Volume 65 Number 3

September 2023

QQ  evised Regional Maximum Flood (RMF)

R
method and regionalisation
QQ The status of water for firefighting in South Africa
QQ E xperimental comparison of active seismic surface
wave tests on shallow and deep bedrock sites
QQ Thermal performance of cavities in 3DPC building façades
Volume 65 No 3 September 2023
ISSN 1021-2019 (print) | ISSN 2309-8775 (online)

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

CONTENTS
EDITOR-IN-CHIEF
Dr Peter Day
Consultant, Jones & Wagener (Pty) Ltd
Extraordinary Professor of Geotechnical Engineering
Stellenbosch University
day@jaws.co.za
CO-EDITOR-IN-CHIEF
Prof Gerhard Heymann 2 Revised Regional Maximum Flood (RMF)
University of Pretoria
gerhard.heymann@up.ac.za method and regionalisation
ADMINISTRATIVE EDITOR
Verelene de Koker J A du Plessis, S Masule
Tel +27 11 805 5947, Mobile +27 83 378 3996
verelene@saice.org.za
SUB-EDITORS
Prof EP Kearsley University of Pretoria
20 The status of water for firefighting in South Africa
Prof CP Roth University of Pretoria
Prof A Visser University of Pretoria A Ilemobade
Prof J Wium Stellenbosch University
Prof M Zuidgeest University of Cape Town
BALANCE OF THE JOURNAL EDITORIAL PANEL
Prof Y Ballim University of the Witwatersrand
27 Experimental comparison of active seismic surface
Prof W Burdzik Retired from the University of Pretoria
Prof N de Koker Stellenbosch University
wave tests on shallow and deep bedrock sites
Prof J du Plessis Stellenbosch University
Prof GC Fanourakis University of Johannesburg
M C Ntaote, G Heymann
Dr T Gaspar Sheffield University
Prof M Gohnert University of the Witwatersrand
Dr C Herold Umfula Wempilo Consulting
Prof A Ilemobade University of the Witwatersrand
39 Thermal performance of cavities
Prof SW Jacobsz
Dr C MacRobert
University of Pretoria
Stellenbosch University
in 3DPC building façades
Dr O Nnene University of Cape Town
Prof S Nwaubani University of the Witwatersrand M Moelich, G van Zijl, W de Villiers
Prof W Steyn University of Pretoria
Dr J van der Merwe University of Pretoria
Mr M van Dijk
Prof G van Zijl
University of Pretoria
Stellenbosch University
54 Guidelines for the preparation of
Prof C Venter
Dr E Vorster
University of Pretoria
Zutari
papers and technical notes
Prof R Walls Stellenbosch University
Prof A Zingoni University of Cape Town
PEER REVIEWING / OPEN ACCESS
The Journal of the South African Institution of Civil Engineering,
which is distributed internationally, is a peer-reviewed,
open-access journal licensed under a Creative Commons
Attribution Licence (CC BY-NC-ND). Readers may therefore
freely use and share the content as long as they credit the
original creator and publisher, do not change the material
in any way, and do not use it commercially.

Creative Commons
Attribution Licence

DESIGN AND REPRODUCTION

Marketing Support Services, Ashlea Gardens, Pretoria
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
and the authors 1
TECHNICAL PAPER
Revised Regional Maximum
Flood (RMF) method
JOURNAL OF THE SOUTH AFRICAN
INSTITUTION OF CIVIL ENGINEERING

and regionalisation
ISSN 1021-2019 (print) | ISSN 2309-8775 (online)
Vol 65 No 3, September 2023, Pages 2–19, Paper 1456

PROF J A (KOBUS) DU PLESSIS (Pr Eng,

FSAICE, FIMESA) has more than 34 years of
experience in water engineering, of which
the past 21 years were spent in the Civil
Engineering Department at Stellenbosch
University, where he is responsible for
Hydrology and Environmental Engineering.
He has a special interest in integrated
management of water resources in South Africa as applied by local
authorities, as well as flood hydrology. He obtained his PhD (Water
Governance), MEng (Water Resource Management) and BEng (Civil) from
Stellenbosch University. He presently serves as an Executive Committee
member of the Institute of Municipal Engineering of Southern Africa
(IMESA), and as a member of the Education and Training Panel of the
South African Institution of Civil Engineering (SAICE).
Contact details:
Department of Civil Engineering
Stellenbosch University
Private Bag X1
Matieland 7600
South Africa
E: jadup@sun.ac.za
SIMENDA MASULE holds a BSc (Civil
Engineering) from the University of Namibia
and an MEng (Civil Engineering) from
Stellenbosch University. His research so far
has focused on the review of a flood
estimation method (Regional Maximum
Flood (RMF)), which included the update of
a historical maximum flood peaks database
with the latest flood peak readings and with a re-regionalisation of the
RMF according to measurable catchment variables.
Contact details:
Department of Civil Engineering
Stellenbosch University
Private Bag X1
Matieland 7600
South Africa
E: simamasule@gmail.com
J A du Plessis, S Masule
South Africa receives an average annual rainfall of about 450 mm. Hydraulic structures are
typically constructed to either store or manage the excess water resulting from runoff. These
hydraulic structures are designed and evaluated to withstand a particular flood peak that can
occur in its catchment area. Adequate flow or rainfall records may often not be available to
enable a reliable flood estimation. In South Africa an empirical estimation method (the Regional
Maximum Flood (RMF)) that utilises regional envelope curves to estimate the maximum
observed flood peaks that can be expected in a region, is available. The RMF method adopted
by Kovács in 1980, and revised in 1988, is robust and simple to use. The current research revisits
the method as applicable to South Africa, and presents an update of the method, taking more
than 30 years of additional data and a revised regionalisation approach into consideration.
Numerous previous researchers evaluated the RMF method and concluded that the method
needs to be updated. It was identified that recently observed flood peaks exceeded the existing
RMF envelopes. It was further identified that the Kovács regionalisation process is inconsistent,
and a revised regionalisation approach was proposed. The revised regionalisation resulted in
15 RMF K regions and their associated envelope curves. The new RMF K regions are smaller,
with the highest K value equal to 5.8 and the lowest value 2.8. The recommended envelope
curves were drawn 15% above the maximum observed flood peaks for each region, allowing for
possible future climate impacts. The revised RMF envelope curves are considered to adequately
represent the RMFs in South Africa and are therefore applicable for determining the expected
maximum regional flood at any site in South Africa.
INTRODUCTION and presents an improved approach in
South Africa has experienced significant developing a new set of regional curves,
floods, including the Laingsburg and using an extended data set, for use to esti-
southeastern Cape floods in 1981, floods mate maximum regional flood peaks as an
from the cyclone Domoina in 1984, upper limit value to be expected.
the KwaZulu-Natal floods in 1987, the
Orange River Basin floods in 1988, and Estimation methods
the Limpopo floods in 2000 (Görgens et al Probabilistic, deterministic and empirical
2006). In addition, the Western Cape expe- flood estimation methods were developed
rienced flooding in 2005, and the Eastern by various institutions from the late 1960s,
Cape and Free State experienced flooding using only data available at that time.
in 2011 (Smithers 2012). It is critical to These methods estimate flood peaks using
understand the magnitude and recurrence either historical flood peaks or rainfall
interval of these extreme floods to plan, as the primary hydrological input data
design and operate hydraulic structures. In (Smithers 2012).
South Africa the methods used to estimate
these flood peaks are based on probabilis- Probabilistic methods
tic, deterministic and/or empirical meth- Probabilistic methods estimate flood peaks
ods (Kovács 1988; Parak & Pegram 2006; by performing a flood frequency analysis
Van der Spuy & Rademeyer 2010; Van Dijk of historically observed streamflow
et al 2013; Gericke & Du Plessis 2013). records, resulting in a direct estimation of
This research focused on the update of flood peaks for various return period or
the RMF as one of the empirical methods, exceedance probabilities. Their application

Keywords: Regional Maximum Flood (RMF), flood peaks,

regionalisation, safety factor, flood zone, transition zone, Du Plessis JA, Masule S. Revised Regional Maximum Flood (RMF) method and regionalisation. J. S. Afr. Inst. Civ. Eng. 2023:65(3),
Art. #1456, 18 pages. http://dx.doi.org/10.17159/2309-8775/2023/v65n3a1
storm zone
The Journal of the South African Institution of Civil Engineering, which is distributed internationally, is a peer-reviewed, open-access journal licensed under a Creative
Commons Attribution Licence (CC BY-NC-ND). Readers may therefore freely use and share the content as long as they credit the original creator and publisher, do
2 not change the material in any way, and do not use it commercially. Copyright of this article remains with the authors.
is limited to catchments with suitable
streamflow data (Van Dijk et al 2013). Total mean flow of all rivers on earth
106
Probabilistic methods perform frequency Storm zone Transition zone Flood zone
k s)
analysis at a single specific site on observed d p ea
d floo
105 rld re
corde

Total drainage area of earth

streamflow data. The method assumes o f wo
e
velop

Runoff – Q (m3/s)
.5 (en
K=6
that the observed streamflow data comes 104 K = 6
5
from a known probability distribution and K=
4
K=
that it is stationary (Gericke & Du Plessis 103 K=
3
2
K= 1
2013). Beven (2000), cited by Smithers K= 0
K=
(2012), identified limitations associated or
d 102
r e c y)
with a direct probabilistic approach. It was w
ld
o r nsi t
( n te
/h i
identified that, while the actual distribu- m f a ll
/h
10
0 m ain
8 0 i nt r 0 mm
tion of flood peaks is unknown, numerous p o 10
1
probability distributions may reasonably /h 1
mm y
10 102 103 104 105 106 107 108
fit the plotted observed streamflow data, 10 si t
l i nt
en Catchment area (km2)
l f
but when extrapolated, these result in nf a o f
r ai r un
n. g
M i usi n
significantly different estimations of flood ca

peaks. Often, observed flood peak records

are short, representing only a small popu-
lation of the flood peaks at a given site.
Additionally, while probabilistic methods
presume that flood propagating vari-
ables are stationary, the statistics of data
Figure 1 Francou and Rodier envelope curves (Kovács 1980)
flood frequency curve that is appropriate namely storm, transition and flood zones
throughout the region (Smithers 2012). (Figure 1). The flood zone is governed by
Empirical methods should be utilised Equation 1.

Q = 106 ⎫⎪
A ⎫1 – 0.1K
could have changed over the period of the exclusively in their calibrated catchments

recorded streamflow data. As a result, it
ignores any changes in the runoff generat-
ing variables associated with incidents of
greater magnitude.
Deterministic methods
Deterministic methods are used to esti-
mate design period flood peaks and the
expected maximum flood peaks, denoted
as the Probable Maximum Flood (PMF).
Deterministic methods suppose that the
probability of the resulting flood peak and
PMF occurring in any year depends on the
probability of the causative rainfall. This is
because it is assumed that the frequency of
the determined flood peak and the causative
rainfall are assumed to be equal, at the same
time being influenced by catchment char-
acteristics and model parameters (Smithers
2012). These methods, however, have
shortcomings that impact significantly on
the anticipated flood peaks. The disadvan-
tages are that they rely on yet-to-be-verified
hypotheses (translated into numerical
parameter values such as runoff reduction
factors) and rainfall coefficients such as
storm rainfall area reduction factors, the
transition between storm and storm losses,
and the validity of unit hydrograph princi-
ples in the presence of significant flooding
(Van der Spuy & Rademeyer 2010).
Empirical methods
Empirical methods are calibrated using
streamflow data from hydrologically com-
parable sites in order to generate a regional
⎪
⎭ 106 ⎭
(Van Dijk et al 2013; Gericke & Du Plessis
(1)
2013). Empirical methods use a regional
approach, which is generally preferred over
the at-site frequency analysis (Kjeldsen et Where:
al 2002), as it employs data from multiple Q = discharge (m3/s)
catchment sites. Kachroo et al (2000) A = catchment area (km2)
stated that a regionally based analysis K = Francou and Rodier regional
provides a complete representation of coefficient.
historical flow data from a homogeneous
region. A number of authors (e.g. Cordery The RMF method estimates flood peaks
& Pilgrim 2000; Nortje 2010; Gericke & using flood envelopes in homogeneous
Du Plessis 2013; Nobert et al 2014) recom- regions, and it is empirically established for
mended the use of regionalised methods, obtaining extreme floods that represent the
as these have benefits over site-specific expected upper realistic limit in a region
methods using short records. Van der Spuy (Kovács 1988). Furthermore, it is used to
and Rademeyer (2010) defined empirical calculate the Safety Evaluation Discharge
approaches as methods “based on observa- (SED), a level pool peak discharge used to
tion or experience rather than theory or determine the associated safety risk for a
pure logic”, in which flood peak frequen- spillway system of a new or existing dam
cies are estimated using mathematical in the event of an extreme flood, in accor-
models established through the analysis dance with the SANCOLD (South African
of existing streamflow data. According to National Committee on Large Dams)
Wang (2000), the methods are based on guidelines (SANCOLD 1990). The RMF is
the concept of transferring hydrological also extensively used by practitioners, and
data from sites located in a homogeneous specifically the Department of Water and
region to sites with little or no historically Sanitation, as a guiding upper limit flood
observed data. Homogeneous regions are peak estimate when comparing different
established through regionalisation, which estimated flood peaks, using different
enables the selection of a frequency distri- approaches. This research does not inves-
bution or envelope curve that is suited for tigate the applicability of the method, but
a specific region (Kachroo et al 2000). The builds on the well-established use thereof
RMF is one such empirical method. The in South Africa and since the method was
method was adopted from Francou and last updated in 1988 (making available more
Rodier (1967) by Kovács (1980) for South than 30 years of additional data for further
Africa. Francou and Rodier (1967) identi- analysis), focusing on the improvement of
fied three zones that define the method, the RMF method for use in South Africa.

Journal of the South African Institution of Civil Engineering Volume 65 Number 3 September 2023 3
 Several reviews on the 1988 RMF
method have been conducted, and Pilon and A 20°0’0”E 25°0’0”E 30°0’0”E
Adamoski (1992), cited by Smithers (2012), N
acclaim the method for its ability to estimate
floods (regional maximum floods and Legend
River gauging stations
design flood peaks associated with a specific
Dam gauging stations
recurrence interval (RI)) in the absence of
25°0’0”S 25°0’0”S
streamflow data. Pegram and Parak (2004)
stated that the RMF method is a robust
and simple-to-use method that estimates
maximum regional flood at any site of inter-
est by considering only the regional envelope
and catchment area. However, various other 30°0’0”S 30°0’0”S
studies have expressed and highlighted
concerns and inconsistencies with the
method. Görgens (2007) asserted that flood
peaks that occurred after 1988 might have
surpassed the RMF envelopes. Van Vuuren
et al (2013) stated that the Kovács 1988 0 130 260 520 kilometres
35°0’0”S 35°0’0”S
analysis of the RMF should be reviewed by
20°0’0”E 25°0’0”E 30°0’0”E
including all the available and applicable data
to verify and/or reproduce the maximum
envelope curves. Smithers (2012) stated that B Flow station’s record length
140
the Kovács (1988) RMF regions must be Station record length
updated and refined due to the availability Average (record length) = 51.5 years
120 Missing data
of additional data. He further added that it Average (missing data = 1.5 years
would be prudent to investigate the use of
100
probability of exceedance associated with
Record length (years)

the RMF. Verwey (2015) stated that Kovács’s

method of regionalisation is inconsistent.
Verwey (2015) then investigated a new
approach that incorporates a safety factor
into the individual K values per station prior
to regionalising them into distinct regions.
Kovács (1980; 1988) added a safety factor ΔK
to the K value that was already regionalised.
Verwey (2015) argued that the added ΔK
has no mathematical or scientific basis for
its values. This argument was made in light
of Kovács’s (1980; 1988) omission of a math-
ematical or statistical explanation for ΔK.
As a result, Verwey (2015), and subsequently
Swanepoel (2017), developed a statistically
sounder new safety factor (ΔQ), applicable
to station flood peaks, before regionalisation.
Du Plessis and Masule (2023) evaluated
the methodology used by the Kovács 1988
RMF method, and concluded that the
method needed to be reviewed, given various
inconsistencies with the data used by Kovács
(1988), and the observation that some of his
RMF values had already been exceeded by
recent observed flood peaks.
Regionalisation
New safety factor approach
80
60
40
20
0
Flow gauging stations
Figure 2 A
 : Flow gauging stations distribution; B: Record length
to each station’s largest historical observed Where:
flood peak (Q1), taking into consideration OV = outlier value (m3/s)
that not all stations used in the analysis RV = station’s record length factor
have the same record length or span the (unitless)
same period of analysis. For each station, QV = flood peak ratio (unitless)
the method results in a new station flood ΔQ = calculated flood peaks (m3/s).
peak (QS). By substituting the QS value and
the corresponding area into the Francou The OV (outlier value) factor (Equation 3)
and Rodier (1967) flood zone equation takes into consideration whether or not the
(Equation 1), a new calculated K station station’s highest recorded flood peak Q1 is
value (KS) is obtained. The ΔQ is composed an outlier value.
of three parameters (OV, RV and QV) that
account for the statistical characteristics of OV = Q1 + xσ (3)
each gauging station’s Annual Maximum
Series (AMS) prior to regionalisation, as The value of Q1 is considered an outlier if it
illustrated in Equation 2 (Verwey 2015). proves to be greater than the value associ-

∆Q = QV ⎫⎪ V ⎫⎪(2)
for regionalisation ated with the outlier threshold determined
O
⎭ RV ⎭
Verwey (2015) proposed a methodology by the adjusted boxplot outlier test using
based on the addition of a calculated ΔQ the AMS.

4 Volume 65 Number 3 September 2023 Journal of the South African Institution of Civil Engineering
 The factor adjusts Q1 based on its
outlier status; if an outlier is observed, one A 20°0’0”E 25°0’0”E 30°0’0”E
standard deviation (x = 1, therefore σ) is N
added to Q1; alternatively, two standard
deviations (2σ) are added. Legend
The RV (station’s record length) factor River gauging stations
(Equation 4) was introduced to account
25°0’0”S 25°0’0”S
for the gauging stations’ varying record
lengths. The RV is expressed as the ratio of
the probability (R1) of a T-year flood occur-
ring within the available record length (n)
of the station to the hydrological risk (R0)
(that a T-year flood will occur in T years) 30°0’0”S 30°0’0”S
(Equation 5) (Swanepoel 2017).

R1
RV = (4)
R0

R1 or R0 = 1 – ⎫⎪1 – ⎫⎪ (5)
n
1 0 130 260 520 kilometres

⎭ ⎭
35°0’0”S 35°0’0”S
T
20°0’0”E 25°0’0”E 30°0’0”E

The QV (flood peak ratio) factor (Equation 6)

was added to scale the station’s highest B Rainfall station’s record length
120
observed flood peak (Q1) to the average of Station record length
the station’s top 25% of flood peaks (referred Average (record length) = 68.9 years
to as QAVE, 25%) (Swanepoel 2017). 100

OAVE,25%
Record length (years)

QV = (6) 80
Q1

60
The QS values can then be determined using
Equation 7, and subsequently KS values
using Equation 8, rewritten from Francou 40

⎪ ⎪
and Rodier’s (1967) flood zone equation.
20
QS = Q1 + ∆Q (7)

⎫ log⎫ QS ⎫ ⎫
⎪ 6⎪
0

⎭10 ⎭ (8)
Rainfall gauging stations (80)

log⎫⎪ ⎫⎪
KS = 10 1 –
A
⎭ ⎭ 8⎭ ⎭
10
The KS values are then used to regionalise
flood peaks, and with the addition of
superimposed catchment variables, an
RMF K regional map of homogeneous
flood regions can be delineated.
Catchment variables for
regionalisation
Agarwal et al (2016), Rosli et al (2019) and
Ahani et al (2020) all defined regionalisation
as the process of transferring hydrologic
information from gauged catchments to
ungauged catchments with similar char-
acteristics, thereby forming homogeneous
regions. Their similarities are determined
by their geographical proximity and shared
hydrologic characteristics and other defin-
ing variables. Catchments identified as
Figure 3 A: Rainfall gauging stations distribution; B: Record length
similar are combined into a single region.
Numerous studies assessed regionalisation
approaches for delineating homogeneous
flood regions. The Groupe de recherche
en hydroclimatologie statistique (GREHYS
1996) investigated four distinct approaches:
region of influence (ROI), canonical cor-
relation analysis, cluster analysis and
L-moments statistics. Parajka et al (2005)
compared four approaches in a similar man-
ner: the ROI, spatial proximity, catchment
variables and flood moments (probability
weighted moments and L-moments). The
catchment variable approach is based on
catchment characteristics and does not
include complex site statistics, and is suited
for regionalisation studies that are based
on an empirical nature (Ahani et al 2020).
Ahani et al (2020) defined catchment
variables as physiographic, meteorological,
geological, geographic location, soil type
and land use. Spatial proximity and physical
similarity that consider catchment variables,
are two methods that have been implement-
ed through models and software applica-
tions to establish homogeneous regions. The
two methods are implemented differently
based on the catchment-level assumptions.
A potentially hydrologically homogeneous
region would then be identified as catch-
ments with similar variables, and at-site
statistics are then used to independently
assess if the region is homogeneous. It is
urged that, despite the variety of approaches
to regionalisation, a careful selection of vari-
ables affecting rainfall runoff is necessary
to effectively distinguish runoff behaviours
within a region (Zhang & Stadnyk 2020).

Journal of the South African Institution of Civil Engineering Volume 65 Number 3 September 2023 5
Van Dijk et al (2013) identified three
variables that have a cumulative effect on a A 20°0’0”E 25°0’0”E 30°0’0”E
region’s potential runoff response: climatic, N
physiographic and topographic. Benson
(1964) identified climatic and catchment Legend
characteristics (catchment size, vegetation, River gauging stations
land use/cover and soil type) as the critical
25°0’0”S 25°0’0”S
hydrological variables affecting runoff.

STUDY AREA
The study area applicable to this research
is limited to the national boundaries of 30°0’0”S 30°0’0”S
South Africa and uses data obtained from
the Department of Water and Sanitation
(DWS) flow gauging stations (which
include dams) and the South African
Weather Service (SAWS) rainfall gauging
stations located within the study area. The 0 130 260 520 kilometres
35°0’0”S 35°0’0”S
dam station data presents calibrated (rout-
20°0’0”E 25°0’0”E 30°0’0”E
ed) dam inflow data, which includes the
maximum flood peaks for South Africa’s
major dams. The river flow station data B 140
Station record length
reflects the data recorded at river gauges. Average (record length) = 114 years
120
Data sources
100
Record length (years)

Flood data
The DWS provided the Annual Maximum 80

Series (AMS) of 494 stations, comprising

60
316 river gauging stations and 178 dam sta-
tions. The locations of the gauging stations
40
and their record lengths are depicted in
Figure 2 on page 4.
20

Rainfall data
Two rainfall data sets were used in the
research:
1. Design rainfall from Smithers and
Schulze’s (2002a) Design Rainfall soft-
ware dataset
Seven-day 1:100-year storm rainfall data
of 80 rainfall stations (Figure 3 on page 5)
with a reasonably even distribution, cho-
sen in accordance with the South African
Weather Service’s Rainfall Stations
Reference Grid, was extracted from the
Design Rainfall software. Alexander
(2006) stated that extreme flood peaks are
the result of prolonged and widespread
extreme rainfall events such as the 7-day
rainfall. The 7-day rainfall was significant
in establishing rainfall isohyets that
assisted in the regionalisation process.
2. Daily rainfall from the SAWS and Lynch
dataset
A consolidated dataset of one-day rain-
fall from SAWS and Lynch was created
(Figure 4). SAWS provided data from 69
gauging stations, and much older data of
0
Rainfall gauging stations
Figure 4 A: Rainfall gauging station distribution; B: Record length
35 gauging stations was extracted from
the Lynch database. The gauging stations
had an average record length of 114
years. The Lynch database provides data
from a spatial database of daily rainfall
in South Africa. The one-day rainfall
data was required to determine the
15-­minute rainfall intensities that enable
the determination of point discharge at
the storm zone. The 15-minute rainfall
data was scaled down from one-day
annual maximum series rainfall using
Smithers and Schulze (2002a) scaling
factors.
METHODOLOGY
This section provides the detailed method-
ology that was followed during the research
to update the 1988 RMF approach.
Updating of the 1988 RMF method
Database of maximum flood peaks
In compiling the observed flood peak data-
base the principles were (i) that only the
Annual Maximum Series (AMS) from the
DWS dataset would be used, and (ii) that at
any one gauging station only the maximum
observed flood peak would be selected. A
new database, containing 494 flood peaks
was then created.
Determination of QS and KS values
Prior to regionalisation, using the AMS
of each gauging station, the statistical
parameters utilised in calculating ΔQ were
determined, and ΔQ was calculated using
Equation 2. The QS values were determined
using Equation 7, which were then used to
calculate the KS values using Equation 8.

6 Volume 65 Number 3 September 2023 Journal of the South African Institution of Civil Engineering

Geographically plot the Ks
values in GIS and perform
Create Ks isolines (lines
of equal Ks values)
Produce a Ks regional map
(delimited from isolines)
Assign Ks regions with a
regional Ks envelope value
Evaluate the Ks envelopes
against the Qs values
Figure 5 Regionalisation process
The KS values were then regionalised
to determine a grouping of KS regions
with similar KS values, and subsequently
delineating final RMF K regions by incor-
porating various geographical, climatic and
relief variables.
Regionalisation
Flood zone
Regionalisation makes it possible to trans-
fer data from gauged catchments to similar
ungauged catchments in homogeneous
regions. The regionalisation process is
illustrated in Figure 5, and explained in
more detail in the following paragraphs.
Several variables were superimposed in
delineating the RMF K regions. The vari-
ables were superimposed sequentially from
the most influential to the least. These were
the KS (stations calculated K-value) values,
drainage regions, rainfall pattern, flood
homogeneous regions (SDF, Veld Types, and
RMF regions), vegetation and soil type, K
values at international boundaries (Regional
Maximum Flood envelopes lines KE value),
river systems and seasonal rainfall regions.
These variables are shown in Figure 6, and
can be expanded on as follows:
Journal of the South African Institution of Civil Engineering
Delineate final
Regionalisation
RMF K regions
Alter regions by superimposing
more catchment variables
(main river systems,
an interpolation international boundary K
values and seasonal rainfall)
Alter the Ks regions by
superimposing catchment
variables (drainage regions,
rainfall, flood regions
and vegetation cover)
QQ KS values (Figure 6A):
The KS isolines were regarded as the
most influential variables in determining
the new RMF regions. The KS values
were plotted geographically in GIS and
spatially interpolated to generate KS iso- QQ
lines (Figure 6A). KS regions were then
directly drawn from the isolines. The
KS regions were used to group gauging
stations with similar KS values, without
any influences from other variables. As
a result, 15 KS regions were delineated.
A regional KS value was assigned to each
region and a representative envelope
curve was drawn. It should be noted that
no objective criterion was used in select- QQ
ing the regional KS envelope value used,
but a base value of KS = 1.8 with a 0.2
interval was chosen based on Kovács’s
original approach. As a result of the
selected interval, a grouping of multiple
KS isolines was required to delineate a QQ
single KS region. It was then necessary to
graphically depict the assigned regional
KS envelopes against the QS values of
stations in each KS region. To further
improve the regionalised 15 KS regions
to achieve a grouping of gauging stations
with QS values not exceeding and not
Volume 65 Number 3 September 2023
Identify the highest observed
flood peak per new RMF K
region; draw a K envelope
15% above the flood peak
Disregard the Ks envelopes
(Qs grouping achieved)
Produce a new RMF
K regional map
significantly plotting below the assigned
regional KS envelopes, the KS regions
were adjusted by including the influence
of physiographic and climatic variables
as outlined below.
Drainage regions and topography
(Figure 6B):
It was anticipated that the regionalisa-
tion of the KS regions could be improved
by reflecting the quaternary catchments
of South Africa. The quaternary catch-
ments (1946 catchments), along with the
topography, were superimposed to pro-
vide guidance regarding the edges of the
flat areas and areas of small catchments.
Rainfall (Figure 6C):
The 7-day 1:100-year rainfall isohyets
were superimposed on the KS regions,
and the KS regional boundary was then
adjusted to follow the isohyets where
few or no KS values were available.
Flood homogeneous regions (Figure 6D):
Three of South Africa’s well-established
flood regional maps that are used for
local practices in flood hydrology were
superimposed in the adjustment of the
KS regions. These include the Standard
Design Flood (SDF), the General Veld
Types and the 1988 RMF regions. The
7
 Legend N

N 5.8 Rivers
3.2 5.45.6 4.6 Primary drinage regions
4.4
3.4 3.6 5.2 5
4.6 Quaternary catchments A
3.8
3.6
3.8 B
4 5.45.2
2
4.2 4.4 4.2
3.6 X
Legend 1.8 2.2 1.41.21491 5
Ks isolines 1.61.82 4.84.64.8
2.222.4 5
2.4
4.64.4
2
2.8 4.2
4 5.8 6.6
6 6.4 C W
2.6 6.2
3.4 3 4.8 4
2.2
3.2
5.65.8
4.855.2 5.8 6
5.4
5.6
V
1.8
5.65.4 4.8 5.2
5.4 D
1.6
5
5.2 5.6 U
2
5
5.2
5.665.8 F
3.8 5.4
4
4.8 5.2
4.6 4.8
5.4 T
4.44.2
3.6 5.2 5.45.2 E S
3.2 4.6 5.4 5 Q
5.2
55.4 3 4 5.45.2 5.6 L N R
5.65.4
G J
M P0
5.2 4.8
4.2 4.8 55.25.4
5.2 5.45.2
5.2
5.4
5.25.4 5.86 5.6
6.2 H K 162.5 325 650
kilometres
A 5 0 135 270 540 kilometres B
Legend
7-day 1:100-year rainfall 225 375 525 675
3

550
Isohyets (mm) 250 400 550 700
125 275 425 575 725
150 300 450 600 750
2 5
175 325 475 625 775
22
200 350 500 650 800 5 1 4
525

825
29
250 300
27

22
5

5
7
0 6
50 13 8
28
75 0
5 4 60
42 75 50 26
5 7 27
20

525

9
0

15
0 11 25
14
5

450 12 10
27
0

15
20

23 24
5
0
625

12
0 35
37 00
40
125

200 225 16 21
3
5

22
250 325 19
5 17
17
0 120 240 480 kilometres
C D 18
20
Legend
N
Rivers Notes:
Simplified Acocks veld types 1. Base data generally captured at a scale of 1:250 000 N
Coastal tropical forest types 2. Scale effects may occur at scales < 1:250 000
3. Some layers show more information when zoomed
False bushveld types
in beyond 1:2 000 000 scale
False grassveld types
False karoo types
False sclerophyllous bush types
Inland tropical forest types
Karoo and karroid types
Pure grassveld types
Sclerophyllous bush types
Temperate and transitional forest and scrub types
Tropical bush and savanna types (bushveld)

E 0 75 150 300 450

kilometres
F 0 135 270 540 kilometres

Legend
N Seasonal rainfall regions
Legend Region N
Rivers No definition
Quaternary catchments Summer rainfall zone 50–75%
Summer rainfall zone >75%
Summer/winter rainfall zone
Winter rainfall zone 50–75%
Winter rainfall zone >75%
Winter/summer rainfall zone
Year rainfall zone 50–75%
Year rainfall zone >75%

Sumer rainfall zone >75%

Winter rainfall zone 50–75%

No definition

Winter rainfall zone >75%

Winter/sumer rainfall zone Sumer rainfall zone 50–75%
Sumer/winter rainfall zone

Year rainfall zone 50–75%

Year rainfall zone >75%
0 135 270 540 kilometres
G H 0 135 270 540 kilometres

Legend (A): KS isolines (B): Drainage regions

(C): Rainfall isohyets (D): SDF drainage basin (Alexander 2002) (E): Vegitation cover (WR 2012)
(F): Boundary K values (G): Main river systems (H): Seasonal rainfall (Roffe 2019)

Figure 6 Variables in the considered regionalisation process

8 Volume 65 Number 3 September 2023 Journal of the South African Institution of Civil Engineering
 Region KS = 3.4 Region KS = 3.6
100 000 100 000

10 000 10 000
Discharge (m3/s)

Discharge (m3/s)
1 000 1 000

100 100

10 10

1 1
1 10 100 1 000 10 000 100 000 1 000 000 1 10 100 1 000 10 000 100 000
Catchment area (km2) Catchment area (km2)

Region KS = 3.8 Region KS = 4.0

100 000 100 000

10 000 10 000
Discharge (m3/s)

Discharge (m3/s)
1 000 1 000

100 100

10 10

1 1
1 10 100 1 000 10 000 100 000 1 000 000 1 10 100 1 000 10 000 100 000
Catchment area (km2) Catchment area (km2)

Region KS = 4.2 Region KS = 4.4

10 000 100 000

10 000
1 000
Discharge (m3/s)

Discharge (m3/s)

1 000
100
100

10
10

1 1
1 10 100 1 000 10 000 1 10 100 1 000 10 000 100 000 1 000 000
Catchment area (km2) Catchment area (km2)

Region KS = 4.6 Region KS = 5.0

100 000 100 000

10 000 10 000
Discharge (m3/s)

Discharge (m3/s)

1 000 1 000

100 100

10 10

1 1
1 10 100 1 000 10 000 100 000 1 10 100 1 000 10 000 100 000
Catchment area (km2) Catchment area (km2)

Qs river stations Qs dam stations Regional KS envelope curve

Figure 7 Regionalised KS regions (continued on next page)

Journal of the South African Institution of Civil Engineering Volume 65 Number 3 September 2023 9
 Region KS = 5.2 Region KS = 5.4
100 000 100 000

10 000 10 000
Discharge (m3/s)

Discharge (m3/s)
1 000 1 000

100 100

10 10

1 1
0.1 1 10 100 1 000 10 000 100 000 0.1 1 10 100 1 000 10 000 100 000
Catchment area (km2) Catchment area (km2)

Region KS = 5.6 Region KS = 5.8

100 000 100 000

10 000 10 000
Discharge (m3/s)

Discharge (m3/s)
1 000 1 000

100 100

10 10

1 1
1 10 100 1 000 10 000 100 000 1 10 100 1 000 10 000 100 000
Catchment area (km2) Catchment area (km2)

Region KS = 6.0 Region KS = 6.2

100 000 100 000

10 000 10 000
Discharge (m3/s)

Discharge (m3/s)

1 000 1 000

100 100

10 10

1 1
0.1 1 10 100 1 000 10 000 1 10 100 1 000 10 000
Catchment area (km2) Catchment area (km2)

Qs river stations Qs dam stations Regional KS envelope curve

Figure 7 (continued from previous page) Regionalised KS regions

eastern escarpment consists of numer-
ous small catchment areas, therefore
the flood homogeneous regional bound-
aries provided guidance for the RMF
regional boundaries to follow.
QQ Vegetation and soil (Figure 6E):
Vegetation and soil types are landscape
characteristics that interact at a catch-
ment scale with climate characteristics
such as rainfall to produce catchment
adjusted KS regions correspond well subsequently delineated. At this point, the
with the vegetation and soil type QS values of each station and the regional KS
boundaries. This was because vegeta- envelope curves were disregarded (region-
tion and soil boundaries are closely alisation using QS had been completed) and
linked to the topography. consideration was now given to the actual
The adjusted KS regions were graphically observed flood peaks Q1. Note that the KS
analysed, plotting the QS values against values were only used for regionalisation.
the KS envelopes. It was then concluded The RMF K regions were further
that regionalisation had been achieved, as enhanced as follows:
depicted in Figure 7. Based on the results QQ K values along international boundaries

runoff. Vegetation and soil types were presented in Figure 7, the systematically (Figure 6F):
superimposed in altering KS regions. adjusted regionalisation was deemed to Boundaries of the RMF K regions where
It was found, however, that the already be completed. As a result, 15 regions were few or no KS values were available,

10 Volume 65 Number 3 September 2023 Journal of the South African Institution of Civil Engineering
 followed the rainfall isohyets. However, Table 1 Cluster group ratios (Smithers & Schulze 2002b)
along the borders of South Africa and Standard Standard
Namibia where, over a large area, no KS Cluster Average Median Cluster Average Median
error error
values were available, the updated RMF 1 1.20 1.20 0.049 9 1.26 1.27 0.111
regions of Namibia from Cloete et al
2 1.21 1.21 0.063 10 1.19 1.18 0.090
(2014) were georeferenced and superim-
posed to adjust the boundaries of the KS 3 1.19 1.18 0.072 11 1.20 1.15 0.087

regions to be consistent with the K values 4 1.21 1.22 0.090 12 1.19 1.18 0.044
for Namibia. 5 1.20 1.17 0.097 13 1.28 1.30 0.139
QQ Rivers (Figure 6G): 6 1.17 1.16 0.055 14 1.24 1.24 0.056
Major rivers were superimposed to
7 1.15 1.14 0.051 15 1.25 1.26 0.096
better understand the fluvial landscape,
8 1.20 1.20 0.032 15 1.25 1.26 0.096
with an emphasis on the flow direction
and length of the river from its catch-
ment boundary to the gauging station. Table 2 Regression coefficients and constants – 15 minutes (Smithers & Schulze 2002b)
QQ Seasonal rainfall (Figure 6H): Cluster 1 2 3 4 5 6 7 8
Rainfall in South Africa is caused by a
Xcoeff. 0.1764 0.0391 0.1041 0.2584 0.0869 11.9385 0.0681 0.1169
variety of weather phenomena that occur
in different regions and at different times Constant 7.0258 16.0588 13.3526 3.1253 11.9385 6.7972 12.5861 7.7585
of the year. The seasonal rainfall distribu- Cluster 9 10 11 12 13 14 15
tions were similarly superimposed. The Xcoeff. 0.0159 -0.0069 0.0855 0.1940 -0.0054 0.1094 0.3055
boundaries of the all-year rainfall zones
Constant 8.6239 11.8186 11.3601 4.8688 14.8076 9.4261 -0.9936
and the winter rainfall zones were used to
justify several delineated final regions.
QQ River K values: Q = 0.278i (9) average, median and standard error. The
Kovács (1988) stated that rivers that ratios are presented in Table 1.
flow through multiple RMF regions To determine the required regional dis- The median ratios were used to
may exhibit distinct flood characteris- charge at the storm zone, the 15-minute convert the highest 1-day rainfall values
tics that differ significantly from those rainfall intensities were calculated from 15 (selected from the annual maximum
found in the delineated K region, thus minutes of rainfall, scaled from the 1-day series dataset of each rainfall station) to
the lower reaches of these rivers have rainfall, using Smithers and Schulze (2002b) 24-hour rainfall values. The required short
a different K value compared to that scaling factors. The rainfall stations used in duration (15-minute) rainfall values were
of their upper reaches. The rivers were this analysis were selected from catchment then scaled down from the 24-hour scaled
investigated and the sections with a dif- areas where the peak discharges depended rainfall, using the regression coefficients
ferent K value were identified. only on rainfall intensity. Flow gauging sta- and regression constants in Table 2, as
tions in these catchment areas had recorded presented by Smithers and Schulze (2002b),
Storm zone extreme flood peaks. The selected rainfall of each corresponding cluster group.
In the storm zone, the peak discharges stations were then geographically located A regional 15-minute rainfall was then
depend only on rainfall intensity. For a using GIS. For each station a predetermined selected as the highest 15-minutes rainfall
catchment area of 1 km2 the discharge cluster group and the corresponding of all gauging stations in each RMF K
is as presented in Equation 9, where i is scaling factors according to Smithers and region. The regional 15-minutes rainfall was
the maximum 15-minute rainfall inten- Schulze (2002b) were identified. Smithers converted into rainfall intensities (mm/h).
sity in mm/h. Kovács (1988) stated that a and Schulze (2002b) presented ratios of The 15-minute regional rainfall intensities
15-­minute period is the approximate time 24-hour:1-day rainfall values corresponding were substituted into storm zone Equation 9
of concentration in a catchment of 1 km2. to each cluster group which included an to determine the regional discharge for a

Table 3 Distributed RMF K regional 15-minute rainfall and discharges

15-minute Regional 15-minute Regional
15-minute rainfall discharge 15-minute rainfall discharge
RMF K region RMF K region
rainfall (mm) intensity (i) (m3/s) rainfall (mm) intensity (i) (m3/s)
(mm/h) (Q = 0.278i) (mm/h) (Q = 0.278i)

2.8 9 36 10 4.6 26.5 106.0 29.5

3.4 11.3 45.2 12.6 4.8 29.5 118.0 32.8

3.6 13.6 54.4 15.1 5.0 32.5 130.0 36.1

3.8 15.9 63.6 17.7 5.2 35.2 140.8 39.1

4 18.2 72.8 20.2 5.4 38.5 154.0 42.8

4.2 20.5 82.0 22.8 5.6 41.5 166.0 46.1

4.4 23.5 94.0 26.1 5.8 44.4 177.6 49.4

Journal of the South African Institution of Civil Engineering Volume 65 Number 3 September 2023 11
 Statistical analysis for A2H007
Recurrence interval in years
2 5 10 20 50 100 1 000 10 000
10 000

1 000
Q (m3/s)

100

10
0.8 0.5 0.2 0.1 0.05 0.02 0.01 0.001 0.0001
Probability of exceedance
LN LP3 GEV GEVpwm Proposed RMF

Figure 8 Statistical analysis

Figure 9 New RMF K regional map

12 Volume 65 Number 3 September 2023 Journal of the South African Institution of Civil Engineering
1 km2 catchment. These regional discharges A
are presented in Table 3 on page 11. 20°0’0”E 25°0’0”E 30°0’0”E
When the regional discharges for the Legend
New RMF K Regions 4.0 5.2
N
storm zone in Table 3 were compared to
K_value 4.2 5.4 5.2
the Kovács 1988 regional discharges, it was 2.8 4.4 5.6 5.4
3.4 4.6 5.8 4.2
observed that Kovács’s regional discharges 3.6 4.8 <2.8 3.8
25°0’0”S 25°0’0”S
were significantly larger. There is, however, 3.8 5.0 4.8
<2.8 4.4
no documented reference to where Kovács
obtained his data from, and the data in 3.6

Table 3 was therefore adopted, being the 2.8 3.8

5.8
best available information. The available 5.0
3.4 5.2 5.6
observed flood peaks from gauging stations
30°0’0”S <2.8 5.2 30°0’0”S
used in this research with catchment areas 3.6
4.6
approximately 1 km2 in size were, however,
4.0
all significantly lower than the flood peaks 4.8
4.4
based on the flood zone equation, provid-
4.6 5.4
ing justification for the decision to adopt 5.2 5.8
4.8
the values from Table 3. 0 135 270 540 kilometres

20°0’0”E 25°0’0”E 30°0’0”E

Transition zone
No regionalisation was carried out in the B
20°0’0”E 25°0’0”E 30°0’0”E
transition zone, but only the determination
of envelope curves. The transition zone Legend 5.0
N
Kovacs 1988 RMF Regions 5.2
envelope curves provide a smooth transi- Key_value 5.4
5.0
tion between the storm and flood zones. 2.8 5.6
3.4 <2.8
To determine the transition zone envelope 4.0 <3.4
25°0’0”S 25°0’0”S
curves, the regional discharge associated 4.6 5.0
with 15-minute rainfall intensity at the <2.8

storm zone, as determined in Table 3, and 2.8 4.6

5.6 5.6
the RMF K envelope curve in the flood 3.4
zone were required.
<3.4 4.0
30°0’0”S 3.4 3.4 30°0’0”S
Homogeneity of the RMF K regions
4.6 5.0 5.4
Due to the numerous adjustments, the 5.2
homogeneity of the newly established RMF
K regions was investigated. The homogene- 4.6

ity of RMF K regions was determined using 5.4

a statistical approach where a coefficient
of variation (CV) was calculated for each
RMF K region and then compared to a CV
threshold, adopted from Kovács (1980). The
purpose of this evaluation was to assess
the dispersion of KS values within a region.
The KS values were used since they were
the primary parameters in the regionalisa-
tion process. Kovács (1980) calculated
the CV of his KR values to determine
the homogeneity of his 1980 RMF flood
regions. According to Kovács, a CV of less
than or equal to 17% satisfied the funda-
mental requirement of regionalisation and
that of homogeneous regions. Based on his
finding, the homogeneity of each RMF K
region was evaluated against a maximum
CV value of 17%. The CV was tested, using
Equation 10.
σK
CVK = S
≤ 0.17
S μK
S
4.6
0 135 270 540 kilometres
20°0’0”E 25°0’0”E 30°0’0”E
Figure 10 A: New RMF K regions; B: Kovács 1988 RMF K regions
CV K = the regional coefficient of varia- catchment areas and drawing the K enve-
S
tion of the KS (unitless) lopes 15% above the highest observed flood
σK = the standard deviation of the KS peaks (Q1) in each K region. These were
S
(unitless) carried out after regionalisation. The highest
μK = the mean value of the KS observed flood peak (“envelope determining
S
(unitless). flood peak”) in each K region was identified,
from which the regional K envelope was then
Determination of envelope curves drawn 15% above the selected flood peak
The regional envelope curves were de­veloped value. The 15% was adopted from Kovács
for the flood zone and the transition zone. (1988), and deemed as the safety factor
above the highest selected flood peak. The
Flood zone Kovács 1980 envelope curves, which were
The envelope curve in the flood zone was drawn above the regional maximum flood
established by considering the regional peak, were in most cases between ΔK = 0.2
boundaries of the RMF K regions, identify- to ΔK = 0.31 above the maximum observed
ing the separation between the flood zone value, which is equivalent to an increase in
and the transition zone, and furthermore discharge of between 15% to 30%, depending
(10)
by plotting flood peaks against respective on the catchment area. Therefore, in the

Journal of the South African Institution of Civil Engineering Volume 65 Number 3 September 2023 13
flood zones for all K regions, the K envelope Table 4 Statistical parameters for homogeneity check
curves were drawn by adding ΔK = 0.2, Number of Standard Mean Kovács 1980
which is 15% of the flood peak. The assigned Region Regional CV
stations deviation (σKs) value µKs maximum CV
regional K envelope values were from < 2.8,
Region 2.8 6 0.41 3.15 0.13 0.17
2.8, 3.4 to 5.8 with a 0.2 increment. The
boundary between the flood zone and the Region 3.4 11 0.45 3.47 0.12 0.17
transition zone (still to be discussed) was Region 3.6 9 0.56 3.65 0.15 0.17
identified as being at the catchment area
Region 3.8 8 0.57 4.08 0.14 0.17
where the trend of the plotted observed
flood peaks changes from aligning with the Region 4.0 3 0.41 4.25 0.10 0.17
direction of the flood zone envelope curve.
Region 4.2 31 0.47 4.14 0.11 0.17

Transition zone Region 4.4 60 0.48 4.31 0.11 0.17

The envelope curves in the transition zone Region 4.6 8 0.27 4.58 0.06 0.17
were drawn between the storm zone and the
Region 4.8 94 0.62 4.52 0.14 0.17
flood zone. The envelope was drawn from
the discharge associated with the 15-minute Region 5.0 169 0.57 4.62 0.12 0.17
duration rainfall over 1 km2 at the storm zone Region 5.2 52 0.53 4.99 0.11 0.17
to the K envelope curve in the flood zone.
Region 5.4 32 0.49 5.18 0.10 0.17

Return period of the RMF Region 5.6 11 0.61 5.23 0.12 0.17
Kovács estimated the return period of the
Region 5.8 3 0.14 6.12 0.02 0.17
RMF to be 200 years (Kovács 1988). To
assess this finding, the return period of the
newly determined RMF values at 452 of the
gauging stations in the DWS dataset were 20°0’0”E 25°0’0”E 30°0’0”E
determined. The AMS of each gauging Legend N
station was arranged in a descending order River reaches with different K values
River
and a plotting position of each value was Modder
Mooi
determined using the Cunnane formular. Orange
25°0’0”S Riet 25°0’0”S
Cunnane was used as it is the suggested Sand
and the preferred plotting position in the Vet
4.6

DWS (Van der Spuy & Rademeyer 2010).

The values were then plotted against 4.6
the probability of occurrence on a log-­ 4.2 5.4 4.4
4 .4 4. 2
probability scale, and the best visually
30°0’0”S 30°0’0”S
fitting probability distribution selected.
The required probability of occurrence
(hence return period) for a specific RMF
was then determined as shown in Figure 8
on page 12 for station A2H007.
0 135 270 540 kilometres

20°0’0”E 25°0’0”E 30°0’0”E

RESULTS
Figure 11 River reaches with different K values
RMF K regions
The regionalisation process resulted in the Table 5 K values in particular reaches of large rivers
delineation of 15 RMF regions as presented RMF K region
in Figure 9 on page 12. Drainage region Upstream
in which the
River (location of the River K value gauging station
The RMF K regions with higher K reaches of river
river section) of the river reach
flow through
values are dominated by hilly to moun-
tainous relief, higher rainfall and smaller Riet C 4.2 3.6 C2R002
catchment areas. RMF K regions with
Orange D 4.4 3.4 D7H002
lower K values are dominated by flat relief,
lower rainfall and larger catchment areas. Sand C 4.6 4.4, 4.2 & 3.8 C4R001
Similar trends were observed in the Kovács
Modder C 4.2 & 5.4 3.8 & 5.0 C5R004
1988 RMF regional map. The highest K
region of 5.8 was delineated for the new Mooi C 4.6 3.6 C2R001
RMF K regions, which was delineated to
Vet C 4.4 3.6 C4R002
accommodate the gauging station that had

14 Volume 65 Number 3 September 2023 Journal of the South African Institution of Civil Engineering
 Region K = 2.8 Region K = 3.4
100 000 100 000

10 000 10 000
Discharge (m3/s)

Discharge (m3/s)
1 000 1 000

100 100

10 10

1 1
1 10 100 1 000 10 000 100 000 1 000 000 1 10 100 1 000 10 000 100 000 1 000 000
Catchment area (km2) Catchment area (km2)

Region K = 3.6 Region K = 3.8

100 000 100 000

10 000 10 000
Discharge (m3/s)

Discharge (m3/s)
1 000 1 000

100 100

10 10

1 1
1 10 100 1 000 10 000 100 000 1 10 100 1 000 10 000 100 000
Catchment area (km2) Catchment area (km2)

Region K = 4.0 Region K = 4.2

100 000 100 000

10 000 10 000
Discharge (m3/s)

Discharge (m3/s)

1 000 1 000

100 100

10 10

1 1
1 10 100 1 000 10 000 100 000 1 10 100 1 000 10 000 100 000
Catchment area (km2) Catchment area (km2)

Region K = 4.4 Region K = 4.6

100 000 100 000

10 000 10 000
Discharge (m3/s)

Discharge (m3/s)

1 000 1 000

100 100

10 10

1 1
1 10 100 1 000 10 000 100 000 1 10 100 1 000 10 000 100 000
Catchment area (km2) Catchment area (km2)

River stations Dam stations Regional discharge 15% safety Flood zone envelope Transition zone envelope

Figure 12 New RMF K regional envelope curves (continued on next page)

Journal of the South African Institution of Civil Engineering Volume 65 Number 3 September 2023 15
 Region K = 4.8 Region K = 5.0
100 000 100 000

10 000 10 000
Discharge (m3/s)

Discharge (m3/s)
1 000 1 000

100 100

10 10

1 1
0.1 1 10 100 1 000 10 000 100 000 1 10 100 1 000 10 000 100 000
Catchment area (km2) Catchment area (km2)

Region K = 5.2 Region K = 5.4

100 000 100 000

10 000 10 000
Discharge (m3/s)

Discharge (m3/s)
1 000
1 000
100
100
10

1 10

0.1 1
0.1 1 10 100 1 000 10 000 1 10 100 1 000 10 000
Catchment area (km2) Catchment area (km2)

Region K = 5.6 Region K = 5.8

100 000 100 000

10 000 10 000
Discharge (m3/s)

Discharge (m3/s)

1 000 1 000

100 100

10 10

1 1
0.1 1 10 100 1 000 10 000 1 10 100 1 000 10 000
Catchment area (km2) Catchment area (km2)

River stations Dam stations Regional discharge 15% safety Flood zone envelope Transition zone envelope

Figure 12 (continued from previous page) New RMF K regional envelope curves

exceeded the Kovács K E Regions 5.4 and
5.6, with the latter the highest K region as
defined by Kovács. The lowest K region
represents a region with K values lower
than K = 2.8; however, in accordance with
Kovács (1988), such a K region was defined
as < 2.8. Kovács (1988) stated that the RMF
in such a region should be determined
using a K value of 2.8.
When the RMF K regions (Figure 10A
on page 13) were compared to Kovács’s
1988 RMF regions (Figure 10B), general
similarities and differences were observed.
The new RMF K regional map showed
that, in contrast to the Kovács 1988 RMF
regional map, a more localised regional
map, with regions having smaller areas,
particularly those with high K values, was
delineated. However, various boundaries of
the new RMF K regional map corresponded
to those of the Kovács 1988 RMF regional
boundaries, and these were observed mostly
in the western and northeastern parts of the
country. The lower (2.8, 3.4) and the higher
(5.0, 5.2, 5.4) K regions were observed to be
consistent with those defined by Kovács,
despite having different sizes. The high
correlation and similarities between the new
RMF and the Kovács 1988 RMF regions
instil confidence in the validity of the new
RMF regions. The new regions were delin-
eated using a much larger dataset and longer
record length compared to the Kovács 1988
regions. As a result, 15 RMF K regions
were delineated in this study compared to
Kovács’s previous 8 RMF regions.
With the much larger dataset it was
required to evaluate the homogeneity of
the RMF K regions.

16 Volume 65 Number 3 September 2023 Journal of the South African Institution of Civil Engineering
 The homogeneity of the RMF K regions Table 6 Regional envelope curve equations
was evaluated using the statistical param- Transition zone Flood zone
eters (σ and μ) listed in Table 4 on page 14.
RMF K region Envelope Areal range Envelope Areal range
The number of gauging stations located
equation (m3/s) (km2) equation (m3/s) (km2)
within a region are shown. Equation 10 was
then used to calculate the regional coef- 2.8 QRMF = 19.3A0.44 1 – 500 QRMF = 1.7A0.72 500 – 600 000
ficient of variation. 3.4 QRMF = 12.6A0.52 1 – 500 QRMF = 5.3A1.66 500 – 600 000
K regions 4.4, 4.8, 5.0 and 5.2 all had
3.6 QRMF = 15.1A0.56 1 – 500 QRMF = 7.6A0.64 500 – 10 000
more than 50 gauging stations per region
and, according to Table 4, a CV of less than 3.8 QRMF = 17.7A0.54 1 – 500 QRMF = 11.0A0.62 500 – 80 000
0.14, which gave credence to the regionali- 4.0 QRMF = 20.2A0.56 1 – 500 QRMF = 15.9A0.60 500 – 80 000
sation approach. Based on the CV of the
4.2 QRMF = 22.8A0.58 1 – 200 QRMF = 22.9A0.58 200 – 80 000
KS values in each new RMF K regions, as
discussed above, the regions were deemed 4.4 QRMF = 26.1A0.61 1 – 200 QRMF = 33.1A0.56 200 – 80 000
to be homogeneous. 4.6 QRMF = 29.5A0.61 1 – 1 000 QRMF = 47.9A0.54 1000 – 80 000
It was, however, observed that certain
reaches of rivers running through the RMF 4.8 QRMF = 32.8A0.69 1 – 80 QRMF = 69.2A0.52 80 – 80 000

K regional boundaries had a markedly dif- 5.0 QRMF = 36.1A0.73 1 – 80 QRMF = 100.0A0.50 80 – 80 000
ferent K value compared to the K value of
5.2 QRMF = 39.1A0.78 1 – 80 QRMF = 144.5A0.48 80 – 10 000
the region, similar to what was observed
by Kovács. The positions of these river 5.4 QRMF = 42.8A0.87 1 – 50 QRMF = 208.9A0.46 50 – 10 000
reaches are shown in Figure 11 on page 14. 5.6 QRMF = 46.1A0.92 1 – 50 QRMF = 302.0A0.44 50 – 10 000
If any flood estimation had to be carried
5.8 QRMF = 49.4A0.98 1 – 50 QRMF = 436.5A0.42 50 – 10 000
out within these six reaches, a river K value
different to the RMF K region in which it is
situated needs to be used. The river reach The regional discharge at the storm zone could not easily be defined due to the
K values and corresponding information increases as the RMF K values increase. The irregular and scatter plotting of the limited
are presented in Table 5 on page 14. transition zone envelope was drawn from available observed flood peaks or a lack of
the regional point discharge to the regional gauging stations. In this regard, the transi-
Envelope curves K envelope curve in the flood zone. tion zone envelope curves were drawn in
The final regional K envelope curves are In several RMF K regions the bound- the context of the Kovács RMF envelope
presented in Figure 12. ary between the flood zone and transition curves where the largest catchment area

80

70

60

50
Number of stations

40

30

20

10

0
0 to 100 to 200 to 500 to 1 000 to 2 000 to 5 000 to 10 000 to 20 000 to 50 000 to 100 000 to 200 000 to
100 200 500 1 000 2 000 5 000 10 000 20 000 50 000 100 000 200 000 500 000
RMF return period range (years)

Figure 13 Histogram of the RMF return period

Journal of the South African Institution of Civil Engineering Volume 65 Number 3 September 2023 17

1.0
0.9
0.8
Exceedence probability
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0 1 000
Figure 14 Cumulative frequency curve of the RI
boundary of 500 km2 was adopted. This
was applied for RMF K regions 2.8, 3.4, 3.6,
3.8 and 4.0. In regions 3.6, 3.8, 4.2 and 4.6 a
clear boundary between the flood zone and
transition zone could be defined.
The storm zone regional discharges
at 1 km2 on the RMF K regional envelope
diagram were significantly lower than
those in the Kovács (1988) RMF. However,
the regional discharges for the storm zone
determined in this study were found to
be significantly higher than the observed
flood peaks for gauging stations plotting
closer to 1 km2. This finding confirmed
that the regional 15-minute rainfall intensi-
ties used were indeed applicable.
It was observed that, for the higher K
regions, most of the plotted maximum
observed flood peaks in the flood zone
follow the trend of the K regional enve-
lope fairly well. The flood peaks plotting
towards the lower limit of the transition
zone were observed to be significantly
lower than the storm zone regional dis-
charge at 1 km2. This provided confidence
that the determined regional discharge at
the storm zone associated with 15-minute
rainfall intensity at 1 km2 is applicable,
even though these values are significantly
lower than the Kovács storm zone values.
Envelope equations
The derived RMF K regional envelope
equations are presented in Table 6 on
page 17 along with their applicable catch-
ment area range. The envelope curve repre-
sents the RMF that can be expected in each
of the K regions for any catchment size.
The applicable areal range corresponds to
the catchment area boundaries of zones on
the envelope curve diagrams in Figure 12.
The applicable area range in the transition
18
2 000 3 000 4 000 5 000 6 000 7 000 8 000 9 000 10 000
RMF return period (years)
zone is defined by the catchment area from
the storm zone to the flood zone.
Return period of the RMF
It was noted that the RMF values of gaug-
ing stations in a K region did not provide a
consistent return period, but varied depend-
ing on the historical flood peaks and the
catchment area. In the further analysis of
the return period of the RMF, the stations
(14 stations) with a return period of greater
than 1 in 500 000 years were stations with
very limited data or short record lengths and
were therefore excluded from the analysis.
Figure 13 on page 17 is a histogram of the
number of gauging stations with RMF fall-
ing in various ranges of return periods.
Figure 13 shows that 69 out of 466
stations had RMF values associated with
a return period in the range of 200 to 500
years, while 109 stations were in the range
of 10 000 to 50 000 years (both scenarios
respectively highlighted by boxes in the
figure). Most of the RMF values were
associated with return periods greater than
1 in 2 000 years. A cumulative frequency
diagram generated from the histogram
in Figure 13 is shown in Figure 14. From
Figure 14 it was observed that, for an
exceedance probability of 0.5 (the median
value), the return period associated with
the RMF is 5 000 years.
Application of the RMF
To calculate the RMF of a site of interest,
the RMF K regional map in Figure 9 and
the RMF K regional envelope equations in
Table 6 are used. This requires the following
actions:
QQ The geographical location and catch-
ment area of the sites of interest must
be determined.
Volume 65 Number 3 September 2023 Journal of the South African Institution of Civil Engineering
QQ The RMF K region of the site must be
identified from Figure 9.
QQ The RMF K regional envelope equa-
tion must be identified from Table 6,
corresponding to the identified RMF K
region. It should also be remembered to
identify the zone (transition or flood) of
the site, using its catchment area.
QQ Lastly, the required RMF should be
calculated, using the identified RMF K
regional equation for the correct zone,
by substituting the catchment area of
the site in the relevant equation.
QQ If the required RMF is for a site located
along the identified river section with a
different K value from that of the RMF
K region, the RMF K regional envelope
curve equation corresponding to the K
value of the river section (see Table 5)
should be used.
CONCLUSIONS
The objectives of this study were to develop
an improved approach to estimate the RMF
as an upper limit flood. A refined region-
alisation approach was used to regionalise
flood peaks, and 15 RMF K regions with
their associated envelope curves were
identified. The lowest RMF K region is 2.8
and the highest RMF K region is 5.8. The
homogeneity of the regions was evaluated,
and it was concluded that the regions were
indeed homogeneous. The envelope curves
in the flood zone corresponding to the
RMF K regions were drawn 15% above the
maximum observed flood peak. The transi-
tion zone envelope curves were drawn from
the K envelope curve in the flood zone to
the regional point discharge at 1 km2 in the
storm zone. It was evident from this study
that the regional discharge associated with
a 15-minute rainfall intensity at 1 km2 was
higher than the observed flood peaks that
plotted closest to the 1 km2 catchment areas.
This supported the regional discharges at
the storm zone. Envelope equations were
generated for each region, which can be
applied in estimating the RMF for any site
of interest. It was concluded that the return
period associated with median RMF peaks in
this study was estimated to be 1:5 000 years,
but ranged from less than 100 to more than
500 000 years, as opposed to the original
Kovács estimation of 1:200 years.
REFERENCES
Agarwal, A, Maheswaran, R, Sehgal, V, Khosa, R,
Sivakumar, B & Bernhofer, C 2016. Hydrologic
 regionalization using wavelet-based multiscale
entropy method. Journal of Hydrology, 53(8): 22–32.
Ahani, A, Nadoushani, S S M, Moridi, A 2020.
Regionalization of watersheds by finite mixture
models. Journal of Hydrology, 583: 124620.
Alexander, W J R 2002. The standard design flood.
Journal of the South African Institution of Civil
Engineering, 44(1): 26–30.
Alexander, W J R 2006. Climate change and its
consequences – An African Perspective. Technical
Report: University of Pretoria.
Benson, M A 1964. Factors Affecting the Occurrence
of Floods in the Southwest. Washington DC: United
States Government Printing Office.
Cloete, G C, Basson, G R & Sinske, S A 2014. Revision
of Regional Maximum Flood (RMF) estimation in
Namibia. Water SA, 40(3): 535–548.
Cordery, I & Pilgrim, D H 2000. The state of the art of
flood prediction. In Parker, D J (Ed). Floods. Vol II.
London: Routledge, 185–197.
Du Plessis, J A & Masule, S 2023. Evaluation of Kovács
1988 Regional Maximum Flood 556 Method.
Journal of the South African Institution of Civil
Engineering, 65(1): 35–44.
Francou, J & Rodier, J A 1967. Essai de classification des
crues maximales observées dans le monde. Cahiers
ORSTOM série Hydrologie, IV(3):19–46. Available
at: http://horizon.documentation.ird.fr/exl-doc/
pleins_textes/pleins_textes_4/hydrologie/14846.
pdf.
Gericke, O J & Du Plessis, J A 2013. Development of a
customised design flood estimation tool to estimate
floods in gauged and ungauged catchments. Water
SA, 39(1): 67–94.
Görgens, A H M, Lyons, S, Hayes, L, Makhabane, M &
Maluleke, D 2006. Modernised South African design
flood practice in the context of dam safety. Water
Research Commission Report, Project No K5/1420.
Pretoria: Water Research Commission.
Görgens, A 2007. Joint Peak-Volume (JPV) Design
Flood Hydrographs for South Africa. WRC
Report No 1420/3/07. Pretoria: Water Research
Commission.
GREHYS (Groupe de recherche en hydroclimatologie
statistique) 1996. Presentation and review of some
methods for regional flood frequency analysis.
Journal of Hydrology, 186(1–4): 63–85.
Journal of the South African Institution of Civil Engineering
Kachroo, R K, Mkhandi, S H & Parida, B P 2000.
Flood frequency analysis of southern Africa. Part I.
Delineation of homogeneous regions. Hydrological
Sciences Journal, 45(3): 437–447.
Kjeldsen, T R, Smithers, J C & Schulze, R E 2002.
Regional flood frequency analysis in the KwaZulu-
Natal province, South Africa, using the index-flood
method. Journal of Hydrology, 255(1–4): 194–211.
Kovács, Z P 1980. Maximum flood peak discharges in
South Africa: An empirical approach. Technical
Report TR 105. Pretoria: Department of Water
Affairs.
Kovács, Z P 1988. Regional maximum flood peaks in
southern Africa. Technical Report TR 137. Pretoria:
Department of Water Affairs.
Nobert, J, Mugo, M & Gadain, H 2014. Estimation
of design floods in ungauged catchments using a
regional index flood method. A case study of Lake
Victoria Basin in Kenya. Physics and Chemistry of
the Earth, 67–69: 4–11.
Nortje, J H 2010. Estimation of extreme flood peaks by
selective statistical analyses of relevant flood peak
data within similar hydrological regions. Journal of
the South African Institution of Civil Engineering,
52(2): 48–57.
Parak, M & Pegram, G 2006. The rational formula
from the run hydrograph. Water SA, 32(2): 163–180.
Parajka, J, Merz, R & Blöschl, G 2005. A comparison
of regionalisation methods for catchment model
parameters. Hydrology and Earth System Sciences
Discussions, 2(2): 509–542.
Pegram, G & Parak, M 2004. A review of the regional
maximum flood and rational formula using
geomorphological information and observed floods.
Water SA, 30(3): 377–392.
Roffe, S J, Fitchett, J M & Curtis, C J 2019. Classifying
and mapping rainfall seasonality in South Africa:
A review. South African Geographical Journal,
101(2): 158–174.
Rosli, A S, Aris, A, Salmiati & Haniffah, M R M 2019.
A review of regionalisation methods for ungauged
watersheds in the SWAT model. Proceedings,
International Conference on Environmental
Sustainability and Resource Security, 5–6
November, Malaysia, pp 183–188.
SANCOLD (South African National Committee on
Large Dams) 1990. Guideline on safety in relation to
Volume 65 Number 3 September 2023
floods. Safety Evaluation of Dams Report. Pretoria:
SANCOLD.
Smithers, J & Schulze, R 2002a. Regionalization for
one- to seven-day design rainfall estimation in
South Africa. Proceedings, FRIEND 2002 – Regional
Hydrology: Bridging the Gap between Research
and Practice. Cape Town: IAHS (International
Association of Hydrological Sciences) Publication
274, pp 173–179.
Smithers, J & Schulze, R 2002b. Design rainfall and
flood estimation in South Africa. Water Research
Commission Report, Project No K5/1060.
Pietermaritzburg, University of KwaZulu-Natal,
Report to the Water Research Commission.
Smithers, J C 2012. Methods for design flood
estimation in South Africa. Water SA, 38(4):
633–646.
Swanepoel, P 2017. The evaluation of regional extreme
flood events. Final year project. Stellenbosch:
Stellenbosch University.
Van der Spuy, D & Rademeyer, P F 2010. Flood
frequency estimation methods as applied in the
Department of Water Affairs. Pretoria: Department
of Water Affairs.
Van Dijk, M, Van Vuuren, S J & Smithers, J 2013. Flood
calculation. In SANRAL Drainage Manual (6th
edition). Pretoria: South African National Roads
Agency, 3-1 – 3-69.
Van Vuuren, S J, Van Dijk, M & Coetzee, G L 2013.
Status review and requirements of overhauling flood
determination methods in South Africa. Report
TT563/13. Pretoria: Water Research Commission.
Verwey, J W 2015. A critical evaluation of the
regionalisation of Kovács K E values. Unpublished
MEng Dissertation. Stellenbosch University.
Wang, Y 2000. Development of methods for regional
flood estimates in the province of British Columbia,
Canada. PhD Thesis. Vancouver: University of
British Columbia.
WR 2012. Water Resources of South Africa (WR2012).
Available at: https://waterresourceswr2012.co.za/
(accessed on 16 August 2020).
Zhang, Z & Stadnyk, T A 2020. Investigation of
attributes for identifying homogeneous flood
regions for regional flood frequency analysis in
Canada. Water Journal, 12(9): 25–70.
19
TECHNICAL PAPER The status of water for
JOURNAL OF THE SOUTH AFRICAN
INSTITUTION OF CIVIL ENGINEERING
firefighting in South Africa
ISSN 1021-2019 (print) | ISSN 2309-8775 (online)
Vol 65 No 3, September 2023, Pages 20–26, Paper 1620 A Ilemobade

PROF ADESHOLA ILEMOBADE (Pr Eng,

FSAICE, FWISA) is currently at the School of It is mandatory in South Africa that water for firefighting is included in the demand estimates
Civil and Environmental Engineering,
input when designing municipal water networks. To this end, guidelines for determining fire
University of the Witwatersrand,
Johannesburg, where he teaches flows for different risk categories have been developed. The most commonly used guidelines
infrastructure planning and water are the South African National Standard (SANS) 10090 of 2018 and the Department of Human
engineering subjects at undergraduate and
Settlements (DHS) guide of 2019 (also called the Red Book). The evolution of the built
postgraduate levels. His research interests
are in the areas of water conservation, water demand and water supply. environment, firefighting technologies and approaches, and the availability of freshwater
He is an NRF-rated researcher, has published extensively, received awards resources all require that the fire flows in the guidelines are current and adequately protect
for research conducted and supervised, has led several multi-disciplinary communities from the devastating impacts of fires – this while conserving depleting freshwater
research and consulting teams, and is an associate editor of several
resources and contributing to optimal water network design. Currently, empirical research into
initiatives.
the appropriateness of the fire flows in SANS 10090 and the Red Book to firefighting efforts
Contact details:
in South African municipalities is limited. This paper therefore compares water for firefighting
School of Civil and Environmental Engineering
University of the Witwatersrand in these two guidelines with some international companions and local firefighting data,
Private Bag 3 and also compares similar data from the six studies published on this topic in South Africa.
Wits Some highlights from this investigation include: (i) the fire risk categories and fire flows in
2050
South Africa
SANS 10090 and the Red Book are misaligned, and are contradictory in at least one instance;
E: adesola.ilemobade@wits.ac.za (ii) when compared with firefighting data, the fire flows in SANS 10090 and the Red Book
are conservative; (iii) a significant percentage of fires were extinguished between 30 and
120 minutes; (iv) 10 kℓ or less of water was employed to extinguish more than 87% of fires; and
(v) the frequencies of fires were positively related to dry and/or cold weather conditions.

BACKGROUND resources, not only for firefighting, but

Firefighting infrastructure, although rarely more importantly, for meeting basic needs,
used when compared to infrastructure sustaining economic development and
that provides basic services, protects com- maintaining environmental flows.
munities from the devastating impacts of South African municipalities are legally
fires. Sadly, fires have been shown to be obligated to provide water for firefighting
devastating on many levels, including socio- (DHS 2019), and thus water for firefight-
economic, financial, environmental and ing is given equal importance as drinking
fatalities. The Department of Cooperative water when designing water networks.
Governance (DCG 2020) estimates that Water networks are therefore designed for
R300 million is spent annually in South both peak water demand plus fire demand,
Africa to care for burns from paraffin stove and as a result this may increase network
incidents, while indirect costs such as lost capacity by as much as 20% (CSIR 2005).
wages, prolonged care for deformities, Over the past thirty years, only six
emotional trauma and commitment of fam- studies have been published in South
ily resources also contribute to the socio- Africa that analyse in detail, water for
economic impacts of fires. In 2021, financial firefighting in various communities. These
losses due to fires in South Africa were studies are: Van Zyl and Haarhoff (1997;
estimated at R23 billion (Walls et al 2022). 2007), Van Zyl et al (2011), Myburgh and
Several agents may be employed to Jacobs (2014), Mac Bean and Ilemobade
extinguish fires, with water being the most (2019), Essack and Ilemobade (2022), and
popular due to several reasons, including Thage and Ilemobade (2023). While these
its ease of availability, its relatively afford- studies have been driven by similar objec-
able cost, and its ability to extinguish most tives, their analyses have not produced
types of fires. Unfortunately, increasing results that can easily be compared and
aridity in many communities is negatively made generic to render firefighting efforts
impacting the availability of freshwater more efficient. This paper therefore

Keywords: water for firefighting, guidelines, fire flows, SANS 10090, Ilemobade A. The status of water for firefighting in South Africa. J. S. Afr. Inst. Civ. Eng. 2023:65(3),
Art. #1620, 7 pages. http://dx.doi.org/10.17159/2309-8775/2023/v65n3a2
Red Book
The Journal of the South African Institution of Civil Engineering, which is distributed internationally, is a peer-reviewed, open-access journal licensed under a Creative
Commons Attribution Licence (CC BY-NC-ND). Readers may therefore freely use and share the content as long as they credit the original creator and publisher, do
20 not change the material in any way, and do not use it commercially. Copyright of this article remains with the authors.
addresses two objectives: (i) it presents
the guidelines popularly used to estimate
water for firefighting in South Africa, and
compares same with some international
companions and local firefighting data, and
(ii) it compares similar data from the six
studies published in South Africa and dis-
cusses some implications for firefighting.
THE SOUTH AFRICAN
WATER FOR FIREFIGHTING
GUIDELINES AND SOME
INTERNATIONAL COMPANIONS
The South African National Standard speci-
fying water for firefighting in different com-
munities, SANS 10090, is titled Community
Protection Against Fire (SANS 2018). The
other well-known guideline specifying water
for firefighting is called the Red Book and
is titled The Neighbourhood Planning and
Design Guide (DHS 2019). When water for
firefighting values in both guidelines are
compared (Table 1), the following is evident:
i. Both guidelines present different fire
risk categories.
ii. The total fire flows in the Red Book are
lower than the minimum values in the
SANS 10090 for similar risk categories.
iii. Column 2 presents descriptions for “Fire
risk categories” and column 3 presents
descriptions for “Possible fire sizes”,
both of which are in Table 8 (Minimum
fire flow) of the SANS 10090. While
the descriptions for residential users
(Category D1, D2, D3 and D4) in col-
umns 2 and 3 are aligned, this is not
the case for non-residential users in
Categories A, B and C.
iv. Determination of the minimum fire flows
for Categories C and D1 in SANS 10090 is
misleading. For example, a residential area
of conventional construction, where hous-
es are spaced further than 30 m apart,
would have SANS 10090 specify two min-
imum fire flows of 6 000 ℓpm (Fire risk
categories, Category C) and 1 900 ℓpm
(Possible fire sizes, Category D1).
v. Both guidelines only address on-scene
water for firefighting. On-scene water
for firefighting refers to fire flow that is
calculated (using simple, scientifically-
based formulas) at the scene of the fire
based on fire area or volume. On-scene
water for firefighting allows firefighters
on the scene to assess whether they
need more hose lines or apparatus to
fight the fire (Benfer & Scheffey 2015).
Benfer and Scheffey (2015) examined several
fire flow calculation methods from the
USA, UK, France, Australia, Germany, the
Netherlands, New Zealand and Canada.
Eleven of the methods specifically addressed
pre-incident building planning, and at least
five addressed on-scene firefighting. The on-
scene firefighting methods were: the Iowa
State University method, ISU (USA); the
Illinois Institute of Technology method, IIT
(USA); the National Fire Academy method,
NFA (USA); the 3D Firefighting method
(USA/UK/Australia); and the Thomas,
Sardqvist and Baldwin methods (UK, UK
and USA respectively).
The fire flow methods were applied to
two differently sized non-residential build-
ings and two differently sized single-family
residential buildings. Their study included
both sprinklered and non-sprinklered
calculations. For on-scene firefighting,
highlights from their study, compared with
SANS 10090 and the Red Book, were: (i) fire
flows varied greatly across the countries
(Figures 1, 2 and 3); (ii) for residential
buildings, all on-scene firefighting methods
required the same fire flows irrespective of
whether the buildings were fitted with or
without sprinklers (Figure 3); and (iii) the
appropriate SANS 10090 and the Red Book
fire flows for non-residential buildings
(929 m2 and 4 645 m2) are lower than
most of the average fire flows determined
by Benfer and Scheffey (2015) (Figures 1
and 2). For residential buildings, while this

Table 1 The SANS 10090 (2018) and the Red Book (DHS 2019) fire flow specifications
The SANS 10090 (SANS 2018) The Red Book (DHS 2019)

Minimum
Total fire
Fire risk categories Possible fire sizes fire flow Risk classification
flow (ℓpm)
(ℓpm)

Non-residential
Central business districts and extensive commercial and High-risk: CBD and high-
A buildings with divisions 13 000 6 000
industrial areas normally found in cities and large towns risk industrial
not > 5 000 m2

Limited central business districts, smaller commercial or Non-residential Moderate-risk 1: Industrial,

B industrial areas normally associated with small towns and buildings with divisions 9 000 business, high-rise flats ≥ 3 000
decentralised areas, cities and large towns not > 2 500 m2 four storeys

Non-residential Moderate-risk 2: Cluster &

C Residential areas of conventional construction buildings with divisions 6 000 low-income housing, high- 1 500
not > 1 250 m2 rise flats ≤ three storeys

Rural area of limited buildings and remote from urban Low-risk: Single residential
D1 Houses > 30 m apart 1 900 900
areas; houses > 30 m apart housing

Rural area of limited buildings and remote from urban Houses 10.1 m to 30 m
D2 2 850 – –
areas; houses 10.1 m to 30 m apart apart

Rural area of limited buildings and remote from urban Houses 3 m to 10 m

D3 3 800 – –
areas; houses 3 m to 10 m apart apart

Rural area of limited buildings and remote from urban

D4 Houses < 3 m apart 5 700 – –
areas; houses < 3 m apart

Special risk areas; individual areas requiring a pre-

determined attendance over and above the predominant As determined
As determined
risk category in an area – these include large shopping/ by individual
E by individual risk – –
entertainment centres, informal settlements, harbours, risk
assessment
hospitals, prisons, large airport buildings and assessment
petrochemical plants

Journal of the South African Institution of Civil Engineering Volume 65 Number 3 September 2023 21
observation is similar for the Red Book total
fire flow for the “Low-risk: Single residential 14 000
housing” category, it is not for the minimum 12 000

Litres per minute (ℓpm)

fire flow for the SANS 10090 Category C
10 000
(Figure 3). In summary, the comparisons
shown in Figures 1, 2 and 3 illustrate that, in 8 000
general, the SANS 10090 and the Red Book 6 000
fire flows are lower than the average fire
4 000
flows determined by Benfer and Scheffey
(2015) for more developed countries. 2 000

0
ISU ITT NFA 3D Thomas Baldwin Sardqvist
Firefighting
STUDIES ON WATER FOR
Average fire flow for a 929 m2 sprinklered or non-sprinklered non-residential building
FIREFIGHTING IN SOUTH AFRICA The SANS 10090 (2018) minimum fire flow for Category C - Non-residential buildings
Van Zyl and Haarhoff (1997; 2007) investi- with divisions not > 1 250 m2
gated fire incidents spanning 1980 to 1991 The Red Book (DHS 2019) total fire flow for high-risk: CBD and high-risk industrial
within a section of the Central Business
District (CBD) in the City of Johannesburg Figure 1 Benfer and Scheffey’s (2015) on-scene fire flows for a 929 m2 non-residential building
(CoJ), Gauteng, and a residential area in the compared with the appropriate fire flows in SANS 10090 and the Red Book
Province of Mpumalanga. Highlights of
their study were: (i) extinguishing large fires
in the CoJ over the 11-year period typically 80 000
took between 30 and 90 minutes, with fire 70 000
Litres per minute (ℓpm)

flows ranging between 300 and 2 100 ℓpm 60 000

(large fires are fires that require more than 50 000
5 kℓ to extinguish); (ii) 90% of the large
40 000
fires investigated were extinguished using
30 000
fire flows of 3 120 ℓpm or less and 440 kℓ
of water or less; and (iii) peak consumer 20 000
demands were observed only at small inter- 10 000
vals during the year, and usually did not 0
coincide with fire events that typically took ISU ITT NFA 3D Thomas Baldwin Sardqvist
Firefighting
place in mid-winter.
Average fire flow for a 4 645 m2 sprinklered or non-sprinklered non-residential building
Van Zyl et al (2011) studied water for
The SANS 10090 (2018) minimum fire flow for Category A - Non-residential buildings
firefighting in the Greater City of Cape with divisions not > 5 000 m2
Town area between 1 January 2005 and The Red Book (DHS 2019) total fire flow for high-risk: CBD and high-risk industrial
21 April 2010. Highlights of their study
include: (i) vegetation fires predominated  enfer and Scheffey’s (2015) on-scene fire flows for a 4 645 m2 non-residential building
Figure 2 B
the categories investigated and occurred compared with the appropriate fire flows in SANS 10090 and the Red Book
mostly during the summer (no rain) months
when weather conditions were conducive to
vegetation fires; and (ii) 98%, 95% and 87% of 14 000
residential, commercial and industrial fires 12 000
Litres per minute (ℓpm)

respectively were extinguished in 120 min-

10 000
utes or less. Industrial fires had the longest
durations and required the largest volumes 8 000
of water to extinguish, while residential fires 6 000
had the shortest durations and required the
4 000
smallest volumes of water.
Myburgh and Jacobs (2014), employing 2 000
a filtered database containing 554 records 0
with water volumes and 546 records with ISU ITT NFA 3D Thomas Baldwin Sardqvist
Firefighting
computed fire flows from 1 January 2010 to
Fire flow for a 325 m2 sprinklered or non-sprinklered single-family home
31 December 2010, analysed fire incidents
Fire flow for a 139 m2 sprinklered or non-sprinklered single-family home
within five towns (i.e. three municipali-
The SANS 10090 (2018) minimum fire flow for Category C - Residential areas of
ties) in the Province of the Western Cape conventional construction
(WC). Prior to filtering, the total number of The Red Book (DHS 2019) total fire flow for low-risk: Single residential housing
records were 2 450. Highlights were: (i) more
than 90% of fires were extinguished using  enfer and Scheffey’s (2015) on-scene fire flows for a 325 m2 and 139 m2 single-family
Figure 3 B
less than 10 kℓ of water while the average home compared with the appropriate fire flows in SANS 10090 and the Red Book

22 Volume 65 Number 3 September 2023 Journal of the South African Institution of Civil Engineering
water volume used to extinguish 77 large
Category 2 (structural and industry) fires
was 8.60 kℓ; and (ii) 91.4% of the 546 fires
were extinguished using non-­reticulated
water supplies, i.e. municipal water supply
was only directly used to extinguish 8.6%
of the total number of fires and this was
achieved because fire fighters used the pre-
filled water tanks of emergency vehicles as
the first line of defence in all fire calls.
Mac Bean and Ilemobade (2019) ana­
lysed about ten years (from 1 January 2006
to 30 September 2017) of 3 859 billable fire
incident reports that occurred within the
CoJ. The initial dataset comprised 4 556
records. Some highlights of their study were:
(i) about 75% of fire incidents were extin-
guished using 6.60 kℓ of water or less; 87% of
fires were extinguished using 10 kℓ of water
or less, and 99% of fires were extinguished
using 100 kℓ of water or less; (ii) while the
frequency of fire occurrence was strongly
related to weather conditions, the volume
of water used to extinguish fires was not;
(iii) fire flows for 99.90% and 99.60% of the
fire incidents were less than the minimum
and total fire flows for the lowest fire risk
categories in SANS 10090 and the Red Book
respectively; and (iv) the peak of the start
times of fires did not coincide with the
typical peak period when residential water
demand occurred (in fact, the least number
of fires occurred during the typical peak
period of residential demand).
Essack and Ilemobade (2022), using
better representative criteria, re-analysed
14 000
SANS 10090 Category A Fire Flow
13 000
12 000
11 000
10 000
SANS 10090 Category B Fire Flow
9 000
Flow rate (ℓ/min)
8 000
7 000
SANS 10090 Category C Fire Flow
6 000
SANS 10090 Category D4 Fire Flow
5 000
4 000 SANS 10090 Category D3 Fire Flow
SANS 10090 Category D2 Fire Flow
3 000
SANS 10090 Category D1 Fire Flow
2 000
1 000
0
at CoJ for large structural and industry fires (Essack & Ilemobade 2022)
at 5 WC towns for large structural fires (Myburgh & Jacobs 2014)
Figure 4 Fire flows from four studies superimposed on the SANS 10090 minimum and the Red Book total fire flows
Journal of the South African Institution of Civil Engineering
the initial CoJ database employed by
Mac Bean and Ilemobade (2019), and in
particular addressed 89 large fires from
an initial dataset of 4 479 records (span-
ning 24 February 2003 to 26 September
2017). Some highlights of their study were:
(i) an average of 32 incidents occurred per
30-day month, with an increase in the
average number of fire incidents during
Johannesburg’s dry and cold months (i.e.
May to September); the higher average
number of fires during the cold months
was attributed to households using unsafe
appliances for heating and cooking, espe-
cially in informal settlements; and (ii) fire
flows ranged between 0 and 180 ℓpm.
With particular reference to large fires,
additional highlights were: (iii) the duration
of 60% of the large fires was between 30
and 120 minutes; (iv) the average volume of
water employed to extinguish the large fires
was 9.63 kℓ; (v) the fire flows employed to
extinguish the 89 large fires were less than
the minimum and total fire flows for the
lowest fire risk categories in SANS 10090
and the Red Book respectively; and (vi)
municipal water was used to supplement
water conveyed in the vehicles in about 99%
of 84 large structural and industry fires.
Thage and Ilemobade (2023) analysed
3 236 fire incident reports at the Sol Plaatje
Municipality (SPM) in the Northern Cape
during the period 21 July 2017 to 27 August
2020. After filtering, 2 060 reports remained.
Some highlights of this study were: (i) the
SPM experienced the highest number of fires
at CoJ for all fires (Mac Bean & Ilemobade 2019)
at SPM for structural and industry fires (Thage & Ilemobade 2023)
Volume 65 Number 3 September 2023
in its predominant land uses, i.e. open veld
(bush, grass, vacant/unspecified) and then
the built environment; (ii) the average fire
flow employed at the SPM to extinguish each
category of fire was notably lower than the
fire flow specified in SANS 10090 and the
Red Book – this highlight is corroborated by
the above studies; (iii) the average volume of
water employed to extinguish all categories
of fires at the SPM was 6 kℓ or less, with
93.3% of SPM fires extinguished using 7 kℓ
of water or less; (iv) the average number of
fire incidents in SPM peaked during the
winter (July and August) and early spring
(September) months. These months experi-
ence the least rain and the highest energy
(electricity, wood, paraffin) consumption for
heating and cooking use; (v) while fires in
the SPM typically occurred between 11h00
and 20h00, the peak of the start times of
fires during the course of a typical day was
observed between 15h00 and 16h00. As a
result, while the peak of the start times of
fires and the peak of the residential water
demand do not coincide in the morning,
this is not the case in the afternoon; and
(vi) about 81% of fires were extinguished in
less than 135 minutes.
COMPARISON OF THE
HIGHLIGHTS OF THE SOUTH
AFRICAN STUDIES
A fire incident report (also called fire infor-
mation report form) is a physical document
that is filled out and typically submitted
Red Book High Risk Fire Flow
Red Book Moderate Risk 1 Fire Flow
Red Book Moderate Risk 2 Fire Flow
Red Book Low Risk Fire Flow
23
 Typically, on-site calculations, based on
50 water meter readings taken from different
45
Average number of incidents
appliances, are carried out to estimate the
40 total volume of water released over the
35
duration of the incident. These calcula-
30
tions account for water obtained from fire
25
trucks, water tankers and hydrants (Mac
20
Bean & Ilemobade 2019). Seeing that physi-
15
10 cal records were employed (and still are)
5 to capture details of fire incidents in the
0 different communities, it is impossible to
verify that the physical records provided
January

February

March

September
April

May

June

July

August

October

November

December
represent all the fire incidents that took
place within those communities.
Month Figure 4 on page 23 shows average
at CoCT (Van Zyl et al 2011) at CoJ (Mac Bean & Ilemobade 2019) fire flows employed to extinguish fires in
at CoJ (Essack & Ilemobade 2022) at SPM (Thage & Ilemobade 2023) Myburg and Jacobs (2014), Mac Bean and
Ilemobade (2019), Essack and Ilemobade
Figure 5 A
 verage number of fire incidents for each representative month of the year over the (2022), and Thage and Ilemobade (2023),
study periods plotted alongside the minimum and total
fire flows in SANS 10090 and the Red
to emergency services after each fire Fire incident reports that were legible, Book, respectively. What is plainly obvious
incident attended to by the fire brigade (a correctly completed, and had entries that is that, over the approximately 15 years of
template is presented in SANS 10090 2018: were considered reasonable, were assumed data spanning 2003 to 2017, the minimum
Annexure B.2). Only billable fire incident to be a true reflection of the fires attended and total fire flows in SANS 10090 and the
reports were employed in the above studies. to by the emergency services. In the stud- Red Book, respectively, are conservative
Billable fire incident reports are incidents ies undertaken in the CoJ (Mac Bean & (i.e. much higher than actual). Since these
that the fire department charges (or should Ilemobade 2019; Essack & Ilemobade 2022), conservative fire flows must be incorpo-
have charged) the property owner for the five WC towns (Myburgh & Jacobs rated into the design of water networks,
firefighting services rendered and is best 2014), and the SPM (Thage & Ilemobade it is inevitable that larger water network
explained by Sections 10.1 and 10.2 (Fees) 2023), random and difficult-to-understand components (and consequently, longer
of the Fire Brigade Services Act No 99 data in some fire incident reports were retention times that negatively impact
(Republic of South Africa 1987). verified with emergency services personnel. water quality) will result.

9 2.50

8 2.25

2.00
7
Demand as ratio of average for day
Percentage of fire incidents (%)

1.75
6
1.50
5
1.25
4
1.00
3
0.75

2
0.50

1 0.25

0 0
00:00

01:00

02:00

03:00

04:00

05:00

06:00

07:00

08:00

09:00

10:00

11:00

12:00

13:00

14:00

15:00

16:00

17:00

18:00

19:00

20:00

21:00

22:00

23:00

24:00

Time
Fire incident start times at CoJ (Mac Bean & Ilemobade 2019) Business/commercial/industrial – large
Residential large stands Residential medium stands Fire incident start time at SPM (Thage 2022)
Residential small stands Low-income housing Business/commercial/industrial – medium

Figure 6 Daily variation of fire start times versus 24-hour demand patterns for a selection of land uses in the Red Book (DHS 2019)

24 Volume 65 Number 3 September 2023 Journal of the South African Institution of Civil Engineering
 18 22.50

16 20.00

14 17.50

12 15.00

Percentage of fires
Number of fires

10 12.50

8 10.00

6 7.50

4 5.00

2 2.50

0 0
0–15

15–30

30–45

45–60

60–75

75–90

90–105

105–120

120–135

135–150

150–165

165–180

180–195

195–210

210–225

225–240

240–255

255–270

270–285

285–300
Duration of fire (minutes)
at CoJ for large structural and industry fires (Essack & Ilemobade 2022) at CoJ for large fires (Van Zyl & Haarhoff 1997; 2007)
at SPM for large structural and industry fires (Thage & Ilemobade 2023)

Figure 7 Distribution of fire durations

The average volume of water employed for the WC displays marginally lower (Van Zyl & Haarhoff 1997, 2007; Essack &
to extinguish 77 large structural fires in the frequencies between March to October Ilemobade 2022) and at the SPM (Thage &
WC study was 8.60 kℓ (Myburgh & Jacobs when compared with frequencies at other Ilemobade 2023). In the three studies, more
2014); 75% of fires were extinguished in the times of the year. These results confirm the than 60% of the large fires were extin-
CoJ using 6.6 kℓ of water or less (Mac Bean & impact dry and/or cold weather has on the guished between 30 and 120 minutes, while
Ilemobade 2019); the average volume of water frequency of fires. It therefore makes sense in the CoJ, Essack and Ilemobade reported
employed to extinguish fires in the CoJ was for emergency services to annually plan and that ~20% of large fires were extinguished
9.63 kℓ (Essack & Ilemobade 2022); and 93.3% prepare for increased fire events occurring between 120 and 165 minutes.
of fires were extinguished in the SPM using during the dry and/or cold periods.
7 kℓ of water or less (Thage & Ilemobade Figure 6 shows that, while the peak
2023). Similarly, in the Davis (2000) study of the start times of fires and the peak CONCLUSION
in New Zealand, the average volume of of the typical residential water demands This paper addressed two objectives. Each
water used to extinguish 89.5% of 4 336 fire (DHS 2019) do not coincide in the morning objective, and key highlights from address-
incidents was 10.228 kℓ. The implication (where the larger of the two residential peak ing each, are presented below:
of these results is that a large percentage of demands occur in the diurnal pattern),
fires in these studies could be extinguished this is not the case in the afternoon, as the Objective 1:
using water conveyed in the tanks of certain start times of fires in the CoJ (Mac Bean To present the SANS 10090 (SANS 2018)
emergency vehicles employed to extinguish & Ilemobade 2019) and SPM (Thage & and the Red Book (DHS 2019)
fires. For example, some water tankers can Ilemobade 2023) peak at 15h00, about an guidelines for estimating water for
transport up to 10 kℓ of water. hour prior to the afternoon peak demand. firefighting in South Africa, and to
Figure 5 shows that at the CoJ (Mac This is likely because heating (during the compare same with some international
Bean & Ilemobade 2019; Essack & Ilemobade dry and cold months) and cooking facilities companions and local firefighting data.
2022) and at the SPM (Thage & Ilemobade are mostly employed in the afternoons when
2023), the frequency of fire incidents children/families start to arrive home from Key highlights:
increases and peaks during the dry months school/work. The peaks of the start times i. The fire risk categories and fire flows in
(April/May to September) and is the low- for the commercial/business/industrial SANS 10090 and the Red Book are not
est during the rainy periods (December to water demands also do not coincide with aligned and are contradictory in one
March). In the WC (Van Zyl et al 2011), the peaks of the start times of fires. instance. It is imperative therefore that
the rainy months are between March and Figure 7 shows the frequency distribu- this misalignment is speedily resolved
November, and this explains why the chart tion of the duration of large fires at the CoJ to prevent the haphazard selection of

Journal of the South African Institution of Civil Engineering Volume 65 Number 3 September 2023 25
 fire flows from either guideline during
water network design.
ii. Minimum and total fire flows in SANS
10090 and the Red Book are conserva-
tive (i.e. much larger than the average
fire flows employed to extinguish
fires in the six studies reviewed) and
therefore do not promote the optimal
design of water networks. For SANS
10090 this may be attributed to the fact
that the fire flows in the first edition
were likely over-estimated for South
Africa since they were compiled with
the assistance of organisations from
the United Kingdom, United States of
America, Canada, New Zealand and
Germany, and have not notably changed
since then. For the Red Book, while fire
flows were recently revised, the empiri-
cal data presented in this paper shows
that these values are still conservative.
While adequate fire protection must be
provided to communities, it is recom-
mended that the current fire flows be
reduced in light of dire water scarcity
and the need to design optimal water
networks.
Objective 2:
To compare data from the six studies
published in South Africa and to
discuss implications for firefighting.
Key highlights:
i. In four of the six studies, the typical
duration of fires did not exceed 120
minutes.
ii. The volume of water employed to extin-
guish 90%, 87% and 96% of fires in five
towns in the WC, the CoJ and the SPM
respectively, was 10 kℓ or less. These
results are especially important for
developed communities that experience
sub-optimal and/or erratic municipal
water supply, as the possession of func-
tional fire engines that can convey at
least 10 kℓ of water to site will result in
extinguishing a significant percentage
of fires in these communities.
iii. The frequencies of fires are related
to weather conditions, with a higher
26
number of fires occurring during the
cold and dry months.
iv. Since there is a legal requirement to
design water networks to cater for both
peak and fire demand, further research
into this aspect is necessary. If empiri-
cal firefighting data in other communi-
ties is similar to that shown in Figure 6
for the CoJ and SPM, this may provide
additional evidence for reducing the
peak factor employed in water network
design.
ACKNOWLEDGEMENTS
Access to data has been fundamental
to this study, and therefore emergency
ser­v ices in the following municipalities
are gratefully acknowledged for provid-
ing access to primary data: the City of
Johannesburg, the City of Cape Town, the
five towns in the Western Cape, and the
Sol Plaatje Local Municipality.
ETHICS CLEARANCE
Ethical clearance for each of the studies
employed in the synthesis undertaken
herein was separately obtained from the
relevant authorities and acknowledged in
each publication.
REFERENCES
Benfer, M E & Scheffey, J L 2015. Evaluation of
Fire Flow Methodologies. Springer Briefs in
Fire Series. New York: Springer. https://doi.
org/10.1007/978-1-4939-2889-7.
CSIR (Council for Scientific and Industrial Research)
2005. Guidelines for human settlement planning and
design. Pretoria: CSIR Building and Construction
Technology.
Davis, S K 2000. A review of firefighting water
requirements. MEng Thesis. New Zealand
University of Canterbury. https://ir.canterbury.
ac.nz/items/88da908b-9520-4883-91e6-3bc87a
494439.
DCG (Department of Cooperative Governance) 2020.
White Paper on Fire Services. Notice 515 of 2020.
Pretoria: DCG. https://www.gov.za/sites/default/
files/gcis_document/202009/43734gen515.pdf.
Volume 65 Number 3 September 2023 Journal of the South African Institution of Civil Engineering
DHS (Department of Human Settlements) 2019.
The Neighbourhood Planning and Design Guide:
Creating Sustainable Human Settlement (the Red
Book). Pretoria: DHS. https://www.csir.co.za/
neighbourhood-planning-and-design-guide-red-
book.
Essack, F & Ilemobade, A 2022. Water for fighting
large structural fires: A South African study. Urban
Water Journal. https://doi.org/10.1080/157306
2X.2022.2050928.
MacBean, C B & Ilemobade, A A 2019. An evaluation
of the primary South African standard and
guideline for the provision of water for firefighting.
Water SA, 45(4): 691–699. https://doi.org/10.17159/
wsa/2019.v45.i4.7551.
Myburgh, H M & Jacobs, H E 2014. Water for
firefighting in five South African towns. Water SA,
40(1): 11–17. https://doi.org/10.4314/wsa.v40i1.2.
Republic of South Africa 1987. Fire Brigade Services
Act 99 of 1987. Government Gazette 11006. Pretoria:
Government Printer. https://www.gov.za/sites/
default/files/gcis_document/201503/act-99-1987.
pdf.
SANS (South African National Standard). SANS
100902018. Community Protection Against Fire
(Edition 3.1). Pretoria: SABS Standards Division.
Thage, T & Ilemobade, A 2023. Water for firefighting
in the Sol Plaatje Municipality. Civil Engineering,
31(5): 19–22. https://www.flipsnack.com/
BE95BDAA9F7/civil-engineering-june-2023/full-
view.html.
Van Zyl, J E & Haarhoff, J 1997. South African fire
water guidelines and their impact on water supply
system cost. Journal of the South African Institution
of Civil Engineering, 39(1): 16–22.
Van Zyl, J E & Haarhoff, J 2007. Reliability analysis
of municipal storage reservoirs using stochastic
analysis. Journal of the South African Institution of
Civil Engineering, 49(3): 27–32. https://saice.org.za/
downloads/journal/vol49-3-2007/vol49_n3_d.pdf.
Van Zyl, J E, Davy, M & Haihambo, F 2011. Water
used for firefighting in Cape Town. Proceedings,
Computing and Control for the Water Industry
(CCWI) Conference, University of Exeter, UK,
pp 107–112.
Walls, R, Evans, J & Quiroz, N F 2022. SAICE
Infrastructure Report Card 2022 – Fire
engineering infrastructure summary report.
Civil Engineering, 30(2): 58–63. https://saice.org.
za/downloads/monthly_publications/2022/
Civil-Engineering-March-2022/59/.
Experimental comparison TECHNICAL PAPER
of active seismic surface JOURNAL OF THE SOUTH AFRICAN
INSTITUTION OF CIVIL ENGINEERING

wave tests on shallow ISSN 1021-2019 (print) | ISSN 2309-8775 (online)

Vol 65 No 3, September 2023, Pages 27–38, Paper 1519

and deep bedrock sites MOSITO NTAOTE (AMSAICE) is a

geotechnical engineer in South Africa and a
member of the SAICE Geotechnical Division.
He holds BEng, BEng(Hons) and MEng
M C Ntaote, G Heymann degrees from the University of Pretoria. He
practises mainly in the field of tailings
where he is involved with design and
assessment of the stability of tailings dams
Active seismic surface wave tests are tests during which signals are generated by artificial and other mine residue deposits. This paper formed part of his MEng
sources such as sledgehammers, drop weights and fixed or variable frequency shakers. research project at the University of Pretoria.
A number of active seismic surface wave tests have become popular in the geotechnical Contact details:
industry. Interestingly, different surface wave tests are popular in different regions of the world, Department of Civil Engineering
presumably because the tests require different equipment, expertise and numerical analysis University of Pretoria
Private Bag X20
techniques. Few studies have been done to compare seismic surface wave tests directly. This
Hatfield, Pretoria
study investigated the performance of three active seismic surface wave tests, namely the 0028
spectral analysis of surface waves (SASW), multi-channel analysis of surface waves (MASW) and South Africa
continuous surface wave (CSW) tests. This paper reports the performance of the three tests E: mosito@ecoe.co.za / mositontaote@gmail.com

with regard to repeatability, susceptibility to near-field effects, and maximum and minimum
PROF GERHARD HEYMANN (Pr Eng, FSAICE)
measurement depths at both a shallow and a deep bedrock site.
is a professor in the Department of Civil
Engineering at the University of Pretoria. He
holds BEng, BEng(Hons) and MEng degrees
INTRODUCTION surface wave methods can be divided into from the University of Pretoria, and a PhD
from the University of Surrey. He has been
Numerous laboratory and in situ tests are two broad categories – active and passive
involved with teaching and research in
used to characterise geomaterial behaviour. types – and they differ by means of the geotechnical engineering for many years.
These include seismic tests which measure signal source. Active tests use a signal His fields of interests include the characterisation of soil behaviour and its
the velocity at which mechanical seismic that is generated artificially by means of application in geotechnical engineering. He is a past chairman of the
SAICE Geotechnical Division. In 2002 and 2016 he received the Jennings
waves are propagated in the material. a sledgehammer, a weight drop or a fixed Award for the best geotechnical paper by a South African author, and in
Parameters determined from these tests or variable frequency shaker. For passive 2018 he received the South African Geotechnical Medal for his
can be utilised in dynamic problems such tests, the signal source is environmental contribution to geotechnical engineering in South Africa.
as site response evaluation, liquefaction and is mainly due to different causes, such Contact details:
potential evaluation, earthquake engineer- as wind, sea wave motion, vibrations from Department of Civil Engineering
University of Pretoria
ing and machine foundations (Strobbia structures and vehicle traffic, and other
Private Bag X20
2003). Another important application of cultural activities (Bignardi 2011). Hatfield, Pretoria
seismic tests is the determination of soil Active seismic surface wave methods 0028
stiffness which can be used in the design have gained popularity internationally and South Africa
E: gerhard.heymann@up.ac.za
of geotechnical structures such as founda- are widely used in geotechnical engineering.
tions, excavations and tunnels to predict These tests have a number of advantages,
ground movements, both during and after including the fact that they do not require
construction (Matthews et al 1996). drilling, which results in significant time
Seismic testing can be done in the and cost savings. In addition, surface wave
field, as well as in the laboratory. When tests avoid any potential environmental
conducted in the field, the tests can be hazards associated with drilling. Also, since
intrusive or non-intrusive. Intrusive meth- they do not require sampling, no sample
ods require a drilled borehole or insertion disturbance occurs, making it an attractive
of a probe into the soil. However, seismic test for hard-to-sample geomaterials. Unlike
surface wave tests are non-intrusive, as the some other geophysical methods, surface
seismic source and receivers are all located wave methods are capable of detecting soft
on the surface of the ground. Methods layers beneath stiffer layers, as well as soft
that measure the propagation of Raleigh layers sandwiched between two stiff layers.
waves have gained significant popularity The three most widely used active seismic
(Foti et al 2018; Stokoe et al 2004). Seismic surface wave tests are: spectral analysis of

Ntaote MC, Heymann G. Experimental comparison of active seismic surface wave tests on shallow and deep bedrock sites. Keywords: SASW, MASW, CSW, near-field effects, measurement
J. S. Afr. Inst. Civ. Eng. 2023:65(3), Art. #1519, 12 pages. http://dx.doi.org/10.17159/2309-8775/2023/v65n3a3
depth
The Journal of the South African Institution of Civil Engineering, which is distributed internationally, is a peer-reviewed, open-access journal licensed under a Creative
Commons Attribution Licence (CC BY-NC-ND). Readers may therefore freely use and share the content as long as they credit the original creator and publisher, do
not change the material in any way, and do not use it commercially. Copyright of this article remains with the authors. 27
surface waves (SASW), multi-channel analysis
of surface waves (MASW), and the continu-
ous surface wave (CSW) method (Stokoe et al
2004). The advantages and limitations of each
of these tests have been well documented
in the literature. However, few direct com-
parisons of these tests have been reported.
In cases where such comparisons have been
made, tests were not performed with the same
equipment, time sampling parameters, array
configurations (spatial sampling range) and
sometimes not on sites of the same geological
nature (Kim et al 2013; Omar et al 2011). This
paper reports the findings of an investigation
comparing the performance of these three
tests as objectively as possible.
SITE DESCRIPTIONS
Shallow bedrock site
SASW, MASW and CSW tests were con-
ducted on the University of Pretoria experi-
mental farm near the Civil Engineering
Laboratory (Engineering 4.0). The site
topography is generally flat, with vegetation
comprising mainly grass and shrubs, with
some trees. This is a shallow bedrock site
with residual lava soils that had formed from
the weathering of andesite lava, with the
bedrock at a depth of about 3.0 m. The site
is relatively quiet, as the only source of noise
is the traffic from the N1 and N4 highways
on the northern boundary of the site. This
background noise was minimal, as the test
site was about 230 m away from the nearest
highway. The coordinate of the test position
is 25°44’36.57”S, 28°15’34.48”E.
Deep bedrock site
The deep bedrock site is located near
Vredefort in the Free State Province of South
Africa. This site is located in the vicinity of
an inactive bentonite mine near positions
where piles had been installed to study their
behaviour in expansive clays. The topography
of this site is flat with vegetation consisting
of dense grass and some scattered small
trees. The bedrock is at a depth of about 12.0
m. The site was also quiet, as it is located
far from any public roads or other activities
that may cause background noise. The coor-
dinate of the test position is 27°15’15.95”S,
27°16’00.90”E.
EQUIPMENT
For the continuous surface wave (CSW)
tests two mechanical shakers were used,
namely a low- and a high-frequency shaker.
28
Table 1 CSW time sampling parameters
Frequency Frequency
range (Hz) increments (Hz)
10 – 20 0.5
20 – 22 1.5
22 – 45 1.5
45 – 50 2 1 000
50 – 70 2 2 000
70 – 90 3 2 000
The low-frequency shaker had a total
mass of 80 kg and a peak force of 6.3 kN
at a frequency of 22 Hz and was operated
between 10 Hz and 22 Hz. The high-
frequency shaker had a total mass of 70 kg
and a peak force of 8.0 kN at a frequency of
90 Hz and was operated in the frequency
range of 22.5 Hz to 90 Hz. A variable
frequency drive (VFD) was used to adjust
the frequency of the shakers as required.
For the SASW and MASW tests, the
seismic energy was produced by impact-
ing a steel plate 185 mm in diameter and
30 mm thick, placed on the ground, with
2.2 kg, 6.3 kg and 9.2 kg sledgehammers.
Seismic signals were acquired using 4.5 Hz
resonant frequency geophones with spikes
to hold them firmly into the ground. The
signals detected by the geophones were
recorded with a PASI GEA 24 seismograph.
EXECUTION OF FIELD TESTS
SASW test execution
Common midpoint two-receiver SASW
tests were conducted for this study. The
tests were done for six sets of receiver spac-
ings which were 1 m, 2 m, 4 m, 8 m, 16 m
and 32 m. For each receiver spacing the
ground was impacted three times on both
sides of the centreline of the receivers using
each of the 2.2 kg, 6.3 kg and 9.2 kg sledge-
hammers. The reason for the three shots
was not for stacking of the data to improve
the signal-to-noise ratio, but rather for
assessment of the repeatability of the SASW
test. Data was recorded for a period of 2
seconds with a logging frequency of 500 Hz.
MASW test execution
Two sets of MASW tests were conducted
using a linear array of 24 geophones. Some
researchers have suggested using a geophone
spacing of 1 m, while others suggested a
spacing of 2 m. For this reason, two sets of
MASW tests were conducted using both
Volume 65 Number 3 September 2023 Journal of the South African Institution of Civil Engineering
Logging
Acquisition time (s)
frequency (Hz)
500 10
500 10
1 000 5
5
3
3
1 m and 2 m geophone spacings. Multiple
source offsets for both sets of tests were
used. These were 1 m, 2 m, 4 m, 8 m and
16 m for the 1 m geophone spacing, and
2 m, 4 m, 8 m, 16 m and 32 m for the 2 m
geophone spacing. The variation of source
offset was for investigation of near-field
effects. Also, each one of the 2.2 kg, 6.3 kg
and 9.2 kg sledgehammers was used at each
source offset, impacting the steel plate three
times for the assessment of the repeatability
of the MASW test. As with the SASW test,
the data was recorded for a period of 2 sec-
onds with a logging frequency of 500 Hz.
CSW test execution
Monotonic and sweep CSW tests were done
using an array of five geophones spaced at
1 m. For a monotonic test the frequency at
which the shaker vibrates is held constant,
and for a sweep test the frequency is varied
from the lowest frequency of interest to the
highest. The low- and high-frequency shak-
ers were used to provide ground vibration
in the frequency ranges of 10 to 22 Hz, and
22.5 to 90 Hz, respectively. Table 1 shows
time sampling parameters for the different
frequency ranges in which the shakers were
operated. The tests were done at three dif-
ferent source offsets, which were 1 m, 2 m
and 4 m for the investigation of near-field
effects.
RESULTS AND DISCUSSION
Repeatability comparison
In this study, SASW data and CSW
data were analysed using Python codes
developed in-house using procedures
that have been well documented in the
literature (McCaskill 2014; Groenewold
2016; Strobbia 2003; Hebeler & Rix 2001;
Heisey et al 1982; Bouazza & Kavazanjian
2000; Heymann 2007). The Python codes
are available in Ntaote (2022). The MASW
data was analysed using Geopsy software
 500
400
Phase velocity (m/s)
300
200
100
0 5 10 15 20
Frequency (Hz)
Figure 1 G
 eopsy dispersion overtone image (left); extracted dispersion curve (right)
(www.geopsy.org) to produce dispersion
overtone images. The dispersion data was
extracted from the overtone images using
the Geopsy extraction algorithm. A typical
Geopsy’s overtone image and the extracted
dispersion data are shown in Figure 1.
The following nomenclature is used for
CSW and MASW throughout the article:
QQ CSW1m – CSW performed with the
source at 1 m offset
QQ CSW2m – CSW performed with the
source at 2 m offset
QQ CSW4m – CSW performed with the
source at 4 m offset
QQ MASW1m – MASW performed with
1 m geophone spacing
QQ MASW2m – MASW performed with
2 m geophone spacing
QQ MASW(geophone spacing)(source
offset)(sledgehammer size) – MASW
test performed at a particular geo-
phone spacing and source offset using
a specific sledgehammer size, e.g.
MASW1m2m2.2kg for MASW per-
formed at 1 m geophone spacing, 2 m
source offset and 2.2 kg sledgehammer.
For this study, only the fundamental mode
dispersion data was used for analysis.
Figures 2, 3 and 4 show the dispersion data
for the three surface wave tests for the
shallow bedrock site, while Figures 5, 6 and
7 show the data for the deep bedrock site.
The figures show that the frequency ranges
for the three tests were different. The
SASW test achieved the highest frequen-
cies of up to 150 Hz. The CSW test meas-
ured frequencies up to 90 Hz; however, at
Journal of the South African Institution of Civil Engineering
600
500
400
Phase velocity (m/s)
300
200
100
0
25 30 35
frequencies above approximately 70 Hz,
higher modes were excited at the shallow
bedrock site, whereas higher modes were
excited above frequencies of about 60 Hz at
the deep bedrock site. Higher modes con-
tain valuable information, but the inversion
of higher mode data requires more sophis-
ticated inversion algorithms than when
using only fundamental mode dispersion
data. As stated earlier, for this investiga-
tion, only fundamental mode dispersion
data was used. The two MASW tests meas-
ured maximum frequencies of between
35 Hz and 74 Hz at both sites. When con-
ducting seismic surface wave tests at low
frequencies, dispersion data points become
scattered as a result of low energy gener-
ated at these low frequencies. The SASW
dispersion data became scattered below
about 8 Hz and the dispersion data for the
MASW tests became scattered below about
10 Hz. As stated earlier, for this investiga-
tion, for the CSW test, the frequency range
used for the two shakers was 10 Hz to
90 Hz. Measurements at lower frequen-
cies would be possible with larger shakers
designed to be used at such low frequen-
cies. For the SASW test, the dispersion data
had very good repeatability between 8 Hz
and 80 Hz. Above about 80 Hz, differences
in the measured phase velocity can be
observed. It is worth mentioning that, for
the investigation of the repeatability of the
SASW test, the data in the frequency range
over which the hammers could produce a
good signal was considered, because the
objective of the repeatability assessment
Volume 65 Number 3 September 2023
0 5 10 15 20 25 30 35 40
Frequency (Hz)
was to investigate if the test is capable
of yielding the same results when done
repeatedly, irrespective of the frequency
range of the data. However, the data that
was considered admissible to build com-
posite dispersion data for determination of
measurement depths is the data that had a
coherence larger than 0.9 and whose plot
of wrapped-phase angle against frequency
showed a saw-tooth pattern. Observation
of a clear saw-tooth pattern, as the phase
angle changes from –π to π radians, is one
of the quality control measures used for the
SASW test (Stokoe et al 2004).
The MASW1m and MASW2 dispersion
data showed good repeatability for most
of the frequencies for which the overtone
image displayed clear fundamental mode
data, except near the upper end of the
frequency range and below about 10 Hz.
The CSW dispersion data showed predomi-
nantly fundamental mode Rayleigh wave
propagation with some higher modes at
frequencies above 60 Hz to 70 Hz. Although
some of the data points are scattered, good
repeatability is observed throughout the
entire frequency range over which the
shakers operated. This can be seen for both
the fundamental mode as well as on higher
modes. The good repeatability of CSW can
be attributed to the fact that the source
(shaker) generates a near sinusoidal wave at
a single frequency, unlike for the SASW and
MASW tests where the frequency content
generated by the sledgehammer could vary
from test to test. It should also be noted
that, due to the high level of repeatability
29
 SASW MASW1m MASW2m CSW
1 000 400 600 500
900
Phase velocity (m/s)

Phase velocity (m/s)

Phase velocity (m/s)

Phase velocity (m/s)

350 500
800 300 400
700 400
600 250 300
500 200 300
400 150 200
300 200
200 100 100
50 100
100
0 0 0 0
0 40 80 120 160 0 10 20 30 40 0 10 20 30 40 50 0 20 40 60 80 100
Frequency (Hz) Frequency (Hz) Frequency (Hz) Frequency (Hz)
Test 1 Test 2 Test 3

Figure 2 T ypical repeatability dispersion curves for SASW, MASW performed with a 2.2 kg sledgehammer and CSW at the shallow bedrock site

SASW MASW1m MASW2m CSW

1 000 400 400 500
900
Phase velocity (m/s)

Phase velocity (m/s)

Phase velocity (m/s)

Phase velocity (m/s)

350 350
800 300 300 400
700
600 250 250 300
500 200 200
400 150 150 200
300 100 100
200 100
100 50 50
0 0 0 0
0 40 80 120 160 0 10 20 30 40 50 0 20 40 60 0 20 40 60 80 100
Frequency (Hz) Frequency (Hz) Frequency (Hz) Frequency (Hz)
Test 1 Test 2 Test 3

Figure 3 T ypical repeatability dispersion curves for SASW, MASW performed with a 6.3 kg sledgehammer and CSW at the shallow bedrock site

SASW MASW1m MASW2m CSW

1 000 400 400 500
900
Phase velocity (m/s)

Phase velocity (m/s)

Phase velocity (m/s)

Phase velocity (m/s)

350 350
800 300 300 400
700
600 250 250 300
500 200 200
400 150 150 200
300 100 100
200 100
100 50 50
0 0 0 0
0 40 80 120 160 0 10 20 30 40 50 0 10 20 30 40 50 0 20 40 60 80 100
Frequency (Hz) Frequency (Hz) Frequency (Hz) Frequency (Hz)
Test 1 Test 2 Test 3

Figure 4 T ypical repeatability dispersion curves for SASW, MASW performed with a 9.2 kg sledgehammer and CSW at the shallow bedrock site

of the CSW test, the dispersion data for the compared with the other two tests which Sanchez-Salinero et al 1987; Roesset et al
three times that the test was repeated plot had smaller frequency ranges. 1990; Hiltunen & Woods 1990; Gucunski &
on top of each other, hence this makes it Woods 1992; Hunter & Crow 2015). Most
appear as if only one data set has been plot- Near-field effects comparison of these authors developed their criteria
ted (Figures 3 and 5). Close to the seismic source, cylindri- using numerical methods with only limited
All three seismic surface wave tests cally propagating Rayleigh waves occur, results from experimental work. Hiltunen
showed good repeatability of the measured and further from the source the waves and Woods (1990) used experimental data
dispersion data at both sites, and the depth become near-plane Rayleigh waves. The to suggest a maximum wavelength of 2d to
of the bedrock appeared not to influence region where the assumption of plane avoid near-field effects. In this study, tests
the repeatability of the data. When ranking Rayleigh waves is not valid is called the were performed with the source placed at
the three tests in order of repeatability near-field, and any adverse effects resulting multiple distances (offsets) from the near-
performance for the measured frequency from the invalid assumption are called est receiver, in order to investigate the sus-
range for each test, the CSW test per- near-field effects. Numerous authors have ceptibility of the SASW, MASW and CSW
formed the best, followed by the MASW suggested near-field effects criteria. The tests to near-field effects. As explained in
test and then the SASW test. However, it maximum wavelength to avoid near-field the test execution sections above, SASW
should be noted that the SASW dispersion effects range from 0.5d to 3d, where d was performed using source offsets of 1 m,
data had a larger frequency range, which is the distance between the source and 2 m, 4 m, 8 m, 16 m and 32 m. MASW1m
may cause a negative bias of the test when the nearest receiver (Heisey et al 1982; tests were performed at source offsets

30 Volume 65 Number 3 September 2023 Journal of the South African Institution of Civil Engineering
 SASW MASW1m MASW2m CSW
500 400 400 500
450 450
Phase velocity (m/s)

Phase velocity (m/s)

Phase velocity (m/s)

Phase velocity (m/s)

350 350
400 300 300 400
350 350
300 250 250 300
250 200 200 250
200 150 150 200
150 100 100 150
100 100
50 50 50 50
0 0 0 0
0 40 80 120 160 0 20 40 60 80 0 20 40 60 80 0 20 40 60 80 100
Frequency (Hz) Frequency (Hz) Frequency (Hz) Frequency (Hz)
Test 1 Test 2 Test 3

Figure 5 T ypical repeatability dispersion curves for SASW, MASW performed with a 2.2 kg sledgehammer and CSW at the deep bedrock site

SASW MASW1m MASW2m CSW

500 600 400 300
450
Phase velocity (m/s)

Phase velocity (m/s)

Phase velocity (m/s)

Phase velocity (m/s)

500 350 250
400 300
350 400 200
300 250
250 300 200 150
200 150
150 200 100
100 100
100 50 50
50
0 0 0 0
0 40 80 120 160 0 10 20 30 40 50 60 70 0 20 40 60 80 0 20 40 60 80 100
Frequency (Hz) Frequency (Hz) Frequency (Hz) Frequency (Hz)
Test 1 Test 2 Test 3

Figure 6 T ypical repeatability dispersion curves for SASW, MASW performed with a 6.3 kg sledgehammer and CSW at the deep bedrock site

SASW MASW1m MASW2m CSW

500 400 400 500
450 450
Phase velocity (m/s)

Phase velocity (m/s)

Phase velocity (m/s)

Phase velocity (m/s)

350 350
400 300 300 400
350 350
300 250 250 300
250 200 200 250
200 150 150 200
150 100 100 150
100 100
50 50 50 50
0 0 0 0
0 40 80 120 160 0 20 40 60 80 0 10 20 30 40 0 20 40 60 80 100
Frequency (Hz) Frequency (Hz) Frequency (Hz) Frequency (Hz)
Test 1 Test 2 Test 3

Figure 7 T ypical repeatability dispersion curves for SASW, MASW performed with a 9.2 kg sledgehammer and CSW at the deep bedrock site

of 1 m, 2 m, 4 m, 8 m and 16 m, whereas Table 2 Near-field effects criteria

MASW2m tests were performed at source
Maximum wavelength (λ) to avoid near-field effects (m)
offsets of 2 m, 4 m, 8 m, 16 m and 32 m.
The CSW tests were performed using off- Source Sanchez- Gucunski Hiltunen
Roeset et al Heisey
sets of 1 m, 2 m and 4 m. offset [d] (m) Salinero & Woods & Woods
(1990) et al 1982
et al (1987) (1992) (1990)
Table 2 shows maximum wavelengths (λ), λ < 2d λ < 3d
λ < 0.5d λ<d λ < 2d
in terms of minimum source offsets, at
which near-field effects will start to occur, 1 0.5 1 2 2 3
according to the criteria suggested by differ-
ent authors. It is apparent that the shortest 2 1 2 4 4 6
wavelength at which near-field effects are
suggested to occur is 0.5 m for a 1 m source 4 2 4 8 8 12

offset, as suggested by Sanchez-Salinero et

8 4 8 16 16 24
al (1987), and may be as much as 96 m for
a test performed at a source offset of 32 m,
16 8 16 32 32 48
as suggested by Heisey et al (1982). When
surface waves propagate in the near field, the 32 16 32 64 64 96
measured Rayleigh wave phase velocity is

Journal of the South African Institution of Civil Engineering Volume 65 Number 3 September 2023 31
 SASW MASW1m MASW2m CSW
500 3
400 500 500
8

24
8

6
2

2
λ = 48

λ = 48
λ = 32
450

λ = 24
Phase velocity (m/s)

Phase velocity (m/s)

Phase velocity (m/s)

Phase velocity (m/s)

6

16

12
12
4
λ=

12

8
λ=1

λ=1

6
350

λ=3
λ=

6
λ=
λ=

λ=

λ=

λ=
λ=
λ=

λ=
λ=

λ=

λ=
λ=
400 300 400 4 400 4
λ= λ=
350
300 250 6 300 3 300 3
2 λ= λ=
250 λ= 200 λ= 4
λ=
200 150 λ=3
200 λ=
2 200 λ=
2
150 λ=1 100 λ=2
100 λ = 0.5 λ=1
100 λ=1 100 λ=1
50 50 λ = 0.5
λ = 0.5 λ = 0.5
0 0 0 0
0 40 80 120 160 0 10 20 30 40 50 0 20 40 60 80 100 0 20 40 60 80 100
Frequency (Hz) Frequency (Hz) Frequency (Hz) Frequency (Hz)
2.2 kg at 1 m 2.2 kg at 2 m 2.2 kg at 4 m 2.2 kg at 8 m 2.2 kg at 16 m λ lines Source at 1 m Source at 2 m
Source at 4 m λ lines

Figure 8 S ASW, MASW and CSW dispersion curves (obtained with a 2.2 kg sledgehammer for SASW and MASW) and their near-field guidelines at the
shallow bedrock site

SASW MASW1m MASW2m CSW

500 3
400 6
500 500
16
2

λ = 48
2
λ = 16

12
λ = 32

450 450
Phase velocity (m/s)

Phase velocity (m/s)

Phase velocity (m/s)

Phase velocity (m/s)

8

2
6
8

24

12
4

16

6
12
λ=1

λ= 350
λ=3

λ=

λ=3
λ=
λ=

λ=
λ=

λ=

λ=
λ=
λ=

λ=
λ=
λ=

λ=
400 300 400 8 400
5 λ=
350 λ= 350
λ = 48

300 λ =2 250 300 6 300 4

λ= λ=
250 200 250 λ=
3
200 150 4 λ=3 200 200 λ=
2
λ=
150 100 λ=2 4 λ=3 150
λ=
100 λ=1 100 λ=2 100 λ=1
λ = 0.5 50 λ=1 λ=1
50 λ = 0.5 λ = 0.5
50 λ = 0.5
0 0 0 0
0 40 80 120 160 0 10 20 30 40 50 0 10 20 30 40 50 0 20 40 60 80 100
Frequency (Hz) Frequency (Hz) Frequency (Hz) Frequency (Hz)
6.3 kg at 1 m 6.3 kg at 2 m 6.3 kg at 4 m 6.3 kg at 8 m 6.3 kg at 16 m λ lines Source at 1 m Source at 2 m
Source at 4 m λ lines

Figure 9 S ASW, MASW and CSW dispersion curves (obtained with a 6.3 kg sledgehammer for SASW and MASW) and their near-field guidelines at the
shallow bedrock site

SASW MASW1m MASW2m CSW

1 000 6
450 500 8
500
4

12

12

900
Phase velocity (m/s)

Phase velocity (m/s)

Phase velocity (m/s)

Phase velocity (m/s)

8
16

16
24

12
24
16

400
2
2

λ = 48
12

6
λ=
λ=2

λ=
8

λ=3

λ=
λ=3

8
λ=

λ=
λ=
λ=

λ=
λ=

λ=
λ=
λ=
λ=

λ=

800 λ= 400 400

350
700 300 λ=
6
λ = 48

4 6
600 λ= λ= 300 300 4 3
250 λ= λ=
500 3 4
400 λ= 200 200
λ=
200 2
λ=3 λ=
300 λ=2 150 λ=
4 λ=3
100 λ=2 λ=2
200 100 λ=1
100 λ=1
50 λ=1 λ = 0.5
100 λ=1 λ = 0.5 λ = 0.5 λ = 0.5
0 0 0 0
0 40 80 120 160 0 10 20 30 40 50 0 10 20 30 40 50 60 0 20 40 60 80 100
Frequency (Hz) Frequency (Hz) Frequency (Hz) Frequency (Hz)
9.2 kg at 1 m 9.2 kg at 2 m 9.2 kg at 4 m 9.2 kg at 8 m 9.2 kg at 16 m λ lines Source at 1 m Source at 2 m
Source at 4 m λ lines

Figure 10 S ASW, MASW and CSW dispersion curves (obtained with a 9.2 kg sledgehammer for SASW and MASW) and their near-field guidelines at
the shallow bedrock site

slower than in the far field. Near-field effect The λ values for diagonal lines range the 6.3 and 9.2 kg sledgehammers it can be
can therefore be identified when the phase from 0.5 m to the maximum λ correspond- seen that the phase velocity was underesti-
velocity is slower than the phase velocity for ing to the maximum source offset used for mated when the tests were performed at a
the same geophone positions, but the source that test, as shown in Table 2 (e.g. λ ranges 1 m source offset, and this effect appears to
placed at a larger offset. For the analysis of from 0.5 m to 24 m for SASW performed be significant starting from wavelengths of
the dispersion data from this study, the data up to a source offset of 8 m, and λ ranges at least 3 m, as can be seen in Figures 9 and
is displayed in Figures 8 to 10 for the shallow from 0.5 m to 48 m for MASW performed 10, respectively. Therefore, near-field effects
bedrock site, and in Figures 11 to 13 for the up to a source offset of 16 m). started to occur when the wavelength was
deep bedrock site. Lines of constant wave- SASW dispersion data obtained with a longer than 3d, where d is the source offset.
length (i.e. λ = 0.5, λ = 1, etc) are shown to 2.2 kg sledgehammer, as shown in Figure 8, MASW1m dispersion data for the
assist with the identification of wavelengths shows no clear signs of near-field effects. source placed at 1 m, 2 m and 4 m is scat-
at which near-field effects start to occur. However, for SASW tests performed with tered and does not show clear fundamental

32 Volume 65 Number 3 September 2023 Journal of the South African Institution of Civil Engineering
 SASW MASW1m MASW2m CSW
500 3
400 500 500

λ = 48

λ = 24
λ = 32
450

λ = 32
Phase velocity (m/s)

Phase velocity (m/s)

Phase velocity (m/s)

λ = 48

Phase velocity (m/s)

12
12
6
λ=

6
2

8
8
12

6
350

6
4

6
6

λ=1
λ=1

λ=1

λ=
λ=

λ=
λ=

λ=
λ=

λ=

λ=
λ=

λ=
λ=
λ=

λ=
400 300 400 λ=
4 400
350 3 λ=
4
300 λ =2 250 λ=
300 3 300 3
λ= λ=
250 200 2
λ=
200 150 200 λ =2 200 λ=
2
150 100 λ=1 λ=1 λ=1
100 100 100
50 λ=1 λ = 0.5 50 λ = 0.5 λ = 0.5 λ = 0.5
0 0 0 0
0 40 80 120 160 0 10 20 30 40 50 60 70 80 90 0 20 40 60 80 100 0 20 40 60 80 100
Frequency (Hz) Frequency (Hz) Frequency (Hz) Frequency (Hz)
2.2 kg at 1 m 2.2 kg at 2 m 2.2 kg at 4 m λ lines Source at 1 m Source at 2 m
2.2 kg at 8 m 2.2 kg at 16 m 2.2 kg at 32 m Source at 4 m λ lines

Figure 11 S ASW, MASW and CSW dispersion curves (obtained with a 2.2 kg sledgehammer for SASW and MASW) and their near-field guidelines at
the deep bedrock site

SASW MASW1m MASW2m CSW

350 500 500 500
λ = 32
λ = 16
λ = 24

λ = 32

λ = 48
λ = 24
Phase velocity (m/s)

Phase velocity (m/s)

Phase velocity (m/s)

Phase velocity (m/s)

6

λ = 32
2

6
4

12
8

8
8
6

2
12

12
8
16
4
λ=1

λ=1

6
λ=2

λ=

6
300
λ=

λ=

λ=
λ=
λ=

λ=
λ=

λ=
λ=
λ=
λ=

λ=
λ=

400 400 4 400

250 λ=
λ = 48

300 =4 300 3 300 3

200 λ
λ= 4 λ=
3 λ=
150 λ=
200 200 λ=
2 200 λ =2
100
=1 λ = 0. 5 100 λ=2 100 λ=1 100 λ=1
λ λ=1
50 λ = 0.5 λ = 0.5 λ = 0.5
0 0 0 0
0 40 80 120 160 0 20 40 60 80 0 20 40 60 80 100 0 20 40 60 80 100
Frequency (Hz) Frequency (Hz) Frequency (Hz) Frequency (Hz)
6.3 kg at 1 m 6.3 kg at 2 m 6.3 kg at 4 m λ lines Source at 1 m Source at 2 m
6.3 kg at 8 m 6.3 kg at 16 m 6.3 kg at 32 m Source at 4 m λ lines

Figure 12 S ASW, MASW and CSW dispersion curves (obtained with a 6.3 kg sledgehammer for SASW and MASW) and their near-field guidelines at
the deep bedrock site

SASW MASW1m MASW2m CSW

500 500 500 500
λ = 48
λ = 32
λ = 24

λ = 48
λ = 32

450 450
Phase velocity (m/s)

Phase velocity (m/s)

Phase velocity (m/s)

Phase velocity (m/s)

λ = 16

12
16
8

8
12
6

16
λ=1

λ=2

12

6
8

λ=
8

λ=2

λ=
λ=

λ=
λ=
λ=

λ=
λ=
λ=

λ=
λ=
λ=

λ=
λ=

400 400 400 λ=

4 400
350 350 λ=
4
300 λ =2 300 λ =4 300 3 300 3
λ= λ=
250 λ=
3 250
200 200 200 λ=
2 200 λ=
2
150 150
100 100 λ=2 100 λ=1 100 λ=1
λ = 0.5 λ=1
λ=1 λ = 0.5 λ = 0.5
50 λ = 0.5 50
0 0 0 0
0 40 80 120 160 0 20 40 60 80 0 20 40 60 80 100 0 20 40 60 80 100
Frequency (Hz) Frequency (Hz) Frequency (Hz) Frequency (Hz)
9.2 kg at 1 m 9.2 kg at 2 m 9.2 kg at 4 m λ lines Source at 1 m Source at 2 m
9.2 kg at 8 m 9.2 kg at 16 m 9.2 kg at 32 m Source at 4 m λ lines

Figure 13 S ASW, MASW and CSW dispersion curves (obtained with a 9.2 kg sledgehammer for SASW and MASW) and their near-field guidelines at
the deep bedrock site

mode trends. For the 6.3 kg and 9.2 kg phase appears to occur for a wavelength occur at wavelengths longer than 2d, where
sledgehammers there is a clear underesti- (λ) of 16 m when comparing the results for d is the source offset.
mation of the phase velocity at wavelengths source offsets of 8 m and 16 m, indicating Figures 11 to 13 compare the dispersion
(λ) of 8 m and 12 m, respectively, when that near-field effects occurred if the wave- data for the tests at the deep bedrock site.
data from the 8 m and 16 m source offsets length was longer than the source offset. Near-field effects appeared to be more
is compared. Near-field effect therefore For the CSW test, phase velocities pronounced at this site than at the shallow
occurred for wavelengths longer than the measured at frequencies below 20 Hz were bedrock site. For the SASW tests per-
source offset. For the MASW2m disper- lower for source offsets of 1 m and 2 m formed with the three hammer sizes, some
sion data, near-field effects were less clear, than for a source offset of 4 m. At 20 Hz higher mode excitation clearly occurred, so
but when again comparing the results a wavelength of 8 m was generated at the it is not possible to compare fundamental
for the 6.3 kg and 9.2 kg sledgehammers, shallow bedrock site. The results therefore mode data at all frequencies. However, it
an underestimation of the Rayleigh wave indicate that near-field effects started to appears that the data for the 1 m source

Journal of the South African Institution of Civil Engineering Volume 65 Number 3 September 2023 33
offset underestimated the measured phase Table 3 M
 inimum and maximum frequencies, wavelengths and measurement depths at the
velocities compared with the other source shallow bedrock site
offsets at wavelengths longer than 1 m, Test fmin (Hz) fmax (Hz) λmin (m) λmax (m) dmin (m) dmax (m)
indicating that near-field effects start to
SASW 9.0 150.0 1.1 57.6 0.6 28.8
occur when the source is less than one
wavelength from the first receiver. For the MASW1m8m2.2kg 10.0 40.0 4.8 19.2 2.4 9.6
MASW1m tests, underestimation of phase MASW1m16m2.2kg 10.0 44.9 4.8 25.0 2.4 12.5
velocities measured at a source offset of
MASW1m1m6.3kg 12.0 20.0 11.2 18.2 5.6 9.1
4 m, occurred at wavelength of between
3 m and 4 m, when compared to the larger MASW1m2m6.3kg 7.5 24.7 11.0 30.6 5.5 15.3
offsets, suggesting that a minimum source MASW1m4m6.3kg 7.9 27.0 7.4 27.4 3.7 13.7
offset of one wavelength is required to
MASW1m8m6.3kg 7.9 43.0 5.4 22.2 2.7 11.1
avoid near-field effects. For the MASW2m,
underestimation of the phase velocities MASW1m16m6.3kg 11.0 45.0 4.0 21.4 2.0 10.7
started to occur at wavelengths larger MASW1m2m9.2kg 7.5 24.0 11.0 26.4 5.5 13.2
than 8 m for offsets less than 8 m, also
MASW1m4m9.2kg 6.7 20.0 11.6 28.0 5.8 14.0
indicating that near-field effects occur for a
source offset less than one wavelength. For MASW1m8m9.2kg 8.9 42.8 5.4 19.8 2.7 9.9
the CSW test, when comparing the data MASW1m16m9.2kg 10.0 45.0 4.0 25.0 2.0 12.5
measured at a source offset of 4 m with the
MASW2m4m2.2kg 9.1 33.3 4.4 27.2 2.2 13.6
data for the 1 m and 2 m offsets, underes-
timation of the phase velocity started to MASW2m8m2.2kg 10.0 54.9 3.6 26.4 1.8 13.2
occur at a wavelength of between 3 m and MASW2m16m2.2kg 10.0 40.0 3.4 28.4 1.7 14.2
4 m. This again suggests that the near-field
MASW2m2m6.3kg 10.6 21.9 9.4 21.8 4.7 10.9
effect becomes important when the source
offset is less than about one wavelength. MASW2m4m6.3kg 8.6 30.0 5.0 27.8 2.5 13.9

MASW2m8m6.3kg 8.5 47.2 3.8 31.4 1.9 15.7

Measurement depth comparison MASW2m16m6.3kg 9.1 40.0 4.8 39.8 2.4 19.9
This section focuses on the maximum
MASW2m2m9.2kg 11.3 21.2 9.4 22.0 4.7 11.0
and minimum measurement depths of the
three tests at the two sites. MASW2m4m9.2kg 8.1 30.3 5.2 30.4 2.6 15.2
When comparing different seismic MASW2m8m9.2kg 8.0 45.0 3.7 33.4 1.9 16.7
surface wave tests with regard to the
MASW2m16m9.2kg 7.9 40.0 4.8 50.0 2.4 25.0
maximum and minimum depth measured,
the key parameters are the frequencies CSW1m 10.0 68.0 2.4 26.0 1.2 13.0
and wavelengths of the generated Rayleigh CSW2m 10.0 74.0 2.2 27.0 1.1 13.5
waves. High frequencies generate short
CSW4m 10.0 89.0 2.4 28.2 1.2 14.1
wavelengths which are required to mea-
sure the shallow material with adequate
resolution. In contrast, to determine the required. However, low frequency measure- The maximum and minimum depths
behaviour of deep geomaterials, long ments require a powerful shaker, as a large measured are shown in Figures 14 and
wavelengths are required which occur at volume of ground has to be excited. Hence, 15 for the shallow bedrock site, and in
low frequencies. When designing an active often the limitation of the shaker is not its Figures 16 and 17 for the deep bedrock
seismic surface wave test, some control is lowest frequency, but the energy that can site. Measurement depths were determined
possible over the desired source frequency be generated at low frequency. using Equation 1 which has been used by
content. For instance, for a CSW test, high For the SASW and MASW test, when many researchers (Olafsdottir et al 2018;
frequencies or low frequencies can be tar- using a sledgehammer, the operator has Hunter & Crow 2015; Park et al 2000; Yuan
geted by appropriate design of the shaker. little control over the frequency content and 2011).
However, there are some complexities that energy level of the source other than the
have to be considered. The CSW test is type and size of hammer and anvil used. λ
z=  (1)
prone to generating higher modes at high Common practice is to use steel-headed 2
frequencies. This can clearly be seen in the hammers of various sizes striking a steel
last graphs of Figures 2 to 13. plate. These factors make it difficult to com- Where: z is the measured depth and λ is
Inversion analysis with dispersion data pare the maximum and minimum depths the measured wavelength.
that contains higher modes is done rou- that can be achieved by each of the three
tinely, but requires an appropriate inversion tests in a completely unbiased way. However, As explained earlier, only fundamental
algorithm and also judgement from the for this investigation the aim was to com- mode dispersion data was used to determine
analyst to correctly identify the modes. For pare the maximum and minimum depth the measurement depths. The points in the
low-frequency deep measurements, a shak- measurements that can be achieved using low frequency range where the dispersion
er capable of generating low frequencies is equipment that is often used in industry. curves started to exhibit scatter were used

34 Volume 65 Number 3 September 2023 Journal of the South African Institution of Civil Engineering
 7

5
Minimum depth (m)

0
SASW

MASW1m8m2.2kg

MASW1m16m2.2kg

MASW1m1m6.3kg

MASW1m2m6.3kg

MASW1m4m6.3kg

MASW1m8m6.3kg

MASW1m16m6.3kg

MASW1m2m9.2kg

MASW1m4m9.2kg

MASW1m8m9.2kg

MASW1m16m9.2kg

MASW2m4m2.2kg

MASW2m8m2.2kg

MASW2m16m2.2kg

MASW2m2m6.3kg

MASW2m4m6.3kg

MASW2m8m6.3kg

MASW2m16m6.3kg

MASW2m2m9.2kg

MASW2m4m9.2kg

MASW2m8m9.2kg

MASW2m16m9.2kg

CSW1m

CSW2m

CSW4m
Test

Figure 14 M
 inimum measurement depths at the shallow bedrock site

35

30

25
Maximum depth (m)

20

15

10

0
SASW

MASW1m8m2.2kg

MASW1m16m2.2kg

MASW1m1m6.3kg

MASW1m2m6.3kg

MASW1m4m6.3kg

MASW1m8m6.3kg

MASW1m16m6.3kg

MASW1m2m9.2kg

MASW1m4m9.2kg

MASW1m8m9.2kg

MASW1m16m9.2kg

MASW2m4m2.2kg

MASW2m8m2.2kg

MASW2m16m2.2kg

MASW2m2m6.3kg

MASW2m4m6.3kg

MASW2m8m6.3kg

MASW2m16m6.3kg

MASW2m2m9.2kg

MASW2m4m9.2kg

MASW2m8m9.2kg

MASW2m16m9.2kg

CSW1m

CSW2m

CSW4m

Test

Figure 15 M
 aximum measurement depths at the shallow bedrock site

to evaluate the maximum measurement explained in the SASW test execution sec- From Figure 14 it is clear that SASW
depths, whereas the points at the high tion, common midpoint two receiver SASW was able to measure shallower than both
frequency end of the fundamental mode tests were conducted for this study. The the MASW and CSW, and the CSW test
data were used to determine the minimum composite dispersion data was determined measured shallower than the MASW test.
measurement depths. The maximum and for six sets of receiver spacings and three From Figure 15 it is also clear that SASW
minimum frequencies, wavelengths and impacts of each hammer. The source and measured deeper than both MASW and
measurement depths for the three tests receiver positions did not change for each CSW. The MASW and CSW tests mea-
at the shallow and deep bedrock site are common midpoint stage of the test, and sured to similar depth, except for very large
shown in Tables 3 and 4 respectively. Depth hence the composite test is regarded as one MASW source offsets, where the MASW
comparisons are shown in Figures 14 and 15 test and not multiple tests. Therefore, in test measured deeper.
for the shallow bedrock site, and in Figures Figures 14 to 17 and in Tables 3 and 4, only Table 4 shows the maximum and
16 and 17 for the deep bedrock site. As one SASW test result is shown. minimum frequencies, wavelengths and

Journal of the South African Institution of Civil Engineering Volume 65 Number 3 September 2023 35
Table 4 Maximum and minimum frequencies, wavelengths and measurement depths at the deep Summary
bedrock site Table 5 gives a summary of the rank-
Test fmin (Hz) fmax (Hz) λmin (m) λmax (m) dmin (m) dmax (m) ing of SASW, MASW (MASW1m and
MASW2m) and CSW in terms of repeat-
SASW 8.0 76.0 1.6 49.4 0.8 24.7
ability, resistance to near-field effects,
MASW1m2m2.2kg 10.9 46.6 6.6 33.0 3.3 16.5 as well as the minimum and maximum
MASW1m4m2.2kg 8.1 75.0 1.8 30.6 0.9 15.3
measurement depths. Near-field effects
were less pronounced at the shallow bed-
MASW1m8m2.2kg 8.1 70.7 2.2 30.2 1.1 15.1
rock site than at the deep bedrock site. All
MASW1m16m2.2kg 8.5 43.1 3.2 27.0 1.6 13.5 three tests performed similarly with regard
MASW1m1m6.3kg 18.0 68.7 2.1 7.8 1.1 3.9
to near-field effects at the deep bedrock site
and are given an equal ranking.
MASW1m2m6.3kg 8.0 60.1 2.3 37.0 1.1 18.5

MASW1m4m6.3kg 8.0 70.3 1.9 31.1 1.0 15.5

MASW1m8m6.3kg 7.5 65.1 2.2 32.5 1.1 16.3

CONCLUSIONS
This investigation compared the perfor-
MASW1m16m6.3kg 8.5 43.3 3.3 40.1 1.6 20.1 mance of three popular seismic surface
MASW1m1m9.2kg 10.0 65.5 2.0 11.4 1.0 5.7 wave tests in terms of repeatability of the
test data, susceptibility to near-field effects,
MASW1m2m9.2kg 8.1 60.6 2.0 35.8 1.0 17.9
as well as the maximum and minimum
MASW1m4m9.2kg 7.3 68.7 1.9 34.6 1.0 17.3 depths that were measured.
MASW1m8m9.2kg 7.1 64.3 2.2 35.2 1.1 17.6 All three surface wave tests performed
well with regard to repeatability of the
MASW1m16m9.2kg 7.8 43.8 3.4 39.0 1.7 19.5
data. The CSW test data was the most
MASW2m2m2.2kg 13.7 60.6 1.5 10.6 0.7 5.3 repeatable. This may be attributed to the
MASW2m4m2.2kg 8.5 85.0 1.4 25.2 0.7 12.6 well-defined characteristics of the source,
which was a mechanical shaker generating
MASW2m8m2.2kg 8.0 70.3 1.8 29.4 0.9 14.7
a monotonic Rayleigh wave. For the SASW
MASW2m16m2.2kg 7.9 34.4 4.4 36.8 2.2 18.4 and MASW tests, sledgehammers were
MASW2m32m2.2kg 9.5 35.1 4.2 23.6 2.1 11.8 used which do not generate an impact with
exactly the same frequency content each
MASW2m2m6.3kg 12.0 74.0 1.6 11.0 0.8 5.5
time. Nonetheless, both tests exhibited
MASW2m4m6.3kg 8.1 80.0 1.4 30.8 0.7 15.4 good repeatability, with the MASW test
MASW2m8m6.3kg 7.3 70.0 1.8 36.4 0.9 18.2 slightly outperforming the SASW test.
However, this has to be seen in the context
MASW2m16m6.3kg 7.3 45.5 3.0 42.4 1.5 21.2
of the frequency range measured by each
MASW2m32m6.3kg 8.1 35.2 4.2 24.4 2.1 12.2 test. The frequency range measured by the
MASW2m2m9.2kg 12.0 75.0 1.5 12.4 0.8 6.2 SASW test was significantly larger than
that measured by the MASW test for the
MASW2m4m9.2kg 7.3 80.0 1.4 35.8 0.7 17.9
two chosen MASW test geometries. This
MASW2m8m9.2kg 7.3 70.3 1.8 37.6 0.9 18.8 may cause a bias against the SASW test
MASW2m16m9.2kg 6.7 45.5 3.0 47.2 1.5 23.6 when evaluating the repeatability of the
data, because at frequencies below about
MASW2m32m9.2kg 8.1 35.2 4.2 23.8 2.1 11.9
80 Hz the SASW test results were more
CSW1m 8.0 58.0 2.2 20.4 1.1 10.2 repeatable than at frequencies above 80Hz.
CSW2m 8.0 53.0 2.4 21.6 1.2 10.8 With regard to near-field effects
observed at the shallow bedrock site, the
CSW4m 8.0 57.0 2.4 24.6 1.2 12.3
SASW test performed best with near-field
effects being observed at wavelengths
measurement depths at the deep bedrock test geometry has to be chosen carefully longer than 3d, where d is the source offset
site. The results of maximum and minimum when targeting shallow measurements. It is (distance between the source and the
measurement depth are plotted as bar noteworthy that SASW measured slightly first receiver). For the CSW test near-field
charts in Figures 16 and 17 respectively. For shallower than CSW regardless of the effects were observed for wavelengths
MASW2m, performed at source offsets of source offset used for CSW. With regard longer than 2d and for the MASW tests
2 m and 4 m, the minimum depths were to maximum measured depth, SASW was at wavelengths longer than 1d. At the
less than those obtained with the SASW able to measure deeper than MASW and deep bedrock site near-field effects were
and CSW tests. However, for offsets beyond CSW. The CSW measured deeper than the observed for all three tests at wavelengths
these, the minimum depths measured by MASW with small offsets of 1 m and 2 m, longer than 1d.
MASW2m were deeper than those of SASW but at larger offsets the MASW test mea- As noted in the discussion, the maxi-
and CSW. Clearly, for the MASW test the sured deeper than the CSW test. mum and minimum depths that can be

36 Volume 65 Number 3 September 2023 Journal of the South African Institution of Civil Engineering
 Maximum depth (m) Minimum depth (m)

Note:
0
5
10
15
20
25
30
0
0.5
1.0
1.5
2.0
2.5
3.0
3.5

Figure 17 M

Repeatability
SASW SASW

– : Inconclusive
1: Performed best
MASW1m2m2.2kg MASW1m2m2.2kg
MASW1m4m2.2kg MASW1m4m2.2kg

parameter
MASW1m8m2.2kg MASW1m8m2.2kg

2: Performed second best, etc

Resistance to near-field effects
Performance measuring
MASW1m16m2.2kg MASW1m16m2.2kg

Minimum measurement depth (m)

Maximum measurement depth (m)

MASW1m1m6.3kg MASW1m1m6.3kg
MASW1m2m6.3kg MASW1m2m6.3kg

Same number: Indicates similar performance

MASW1m4m6.3kg MASW1m4m6.3kg

1
1
1
3

Journal of the South African Institution of Civil Engineering

SASW
MASW1m8m6.3kg MASW1m8m6.3kg
MASW1m16m6.3kg MASW1m16m6.3kg
MASW1m1m9.2kg MASW1m1m9.2kg

Table 5 Performance ranking of SASW, MASW and CSW

2
3
3
2
MASW1m2m9.2kg MASW1m2m9.2kg

MASW1m
Figure 16 Minimum measured depths at the deep bedrock site
MASW1m4m9.2kg MASW1m4m9.2kg

 aximum measured depths at the deep bedrock site

MASW1m8m9.2kg MASW1m8m9.2kg
MASW1m16m9.2kg

2
3
3
2
MASW1m16m9.2kg

Volume 65 Number 3 September 2023

Shallow bedrock site
MASW2m2m2.2kg MASW2m2m2.2kg

MASW2m
MASW2m4m2.2kg MASW2m4m2.2kg

Test
Test

MASW2m8m2.2kg MASW2m8m2.2kg

2
2
2
1
CSW
MASW2m16m2.2kg MASW2m16m2.2kg
MASW2m32m2.2kg MASW2m32m2.2kg
MASW2m2m6.3kg MASW2m2m6.3kg
MASW2m4m6.3kg

1
1
1
3
MASW2m4m6.3kg

SASW
MASW2m8m6.3kg MASW2m8m6.3kg
MASW2m16m6.3kg MASW2m16m6.3kg
MASW2m32m6.3kg MASW2m32m6.3kg

2
1
2

–
MASW2m2m9.2kg MASW2m2m9.2kg

MASW1m
MASW2m4m9.2kg MASW2m4m9.2kg
MASW2m8m9.2kg MASW2m8m9.2kg

1
1
2

–
MASW2m16m9.2kg MASW2m16m9.2kg

Deep bedrock site

MASW2m32m9.2kg MASW2m32m9.2kg

MASW2m
CSW1m CSW1m
CSW2m CSW2m

2
1
1

–
CSW
CSW4m CSW4m

37
measured by each test are dependent on
the choice of equipment, making it difficult
to compare the tests in a completely unbi-
ased way. However, for this investigation
the aim was to compare the maximum
and minimum measurement depths that
can be achieved using equipment that is
often used in industry. The SASW test was
able to measure shallower and deeper than
both the MASW and CSW tests at both
sites. The reason for this is the fact that
the SASW test is conducted with multiple
test geometries. Measurements are taken at
small geophone spacings and source offset
to capture the short Rayleigh waves, and
the spacings are progressively increased
to also capture long Rayleigh waves. This
benefit comes at the time cost of having
to change the setup several times during
the test. At the shallow bedrock site, the
CSW test was able to measure to shallower
depths than the MASW test; however, at
the deep bedrock site the MASW mea-
sured to shallower depths than the CSW
test when using short source offsets of less
than 4 m. When using source offsets of
more than 4 m, the CSW test measured to
shallower depths than the MASW test.
At both the shallow and deep bedrock
sites, the MASW test measured to greater
depths than the CSW test when long
source offsets were used for the MASW
test. When short MASW source offsets
were used, the CSW test was able to mea-
sure to greater depths than the MASW
test. This indicates the importance of
selecting the appropriate test geometry,
according to the measurement depth of
interest, when conducting the MASW test.
This investigation attempted to com-
pare the performance of the three popular
seismic surface wave tests (SASW, MASW
and CSW) when using equipment that is
often used in practice. The tests were only
compared with regard to repeatability of
the test results, susceptibility to near-field
effects and depth of measurement. An
aspect that was not considered was produc-
tion rate of the tests, because this is depen-
dent on the number of personnel on site,
their experience, and ease of site access.
The SASW and MASW test require sig-
nificant space to conduct the tests, whereas
the CSW test uses only a small number of
geophones that do not have to be moved
38
during the test and therefore require much
less space. This is clearly an advantage on
congested sites. Also, the CSW test allows
for easy automation of the test, by stepping
or sweeping through the frequencies of
interest, significantly reducing the time
required to conduct the CSW test.
REFERENCES
Bignardi, S 2011. Complete waveform inversion
approach to seismic surface waves and adjoint
active surfaces. Doctoral Thesis. Ferrara, Italy:
Universitàdegli Studi di Ferrara.
Bouazza, A & Kavazanjian, E 2000. Characterization
of municipal solid waste sites using the continuous
surface wave method. Proceedings, International
Society for Rock Mechanics and Rock Engineering
(ISRM) International Symposium, November,
Melbourne, Australia.
Foti, S, Hollender, F, Garofalo, F, Albarello, D,
Asten, M, Bard, P Y et al 2018. Guidelines for the
good practice of surface wave analysis: A product
of the InterPACIFIC project. Bulletin of Earthquake
Engineering, 16(6): 2367–2420.
Groenewold, F 2016. Comparison of SASW systems
for coastal and offshore applications. Master’s
Dissertation. South Kingston, RI: University of
Rhode Island.
Gucunski, N & Woods, RD 1992. Numerical simulation
of the SASW test. Soil Dynamics and Earthquake
Engineering, 11(4): 213–227.
Hebeler, G L & Rix, G J 2001. Site characterization in
Shelby County, Tennessee, using advanced surface
wave methods. Urbana, IL: MAE Center CD Release
06-02.
Heisey, J S, Stokoe, K H, Hudson, W R & Meyer, A H
1982. Determination of in situ shear wave velocities
from spectral analysis of surface waves. Master’s
Dissertation. Austin, TX: University of Texas at
Austin.
Heymann, G 2007. Ground stiffness measurement
by the continuous surface wave test. Journal of the
South African Institution of Civil Engineering, 49(1):
25–31.
Hiltunen, D R & Woods, R D 1990. Variables affecting
the testing of pavements by the surface wave
method. Transportation Research Record, 1260:
42–52.
Hunter, J A M & Crow, H L 2015. Shear wave velocity
measurement guidelines for Canadian seismic site
characterization in soil and rock. Ottawa: Natural
Resources Canada, Geological Survey of Canada.
Kim, D S, Park, H J & Bang, E S 2013. Round Robin
test for comparative study of in situ seismic tests.
Volume 65 Number 3 September 2023 Journal of the South African Institution of Civil Engineering
Geotechnical and geophysical site characterization:
Proceedings, 4th International Conference on Site
Characterization (ISC-4), Vol 1. Abingdon, UK:
Taylor & Francis, 1427–1434.
Matthews, M C, Hope, V S & Clayton, C R I 1996. The
use of surface waves in the determination of ground
stiffness profiles. International Journal of Rock
Mechanics and Mining Sciences and Geomechanics
Abstracts, 8(33): 360A–361A.
McCaskill, A 2014. A study on the benefits of including
near-field effects in active-source surface wave data
collection and interpretation. Doctoral Thesis.
Columbia, MI: University of Missouri–Columbia.
Ntaote, M 2022. Experimental comparison of active
seismic surface wave tests on shallow and deep
bedrock sites. MEng Dissertation. University of
Pretoria.
Olafsdottir, E A, Erlingsson, S & Bessason, B 2018.
Tool for analysis of multichannel analysis of
surface waves (MASW) field data and evaluation
of shear wave velocity profiles of soils. Canadian
Geotechnical Journal, 55(2): 217–233.
Omar, M N, Abbiss, C P, Taha, M R & Nayan, K A M
2011. Prediction of long-term settlement on
soft clay using shear wave velocity and damping
characteristics. Engineering Geology, 123(4): 259–270.
Park, C B, Miller, R D, Xia, J & Ivanov, J 2000.
Multichannel seismic surface-wave methods
for geotechnical applications. Proceedings, First
International Conference on the Application of
Geophysical Methodologies to Transportation
Facilities and Infrastructure, 11–15 December 2000,
St. Louis, MO, pp 11–15.
Roesset, J M, Chang, W, Stokoe, K H & Aouad, M 1990.
Modulus and thickness of the pavement surface
layer from SASW tests. Transportation Research
Record, 1260, 53–63.
Sanchez-Salinero, I, Roesset, J M, Shao, K Y, Stokoe, K H
& Rix, G J 1987. Analytical evaluation of variables
affecting surface wave testing of pavements.
Transportation Research Record, 1136.
Stokoe, K H, Joh, S H & Woods, R D 2004. Some
contributions of in situ geophysical measurements
to solving geotechnical engineering problems.
Proceedings, International Conference on Site
Characterization (ISC-2), 19–22 September 2004,
Porto, Portugal, pp 97–132.
Strobbia, C 2003. Surface wave methods: Acquisition,
processing and inversion. PhD Thesis. Turin, Italy:
Politecnico di Torino.
Yuan, J 2011. Field studies comparing SASW,
beamforming and MASW test methods and
characterization of geotechnical materials based on
Vs. Doctoral Thesis. Austin, TX: University of Texas
at Austin.
 TECHNICAL PAPER
Thermal performance
of cavities in 3DPC
JOURNAL OF THE SOUTH AFRICAN
INSTITUTION OF CIVIL ENGINEERING

building façades
ISSN 1021-2019 (print) | ISSN 2309-8775 (online)
Vol 65 No 3, September 2023, Pages 39–53, Paper 1588

MIGAEL MOELICH (MEng) is a Candidate

M Moelich, G van Zijl, W de Villiers Structural Design Engineer at AECOM in
Bellville, where he has worked for the past
year. He graduated with a Bachelor and a
Master’s in Civil Engineering from
In the quest for sufficient and comfortable housing in South Africa, attention to building Stellenbosch University in 2020 and 2022
respectively.
orientation and thermal transmittance principles has significantly improved thermal
performance in local houses. Thermal simulation and monitoring of houses in the Joe Slovo Contact details:
Phase Three development in Cape Town confirms improved performance by solar radiation Department of Civil Engineering AECOM
blockage and shading. This paper simulates the thermal performance of a housing unit in this Stellenbosch University PO Box 112
Private Bag X1 Bellville 7535
development. The house was instrumented with thermocouples, and monitored from March Matieland 7600 South Africa
to December 2013 by Sustainable Energy Africa in collaboration with the National Department South Africa
of Human Settlements. Sustainable Energy Africa reported significant improvement when E: migael.moelich@aecom.com
compared to a nearby traditional Reconstruction and Development Programme house that
had been monitored during the same period. With the benefit of the monitored thermal data, PROF GIDEON VAN ZIJL (DEng, Pr Eng,
FSAICE) is Distinguished Professor of
complete drawings, and specification details of the Joe Slovo Phase Three unit, the current Structural Engineering at Stellenbosch
research modelled and calibrated a simulation model in DesignBuilder. DesignBuilder was University (SU). After graduating with a
selected given its accreditation status for Green Star Rating of buildings in South Africa. An Bachelor and a Master’s in Civil Engineering
at SU, and a PhD in Civil Engineering at Delft
additional motivation for investigating this unit was the availability of recorded weather data
University of Technology, The Netherlands,
from the nearby Cape Town International Airport, captured in the DesignBuilder climate data he joined SU in 2001 as professor of
base for 2013, which was assumed relevant in lieu of complete weather station data measured structural engineering. He established the Centre for Development of
at the location of the housing unit. The study aimed to simulate the monitored temperatures Sustainable Infrastructure (CDSI) in 2002, with research interests in
structural, computational and durability mechanics, currently applied in
in the house with acceptable agreement, and to investigate further potential improvement in
3D concrete printing.
occupant thermal comfort by alternative 3D-printed concrete walling developed by the authors.
Contact details:
Systematic finite element analysis (FEA) iteratively solved cavity radiation and convection
Department of Civil Engineering
in wall cavities to calibrate thermal transmittance parameters for DesignBuilder. The FEA Stellenbosch University
results acceptably simulated temperatures monitored in this physical wall experiment. The Private Bag X1
calibrated DesignBuilder model simulated the indoor temperatures of the monitored house with Matieland 7600
South Africa
acceptable agreement, and predicted significant improvement in occupant thermal comfort if E: gvanzijl@sun.ac.za
the walls were 3D-printed with a particular cross-sectional design.
DR WIBKE DE VILLIERS is a senior lecturer at
Stellenbosch University, where she obtained
INTRODUCTION inexpensive to build and uncomfortable to her BEng (2006), MScEng (2008) and PhD
(2019), the latter in the development of
According to the Bill of Rights in the live in, is not sustainable. performance-based specifications for
Constitution of the Republic of South The National Department of Human alternative masonry units. As a member of
Africa (RSA 1996), “Everyone has the Settlements (NDHS) in South Africa the Unit of Construction Materials, her
research interests lie in alternative and
right to have access to adequate housing”. launched a new approach for the sus-
sustainable construction materials and their structural and thermal
To this end the government introduced tainable delivery of housing under the performance. She has supervised a number of postgraduate students and
the Reconstruction and Development Integrated Reconstruction and Development authored or co-authored several publications over the past nine years.
Programme (RDP). With an increase Programme (IRDP) initiative. The third Prior to this, she spent two years at Aurecon (now Zutari) as a structural
design engineer.
in population, housing is becoming a phase of the Joe Slovo project was the flag-
significant problem. The total number ship of this initiative for more sustainable Contact details:
Department of Civil Engineering
of households increased from 9 million housing, with 2 886 subsidised homes on 27
Stellenbosch University
in 1996 to 16.9 million in 2016, of which hectares (NDHS 2013). Despite the improve- Private Bag X1
79.2% lived in formal dwellings, 7% in ment in living conditions, illustrated in Matieland 7600
traditional dwellings and 14% in informal Figure 1, the project also focused on sustain- South Africa
E: wdv@sun.ac.za
dwellings, as classified by Statistics South able energy interventions, such as solar water
Africa (2016). It is clear that there is a heaters and energy-efficient lighting. The
housing shortage in the country. However, most significant change, however, was the
developing low-density buildings that are optimised thermal design, which included:

Moelich M, Van Zijl G, De Villiers W. Thermal performance of cavities in 3DPC building façades. Keywords: 3D-printed concrete, housing, thermal simulation,
J. S. Afr. Inst. Civ. Eng. 2023:65(3), Art. #1588, 15 pages. http://dx.doi.org/10.17159/2309-8775/2023/v65n3a4
thermal monitoring
The Journal of the South African Institution of Civil Engineering, which is distributed internationally, is a peer-reviewed, open-access journal licensed under a Creative
Commons Attribution Licence (CC BY-NC-ND). Readers may therefore freely use and share the content as long as they credit the original creator and publisher, do
not change the material in any way, and do not use it commercially. Copyright of this article remains with the authors. 39
QQ A double-storey design that improves
population density and allows for
dual living spaces, with the upstairs
space being slightly warmer than the
downstairs.
QQ Roof overhangs that provide partial
shade on windows during summer
while allowing solar gain during winter.
QQ Inside living spaces that are oriented to
be north-facing as per the SANS 10400-
XA:2021 recommendations (SANS
2021). This orientation allows other
rooms to screen the unwanted western
sun and prevent heat loss on the south-
facing façades.
QQ Ceiling insulation and external façade
plaster and paint to improve the ther-
mal comfort for occupants by reducing
the global thermal transmittance of the
building.
The NDHS (2013) confirmed improved
indoor thermal conditions by instrument-
ing and monitoring a Joe Slovo house and
an older RDP house in close vicinity for
the periods March to December 2013,
and March to October 2013 respectively.
The research leading to this paper aimed
at developing a computational model-
ling strategy for the prediction of indoor
thermal conditions in local houses, towards
proposing further potential improve-
ments. The objectives of the study were:
(i) to model thermal conduction, radiation
and convection to reasonable accuracy,
including those in wall cavities typical of
these structures, (ii) to compute equivalent
thermal transmittance coefficients for
cavity wall systems, required (iii) for use in
a commercial building thermal simulation
package (DesignBuilder 2022) that enables
realistic modelling of a complete house,
its surroundings relevant to its thermal
performance, as well as weather condi-
tions over the period of study, and (iv) to
quantify the occupant thermal comfort of
the Joe Slovo house and further potential
improvement should 3D-printed concrete
walls replace the actual cavity walls of
this house.
BACKGROUND
Thermal comfort for occupants
Thermal comfort is a duration value of the
period of time that occupants feel com-
fortable indoors, based on humidity and
temperature. It was used in this research
to quantify the possible improvement in
comfort for occupants of the proposed
40
Informal shacks
Figure 1 S ignificantly improved living conditions in South Africa (NDHS 2013)
32
30
Operative temperature (°C)
28
26
24
y = 0.255x + 18.9
20
18
16
14
5 10
Mean outdoor effective temperature (°C)
Figure 2 A
 daptive model for naturally ventilated buildings (reworked from De Dear & Brager 1998)
3DPC (three-dimensional printed
concrete) housing. Thermal comfort in
buildings is challenging to define since it
depends on the occupant’s perception of
comfort. While temperatures in the range
18°C to 24°C are generally considered
appropriate for health and thermal com-
fort (Ormandy & Ezratty 2012), adaptive
thermal models have been proposed by De
Dear and Brager (1998) and standardised
in ASHRAE Standard 55 (2020) and BS
EN 16798-1:2019 (BS 2019).
ASHRAE Standard 55 (ASHRAE
2020) – on the Thermal Environmental
Conditions for Human Occupancy – con-
cludes that thermal comfort for occupants
in an environment depends on six factors,
namely metabolic rate, clothing insulation,
air temperature, radiant temperature, air
speed, and humidity. The metabolic rate
of occupants is their effective sensible heat
loss to the environment. On average, this
is 75 W per person according to ASHRAE
Standard 55 and can range from 235 W for
intense physical exercise to 40 W during
sleep. Clothing worn by occupants can
act as insulation. Walking shorts with a
short-sleeved shirt have a factor of 0.36
insulation, while trousers with a T-shirt
and long-sleeved sweater have a factor
of 1.01 (ASHRAE 2020). These first two
Volume 65 Number 3 September 2023 Journal of the South African Institution of Civil Engineering
IRDP improved
RDP buildings
buildings
90% accept
80% accept
15 20 25 30 35
factors are characteristics of the occupant
and can be self-adjusted. The remaining
four factors are conditions of the thermal
environment and are generally dependent
on the building envelope.
De Dear and Brager (1998) developed an
adaptive thermal comfort model. The pri-
mary assumption was that people in warm
climate zones would prefer warmer indoor
temperatures than people in cold climates,
i.e. a reduced difference between indoor
and outdoor temperatures. For naturally
ventilated buildings, an environment is
seen as comfortable when the operative
temperature is within either the 80% or
90% acceptability range of a linear regres-
sion predicted ideal operative temperature,
as per Figure 2.
In South Africa, there is a lack of guid-
ance as to designing for occupant comfort.
The South African Bureau of Standards set
out SANS 204:2011 for energy efficiency in
buildings (SANS 2011) and SANS 10400-
XA:2021 for energy usage in buildings
(SANS 2021). These regulations make
design recommendations for the building
orientation, thermal performance of floors,
walls, windows and roof elements, building
sealing and air infiltration, and building
services and air conditioning systems. The
building should ideally be orientated with
the most used living spaces facing north.
For the coastal region from Cape Town
to Gqeberha (previously Port Elizabeth),
the minimum thermal resistance of
1.9 m2K/W and a thermal capacity and
resistance product of 80 hours are pro-
posed for façades with a surface density of
less than 270 kg/m2, such as the material
used in this research. Modelling conditions
such as ambient temperatures ranging from
19°C to 25°C, average sensible heat gain per
person of 75 W, and occupant behaviour
are also proposed.
Thermal energy transfer
The principle of energy conservation is used
in finite element analysis (FEA) to model
the heat transfer through a wall with cavi-
ties. Thermal energy in the form of ambient
temperature and solar radiation is applied
to the external face of the wall. The energy
is absorbed by the wall dependent on the
surface area, surface roughness, solar emis-
sivity and thermal conduction. Through the
three heat transfer mechanisms (conduction,
convection and radiation), the energy is then
transferred to the inner face and internal
environment. The principle of energy balance
should hold throughout this process without
any energy loss in a controlled environment,
as per Equation 1, where q is the heat flux
through a unit area (Bergman et al 2011).
qexternal = qwall = qinternal
qsolar rad + qext temp = qconduction +
qconvection + qradiation = qint temp + qint rad(1)
A 2D approach is taken to simplify the
problem. The section is homogenous
in height and the only variable to be
accounted for is the convection inside the
cavities. Due to the height of the walls,
natural convection will occur with warmer
and less dense air moving up in the cavity,
and colder, more dense air moving down,
accounted for by calculations of the con-
vective heat transfer coefficient.
dT
qconduction = –kA (2)
dx
Where: q is the heat flux (W), A is the
surface area that is exposed to the heat flux
dT
(m2), and is the negative temperature
dx
gradient due to thermal energy dissipation
through the solid (K/m).
Convection interaction
The convection interaction of moving fluid
in contact with a surface or in cavities can
be represented by Equation 3 (Bergman et
al 2011):
qconvection = hA(T∞ – Ts)(3)
Where: h is the convection heat transfer
coefficient (W/m2K) and (T∞ – Ts) refers
to the difference between surface and fluid
temperatures, respectively (K). The convec-
tion heat transfer coefficient is the only
unknown and depends on the conditions in
the boundary layer, influenced by surface
geometry, the nature of fluid motion,
and fluid thermodynamic and transport
properties.
For calculation of the convection heat
transfer coefficient, two conditions exist –
either forced convection where air is driven
by external factors (such as an HVAC
system, frontal wind, and convective macro
environmental wind), or natural convec-
tion where air is driven by temperature
differences. On the external surface of the
façade air is driven by the wind and forced
convection is used. In the façade internal
cavities, no external driving factors are
present and natural convection is used. On
the internal face where no external driving
factors or cavities are present, a maximum
heat transfer coefficient of 25 W/m2K is
applied according to Bergman et al (2011).
The coefficient can be solved from the
Table 1 Calculation of Nusselt number (Bergman et al 2011)
Situation
Nusselt (Nu) number according to Table 1,
dependent on either the Reynolds number
(Re), or the Grashof (Gr) and Prandlt (Pr)
numbers. These can be calculated from
Equations 4 to 7:
hl
Nu =  (4)
k
pu∞l
Re = (5)
μ
glβ∆T
Gr = (6)
v2
v
Pr = (7)
α
Where: l is the height of the wall (m), u∞ is
the average wind velocity on the external
face of the façade (m/s), β is the inverse of
air temperature (K), ∆T is the difference
in air and surface temperatures (K), and
k, v, α, ρ and μ are the thermophysical
properties of air at atmospheric pressure
and 300K.
Radiation interaction
The radiation interaction is presented by
Equation 8 (Bergman et al 2011):
qradiation = σAε(T14 – T24)(8)
Where: σ is the Stefan-Boltzmann constant
of 5.67 × 10 –8 W/m2K4, ε is the emissivity
of the surface from which the radiation is
emitted, and T1 and T2 are the surface or
environment temperatures (K).
Steady state
When a system is in steady state, the
conservation-of-energy principle is satisfied
where the energy into a system is equal to
the energy out of the system and no energy
is created or lost. Equation 1 must therefore
Nu

Conduction interaction Forced convection

The conduction interaction can be derived Laminar flow, parallel to flat plate (20 < Re < 3 × 105) 0.66 Re ¹∕₂ Pr ¹∕₃
as thermal energy flowing through a solid,
Turbulent flow, parallel to flat plate (Re > 3 × 105) 0.037 Re⁴ ∕₅ Pr ¹∕₃
or through stationary fluid. Fourier’s law
describes the thermal energy flowing Natural convection
through the solid parts of the section Laminar flow, vertical flat plate (104 < Gr ∙ Pr < 109) 0.59 (GrPr)¹∕₄
as expressed in Equation 2 (Bergman et
al 2011), where the product of thermal Turbulent flow, vertical flat plate (Gr ∙ Pr > 109) 0.13 (GrPr)¹∕₃

conductivity and temperature difference Laminar flow, hot horizontal plate (105 < Gr ∙ Pr < 2 × 107) 0.54 (GrPr)¹∕₄
between two connected nodes in the sec-
Turbulent flow, hot horizontal plate ( 2 × 107 < Gr ∙ Pr < 3 × 1010) 0.14 (GrPr)¹∕₃
tion is calculated.

Journal of the South African Institution of Civil Engineering Volume 65 Number 3 September 2023 41
be satisfied, but with the time-dependent
specific heat term included. Without, 32
the problem has no intrinsic meaningful 30
28
timescale. This is why a transient analysis
26

Temperature (°C)
is used with the backward Euler method
24
which is unconditionally stable for linear 22
problems to determine when steady state is 20
reached (Dassault Systemes Simulia 2020). 18
In this transient analysis, external tempera- 16
ture and solar flux are applied at specific 14
time intervals, but since the heat flux in the 12
cavities is unknown, an iterative approach 10

00:00
01:00
02:00
03:00
04:00
05:00
06:00
07:00
08:00
09:00
10:00
11:00
12:00
13:00
14:00
15:00
16:00
17:00
18:00
19:00
20:00
21:00
22:00
23:00
is required.
Time
Determining thermal transmittance (a) Outside IDRP downstairs IDRP upstairs RDP inside
The thermal transmittance was calculated
by computationally simulating the Hot Box
32
procedure according to ASTM C1363:2014
30
(ASTM 2014). This value effectively quanti-
28
fies the amount of thermal energy that is 26
Temperature (°C)

transferred through the section. It is calcu- 24

lated from Equation 9: 24
20
Q 18
U= (9)
A ∙ (tenv,h – tenv,c ) 16
14
12
Where: U is the thermal transmittance
10
of the section (W/m2K), Q is the average
00:00
01:00
02:00
03:00
04:00
05:00
06:00
07:00
08:00
09:00
10:00
11:00
12:00
13:00
14:00
15:00
16:00
17:00
18:00
19:00
20:00
21:00
22:00
23:00
time rate of heat flow (W), A is the area on
which the external temperature and solar Time
flux is applied (m2), and tenv,h and tenv,c are (b) Outside IDRP downstairs IDRP upstairs RDP inside
the ambient temperatures on the hot and
cold surfaces, respectively. The thermal Figure 3 D
 aily average measured temperature during (a) the summer and (b) the winter period
resistance (R) is the reciprocal value of (NDHS 2013)
thermal transmittance.
finish. All overhangs are at least 500 mm to Thermal performance
Joe Slovo Phase Three provide additional shade. The 4 mm Nutec and monitored data
Block 6 on the northernmost land was ceilings are at the minimum height of In December 2013 the Department of
selected for analysis in this research. The 2.4 m and insulated with 40 mm Isotherm Human Settlements in South Africa, in col-
building was constructed to comply with mineral wool. laboration with Sustainable Energy Africa,
the regulations of the National Home released a case study report on the lessons
Builders Registration Council (NHBRC) Windows and doors learnt from the project after the thermal
of South Africa. It consists of six double- All windows consist of 3 mm thick plane performance was monitored to evaluate the
storey 50 m2 apartments, each comprising glass following SANS 10400-N:2012 improvement over traditional RDP build-
a living area, kitchen and bathroom down- (SANS 2012) with 10 mm Cliscoe-type ings (NDHS 2013). From the study, it was
stairs, and two bedrooms upstairs. steel window frames. The external doors clear that the thermal performance had
are of 40 mm hardwood and the internal significantly improved under the new IRDP
Floors doors are of 40 mm masonite-faced hollow initiative. The temperature peaks were
The 75 mm thick ground-floor slab is con- flush panels. reduced by between 4°C and 5°C for both
structed of 10 MPa cube strength concrete summer and winter periods, as illustrated
on 250-micron membrane and a 150 mm Walls in Figure 3.
layer of compacted subsurface. The first- The external ground-floor loadbearing
floor slab is a hollow core precast class walls are compressive strength class 7 MPa, 3D-printed concrete
25 MPa concrete slab with a maximum 140 mm wide hollow blocks, while the Typical RDP buildings are constructed
stone size of 6 mm. The slab bears on the first-floor walls are class 3.5 MPa following using traditional masonry methods, which
walls with a 90 mm overlap. SANS 10400-K:2015 (SANS 2015). The walls are unproductive in terms of labour and
are plastered with a 12 mm mortar and time. Three-dimensional (3D) printed
Roof painted with two coats of water-­resistant concrete (3DPC) construction and the
The roof is covered with Widedek acrylic paint externally. The internal parti- rapid shift to 3D building information
Zincalume sheets with a Colourbond tions are 90 mm wide masonry blocks. modelling (BIM) mitigate these problems

42 Volume 65 Number 3 September 2023 Journal of the South African Institution of Civil Engineering
by the high-precision and versatile nature Table 2 LWFC and dense concrete thermal properties (CIBSE 2006; Marais et al 2021)
of construction from digital models that LWFC Dense concrete
inherit all construction details. This can be
Density 1 400 kg/m3 2 300 kg/m3
seen as the fourth industrial revolution for
construction (Kruger 2019). Specific heat 1 100 J/kg°C 1 000 J/kg°C
3DPC construction is generally per- Thermal conductivity 0.3708 W/mK 1.63 W/mK
formed by depositing printable concrete
Surface emissivity 0.85 0.9
through a nozzle in a three-dimensional
environment by moving on a gantry sys-
tem. This enables a wide range of intricate
wall cross-section geometries that can be (a) (b) (c) (d)
constructed, each with a different thermal
transmittance. Marais et al (2021) found
that a façade geometry with cavities of
20 mm in width performs better thermally
than a solid façade of the same lightweight
foamed concrete (LWFC), due to the low
conductivity relative to heat transferred (e) (f) (g)
by cavity convection and radiation. The
difference in thermal properties of LWFC
(Marais et al 2021) and typical dense con-
crete (CIBSE 2006) is illustrated in Table 2.
This renders a façade section with a higher
number of thin cavities to perform better
than one with a few larger cavities. An Figure 4 D
 ifferent sections analysed: (a) solid reference, (b) hollow concrete block,
overall void-to-façade footprint ratio of less (c) symmetrical triangles, (d) single triangle, (e) single triangle rotated by 180°, (f) double
than 50% is recommended. triangle, (g) double triangle rotated by 180° (not to scale)

FINITE ELEMENT ANALYSIS
OF HEAT TRANSFER
THROUGH CAVITY WALLS
This section covers the development of
an appropriate finite element analysis
(FEA) approach for heat transfer through
cavity walls, appropriately incorporating
conduction, as well as cavity radiation and
convection. The simulation is validated
against measured data from a physical
experiment on a 3DPC wall containing large
cavities. Subsequently, equivalent thermal
transmittance values of cavity walls are
determined computationally with the FEA
approach. Commercial thermal simulation
packages that enable simulation of complete
buildings and relevant surroundings require
transmittance values. DesignBuilder (2022)
has predetermined transmittance values for
typical wall types, most appropriately deter-
mined by physical testing. However, those
of non-standard wall types, in this case
3DPC walls with intricate cross-sectional
shapes containing non-standard cavities and
arrangements, are not available.
Methodology
In this research, different wall geometries
from literature and a reference solid sec-
tion illustrated in Figure 4 were analysed,
all manufactured from lightweight foam
concrete (LWFC). For consistency, ±140
mm thick sections were used for all geom-
etries, similar to the analysed hollow con-
crete blocks used for the Joe Slovo house.
Applied ambient temperature
and solar heat flux
From the nearby Cape Town International
Airport’s measured meteonorm weather
data, included in the DesignBuilder (2022)
database, a three-day average ambient
temperature and solar heat flux were
calculated for 15-minute increments. This
data could then be applied in a transient
response analysis where environmental
loads are applied at specific time incre-
ments. The ambient temperature was used
as reference for convection and radiation
(qconvection + qradiation), while the solar
heat flux was applied as a surface heat flux
(qsolar rad) on the external face.
Steady state
For the first iteration, the external tempera­
ture and solar flux were applied to a system
that had a predefined temperature of 20°C,
but the convection interaction could not be
applied to the internal cavities since their
temperatures were unknown. The radiation
interaction was still applied to internal
closed cavities because the temperature
changes linearly through the cavity based on
radiation viewfactors. The resultant surface
temperature could then be used for input of
the convection interaction of internal cavi-
ties in the second iteration. This procedure
could be followed until the internal surface
of the section reached steady state, occur-
ring when there was no, or an acceptably
small, change in temperature at the internal
surface between successive iterations. A
Python script was developed for this itera-
tive process.
Results and discussion
An example of steady state reached after
five iterations for configuration g during
summer is given in Figure 5. It can be seen
that the maximum internal temperatures
of trial four and trial five are consistent
when rounded to one decimal.
The thermal transmittance of each
configuration wall cross-section, calculated
with the FEA procedure described above, is
summarised in Table 3. The configurations
are as per Figure 4 for a summer period,
winter period, and an average with differ-
ent thicknesses and rotations. The thermal
transmittance, thermal resistance, and
time constant are indicated on three-point
colour scales where green is the best, and
red the worst, respectively.
Closer investigation was done on con-
figuration g since it had the lowest thermal

Journal of the South African Institution of Civil Engineering Volume 65 Number 3 September 2023 43
 32
30.9564
30.7831
30 30.7707
30.7147
28
Temperature (°C)

26

24 23.1094

24

20

18
172 800 182 800 192 800 202 800 212 800 222 800 232 800 242 800 252 800
Time step (sec)
Trial 1 Trial 2 Trial 3 Trial 4 Trial 5

Figure 5 E xample of steady state reached after five iterations

Table 3 FEA results for each configuration

Façade thickness 3DPC-nozzle
Configuration (Figure 4) U (W/m2K) R (m2K/W) CR (hrs)
(mm) thickness (mm)
a – Summer 1.4350 0.6969 74.5279
a – Winter 250 Solid 1.4307 0.6990 74.7490
Average 1.4328 0.6979 74.6383
b – Summer 5.6140 0.1781 10.6677
b – Winter 140 35 5.4992 0.1818 10.8905
Average 5.5566 0.1800 10.7779
c – Summer 1.4249 0.7018 75.0562
c – Winter 250 35 1.2781 0.7824 83.6744
Average 1.3515 0.7399 79.1313
c – Summer 2.0025 0.4994 30.9748
c – Winter 145 20 1.7793 0.5620 34.8612
Average 1.8909 0.5288 32.8033
d – Summer 1.7204 0.5812 44.7558
d – Winter 180 35 1.5945 0.6272 48.2925
Average 1.6574 0.6033 46.4569
d – Summer 2.3990 0.4168 25.8561
d – Winter 145 20 2.0981 0.4766 29.5635
Average 2.2485 0.4447 27.5858
e – Summer 2.4274 0.4120 25.5534
e – Winter 145 20 2.1189 0.4719 29.2734
Average 2.2731 0.4399 27.2872
f – Summer 1.1981 0.8347 89.2630
f – Winter 250 20 1.0417 0.9600 102.6662
Average 1.1199 0.8930 95.4966
f – Summer 1.6212 0.6168 38.2594
f – Winter 145 20 1.5373 0.6505 40.3481
Average 1.5793 0.6332 39.2760
g – Summer 1.6014 0.6244 38.7324
g – Winter 145 20 1.5383 0.6501 40.3221
Average 1.5699 0.6370 39.5113

44 Volume 65 Number 3 September 2023 Journal of the South African Institution of Civil Engineering
 50
46.70

45
43.20

39.40
40
Temperature (°C)

35 33.66

29.80
30

25.78
25

20
0 10 800 21 600 32 400 43 200 54 000 64 800 75 600 86 400 97 200 108 000
Time (sec) (1) (3) (5) (6)

(1) Outside average temperature (2) Left average temperature (3) Top left average temperature (2) (4)

(a) (4) Right average temperature (5) Top right average temperature (6) Inside average temperature (3) (5)

26

25 24.84

24.33

24 23.76
Temperature (°C)

22.91
23
22.34

22 21.76

21

20
0 10 800 21 600 32 400 43 200 54 000 64 800 75 600 86 400 97 200 108 000
Time (sec) (1) (3) (5) (6)

(1) Outside average temperature (2) Left average temperature (3) Top left average temperature (2) (4)

(b) (4) Right average temperature (5) Top right average temperature (6) Inside average temperature (3) (5)

Figure 6 T emperatures in configuration g for (a) the summer and (b) the winter period

transmittance compared to all ±140 mm
thick configurations. The temperature
curves illustrated in Figure 6 can be com-
piled for locations throughout the section
to clearly see the reduction in temperature
from the outside to the inside surface.
From Table 3 it is clear that configura-
tion g performed the best when comparing
sections similar in thickness to the existing
140 mm hollow concrete blocks of the
selected baseline building. The thermal
transmittance of the existing building
façades, configuration b, can be reduced
by 71.5% from 5.6 to 1.57 W/m2 if replaced
with configuration g. Comparing configura-
tion d and e concludes that more cavities
throughout the section and layering sections
on each other reduce the thermal resistance.
From Figure 6 it is also clear that all
cavities and surfaces of configuration g
reach steady state after about three hours.
From comparing the temperature differ-
ence between the exterior and cavities
numbered 1 and 2 respectively in Figure 6,
and the temperature difference between
cavities 1 and 3, it is clear that there is a
greater temperature reduction to cavity 3
than to cavity 2. The concrete filaments
bounding the number 2 cavity conducts
heat at 45° angles through the section. The
diagonal concrete bridge affects cavity 3
less, and thus results in a greater tem-
perature reduction. The large temperature
reduction between the node 2 and 4 cavi-
ties proves that the layering of two sections
is effective.

Journal of the South African Institution of Civil Engineering Volume 65 Number 3 September 2023 45
Validation
(a)
Experimental setup
Physical testing of configuration c (see
Table 3) was done to validate the FEA in
collaboration with colleague Heidi Christen
at Stellenbosch University (Christen et al
2023). The test setup was designed with
one surface of the façade exposed to solar
radiation and external ambient tempera-
ture, while the other was completely insu-
lated as per the cross-section illustrated in
Figure 7(a). For the testing to be as accurate
as possible, the insulated part should add
as little heat gain as possible. A 40 mm
polystyrene insulation was used all around
to reduce additional heat gain or loss by
conduction through these sides. Type T
thermocouples (Energi Solutions @ Africa
2015) were placed in the cross-section, as
seen in Figure 7b. These thermocouples
had a measuring accuracy of 0.5°C when
connected to a 34980A BenchLink
(b) Inside 11 10
Data Logger.
The FEA methodology described in
the previous section was then followed to 3 4
perform an analysis under similar circum-
13
stances using the measured solar radiation
and external ambient temperature from
the adjacent Sonbesie Weather Station 5
(Meijers 2021). The temperatures of the 1 12
6
same nodes were then compared using the
Pearson correlation coefficient according to
Equation 10: 14
2 7

∑(xi – x)(yi – y)
r= (10)
√∑(xi – x)2 ∑(yi – y)2 Outside 8 9

Where: r is the correlation coefficient, xi
are the values of the test sample, x is the
mean of the test sample, yi are the com-
puted values of the FEA sample, and y is
the mean of the FEA sample.
Results
The physical test data and results from
the last 24 hours of FEA are presented in
Figure 8 for comparison. Only nodes 8,
2, 12, 3 and 11 (refer to Figure 7(b)) are
compared, since they represent a straight
line of thermal energy transference through
the cross-section. The same colour and line
markers represent the same nodes. The solid
lines with abbreviation T represent the data
from physical testing, and dashed lines with
abbreviation A represent the data from FEA.
The calculated Pearson correlation coef-
ficient for the last 24 hours is summarised
in Table 4. The correlation is shown on a
Figure 7 T hermal testing: (a) setup of 3DPC façade element shape c, and (b) placement of
thermocouples central in the 3DPC element
three-point colour scale where green is the
best correlation and red the worst.
Discussion
A reasonable correlation was found
between the physical test data and FEA,
with average Pearson correlation coefficient
of 0.83. From Figure 8 the temperatures for
the tested and analysed data sets correlated
well, with maximum and minimum peaks
accounted for. The gradients, representing
the change in temperature between time
steps, were constant and of similar shape.
From Table 4 it is clear that the tempera-
tures on the external surface (nodes 8 and
9) correlate well with a correlation coef-
ficient of 0.95. This is expected, since the
ambient conditions were applied there. As
the thermal energy was conducted to the
internal surface, the correlation reduced to
0.79 at node 11, which is still acceptable.
DESIGNBUILDER SIMULATION
OF THE JOE SLOVO HOUSE
Following from the FEA of integral façade
parts, the resulting equivalent transmittance
properties were incorporated in full build-
ing simulations to conclude on the thermal
comfort levels in the Joe Slovo house, and for
the case of replacing the cavity walls with
3DPC façades. State-of-the-art DesignBuilder
software can be used to determine thermal
comfort for occupants of buildings, approved
by SANS 10400-XA in South Africa
(DesignBuilder 2022; SANS 2021).

46 Volume 65 Number 3 September 2023 Journal of the South African Institution of Civil Engineering
 50

45

40

35
Temperature (°C)

30

25

20

15

10
16:00 18:00 20:00 22:00 00:00 02:00 04:00 06:00 08:00 10:00 12:00 14:00 16:00
Time
T8 T2 T12 T3 T11 Ambient
A8 A2 A12 A3 A11

Figure 8 T emperature comparison between testing (T) and FEA (A) results

Table 4 Pearson correlation coefficient calculated per node

Testing Abaqus
Pearson
Average Standard Average Standard correlation
Node Node
(°C) deviation (°C) (°C) deviation (°C)

T1 21.5232 3.4882 A1 22.1916 3.3038 0.8231

T2 23.4041 7.2287 A2 24.0544 7.7907 0.9225

T3 20.9472 1.9899 A3 20.9203 1.4780 0.7686

T4 20.6220 1.6972 A4 20.6024 1.2116 0.7461

T5 20.7425 2.7265 A5 21.7719 2.6224 0.7327

T6 21.4905 4.6061 A6 22.6743 4.1950 0.8158

T7 23.0884 8.9471 A7 24.6098 9.1879 0.9180

T8 24.2907 12.1098 A8 25.6622 11.9988 0.9462

T9 25.0053 11.7339 A9 25.9077 12.8926 0.9495

T11 21.2114 1.6709 A11 20.2202 0.9441 0.7893

T12 22.1620 3.6243 A12 22.3199 3.6195 0.8329

T13 20.8611 2.1706 A13 20.8931 1.5356 0.6969

T14 23.4890 8.3177 A14 24.1003 7.6985 0.8955

Average 0.8336

Methodology influencing the weather conditions. Since Construction

Environment modelling
Site modelling involves modelling of the
building layout, orientation and location,
ground temperatures, weather condi-
tions, and stipulating adjustment factors
the meteonorm weather data supplied
by DesignBuilder (2022) was measured,
no weather adjustment factors had to be
applied. The building and neighbouring
buildings that could cast shadows or radiate
heat, as seen in Figure 9, were also modelled.
Specific details of the existing construc-
tion elements according to the construc-
tion drawings were reported and modelled
according to CIBSE Guide A 2006 (CIBSE
2006) for the baseline building. This
guide is useful, with a wide range of

Journal of the South African Institution of Civil Engineering Volume 65 Number 3 September 2023 47
 (a) (b)

Figure 9 M
 odel layout and shadow patterns of Block 6 and neighbouring buildings for (a) 22 December at 12:00 and (b) 10 June at 12:00

Table 5 Layered construction of elements

Layer 1 Layer 2 Layer 3
Element Thickness Thickness Thickness
Material Material Material
(mm) (mm) (mm)

Doors (external) Hardwood 40

Doors (internal) Hardwood 6 Air gap 28 Hardwood 6

First floor Concrete reinforced 75 Floor screed 20

Ground floor PVC membrane 0.3 Concrete reinforced 75 Floor screed 20

Roof Zinc sheet 0.7 Mineral wool 40 Nutec plasterboard 4

Wall (Hollow block) Cement plaster 12 Hollow block 140

Wall (Masonry) Masonry 90

Windows Glass 3

material properties to be used, and is also
adopted by the DesignBuilder software for
default parameter values. The elements
were then constructed in layer formats
according to Table 5 where layer 1 is the
outermost layer. These layers were parallel
to each other with different thicknesses
perpendicular to the thermal energy
transfer direction. In Table 6 the mate-
rial properties are quantified to be used
in the layered construction. The 3 mm
thick glass, excluding the window steel
frame, had a solar transmittance factor of
0.837, and all other material had a surface
emissivity of 0.9 according to the CIBSE
Guide A (CIBSE 2006). The steel frames
of relatively high thermal conductivity
were a concern for thermal bridging.
3DPC walls
Since the geometry of the 3DPC walls can-
not be layered parallel as easily, alternative
methods were used to transform the configu-
rations into a layered format, as required by
DesignBuilder. The thermal transmittance
values were calculated for these methods
in DesignBuilder and compared with FEA,
combining the sections in a useful tool.
Three-, five- and nine- layered alterna-
tives were used, as presented in Figure 10.
Table 6 Material properties (CIBSE 2006)
Conductivity
Material
(W/mK)
Cement plaster 0.72
Concrete reinforced 1.9
In each case the internal and external
flanges of the section were kept constant
at 20 mm thick, since these were already
parallel to each other, while the thickness
of triangular cavity parts was calculated
by 3DPC-to-cavity cross-sectional area
ratios. Since the unusual shape of the
triangular cavity part would not transfer
thermal energy linearly, the thickness of
this part was iterated until the calculated
thermal transmittance in DesignBuilder
Specific heat Density
(J/kgK) (kg/m3)
840 1 860
840 2 300

Nevertheless, their equivalent transmit- Floor screed 0.41 840 1 200

tance was simulated in the model. Glass 0.9 – –
Natural ventilation from open win-
Hollow block 1.35 840 1 220
dows and doors can be controlled in
the software, but was not specified for Hardwood 0.17 1880 700
this simulation. The opening of doors Masonry 0.2 840 520
was assumed to follow occupancy, while
Mineral wool 0.038 840 140
windows were assumed to be closed at all
times. Air could still filter into or out of the Nutec plasterboard 0.16 840 950
building through construction gaps and PVC membrane 0.14 1000 1 200
was assumed to be constant at 7 m3/hm2,
Zinc sheet 110 380 7 200
similar to Meyer (2022).

48 Volume 65 Number 3 September 2023 Journal of the South African Institution of Civil Engineering
 Outer surface Outer surface Outer surface

20 mm 3DP foamed concrete 20 mm 3DP foamed concrete 20 mm 3DP foamed concrete

2.5 mm 1. Air layer

6.3 mm 1. Air layer
13.9 mm 1. Air layer 6.2 mm 3DP foamed concrete
3.3 mm 1. Air layer

20 mm 3DP foamed concrete

20 mm 3DP foamed concrete
20 mm 3DP foamed concrete
6.3 mm 1. Air layer
2.5 mm 1. Air layer
Inner surface 6.2 mm 3DP foamed concrete
20 mm 3DP foamed concrete 3.3 mm 1. Air layer

Inner surface
20 mm 3DP foamed concrete

Inner surface

Figure 10 T hree-, five- and nine-layered alternatives (not to scale)

was equal to 1.57 W/m2K as calculated by

the FEA. The three-layered alternative was 30
represented by two parallel 3DPC external 28
elements, but the centre parts were repre- 26
sented by only one air layer. Similarly, the 24 28.54
Temperature (°C)

five-layered alternative had an additional 26.88

22
24.38 26.56
parallel 3DPC element in the centre. For 20
the nine-layered alternative, however, the 23.95
18 22.94
thickness of parts was calculated accord-
16
ing to ratios. It had two triangular parts
14
between parallel 3DPC elements with an
12
area of 3 950 mm2, of which 21% was the
10
first air cavity, 51% 3DPC, and 28% the 00:00 02:00 04:00 06:00 08:00 10:00 12:00 14:00 16:00 18:00 20:00 22:00 00:00
second air cavity. Multiplying these ratios Time
with the triangular cavity part thickness of (a) Baseline Measured upstairs Measured downstairs
12 mm, calculated by iteration, resulted in
the nine-layered alternative which was used
30
for further analysis.
28

Simulation parameters 26
21.25
A wide range of outputs are generated by 24
Temperature (°C)

18.98
DesignBuilder, including surface and ambi- 22 16.81
15.19
ent temperatures, thermal heat gains for 20 13.50
elements, internal heat gains and thermal 18 12.46
comfort for occupants. All of these outputs 16
can be useful, but for this research the 14
focus was mainly on the thermal comfort
12
for occupants. This can be calculated by
10
ASHRAE Standard 55 Adaptive models 00:00 02:00 04:00 06:00 08:00 10:00 12:00 14:00 16:00 18:00 20:00 22:00 00:00
discussed earlier. Based on the validation Time
done below, well-correlated representa- (b) Baseline Measured upstairs Measured downstairs
tive days for both a summer month, 22
December, and a winter month, 10 June, Figure 11 H
 ourly internal ambient temperature from simulation and measured data for
were chosen. (a) 22 December 2013 and (b) 10 June 2013

Journal of the South African Institution of Civil Engineering Volume 65 Number 3 September 2023 49
 data from Sustainable Energy Africa (2014)
30 for both the summer and winter periods
28.47
28 in Figure 11. The Pearson correlation coef-
26 ficient according to Equation 10 was cal-
28.52
24 culated to be 0.88 and 0.79 for the summer
Temperature (°C)

25.26 period, and 0.93 and 0.96 for the winter

22
23.93
18.98 period, for the downstairs and upstairs
20
18.64
18
internal ambient temperature, respectively.
The Pearson correlation coefficient of the
16
outside surface temperature was 0.94 for
14
14.23 both periods, showing good correlation of
12
12.46 the façade’s ability to absorb ambient heat.
10
00:00 02:00 04:00 06:00 08:00 10:00 12:00 14:00 16:00 18:00 20:00 22:00 00:00 A further validation was done by checking
Time that the thermal transmittance values,
Baseline – summer 3DPC – summer according to Equation 9, of all façades were
Baseline – winter 3DPC – winter equal to those calculated in the FEA.

Figure 12 H
 ourly internal ambient temperature from simulation Results and discussion

Validation Temperature
(a) The results of the baseline simulation The computed internal ambient tempera-
model were validated against the measured tures for the baseline building with hollow

4
20.4 27.5
2
Temperature (°C)

–2

–4

20.4 27.5
–6
00:00 02:00 04:00 06:00 08:00 10:00 12:00 14:00 16:00 18:00 20:00 22:00 00:00
(b) Time
3DPC walls First floors Roofs
(a) Baseline walls Ground floors Windows

4
12.7 18.0

2
Temperature (°C)

–2

–4

12.7 18.0
–6
00:00 02:00 04:00 06:00 08:00 10:00 12:00 14:00 16:00 18:00 20:00 22:00 00:00
Figure 13 E xternal surface temperature (°C) Time
for (a) December and (b) June of the 3DPC walls First floors Roofs
3DPC model’s north-facing (a and b (b) Baseline walls Ground floors Windows
top images) and south-facing (a and
b bottom images) façades Figure 14 H
 eat gains of elements on (a) 22 December 2013 and (b) 10 June 2013

50 Volume 65 Number 3 September 2023 Journal of the South African Institution of Civil Engineering
concrete blocks were compared to the
building with 3DPC façades in Figure 12 1.0
for 22 December and 10 June 2013. These 0.9
temperatures were averaged over all 0.8
internal spaces. The thermal comfort for
0.7

Discomfort factor
occupants is not dependent only on the
0.6
internal ambient temperatures, but they do
give a good indication thereof. 0.5
The external monthly average surface 0.4
temperatures of the 3DPC building are 0.3
presented on a colour scale in Figure 13 for
0.2
both the months of December and June.
0.1
Similar results were found for the 3DPC
and baseline buildings, with the latter 0
00:00 02:00 04:00 06:00 08:00 10:00 12:00 14:00 16:00 18:00 20:00 22:00 00:00
having 1% cooler surface temperatures. (a) Time
The thermal absorptivity of both façade 3DPC – simplified 3DPC – 80% adaptive 3DPC – 90% adaptive
materials was similar, but the difference in Baseline – simplified Baseline – 80% adaptive Baseline – 90% adaptive
temperature was because of the convection
close to the rougher, layered 3DPC surface.
1.0
From the external surface temperature in
Figure 13 it is useful to see which wall and 0.9
roof surfaces experienced comfort. 0.8
0.7
Discomfort factor

Heat gains by elements

The thermal heat gains and losses per
element are presented in Figure 14 for
22 December and 10 June 2013. Presented
are all major external elements, including
the 3DPC and baseline building walls. A
positive heat gain is when the heat flux
through an element increases the internal
ambient temperature, and vice versa. Solar
heat gains by the windows are the largest
contributor to heat gains of the building,
followed closely by roofs in the summer
and walls in the winter months. The opti-
mised design of the building can clearly
be seen by the solar heat gain curve of the
windows. The solar heat gain is reduced
during the peak temperature hours of 12:00
to 14:00 by shadowing solar heat gains
from the windows. The solar heat gains are
also higher during the winter period when
additional heat is required, which further
improves the design. The heat gains and
losses from the ground floor interaction
are beneficial for thermal comfort during
winter and summer periods, respectively.
The heat gain data of elements confirms
that the walls are not the largest contribu-
tor to heat gains in this building model.
The positive heat gain of wall elements
is outweighed by both solar heat gains
through windows and heat gains via the
roof element during the summer period,
but not the winter period. This can be seen
as a positive attribute because for summer
periods additional heat gains will have a
negative influence on the thermal comfort,
while for winter periods additional heat
0.6
0.5
0.4
0.3
0.2
0.1
0
00:00 02:00 04:00 06:00 08:00 10:00 12:00 14:00 16:00 18:00 20:00 22:00 00:00
(b) Time
3DPC – simplified
Baseline – simplified
Figure 15 H
 ourly discomfort factor for (a) 22 December 2013 and (b) 10 June 2013
gains are beneficial. The heat gains of the
3DPC wall follow the same trend as the
baseline wall, but reduce the maximum and
minimum heat gain range. The occupant
thermal comfort was improved by reducing
the temperature range.
Thermal comfort
The thermal comfort is represented by
the discomfort hours for occupants of the
building for the simplified, 80% adaptive
and 90% adaptive methods. The hourly
discomfort factor for both baseline and
3DPC façade buildings for summer and
winter representative days are presented
in Figure 15. The vertical axis ranges from
zero to one representing the hours of
discomfort per hour interval. Zero would
represent zero minutes of discomfort per
hour interval, while one represents 60
minutes of discomfort per hour interval.
The solid lighter-coloured line represents
3DPC – 80% adaptive 3DPC – 90% adaptive
Baseline – 80% adaptive Baseline – 90% adaptive
the 3DPC façade, and the dashed darker-
coloured lines represent the baseline façade
for all three thermal comfort methods.
At a quick glance, one can observe that
the 3DPC façade is less comfortable for
occupants than the baseline façade for the
summer day, but the opposite is true for
the winter day.
The thermal discomfort hour results for
all three ASHRAE methods are summarised
in Table 7 for the whole year, i.e. the summer,
autumn, winter and spring periods. The table
includes the percentage of improvement for
each method between the baseline and 3DPC
façade buildings on a three-point colour scale
where green is the greatest improvement and
red the least improvement for the simple,
adaptive 80% and adaptive 90%, respectively.
During the summer and winter months the
thermal comfort improved slightly for occu-
pants of the 3DPC façade building, but these
improvements were outweighed over the

Journal of the South African Institution of Civil Engineering Volume 65 Number 3 September 2023 51
Table 7 Thermal discomfort hours summary
ASHRAE
Period
method Baseline
Simple
Summer Adaptive 80%
Adaptive 90%
Simple
Autumn Adaptive 80%
Adaptive 90%
Simple
Winter Adaptive 80%
Adaptive 90%
Simple
Spring Adaptive 80%
Adaptive 90%
Simple
Year Adaptive 80%
Adaptive 90%
whole year by the significant improvement
in comfort during the autumn and spring
transition months when temperatures are not
at extremes.
The thermal comfort data is plotted for
all three thermal comfort methods. From the
hourly data it can be seen that the simplified
and 90% adaptive methods are more sensitive
to sudden changes in temperature, which can
lead to inaccurate and non-uniform results.
Since the 80% adaptive method also takes
into account a wider range of data points
with 80% acceptability, it would be the meth-
od referred to for results. The winter month
of the existing building proved to be the least
comfortable, followed by spring, autumn,
then summer the most comfortable. A 2.4%
improvement in thermal comfort summer
periods, and a 14.1% improvement for winter
periods can be seen from the results. The
transition seasons of autumn and spring,
where the maximum and minimum temper-
atures are not as extreme, had a much greater
improvement of 83% and 80%, respectively.
The discomfort hours for occupants reduced
form 4 400 hours per annum for the baseline
façade to 2 244 hours per annum for the
3DPC façade. This yields a 49% improvement
in thermal comfort for occupants according
to the 80% adaptive method.
CONCLUSION
There is a need for better thermal per-
formance of buildings in South Africa to
52
QQ The validated computational procedure
Thermal discomfort hours justified its use to compute equivalent
thermal transmittance values for differ-
3DPC Improvement
ent 3DPC façade configurations and the
294.85 389.15 -32.0% existing façade of the Joe Slovo phase
128.84 125.81 2.4% house, in lieu of thermal transmittance
test data from the standardised Hot Box
251.06 267.33 -6.5%
procedure of ASTM C1363:2014.
420.30 397.44 5.4% QQ From the FEA, the thermal transmit-
212.99 37.05 82.6% tance of the existing building façades
can be reduced by 71.5% from 5.6 to
357.78 101.99 71.5%
1.57 W/m2 if replaced with a 3DPC
691.88 684.04 1.1% façade alternative. The 3DPC façade
654.20 562.12 14.1% also performed better than a double-
layered masonry wall with 1 m2K/W
679.60 641.20 5.7%
insulation according to SANS 204:2011.
613.61 532.65 13.2% QQ DesignBuilder, a commercial simula-
482.75 96.23 80.1% tion software accredited for Green Star
Rating analysis, was used to simulate
584.90 242.10 58.6%
the building as a whole. The level of
5 988.02 5 870.96 2.0% detail on the material parameters, occu-
4 400.92 2 244.19 49.0% pancy usage, and surrounding building
modelling considered by the software
5 576.31 3 626.85 35.0%
reflects the complexity of simulating
an existing, occupied building in its
improve occupant comfort and to reduce local environment. However, the results
the energy consumption in residential helped in reaching conclusions on the
buildings. Significant progress has already thermal comfort of occupants, as well
been made, as suggested in literature, as understanding and designing for a
concluding that this approach is the future building to perform thermally efficient.
for construction. Significantly improved With the weather data for a nearby
indoor thermal conditions were recently location, and no attempt made to
reported in a Joe Slovo Phase Three house improve the agreement between results,
compared to a traditional RDP house in the an average Pearson correlation coef-
vicinity. The Joe Slovo house was selected ficient of 0.89 was obtained, represent-
for this research on thermal comfort, since ing a good correlation between model
all necessary information on construction results and measured data.
details, materials, and thermal performance QQ Three methods for calculating the
monitored data was available. This building thermal comfort are available according
represents the new improved approach for to ASHRAE Standard 55, but the 80%
construction of RDP buildings in South adaptive method was chosen, since it is
Africa, designed to be thermally efficient not as sensitive to sudden temperature
and leaning towards higher-density housing. changes and takes into account a larger
The computational approach in this paper range of data points. For the existing
aimed at developing a predictive capacity building the winter period proved to
for further thermal improvement by 3DPC be the least comfortable, followed by
walls, with careful consideration for heat spring, autumn, and then summer being
transfer through the 3DPC façade configu- the most comfortable. The simulation
rations with cavities. From this research the predicts a 49% improvement in ther-
following conclusions are drawn: mal comfort hours for occupants per
QQ Finite element heat flow analysis per- annum, with peak improvements of 83%
formed on a 3DPC wall section with and 80% during transition seasons. A
an appropriate iterative procedure to 14.1% improvement for the least com-
capture cavity radiation and convection fortable winter period was noted. This
was validated with physical tests on improvement relates directly to the
an instrumented 3DPC wall part. The energy-saving potential of 3DPC build-
measured and computed temperatures ings, since occupants would not need
were in reasonable agreement, with an to rely on active energy-consuming
average Pearson correlation coefficient heating and cooling devices to keep
of 0.83. themselves comfortable.
Volume 65 Number 3 September 2023 Journal of the South African Institution of Civil Engineering
This work contributes to the success of
3DPC construction and creates a frame-
work for continued research to further
improve the thermal comfort for occu-
pants. In a broader context, the research
brings understanding of how building
design may improve thermal comfort in
South African housing, and potentially
reduce energy consumption and associated
harmful emissions for heating and cooling.
REFERENCES
ASHRAE (American Society of Heating Refrigerating
and Air-Conditioning Engineers) 2020. ANSI/
ASHRAE Standard 55. Thermal Environmental
Conditions for Human Occupancy. Peachtree
Corners, GA: ASHRAE. Available at: www.ashrae.
org.
ASTM (American Society for Testing and Materials)
2014. ASTM C1363:2014. Standard Test Method
for Thermal Performance of Building Materials
and Envelope Assemblies by Means of a Hot Box
Apparatus. West Conshohocken, PA: ASTM.
Bergman, T L, Lavine, S A, Incropera, F P &
Dewitt, D P 2011. Fundamentals of Heat and
Mass Transfer (7th ed). Hoboken, NJ: Wiley.
BS (British Standard) 2019. BS EN 16798-1:2019 Energy
Performance of Buildings. Ventilation for Buildings.
London: British Standards Institution.
Christen, H E, Van Zijl, G P A G & De Villiers, W I
2023. Improving building thermal comfort
through passive design: An experimental analysis
of phase-change material 3D printed concrete.
Journal of the South African Institution of Civil Engineering
Journal for Cleaner Production, 392(5). https://doi.
org/10.1016/j.jclepro.2023.136247.
CIBSE (Chartered Institution of Building Services
Engineers) 2006. Guide A: Environmental Design
(7th ed). London: CIBSE.
Dassault Systemes Simulia 2020. ABAQUS. Johnston,
IO: Dassault Systemes. Available at: https://
www.3ds.com/products-services/simulia/
products/abaqus.
De Dear, R J & Brager, G S 1998. Developing an
adaptive model of thermal comfort and preference.
ASHRAE Transactions, 104(Pt 1A): 145–167.
DesignBuilder 2022. DesignBuilder Release Software.
Available at: https://DesignBuilder.co.uk/
download/release-software (accessed on 8 June
2022).
Energi Solutions @ Africa 2015. Temperature Sensors.
Available at: https://www.thermon.co.za/
catalogue/temperature-sensors (accessed on
7 June 2022).
Kruger, P J 2019. Rheo-mechanics modelling of 3D
concrete printing constructability. Stellenbosch
University. Available at: https://scholar.sun.ac.za.
Marais, H , Christen, H, Cho, S, Villiers, W I D &
Van Zijl G 2021. Computational assessment of
thermal performance of 3D printed concrete
wall structures with cavities. Journal of
Building Engineering, 41: 1–21. doi: 10.1016/j.
jobe.2021.102431.
Meijers, J 2021. Stellenbosch Weather. Stellenbosch
University. Available at: http://weather.sun.ac.za
(accessed on 30 September 2021).
Meyer, E S 2022. South African Building Envelope
Thermal Performance Simulation : Parameters
Volume 65 Number 3 September 2023
and the Role of Moisture Content. Stellenbosch
University. Available at: http://hdl.handle.
net/10019.1/124504.
NDHS (National Department of Human Settlements)
2013. Joe Slovo Phase 3 Low-Income Housing
Development: Sustainable Housing Delivery Lessons
Learnt Document. Cape Town: NDHS.
Ormandy, D & Ezratty, V 2012. Health and thermal
comfort: From WHO guidance to housing
strategies. Energy Policy, 49: 116–121. https://doi.
org/10.1016/j.enpol.2011.09.003.
RSA (Republic of South Africa) 1996. The Constitution
of the Republic of South Africa, 1996. Available at:
https://www.gov.za.
SANS (South African National Standard) 2011.
SANS 204:2011. Energy Efficiency in Buildings.
Pretoria: SABS Standards Division.
SANS 2012. SANS 10400-N:2012. The Application of
the National Building Regulations. Part N: Glazing.
Pretoria: SABS Standards Division.
SANS 2015. SANS 10400-K: 2015. The Application of
the National Building Regulations. Part K: Walls.
Pretoria: SABS Standards Division.
SANS 2021. SANS 10400-XA:2021. Energy Usage
In Buildings (2nd ed). Pretoria: SABS Standards
Division. Available at: https://store.sabs.co.za.
Statistics South Africa 2016. Community Survey,
2016. Pretoria: StatsSA. Available at: http://cs2016.
statssa.gov.za.
Sustainable Energy Africa 2014. Joe Slovo Phase 3:
Sustainable densification in well located areas,
Cape Town. Available at: http://sustainable.org.za/
sustainable.org.za/uploads/files/file79.pdf.
53
 Guidelines for the preparation of
papers and technical notes
INTRODUCTION
The Journal of the South African Institution of Civil Engineering,
which is a peer-reviewed, open-access journal, is published quar-
terly in March, June, September and December. Articles submitted
for publication are reviewed by a panel of referees under the guid-
ance of the SAICE Journal Editorial Panel. The journal publishes
research papers covering all the disciplines of civil engineering
(structural, geotechnical, railway, coastal/marine, water, construc-
tion, 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 arti-
cles for publication, authors should please take note of, and comply
with, the guidelines as set out below.
CLASSIFICATION OF ARTICLES
CONSIDERED FOR PUBLICATION
QQ Technical papers are well-researched, in-depth, fully refer-
enced 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 upliftment, etc) are accepted if they are of a techni-
cal nature and of particular interest to the civil engineering
profession. The latter type of paper will be subject not only
to peer-review by civil engineers, but also to review by non-
engineering specialists in the field covered by the paper.
QQ Technical notes are short, fully referenced technical articles
that do not exceed 2 000 words. A typical technical note will
have limited scope often dealing with a single technical issue
of particular importance to civil engineering.
QQ Review papers are considered for publication as either techni-
cal papers or technical notes on condition 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 must contribute by synthesis-
ing the information and drawing new conclusions from the
dissemination of the previously published work.
QQ Discussions on published articles are welcomed up to six
months after publication. The length of discussion contribu-
tions is limited to 1 500 words. Where appropriate, discussion
contributions will be subject to the normal reviewing process
and will be forwarded to the authors of the original article
for reply.
POLICY REGARDING LANGUAGE AND
ORIGINALITY OF SUBMITTED ARTICLES
QQ Language: Manuscripts should be presented in English, as the
journal is distributed internationally.
QQ Original work: Papers and technical notes must be original
contributions. Authors of JSAICE articles must affirm that
54
submitted 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 publication of any paper in the journal will not
constitute a breach of any agreement or the transgression of
any law. The corresponding author should confirm that all
co-authors have read and approved the manuscript and accept
these conditions. Authors are responsible for obtaining permis-
sion to publish experimental data and other information that
may be confidential or sensitive. Authors are also responsible
for obtaining written permission from copyright owners (usu-
ally the publishers) when reproducing material that had been
published elsewhere. If such material had been published under
a Creative Commons Licence, appropriate recognition must be
included in the submitted manuscript.
SUBMISSION PROCEDURES AND
REQUIRED FORMAT
QQ Online submission: Manuscripts must be uploaded as PDF
files (http://journal.saice.org.za). Individual file sizes may
not exceed 10 MB. Should you experience problems when
uploading your paper, please contact the administrative editor
(verelene@saice.org.za).
QQ Format: Manuscripts should be prepared in MS Word and
presented in double-line spacing, single-column layout with
25 mm wide margins. Line numbers must be applied to the
whole document. All pages should bear the authors’ names and
be numbered at the bottom of the page. With the exception of
tables and figures (see below) the document should be typed in
Times New Roman 12 pt font. Contributions should be accom-
panied by an abstract of not more than 200 words.
QQ First page: The first page of the manuscript should include
the title of the paper, the number of words of the main text
(i.e. excluding figures, tables and the list of references), the
initials, titles and surnames of the authors, professional status
(if applicable), SAICE affiliation (Member, Fellow, Visitor,
etc), ­telephone numbers (landline and mobile), and e-mail
and postal addresses. The name of the corresponding author
should be underlined. Five keywords should be suggested.
QQ Figures, tables, photos and illustrations:
QQ These should preferably be submitted in colour, as the jour-
nal is a full-colour publication.
QQ Their positions should be clearly marked in the text as fol-
lows: [Insert Figure 1]
QQ 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.
QQ Illustrations must be accompanied by appropriate captions.
Captions for tables should appear above the table. All other
captions should appear below the illustration (figures,
graphs, photos).
Volume 65 Number 3 September 2023 Journal of the South African Institution of Civil Engineering
 QQ Only those figures and photographs essential to the under-
standing of the text should be included. All illustrations
should be referred to in the text.
QQ Figures should be produced using computer graphics. Hand-
drafted figures will not be accepted. 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.
QQ Tables should be typed in Times New Roman 9 pt font. They
should not duplicate information already given in the text,
nor contain material that would be better presented graphi-
cally. Tabular matter should be as simple as possible, with
brief column headings and a minimum number of columns.
QQ Mathematical expressions and presentation of symbols:
QQ Equations should be presented in a clear form which can
easily be read by non-mathematicians. Each equation
should appear on a separate line and should be numbered
consecutively.
QQ Symbols should preferably reflect those used in Microsoft
Word Equation Editor or Mathtype, or should be typed using
the Times New Roman symbol set.
QQ Variables in equations (x, y, z, etc, as well as lower-case Greek
letters) should be presented in italics. Numbers (digits),
upper-case Greek letters, symbols of metric measurement
units (m for metres, s for seconds, etc) and mathematical/
trigonometrical functions (such as sin, cos and tan) are not
written in italics, but in upright type (Roman). Variables
and symbols used in the body of the text should match the
format used in the equations, i.e. upright or italics, whichever
is applicable.
QQ Metric measurement abbreviations/units should conform to
international usage – the SI system of units should be used.
QQ Decimal commas may be used, but decimal points are pre-
ferred in accordance with accepted international usage.
QQ Symbols should preferably be defined in the text, but if this is
not feasible, a list of notations may be provided for inclusion
at the end of the paper.
QQ Headings: Sections and paragraphs should not be numbered.
The following hierarchy of headings should be followed:
HEADING OF MAIN SECTION
Heading of subsection
Heading of sub-subsection
QQ References: References should follow the Harvard system. The
format of text citations should be as follows: “Jones (1999) dis-
covered 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
Journal of the South African Institution of Civil Engineering Volume 65 Number 3 September 2023
should be in chronological order. The following are examples of
a journal article, a book and a conference paper:
QQ Donald, A W 1954. Pore pressure coefficients.
Géotechnique, 4(4): 143–147.
QQ Terzaghi, K 1943. Practical soil mechanics. New York: Wiley.
QQ Smith, A W, Black, C & White, A 1999. Factors affecting the
strength of soils. Proceedings, ASCE Conference on Shear
Strength of Soils, Colorado, pp 503–532.
Papers published previously in the Journal of the South African
Institution of Civil Engineering should be cited if applicable.
QQ Footnotes, trade names, acronyms, abbreviations: These
should be avoided. If acronyms are used, they should be defined
when they first appear in the text. Do not use full stops after
abbreviations or acronyms.
QQ Return of amended papers: Papers requiring amendments will
be accepted up to six months after the referee reports had been
sent to authors, after which time the paper will be withdrawn
from the system.
FINAL ARTICLE
QQ Proofs: First proofs of accepted papers will be sent to authors
in PDF format for verification before publication. No major
re-writes will be allowed at this stage, only essential minor
corrections.
QQ Photos of authors: The final corrected version of the paper
should be accompanied by recent, high-resolution head and
shoulders colour photographs and a profile not exceeding 100
words for each of the authors.
QQ Copyright: Copyright remains with the authors. As a peer-
reviewed, open-access journal licensed under a Creative
Commons Attribution Licence (CC BY-NC-ND) (https://
creativecommons.org/licenses/by-nc-nd/4.0/) the content
of this journal is available to users free of charge. Users are
therefore allowed to copy and distribute the material in any
medium or format, but such material may only be used in
unadapted form and for non-commercial purposes, and due
recognition must be given to the authors, as well as to the
Journal of the South African Institution of Civil Engineering as
the original publisher.
QQ Repository policy (depositing of articles by authors after
publication): After publication a PDF of their published article
is sent to authors for their records. Authors may then upload the
published article to their work/institution website, or any other
repository of their choice, such as, for example, ResearchGate
and/or similar. Only the published version of the paper may be
used/uploaded in this manner, i.e. not the originally submitted
or the accepted/pre-published version.
55