel
design provisions in SANS 10160‑4 addressed
above, there appears to be a contradiction
between SANS 10160 and SANS 10162.
The SAISC proposed modifications to
both SANS 10160‑4 and SANS 10162‑1
(SANS 2011b) in order to resolve this contra-
diction while addressing the omission.
As stated previously, no specific detailing
requirements are specified in SANS 10160‑4
other than references to SANS 10162‑1,
SANS 10162‑2 (SANS 2011c) and EN 1998‑1.
However, Clause 27 of SANS 10162‑1:2011
refers the reader to Canadian standard
CSA S16 (CSA S16 2014) for seismic design
of steel structures. This contradiction has
arisen because SANS 10162‑1 is based on
Canadian standards, while SANS 10160 is
primarily based on European standards.
For steel structures where no specific
detailing rules are applied, a behaviour
factor of 1.5 is recommended in CSA S16.
This value can be raised to 1.95 in some
cases. These Canadian codes in turn rely
on AISC 360/341/358 (AISC 360/341/358
2010) and the associated American stan-
dards ASCE 7 (ASCE 7 2010) (referred to as
“ASCE 7” in the remainder of the note) and
IBC (IBC 2015) to form the basis for their
structural steel seismic design provisions.
ASCE 7 in turn allows the use of a beha­
viour factor of up to 3 for structural steel
systems that are not specifically detailed
to achieve high ductility. This level of con-
fidence in the inherent ductility of typical
steel structures is borne out by a century
of experience in the western parts of the
United States.
The dominance of American codes in
general when it comes to seismic design can
be attributed to the efforts of the Federal
Emergency Management Agency in exten-
sively collecting, analysing and standardising
seismic response data over the past several
decades. As such South Africa, Canada
and Europe continue to look to American
approaches and standards as a basis for their
own standards.
46 Journal of the South African Institution of Civil Engineering • Volume 59 Number 1 March 2017
PROPOSED CODE FOR BUILDINGS
The intention of the proposed code is not to
provide steel detailing guidelines, but rather
to deal more rationally with steel structures
not specifically detailed for seismic resis-
tance. A conservative option would be to use
a behaviour factor of 1.5 in SANS 10160-4
for such steel structures, as recommended
in EN 1998‑1 and CSA S16. However, it is
clear from the discussion above that such
conservatism is not warranted when there is
sufficient experience and data to support the
more economical behaviour factors that are
used in American codes such as ASCE 7.
Table 12.2‑1 of ASCE 7 defines Response
Modification Coefficients, Overstrength
Factors and Deflection Amplification Factors
for various types of structural systems.
The Response Modification Coefficient
is equivalent to the behaviour factor in
SANS 10160‑4. A value of 3.0 is given for
“Steel systems not specifically detailed for
seismic resistance, excluding cantilever
column systems”, for Seismic Design
Categories B and C.
Cantilever column systems resist lateral
forces solely by the columns acting as canti-
levers supported at the base. Their exclusion
in ASCE 7 is not relevant to the proposed
code, as SANS 10160‑4 Table 4 only includes
braced and moment-resisting frames for steel
structures.
To relate Seismic Design Categories B
or C to SANS 10160‑4 terminology,
the clauses must be examined in more
detail. Table 11.6‑1 of ASCE defines the
Seismic Design Categories in terms of
Risk Categories I–IV and SDS, the “design,
5 percent damped, spectral response
acceleration parameter at short periods”. In
SANS 10160‑4, SDS is equivalent to Sd(T),
TB ≤ T < TC with q = 1, which is graphically
the value of the flat portion of the design
response spectrum in Figure 2. A structure
in Risk Categories I–III falls under Seismic
Design Categories B or C for SDS < 0.50,
and a structure in Risk Category IV falls
under Seismic Design Categories B or C for
SDS < 0.33. These Risk Categories roughly
correspond to the Importance Classes in
Table 3 of SANS 10160‑4. While there are
some differences between classes II and III in
the two codes, the distinction between I–III
on one hand and IV on the other hand is the
same in SANS 10160 and ASCE 7.
In seismic zone 1 as defined in
SANS 10160‑4, ag = 0.1 g and the resulting
Sd(T) for TB ≤ T < TC is shown in Table 1 for
the four different ground types. However, a
more accurate comparison should include
the effect of the redundancy factor ρ of
clause 7.4, which can be taken as 1.2. This
minimum value of the factor was included to
provide an additional factor to increase the
seismic forces, as the specified value for ag,
0.1 g, is lower than the values on the latest
seismic hazard maps (Wium 2010). Values of
ρSd are also included in Table 1.
Table 1 Values of design spectral acceleration
Ground
Sd ρSd
type
1 0.25 0.3
2 0.30 0.36
3 0.29 0.35
4 0.34 0.41
It can be seen that if ρ is excluded, the
maximum value of Sd is 0.34, which is well
under the 0.5 limit for classes I–III to fall in
Design Categories B or C, and very close to
the 0.33 limit for class IV. It may be argued
that essentially all classes fall in Design
Categories B or C, so the value of 3.0 sug-
gested by ASCE 7 is applicable.
If ρ is included, then the maximum
value of ρSd is 0.41, so classes I–III still fall
in Design Categories B or C. Class IV does
not fall in Design Categories B or C, as 0.41
exceeds 0.33.
As the emphasis in the SANS 10160-4
code is more on instilling correct structural
concepts than on extensive calculations
(Wium 2010), it was concluded that the
vast majority of structures fall in Design
Categories B and C, though possibly more
consideration should be given to class IV
structures that are founded on very loose soil
and a lower behaviour factor should be pre-
scribed in such cases. As a behavior factor of
3.0 is already more conservative and realistic
than the existing code values, it was decided
to adopt this value until more research on
South African conditions becomes available.
With q = 3.0, the inelastic displacement
ds according to SANS 10160-4 clause 9.2
would be 0.7q × de = 2.1 × de. From the
Deflection Amplification Factor given in
Table 12.2-1, ASCE 7 prescribes 3.0 × de. For
consistency with ASCE 7 it is recommended
that ds = 3.0 × de be specified in D.2.2 of the
proposed code. The factor in ASCE 7 can be
reduced by the redundancy and importance
factors in certain cases, but it was felt that it
is conservative and simple to leave it as 3.0.
Clause D.2.3 of the proposed code deals
with elements that transfer load from floor or
roof diaphragms to the vertical-lateral force
resisting system and is based on clause 12.10.2
of ASCE 7, which requires “collector elements
and their connections to vertical elements” to
be designed for seismic forces including the
Overstrength Factor, which is given as 3 in
Table 12.2‑1. Clause 12.10.2 of ASCE 7 applies
to structures in Design Category C, and
not to Category B, but no distinction is
made in the proposed code to keep it simple
and conservative.
Support structures for
heavy rigid objects
Guidelines for steel support structures for
heavy rigid objects are included in clause D.3.
They are based on section 15.3 of ASCE 7,
“Non-building structures supported by other
structures” from which the definition of rigid
objects as “objects that have a fundamental
period of less than 0.06 s and weigh over
25% of the combined weight of the object
and supporting structure” is taken. Values
of the Response Modification Coefficients,
Overstrength Factors and Deflection
Amplification Factors used in clauses D.3.2,
D.3.3 and D.3.4 are from Table 15.4-2.
According to section 15.3.2, “The
supporting structure shall be designed
in accordance with the requirements of
Chapter 12 or Section 15.5 as appropriate,
and the R value of the combined system is
permitted to be taken as the R value of the
supporting structural system.” The R value
obtained from Table 15.4-2 for “Elevated
tanks, vessels, bins or hoppers” can cover the
vast majority of support structures that are
designed in South Africa.
Clause D.3.4 deals with connections
between the object and the support structure
and is based on section 15.7.3 of ASCE 7,
with some modifications to the language to
Journal of the South African Institution of Civil Engineering • Volume 59 Number 1 March 2017
make it easier to follow in the SANS context.
In ASCE 7 the factor by which forces are to
be multiplied is given as the “overstrength
factor”, which is given as 2 in Table 15.4-2.
Clause D.3.5 deals with a heavy rigid
object on grillage beams and is based on
section 15.5.5 of ASCE 7, also with some
modifications to the language to make it
easier to follow.
CONCLUSION
A forthcoming revision of SANS 10160-4
addresses the omission of structural steel
design provisions from the standard, as well
as contradictions between SANS 10160 and
SANS 10162, the steel design standard. The
proposed revision is based on the ASCE 7
standard. Relevant parts of ASCE 7 have
been adapted for the new SANS standard.
References
AISC 341 2010. Seismic Provisions for Structural Steel
Buildings. Chicago, IL: American Institute of Steel
Construction.
AISC 358 2010. Prequalified Connections for Special
and Intermediate Steel Moment Frames for Seismic
Applications. Chicago, IL: American Institute of
Steel Construction.
AISC 360 2010. Specification for Structural Steel
Buildings. Chicago, IL: American Institute of Steel
Construction.
ASCE 7 2010. Minimum Design Loads for Buildings and
Other Structures. Reston, VA: American Society of
Civil Engineers.
CSA S16 2014. Design of Steel Structures. Toronto,
Canada: CSA Group.
EN (European Standard) 1998. EN 1998-1 2004.
Eurocode 8: Design of Structures for Earthquake
Resistance. Part 1: General Rules, Seismic Actions,
and Rules for Buildings. Brussels, Belgium: European
Committee for Standardization (CEN).
IBC 2015. International Building Code. Washington,
DC: International Code Council (ICC).
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.
SANS (South African National Standard) 2011a. SANS
10160-4: Basis of Structural Design and Actions for
Buildings and Industrial Structures. Part 4: Seismic
Actions and General Requirements for Buildings,
Amendment 1. Pretoria: SABS Standards Division.
SANS (South African National Standard) 2011b. SANS
10162-1: The Structural Use of Steel. Part 1: Limit-
States Design of Hot-Rolled Steelwork. Pretoria: SABS
Standards Division.
SANS (South African National Standard) 2011c. SANS
10162-2: The Structural Use of Steel. Part 2: Cold-
Formed Steel Structures. Pretoria: SABS Standards
Division.
Wium, J 2009. Background to the development of
procedures for seismic design. In: Retief, J V &
Dunaiski, P E (Eds.), Background to SANS 10160.
Stellenbosch: SUN Press, 167–188.
Wium, J 2010. Background to Draft SANS 20160 (2009):
Part 4. Seismic loading. Journal of the South African
Institution of Civil Engineering, 52(1): 20–27.
47
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 jour-
nal publishes research papers covering all the disciplines
of civil engineering (structural, geotechnical, railway,
coastal/marine, water, construction, environmental,
municipal, transportation) and associated topics that
are relevant to the civil engineering profession, and that
preferably have particular relevance to civil engineering
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
■■ Technical papers are well-researched, in-depth,
fully referenced technical articles not exceeding
6 000 words in length (excluding tables, illustrations
and the list of references). Related papers that deal
with ‘softer sciences’ (e.g. education, social uplift-
ment, etc) are accepted if they are of a technical
nature and of particular interest to the civil engineer-
ing 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.
■■ 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.
■■ 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.
■■ 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
■■ 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 accompa-
nied by an expanded abstract in English.
■■ Original work: Papers and technical notes must be
original contributions. Authors must affirm that sub-
mitted material has not been published previously,
is not under consideration for publication 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 pub-
lication 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 obtain-
ing permission from copyright owners when repro-
ducing material that has been published elsewhere.
Proof of such permission must be supplied.
SUBMISSION PROCEDURES
AND REQUIRED FORMAT
■■ Electronic submission: Manuscripts should be
e-mailed to the editor at verelene@saice.org.za. File
sizes should not exceed 4 MB per e-mail – figures
may for example be sent one by one or in groups not
larger than 4 MB. Manuscripts should not be sent in
PDF format as this precludes reviewing of papers per
track changes.
■■ 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.
■■ 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.
■■ Figures, tables, photos and illustrations: These
should preferably be submitted in colour, as the
journal is a full-colour publication.
■■ Their positions should be clearly marked in the
text as follows: [Insert Figure 1]
■■ Figures, tables, photos, illustrations and equations
should be numbered consecutively and should
appear in the text directly after they have been
referred to for the first time.
■■ 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).
■■ Only those figures and photographs essential to
the understanding of the text should be included.
All illustrations should be referred to in the text.
■■ Figures should be produced using computer
graphics. Hand-drafted figures will not be accept-
ed. Lettering on figures should be equivalent 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.
■■ 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.
■■ Mathematical expressions and presentation of
symbols:
■■ 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.
■■ Symbols should preferably reflect those used in
Microsoft Word Equation Editor or Mathtype,
or should be typed using the Times New Roman
symbol set.
■■ 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, sym-
bols of metric measurement units (m for metres,
s for seconds, etc) and mathematical/trigono-
metrical 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.
■■ Metric measurement abbreviations/units should
conform to international usage – the SI system of
units should be used.
■■ Decimal commas may be used, but decimal points
are preferred.
■■ 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.
■■ 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
■■ 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:
■■ Donald, A W 1954. Pore pressure coefficients.
Géotechnique, 4(4): 143–147.
■■ Terzaghi, K 1943. Practical soil mechanics. New
York: Wiley.
■■ 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.
■■ 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.
■■ Return of amended papers: Papers requiring
amendments will be accepted up to six months after
the referee reports have been sent to authors, after
which the paper will be withdrawn from the system.
FINAL ARTICLE
■■ 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.
■■ 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.
■■ Proofs: First proofs of papers will be sent to authors
in PDF format for verification before publication. No
major rewrites will be allowed, only essential minor
corrections.
 Thanks to our Referees!
The SAICE Journal Editorial Panel would like to thank the persons listed below, all of whom served as referees
during 2016. The quality of our journal is not only a reflection of the level of expertise of participating authors,
but certainly also of the high standard set by our referees.

Dr Elhadi Adam Dr Graham Howell Ms Nicole Nel

Prof Mark Alexander Prof Adesola Ilemobade Mr Iiya-Khairanis Othman
Dr Johann Andersen Dr Hans Ittmann Dr Phil Paige-Green
Dr Joseph Anochie-Boateng Prof Chris James Prof Marius Pretorius
Dr MC Giel Bekker Mr Garth James Dr Pieter Pretorius
Prof Hans Beushausen Mr Malcolm Jaros Dr LDV Roodt
Dr Precious Biyela Prof Kim Jenkins Prof Chris Roth
Mr Deon Bonthuys Dr Petrina Johannes Dr Martin Rust
Prof Walter Burdzik Dr Gary Jones Prof Manu Santhanam
Mr Gavin Byrne Prof Gerrit Jordaan Prof Corné Schutte
Dr Nicol Chang Mr Chris Jurgens Prof Jeff Smithers
Mr Arthur Chapman Dr Denis Kalumba Dr Izak Snyman
Prof Chris Clayton Prof Elsabé Kearsley Dr Nina Steffani
Dr Stefano Corbella Mr John Kelly Mr Chris Swartz
Dr Peter Day Dr Geoff Krige Mr Arthur Taute
Mr Johan de Koker Dr Maxim Kovtun Dr Michelle Theron
Dr Erik Denneman Mr Peter Legg Prof Hlanganani Tutu
Mr Jaco de Villiers Dr Kuinian Li Dr Eunice Ubomba-Jaswa
Dr Hardus Diedericks Mr Jan Louw Mr Danie van der Spuy
Mr Douglas Dorren Mr Pieter Louw Mr Jackie van der Westhuizen
Mr Jonathan Duvel Prof James Maina Mr Anton van Langelaar
Dr OJ Gericke Mr Stephen Mallory Ms Samantha van Staden
Prof Mitchell Gohnert Dr Stefania Merlo Prof Kobus van Zyl
Mr Anton Groenewald Dr Gary Mocke Prof Christo Venter
Mr DJ Hagen Dr Mathetha Mokonyama Dr Nico Vermeulen
Dr Brian Harrison Prof Lavern Moodley Dr Eduard Vorster
Dr Anton Hartman Prof Pilate Moyo Mr Jaco Vorster
Dr Francois Heyns Mr David Mpye Dr Abimbola Windapo
Dr Emile Horak Dr Rachel Mugai Prof Mark Zuidgeest