Journal Vol 60 4 2018 December

Volume 60 Number 4
of the South African Institution of Civil Engineering December 2018
■■ tudy of the Durban Bight shoreline evolution under

S
schematised climate change and sand‑bypassing scenarios
■■ Risk-based member reliability in structural design
■■ he relationship between project performance of emerging
T
contractors in government infrastructure projects
and their experience and technical qualifications
■■ he influence of concrete compressive strength
T
and specimen size on the compression stress
block parameters of reinforced concrete
■■ E ffect of chlorine and chloramine disinfection and the
presence of phosphorous and nitrogen on biofilm growth
in dead zones on PVC pipes in drinking water systems
■■ atchment response time and design rainfall:
C
the key input parameters for design flood
estimation in ungauged catchments
■■ iscussion: Comparative analysis of methods of pile-bearing
D
capacity evaluation using CPT logs from tropical soils
Guidelines for the preparation of papers and technical notes
Authors should comply with the following guidelines when preparing papers for publication in the journal
The Journal of the South African Institution of Civil permission to publish experimental data and other or should be typed using the Times New Roman
Engineering is published quarterly in March, June, information that may be confidential or sensitive. symbol set.
September and December. Articles submitted for Authors are also responsible for obtaining permis- ■■ Variables in equations (x, y, z, etc, as well as
publication are reviewed by a panel of referees under sion from copyright owners when reproducing mate- lower case Greek letters) should be presented
the guidance of the SAICE Journal Editorial Panel. rial that had been published elsewhere. Proof of such in italics. Numbers (digits), upper case Greek
The journal publishes research papers covering all the permission must be supplied. letters, symbols of metric measurement units
disciplines of civil engineering (structural, geotechnical, (m for metres, s for seconds, etc) and mathemati-
railway, coastal/marine, water, construction, environ- SUBMISSION PROCEDURES cal/trigonometrical functions (such as sin, cos
mental, municipal, transportation) and associated topics AND REQUIRED FORMAT and tan) are not written in italics, but in upright
that are relevant to the civil engineering profession, and ■■ Online submission: Manuscripts must be uploaded type (Roman). Variables and symbols used in the
that preferably have particular relevance to civil engi- as PDF files (http://journal.saice.org.za). Individual body of the text should match the format used
neering in southern Africa and the African continent. file sizes may not exceed 4 MB. Should you experi- in the equations, i.e. upright or italics, whichever
When preparing articles for publication, authors ence problems uploading your paper, please contact is applicable.
should please take note of the following and comply the editor (verelene@saice.org.za). ■■ Metric measurement abbreviations/units should
with the guidelines as set out: ■■ Format: Manuscripts should be prepared in MS conform to international usage – the SI system of
Word and presented in double line spacing, single units should be used.
CLASSIFICATION OF ARTICLES column layout with 25 mm wide margins. Line ■■ Decimal commas may be used, but decimal
CONSIDERED FOR PUBLICATION numbers must be applied to the whole document. points are preferred.
■■ Technical papers are well-researched, in-depth, All pages should bear the authors’ names and be ■■ Symbols should preferably be defined in the text,
fully referenced technical articles not exceeding numbered at the bottom of the page. With the excep- but if this is not feasible, a list of notations may be
6 000 words in length (excluding tables, illustration of tables and figures (see below) the document provided for inclusion at the end of the paper.
tions and the list of references). Related papers that should be typed in Times New Roman 12 pt font. ■■ Headings: Sections and paragraphs should not be
deal with ‘softer sciences’ (e.g. education, social Contributions should be accompanied by an abstract numbered. The following hierarchy of headings
upliftment, etc) are accepted if they are of a techni- of not more than 200 words. should be followed:
cal nature and of particular interest to the civil ■■ First page: The first page of the manuscript should HEADING OF MAIN SECTION
engineering profession. The latter type of paper will include the title of the paper, the number of words of Heading of subsection
be subject not only to peer-review by civil engineers, the main text (i.e. excluding figures, tables and the Heading of sub-subsection
but also to review by non-engineering specialists in list of references), the initials and surnames of the ■■ References: References should follow the Harvard
the field covered by the paper. authors, professional status (if applicable), SAICE system. The format of text citations should be as
■■ Technical notes are short, fully referenced technical affiliation (Member, Fellow, Visitor, etc), telephone follows: “Jones (1999) discovered that …” or “recent
articles that do not exceed 2 000 words. A typical numbers (landline and mobile), and e-mail and results (Brown & Carter 1985; Green et al 1999)
technical note will have limited scope often dealing postal addresses. The name of the corresponding indicated that …”
with a single technical issue of particular importance author should be underlined. Five keywords should References cited in the text should be listed in
to civil engineering. be suggested. alphabetical order at the end of the paper. References
■■ Review papers are considered for publication as ■■ Figures, tables, photos and illustrations: These by the same author should be in chronological order.
either technical papers or technical notes on condi- should preferably be submitted in colour, as the The following are examples of a journal article, a
tion that they are the original work of the author journal is a full-colour publication. book and a conference paper:
and will assist the reader with the understanding, ■■ Their positions should be clearly marked in the ■■ Donald, A W 1954. Pore pressure coefficients.
interpreting or applying of the subject under review. text as follows: [Insert Figure 1]. Géotechnique, 4(4):143–147.
A review paper must contain criteria by which the ■■ Figures, tables, photos, illustrations and equa- ■■ Terzaghi, K 1943. Practical soil mechanics. New
work under review was evaluated, and contribute tions should be numbered consecutively and York: Wiley.
by synthesising the information and drawing new should appear in the text directly after they have ■■ Smith, A W, Black, C & White, A 1999. Factors
conclusions from the dissemination of the previously been referred to for the first time. affecting the strength of soils. Proceedings, ASCE
published work. ■■ Illustrations must be accompanied by appropriate Conference on Shear Strength of Soils, Colorado,
■■ Discussion on published articles is welcomed up to captions. Captions for tables should appear above pp 503–532.
six months after publication. The length of discus- the table. All other captions should appear below Papers published previously in the Journal of the
sion contributions is limited to 1 500 words. Where the illustration (figures, graphs, photos). South African Institution of Civil Engineering should
appropriate, discussion contributions will be subject ■■ Only those figures and photographs essential to be cited if applicable.
to the normal reviewing process and will be for- the understanding of the text should be included. ■■ Footnotes, trade names, acronyms, abbreviations:
warded to the authors of the original article for reply. All illustrations should be referred to in the text. These should be avoided. If acronyms are used, they
■■ Figures should be produced using com- should be defined when they first appear in the text.
POLICY REGARDING LANGUAGE AND puter graphics. Hand-drafted figures will not be Do not use full stops after abbreviations or acronyms.
ORIGINALITY OF SUBMITTED ARTICLES accepted. Lettering on figures should be equiva- ■■ Return of amended papers: Papers requiring
■■ Language: Manuscripts should preferably be lent to a Times New Roman 9 pt font or slightly amendments will be accepted up to six months after
presented in English, as the journal is distributed larger (up to 12 pt) if desired. Lettering smaller the referee reports had been sent to authors, after
internationally. Articles submitted in any of the than 9 pt is not acceptable. which the paper will be withdrawn from the system.
other official South African languages should be ■■ Tables should be typed in Times New Roman
accompanied by an expanded abstract in English. 9 pt font. They should not duplicate information FINAL ARTICLE
■■ Original work: Papers and technical notes must already given in the text, nor contain material ■■ Copyright: On acceptance of the paper or technical
be original contributions. Authors must affirm that would be better presented graphically. note, copyright must be transferred by the author/s
that submitted material has not been published Tabular matter should be as simple as possible, to the South African Institution of Civil Engineering
previously, is not under consideration for publica- with brief column headings and a minimum on the form that will be provided by the Institution.
tion elsewhere and will not be submitted elsewhere number of columns. ■■ Photos of authors: The final corrected version of
while under consideration by the SAICE Journal ■■ Mathematical expressions and presentation the paper should be accompanied by recent, high-
Editorial Panel. It is the responsibility of the authors of symbols: resolution head and shoulders colour photographs
to ensure that publication of any paper in the journal ■■ Equations should be presented in a clear form and a profile not exceeding 100 words for each of
will not constitute a breach of any agreement or which can easily be read by non-mathematicians. the authors.
the transgression of any law. The corresponding Each equation should appear on a separate line ■■ Proofs: First proofs of papers will be sent to authors
author should confirm that all co-authors have and should be numbered consecutively. in PDF format for verification before publication.
read and approved the manuscript and accept these ■■ Symbols should preferably reflect those used in No major re-writes will be allowed, only essential
conditions. Authors are responsible for obtaining Microsoft Word Equation Editor or Mathtype, minor corrections.
Volume 60 No 4 December 2018 ISSN 1021-2019
Publisher
South African Institution of Civil Engineering
Block 19, Thornhill Office Park,
Bekker Street, Vorna Valley, Midrand, South Africa
Private Bag X200, Halfway House, 1685, South Africa
Tel +27 (0)11 805 5947/48, Fax +27 (0)11 805 5971
http://www.saice.org.za
civilinfo@saice.org.za
Contents
Editor-in-chief
Prof Gerhard Heymann
University of Pretoria
Tel +27 (0)12 420 3627
gerhard.heymann@up.ac.za
joint Editor-in-chief
Prof Chris Clayton 2 Study of the Durban Bight shoreline
University of Southampton
c.clayton@soton.ac.uk
evolution under schematised climate
MANAGING Editor
change and sand‑bypassing scenarios
Verelene de Koker C Rautenbach, A K Theron DOI: 10.17159/2309-8775/2018/v60n4a1
Tel +27 (0)11 805 5947, Cell +27 (0)83 378 3996
verelene@saice.org.za
journal editorial Panel
16 Risk-based member reliability in structural design
Prof G Heymann – University of Pretoria N de Koker, A A Elvin DOI: 10.17159/2309-8775/2018/v60n4a2
Prof CRI Clayton – University of Southampton
Prof Y Ballim – University of the Witwatersrand
Prof W Burdzik – University of Pretoria 25 The relationship between project performance of
Dr P Day – Jones & Wagener (Pty) Ltd
Prof J du Plessis – University of Stellenbosch emerging contractors in government infrastructure
Prof GC Fanourakis – University of Johannesburg
Prof M Gohnert – University of the Witwatersrand
projects and their experience and technical qualifications
Prof PJ Gräbe – University of Pretoria F T Muzondo, R T McCutcheon DOI: 10.17159/2309-8775/2018/v60n4a3
Dr C Herold – Umfula Wempilo Consulting
Prof A Ilemobade – University of the Witwatersrand
Prof SW Jacobsz – University of Pretoria 34 The influence of concrete compressive strength
Prof EP Kearsley – University of Pretoria
Prof C Roth – University of Pretoria and specimen size on the compression stress
Prof W Steyn – University of Pretoria block parameters of reinforced concrete
Mr M van Dijk – University of Pretoria
Prof C Venter – University of Pretoria F van Schalkwyk, E P Kearsley DOI: 10.17159/2309-8775/2018/v60n4a4
Prof A Visser – University of Pretoria
Dr E Vorster – Aurecon South Africa (Pty) Ltd
Prof J Wium – University of Stellenbosch 45 Effect of chlorine and chloramine disinfection
Prof A Zingoni – University of Cape Town
Prof M Zuidgeest – University of Cape Town
and the presence of phosphorous and
Peer reviewing
nitrogen on biofilm growth in dead zones
The Journal of the South African Institution of on PVC pipes in drinking water systems
Civil Engineering is a peer-reviewed journal
that is distributed internationally
U Kadwa, M V Kumarasamy, D Stretch DOI: 10.17159/2309-8775/2018/v60n4a5
Design and reproduction

Marketing Support Services, Ashlea Gardens, Pretoria 51 Catchment response time and design rainfall:
Printing the key input parameters for design flood
Fishwicks, Pretoria
estimation in ungauged catchments
Papers for consideration should be submitted
online at: http://journal.saice.org.za O J Gericke DOI: 10.17159/2309-8775/2018/v60n4a6
The South African Institution of Civil Engineering accepts no
responsibility for any statement made or opinion expressed
in this publication. Consequently, nobody connected with
68 Discussion: Comparative analysis of
the publication of this journal, in particular the proprietor, methods of pile-bearing capacity evaluation
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
using CPT logs from tropical soils
upon any statement or opinion published in this journal. I N Obeta, M E Onyia, D A Obiekwe DOI: 10.17159/2309-8775/2018/v60n4a7
© South African Institution of Civil Engineering 1

TECHNICAL PAPER Study of the Durban Bight
Journal of the South African
Institution of Civil Engineering shoreline evolution under
ISSN 1021-2019
Vol 60 No 4, December 2018, Pages 2–15, Paper 0273 schematised climate change
DR CHRISTO RAUTENBACH obtained his PhD
at the University College of Southeast
and sand‑bypassing scenarios
Norway in 2012 and his MSc cum laude from
the Applied Mathematics Department at
Stellenbosch University. Immediately after C Rautenbach, A K Theron
completion of his PhD he started working at
the Coastal Systems research group at the
Council for Scientific and Industrial Research
(CSIR) as a numerical modeller. Ever since he has had both private The Durban Bight beaches, on the KwaZulu-Natal coast of South Africa, have been the topic of
engineering consulting and public research experience, with emphasis on several studies in the past, because of their recreational and economic value. These beaches are
ocean and coastal waves, shoreline morphodynamics, water quality, mainly nourished via a sand-pumping scheme, which has been the case since the 1980s. The
integrated coastal management, coastal and estuarine hydrodynamics
present study investigated the longer-term dynamic stability of the Durban Bight beaches using
and marine forecasting. He currently works at the South African Weather
Service as the Chief Marine Scientist, with a research, project a one-dimensional numerical shoreline model. The wave conditions that drive the longshore
management and product development role. He also holds an Honorary transport of sand were simulated using the numerical spectral wave model Simulation WAves
Research Associate position at the Marine Research Unit at the University in the Nearshore (SWAN). The shoreline model was calibrated and validated against existing
of Cape Town, where he is an active partner in postgraduate lecturing and
student supervision.
historical data. The calibrated model was then employed to simulate possible future scenarios.
These included climate change phenomena and the influence of anthropogenic changes. Each
Contact details:
Marine Research Unit
of the simulations predicts a twenty-year period, with results extracted at five-year intervals.
South African Weather Service The results are presented and discussed, and the emphasis is placed on the importance of
PO Box 2981 maintaining the required sand-bypassing rates and the awareness of the possible effects of
Paarl 7620
climate change on shoreline maintenance and management.
South Africa
T: +27 21 935 5700
E: christo.rautenbach@weathersa.co.za
INTRODUCTION (Mather & Stretch 2012; Theron 2007;
DR ANDRÉ THERON is a senior lecturer in Mather et al 2003; Barnett 1999).
port, coastal and water engineering at Background One of the most significant alterations
Stellenbosch University. He holds a PhD and
MEng from the same university, and has
The Durban Bight, on the east coast to the natural beaches in the Durban Bight
specialised in port and coastal engineering of South Africa in the KwaZulu-Natal is the sand-bypassing scheme located at
for the past 27 years. His fields of expertise Province, has been the topic of numerous the harbour entrance. This scheme was
are nearshore hydrodynamics, sediment
studies in the past (Mather et al 2003; implemented to compensate for the loss
transport, dredging, coastal erosion,
planning and design of port/coastal developments, coastal protection and Cawthra et al 2012; Campbell et al 1985; of sediment transported from the south
flood defence, estuarine hydro- and sediment-dynamics, and coastal Soltau & Smith 2003; Mather & Stretch to the north, due to the construction of
physical environmental studies. He started his career at the Department 2012; Green et al 2012; Habets 2015). the Durban Harbour southern breakwater.
of Water Affairs, then moved to the Council for Scientific and Industrial
Durban Harbour is one of South Africa’s The longshore sediment transport from
Research (CSIR), and joined Stellenbosch University in 2015.
and East Africa’s largest harbours, which the south was cut off, resulting in severe
Contact details:
mainly functions as import and export erosion of the Bight beaches located to the
Department Civil Engineering
Stellenbosch University harbour for containers, oil and food north of the harbour entrance, prior to the
Private Bag X1 (Rossouw & Theron 2012). Import and 1980s. Currently the beaches are mainly
Matieland 7602 nourished via the sand-bypassing scheme
export container throughput per annum is
South Africa
T: +27 21 808 4353 estimated at 41.9 Megatons (2.48 Million consisting of a sand trap (just south of
E: aktheron@sun.ac.za Twenty-foot Equivalent Units (MTEU)) the breakwater) and a dredger that pumps
(Rossouw & Theron 2012). Durban sand into a land-bound hopper system that
Harbour is the largest non-bulk terminal transports sediment in a fluidised state to
in South Africa, and up to 2010 was also the Bight beaches. The sand-pumping pipe-
the second-busiest port in the southern lines consist of approximately 3.5 km of
hemisphere (eThekwini Municipality conduits (Mather et al 2003). A schematic
2010). In close proximity to the harbour, illustration of this scheme is presented
Durban also boasts some of South Africa’s in Figure 1. With historical data of the
most well-known recreational beaches, sand-pumping volumes, this anthropogenic
which attract both national and inter- influence on the Durban Bight beaches can
national tourists. Due to this hotspot of be studied and modelled.
coastal activity, the Durban Bight forms The focus of the present work is on
Keywords: long-term shoreline modelling, sand-bypassing scheme, an ideal study location for various aspects the long-term shoreline dynamics of the
climate change, coastal management of natural and anthropogenic changes Durban Bight associated with:
Rautenbach C, Theron AK. Study of the Durban Bight shoreline evolution under schematised climate change and sand-bypassing scenarios.
2 J. S. Afr. Inst. Civ. Eng. 2018:60(4), Art. #0273, 14 pages. http://dx.doi.org/10.17159/2309-8775/2018/v60n4a1
■■ global climate change
Legend ■■ the possible construction of a new
dig-out harbour south of the existing
Longshore transport
harbour (Habets 2015)
Sand pumping lines ■■ changes in sand-bypassing volumes and
sediment characteristics, and
Sand trap
■■ the effect of dam building in rivers that
mouth south of the harbour.
These factors can influence the quality
of the Durban beaches significantly, and
investigating these scenarios will contribute
to effective planning and coastal hazard
assessment by the eThekwini Municipality
(greater Durban area) (Mather & Stretch
2012). In Figure 2 the layout of the greater
Durban Bight is illustrated, together with
the cross-shore measurement beacon
locations utilised in the present study for
calibration and validation. Relatively limited
studies have been performed regarding
climate change in southern Africa and the
associated vulnerability and mitigation
options, with the exception of some stud-
ies in South Africa (Theron 2007; Mather
& Stretch 2012; Theron et al 2012). The
present study will supplement existing
Figure 1 I mage of Durban Harbour, indicating the locations of the sand trap, harbour entrance studies to quantify the influence of certain
and Durban Bight, with (inset) a map of southern Africa showing the location of Durban future scenarios on the long-term shoreline
dynamics in the Durban area, with the aim
to inform local coastal management.
Legend
With knowledge of such potential
Cross-shore beach
measurement changes, the eThekwini Municipality will
beacons be able to investigate ways to manage
and mitigate potential impacts of wider
or narrower beaches, such as blockage
of stormwater outlets, wind-blown sand
problems, reduced available beach space for
recreational activities and possible effects
on the Umgeni River mouth.
Aim
The aim of this study is to use a long-term
shoreline model to provide updated pre-
dictions of the potential future shoreline
B3 changes resulting from four future scenar-
ios. To this effect the following approaches
were followed:
■■ Employ the Simulation WAves in the
Nearshore (SWAN) wave propagation
B2 model to transform offshore wave con-
Bell ditions to the nearshore.
Street ■■ Calibrate a Unibest (Deltares 2005)
B1 shoreline model, employing 200 to 400

representative wave conditions.
Short ■■ Use a model domain extending up to
line
Umhlanga Rocks (Beacon NC7) in the
Figure 2 L ayout of the greater Durban Bight indicating the beach surveying beacon locations and north.
their names; this illustration also indicates the extent of the numerical model employed ■■ Include in the shoreline model the sedi-
for the current study (Vetch’s Reef to beacon NC7) and the sand-bypassing outlet ment inputs from the Umgeni River and
locations in the zoomed-in view around Vetch’s Reef the beach nourishment scheme.
Journal of the South African Institution of Civil Engineering Volume 60 Number 4 December 2018 3
The model will also be employed to investi-
gate possible effects on the shoreline due to: Summer – 63 831 records Autumn – 62 145 records
■■ Climate change North North
■■ The construction of a new port
60% 60%
■■ The construction of dams in the rivers
45% 45%
feeding the central KwaZulu-Natal coast
30% 30%
■■ Changes in the particle size distribution
15% 15%
(PSD) of the dredged sand.
To be able to predict the shoreline dyna West East West East
mics accurately the representative wave
conditions had to be analysed first. This
includes the setting up of the wave trans-
(m)
formation model, analysis of wave data and 3.0–3.5
determination of the conditions that are South 2.5–3.0 South
2.0–2.5
most representative for the Durban Bight. 1.5–2.0
Spring – 62 192 records Winter – 61 725 records
The predictions of potential future shore- 1.0–1.5
North 0.5–1.0 North
line responses to the future scenarios are 0–0.5
provided and discussed. Conclusions are
60% 60%
made and recommendations are given. 45% 45%
30% 30%
15% 15%
SHORELINE MODEL SET UP
West East West East
Model description
The Unibest CL+ numerical shoreline
model (Deltares 2005) was employed in
this study to simulate shoreline evolution
based on wave-driven longshore transport. South South
The model incorporates two sub-models
that run in tandem: In the Unibest LT sub- Figure 3 S easonal wave roses for the directional wave rider buoy records at Richards Bay (at
model, linear nearshore wave refraction, 28.8265 S, 32.104 E), in 23 m water depth; these roses represent all the data from
shoaling and breaking are predicted. The 8 November 2002 to 23 October 2013
resulting longshore current and sediment
transport are then calculated. This infor- Three-dimensional effects are not included These wave data sources were evalu-
mation serves as input (at several locations in the model formulation and therefore ated and transformed from their in-situ
along the shoreline) to the Unibest CL the cross-shore erosion/accretion, due to location to the offshore model boundary
sub-model, with which one-dimensional e.g. storms (episodic erosion), is not repre- location. Transformation tables were used
shoreline evolution is simulated. The latter sented in the model process. In a shoreline to avoid simulating recurring conditions.
sub-model allows for the incorporation model, such factors are assumed to average Datasets were transformed back and forth
of varying sediment sources, such as the out over the long time periods (Hugo 2013). from these measurement locations to the
sand-feeding along the Durban beaches The model is thus appropriately used offshore model boundary to assure the
from natural and anthropogenic sources. to simulate structural erosion/accretion accuracy of the transformation tables. A
The shoreline model functions based and not episodic erosion associated with near 1:1 comparison was obtained between
on single-line theory, which assumes the cross-shore movement of sediments the twice-transformed wave conditions and
the beach and nearshore profile to be (Bosboom & Stive 2015). the original measurement datasets.
represented by a cross-shore profile shape After transforming several wave datasets
that does not change in time (Bosboom Wave input to the offshore model boundary, as a virtual
& Stive 2015). As the shoreline erodes or During the shoreline model calibration time series, it was found that the most
accretes, the entire equilibrium cross-shore period (1990 to 2006) waves were recorded representative transport rates are provided
profile moves either landward or seaward off Durban by means of a Waverider buoy by the data from the directional wave buoys
(Bosboom & Stive 2015). The model is ideal (1992–2001) and an Acoustic Doppler at the Port of Richards Bay. In Figure 3 the
for medium- to long-term predictions of Current Profiler (ADCP 2002–2007) seasonal directional wave roses at Richards
shoreline evolution. Generally, the model is (Rossouw et al 1999). However, the older Bay are given. The predominant wave
run with a climate of representative wave wave rider recordings off the Durban direction is from the south-southeast with
conditions. The model can also simulate Bluff lacked wave directions, while the a directional spreading difference between
the temporal response of the shoreline shallow ADCP wave data suffered from the various seasons. Summer was the only
to varying wave conditions. For periods intermittent coverage and accuracy issues season presenting a stronger east-southeast
shorter than seasonal scale, observed due to various challenges, for example wave directional component.
shoreline behaviour is dominated by on/ episodic sand inundation of the instrument In the Bight an approximate annual
offshore processes (storms/calm periods). (Mather & Theron 2011). sediment transport rate of 265 000 m3 can
4 Volume 60 Number 4 December 2018 Journal of the South African Institution of Civil Engineering
be expected 305 000 m3 further north of of approximately 16 km (refer to Figure 1). therefore extracted from the bathymetry.
the Umgeni River mouth (Theron 2015; The coastline between Vetch’s Reef and Due to some uniformity in the nearshore
Wells 2015) and the net north-eastward the North Breakwater of the port was not bathymetry, the profiles were schematised
annual longshore sediment transport included in the model, as the focus of the to six typical profiles based on meas-
rate of about 500 000 m3 along the study is on beach accretion of the main urements provided by the eThekwini
Durban Bluff (Schoonees 2000; Theron & beaches (and secondarily because modelled Municipality.
Rautenbach 2014). Both resulting transport wave conditions are less accurate in this
directions and magnitudes compare well partial diffraction zone). Sediment size
to the known estimates, and the nearshore At the boundaries of the model the Limited samples are available of the mate-
refracted wave parameters were con- transport (or shoreline behaviour) needs rial dredged from the sand trap. From 2000
firmed with the nearshore ADCP dataset. to be specified. No sand enters the Vetch’s to 2004, samples from the sand-pumping
Transforming only local ADCP data from Bight area from the south, due to the inter- scheme hopper indicated a profusion of
the Bight failed in reproducing the wave vening harbour breakwaters and deeper coarse material. This was in contradiction
conditions offshore (and thus nearshore dredged entrance channel. At the southern to previous samples (1999), which indicated
transport rates) for the greater extent of model boundary, a variable input of sedi- an average of 253 µm, once outliers were
the shoreline, due to a dominant easterly ment was used, this being the combined removed. Earlier sampling indicated sizes
wave component. The peak wave period, historical discharge rates, from the so-called between 200 and 250 µm (Campbell et al
significant wave height and directional Short Line and Long Line, which outlets 1985). The long-term averages, of dredging
time series correspond to the ADCP data formed part of the beach nourishment samples collected near-annually by eThe-
less than a kilometre away from the tip of scheme. Both discharge close to Vetch’s kwini Municipality, indicate D50 values
the Durban southern breakwater, in 17 m Reef, in the corner of the Bight, close to the in the range between 250 µm and 300 µm
water depth (the data was provided by the harbour entrance. This was assumed to be (Theron et al 2013). A value of 250 µm was
local port authorities) (Diedericks et al a reasonable discretisation, as the Vetch’s thus applied as an approximation in the
2011). This data was therefore utilised in Bight cell appears to be neither a substantial shoreline model calibration, validation and
the present study. The Richards Bay wave sink nor a source of sand (CSIR 2003; baseline scenario. For the future scenarios
buoy was in a water depth of 23 m. Theron & Rautenbach 2014). Discharged two extreme cases were investigated – one
Wave conditions were extracted from sand, e.g. into the Bight by means of the with D50 a third less than the baseline
the 437 refraction simulation outputs, just Short Line, is therefore likely to exit north- scenario (166 µm), and one with D50 a third
seaward (between 8 m and 6 m, depending ward towards the Durban beaches within a more than the baseline scenario (333 µm).
on wave exposure) of wave breaking. The reasonably short period of time.
SWAN model consisted of three compu- The northern boundary of the model Model effective depth
tational grids. The parent domain had a was located at Umhlanga Rocks. At this The model effective depth parameter,
resolution of 1 km, the intermediate grid point the rocky coastline forms a rocky or active height of the profile, describes
resolution was approximately 150 m and headland protruding from the sandy beach the height of the mobile sand layer that
the final coastal nested grid resolution was that lies to the south, forming a non- responds to changes in shoreline orienta-
approximately 20 m. The coastal nested erodible coast with an embayment to the tion. Based on previous modelling inves-
domain was a curvilinear grid, while the north. Northward transport can thus read- tigations at the Council for Scientific and
other two domains were rectangular. The ily occur while southward return transport Industrial Research (CSIR 2007), a variable
parent domain covered approximately is restricted by the embayment and head- depth was used, ranging from 10 m in the
150 km along the coastline and the offshore land, as well as by the local coastal orien north to 6 m in the more sheltered extreme
boundary of the parent grid was approxi- tation relative to the incident wave direc- south. During the predictions of future
mately 60 km away from the coastline. tions resulting in virtually unidirectional shoreline evolution, this parameter was
To ensure that alongshore changes in longshore currents and transport towards increased to account for areas where mate-
wave conditions are represented throughout the north. This situation is reproduced by rial would be deposited into deeper water
the Bight, waves were input to the shoreline schematising the model boundary with if the shoreline accreted substantially. It
model at 48 locations. Approximately half a short groyne at Umhlanga Rocks and a should be noted that the model effective
of these were located north of the Umgeni fixed northward transport on its northern depth has virtually no influence on the
River and half south of this point (refer to side (implying no southward transport configuration of the long-term equilibrium
Figure 2). This gave greater computational across the groyne). Transport rates south shoreline, only the rate at which shoreline
density in the area of greater wave and of the boundary are influenced only by the equilibrium is attained in response to a
bathymetric variability, with lower density transports and shoreline behaviour occur- steady rate of sand discharge.
along the northern part where the shore- ring in the remainder of the model domain.
line and bathymetry are fairly uniform. These determine the actual transport past Sediment input from the Umgeni River
Transport conditions were interpolated the groyne and thus the model boundary. Data on the amount of sediment dis-
between locations in the shoreline model. charged to the sea by the Umgeni River is
Cross-shore profiles poor. However, the present typical annual
Boundaries Beach and nearshore profiles are required input is very low, as there are limited
The modelled domain extends from the in the shoreline model to calculate wave sources of sandy material below the Inanda
present Vetch’s Reef (or pier) in the south transformation and longshore transport Dam (Garland & Moleko 2000). Sediment
to Umhlanga Rocks in the north, a distance processes. The nearshore profiles were inputs occur mostly during episodic
events. It has been estimated that the last
extreme flood (1987) discharged between 500 000
720 000 m3 and 1.24 million m3 of sand.
It is assumed that the present annual 400 000
discharge is unlikely to be more than a few
Annual rate (m3)

percent of this, due to five large dams trap-
300 000
ping virtually all of the sediments (Theron
et al 2008).
Longshore transport rates calcu- 200 000
lated north and south of the Umgeni
area indicate a difference in the order 100 000
of 70 000 m3/a (Soltau & Theron 2007).
However, shoreline changes (net erosion)
north and south of the mouth suggest that 0
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
at present this magnitude of material is not
being supplied from river inputs (i.e. the Time (years)
difference in transport is being supplied Annual Average 265 000
by beach erosion). Tests with the shoreline
model indicate that it is fairly insensitive Figure 4 L ongterm total sand-pumping rate in cubic metres (Theron & Rautenbach 2014)
to the input rate. The calibration suggested
that a sediment input in the order of
20 000 m3/a was appropriate. This repre- 3 500 000
sents the volumes contributed by the river Bay of
B4 Plenty
itself and any remnant contribution from 3 000 000 B3
sub-tidal flood deposits. This quantity was B2 Dairy
Average annual pumping rate (m3/a)
applied for the simulations. Beach

2 500 000
SHORELINE MODEL CALIBRATION 2 000 000

B1 Bell Street
AND VERIFICATION
With the purpose of the study being to sim- 1 500 000
ulate future shoreline changes, the shoreline
model was calibrated against actual meas-
ured shoreline changes. The model was set 1 000 000
Short Line
up with the appropriate sediment sources to
represent the historic sand-pumping rates, 500 000
and calibrated against measured shoreline
changes at the municipality’s beach survey
0
beacons (which data is sourced from the 0 200 350 800 1 600 1 700 2 250 2 550
eThekwini Municipality’s extensive coastal Estimated distance northward from Vetch's Pier (m)
monitoring programme, as for example
reported on in Theron et al (2013). Figure 5 A
verage sand discharge rates along the shoreline from 1989 to 2005
Model calibration 265 000 m3. The lowest annual pumping has been discharged south of the Dairy
The period of model calibration was the rate during this period was 99 000 m3 Beach outlet.
sixteen-year period from 1990 to 2006. (2005) and the highest was 444 000 m3
This period was selected as the groynes/ (1994). The average rate has varied con- Calibration results
piers were in place (constructed 1982 to siderably over time. The running average The calibration results are depicted in
1988) and the sand-pumping scheme had annual rate is illustrated in Figure 4 for Figures 6(a) to (f). In these figures the mea-
been in operation for long enough (since the period 1989 to 2013. During this sured horizontal offset data of the +2 m
1982) to allow the shoreline to adapt to period, the average annual rate has been Chart Datum (CD) contour is depicted
these changes. Comprehensive survey 265 000 m3. In general, the annual rate has against the simulation model results (also
and pumping data was available (through been decreasing in the past three decades. at the +2 m CD contour) for the period of
eThekwini Municipality’s coastal monitor- The average distribution of pumping interest. The results depicted are given only
ing programme, e.g. Theron et al (2013)). volumes between the different beach for some of the beacon locations along the
outlets is given in Figure 5, noting that the Bight (refer to Figure 2).
Discharge data Short Line and Long Line volumes have The accuracy of the calibration results
From 1990 to 2006 a volume of approxi- been combined in this figure, as this is how was determined via the Root Mean Square
mately 4 207 000 m3 was pumped onto they were applied in the model (refer to Error of Prediction (RMSEP), which is
the beaches at an average annual rate of Figure 2). Most of the sand (almost 90%) defined as:
Beacon 2/3 (RMSEP = 16 m) Beacon 9 (RMSEP = 10 m)
100 100
90 90
80 80
Offset of +2 m CD contour (m)

70 70
60 60
50 50
40 40
30 30
20 Measured offset data 20 Measured offset data

y mean y mean
10 y standard 10 y standard
Simulation model results Simulation model results
0 0
1 Jan 2 Jul 1 Jan 2 Jul 1 Jan 2 Jul 1 Jan 1 Jan 1 Jan 1 Jan 1 Jan
1990 1992 1995 1997 2000 2002 2005 1990 1995 2000 2005
(a) Date (b) Date
Beacon 13 (RMSEP = 12 m) Beacon D (RMSEP = 14 m)

100 100
90 90
80 80
70 70
60 60
50 50
40 40
30 30

y mean y mean
0 0
1 Jan 1 Jan 1 Jan 1 Jan 1 Jan 2 Jul 1 Jan 2 Jul 1 Jan 2 Jul 1 Jan
1990 1995 2000 2005 1990 1992 1995 1997 2000 2002 2005
(c) Date (d) Date
Beacon DN8 (RMSEP = 10 m) Beacon DN12 (RMSEP = 11 m)

100 100
90 90
80 80
70 70
60 60
50 50
40 40
30 30

y mean y mean
0 0
1 Jan 1 Jan 1 Jan 1 Jan 1 Jan 2 Jul 1 Jan 2 Jul 1 Jan 2 Jul 1 Jan
1990 1995 2000 2005 1990 1992 1995 1997 2000 2002 2005
(e) Date (f) Date
Figure 6(a)–(f) Calibration results for intermitted beacons along the Bight; each result indicates the measured data, simulation results and the mean
and standard deviation of the measured data (refer to Figure 2)
n
Σ (ŷi – yi )2
× 105
i=1 Model 19/03/1992
RMSEP = 1.970
n Beacon 2/3 Model 01/01/1990
Closest data to
Where ŷi is the predicted values from the 19/03/1992
model, yi is the measured data values and n 1.965
is the number of measured data points. Beacon 9
The RMSEP is also given in metres and
Y-coordinate (m)
is thus an easy-to-understand method of 1.960
quantifying the model prediction error. Indian Ocean
It should be emphasised that the Unibest
model does not consider cross-shore
1.955 Beacon 13
effects. This is also clear in the compari-
sons presented in Figure 6. The variation Land
in the measured data can, to a large extent,

be attributed to cross-shore effect (for 1.950
example storm events). The total average
Beacon 18
long-term behaviour is, however, dominat-
ed by the longshore transport of sediment. 1.945
Thus, the ability of the model to predict 1.030 1.035 1.040 1.045 1.050 1.055 1.060
the long-term behaviour of the beaches is X-coordinate (m) × 105
adequate. It is also important to notice that
the RMSEP value varies as the standard Figure 7 A
rea of the Durban Bight close to the piers; the model results at 19.03.1992 and 01.01.1990
deviation of the measured data varies. The are presented together with the data points measured closest to the date 19.03.1992
Beacon 12 (RMSEP = 8 m) Beacon 13 (RMSEP = 11 m)

140 140
120 120
100 100
80 80
60 60
40 40
20 Measured shoreline position 20 Measured shoreline position

Model simulated position Model simulated position
0 0
1 Jan 1 Jan 1 Jan 1 Jan 1 Jan 1 Jan
1982 1983 1984 1982 1983 1984
(a) Date (b) Date
Beacon 15 (RMSEP = 18 m) Beacon 18 (RMSEP = 11 m)

140 100
90
120
80
100 70
80 60
50
60 40
40 30
20
20 Measured shoreline position Measured shoreline position
10
Model simulated position Model simulated position
0 0
1 Jan 1 Jan 1 Jan 1 Jan 1 Jan 1 Jan
1982 1983 1984 1982 1983 1984
(c) Date (d) Date
Figure 8(a)–(d) V
alidation results for intermitted beacons between the harbour entrance and the piers; each result indicates the measured
cross‑shore data along with the simulation results
best with the large changes in shoreline
× 105 location. The conclusion is that the model
Model 01/01/1982 (initial shoreline)
1.968 performs adequately and can thus be used
Model 27/09/1982
Model 24/02/1983
to simulate predictions of possible future
1.966
Model 01/10/1982 scenarios.
1.964 Data 01/03/1983
1.962
FUTURE SCENARIOS
Y-coordinate (m)
1.960 The first and most probable future scenario

Beacon 12 Indian Ocean is the decrease in the amount of sand
1.958
pumped onto the Durban beaches. There
1.956 can be many reasons for the decrease in
sand supply and the change in sediment
1.954 Durban characteristics:
1.952 ■■ The dredger might not supply enough
sand to the sand pumping scheme.
1.950 ■■ Due to dam building in the rivers in
1.948 the proximity of Durban, there might

Beacon 18
be a shortage of sand supply to the
1.030 1.035 1.040 1.045 1.050 1.055 sand trap south of the Durban Harbour
X-coordinate (m) × 105 breakwater, and thus a shortage of sand
supply for the Durban beaches. Initially
Figure 9 P
lan view of some of the validation results; measured data is compared with model results the sand shortage will manifest as
erosion on the Bluff beaches, and once
accuracy of the model prediction is thus company. A sand volume of approximately these beaches get close to depletion, the
directly related to the spread in the data. 600 000 m3 was placed on the beach in a sand supply to the sand trap will not be
A large spread in the measured data is also period of six weeks between August and sufficient.
an indication of a more dynamic beach September 1982. Limited data is available ■■ The physical characteristics of the
regarding cross-shore processes. To resolve on the precise location or rate of nourish- sand dredged from the sand trap might
these processes a fully three-dimensional ment. The shoreline model was set up to also change due to anthropogenic
or cross-shore model would have to be simulate this nourishment event. During interventions.
employed over a period usually shorter the period from July 1982 to November Each of these scenarios was investigated
than seasonal scale. 1983, routine sand nourishments also individually. With each scenario only the
A shoreline model with an RMSEP occurred, amounting to 266 000 m3. These parameter of interest was varied, while all
of less than 10 m can be regarded as were also included in the model. the other parameters were maintained at
extremely accurate. Figure 7 is provided Regular surveys of the nourishment area the baseline scenario settings. A summary
to illustrate the results of Figure 6 in a were conducted. The shoreline accreted by of these scenarios is provided in Table 1.
more realistic perspective. Here it can be between 30 m and 80 m. The surveys were The main aim of simulating these future
seen clearly that even discrepancies that used to validate the model predictions of scenarios is to elucidate the consequences
seem large in Figure 6 are in fact adequate the shoreline accretion resulting from the of not maintaining the baseline sand
in predicting shoreline variation. The nourishment. supply amount and characteristics (grain
South African Lo 31 WGS84 coordinate size) to the Durban beaches. All the future
system was used in the model, with the Validation results scenarios were simulated over a period of
following transformation to local model Similar to Figure 6 the results of the model twenty years, and the results are given in
coordinates: Xmodel = 100 000 – YLo31 and validation are presented in Figures 8(a) to five-year intervals. For all the simulations
Ymodel = 3 500 000 – XLo31. (d). The RMSEP was utilised to quantify done in the future scenario study, the
the model prediction accuracy. Not only initial shoreline was assumed to be the
Model validation did the model simulate the large nourish- long-term average existing shoreline.
Validation of a model calibration provides ment event accurately, but it also indicated The climate change scenarios were based
a check on the model accuracy. Such vali- acceptable RMSEP model correlations. In on a sensitivity analysis approach. Rather
dation data should ideally be of a similar Figure 8 only beacons between the harbour than attempting to predict the exact effect
nature to that for which the model is to be entrance and the piers are given, as this of climate change on the incoming wave
applied, i.e. similar shoreline responses or was the area with the largest changes in parameters, the present study aims to inves-
discharge rates. shoreline location. tigate the sensitivity of the Durban Bight to
The best way to illustrate the accuracy changes in only some of the incoming wave
Discharge data of a longshore transport model is by view- characteristics. These types of scenarios
In 1982 a large nourishment of South Beach ing the results in the form of scaled shore- can occur if climate change or other physi-
(approximately Beacon 13) and Addington line changes. In Figure 9 the model results cal phenomena change the net movement
Beach (approximately Beacon 6) was are compared with the measured data. of weather fronts in the Southern Indian
carried out by a contracted dredging Two dates were chosen which correspond Ocean (Weldon & Reason 2014; Kohfeld
et al 2013; Langlais et al 2011; Kostianoy Table 1 F uture sediment pumping volume and sediment characteristics scenario summary
et al 2004). Current research between the (refer to Figure 2)
South African Weather Service and the Pumping volumes and sediment characteristics scenarios
South African Environmental Observation
Baseline A
Network is quantifying the magnitude of Discharge point B C
(m3/year) (m3/year)
the expected wave height changes and wave
directional approach angle changes. To date Booster B1 – 200 m south of Beacon 18 163 968 122 831
these estimates are unclear for the southern Booster B2 – Beacon 15 66 223 49 521 Baseline Baseline
African coastlines, and thus a range of sce- Booster B3 – Beacon 12 7 541 5 509
scenario scenario
narios are presented here. A summary of the but with but with
Dairy Beach – Beacon 11 20 336 15 106 D50 = 333 µm D50 = 167 µm
scenarios investigated in the present study and and
is provided in Table 2. Sea Level Rise (SLR) Booster B4 – Beacon 7 9 573 7 033 D90 = 813 µm D90 = 407 µm
will ultimately also worsen the effects of the
Total 267 641 200 000
scenarios presented here, as the reach of the
waves will extend further inland. Studying
the effects of SLR is therefore suggested for Table 2 Future climate change scenarios summary
future shoreline studies (Habets 2015). Climate change scenarios
The wave climate changes are applied at
Baseline
the point of input into the Unibest model. Discharge point D E
(m3/year)
In all cases this is before wave breaking is
expected to occur and the start of what Booster B1 – 200 m south of Beacon 18 163 968
Baseline scenario Baseline scenario
is known as the dynamic zone. In the Booster B2 – Beacon 15 66 223 but with a: but with a:
dynamic zone the depth to which sediment Booster B3 – Beacon 12 7 541
a. 5% a. 1°
can move is specified. In areas where larger b. 10% b. 2.5°
Dairy Beach – Beacon 11 20 336 c. 20% c. 5°
wave energy is expected (towards the north increase in increase and decrease
of the Bight, refer to Figure 2) a deeper Booster B4 – Beacon 7 9 573 significant wave in wave approach
contour was chosen as dynamic depth and height direction (TN)
Total 267 641
thus a deeper near-wave breaking depth.
RESULTS
The baseline scenario is depicted in
Figure 10. For the most part the baseline
scenario maintains the Durban beaches at
a relatively stable state. The shoreline does
not change much from its initial position.
The baseline scenario thus establishes a
good reference for the future scenarios.
In Figure 11 the shorelines are given
from January 2017 to January 2037 if
scenario E(c) were to be implemented every
year with an increased wave approach
angle (i.e. more northerly or southerly).
In Figure 11 the erosion and/or accre-
tion are not clearly discernible, even with
a zoomed-in map view. Scenario E(c) was
chosen to illustrate this, because it was one
of the scenarios with the greatest effect on
the shoreline movement and yet the shoreline
dynamics are still not readily observed. To
make the change more visible Figure 12 was
created. Here the erosion and accretion are
illustrated for the entire area of interest as a
function of the distance north of Vetch’s Reef. Figure 10 P
lan view of baseline scenario together with applicable beacon locations close to the
From Figure 12 it is clear where the most Durban Harbour entrance
erosion and accretion occurred (refer to
Figure 2 for beacon locations in the Bight). (Holthuijsen 2007). To the north accretion is rates are decreased due to the wave
Since the wave approach angle will cause observed as the longshore current deposits approach angle. It should be noted that this
the coastline to align with it, the most ero- the sediment that was removed close to study is only related to long-term shoreline
sion is observed in the corner of the Bight Vetch’s Reef and the sediment transport prediction and that cross-shore processes
in this simulation. In the model the beach
× 105 width is not constrained, but in reality
2017
significant areas of the Bight beaches are
G 2022
1.975 2027
backed by hard structures along their land-
2032 ward edges (e.g. by the promenade, parking
2037 areas, revetments, etc). Thus, the model
1 (Somtseu Road)
1.970 may in certain cases predict potential
2/3
shoreline erosion extending further land-
5 (Bay of Plenty Pier) ward than the actual extent of the existing
Distance (m)
1.965 6
8 Indian Ocean sandy beach area. In such instances the
9 actual shoreline response will necessarily
10 (Dairy Beach Pier) differ from the simulated response (also
1.960 11 depending on the backshore area and pos-
Durban 12 (West Street Jetty) sible preemptive intervention measures
implemented by the municipality).
1.955 13
In Figure 13(b) the scenario with an
14 increase in sediment size is depicted. In
15 this case accretion is observed in the area
1.950 16
close to where the pumping takes place.
1.020 1.025 1.030 1.035 1.040 1.045 1.050 1.055 Further to the north of the Bight erosion
Distance (m) × 105 is observed, as the heavier sediment can-
not readily be transported to this area
Figure 11 P
lan view of scenario E(c) with increasing wave angle (refer to Table 2) by means of the longshore currents. The
inverse is observed in Figure 13(c), as the
finer sediment assumed in scenario C
2.10
is readily transported northward, thus
× 105
NC7
2017
causing erosion in the corner of the Bight
2022 NC8
2.08 NC9
and accretion further northward due to
2027
2032 NC10 the presence of groynes. In most of these
NC11
Distance north from Vetch's Reef (m)
2.06 2037
DN13 results some areas reach the new equilib-
DN12
DN11 rium more rapidly than other areas.
DN10
2.04 DN9
If consecutive years’ predictions overlap,
DN8
Erosion DN7
Accretion it implies that those areas’ shorelines have
DN6
NC15 reached its new equilibrium position. From
2.02
NC16
a coastal management point of view these
A
areas will require the least amount of
2.00 B
mitigation and/or intervention to maintain
C
CD
D
E (Walter Gilbert Road)
future beach stability. In areas where
1.98 F continued erosion or accretion is observed,
G
1 (Somtseu Road) the gradient of erosion and accretion is still
5 (Bay Plenty Pier)
1.96 10 (Dairy Beach Pier) present over time. A decreasing steepness in

12 (West Street Jetty)
13
the gradient is an indication that the shore-
17
20 line is realigning itself towards its new equi-
1.94
–150 –100 –50 0 50 100 150 librium position. The simulations were not
Shoreline movement (m) run long enough to illustrate the total new
equilibrium position of the entire coastline,
Figure 12 T otal erosion and accretion for the entire area of interest but it was assumed that the gradient in ero-
sion and accretion provides useful additional
are not considered. The results, especially DISCUSSION assistance to coastal management and gov-
between the piers, should be interpreted In Figure 13(a) the scenario is simulated erning authorities, for example to identify
with care. The shoreline stability between of pumping 25% less sand to the Bight. As the areas where the greatest potential future
the piers also indicate the effectiveness of expected, erosion occurred especially in shoreline impacts are expected. The new
these structures in slowing down struc- the area at the down-drift side of the piers. beach equilibrium may never be reached in
tural erosion (Bosboom & Stive 2015). The After twenty years a maximum erosion in the model. In reality the beach profile will
groynes close to beacon A and DN10 also the order of 40 m is expected. Erosion is flatten and the shallow water waves will
caused local accretion on the down-drift first expected at the southern side of the dissipate differently towards the shore. This
side of these structures. model, while erosion to the northern part is a limitation of the model that must also
Figure 12 and Figures 13(a) to (k) give of the model is expected once the southern be kept in mind, and the beach profile used
all the scenarios summarised in Tables 1 beaches are depleted of sediments. The must be updated with large erosion and/or
and 2. stabilisation effect of the piers is again clear accretion events.
Scenario A Scenario B
2.10 2.10
× 105
× 105
NC7 NC7
2017 2017
2.08 2022 NC8
NC9 2.08 2022 NC8
NC9
2027 2027

NC10 NC10
NC11
2.06 2032 NC11
DN13
2.06 2032 DN13
DN12 DN12
2037 DN11 2037 DN11
DN10 DN10
2.04 DN9 Accretion 2.04 DN9

DN8
Accretion
DN8
DN7 DN7
DN6 DN6
Erosion NC15 Erosion NC15
2.02 2.02
NC16 NC16
A A
2.00 B 2.00 B
C
C CD
CD
D D
E (Walter Gilbert Road) E (Walter Gilbert Road)
1.98 F 1.98 F
G
G
1 (Somtseu Road) 1 (Somtseu Road)
5 (Bay Plenty Pier)
8 8
1.96 10 (Dairy Beach Pier)
13
14 15
–150 –100 –50 0 50 100 150 –150 –100 –50 0 50 100 150
(a) Shoreline movement (m) (b) Shoreline movement (m)
Scenario D (b) Scenario D (c)

2.10 2.10
× 105
× 105
NC7 NC7
2017 2017
2.08 2022 NC8
NC9 2.08 2022 NC8
NC9
2027 2027

NC10 NC10
NC11 NC11
2.06 2032 DN13
DN12
2.06 2032 DN13
DN12
2037 DN11
DN10
2037 DN11
DN10
2.04 DN9
DN8 Accretion 2.04 DN9
DN8 Accretion
DN7 DN7
DN6 DN6
NC15 NC15
2.02 Erosion 2.02 Erosion

NC16 NC16
A A
2.00 B
C
2.00 B
C
CD CD
D D
1.98 G
F 1.98 F
G
5 (Bay Plenty Pier) 5 (Bay Plenty Pier)

12 (WestStreet Jetty) 1.96 10 (Dairy Beach Pier)
14 14
15
17
20
1.94 1.94
–150 –100 –50 0 50 100 150 –150 –100 –50 0 50 100 150
(e) Shoreline movement (m) (f) Shoreline movement (m)
Scenario E (c) decreasing Scenario E (a) increasing

2.10 2.10
× 105
× 105
NC7 NC7
2017 2017
2.08 2022 NC8
NC9 2.08 2022 NC8
NC9
2027 2027
NC10 NC10
NC11 NC11
2.06 2032 DN13
DN12
2.06 2032 DN13
DN12
2037 DN11
DN10
2037 DN11
DN10
2.04 DN9
DN8 Accretion
DN7 DN7
DN6 DN6
NC15 NC15

NC16 NC16
A A
2.00 B
C
2.00 B
CD
D D
1.98 F
G
1.98 F
G
1.96 12 (West Street Jetty) 1.96 10 (Dairy Beach Pier)

13
15 14
17 17
20
1.96 1.94
–150 –100 –50 0 50 100 150 –150 –100 –50 0 50 100 150
(i) Shoreline movement (m) (j) Shoreline movement (m)
Figure 13(a)–(k) R
esults of long-term numerical modelling grouped per scenario, as described in Tables 1 and 2; the title of each figure describes the scenario
Scenario C Scenario D (a)
2.10 2.10
× 105
× 105
NC7
2017 2017
NC8
2.08 2022 NC9
2.08 2022 NC8
NC9
2027 2027

NC10 NC10
NC11 NC11
2.06 2032 DN13
DN12 2.06 2032 DN13
DN12
2037 DN11
DN10
2037 DN11
DN10
2.04 DN9
DN8 Accretion
DN7
DN7 DN6
DN6 NC15
Erosion NC15
2.02 Erosion
2.02 NC16
NC16
A
A 2.00 B
C
2.00 C
B
D
CD
CD
E (Walter Gilbert Road)
E (Walter Gilbert Road) 1.98 G

F
1.98 F
G
1 (Somtseu Road)
5 (Bay Plenty Pier)
1 (Somtseu Road)
5 (Bay Plenty Pier) 13
15
13 20
14 1.94
–150 –100 –50 0 50 100 150 –150 –100 –50 0 50 100 150
(c) Shoreline movement (m) (d) Shoreline movement (m)
Scenario E (a) decreasing Scenario E (b) decreasing)

2.10 2.10
× 105
× 105
NC7 NC7
2017 2017
2.08 2022 NC8
NC9 2.08 2022 NC8
NC9
2027 2027
NC10 NC10
NC11 NC11
2.06 2032 DN13
DN12
2.06 2032 DN13
DN12
2037 DN11
DN10
2037 DN11
DN10
2.04 DN9
DN8 Accretion
DN7 DN7
DN6 DN6
NC15 NC15

NC16 NC16
A A
2.00 B
C
2.00 B
C
CD CD
D D
1.98 F
G
1.98 F
G

12 (West Street Jetty) 1.96 10 (Dairy Beach Pier)
13
14 15
18 19
1.94 1.94
–150 –100 –50 0 50 100 150 –150 –100 –50 0 50 100 150
(g) Shoreline movement (m) (h) Shoreline movement (m)
Scenario E (b) increasing

2.10
In Figures 13(d), (e) and (f) an increase in wave height has the
× 105
2017
2.08 2022
NC8
NC9 net effect of erosion in the corner of the Bight. The increase in wave
2027 height can be related to an increase in wave power, and thus the ero-
NC10
NC11
2.06 2032
DN13
DN12
sion potential increases on the beaches where the longshore current
2037
DN11
DN10
is strengthened. All three percentage increases under scenario D had
2.04 DN9
Accretion the same outcome over a twenty-year period. The only aspect that
DN7
DN6
NC15
differed was the rate at which the final coastline was reached. The
2.02 Erosion higher the wave energy, the quicker a seemingly new stable coastline
NC16
A
was obtained. The area down-drift of the piers, experiencing erosion,
2.00 B
also eroded more, and the extent of the erosion increased with an
CD
D
E (Walter Gilbert Road) increase in wave height. The amount of accretion observed at the
1.98 F
G northern extent of the model also increased with increase in wave
1 (Somtseu Road)
5 (Bay Plenty Pier)
height. This is a sign that the amount of northwardly transported
sediment is increased to such an extent that the sediment will start to
15
20 build up on the northern beaches. The amount of accretion caused by
1.94
–150 –100 –50 0 50 100 150 the groynes also decreased with increase in wave energy.
(k) Shoreline movement (m) In Figures 13(g) to (i) the effects of a decreasing wave approach
angle (anti-clockwise) is illustrated. Here it can be seen how the
corresponding to the table descriptions waves are reshaping the beaches to adapt to the new approach angle.
The net effect of this scenario is accretion areas, and therefore require earlier sand the scenario and beach location. It is thus
in the corner of the Bight and erosion at nourishment if such progressive erosional also clear that more research is needed to
the northern beaches. The locations of the trends are observed. quantify the exact directional and wave
erosion and accretion are associated with A small accreting trend (approximately energy changes southern Africa can expect
the increase and decrease in the longshore 1 m per annum) was also observed in in the future. These investigations should
transport rate associated with the change both the model and the data at most of also include sea level rise (SLR), as this will
in wave approach angle. the monitoring beacons (refer to Figure 6 increase the reach of the prevailing wave
For the last scenario, the opposite for historical records). Analysing these climate and bring about cross-shore beach
change in wave approach angles was inves- trends fell outside the scope of the cur- profile shifts. Given more accurate climato-
tigated and is depicted in Figures 13(j) and rent study and is therefore suggested for logical prediction, the appropriate shoreline
(k) and Figure 12. In this scenario accretion future research. evolution scenario may be utilised in the
is observed with an increase of one degree Durban coastal management plan, both
in wave approach angle (clockwise). For an to protect existing infrastructure and to
increase of two and a half degrees a new CONCLUSION design and build new infrastructure with
equilibrium scenario is reached from 2022 The shoreline dynamics of the Durban an appropriate setback distance. This study
onwards. If the approach angle is increased Bight were investigated using numerical may also inform future studies regarding
even further to five degrees (scenario models. Calibration and validation were the general sand budget of the east coast of
E(c)) progressive erosion is observed. This done using existing historical data. The South Africa, which includes the building
scenario illustrates some similarities to the accuracy of the model predictions was of more dams on the sediment-supplying
increasing wave height scenario D, with quantified using the RMSEP. Once the rivers of the east coast. The present
similar accretion to the north of the Bight. model was adequately calibrated, the study indicated which areas of the Bight
An increasing wave approach angle of two possible future scenarios were simulated. were least susceptible to severe shoreline
and a half degrees is thus a critical point Each parameter was investigated by vary- changes, and thus assists the governing
between a future progressively eroding or ing only that parameter and keeping the authorities in focusing their management
accreting coast (in the southern corner of rest of the input criteria identical to the strategies to the areas most at risk of
the Bight for scenarios with an increasing baseline scenario. future shoreline changes. With knowledge
wave approach angle). The effects of the four future scenarios of such potential changes, the eThekwini
The results of the present study were clear. These results were compared Municipality will be able to investigate
illustrate the relative changes in future with the baseline scenario in which the ways to manage and mitigate the potential
shorelines based on possible future sce- beaches remained reasonably stable for the impacts of wider/narrower beaches, such as
narios. These results must be interpreted twenty-year simulation period. The effect blockage of stormwater outlets, windblown
as indicative and may be used as a decision of reducing the sand-pumping volumes sand problems, reduced available beach
support tool for the management of the was a reduction in beach width. The ero- space for recreational activities and pos-
beaches at Durban. The exact amount sion was particularly extensive in the area sible effects on the Umgeni River mouth.
of erosion and accretion observed in between the harbour entrance and the Generally, small alterations of the Bight
the future might differ from the results piers. The erosion lessened further north conditions may lead to adverse effects,
presented here, depending on the actual in the Bight, but it can be expected that and thus maintaining the current sand-
changes in Southern and Indian ocean these beaches will also erode once the sand pumping rates and sediment characteristics
(and atmospheric) dynamics (considering supply from the southern beaches dimin- are of utmost importance.
model inaccuracies). Once more research ishes in time. In all the future scenarios
is available to quantify these changes, the the piers seemed to reduce the amount
present study scenario closest to the pre- of erosion of the beaches resulting from ACKNOWLEDGEMENTS
dicted offshore changes can be readily used longshore transport. The authors would like to acknowledge
by coastal managers. The present study The other two scenarios involved the eThekwini Municipality for providing the
also identifies the areas where potential alteration of the grain size of sand being historical horizontal offset data used for
future intervention might be required, and pumped onto the Durban beaches. A gen- the calibration and validation of the present
thus helps coastal managers in planning eral increase of 25% in grain size, as well shoreline model. We would also like to thank
where to focus their attention and which as a decrease of 25% in grain size, had the the Council for Scientific and Industrial
valuable coastal infrastructure would be total effect of erosion on the Bight beaches. Research (CSIR) for providing funding for
most threatened. The potential severity of The way the erosion occurred was differ- the development of the shoreline model. We
some of the predicted results should not ent in that the finer sand particles eroded furthermore acknowledge the original report
be underestimated. With dry beach widths quickly and then remained relatively stable, by Christoph Soltau on which this work was
along the Durban Bight currently ranging while the coarser sand eroded consist- based (Soltau & Theron 2007), and we thank
from as little as about 15 m up to about ently. The dynamics south and north of Transnet National Port Authorities for pro-
90 m (measured on Google Earth imagery the piers were also different in both cases, viding us with appropriate data.
of September 2017), very little or not much with the finer lighter particles indicating
buffer is available on the Durban beaches more extensive erosion south and north of
to tolerate several of the erosion predic- the piers. REFERENCES
tions. Erosion of 40 m or 50 m of the upper The effect of climate change might be Barnett, K A 1999. The management of Durban’s
beach would be catastrophic in many Bight either erosion or accretion, depending on beaches: An historical perspective. Proceedings,
5th International Conference on Coastal and Port Holthuijsen, L H 2007. Waves in Oceanic and Coastal Schoonees, J S 2000. Annual variation in the net
Engineering in Developing Countries, COPEDEC V, Waters. New York: Cambridge University Press. longshore sediment transport rate. Coastal
19–23 April, Cape Town, pp 695–708. Hugo, P 2013. The impact of climate change effects Engineering, 40: 141–160.
Bosboom, J & Stive, M J 2015. Coastal Dynamics 1. on the planform of a headland-bay beach on the Soltau, C & Smith, G 2003. Shoreline modelling
Lecture Notes CIE4305, Vol 0.5. Delft, Netherlands: southern coast of South Africa. MSc Dissertation, applications in southern Africa. Proceedings, 6th
Delft Academic Press. Stellenbosch: Stellenbosch University. International Conference on Coastal and Port
Campbell, N P, MacLeod, D C & Swart, D H 1985. Kohfeld, K E, Graham, R M, de Boer, A M, Sime, L C, Engineering in Developing Countries, COPEDEC
Bypassing and beach nourishment scheme at Wolff, E W, Le Quere, C & Bopp, L 2013. Southern VI, 15–19 September, Colombo, Sri Lanka, p 116.
Durban. Proceedings, 26th International Navigation hemisphere westerly wind changes during the Soltau, C & Theron, A K 2007. Shoreline accretion
Congress, 16–29 June, Brussels, Belgium, Section II, Last Glacial Maximum: Paleo-data synthesis. prediction in the Durban Bight. CSIR Report No
Subject 3, pp 7–18. Quaternary Science Review, 68: 76–95. ENV-S-C 2007, Stellenbosch: CSIR.
Cawthra, H, Uken, R & Ovechkina, M 2012. New Kostianoy, A G, Ginzburg, A I, Frankignoulle, M Theron, A K 2007. Analysis of potential coastal zone
insights into the geological evolution of the Durban & Delille, B 2004. Fronts in the Southern Indian climate changes and possible response options in
Bluff and adjacent blood reef. African Journal of Ocean as inferred from satellite sea surface the Southern African region. Proceedings, IPCC/
Geology, 115(3): 291–308. temperature data. Journal of Marine Systems, TGICA Regional Experts Meeting on Climate
CSIR (Council for Scientific and Industrial Research) 45: 55–73. Change, 20–22 June, Nadi, Fiji.
2003. Sediment dynamics in the Durban Vetch’s Langlais, C, Rintoul, S & Schiller, A 2011. Variability Theron, A K 2015. Dams and sand mining in KZN
Bight region: Response to proposed harbour entrance and mesoscale activity of the Southern Ocean rivers. Implications for coastal sand budget. Three-
developments. CSIR Report No ENV-S-C 2003-087, fronts: Identification of a circumpolar coordinate day Course on Stormwater, River and Estuary
Stellenbosch: CSIR. system. Ocean Modelling, 39(1): 79–96. Hydraulics. Stellenbosch University, Department
CSIR (Council for Scientific and Industrial Research) Mather, A A & Stretch, D D 2012. A perspective on sea Civil Engineering.
2007. Shoreline accretion predictions in the Durban level rise and coastal storm surge from southern and Theron, A K & Rautenbach, C 2014. Proposed Durban
Bight. CSIR Report, Stellenbosch: CSIR. eastern Africa: A case study near Durban, South dig out port FEL 2 study. Assessment of sediment
Deltares 2005. UNIBEST-CL+6.0. User and Theoretical Africa. Water, 4: 237–259. doi:10.3390/w4010237 transport issues. CSIR Report No CSIR/NRE/
Manual. Utrecht, Netherlands. Mather, A A & Theron, A K 2011. Recent extreme COAST/2014/ECCS106. Stellenbosch: CSIR.
Diedericks, G P, De Wet, P & Theron, A K 2011. On events along the coast of South Africa. Paper Theron, A K, Rautenbach, C, Watt, L J & Von St Ange,
cross-shelf sediment mobility along the Kwazulu- presented at the National Storm Surge Workshop. U 2013. Durban beach monitoring progress report:
Natal coast. CSIR Report No CSIR/NRE/CO/ Knysna: Oceans and Coasts Directorate, July 2009 to June 2011. CSIR report number: CSIR/
IR/0047/A, Stellenbosch: CSIR. Department of Environmental Affairs. NRE/CO/ER/2013/C Vol 1. Stellenbosch: CSIR.
eThekwini Municipality 2010. Official website of the Mather, A, Kasserchun, R & Wenlock, H 2003. Theron, A, Barwell, L, Maherry, A, Luck-Vogel, M,
eThekwini Municipality. Available at: http://www. City of Durban sand bypassing scheme: 20-year Robberts, W, Terblanche, L, de Wet, P & Araujo, F
durban.gov.za/Documents/Invest_Durban/ performance evaluation. Proceedings, 6th 2012. Responding to climate change in Mozambique.
Economic%20Development/3_Port_Summary_ International Conference on Coastal and Port Theme 2: Coastal planning and adaptation
Doc.pdf [accessed in 2015]. Engineering in Developing Countries, COPEDEC to mitigate climate change impacts. Maputo,
Garland, G & Moleko, L 2000. Geomorphological VI, 15–19 September, Colombo, Sri Lanka, Mozambique: Ministry of State Administration,
impacts of Inanda Dam on the Mgeni estuary, pp 401–402. National Institute for Disaster Management (INGC).
north of Durban, South Africa. Bulletin of Rossouw, M & Theron, A 2012. Investigation of Theron, A, de Lange, W, Nahman, H & Hardwick, D
Engineering Geology and the Environment, 59(2): potential climate change impacts on ports and 2008. Sand supply from rivers within the Ethekweni
119–126. maritime operations around the southern African jurisdiction. Implications for coastal sand budgets
Green, A N, Ovechkina, M N & Mostovski, M B 2012. coast. Proceedings, Maritime Transport and and resource economics. CSIR Report No CSIR/
Late Holocene shoreface evolution of the wave- the Climate Change Challenge, United Nations NRE/ECO/ER/2008/0096/C, Stellenbosch: CSIR.
dominated Durban Bight, KwaZulu-Natal, South Conference on Trade and Development (UNCTAD), Weldon, D & Reason, C J 2014. Variability of rainfall
Africa: A mixed storm and current-driven system. 14 June, London: Routledge, pp 286–298. characteristics over the south coast region of South
Continental Shelf Research, 49: 56–64. Rossouw, M, Davies, J, Coetzee, L & Kuipers, J 1999. The Africa. Theory of Applied Climatology, 115, 117–185.
Habets, F C 2015. Understanding the sediment wave recording network around the South African doi:10.1007/s00704-013-0882-4
transport and budgets around the Durban Digout coast. Proceedings, 5th International Conference Wells, C C 2015. Modelling sand bypass schemes on
port, South Africa. MSc Dissertation, Delft, on Coastal and Port Engineering in Developing the KwaZulu-Natal coastline (MSc ed.). Durban:
Netherlands: Delft University of Technology. Countries, COPEDEC V, 19–23 April, Cape Town. University of KwaZulu-Natal.
TECHNICAL PAPER Risk-based member
Journal of the South African
Institution of Civil Engineering
reliability in structural design
ISSN 1021-2019
Vol 60 No 4, December 2018, Pages 16–24, Paper 0243 N de Koker, A A Elvin
DR NICO DE KOKER, Associate Member of

SAICE, received a PhD in geophysics, with a The balance between safety and economy in structural design was explored in the context
focus on the physics of materials at extreme
of the member cost, liability and location in a structure. A model was developed giving
conditions, from the University of Michigan
in 2008. Following a number of years as a the optimal reliability of a member, taking account of the tradeoff between cost and risk in
post-doctoral researcher, he changed his maximising the long-term expected benefit derived from the structure.
focus to civil engineering, graduating from
The model was first applied to a single independent member to derive a relationship which
the University of the Witwatersrand in 2017.
His research interests broadly focus on the analysis of risk and reliability in expresses the reliability required for optimal benefit as a function of the liability-cost ratio.
structural and infrastructure engineering. He is currently affiliated with Next the model was applied to two test structures: a determinate steel truss and a multi-storey
the fire engineering research unit at Stellenbosch University. reinforced concrete frame. Reliability analysis for both structures revealed that members can
Affiliation when completing research presented: be treated as independent, and that marginal benefit is greatest for members with the highest
School of Civil and Environmental Engineering liability-cost ratio values.
University of the Witwatersrand
It was shown that the relationship of liability-cost ratio versus optimal reliability provides
Private Bag 3
Wits 2050 a guideline for the improvement of existing structural design. Structures with reliabilities less
South Africa than the optimal value can most effectively be improved by strengthening members with the
Current contact details: highest liability-cost ratio values, while structures with reliabilities greater than optimal are
Department Civil Engineering improved by economising on members with the lowest liability-cost ratio values.
Stellenbosch University
Private Bag X1
Matieland 7602
South Africa INTRODUCTION For small pf a member’s expected lifespan
T: +27 21 808 4434 The primary goal of engineering design will far exceed the design lifetime, and
E: ndekoker@sun.ac.za
is to balance safety and economy. The the probability of failure can be taken as
simplest approach to avoid failure is to constant during the design life.
PROF ALEX ELVIN Pr Eng, Member of SAICE,
overcompensate for the expected load- Calibration studies that aim to deter-
graduated as a civil engineer from the
University of the Witwatersrand in 1989. ing conditions by means of safety factors mine partial factors of safety from a
Working on non-destructive behaviour of (load and material factors), usually at the statistical basis (Milford 1988; Holický et
concrete bridges he completed his Master’s expense of economy. al 2010) indicate that the long-standing
in 1991 at the Massachusetts Institute of
Technology (MIT), and in 1996 obtained his
Safety factors can be intuitively under- empirical range of factors used in permis-
PhD from MIT in numerical modelling of stood to decrease the probability of failure sible stress design correspond to β values in
fracture of brittle high-temperature materials. He worked in industry and of a design, as it accounts for reasonable the range of 3 to 5.
taught several classes at MIT. From 1998 to 2002 he was a junior faculty
variation in applied loads and member As part of an effort to establish a
member at Harvard Medical School doing finite element analysis of
implants. He was promoted first to associate (2007) and then to full resistance values. The statistical inter- robust reliability basis for structural
professor (2010) in structural mechanics at the University of the pretation of safety factors was formalised design, the Joint Committee on Structural
Witwatersrand. His research interests are focused on theoretical with the development of mathematical Safety (JCSS 2008) considered the tradeoff
modelling (i.e. numerical modelling), finite element analysis, loading,
statistics, which enabled the development between the cost of a safety measure and
dynamic simulations, behaviour of structures, instrumentation and sensor
networks in engineering, as well as health monitoring of structures. of the theory for structural reliability the risk associated with fatalities due to
(Freudenthal 1947; Freudenthal & Gumbel structural failure. Based on this analysis,
Contact details:
School of Civil and Environmental Engineering 1953; Pugsley 1955). ISO 2394:2014 provides 50-year reference
University of the Witwatersrand It is customary to quantify structural period target structural reliability values
Private Bag 3 reliability via the reliability index β, which ranging between β = 2.0 and 3.8, for
Wits 2050
South Africa
expresses the separation between expected structural classes depending on the cost
T: +27 11 717 7145 failure and the mean loading and resistance of safety and the consequences of failure.
E: alex.elvin@wits.ac.za conditions in units of standard deviations. Many modern limit-states design stan-
β is associated with the failure probability dards for loads acting on structures use
within a given reference period via the target β that comply with this range. In
standard normal distribution as (Rackwitz particular, SANS 10160:2011 uses β = 3.0
& Fiessler 1978): (Retief & Dunaiski 2010, also assuming a
50-year reference period). For consistency
β = √2 erf–1(1 – 2pf )(1) with these values, 50-year reference
periods will be assumed throughout
where pf is the probability of failure of this study.
Keywords: structural reliability, risk optimisation, reliability index, the member in its design life, and erf(·) is Material-focused design codes (e.g.
structural member design the error function (e.g. McQuarrie 2003). SANS 10162:2011 and SANS 10100:2000)
De Koker N, Elvin AA. Risk-based member reliability in structural design.

all focus on failure of individual members,
specifying partial factors calibrated to ζG
F
target reliability values that depend on ζF
the dominant mode of failure. Milford ζG
xG
(1988) recommends that, for South African ζF
materials codes, β = 3.0 be used for ductile G
failure region
failure modes, β = 4.0 for brittle modes and F=
β = 4.5 for connections. ζG
G
However, these factors do not take the ζF
θ β΄1
location of a member in the structure into θ β1
β΄0
consideration. Almost all structures have
β0
members of greater and lesser importance; ζG
F
determinate structures being an exception. ζF
Yet, when designed according to current
building standards, all members will tend
to have a similar target reliability level, f N(G, F)
ζG
so that some members may be under- or θ = tan–1
ζF
over-designed from the perspective of risk.
If the relative importance of members in a G xG G
structure in the context of reducing overall
risk exposure is taken into account, the eometrical representation of the derived scaling relationship between β0 and β1
Figure 1 G
appropriate adjustment of the design value (Equation 10)
for member resistance capacity needs to
be investigated. where I is the total revenue generated from member i of pfi within the design life of the
This work explores the tradeoff between the structure over its lifetime, C is the cost structure. The total risk over the lifetime of
safety cost and failure risk in determin- of construction and commissioning, and R the structure is then:
ing the member reliability that is most is the risk due to failure of any part of the
N N N
favourable in terms of the total expected structure. Risk, defined as the probability R= Σ pfi cfi = Σ pfi di + Σ Γij cj (5)
benefit over the lifetime of the structure. It of an event times the potential loss result- i i j
develops the theory that considers optimal ing from it, accounts for both the magni-
adjusted reliability for each member in a tude and the likelihood of payable damages. Member reliability adjustment
structure taking risk into account. The Member i costs ci to construct, so that The characteristic internal force Fk and
theory is then applied to two types of the total cost of the structure C can be resistance Gk values used in the design of
structures: determinate structures in which taken as: a member reflect conservative upper and
failure of one member implies failure of lower bounds on these design parameters,
N
all, and hierarchical structures where respectively. For simplicity, F and G are
C= Σ ci(3)
the consequences of member failure vary i assumed to be normally distributed with
with position. means F and G , and coefficients of varia-
Secondary factors, such as maintenance tion ζF and ζG.
cost, construction time and deprecia- Now suppose that the resistance
THEORETICAL DEVELOPMENT tion can be taken into account. However, capacity of a member with reliability β 0
they are ignored here to keep the is adjusted by a factor x, that is G1 = G0 x.
Expected benefit theory tractable. A relation is required for the new reliability
Consider a structure with a design life In the event of failure of member i, a β1 of the adjusted member.
of τ years. The structure consists of N set of dependent members will be affected, Figure 1 illustrates geometrically the
members, grouped into n member types. described by matrix Γ. Liability di will be derivation of x in terms of β 0 and β1 that
Members of a given type have identical payable by the owner of the structure, and follows. The vertical axis is scaled by the
design specification, reliability, cost, and the failed member together with all depen- ratio of coefficients of variation to indicate
liability payable upon failure. Member dent members will have to be replaced. The values of FζG /ζF . This ensures that the
dependence is specified via the N × N total cost of the failure would then be: contours of the bi-variate normal density
matrix Γ, in which entry Γij is 1 if member distribution f N (G, F) are circular, so that
N
i supports member j, and 0 otherwise. For lines β 0 and β1 are normal to the failure
Cfi = di + Σ Γij cj(4)
a determinate structure, all entries of Γ i boundary (F = G), which is inclined at:
will be 1; a structure with a high degree of
redundancy will have a sparse Γ. Members are assumed to have been FζG /ζF ζG
tan θ = = (6)
The total expected benefit Ψ derived designed to meet a minimum target reli- G ζF
from the use of the structure over the ability β 0, for example by adhering to the
course of its design life is given by: specifications set out in SANS 10100:2000, According to its multidimensional geo-
SANS 10162:2011 and SANS 10160:2011. metrical interpretation (Rackwitz &
Ψ = I – C – R(2) This implies a failure probability for Fiessler 1978), β 0 is the distance (in units of
Although only exact for a few special
Rands cases, the preceding discussion suggests
Cu + Ru that m values can be expected to range
between 0.5 and 1.0 for basic member
failure modes.
xopt Marginal benefit of increasing

member reliability
slope ∝ marginal benefit

Let the G resistance of all members of type
u in the structure be scaled by a factor x,
resulting in new member reliabilities of β1x
tange and failure probabilities pfx .
nt at
x = 1.0 Changes in the expected benefit will
Cu Ru
result only from changes in the unit cost
and failure probability of members of type
0 1.0 x u. The contribution to the total cost and
risk from members of type u is:
n illustration of the tradeoff between increasing cost Cu and decreasing risk Ru as
Figure 2 A
members of type u are strengthened by a factor x (not to scale; m = 1) Cu = Σ cixm = Cu0 xm(16)
i∈u
Σ pfx ⎛⎜di + Σ Γij cj xjm ⎛⎜(17)

N
⎝ ⎝
standard deviation) from the mean to the cix = ci xm(12) Ru =
point on the failure boundary where the i∈u j
where xj = x for j ∈ u and xj = 1 otherwise.

multivariate probability density function is where m is a constant.
a maximum. For example, in simple tensile failure,
For G = G fixed, the marginal reliability where the member resistance is related to With increasing x, Cu increases linearly,
index is then: the yield stress σ y via the section area A as: while Ru decreases asymptotically to zero
= ⎛⎜ – 1⎛⎜(7)
as the probability of failure pfx decreases.
1 G
ζF ⎝ F ⎝
G –F Gtension = σ y A(13) These trends are schematically illustrat-
β΄0 =
ζF F ed in Figure 2. As a result of these opposing
increasing G by a factor x would imply trends, an x value xopt exists where Cu + Ru
If the member resistance capacity is now increasing the cross-sectional area and is a minimum, i.e. where Ψ is a maximum.
adjusted, the mean shifts, and thus the volume by the same factor. If cost At x opt the member design represents
is taken to be proportional to member an optimal balance between safety and
Gx – F x–1 mass, this type of failure would imply economy (see Equation 2). If the member
β΄1 = = xβ΄0 + (8)
ζF F ζF m = 1. group is under-designed (with respect
Failure in bending of a square or cir- to Ψ), the benefit can be increased by
Defining the total coefficient of variation as cular sectioned member implies moment making members more reliable, so that
ζT = √ζG2 + ζF2, it is clear from the geometry resistance: x opt > 1; if the member group is over-
Gmom ∝ σ y A3/2(14)
that: designed, members can be more affordable
and x opt < 1.
β0 β1 ζF The extent to which a member group
cos θ = = = (9)
β΄0 β΄1 ζT Scaling G by a factor x now implies is over- or under-designed is quantified
increasing the volume by a factor of x2/3. via the effect of spending (or saving) on Ψ.
so that the new relationship is: For cost proportional to member mass, this This ‘marginal benefit’ ψ is given by:
therefore gives m = 2/3. Sections of more
x–1 complex geometry can only be approxi- dΨ 1 dΨ
β1 = xβ 0 + (10) ψ= = (18)
ζT mately represented via Equation 12. dCu x=1 mCu0 dx x=1
The empirical relations used in design
which can be rearranged to give: against buckling failure (e.g. SANS Under-designed members require addition-
10162:2011) cannot be directly adapted to al spending, and so ψ > 0; over-designed
β1 + 1/ζT conform with Equation 12. However, if members imply too much has been spent,
x= (11)
β 0 + 1/ζT buckling failure is described by the Euler so ψ < 0.
equation, resistance of a square or circular Comparison of ψ among member types
sectioned member implies: in a structure indicates where in a design
Gbuckle ∝ I/L2 ∝ A2/L2(15)

Member cost adjustment the greatest change in Ψ can be affected
Let the resistance G of member i be scaled for a unit amount of expenditure/savings.
by factor x. Given this linear scaling of In addition ψ = 0, implies that the member
resistance, it will be assumed that the so that increasing G by a factor x implies type design is optimal. This special sce-
effect on member cost ci can be repre- an increase in the volume by a factor x1/2, nario is described by βopt, the value of β 0
sented as: that is m = 1/2. for which x opt = 1.
Optimal reliability of an
independent member 7
Consider now a single, independent mem-
ζT = 0.32
ber designed to reliability β 0 with cost c
and liability d. Given coefficients of varia-
tion ζG and ζF for the resistance and inter- 6 m=1
nal forces, what is the optimal reliability of
the member βopt? m = 2/3
If member resistance is increased by
a factor x, the reliability becomes β1 with 5
m = 1/2
failure probability p1, and the expected
benefit is:
βopt
4
Ψ = I –xmc – p1c [d/c + xm](19)
5.4 d/c = 105
At maximum expected benefit, x = xopt 5.3

and dΨ/dx = 0. At the point of optimal
3
βopt
marginal benefit where ψ = 0 and xopt = 1, 5.2
the design reliability is also optimal, so m = 1/2
that β 0 = β1 = βopt, and p1 = popt. Taking 5.1 m = 2/3
dΨ/‌d x = 0 and setting xopt = 1 yields:
2 5.0 m=1
d/c + 1
1 + popt + p΄opt = 0 (20) 0 0.2 0.4 0.6 0.8 1.0
m ζT
1
where, from the density function of 101 102 103 104 105 106 107 108
the standard normal distribution (e.g. d/c
McQuarrie, 2003):
popt = 1 + erf ⎛⎜– opt ⎛⎜

ptimal reliability βopt of an independent member; main plot determined for ζT = 0.32;
Figure 3 O
1 β
⎝ √2 ⎝
(21) inset for d/c = 105
2
a given failure liability, greater member cost The primary aim of this reliability
dp results in lower optimal member reliability. analysis is to explore the potential spread
p΄opt =
dx x=xopt That is, increased safety is more affordable of optimal member reliabilities βopt within
exp ⎛⎜– opt ⎛⎜

for less expensive members. a structure. In addition, the extent of
βopt + 1/ζT β2
⎝ √2 ⎝
(22) The relationship is not exact for strengthening required to upgrade mem-
=
– √2π members forming part of a structure, as bers in a structure from design reliability
If d/c ≫ 1, the relation becomes:

the risk includes the cost of dependent β 0 to optimal βopt is of interest, together
if di ≫ ΣjΓijcj the coefficient of p΄opt

members (see Equation 17). Nonetheless, with the most effective modification by
which the expected benefit of an existing
d/c would again tend to (d/c)/m. Therefore, design can be increased.
1 + popt + p΄opt = 0 (23)
m Equation 20 can be used for members that To perform the analysis, coefficients of
are part of structures as well, provided that variation ζ for applied loads and member
With βopt known, the required adjustment the liability due to member failure exceeds resistance capacities must be assigned.
to the design resistance is then: the cost of repair by a sufficient margin for Table 1 summarises the range of values
the latter to be negligible. suggested in the literature. To conform with
βopt + 1/ζT values used in the calibration of limit-states
xadjust = (24)
β 0 + 1/ζT design codes (Holický et al 2010; Holický &
APPLICATION TO TEST Retief 2005; Kemp et al 1987; Milford 1988),
Equation 20 provides an implicit relation- STRUCTURES ζGsteel = 0.10 and ζGconcrete = 0.20, together
ship for βopt of an independent member in The theory developed in the preceding sec- with imposed loads ζF = 0.25, were used in
terms of d/c, ζT , and m. As can be seen in tion is now applied to two different exam- the test examples.
Figure 3, the dominant factor determin- ple test structures. First, a determinate As failure modes are not specified, it
ing the value of βopt is the liability-cost steel truss, where failure of any member will be assumed that the cost adjustment
ratio d/c. compromises the entire structure, so that relation (Equation 12) is linear, that is
This dependence implies two important the failure liability is the same for all mem- m = 1 and cix = xci . As noted in Figure 3,
concepts. Firstly, for the same member cost, bers. Second, a three-storey reinforced con- the effect of this assumption on βopt values
a greater failure liability requires a greater crete frame, in which the failure liability of is expected to be minor.
member reliability. That is, a greater risk a member depends on its position within Finally, the analysis requires dΨ/dx to
warrants higher safety levels. Secondly, for the structure. be evaluated, but direct values of Ψ are not
Table 1 V
alues for the coefficient of variation
of loads ζF and material strength ζG 3
reported in the literature 1 4
3 000
5
Loads ζG
Dead 0.1a,b
2
Imposed 0.20–0.25a,b,c
6 @ 3 000
Wind 0.25–0.52a,b,c
Elevation
Materials ζF
Figure 4 T est structure made up of a Pratt-type truss; the five member types (numbered) are
Steel 0.10–0.13d,e
identified by the line colours
Reinforced concrete 0.20–0.25f,g
a – Holický & Retief 2005; b – Ellingwood 1982; Table 2 Section design and reliability analysis parameters for the truss test structure
c – Retief & Dunaiski 2010; d – Kemp et al 1987;
Member design
e – Galambos 1990; f – Holický et al 2010;
g – MacGregor 1983 Section
Member Fcompr (max) Ftension(max) Areq Asection
(circular
type (kN) (kN) (mm2) (mm2) hollow)
determined. The revenue I earned from use
1 430 – 2 260 2 270 165.1 × 4.5
of the structure is therefore not needed,
and will not be set. 2 – 470 1 470 2 270 165.1 × 4.5
3 510 – 2 050 2 270 165.1 × 4.5

Determinate steel truss bridge
4 135 330 1 360 1 370 101.6 × 4.5
The first test structure is a Pratt-type
determinate truss (Figure 4), in which fail- 5 165 200 1 030 1 370 101.6 × 4.5
ure of any member results in failure of the
Structure-level reliability parameters
entire structure.
In the current analysis only the primary ζF = 0.25 ζG = 0.10 ζT = 0.26 m=1 τ = 50 years
support trusses are considered, with the Member-level reliability parameters
lateral bracing and the bridge deck exclud-
Number of
ed. These trusses support a deck of width Member type c d d/c
members
sufficient to accommodate a single vehicle
lane, so that only one vehicle would use the 1 2 R 1.5k R 2.0m 1.33 × 103
bridge at a given time. It is further assumed 2 6 R 1.2k† R 2.0m 1.67 × 103
that connections are significantly stronger
3 4 R 1.0k R 2.0m 2.00 × 103
than the members, so that only member
reliability needs to be considered. 4 4 R 0.75k R 2.0m 2.67 × 103
The bridge is designed assuming a 5 5 R 0.5k R 2.0m 4.00 × 103

travelling design load of two 150 kN point
Total: R 19.7k
loads 5 m apart (TMH7:1981). Sections are
designed according to the SANS 10162:2011 † 20% additional cost for fabrication necessary to allow erection of the bridge deck, based on the
number of joints along the bottom chord and the manufacturing cost fraction
specification, with sections of similar
1.4 1.00
4.4
1.3 0.75 least expensive
independent member type
4.2 single 1
2
Marginal benefit
member 1.2 3 0.50

4
xadjust
5
βopt
4.0 1.1 0.25

1 23 4 5
under-
designed
1.0 0
3.8 3 over-
1 4 designed most
3 000
5
0.9 –0.25 expensive
3.6 member type
2
0.8 –0.50
1 × 103 1 × 104 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 3.8 4.0 4.2 4.4 4.6
d/c β0 β0
(a) (b) (c)
Figure 5 Analysis results for the truss test structure; independent single member trend determined using m = 1 and ζT = 0.26
with structural design (3.0 to 4.5, see
slab Introduction).
500 Marginal benefit derived from improv-
beam
400 foot ing any member with reliability β 0 towards
beam its βopt value is shown in Figure 5(c).
top
column slab Marginal benefit decreases with increasing
β 0, becoming negative for β 0 > βopt . At a
3 @ 4 000
300
500
450 given β 0 , the marginal benefit is higher for
more affordable members, corresponding
middle
column 400 to an increase in ψ with d/c values.
A A
Multi-storey reinforced
500 concrete building
bottom column The second test structure is a three-
foot storey frame building constructed from
reinforced concrete, consisting of a series
8 000 2 700 of slab-beam-columns, and supported
Elevation Section A-A by square footing shallow foundations
(Figure 6).
Figure 6 Test structure of a three-storey frame; the six member types are labelled The structure is analysed assuming an
imposed load of 4 kN/m2 on each floor
loading grouped into five member types for reliabilities (βopt) that maximise the and a peak wind speed pressure of 1.3 kPa,
ease of construction. Analysis results and expected benefit correspond closely to the as specified by SANS 10160:2011. The six
section design are summarised in Table 2. independent member values predicted from reinforced concrete member types are
Estimated costs and liabilities for each their liability-cost ratios (d/c values), as designed according to SANS 10100:2000.
member group are given in Table 2. Costs determined using Equation 23. Analysis results and section design are
are assumed to be R16k/tonne, based on a The range of xadjust values needed summarised in Table 3.
cost breakdown of 40% material, 40% fabri- to adjust members to βopt falls between Estimated costs and liabilities for the
cation and transport, and 20% construction 1.2 and 0.9 (Figure 5(b)) for reasonable member groups are also given in Table 3.
and labour (McNamara 2017). Taking both design reliability values β 0 associated Costs are determined using a unit cost of
primary support trusses into account, the
total cost of the structure is calculated to Table 3 S ection design and reliability analysis parameters for the three-storey frame test structure
be R39.4k. Member design (30 MPa concrete)
To determine the associated liability,
FM1(max) FM2(max) Fcompr (max) Dimensions
the legal damages due to injury/death Member type Reinforcing
(kNm) (kNm) (kN) (mm)
of individuals using the structure must
be accounted for. Given the size of the slab 21 –38 – 300 × 8 0002 Y10-150
bridge, it is unlikely that more than one beam 160 –180 – 400 × 500 3Y25 B&T
vehicle would be on the bridge if it were
to fail. Damages should then be expected foot 520 – 1 240 500 × 2 7002 Y16-150
to be payable for two persons, at a sum of top column 50 50 360 400 × 400 8Y12
R2.0m. This value is based on an assess-
middle column 110 140 760 450 × 450 12Y20
ment of reasonable damages for injury/
death in the context of South African law, bottom column 390 0 1 240 500 × 500 12Y32
performed by Koch (2011); occupants are
Structure-level reliability parameters
assumed to be one breadwinner and one
non-breadwinner. ζF = 0.25 ζG = 0.20 ζT = 0.32 m=1 τ = 50 years
Starting from the assumption that the Member-level reliability parameters

structure is of sound design, i.e. every Number of
member satisfies a minimum design reli- Member type c d d/c
members
ability β 0, the optimal reliability value
slab 3 R100k R10m 1.00 × 102
βopt for each member type is determined
numerically by finding the β 0 value for beam 12 R7k R10m 1.43 × 103
which dΨ/dx = 0 at x = 1. From this, the foot 4 R15k R30m 2.00 × 103
required resistance adjustment f actor
top column 4 R4k R10m 2.50 × 103
xadjust is determined using Equation 24,
and marginal benefit ψ as defined in middle column 4 R5k R20m 4.00 × 103
Equation 18. bottom column 4 R6k R30m 5.00 × 103
Results of the analysis are shown in
Total: R504k
Figure 5. As seen in Figure 5(a), member
4.6 1.4 2.0
bottom column highest
4.4 middle column d/c
1.3 bottom column 1.5 member
top column type
4.2
middle column
Marginal benefit
independent 1.2 1.0
4.0
single member top column foot
xadjust
βopt
3.8 1.1 0.5 beam

slab under-
3.6 designed foot slab
beam 1.0 0
foot over- beam
3.4 top column designed lowest
middle column 0.9 –0.5 d/c
3.2 bottom column slab member
types
3.0 0.8 –1.0
1 × 102 1 × 103 1 × 104 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 3.0 3.2 3.4 3.6 3.8 4.0 4.2 4.4 4.6
d/c β0 β0
(a) (b) (c)
Figure 7 Analysis results for the three-storey frame test structure; independent single member trend determined using m = 1 and ζT = 0.32
R6k/m3 (Robberts & Marshall 2010, adjust- reliabilities βopt correspond closely to the the coefficients of variation can provide an
ed for inflation), based on a breakdown of predicted independent member values indication of the optimal design reliability
35% concrete, 35% reinforcing steel, 15% (Figure 7(a)). xadjust values needed to adjust of a member.
formwork and 15% labour. The total cost of members to βopt vary from about 1.25 For South African design standards,
the structure is calculated to be R504k. for β 0 = 3.0 to 0.95 for β 0 = 4.5. Marginal Milford (1988) recommends β 0 = 3.0 for
In contrast to the truss example, this benefit again decreases with increasing βopt ductile failure modes and β 0 = 4.0 for
structure has a unique set of supported values, being negative for β 0 > βopt. At a brittle failure modes. Ductility allows for
members associated with each individual constant β 0 , members with higher d/c val- load redistribution and provides time for
member. Failure of a given member is ues have higher marginal benefit; however, remedial action to be taken prior to col-
assumed to compromise only its supported there is no longer a simple correlation with lapse. The reliability analysis of the two
members (as described via the matrix Γ). member cost. test structures considered here assumes
For example, failure of a beam would that failure leads to collapse, implying that
compromise its supported slab, but leave its either brittle failure occurs, or no remedial
supporting column unaffected. Failure of a DISCUSSION action had been taken when ductile failure
column would compromise the members it started. In this sense, the calculated βopt
supports: two beams, their supported slab, Basic trends values are broadly consistent with the
and recursively the column on the next The close correspondence of member val- recommended β 0 value for brittle failure of
level up with all the members it supports. ues to the independent member βopt values Milford (1988).
The number of compromised members predicted from their d/c ratios (Figures 5(a) The direct relationship between
due to failure of a column will therefore and 7(a)) indicates that the relationship for marginal benefit ψ and the d/c ratio of a
increase towards the base of the structure: optimal reliability of independent members member can be understood as follows. The
if a ground floor column fails, all columns (Equation 20) can be applied for members marginal benefit decreases with increasing
and their associated beams and slabs above in structures, provided that the liability d β 0 (Figures 5(c) and 7(c)) and is zero for
it will no longer have sufficient support; if of the member in question is much greater β 0 = βopt. Members with higher βopt values,
a column in the topmost storey fails, only than the cost of its supported members. and thus higher d/c ratios, will therefore
two top beams together with their sup- For example, in the truss test structure have higher marginal benefit values at a
ported slab are compromised. d = R2.0m for all members, which is much given β 0 value.
Assuming the building to be residential, larger than the cost of repair, as the cost of This suggests a simple guideline for
with a normal-use occupation of 10 persons the entire structure is R39.4k. In the frame, improving the expected benefit of an exist-
per storey, legal damages due to serious the liability of the slab is R10m, while the ing structural design: for members that are
injury or death is estimated at R10m per cost of repair is only the cost of the slab, under-designed relative to βopt, the greatest
floor (Koch 2011). R100k; the liability of the bottom column impact on expected benefit is obtained by
Starting from the assumption that is R30m, while the cost of replacement is strengthening the member with the high-
every member in the structure satisfies a R357k. This comparison ignores demoli- est d/c value; for over-designed members,
minimum target reliability β 0, the optimal tion costs, and loss of revenue is also not the greatest impact on expected benefit is
reliability βopt is determined numerically accounted for in the model. Both these obtained by reducing the size of the mem-
for each member type, together with the factors will increase the liability expenses bers with the lowest d/c value.
required resistance scaling xadjust and the by amounts comparable to or smaller than This principle is illustrated in the two
marginal benefit ψ. estimates used here. examples. In the truss, all members have
Results of the analysis are presented in This result suggests that knowledge the same liability, so that the highest d/c
Figure 7. Similar to the results obtained for of the cost of a member, the liability value corresponds to the least expensive
the truss test structure, optimal member implications of failure, and estimates of member type. Strengthening this member
type would be the most affordable way greater than the common 50 to 100-year various assumptions made. Most notable
to increase the expected benefit. In the design lifetimes. of these are: (a) the normal distribution
frame, the columns have similar costs, but The analyses performed in this work of applied loads and material resistances,
very different liabilities. Strengthening the focus on the member level only; system (b) the power-law scaling of member cost
columns with the greatest liability, i.e. the level reliability is not accounted for, and is with resistance adjustment, (c) the focus on
bottom columns, thus reduces the risk by assumed to be a cumulative, linear result reliability solely at the member level, and
the greatest amount, bringing about the of the individual members. Similarly, it (d) neglecting connections. Exploring each
largest increase in expected benefit. is assumed that failure liabilities are the of these assumptions will be the focus of
cumulative result of individual member future research work.
Effect of assumptions failure liabilities. Member interaction
As shown by Equation 23, m affects effects at system level are therefore not
βopt by acting as an adjustment to d/c. accounted for. For example, redundancy ACKNOWLEDGEMENT
As seen in Figure 3, the effect is rela- in the structure due to ductility and load Discussion with Ammanuel Gebremeskel
tively small: assuming m = 1 results in redistribution can limit the cumulative of the Southern African Institute of Steel
βopt about 2% lower than for m = 1/2. For increase in liability, and thus reduce βopt Construction (SAISC) was of great value to
the two test structures it was assumed for some members in the structure. the research presented here.
that m = 1, as the nature of failure is not
specified in either example. As shown in
Equations 13–15, m values can be expected CONCLUSION REFERENCES
to vary between 1/2 and 1. The effect of A model was developed for the adjusted Ang, A H & Tang, W H 1984. Probability Concepts
m ≠ 1 would therefore be at most a 2% reliability of a member strengthened by a in Engineering Planning and Design. Vol II.
increase in the values of βopt . multiplicative factor. The model allows the Decision, Risk, and Reliability, 1st ed. New York:
The relationship for the reliability of the degree of strengthening required to max- Wiley.
adjusted member (Equation 10) was derived imise overall expected benefit derived from Ellingwood, B 1982. Probability-based criteria for
assuming that both the internal force F the structure to be determined together structural design. Structural Safety, 1: 15–26.
and the member resistance G are normally with the marginal benefit. This provides an Freudenthal, A M 1947. The safety of structures. ASCE
distributed. This assumption allows the indication of the greatest change in benefit Transactions, 112: 125–159.
simple form of Equation 7 upon which brought about by strengthening a given Freudenthal, A M & Gumbel, E J 1953. On the
the derivation is based. Depending on the member type. statistical interpretation of fatigue tests. Proceedings
nature of loading, this assumption does The model was first applied to a of the Royal Society of London. A: Mathematical,
not always hold. The Gumbel distribution single independent member to derive a Physical and Engineering Sciences, 216: 309–332.
is generally used for wind loading, while relationship expressing the reliability βopt Galambos, T 1990. Systems reliability and structural
imposed loading is often represented via a required for optimal benefit as a function design. Structural Safety, 7: 101–108.
log-normal distribution (Retief & Dunaiski of the liability-cost ratio d/c for the mem- Holický, M & Retief, J V 2005. Reliability assessment
2010). These distributions are all positively ber. Other parameters in the equation, i.e. of alternative Eurocode and South African load
skewed (asymmetrical with positive tails), the coefficients of variation and the cost combination schemes for structural design. Journal
so that transformation to normal space scaling power, have only a minor effect of the South African Institution of Civil Engineering,
(via the Rosenblatt transformation equa- on βopt. 47(1): 15–20.
tions, Ang & Tang 1984) would distort the Next the model was applied to analyse Holický, M, Retief, J V & Wium, J 2010. Partial
F–G line to be concave down. The result the reliability of members in two example factors for selected reinforced concrete members:
would be that, for a given β 0 value, the cor- test structures: a determinate steel truss Background to a revision of SANS 10100-1. Journal
responding β1 would be somewhat smaller and a three-storey reinforced concrete of the South African Institution of Civil Engineering,
than predicted by Equation 10. βopt values frame. Reliability analysis of members in 52(1): 36–44.
determined in this work can therefore both structures reveals that the indepen- ISO (International Standards Organization) 2014.
be viewed to represent upper bounds on dent member d/c – βopt relationship is ISO 2394. General Principles on Reliability for
the values for non-normally distributed applicable to members that form part of Structures. Geneva: ISO.
parameters. a structure, while marginal benefit was JCSS (Joint Committee on Structural Safety) 2008.
The effect of time on the value of found to be greatest for members with the Risk Assessment in Engineering: Principles, System
money via interest rates and inflation is highest βopt and d/c values. Representation & Risk Criteria. Stellenbosch: JCSS.
ignored in the model. This effect would The relationship therefore provides a Available at: http://www.jcss.ethz.ch.
enter the model in the revenue derived guideline for the improvement of existing Kemp, A R, Milford, R V & Laurie, J A P 1987.
from the structure, and would also be structural designs. Structures that are Proposals for a comprehensive limit states
needed if the probability of failure is not under-designed with respect to βopt can be formulation for South African structural codes. The
taken as constant during the lifetime of most effectively improved by strengthening Civil Engineer in South Africa, 29(9): 351–360.
the structure. The latter would be the case members with the highest d/c values; over- Koch, R 2011. Damages for personal injury and death:
if the lifetime was similar to the mean designed structures are most effectively Legal aspects relevant to actuarial assessments.
time to failure. However, for the failure improved by reducing member sizes with South African Actuarial Journal, 11: 111–133.
probabilities associated with structural the lowest d/c values. MacGregor, J 1983. Load and resistance factors for
members, mean time to failure is in the While the model provides useful trends, concrete design. American Concrete Institute
order of 103 to 105 years, which is far its quantitative value is limited by the Journal, 80(4): 279–287.
McNamara, P 2017. Costing steelwork. Building Retief, J V & Dunaiski, P E 2010. The Limit States for Buildings and Industrial Structures, Parts 1–8.
Magazine, Steelwork Feature, April. Available at Basis of Structural Design for SANS 10160-1. Pretoria: SABS Standards Division.
http://www.steelconstruction.info. Background to SANS 10160 – Basis of Structural SANS (South African National Standard) 2011.
McQuarrie, D 2003. Mathematical Methods for Design and Actions for Buildings and Industrial SANS 10162. The Structural Use of Steel, Parts 1
Scientists and Engineers. Sausalito, CA: University Structures, Parts 1 and 2. Stellenbosch: SUN and 2. Pretoria: SABS Standards Division.
Science Books. MeDIA. TMH7 (Technical Methods for Highways) 1981. Code
Milford, R V 1988. Target safety and SABS 0160 load Robberts, J & Marshall, V 2010. Analysis and Design of Practice for the Design of Highway Bridges and
factors. The Civil Engineer in South Africa, 30(10): of Concrete Structures. Johannesburg: Cement and Culverts in South Africa. Pretoria: Committee of
475–481. Concrete Institute. State Road Authorities.
Pugsley, A 1955. Report on structural safety. Structural SANS (South African National Standard) 2000.
Engineer, 33: 141–149. SANS 10100. The Structural Use of Concrete, Parts 1
Rackwitz, R & Fiessler, B 1978. Structural reliability and 2. Pretoria: SABS Standards Division.
under combined random load sequences. Computers SANS (South African National Standard) 2011.
and Structures, 9: 484–494. SANS 10160. Basis of Structural Design and Actions
TECHNICAL PAPER
The relationship between
project performance of Journal of the South African
emerging contractors in
ISSN 1021-2019
Vol 60 No 4, December 2018, Pages 25–33, Paper 1664
government infrastructure FATE T MUZONDO holds a BSc in Agricultural

Engineering and an MSc in Civil
projects and their experience

Engineering. She has worked as an engineer
and project manager bringing infrastructure
to subsistence farmers in Mpumalanga
and technical qualifications

Province of South Africa. She currently
works in the regulatory environment at one
of Australia’s largest banks. Her areas of
interest are in the development of South African emerging contractors.
She is a member of the Australian Institute of Engineers and the South
F T Muzondo, R T McCutcheon African Institute of Agricultural Engineers, and is currently completing her
PhD at the University of the Witwatersrand.
Contact details:
4/66 Mandurah Terrace
Various studies have been conducted to investigate the reasons for the comparative failure
Mandurah
of small contractors. Many of these studies have found the reasons for failure to be primarily Western Australia 6210
related to factors that are beyond the control of the contractor or business management. Little T: +61 401 341 879
attention has been paid to technical factors. This study sought to investigate the correlation E: fatemohlala@yahoo.com
between the performance of emerging contractors in government infrastructure projects to

PROF ROBERT McCUTCHEON Pr Eng,
their technical qualifications and experience. An archive research methodology was adopted
Professor Emeritus and Honorary Professor
where contractor performance information was collected on 30 CE and GB projects conducted in the School of Civil and Environmental
in Mpumalanga Province. The project data was then used alongside contractor qualification Engineering at the University of the
and experience data to investigate their relationship. When evaluating the qualification level, Witwatersrand, is Employment Creation and
Development Specialist at Malani
it was found that contractors with higher qualifications show better performance. It was also
Padayachee & Associates (Pty) Ltd,
found that contractors with more technical qualifications perform better than those without. Consulting Civil and Structural Engineers,
This study also concluded that contractors with more years of experience in the construction Randburg. In addition he is Research Associate in the Centre of Full
industry show better project performance. It is recommended that a much broader Employment and Equity at the University of Newcastle, Australia. His
work focuses on skills development, employment creation and
investigation must be carried out to examine to what extent these conclusions are applicable programme management in the field of public works, with special
throughout South Africa. If so, there are important implications for the modification of existing reference to labour-intensive construction and small contractor
procurement policy and procurement practice. development. He is currently engaged in research, consulting and
implementation in this field in various countries.
Contact details:
PO Box 43
introduction performance in a particular project. As
Wits 2050
Economic development can be measured by such, it is conceivable that each project South Africa
the physical development of infrastructure would have measures appropriate to its T: +27 83 629 4783
such as bridges, roads and buildings, and goals. Determining project goals and the E: roberttmccutcheon@gmail.com
job creation (Alzahrani & Emsley 2013), indicators to be used to measure project
which are mainly offered by the civil performance is vital to assessing and
engineering and general building construc- quantifying project performance. This
tion sectors. Emerging contractors, or then forms a basis for determining the
“small, medium and micro enterprises” factors that affect project performance.
(SMMEs), are essential for job creation and Many studies have been conducted to
poverty alleviation in African countries determine the factors that affect contrac-
(Tushabomwe-Kazooba 2006; Okpara & tor performance in developing countries
Wynn 2007; Okpara & Kabongo 2009). (Sweis et al 2014). A study done by Faridi
Therefore, the development of emerging and El-Sayegh (2006) in the United Arab
contractors contributes to the development Emirates (UAE) shows that shortage of
of South Africa. Given that the primary cli- skilled labour, poor supervision and site
ent of these contractors is the government management, inadequate leadership and
(CIDB 2011a), there is a need to investigate equipment failure have contributed to
their project performance in government delays in construction projects. According
infrastructure projects. to Hanson et al (2003) contractor project
It is evident that there are a number performance in South Africa is affected by
of performance measures, and it may poor workmanship and contractor incom- Keywords: Emerging contractors, technical competence, experience,
be that not all can be used to measure petence, while Gharakhani et al (2013) project performance, policy
Muzondo FT, McCutcheon R. The relationship between project performance of emerging contractors in government infrastructure projects and their
experience and technical qualifications. J. S. Afr. Inst. Civ. Eng. 2018:60(4), Art. #1664, 9 pages. http://dx.doi.org/10.17159/2309-8775/2018/v60n4a3 25
found that reputation affects client satisfac- 1976; Van Witteloostujin 1998; Temtime cost, quality, health and safety, site and
tion and, therefore, perceived contractor & Pansiri 2004). The high failure rate of sub-contractors (CIDB 2013).
performance. SMMEs in South Africa is of great concern. Thus, project performance can be
This research was conducted to For economic growth purposes in South defined as the actions taken through
explore the relationship between technical Africa, emerging contractor businesses the life cycle of a project to meet pre-
qualifications and experience of contrac- should perform on a project and business determined goals. These goals may be one
tors within the construction industry, level. Key performance indicators need of the seven key performance indicators
specifically in the South African public to be identified and defined in order to identified by the CIDB (2013), those pro-
sector and contractor project perfor- evaluate emerging contractor performance. posed by Cheung et al (2004), or a combi-
mance. The investigation was conducted The outcome will then be used to develop nation thereof.
in Mpumalanga Province with 30 public recommendations for improving contractor
sector projects carried out by CE and GB project and business performance in order Factors related to project performance
contractors from CIDB grades 1 to 7. This to aid sustainable development. According to Van Wyk (2003), Mbande
research, therefore, argues that in projects, (2010) and Milford (2010), one of the
performance is directly related to the tech- Definition of “emerging contractor” challenges facing the South African
nical competence of the contractors and There is no general definition of an emerg- construction industry is the lack of insti-
their level of experience in the construction ing contractor (Storey 1994; Eyiah 2001). tutional capacity within the public sector
industry. It is expected that, in order for an In the South African context, an “emerging or government organisations. Operational
emerging contractor to develop on a busi- contractor” can be defined as a “person or skills which include project management
ness level, there is a need to perform at the enterprise which is owned, managed and and business skills such as planning and
project level (Mohlala 2015). Completion of controlled by previously disadvantaged financial accounting are crucial for the suc-
a project in the specified time is considered persons and which is overcoming business cess of a business (Thwala & Mvubu 2005).
to be a major criterion for the measure- impediments arising from the legacy of Mofokeng (2012) also found that small and
ment of project success (Rwelamila & Hall apartheid” (CIDB 2011b). These enter- medium construction companies lack busi-
1995), which measurement was used in this prises are also termed “small construction ness and managerial experience. Previous
research to assess project performance of enterprises” and “small-scale contractors”. literature (Fredland & Morris 1976; Yusoff
emerging contractors. Emerging contractors are generally charac- 1995; Jo & Lee 1996; Lin 1998; Jaafar &
terised by limited capital resources, plant Abdul-Aziz 2005) has shown that manage-
and equipment, and managerial support, rial expertise and experience are vital for
literature review all of which affect their ability to acquire the success of an enterprise and have a sig-
A literature survey was conducted with skilled labour and employ professionals nificantly positive relationship to the per-
the aim of finding out what work had been (Eyiah 2001). formance of the enterprise. According to
done in the area of emerging contractor The CIDB does not classify them Croswell and McCutcheon (2001), some of
performance in public sector projects, with according to their financial capabilities the difficulties facing emerging contractors
attention drawn to the South African con- (Mohlala 2015). As such, “emerging” is include lack of entrepreneurial, managerial,
text. The literature was analysed in order not necessarily a reflection of financial technical and administrative expertise.
to answer the first primary research ques- capability. Much of the literature related to con-
tion, which is: What are the factors that This study, therefore, defines “emerging tractor performance and the challenges
affect emerging contractor performance in contractors” as: facing the South African construction
government infrastructure projects? The Small to medium contracting enterprises industry in general focus on the business
secondary objective was to establish the that are owned by individuals previously and financial functions, with some also
limitations of previous work done in the disadvantaged by the apartheid system of citing clients (government organisa-
study area in order to establish a methodol- the pre-1994 South Africa and are regis- tions) as a challenge. However, not much
ogy for the empirical study. tered with the CIDB. emphasis is placed on the technical aspect.
Even though it is mentioned in the study
Success rate of emerging contractors Definition of “project performance” conducted by Croswell and McCutcheon
Small businesses, or SMMEs such as Performance measurement is a systematic (2001), not many studies in South Africa
emerging contractor companies, contribute process of assessing and quantifying past have conducted an in-depth investiga-
significantly to economic growth and behaviours and activities (Neely 1998). tion on the importance of the technical
job creation in South Africa (Van Eeden Project performance measurement is essen- aspect. It has, however, been explored on
et al 2003). The SMME sector contrib- tial in order to determine whether or not an international level, as presented in the
utes approximately 67% of employment project goals have been met for both the sections below.
opportunities in South Africa (DTI 2004). client and the contractor. A study conduct-
According to Pretorius (2009), 50–90% of ed by Cheung et al (2004) found that there Emerging contractor competencies
South African small businesses fail, 32% are seven main key performance indicators, Sweis et al (2014) conducted a study on
of which fail in the first seven years of which include time, cost, quality, client contractor performance at an international
operation (Nemaenzhe 2010). Small busi- satisfaction, client changes, business per- level in order to determine the most critical
ness failure can result in financial losses formance, and health and safety. According contractor performance factors in public
(Shepherd et al 2009), loss of resources to the CIDB, project performance can be infrastructure projects according to the
(Peacock 2000) and job losses (Argenti measured against the management of time, perspectives of contractors, clients and
engineers. It was found that the most com- labour force (Richter 1998). It was also of projects, preparation and approval of
mon factors affecting contractor perfor- found that Germany has a high number of drawings, poor supervision and site man-
mance are related to the contractors’ ability technically qualified construction workers agement, and shortage of manpower. The
to perform the work, that is, contractor even up to Master’s level. According to lowest ranking causes from the same study
competence. It should be noted that, due Richter (1998), this is the support system were found to be shortage or delays in
to the uniqueness of projects, it is difficult and essential factor for entrepreneurship delivery of materials, poor leadership of the
to generalise these factors. These findings and self-employment in the German construction manager or project manager,
do, however, form a basis for the need to construction industry. The lack of techni- incomplete drawings, specifications or
assess contractor performance in the South cally qualified contractors was found to documents and productivity of manpower.
African public sector. be the primary cause of low productivity Similar studies were done by Assaf et al
Competence can be defined as the ability in the United Kingdom (UK) construction (1995), and Mezher and Tawil (1998) in
or potential to perform a particular task or industry (Prais & Steedman 1986; Clarke Saudi Arabia and Lebanon, respectively.
job effectively (Rozewski & Malachowski & Wall 1996). Croswell and McCutcheon According to Mezher and Tawil (1998),
2009) or the capacity to perform a given (2001) also identified the lack of technical the most common causes of project delays
activity (Guillaume et al 2014) to achieve a expertise as a contributory factor to the include poor leadership of construction
predetermined goal to a particular standard poor performance of emerging contractors. or project management, and preparation
(Peters & Zelewski 2007). This involves the These studies highlight the importance and approval of drawings. Assaf et al
application of knowledge and skills and is of technical qualifications and experience in (1995) concluded that project delays in the
likely to vary from one organisation or con- the construction industry. Most literature construction industry are closely related to
tractor to the next (Guillaume et al 2014). in the area of emerging contractor perfor- availability and productivity of manpower,
It is clear that competence should involve mance concentrate on financial and project which are linked to skills and competencies
not only the ability to perform a given task, management competencies. However, this of contractors. Skills and competencies in
but the application of the necessary skills section suggests that technical qualifications the construction industry place contrac-
and knowledge which are acquired before and experience contribute to performance. tors in a better position for getting work
the given task. This is an indication that In addition, to ensure sustainability, institu- and, as a result, experience. This supports
there may be a need to assess the level of tions and construction companies should the proposition that successful contractor
emerging contractor competence before the pass on their skills, knowledge and experi- performance in South Africa is linked to
awarding of every contract, as it may affect ence through training of incoming members skills, competencies and experience; a fac-
the performance during the construction into the industry (Richter 1998). This can tor that has been given much attention in
phase of the project. be achieved by collaboration between the the construction industry.
According to Hanson et al (2003), government (public sector) and the private Project delays can lead to cost over-
contractor performance in South Africa sector. Richter (1998) also highlights the runs (Sambasivan & Yau 2007), low
is affected by poor workmanship and importance of training, and its relationship productivity, contract termination (Arditi
contractor incompetence. This affects the with competence and experience in the & Pattanakitchamroon 2006) and hinder
contractor’s track record or reputation and, construction industry. business development for the contractor
therefore, the eligibility for future con- (Benson 2006). According to Arditi et al
tracts. Gharakhani et al (2013) found that Project delays (1985) and Lo et al (2006), project delays
reputation affects client or customer sat- Project delays or timely completion of can lead to slow economic growth since,
isfaction. Relevant skills and expertise are projects can be used as another indicator of according to Adnan et al (2012), the
vital for emerging contractors’ performance emerging contractors’ project performance. construction industry plays a major role in
through the life cycle of a project, and from It should be noted that not all delays are infrastructure development and, therefore,
one project to the next. The acquisition of a reflection of a contractor’s competence, the growth of the economy.
competence is thus an on-going process, project management skills and perfor- It can be argued from the above studies
especially in project-based enterprises (i.e. mance. It is important to investigate the that have been conducted around the sub-
emerging contractor businesses), since each different causes of construction delays in ject of emerging contractor performance
project is inherently unique. Thus, empha- order to determine the causes. The causes that very little work has been done to find
sis should also be put on continuous learn- of project delays that are attributed to the out whether there is a correlation between
ing from one project to the next in order to contractor’s actions or lack thereof can contractors’ project performance, contrac-
improve emerging contractor competence then be used to determine the solutions tor experience and technical qualifications.
in government infrastructure projects. that can be used to limit such delays and, This investigation is aimed at finding the
as a result, improve contractor perfor- relationship between emerging contractors’
Technical qualifications and experience mance. Project delays that are attributed to project performance in public sector pro-
A study by Sweis et al (2014) found that government organisations (the client) and jects, and their qualifications and experi-
clients and engineers attribute contractors’ the built environment professionals (engi- ence in Mpumalanga Province.
poor performance to lack of technical neers, consultants and the client’s agents)
professionals within their organisa- should also be noted.
tions. According to a study conducted in Faridi and El-Sayegh (2006) found that methodology
Germany, a combination of vocational the most prevalent causes of project delays, A literature review was performed in
training and formal qualifications is a as ranked by consultants and contractors order to clearly state and substantiate the
great strength of the German construction in the UAE, are inadequate early planning problem, which is that there is generally
low performance of emerging contractors data was collected from the database of throughout the life cycle of each project.
in the public sector (Zulu & Chileshe one of the government organisations (in It is, therefore, assumed that all projects
2010). The literature review was used to Mpumalanga Province) that run projects that were completed attained the qual-
determine the causes of low performance which require the appointment of CE ity standards and client satisfaction
or failure of emerging contractors in gov- and GB registered contractors. This data measures. This then ruled out lack of
ernment infrastructure projects. The body included tender and project documents, quality and client dissatisfaction.
of knowledge helped to define “emerging and contractor contact details. The data ■■ The beneficiaries of the government
contractor” and “performance” and this collected in Part 1 was also used to quan- infrastructure projects sign a comple-
helped to determine the gaps in literature, titatively represent and analyse the perfor- tion certificate at the end of every
the types of questions not addressed by mance indicators. project as an indication of satisfaction
literature and the type of empirical data The contact information gathered from and agreement of all work done by the
that was collected in order to address the Part 1 was used to contact the contractors contractor. Thus, the presence of a
unanswered questions. to gather Part 2 of the data, which included completion certificate usually indicates
An archival research methodology was Curriculum Vitae (CVs) and company pro- client and beneficiary satisfaction.
utilised to collect the empirical data used for files. This was then used to confirm what ■■ Timely completion of projects is com-
this research, where contractor performance was proposed in the research statements monly regarded as the major indicator
information based on 30 CE and GB pro- (i.e. determine whether or not there is a of project success (Rwelamila & Hall
jects conducted in Mpumalanga Province relationship between project performance 1995).
between 2011 and 2013 was collected. One and experience and technical qualification). Although cost and quality are considered
of the advantages of this research method to be two of the most important compo-
is that it allows for collection of data that nents of project management, for the rea-
spans over a long period of time, and allows results and analysis sons given above they were not used in the
for a broader view of trends. This study also study as measures of project performance.
acknowledges the disadvantages of using Description of data However, it was possible to investigate the
archival data, one major one being that Data on thirty projects that were conduct- “time factor” in greater detail.
the researcher has no control over how the ed between 2011 and 2014 was collected, Mohlala (2015) developed the follow-
data was archived, and as such, data may be representing a total of 25 contractors. An ing approach for a deeper consideration
incomplete and may fail to address certain email was sent to these 25 contractors of “time”. In order to assess project
issues. This implies that the findings in this requesting their company profiles and performance, that is the timely comple-
study cannot be regarded as generalisable the CVs of the core staff of the company tion of projects, which is the performance
to all contractors in South Africa, thus (mainly the owner). Contact details, CIDB indicator used in this research, the degree
recommending that further study be done grades, payment and completion certifi- of delay was calculated. This provides a
on a larger scale involving projects in the cates, and project start and completion uniform measure for performance and
whole country. dates were collected for all 30 projects. allows for comparison of the different
A positivist philosophy was used for this Therefore, 100% of Part 1 of data collection contractors and projects with different
research, and focused on the factors of fail- was collected. A total of 64% of the 25 con- construction periods. The degree of delay
ure and success of emerging contractors in tractors responded to the request to submit (DoD) is calculated as a percentage over
government infrastructure projects in the their CVs and company profiles for Part 2. the specified construction period as follows
South African construction industry. The (Mohlala 2015):
study was based on facts from the body Project performance
of knowledge and from the empirical data The four most commonly used project a–b
Degree of delay = × 100
that was collected. The research statements performance indicators are time, cost, b
propose that project performance is related quality and client satisfaction (Harrison c
= × 100
to technical qualifications and experience & Lock 2004; Turner 1999; Cheung et al b
prior to a particular project. 2004). The four indicators were evaluated
A qualitative approach was used for in relation to the government institution Where
the collection of data in order to study under study in order to determine which a = actual construction period
the indicators of performance and fac- would be the most suitable indicator for b = specified construction period
tors affecting performance of emerging this study. Mohlala (2015) found that: c = time overrun
contractors in government infrastructure ■■ The government institution under study
projects. “Qualitative data sources include does not impose penalties that result For example, the DoD for a three-month
observation and participant observation in financial losses for the contractor project that took six months to complete is:
(fieldwork), interviews and questionnaires, and the contractors generally to com-
documents and texts, and the researcher’s plete their work within the tendered 6–3
Degree of delay =× 100
impressions and reactions” (Myers & amount. Therefore, there are rarely any 3
Avison 1997). cost overruns, and as such, cost could = 100 %
not be used as a measure of project
that is, the project went
Data collection performance. over the specified construc-
The data was collected in two parts and ■■ The engineers from the government tion period by 100% (three
was project-based. Part 1 of the empirical institution conduct quality checks months in this case)
To further explain this factor, projects that non-technical qualifications. The technical
0–50% have a DoD of zero (0%) are projects that qualifications that were found were in the
∞ DoD
16% 10% have been completed on time (within the civil and agricultural engineering, quantity
specified construction period). surveying, instrumentation and construc-
51–100% For projects that were incomplete or tion fields.
>200% DoD had been abandoned at the time of data
DoD 30%
27% collection, the delay period and degree of Relationship between qualifications
delay are represented by the symbol ∞. and project performance
Figure 1 shows that only 10% of the pro- Further investigation was carried out
151–200% 101–150% jects had a DoD of less than 50%. Figure 1 to determine the relationship between
DoD DoD
7% 10% also shows that 30% of the projects had the types and levels of qualification, and
a DoD that falls between 51% and 100%. project performance. For analysis, the
Figure 1 P
ercentage of projects to the Thus, 60% of the contractors exceeded minimum, maximum and average DoDs
corresponding DoD range the specified construction period by over for the different types of qualifications and
double the time. qualification bands were compared.
The latter, together with the fact that
Unknown Matric 27% of the projects have a DoD of over Qualification type and
11% 11%
200% and that 16% of the projects did not project performance
even reach a commissioning or close-out The results show that projects with the
stage (DoD is ∞), shows that project per- highest DoD and uncompleted projects were
Honours/ National
Certificate formance in government infrastructure undertaken by contractors with non-techni-
Honours
equivalent 34% projects is a cause for concern. cal qualifications. This supports the initial
33% This research is aimed at exploring and proposition that the possession of technical
demonstrating the relationship between qualifications will result in better project
Diploma
11% project performance and the contractors’ performance when compared to contractors
technical qualifications and experience. with non-technical qualifications.
Figure 2 P
ercentage of contractors to their Therefore, the projects that have a DoD Figure 4 shows that project performance
qualification bands range of 0–50% need to be studied at great improves with the possession of technical
length and compared to the projects with a qualifications within the organisation.
higher DoD. The study found that the lowest DoD in
Unknown the results was from a project undertaken
11%
Contractor educational profile by a contractor with a non-technical quali-
Technical
qualifications The highest qualifications within the fication. However, this may be an indica-
44% organisation were extracted from the CVs tion that project success requires a number
that had been gathered from the respond- of factors or prerequisites, which this study
ents (i.e. the highest qualifications of the recommends for further investigation. The
Non- core staff). maximum and average DoD from contrac-
technical Figure 2 shows the proportions of the tors with technical qualifications are both
qualifications
45% qualification bands as the percentage of the lower than those of contractors with non-
respondents. technical qualifications. This implies that,
Figure 3 P
roportion of technical to A total of 34% of the contractors have from the sample considered, on average
non‑technical qualifications National Certificate qualifications from or in general, contractors with technical
institutions such as FET colleges. It was qualifications perform better, in terms of
found that some of the respondents had time, than contractors without technical
600 university degrees; those who went to uni- qualification.
500 versity either attained diplomas, honour’s
500
degrees or honour’s equivalent. This means Qualification level and
400 300 that none of the respondents attained project performance
qualifications between diploma level and The results show that the lowest qualifica-
DoD (%)
300
honour’s level. Honour’s equivalent degrees tion level (matric) has the poorest project
200 170 are four-year degrees, such as engineering, performance and the highest qualification
134 whereby one can continue from that degree level shows better project performance. The
100 67 level to a master’s degree. results also show that the DoD decreases as
25
0
Figure 3 shows the proportion of qualification level increases, which shows
Technical Non-technical respondents who have qualifications that that project performance improves with
Qualification type are in the technical or construction fields the acquisition of higher qualifications (as
Minimum Maximum Average to those with non-technical qualifications. demonstrated in Figure 5).
The qualifications shown on the CVs Apart from technical qualifications,
Figure 4 C
omparison of project performance received from the respondents were the level of qualifications is also essential
by qualification type divided into two types – technical and for project performance. The possession
supervision. Other contractors and their
600 core staff worked in other fields that are
500 not related to the construction industry,
500
420 which were found to be teaching, sales,
400 administration, customer services, and
DoD (%)
300 public relations.

300
210
200 178 Relationship between experience
133 133 133 and project performance
100
100 68
33 25
0
Personal experience and
Matric National Certificate Diploma Honours project performance
Qualification level It was expected that contractors with a
Minimum Maximum Average staff complement who have personal expe-
rience in the construction industry will
Figure 5 Comparison of project performance by qualification level generally show better project performance.
The results, however, do not reflect that.
Figure 7 demonstrates the relationship
Unknown 350 between personal experience and project
11% 300
300 performance. Further analysis was done
to determine the reasons why the results
250
Average DoD (%)
contradicted the initial expectation. This

Related
Non-related 56% 200 was done by comparing the experience of
33% 167
150
the contractors’ staff complement to the
contractors’ qualifications per project.
100 79 The maximum DoD observed from
71
Figure 6 P
roportion of contractors with 50 the contractors with staff who have
construction-related experience construction-related experience is greater
0
1–5 6–10 11–15 16–20 than 500% with some incomplete projects.
No of CE and GB related projects These contractors were also found to have
600 no technical qualifications. This may be
500 Figure 8 R
elationship between company the reason for the poor performance and
500
experience and project performance may mean that technical qualifications are
400 (number of projects conducted) more critical to project performance than
personal experience. This assumption is,
DoD (%)
300
even without formal qualifications. The however, tentative and requires further
182 200 type of experience and the years of experi- investigation of other factors that affect
200
ence are important measures of compe- project performance, which were outside
100 71 86
tence. It is assumed that CIDB grading is the scope of this research.
25
0
a measure of experience, since registration
Related Not related and progression from one grade to another Company experience and
Personal construction requires evidence of previous work done project performance
experience (CIDB 2011a). This research, therefore, Business experience is necessary to build
Minimum Maximum Average compares the project performance of con- competence; it also builds client confidence
tractors by CIDB grades as well. This also in the ability of the business to achieve
Figure 7 R
elationship between personal supplements the data where contractors’ pre-set goals. This research explored the
experience and project performance CVs are not available, since the CIDB grad- relationship between emerging contrac-
ing of all contractors were collected from tors’ business experience and project
of technical qualifications in the construc- the CIDB website. performance. Projects undertaken by the
tion industry can be used as a measure or Figure 6 shows the proportion of con- respondents were evaluated to determine
evidence of contractor competence. tractors with staff who have experience in the types of projects each business has
It is worth noting that none of the construction-related fields to those who do undertaken in the civil engineering and
respondents have any project management not. general building-related fields.
qualifications. This may be another reason The construction-related positions Figure 8 shows the relationship between
for the poor performance of the contrac- that were held by the core staff of the contractors’ company experience in CE and
tors in all the projects they undertake. company before establishment or joining GB projects, in terms of number of projects
of the respective contractor companies conducted, to project performance.
Contractors’ experience profile were considered. The construction-related The results show that companies
Literature claimed that experience provides fields were civil engineering, agricultural which have undertaken more projects have
competence in the construction industry, engineering, site management and site lower DoDs and, therefore, better project
250 500
221 439
200 400 384
331
Average DoD (%)
Average DoD (%)

311
150 133 300 258
100 200
66
50 100 85
0
0 0
1–5 6–10 11–15 1 2 3 4 5 6 7
Years in operation CIDB grade
Figure 9 R
elationship between years in
operation and project performance Figure 10 Relationship between CIDB grading and project performance
performance. This supports the initial propo- in the construction industry. However, little research work has been carried out to
sition that the experience of the construction it would also be expected that the longer determine the correlation between contrac-
company as a business or organisation in the company is in operation, the more tor performance, and contractor experience
construction-related fields prior to a particu- projects they would be conducting. Since and qualifications, especially in the South
lar project will result in better performance the average DoD of contractors with one to African context. This investigation was
in that project. Further analysis was carried five years’ experience was predominately aimed at finding the relationship between
out, and it was found that the contractor with lowered by one Grade 7 contractor, a larger emerging contractor project performance
the highest number of CE and GB-related data sample would allow for a more reliable in public sector projects and their qualifi-
projects (19 projects) has the lowest DoD conclusion to be drawn. cations and experience.
and has a staff complement with technical Project information was collected along
qualifications. This further demonstrates the CIDB grade and project performance with the respective contractors’ experience
importance of the combination of qualifica- An investigation was carried out to test and the qualifications of the key person-
tions and business experience. the assumption that the CIDB grading nel of the respective emerging contractor
is a reflection of company experience organisations. A comparison was then
Years in operation and and, therefore, the higher the grading made between contractors with shorter
project performance the better the project performance. The delays to those with significantly longer
The number of years that the contractor highest CIDB grading was selected for delays. It was concluded that contractors
has been in operation in the industry may contractors registered under both GB and with relevant construction experience and
also be used as a measure of experience. CE categories. The results do not reflect a technical qualifications experience shorter
The relationship between the number clear relationship. It was found, however, to no delays in projects – therefore, better
of years in the industry and project that the projects that were conducted by project performance.
performance was investigated in order to contractors who fall under CIDB Grade 1 Contractors with no experience and
determine whether or not the same results generally had high DoDs and had projects qualifications show poor performance when
that are represented in Figure 8 would be that were incomplete. For the lowest DoD, measured against the time factor. It was also
observed. However, a different result was 25% was achieved by a CIDB Grade 7 con- found that, where technical qualifications
found, as represented in Figure 9. tractor. The result may be due to the use of are lacking, higher education in any other
The results show that the number of an average and may only be reflecting the field is an advantage to project performance.
years in the industry do not necessar- fact that there are more Grade 1 contrac- The number of construction-related pro-
ily result in better project performance. tors than the higher grades. The number of jects undertaken and implemented by the
Further investigation was done to find out respondents in each grade may also be too contractor is a better measure of experience,
why this is. It was found that the lowest small to draw a conclusion on this. promotes competence and, therefore, aids
DoD was achieved by the contractor with a Figure 10 highlights the relationship better project performance.
staff complement with technical qualifica- between the CIDB grading and contractor There is a definite need to develop con-
tions. This may be the reason for the above project performance. tractor development programmes that will
results. This contractor also had the sec- focus on improving emerging contractors’
ond highest number of CE and GB-related project performance. Such programmes
projects, 15 in total. It may, therefore, be conclusion and should focus on developing the technical
said that the number of projects conducted recommendations qualifications and skills of contractors
is more important, or adds more value, and provide construction experience. It
than the number of years in operation. The Conclusion is further recommended that contractors
number of projects undertaken and imple- From the studies that have been conducted who have participated in such programmes
mented develops more competence than around the subject of emerging contractor should be given preference in the tendering
the number of years the business has been performance, it may be argued that very and procurement processes.
Lastly, the technical capacity and com- Collaboration References
petence of contractors, along with their Government organisations that par- Adnan, H, Hashim, N, Yusuwan, N M & Ahmad, N
experience, should be evaluated during the ticipate in public procurement and the 2012. Ethical issues in the construction industry:
selection process and, as such, contrac- use of emerging contractors along with Contractor’s perspective. Procedia – Social and
tors should be appointed on the basis of private sector clients and practition- Behavioral Sciences, 32: 719–727.
their technical ability to “do the job”. The ers should collaborate in developing Alzahrani, J I & Emsley, M W 2013. The impact
authors acknowledge that the research was standards and codes of practice for the of contractors’ attributes on construction
solely based on the project information construction industry. This collabora- project success: A post construction evaluation.
that was collected and, as such, conclu- tion may also require the involvement International Journal of Project Management, 31(2):
sions should not be generalised across all of statutory bodies such as the CIDB, 313–322.
contractors in South Africa without further the Engineering Council of South Africa Arditi, D, Akan, G T & Gurdamar, S 1985. Reasons for
study on a larger scale and, in addition, an (ECSA), and the South African Council delays in public projects in Turkey. Construction
investigation of other factors that deter- for Project and Construction Management Management and Economics, 3(2): 171–181.
mine project success which were outside Professions (SACPCMP). Arditi, D & Pattanakitchamroon, T 2006. Selecting
the delineation of this study. In addition, Project performance of emerging a delay analysis method in resolving construction
due to the sample size and the fact that the contractors should be recorded in order claims. International Journal of Project
data was collected from a single govern- to evaluate compliance with best practice Management, 24(2): 145–55.
ment department, the data should also not guidelines and standards. This information Argenti, J 1976. Corporate planning and corporate
be considered as a general representation of should then be shared across the board to collapse. Long Range Planning, 9(6): 12–17.
contractors in the Mpumalanga Province. ensure that all parties have an idea of the Assaf, S A, Al-Khalil, M & Al-Hazmi, M 1995. Causes
potential project performance of prospec- of delay in large building construction projects.
Recommendations tive contractors prior to appointment. This Journal of Management in Engineering, 11(2): 45–50.
can also be used as part of the selection Benson, R W 2006. Key issues for the construction
Tendering and procurement process process during procurement in the public industry and the economy. MidMarket Advantage,
This research found that all 30 projects sector. A platform such as the CIDB may 16–18.
under investigation had been delayed, be used for such information. It should also Cheung, S O, Suen, H C H & Cheung, K K W
showing poor project time performance. be noted that the CIDB has recommended 2004. PPMS: A Web-based construction project
This may be evidence that the time that is a similar approach in its 2012/2013 Annual performance monitoring system. Automation in
specified for completion of the work by the Report (CIDB 2013). Construction, 13(3): 361–376.
client’s engineers or consultants during the Clarke, L & Wall, C 1996. Skills in the Construction
design and tendering stage is insufficient. It Contracting companies Process: A Comparative Study of Vocational
is, therefore, recommended that government Emerging contractor companies that are Training and Quality in Social House building.
organisations revise the specified construc- already established should take advantage Bristol, UK: Policy Press, University of Bristol.
tion period during the tendering stage or of the interventions of the CIDB and other CIDB (Construction Industry Development Board)
revise the criteria used to set the period. training institutes in order to develop their 2011a. Guidelines for Contractor Registration.
technical knowledge. Contractors who Available at: http://www.cidb.org.za/documents/
Policy wish to enter into the industry should seek kc/cidb_publications/brochures/brochure_
Procurement policy should be amended to acquire higher education or tertiary contractor_registration_guidelines.pdf [accessed
such that the appointment process favours qualifications, ideally in technical fields. on 15 October 2014].
contractors with technical qualifica- CIDB (Construction Industry Development
tions and experience in the construction Recommendations for further study Board) 2011b. Framework: National Contractor
industry. It may be beneficial that highly ■■ Further study in this area is required on Development Programme. Pretoria: CIDB.
technical projects be reserved for expe- a bigger sample and on a broader scale CIDB (Construction Industry Development Board)
rienced contractors. Contractors with no (focusing on the whole of South Africa) 2013. Annual Report 2012/2013. Pretoria: CIDB.
experience should also be encouraged to in order for the results to be more Croswell, J & McCutcheon, R 2001. Small contractor
enter into joint-ventures with experienced generalisable. development and employment – A brief survey
contractors for highly technical projects in ■■ A training programme needs to be of sub-Saharan experience in relation to civil
order to develop their skills. However, this developed for emerging contractors construction. Urban Forum, 12(3): 365–379.
is no substitute for the need for contractors in the public sector. The programme DTI (Department of Trade and Industry) 2004.
to develop their technical skills. needs to deal with the issues addressed Annual Review of Small Business in South Africa
The qualifications that need to be in this paper and develop competencies 2003. Pretoria: Department of Trade and Industry,
focused on are those of the core staff of that will ensure the success of emerging Enterprise Development Unit.
the business. Experience may be that of contractors. Eyiah, A K 2001. An integrated approach to financing
the staff in the business or the overall ■■ The field of construction development small contractors in developing countries: A
business experience. This also means that in South Africa needs to be investigated conceptual model. Construction Management and
the number of projects undertaken and in order to formulate “new” contractor Economics, 19(5): 511–518.
implemented by the company should weigh development programmes aimed at Faridi, A & El-Sayegh, S 2006. Significant factors
heavier when appointing contractors than improving the competencies, and there- causing delay in the UAE construction industry.
the number of years their business has fore, the performance, of contractors on Construction Management and Economics, 24(11):
been in the construction industry. a more technical level. 1167–1176.
Fredland, E J & Morris, C E 1976. A cross-section companies in the Free State Province. PhD Thesis, cost and quality management. Construction
analysis of small business failure. American Journal Johannesburg: University of Johannesburg. Management and Economics, 13(3): 235–41.
of Small Business, 1(1): 7–18. Mohlala, F T 2015. The relationship between project Sambasivan, M & Yau, W S 2007. Causes and effects
Gharakhani, D, Sinaki, M T, Dobakhshari, M A & performance of emerging contractors in government of delays in Malaysian construction industry.
Rahmati, H 2013. The relationship of customer infrastructure projects and their experience and International Journal of Project Management, 25(5):
orientation, customer satisfaction, customer loyalty technical qualifications: An analysis of 30 projects 517–526.
and innovation in small and medium enterprises. conducted in the Mpumalanga Province over Shepherd, D A, Wiklund, J & Haynie, J M 2009.
Life Science Journal, 10(6s): 684–689. the 2011–2013 period. MSc Research Report, Moving forward: Balancing the financial and
Guillaume, R, Houe, R & Grabot, B 2014. Robust Johannesburg: University of the Witwatersrand. emotional costs of business failure. Journal of
competence assessment for job assignment. Myers, M D & Avison, D 1997. Qualitative research Business Venturing, 24(2): 134–148.
European Journal of Operational Research, 238(2): in information systems. Management Information Storey, D J 1994. Understanding the Small Business
630–644. Systems Quarterly, 21: 241–242. Sector. London: Routledge,
Hanson, D, Mbachu, J & Nkando, R 2003. Causes of Neely, A 1998. Measuring Business Performance. Sweis, R J, Bisharat, S M, Bisharat, L & Sweis, G 2014.
client dissatisfaction in the South African Building London: Economist Books. Factors affecting contractor performance on public
industry and ways of improvement: The contractors’ Nemaenzhe, P P 2010. Retrospective analysis of failure construction projects. Life Science Journal, 11(4s):
perspectives. Pretoria: CIDB. causes in South African small businesses. PhD 28–39.
Harrison, F L & Lock, D 2004. Advanced project Thesis, Pretoria: University of Pretoria. Temtime, Z T & Pansiri, J 2004. Small business critical
management: A structured approach. Aldershot, Okpara, J O & Kabongo, J D 2009. An empirical success/failure factors in developing economies:
UK: Gower Publishing. evaluation of barriers hindering the growth of SME evidence from Botswana. American Journal of
Jaafar, M & Abdul-Aziz, A R 2005. Resource-based small and medium sized enterprises (SMEs) in a Applied Sciences, 1(1): 18–25.
view and critical success factors: A study on small developing economy. African Journal of Business Thwala, W D & Mvubu, M 2005. Current challenges
and medium sized contracting enterprises (SMCEs) and Economic Research, 4(1): 7–21. and problems facing small and medium size
in Malaysia. International Journal of Construction Okpara, J O & Wynn, P 2007. Determinants of small contractors in Swaziland. African Journal of
Management, 5(2): 61–77. business growth constraints in a sub-Saharan Business Management, 2(5): 93–98.
Jo, H & Lee, J 1996. The relationship between an African economy. S.A.M. Advanced Management Turner, J R 1999. Handbook of Project-Based
entrepreneur’s background and performance in a Journal, 72(2): 24–35. Management. London: McGraw Hill.
new venture. Technovation, 16(4): 161–211. Peacock, R 2000. Failure and assistance of small firms. Tushabomwe-Kazooba, C 2006. Causes of small
Lin, C Y Y 1998. Success factors of small-and medium- Small Enterprise Series, 1: 1–56. business failure in Uganda: A case study from
sized enterprises in Taiwan: An analysis of cases. Peters, L M & Zelewski, S 2007. Assignment of Bushenyi and Mbarara towns. African Studies
Journal of Small Business Management, 36(4): 43–56. employees to workplaces under consideration Quarterly, 8(4): 1–13.
Lo, T Y, Fung, I W H & Tung, K C F 2006. of employee competences and preferences. Van Eeden, S, Viviers, S & Venter, D 2003.
Construction delays in Hong Kong civil engineering Management Research News, 30(2): 84–99. A comparative study of selected problems
projects. Journal of Construction Engineering and Prais, S J & Steedman, H 1986. Vocational training in encountered by small businesses in the Nelson
Management, 132(6): 636–49. France and Britain: The building trades. National Mandela, Cape Town and Egoli metropoles.
Mbande, C 2010. Overcoming construction constraints Institute Economic Review, 116(1): 45–55. Management Dynamics, 12(3): 13–23.
through infrastructure delivery. Proceedings, 5th Pretorius, M 2009. Defining business decline, failure Van Witteloostujin, A 1998. Bridging behavioural
Built Environment Conference, Association of and turnaround: A content analysis. South African and economic theories of decline: Organizational
Schools of Construction of Southern Africa, 18–20 Journal of Entrepreneurship and Small Business inertia, strategic competition, and chronic failure.
July, Durban, Vol 18, p 20. Management, 2(1): 1–16. Management Science, 44(4): 501–519.
Mezher, T M & Tawil, W 1998. Causes of delays in Richter, A 1998. Qualifications in the German Van Wyk, L 2003. A review of the South African
the construction industry in Lebanon. Engineering, construction industry: Stocks, flows and construction industry. Part 1: Economic, regulatory
Construction and Architectural Management, 5(3): comparisons with the British construction sector. and public sector capacity influences on the
252–260. Construction Management and Economics, 16(5): construction industry. Pretoria: CSIR, Boutek.
Milford, R 2010. Public capacity payment and 581–592. Yusuf, A 1995. Critical success factors for
procurement issues should be a challenge to Rozewski, P & Malachowski, B 2009. Competence small business: Perceptions of South Pacific
the operations of contractors in South Africa. management in knowledge-based organisation: entrepreneurs. Journal of Small Business
Proceedings, 5th Built Environment Conference, Case study based on higher education organisation. Management, 33(2): 68–73.
Association of Schools of Construction of Southern Knowledge Science Engineering and Management, Zulu, S & Chileshe, N 2010. Service quality of building
Africa, 18–20 July, Durban. 5914: 358–369. maintenance contractors in Zambia: A pilot study.
Mofokeng, T G 2012. Assessment of the causes of Rwelamila, P D & Hall, K A 1995. Total systems International Journal of Construction Management,
failure among small and medium sized construction intervention: An integrated approach to time, 10(3): 63–81.
TECHNICAL PAPER
Journal of the South African The influence of concrete
ISSN 1021-2019 compressive strength
and specimen size on
the compression stress

FRANCOIS VAN SCHALKWYK received his
BEng in Civil Engineering cum laude in 2014
and BEng Hons (Structural Engineering)
block parameters of
cum laude in 2015 from the University of
Pretoria. This paper is based on research
conducted for his Master’s degree in
reinforced concrete
Structural Engineering.
Contact details:
PO Box 40772
Moreletapark
Pretoria 0044 F van Schalkwyk, E P Kearsley
Gauteng
South Africa
T: +27 82 574 8165
E: frannavschalkwyk@gmail.com This paper describes the influence of concrete compressive strength and specimen size
on the fundamental characteristics of the flexural compressive stress-strain distribution.
PROF ELSABÉ KEARSLEY is a Fellow of the The main variables were specimen size and cylinder compressive strength. A total of 27
South African Institution of Civil Engineering concrete specimens were subjected to flexural tests, with specific reference to analysis of the
(SAICE), of which she was also the president
compression stresses, produced by varying two independent loads in a configuration aimed at
in 2009. She holds a PhD from the University
of Leeds and she is currently a professor in controlling the strain distribution. These loads generated a condition of zero strain on the one
Civil Engineering at the University of face of the specimen, and a condition of maximum flexural compression strain on the opposite
Pretoria. She worked as a structural design face. From the strain distribution, the stress-strain curves and stress block parameters were
engineer in both South Africa and the
determined, and the influence of specimen size on the stress block parameters described using
United Kingdom before becoming a staff member at the University of
Pretoria. For the last 23 years she has been involved with cement and the Modified Size Effect Law (MSLE). Using a modified form of the moment-axial force (M-N)
concrete materials research. interaction diagram the BS 8110-1 (1997), SANS 0100-1 (2000), ACI-318 (2014) and EN 1992-1-1
Contact details: (2004) codes of practice were compared for the design of reinforced concrete beams containing
Department of Civil Engineering South African materials.
University of Pretoria
Pretoria
0002
South Africa INTRODUCTION In this study, the influence of concrete
T: +27 12 420 2176 Even though high-strength reinforced compressive strength and specimen size
E: elsabe.kearsley@up.ac.za
concrete may be the most cost-effective on the flexural compression capacity of
solution for many structures, most design concrete was assessed. In total, 27 concrete
codes historically did not cover the design specimens were tested in flexure, with
of concrete with strengths in excess of specific reference to analysis of the com-
about 60 MPa. The South African code of pression stresses (produced by applying
practice, SANS 0100-1 (2000), limits the two sets of loads in a configuration aimed
characteristic cube compressive strength at controlling the strain distribution), the
to 60 MPa, while Eurocode 2 (EN 1992- stress block parameters derived, and the
1-1 (2004)) can be used for the design of influence of specimen size on the stress
reinforced concrete elements with charac- block parameters evaluated. Along with the
teristic cube compressive strengths up to size effect in flexural tests, with specific
105 MPa. Using concrete with strengths in reference to an analysis of the compressive
excess of 60 MPa in flexural elements will stresses, the size effect for cubes and cylin-
require less concrete in the compression ders was also evaluated, and the size effect
zone of beams, which could lead to a sig- for cylinders used to eliminate the size
nificant reduction in own weight of struc- effect of the stress block parameters.
tures as a result of minimised structural Finally, the difference between the
element sizes. However, before adopting the measured moment-axial force (M-N)
European code of practice for the design of interaction diagram and the M-N interac-
high-strength concrete elements, it is neces- tion diagrams obtained using the BS 8110-1
sary to ensure that the code design assump- (1997), SANS 0100-1 (2000), ACI-318 (2014)
Keywords: stress block parameters, rectangular stress block, tions for the stress block parameters are and EN 1992-1-1 (2004) code recommenda-
size effect, SANS 0100-1, EN 1992-1-1 applicable to local conditions and materials. tions was compared.
Van Schalkwyk F, Kearsley EP. The influence of concrete compressive strength and specimen size on the compression stress block parameters
34 of reinforced concrete. J. S. Afr. Inst. Civ. Eng. 2018:60(4), Art. #0013, 11 pages. http://dx.doi.org/10.17159/2309-8775/2018/v60n4a4
Background
A beam or a pure flexural member may be b
defined as an element that transfers loads, εcu k 3 fc΄ α1 fc΄
applied normally to its longitudinal axis,
k 2c β1c/2
to support points (Mahachi 2013). When
β 1c
a reinforced concrete beam is loaded to c C = k 1 k 3 fc΄bc C = α1 β1 fc΄bc
failure, the ultimate limit state (ULS) is
initiated when the strain at the extreme
compression fibre reaches a maximum. d
The analysis and design of reinforced con-
crete beams are normally performed at the
ULS, and require four basic assumptions
(Hognestad et al 1955):
■■ Plane sections before bending remain As
plane after bending. This assumption
implies that the longitudinal strain (at
various points in the cross section) is
proportional to the distance from the Figure 1 S train, three-parameter and two-parameter equivalent rectangular stress block
neutral axis. (Hognestad et al 1955)
■■ The stress-strain curve for the steel is
known, and the strain in the steel and compression member. These strains, when for an over-reinforced beam the size effect
concrete at the same depth of the cross- combined with the loads and numeri- for compressive strength is directly related
section is assumed to be equal. cal differentiation, make it possible to to the flexural compressive strength in the
■■ The tensile strength of the concrete determine the stress-strain behaviour of compressive region of the beam. For an
below the neutral axis may be neglected, concrete in flexure, with specific reference under-reinforced member, the increased
as both the force and the lever arm are to an analysis of the compressive stresses. flexural compressive strength results in
small. Numerical integration of this stress-strain a longer lever arm, and thus an increased
■■ The stress-strain curve of concrete behaviour, combined with the uniaxial moment capacity.
(defining the magnitude and distribu- compressive strength of a 152 × 305 mm Concrete is a quasi-brittle material that
tion of the compressive stress) is known, cylinder, enabled the stress block fails through the formation and propaga-
and is a function of the strain only. parameters to be calculated. tion of cracks induced by external loads.
At ULS the nonlinear stress-strain distri- Subsequent investigations done by The formation and propagation of cracks
bution in the concrete compression zone Nedderman (1973), Kaar et al (1978a), Kaar release strain energy, and thus the size of
may mathematically be defined with three et al (1978b), Swartz et al (1985), Pastor the specimen (D) will have an influence on
parameters k 1, k 2 and k 3, and replaced with (1986) and Ibrahim (1994) all utilised the the nominal strength (σ n). Bazant (1984)
the equivalent rectangular stress block “C-shaped” specimen. Researchers such as derived the Size Effect Law (SEL) for
defined by two parameters α1 (k 1k 3/2k 2) Soliman and Yu (1967), Sargin et al (1971), concrete members subjected to uniaxial
and β1 (2k 2), as shown in Figure 1 Schade (1992), Tan and Nguyen (2005), tensile stress ( f t΄) by considering the energy
(Hognestad et al 1955). Mertol et al (2008), and Khadiranaikar balance at the instant of crack propagation.
The ratio between the average and and Awati (2012) used detachable steel The SEL, indicated in Equation 1, assumes
maximum flexural compressive strength lever arms to apply the secondary load to that the energy released is proportional
is defined as k 1, the ratio between the concrete column sections. This type of to ndaa, with n a constant, a the crack
depth of the resultant compressive force experimental procedure is more practical length and da the maximum aggregate
and the depth of the neutral axis is and efficient, reducing reinforcing and size. λ0 = na = 2 is normally assumed,
defined as k 2 , and k 3 is defined as the concrete costs. and the parameter β is determined from
ratio between the maximum stress in Utilising their own and previous test regression analysis.
flexural compression and the uniaxial results, researchers such as Mattock
cylinder compressive strength. The height et al (1961) expressed the stress block βf t΄
σn = (1)
of the rectangular stress block is α1, while parameters as linear step functions depend- D
1+
β1 is the base width of the rectangular ent on the cylinder compressive strength. λ0 d a
stress block. However, the influence of specimen size
It is difficult to measure the stress was not considered in their formulation. Kim and Eo (1990) proposed the Modified
in the compression zone of a reinforced The size effect in flexure, with specific Size Effect Law (MSEL) shown in
concrete flexural member through direct reference to an analysis of the compressive Equation 2, where it is assumed that a
experimentation. The ground-breaking stresses, is important, as an over-reinforced certain component of strength (a f t΄) is
work published by Hognestad et al (1955) concrete beam will have its flexural independent of specimen size.
proposed the use of a 203 × 152 mm strength directly related to the concrete
“C-shaped” specimen to simulate the strain strength, while the flexural strength of an βf t΄
σn = + α f t΄(2)
distribution in the compression zone of under-reinforced beam will be dictated by D
1+
a reinforced concrete flexural or flexural the amount of tensile reinforcing. Thus, λ0 d a
Even though Equations 1 and 2 were Table 1 Mix proportions
derived for cracks opening parallel to the Mix 1 Mix 2 Mix 3
applied load, the compression failure of (40 MPa) (65 MPa) (80 MPa)
members is also related to the formation
PPC CEM 52.5 N (kg/m3) 269 400 536
and propagation of cracks, but perpendicu-
lar to the applied load. The SEL and MSEL Total water (kg/m3) 222 227 231
can thus be used to describe the specimen
Dolomite sand (kg/m3) 1 980 1 848 1 711
size effect in compression (Bazant 1999).
Kim et al (2000) attempted to quan- Total mix weight (kg/m3) 2 471 2 475 2 478
tify the influence of specimen size on
Water/cement 0.8 0.57 0.43
flexural compressive behaviour and the
ultimate compressive strain utilising the
same “C-shaped” specimen proposed by Table 2 Mix quantities
Hognestad et al (1955). The main test Designation Dimensions Quantity
variable was the specimen size ratio (1:2:4),
with the length and depth of the specimens 50 × 50 × 50 mm 12
proportionally changed, but the width of 100 × 100 × 100 mm 12

the specimen kept constant. A total of Cubes
150 × 150 × 150 mm 12
twenty-one 52.0 MPa “C-shaped” speci-
mens were tested, and it was concluded 200 × 200 × 200 mm 5
that the size effect is more pronounced 50 × 100 mm 12
in flexure, with specific reference to an
100 × 200 mm 12
analysis of the compressive stresses, than Cylinders
in uniaxial compression. 150 × 300 mm 12
Although Kim et al (2000) investigated
200 × 400 mm 5
the specimen size effect, neither the inter-
action between the specimen size effect 50 × 50 × 300 mm 3
and compressive strength, nor the effect Flexural compressive members 100 × 100 × 500 mm 3
on the individual stress block parameters
200 × 200 × 1 000 mm 3
was considered, providing the basis for this
investigation.
200 mm 150 mm
EXPERIMENTAL PROGRAMME
This experimental programme aimed to
280 mm
quantify the effect of compressive strength

100 mm 70 mm
and specimen size on the stress block
parameters of concrete containing South 30 mm 50 mm
150 mm
African materials. The mix designs and
100 mm
elements cast for each mix are presented
400 mm
in Tables 1 and 2 respectively. As high- 200 mm 100 mm
strength concrete elements normally have 100 mm

reduced cross-sectional areas, the effect of 150 mm Front view Side view
specimen size was taken into account by
300 mm
Front view Side view Scale 1:4

casting small elements with dimensions
Scale 1:2
comparable to the compression block depth
of high-strength concrete beams. Crushed
dolomitic aggregate with a maximum size Front view Side view
of 4.75 mm, and CEM I 52.5 N cement Scale 1:1
were used as the sole aggregate and binder.
After casting and demoulding, the Figure 2 Configuration of longitudinal and shear reinforcing in flexural compression members
cubes, cylinders and flexural compression
members were water-cured, during which The cubes were crushed in uniaxial To prevent premature localised failure,
time surface-grinding of the cylinders compression, and the uniaxial stress-strain both ends of the flexural compression
and the compression face of the flexural relationship of the cylinders was monitored members were heavily reinforced. The rein-
compression members was carried out. using suitably sized compressometers. forcing configuration in the end regions of
After 28 days, the specimens were placed The 150 mm cylinder compressometer each of the concrete sections is illustrated
in an oven at 50°C for two days (to remove measured both longitudinal and transverse in Figure 2. The longitudinal reinforcing
any moisture), and subsequently stored in a deformation, making it possible to compute for the 1:1, 1:2 and 1:4 scale specimens
room with a constant temperature. Poisson’s ratio. consisted of three Y12, two Y8 and two Y5
(a) Scale 1:1 (b) Scale 1:2 (c) Scale 1:4
Figure 3 Test setup for (a) 1:1, (b) 1:2 and (c) 1:4 scale specimens
bars respectively. The transverse reinforc- 150 × 85 × 16 mm end plate, and connected of the specimen to monitor the deflected
ing, used in the 1:1 and 1:2 scale specimens via four Gr 8.8 M16 threaded rods to a shape, thus enabling the incorporation of
consisted of three Y10 and three Y6 bars, 150 mm × 85 mm × 16 mm back plate. second-order moments.
with 2 mm steel wire used to enclosed the The primary axial load (P1) was applied
1:4 scale specimen cages. to the 1:1 and 1:2 scale specimens with a
After being removed from the oven, the 3 000 kN Amsler compression machine, Discussion and
central unreinforced region of the flexural while a 250 kN MTS 810 compression Analysis of Results
compression members was instrumented machine was used for the 1:4 scale
with 30 mm long BK 2/120 strain gauges. specimens. For the 1:1 scale specimens, a Cubes and cylinders
Six strain gauges were attached to the 1:2 100 ton Simplex hydraulic jack was used to A summary of the cube and cylinder
and 1:4 scale flexural compressive speci- apply the secondary load (P 2), and a 10 ton compressive data can be seen in Table 3.
mens, and nine to the 1:1 scale specimens. Euro press pack hydraulic jack was used By assuming λ0 = 2, da = 0.475 cm,
For all specimen scales, three of these for the 1:2 and 1:4 scale specimens. For all λ0 da = 0.95 cm and standardising the
strain gauges were located on the flexural specimens, P 1 was applied at a constant compressive strength for each specimen
compressive surface, two monitoring the rate of 2 MPa/min, and P 2 was constantly size with respect to the average 150 mm
longitudinal compressive strain and the adjusted to maintain a condition of zero cylinder strength ( fc’) of that particular
remaining, the transverse tensile strain. strain on the back exterior face of the mix, the remaining MSEL regression coef-
Figure 3 shows the experimental specimen, while the opposing face was sub- ficients could be calculated. Table 4 shows
setup for the 1:1, 1:2 and 1:4 scale jected to a monolithically increasing com- a numerical presentation of the MSEL
specimens respectively. The 1:1 scale pressive stain. Three 50 mm linear variable coefficients with a visual presentation of
lever arms consisted of two 600 mm long displacement transducers (LVDTs) were the cube and cylinder size effect provided
PC 200 × 75 channel sections, connected placed at the top, bottom and mid-section in Figures 4 and 5 respectively.
via their top and bottom flanges with two
180 × 10 × 600 mm steel plates. The assem- Table 3 Cube and cylinder (h/d = 2) compressive strength summary
bly was welded to a 400 × 260 × 25 mm Cube (MPa) Cylinder (MPa)
Dimension (mm)
end plate, strengthened with four 10 mm Parameter
cube and cylinder Mix 1 Mix 2 Mix 3 Mix 1 Mix 2 Mix 3
stiffeners, and connected to a 25 mm back
plate (reinforced with two PC 200 × 75 Min 43.3 75.3 86.8 31.4 48.2 65.8
channel sections) through six Gr 8.8 M33 50 Mean 47.3 79.2 96.7 32.6 54.6 68.7
threaded rods.
Max 49.2 89.3 101.3 34.4 58.8 71.7
The 1:2 scale lever arms consisted of
two 400 mm long C 100 × 50 channel Min 41.3 69.9 81.5 38.1 64 76.3
sections connected via their top and bot- 100 Mean 44.8 74.6 86.8 41.9 68.1 80.2
tom flanges with two 90 × 5 × 400 mm
Max 49.5 79.9 92.4 43.3 71.7 85.7
steel plates. The assembly was welded to a
250 × 130 × 20 mm end plate, strengthened Min 37.6 65 76.7 38 60.1 70.8
with two 10 mm stiffeners and connected 150 Mean 40.8 69.7 80.6 40.1 64.1 76.4
to a 20 mm end plate (reinforced with two
Max 43.6 73.3 89.9 42.9 67.4 83.7
C 100 × 50 channel sections) with four
Gr 8.8 M20 threaded rods. Min 36.1 59.4 72.5 32.1 56 63.1
The 1:4 scale lever arms consisted 200 Mean 37.4 61.2 72.3 36.9 60.2 67.3
of a 255 mm long, 50 × 50 × 4.5 mm
Max 41 64.5 78.5 40.6 67.2 71.6
square tube section, welded to a
The values presented in Table 4 indicate Table 4 MSEL coefficients for cubes and cylinders
that the size-independent-component- Mix 1 Mix 2 Mix 3
coefficient (α) for cubes appears to be Average
(40 MPa) (65 MPa) (80 MPa)
independent of compressive strength, but
β 1.17 1.41 1.41 1.33
for cylinders reduces with an increase in Cube
compressive strength. However, β becomes α 0.72 0.71 0.71 0.72
more prominent with an increase in β 0.81 1.21 1.42 1.15
compressive strength for both cubes and Cylinder
α 0.78 0.7 0.62 0.7
cylinders.
Flexural compression specimens

The typical longitudinal strain gauge 1.6
measurements, shown in Figure 6 indicate
that there is a slight deviation in the strain 1.2
gauge readings. However, these deviations
are not significant. Thus, the strain gauge fcube /fc΄
measurements provide confirmation of the 0.8
assumption that planar sections remain

plane during bending. It can therefore be 0.4
assumed that all fibres in the cross-section
follow the same stress-strain curve, and
0
the strain at the extreme compression 0 2.5 5.0 7.5 10.0 12.5 15.0 17.5 20.0 22.5 25.0
fibre may be used to describe the complete Cube size (cm)
stress-strain curve. Mix 1 Mix 2 Mix 3 Average
By combining the longitudinal and
transverse extreme fibre compressive Figure 4 Size effect for cubes
strains, estimates of Poisson’s ratio in flex-
ure, with specific reference to an analysis
of the compressive stresses, was made. 1.6
Published results for Poisson’s ratio, as well
as the Poisson ratios obtained in this study, 1.2
can be seen in Figure 7. Figure 7 indicates
fcylinder/fc΄
that the average Poisson’s ratio is 0.2, and

that the 150 mm cylinder Poisson’s ratio 0.8
was independent of concrete strength,

with an average value of 0.23. The flexural 0.4
compression members, on the other hand,
showed an increase in Poisson’s ratio,
0
with an increase in compressive strength. 0 2.5 5.0 7.5 10.0 12.5 15.0 17.5 20.0 22.5 25.0
However, Figure 7 also illustrates the large Cylinder diameter (cm)
range of Poisson’s ratio values for different Mix 1 Mix 2 Mix 3 Average
concrete strengths. Although most design-
ers use a value ranging from 0.15 to 0.2 (0.2 Figure 5 Size effect for cylinders
being the most common), it can be as high
as 0.28. Since a high Poisson’s ratio can have
a profound effect on the stress distribution 5 000
in structures and structural elements, the
possible consequence of a higher Poisson’s 4 000
Compressive strain
ratio value should be taken into account.

Using the free-body diagram indicated 3 000
in Figure 8, the procedure proposed by
Hognestad et al (1955) was used to calcu- 2 000
late the concrete stress (σc) as a function of

1 000
the extreme fibre longitudinal compressive
strain (εc), and the applied stresses ( fo
0
and mo ) as indicated in Equations 3 and 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
4. Stresses fo and mo are determined from Strain gauge location (fraction of depth)
the applied external loads (P1 and P 2), the 50 mm 100 mm 200 mm
eccentricity of these loads (a1 and a2), the
section width (b), and section depth (c). Figure 6 Strain distribution through depth of flexural compression specimens
bc εc
0.30
C = P 1 + P 2 = fobc =
εc ∫
0
σc(ε x)dε x(3)
0.25
M = P 1a1 + P 2 a2 = mobc 2
Poisson's ratio (υ)
0.20
bc 2 εc
0.15
=
εc2 ∫0
σc(ε x)ε xdε x(4)
0.10 With:
0.05 P1 + P 2
fo = (5)
0 bc
0 10 20 30 40 50 60 70 80 90 100 110 120 130 140
Compressive strength (MPa) P 1 a1 + P 2 a 2
M1/200 mm M2/200 mm mo = (6)
bc 2
M3/200 mm M1/100 mm
M2/100 mm M3/100 mm
M1/50 mm M2/50 mm
Equations 7 and 8 are obtained by dif-
M3/50 mm Other researchers ferentiating and rearranging the last terms
Cylinder results LSM Global of Equations 3 and 4 with respect to εc.
Excentric specimens 150 × 300 mm Using small enough increments, the dif-
ferentials of dfo /dεc and dmo /dεc may be
Figure 7 Poisson’s ratio approximated by ∆fo /∆εc and ∆mo /∆εc , and
Equations 7 and 8 can be used to generate
two similar stress-strain curves. The aver-
age of the two relationships was used to
P1 represent the flexural stress-strain relation-
ship, with specific reference to the analysis
of compression stresses (Hognestad
et al 1955). See Figure 9.
dfo
σc1 = εc + fo(7)
dεc
dmo
a1 σc2 = εc + 2mo(8)
dεc
a2
c
b With the stress-strain relationships in flex-
C P2
M
ure, with specific reference to an analysis
Side view Front view of the compressive stresses defined, the
generalised stress block parameters and
Figure 8 Free-body diagram of scale 1:1 eccentric bracket specimen ultimate strains were calculated. For each
specimen size, the stress block p arameters
were calculated using the 150 mm
100 c ylinder strength, as well as the cylinder
Mix 3 strength of samples with cross-section
80 dimensions similar to that of the scaled
flexural samples.
Mix 2
From Figures 10 and 11 it can be seen
Stress (MPa)
60
that the stress block parameters k 1 and k 2
(and thus β1) are independent of specimen
40 Mix 1 size. As in the case of the cubes and cylin-
ders, the influence of size on the remaining
20 stress block parameters was established by
assuming λ0 da = 0.95. Figure 12 shows the
size effect for k 1k 3 if the 150 mm cylinder
0
0 500 1 000 1 500 2 000 2 500 3 000 3 500 4 000 4 500 5 000 strength is used for calculation.
Strain (με) As the 200 mm (20 cm) specimen is
50 mm 100 mm 200 mm the standard specimen size used to obtain
the stress block parameters, the stress
Figure 9 Typical flexural compression stress strain relationships block parameters for other specimen
Table 5 M
SEL coefficients for relevant stress
1.2
block parameters
1.0
Mix 1 Mix 2 Mix 3
0.8
β 1.11 0.49 0.65
k3 0.6
k1
α 0.79 0.88 0.88
0.4
β 0.94 0.81 0.66
k 1k 3 0.2
α 0.82 0.83 0.87
0
β 1.07 0.63 0.61 0 20 40 60 80 100 120 140
α1
Compressive strength (MPa)
α 0.79 0.86 0.88
Other researchers 50 mm 100 mm 200 mm
β 0.30 0.31 0.41
εcu
α 0.94 0.94 0.89 Figure 10 Effect of 150 mm cylinder compressive strength on k1
sizes are standardised with respect to 0.6

the average 20 cm specimen stress block 0.5
parameter. The direct influence of the
0.4
cylinder strength on the calculation of
k 3, k 1k 3 and α1 (k 1k 3/2k 2), as well as the 0.3
k2
known size effect associated with the 0.2

ultimate compressive strain (ε cu), produces
the size effect coefficients illustrated in 0.1
Table 5. The average of these relationships 0

was used to describe the size effect for 0 20 40 60 80 100 120 140
the stress block parameters, as shown in
Other researchers 50 mm 100 mm 200 mm
Equations 9 to 12.
0.75k 3(20cm) Figure 11 Effect of 150 mm cylinder compressive strength on k2

k3 = + 0.85k 3(20cm)(9)
c
1+
0.95 1.2
0.80k 1k 3(20cm) 1.0

k 1k 3 = + 0.84k 1k 3(20cm) (10)
c 0.8
1+
0.95
k1k 3
0.6
0.77α1(20cm) 0.4
α1 = + 0.84α1(20cm)(11)
c 0.2
1+
0.95 0
0 20 40 60 80 100 120 140
0.34εcu(20cm) Compressive strength (MPa)
εcu = + 0.92εcu(20cm) (12)
c Other researchers 50 mm 100 mm 200 mm
1+
0.95
Figure 12 Effect of 150 mm cylinder compressive strength on k1k3
Graphs for k 1k 3(15cm)/k 1k 3ave(20cm) calcu-
lated using the 150 mm cylinder strength,
1.4
and k 1k 3(ass.)/k 1k 3ave(20cm) calculated using
k1k 3(15cm)/k1k 3ave(20 cm)
the associated cylinder strength (for the 1.2

50 mm and 100 mm specimens) are shown
1.0
in Figures 13 and 14 respectively. Figure 13
illustrates the specimen size effect associ- 0.8
ated with k 1k 3, while Figure 14 illustrates
that the size effect for k 1k 3 (and thus k 3 0.6
and α1) may be eliminated if the associated
0.4
cylinder strength is used in the calcula- 0 2.5 5.0 7.5 10.0 12.5 15.0 17.5 20.0 22.5 25.0
tion of the 50 mm and 100 mm stress Specimen size (cm)
block parameters. Mix 1 Mix 2 Mix 3 Average
As the size effect in the 50 mm and
100 mm flexural compression specimens Figure 13 Effect of specimen size on k1k3(15cm)/k1k3ave(20cm)
1
1.2 tan θ = (16)
0.5 – ζk 2
1.0
To generate the normalised M-N interaction
k1k 3(ass.)/k1k 3ave(20 cm)
0.8
diagram using the code recommendations,
the k1k 3 (α1 × β1) and k 2 (0.5 × β1) value
0.6 obtained from the relevant recommenda-
tion is used in conjunction with a value
0.4 of ζ = ζd = k 2t/k 2d in Equations 13 and 14.
Note that ζ = ζd = k 2t/k 2d implies that the
0.2 angle between the normalised moment axis
and the line connecting the origin to the
0
0 2.5 5.0 7.5 10.0 12.5 15.0 17.5 20.0 22.5 25.0
test point needs to be the same as the angle
Specimen size (cm) between the normalised moment axis and
Mix 1 Mix 2 Mix 3 the line connecting the origin to the point
Minimum Maximum Average generated using the design model. In a simi-
lar fashion, the point obtained by using the
Figure 14 Effect of associated cylinder strength on k1k3(ass.)/k1k3ave(20cm) design recommendation can be presented by
a radial distance (Rd) and angle (θd).
can be eliminated by using the associated on the normalised M-N interaction dia- The experimental points were com-
cylinder strength, the same stress block gram. These coordinates are obtained by pared to the points obtained by using the
parameter models proposed for the con- substituting the k 1k 3 and k 2 values of the equations published in the design codes of
ventional sized specimens can be used for experimental points into Equations 13 and practice, and the difference between the
the design of the flexural members with 14 (derived from the equilibrium of forces measured and predicted values used to
reduced cross-sectional areas. If this is the and moments). For the experimental values determine δ (Equation 17) as a fraction of
case, the increase in compressive strength ζ = ζt = 1, as the depth of the neutral axis the measured values. If δ is larger than one,
(owing to a reduction in specimen size) corresponds to the depth of the section. the design recommendations under-predict
may be taken into consideration using the Each of these points can also be defined the strength. The δ values for the BS 8110-1
size effect for cylinders in uniaxial com- in a radial coordinate system, consisting (1997), SANS 0100-1 (2000), ACI-318 (2014)
pression. Furthermore, if the 50 mm and of radial distance (Rt ) and an angle (θt ), and EN 1992-1-1 (2004) design codes are
100 mm flexural compression specimens calculated by using Equations 15 and 16 shown in Figures 15 to 17. The average of
are used in conjunction with the associ- (Ibrahim 1994) respectively. these differences was calculated in 10 MPa
ated cylinder strength, a much easier and increments, and the histogram shown in
more cost-effective investigation into M Figure 18 constructed.
m= = ζk 1k 3(0.5 – ζk 2)(13)
the conventional stress block parameters bc 2fc΄
can be made. The reduction in flexural- Rt – Rd
δ= × 100 (17)
compression-test-specimen size has the C Rt
n= = ζ × k 1k 3(14)
benefit of not only improving the handling bcfc΄
ability, but also paves a way for further Figures 15 and 18 show that the BS 8110-1
investigations into the flexural behaviour, R = √n2 + m2(15) (1997) and SANS 0100-1 (2000) codes do
with specific reference to an analysis of
the compressive stresses, of high-strength
70
concrete, a study field previously limited
due to compression testing machine 50
capacity limitations.
30
Comparison of design codes 10

δ (%)
The BS 8110-1 (1997), SANS 0100-1 (2000),

–10
ACI-318 (2014) and EN 1992-1-1 (2004)
design codes were compared both from a –30
global perspective and with respect to their
–50
ability to predict the flexural behaviour
of concrete containing South African –70
0 10 20 30 40 50 60 70 80 90 100 110 120 130 140
materials. The basis for this comparison
is a moment-axial force (M-N) interaction Hognestad et al (1955) Tan & Nguyen (2004) Pastor (1986)
diagram that has been normalised with Kaar et al (1978a) Khadiranaikar & Awati (2012) Ibrahim (1994)
respect to the section dimensions and the Swartz et al (1985) Nedderman (1973) Mertol et al (2008)
compressive strength. Schade (1992) Kaar et al (1978b) Study results
Each of the experimental points can
be presented by a coordinate pair (mt , nt ) Figure 15 Percentage difference: BS 8110-1 (1997) and SANS 0100-1 (2000) and experimental results
not provide a constant difference, and
70
the prediction becomes less conservative
50 with an increase in compressive strength.
On average the BS 8110-1 (1997) and
30
SANS 0100-1 (2000) codes underestimate
10 the strength by 0.8%. For cylinder strengths
δ (%)
in excess of 80 MPa, the BS 8110-1 (1997)

–10
and SANS 0100-1 (2000) codes overesti-
–30 mate by as much as 6.5%. In total 48.6% of
results were smaller than the code predic-
–50
tion. On average the BS 8110-1 (1997) and
–70 SANS 0100-1 (2000) codes underestimate
0 10 20 30 40 50 60 70 80 90 100 110 120 130 140
the study results by 5.2%, 4.0% and 9.0% for
Hognestad et al (1955) Tan & Nguyen (2004) Pastor (1986)
Mix 1, Mix 2, and Mix 3 respectively.
Kaar et al (1978a) Khadiranaikar & Awati (2012) Ibrahim (1994) Figures 16 and 18 show that the ACI‑318
Swartz et al (1985) Nedderman (1973) Mertol et al (2008) (2014) code does not provide a constant
Schade (1992) Kaar et al (1978b) Study results difference, becoming less conservative with
an increase in compressive strength. On
Figure 16 Percentage difference: ACI-318 (2014) and experimental results average, the ACI-318 (2014) code underes-
timates the experimental values by 1.8%.
For cylinder strengths in excess of 80 MPa,
70
the ACI-318 (2014) code overestimates by
50 as much as 6.1%. In total, 48.6% of values
are smaller than the code prediction. On
30
average, the ACI-318 (2014) code under-
10 estimates the study results by 6.0%, 4.3%
δ (%)
and 10.1% for Mix 1, Mix 2 and Mix 3

–10
respectively.
–30 Figures 17 and 18 show that the
EN 1992-1-1 (2004) code provides a rela-
–50
tively constant difference for strengths
–70 less than the design limit of 90 Mpa, on
0 10 20 30 40 50 60 70 80 90 100 110 120 130 140
average underestimating by 18.6%. In total
Hognestad et al (1955) Tan & Nguyen (2004) Pastor (1986)
10.1% of values are smaller than the code
Kaar et al (1978a) Khadiranaikar & Awati (2012) Ibrahim (1994) prediction. Compared to the study results,
Swartz et al (1985) Nedderman (1973) Mertol et al (2008) the EN 1992-1-1 (2004) code on average
Schade (1992) Kaar et al (1978b) Study results underestimates by 12.6%, 16.0%, and 27.1%
for Mix 1, Mix 2, and Mix 3 respectively.
Figure 17 Percentage difference: EN 1992-1-1 (2004) and experimental results If the same values for α1 and β1 assumed
for fc’=90 MPa, i.e. α1=0.70 and β1=0.8, are
assumed for strengths in excess of 90 MPa,
70
the EN 1992-1-1 (2004) code calculations
60 result in an ever increasing underestima-
50 tion of strength.
From Figure 18 it can be seen that the
Average δ (%)
40
BS 8110-1 (1997), SANS 0100-1 (2000) and
30
ACI-318 (2014) codes tend to be less con-
20 servative than the EN 1992-1-1 (2004) code
10 for all concrete grades, particularly for con-
0 crete with strengths in excess of 80 MPa.
All three design codes are calibrated to
–10
provide a suitable factor of safety against
0–10
10–20
20–30
100–110
110–120
30–40
40–50
50–60
60–70
70–80
80–90
90–100
>120
collapse, and the study results were mostly

underestimated. It can thus be concluded
Compressive strength (MPa) that the BS 8110-1 (1997), SANS 0100-1
BS 8110-1 (1997) & SABS 0100-1 (2000) difference (2000) and ACI-318 (2014) codes can be
ACI-318 (2014) difference used for the safe design of flexural mem-
EN 1992-1-1 (2004) difference bers and flexural compression members up
to 80 MPa, while the EN 1992‑1-1 (2004)
Figure 18 Effect of compressive strength on average percentage difference code can safely be used for the design
of flexural members with even higher for concrete strengths in excess of REFERENCES
strengths. 80 MPa. ACI (American Concrete Institute) 2014. ACI 318.
■■ The difference between the experi- Building Code Requirements for Structural Concrete
mental M-N interaction diagram and (ACI 318-95) and Commentary (ACI 318R-95).
Conclusions and the M-N interaction diagram gener- Farmington Hills, MI: ACI.
Recommendations ated by using the EN 1992-1-1 (2004) Bazant, Z 1984. Size effect in blunt fracture: Concrete,
code recommendations appears to be rock, metal. Journal of Engineering Mechanics,
Conclusions constant, underestimating the flexural 110(4): 518–535.
The observations and conclusions sum- strength for all concrete strengths. Bazant, Z 1999. Size effect on structural strength:
marised in this section were based on both For strengths in excess of 90 MPa, the A review. Applied Mechanics, 69: 703–725.
the analysis of the results obtained during EN 1992-1-1 (2004) code produces an BS (British Standard). 1997. BS 8110-1 1997. Structural
this study and results obtained by other ever increasing underestimation of Use of Concrete. Part 1: Code of Practice for Design and
researchers. flexural strength. Construction. London: British Standards Institution.
■■ The assumption that plane sections ■■ As all four design codes are calibrated to EN (European Standard) 2004. EN 1992-1-1 2004.
remain plane during bending holds true provide a suitable factor of safety against Eurocode 2: Design of Concrete Structures. Part 1-1:
for concrete containing dolomitic aggre- collapse, and the study results were General Rules and Rules for Buildings. Brussels:
gate, irrespective of the compressive mostly underestimated, the BS 8110-1 European Committee for Standardization (CEN).
strengths or specimen sizes considered (1997), SANS 0100-1 (2000) and ACI-318 Hognestad, E, Hanson, N & McHenry, D 1955.
in this study. (2014) codes can be used for the design Concrete stress distribution in ultimate strength
■■ The results confirm that a value of of flexural members and flexural com- design. Journal of the American Concrete Institute,
0.2 for the Poisson’s ratio is justifiable. pression members up to 80 MPa, while 52: 455–479.
However, a value of 0.28 is recommend- EN 1992-1-1 (2004) code is suitable for Ibrahim, H 1994. Flexural behavior of high-strength
ed owing to the large range of Poisson’s the safe design of flexural members with concrete columns. Edmonton, Canada: University
ratios. Additionally the Poisson’s ratio of even higher strengths. of Alberta, Department of Civil and Environmental
150 mm cylinders tends to remain con- Engineering.
stant at 0.23, while the Poisson’s ratio Recommendations Kaar, P, Hanson, N & Capell, H 1978a. Stress-strain
of the flexural compression specimens, The results presented in this study were characteristics of high strength concrete. ACI
increases with an increase in compres- based on a limited number of test results Special Publication, 55: 161–186.
sive strength. obtained from concrete containing a single Kaar, P, Fiorato, A, Carpenter, J & Corley, W 1978b.
■■ k 1, k 2 , and β1 were all found to be inde- aggregate type. Further investigation is Limiting strains of concrete confined by rectangular
pendent of the specimen size. required into the influence of specimen size hoops. Skokie, IL: Portland Cement Association,
■■ k 3, k 1k 3, α1, and εcu all showed a clear on the stress block parameters of concrete Research and Development/Construction
specimen size effect when calculated to confirm the trends observed in this study. Technologies Laboratories.
using the 150 mm cylinder strength. Based on the results obtained in Khadiranaikar, R & Awati, M 2012. Concrete stress
The size effect for k 3, k 1k 3, α1 are all this study, the 50 × 50 × 300 mm and distribution factors for high-performance concrete.
dependent on the cylinder strength. 100 × 100 × 500 mm flexural compression Journal of Structural Engineering, 138(3): 402–415.
■■ The size effect of non-standard (50 mm specimens can be used to determine the Kim, J & Eo, S 1990. Size effect in concrete specimens
or 100 mm) flexural compression mem- compression stress block parameters, pro- with dissimilar initial cracks. Magazine of Concrete
bers may be taken into account by using vided that the associated cylinder strength Research, 42(153): 233–238.
the associated cylinder strength. By is used in calculations. This paves the way Kim, J-K, Yi, S-T, & Yang, E-I 2000. Size effect
testing cylinders with a diameter similar for determining the stress block parameters on flexural compressive strength of concrete
to the dimensions of the flexural com- of high-strength concrete, a study field specimens. American Concrete Institute Structural
pression member (structural member), previously limited by compression machine Journal, 97(2): 291–296.
the size effect observed for k 3, k 1k 3 and capacity limitations. Mahachi, J 2013. Design of Structural Steelwork to
α1 can be eliminated. This means that From a flexural and flexural compres- SANS 10162, 3rd ed. Randburg: Xsi-tec.
the stress block parameters obtained by sion design perspective, the introduction of Mattock, A, Kriz, L & Hognestad, E 1961. Rectangular
testing 200 mm specimens may be used the EN 1992-1-1 (2004) may be welcomed, concrete stress distribution in ultimate strength
for all specimen sizes and the increased and this code of practice can safely be design. American Concrete Institute Journal, 57(2):
or decreased strength of the specimen used to design flexural compression ele- 875–928.
taken into consideration by using the ments with concrete strengths in excess of Mertol, H, Rizkalla, S, Zia, P & Mirmiran, A 2008.
size effect for cylinders in uniaxial 60 MPa. Characteristics of compressive stress distribution in
compression. high-strength concrete. American Concrete Institute
■■ The difference between the experimen- Structural Journal, 105(2): 626–633.
tal M-N interaction diagram and the Acknowledgements Nedderman, H 1973. Flexural stress distribution
M-N interaction diagram generated The authors would like to express in very-high strength concrete. Arlington, TX:
by using the BS 8110-1 (1997), SANS their gratitude to the Civil Engineering University of Texas at Arlington, Department of
0100-1 (2000) and ACI-318 (2014) code Department of the University of Pretoria Civil Engineering.
recommendations is not constant, for the use of their facilities, materials and Pastor, J 1986. High strength concrete beams. Ithaca,
underestimating for concrete strengths testing equipment, as well as for the sup- NY: Cornell University, Department of Civil
less than 80 MPa, and overestimating port received from laboratory staff. Engineering.
SANS (South African National Standard) Schade, J 1992. Flexural concrete stress in high strength bending properties of higher strength concrete.
2000. SANS 0100-1. The Structural Use of concrete columns. Calgary, Canada: University of ACI Special Publication, 87: 145–178.
Concrete, Parts 1 and 2. Pretoria: SABS Standards Calgary, Department of Civil Engineering. Tan, T H & Nguyen, N 2005. Flexural behavior
Division. Soliman, M T & Yu, C W 1967 The flexural stress- of confined high-strength concrete columns.
Sargin, M, Ghosh, S & Handa, V K 1971. Effects strain relationship of concrete. Magazine of American Concrete Institute Structural Journal,
of lateral reinforcement upon the strength and Concrete Research, 19(61): 223–238. 102(2): 198–205.
deformation properties of concrete. Magazine of Swartz, S, Nikaeen, A, Narayan Babu, H,
Concrete Research, 23(75–76): 99–110. Periyakaruppan, N & Refai, T 1985. Structural
Effect of chlorine and TECHNICAL PAPER
chloramine disinfection Journal of the South African
and the presence of ISSN 1021-2019

phosphorous and nitrogen UNAIS KADWA obtained his MSc Eng in Civil
on biofilm growth in dead

Engineering from the University of
KwaZulu-Natal as a research student in the
Centre for Research in Environmental,
zones on PVC pipes in

Coastal and Hydrological Engineering
(CRECHE) in 2015. He is currently a graduate
civil engineer in JG Afrika’s Hilton office
drinking water systems

where he specialises in the design and
implementation of water and irrigation infrastructure. His research
interests include biofilm inhibition, rainwater harvesting and hydraulic
modelling.
Contact details:
U Kadwa, M V Kumarasamy, D Stretch 6 Pin Oak Avenue
Hilton 3245
South Africa
T: +27 33 343 6700
In drinking water systems, ‘dead zones’ may have higher biofilm counts than areas of higher E: kadwau@jgafrika.com
flow rates, as there are limited/no shear stresses removing biofilms from the pipe material, and
disinfectant concentrations are significantly lower due to no/low supply of disinfected water DR MUTHUKRISHNA VELLAISAMY
KUMARASAMY (BE, ME (AU), PhD (IIT
from the reservoir. Biofilms affect the quality of drinking water and may lead to severe health
Roorkee)) is a Senior Lecturer in Civil
issues for downstream consumers. Biofilms can be controlled by disinfection, two common Engineering at the University of
disinfectants being chlorine and monochloramine. The growth of biofilms is also affected KwaZulu-Natal. His areas of research
by the presence of nutrients, as nutrients act as a food source to bacteria in the biofilms. To interests are water quality modelling, water
distribution system analysis, energy from
compare the growth of biofilms in ‘dead zones’ under different disinfectant (chlorine and
distribution networks and biofilm
chloramine) conditions, PVC coupons were placed in stagnant distilled disinfected waters growth control.
inoculated with 10% pond water (environmental source), and the biofilm growth was monitored
Contact details:
on the coupons using visual analyses. Chloramine has a better disinfectant inhibiting potential School of Engineering
than chlorine. To compare the influence of nutrients on biofilms in ‘dead zones’, coupons were University of KwaZulu-Natal
placed in distilled water inoculated with pond water from the same environmental source, Durban 4041
South Africa
and nitrogen and phosphorous concentrations were monitored. It was found that for ‘dead T: +27 31 260 7571
zones’ in drinking water networks, chloramine has a better biofilm inhibiting and inactivation E: kumarasamy@ukzn.ac.za
potential, and the presence of nitrates in water influence biofilm growth, and to a lesser extent
so does phosphorous. PROF DEREK STRETCH (BSc Eng, MSc Eng,
PhD (Cantab)) lectures Hydraulics and
Environmental Fluid Mechanics at the
University of KwaZulu-Natal. He currently
INTRODUCTION stagnant (non-flow) conditions may be occupies the eThekwini-sponsored Chair in
Biofilms are an encumbrance to drinking encountered where the consumption is low, Civil Engineering and is Director of the
Centre for Research in Environmental,
water systems, since they increase the pipe as well as in reservoirs and in buildings
Coastal and Hydrological Engineering. His
roughness (Babcock & Walton 2008) and (Manuel et al 2007). Research on drinking research group focuses on the bio-hydrodynamics of estuarine systems,
lead to deterioration of the quality of water water biofilms in stagnant biofilms is rare coastal and shoreline processes, and understanding turbulence and
(Kerr et al (2003). Biofilms may contain (Manuel et al 2007). Momba and Kaleni mixing in environmental flows.
pathogenic micro-organisms which, if not (2002) examined the effect of biofilm Contact details:
removed by disinfection, may reach end formation and growth on polyethylene and School of Engineering
University of KwaZulu-Natal
users and may cause outbreaks of disease galvanised steel containers used for water
Durban 4041
within a community (Simoes & Simoes storage by rural communities in South South Africa
2013). The most identified disease associ- Africa, and found that water stored for T: +27 31 260 1064
ated with waterborne outbreaks in devel- longer than 24 hours accumulated biofilms. E: stretchd@ukzn.ac.za
oped countries is gastroenteritis (Simoes Manuel et al (2010) found that periods

& Simoes 2013). The health effects vary in of stagnation promote biofilm growth.
severity and can range from mild gastro- Increased residence times in pipe networks
enteritis to severe (and sometimes fatal) increase the biofilm potentials. Lipphaus
diarrhoea, dysentery, hepatitis and typhoid et al (2014) found that taps opened after
fever (WHO 2011). periods of inactivity had higher biofilm
In a pipe network, flow conditions and bacterial counts. Disinfectant is Keywords: biofilm, drinking water, water quality, nutrients,
range from laminar to turbulent flow, but added to water to inhibit the growth of disinfectant
Kadwa U, Kumarasamy MV, Stretch D. Effect of chlorine and chloramine disinfection and the presence of phosphorous and nitrogen on biofilm growth in
dead zones on PVC pipes in drinking water systems. J. S. Afr. Inst. Civ. Eng. 2018:60(4), Art. #1629, 6 pages. http://dx.doi.org/10.17159/2309-8775/2018/v60n4a5 45
biofilms. Chlorine and monochloramine
are the most commonly used disinfect-
ants in drinking water systems (Momba
et al 2002). Monochloramine disinfection
occurs when ammonia is added to chlorine.
Although monochloramine is less reactive
than chlorine, monochloramine is a more
effective disinfectant (Turetgen 2007; Park
(a) (b)
& Kim 2008; WHO 2000), since it is more
persistent and maintains a higher residual Figure 1 ( a) Raw SEM image before image processing, and (b) after noise removal, background
disinfectant throughout the network and removal and edge detection
also penetrates the biofilm more effectively
(Pressman et al 2012). However, a potential surface roughness enhances biofilm adhe- up of pond water (from an environmental
problem with monochloramine is that sion to substrates during the initial steps source). (A pilot test with pure distilled
ammonia is a by-product of monochlora- of colonisation, as it provides more surface water showed insufficient growth over a
mine decomposition (Dvorak & Skipton area for cell attachment and reduces the 40-day period for any useful analyses.) The
2013). Ammonia is a primary nutrient for shear force of attached bacterial cells (Lars tests were carried out over a period of 12 to
biofilm growth and its presence in water & Douglas 2011). For this research PVC 14 days, since initial tests on disinfectant
will encourage biological growth and bac- pipes were chosen, as PVC is the most decay showed that chloramine (the slower
terial resistance (McVay 2009). commonly used pipe in municipal reticula- decaying one of the two disinfectants test-
Nutrients present in drinking water tion networks in South Africa (Van Zyl ed) takes approximately 11 days to decay to
also have an influence on the biofilm 2014). Furthermore, the biofilm-forming below detectable limits (below 0.05 mg/ℓ),
formation as they are a ‘food source’ potential for plastic pipes has not been fully after which time the behaviour of the
(Conovera & Corner 1968). TOC (total investigated yet (Momba et al 2000). biofilms could be observed. Beakers were
organic carbon) is the controlling nutri- Another challenge with steel pipes sealed off and covered in aluminium foil
ent, followed by nitrogen (N) and then is microbiologically induced corrosion to prevent interaction with the atmosphere
phosphorous (P) (Kerr et al 2003). Carbon (MIC), also known as bio-corrosion. MIC and UV light entering.
is present in the water network in humic happens when electrochemical processes The growths of biofilms were moni-
acids, carbohydrates, carboxylic acids in the presence of micro-organisms are tored by SEM. At the given times, coupons
and proteins (LeChevallier et al 1997). able to initiate, facilitate or accelerate the were removed from the test waters and
The optimum nutrient condition for corrosion reaction in the pipe material prepared for SEM imaging. Six pictures of
biofilm growth is a C:N:P ratio of around (Usher et al 2014). For this research, visual each coupon were taken at a 400x zoom
106:10:1 (Zhang & DiGiano 2002). From analyses on pipe coupons were carried out level. The images obtained were analysed
the ratio, it can be seen that after carbon, from SEM (scanning electron microscopy) using ImageJ software. The biofilm growth
nitrogen is the most important nutrient imaging, and the area of biofilm cover on was quantified by image-processing tech-
for biofilm growth. Nitrogen is used by each coupon was computed to compare niques where biofilm growth present on
microorganisms to build genetic material biofilm growth observed on coupons in the coupon substrate was traced out and
and amino acid; however, the exact role different disinfectant regimes (chlorine and the area coverage of biofilm was calculated.
of nitrogen in bacterial growth is unclear monochloramine) and different nutrient By use of software, each SEM image was
(US EPA 1992). Nitrification is a process conditions (nitrogen and phosphorous). processed by noise removal, background
whereby ammonia – a reduced form of removal and then edge detection. Figure 1
nitrogen – is oxidised into nitrites and then shows an example of how images were
nitrates. The presence of ammonia and MATERIALS AND METHODS processed and biofilms were outlined to
nitrite promotes the growth of nitrifying The lab experiments for this research were calculate the biofilm area covered.
bacteria (US EPA 1992). The ammonia that carried out in two parts – Experiment A A VIS (visible light) spectrophotometer
is present can react with the free chlorine (that looked at the biofilm-disinfectant (Hach DR3900®) was used to determine the
disinfectant generating chloramines, and relationship) and Experiment B (that looked concentration of chlorine and chloramine
reduce the disinfection effectiveness. The at the biofilm-nutrient relationship). Plastic disinfectants (experiment A), and inorganic
pipe material that is used also influences coupons 10 mm × 10 mm were cut from a nitrogen and phosphorous (Experiment B).
the growth of biofilms (Jang et al 2011; PVC pipe and placed in a beaker that was The method entails that standard test pow-
Morvay et al 2011). Biofilms develop more then filled with the respective samples of der pillows are added to a 50 mℓ sample
quickly and support a more diverse biofilm prepared test water. taken at given times from the test waters.
population on iron pipe surfaces than For Experiment A, the test waters The powder reacts with the disinfectant
PVC plastic pipes, while cement-based were chlorinated distilled water (initial or the nutrients in the sample. The sample
pipes have higher biofilm counts than concentration of 2.4 mg/ℓ Cl2) and mono- is then placed in the spectrophotometer,
stainless steel pipes (Momba et al 1998). chloraminated distilled water (initial and light of a certain wavelength is passed
Compared to other materials, PVC pipes concentration of 2.4 mg/ℓ NH2Cl), and for through the sample. The spectrophoto
have a smoother surface, and hence biofilm experiment B, the test water was distilled meter measures how much light has been
adhesion is more difficult (Kerr et al 2003; water with no addition of disinfectant. All absorbed by the sample and calculates the
Momba et al 1998). Nano- and micro-scale tests had 10% of the working volume made concentrations based on light absorption.
16
14
12
Disinfectant residual (mg/ℓ)

10
Biofilm cover (%)
0.5
0 0.05
Day 0 Day 1 Day 4 Day 7 Day 10 Day 14
Times (days)
Figure 2 B
iofilm growth on coupons placed in chlorine-disinfected (orange) and chloramine disinfected (blue) water conditions; the solid line ( )
with squares (■) represents decay of chlorine disinfectant, while the dashed line ( ) with crosses (x) represents the decay of chloramine
disinfectant, and the error bars represent standard error
The concentrations of nutrients and chloramine is a significantly more stable cover on chlorine coupons reached an
disinfectants were taken four times from disinfectant. The stability of chloramine average of 12%. At a very low concentra-
each sample at the chosen times. At times over chlorine has also been observed by tion of monochloramine coupons, the
less than four values were shown (see other researchers (Turetgen 2007; Park & average biofilm cover present had not
Figure 1). This is due to the closeness of the Kim 2008). changed significantly since Day 7, going
concentrations, hence symbols represent- From Figure 2 it can be seen that the up from 5% on Day 7 to 6% on Day 10.
ing values may be overlapping or ‘on top’ of difference in biofilm inhibition between The choice of disinfectants has an effect
one another. chlorine and monochloramine is not on ‘dead zones’ with respect to the bio-
significantly different over the first seven film formation, but the difference is not
days. At Day 1 there is no difference in substantial.
RESULTS AND DISCUSSION the biofilm cover between chlorine and On Day 10, the biofilm cover on chlo-
monochloramine (both at around 1% rine coupons ‘spiked’ due to the chlorine
Experiment A cover). At Day 4 there is an average cover being below detection limits (below
The chlorine and monochloramine concen- of 2% on the chlorine coupons, whilst there 0.05 mg/ℓ). However, after the monochlo-
trations were both 2.4 mg/ℓ at the start of is a 3.5% cover on the monochloramine ramine reached very low levels (on Day 10
the experiment. As can be seen in Figure coupons. At Day 7 the difference in biofilm and later), the biofilm cover on coupons did
2, in line with other published research, coupons becomes slightly more visible, not ‘spike’, but rather remained constant.
there is a first-order decay relationship with chlorine coupons having an average Monochloramine is a better disinfectant
between chlorine disinfectant and time, biofilm cover of 3% while monochloramine at penetrating and inactivating the biofilm
and monochloramine disinfectant and time coupons have an average biofilm cover of present, as also found by other authors
(Chambers et al 1995; Hua et al 1999). The 5%. The biofilm covers observed are not a (LeChevallier et al 1990; LeChevallier et al
concentration of disinfectant at time t is representation of actual effectiveness of the 1991). Due to it being more stable and less
represented mathematically in Equation 1. disinfectants at constant residuals, as the reactive than chlorine, monochloramine
disinfectant concentrations were allowed to can diffuse into the biofilm and eventually
Ct = C0 e–λt(1) decay. The reasons for allowing the disin- inactivate attached bacteria (Coniglio et al
fectants to decay are: 2015). Free chlorine is consumed before
where λ is the decay constant. ■■ To compare the decay kinetics and it has a chance to react with the bacte-
stability of both disinfectants. rial components of the biofilm (Chen &
From Figure 2, the decay constant λ for ■■ To simulate conditions similar to that of Stewart 1996). Momba et al (2002) have
chlorine and chloramine is 0.711 and 0.556 a ‘dead zone’ in a pipe network (such as suggested using a dual chlorine-mono-
respectively. Chlorine reached to below a tank/reservoir or pipes where stagna- chloramine system for effective limitation
detectable limits (0.05 mg/ℓ) much sooner tion occurs). of biofilm to combine the effects of a
(7 days) compared to chloramine (11 days). The most significant difference in biofilm quick-reacting chlorine disinfectant with
The decay constant λ of chloramine is growth occurred on Day 10. After the a more persistent, penetrating and stable
78% of the λ of chlorine, showing that depletion of chlorine on Day 7, the biofilm chloramine disinfectant.
18 3.5
16
Nutrien concentration (mg/ℓ)

3.0
14
2.5
Biofilm cover (%)
12
10 2.0
8 1.5
6
1.0
4
2 0.5
0 0
Day 0 Day 3 Day 6 Day 9 Day 12
Times (days)
Figure 3 B
iofilm growth and presence of inorganic nitrogen in the 12-day lab tests; crosses (x) represent ammonia, circles (●) represent nitrates,
diamonds (◆) represent nitrites, and error bars represent standard error
18 0.09
16 0.08
Nutrien concentration (mg/ℓ)

14 0.07
Biofilm cover (%)
12 0.06
10 0.05
8 0.04
6 0.03
4 0.02
2 0.01
0 0
Day 0 Day 3 Day 6 Day 9 Day 12
Times (days)
Figure 4 B
iofilm growth and presence of phosphorous for the 12-day lab tests; circles (●) represent phosphorous, and error bars represent standard error
There is a ‘crash’ in the biofilm cover effectiveness as a disinfectant (Pressman et growth, while greater biofilms and bacteria
on chlorine coupons after Day 10. This is al 2012). present release more ammonia through
due to the large population of cells in the There is a correlation between nitrate respiration and excretion (Conovera &
biofilm matrix consuming the available concentrations and biofilm growth. Corner 1968) and will ultimately act as
nutrients, thereby leading to a shortage of Figure 3 shows that nitrates support and a ‘food source’ for bacteria and biofilms.
nutrients that act as a ‘food source’ to the promote biofilm growth. From Day 0 to After the peak on Day 9, the large per-
bacterial cells. In the closed environment, Day 9 the biofilm cover increased from 0% centage of biofilm cover meant that the
and with limited nutrient availability, the to an average of 13%, peaking on Day 9. nutrients present were unable to provide
biofilm cover ‘crashes.’ This leads to the After Day 9 the biofilm cover decreased ‘food’ for all the micro-organisms present
dead cells being decomposed and releasing sharply. The peak biofilm cover on the pipe and that a very large portion of the nitrates
ammonia, which acts as a food source and coupons occurred simultaneously with present had been depleted by the large pop-
hence causes an increase in the biofilm the peak nitrate concentrations in the test ulation of biofilms. This led to the dying
cover present (Baibeau 2011). water. The peak nitrate concentration on off or ‘crash’ of biofilms due to there being
Day 9 was 3.22 mg/ℓ, coinciding with the insufficient nutrients. Dead biofilms lose
Experiment B peak biofilm cover of 13%. After the ‘crash’ their adhesive properties and are detached
The nitrogen cycle observed for in the biofilms, there was also a ‘crash’ from the coupon substrate (Winstanley
Experiment B is shown in Figure 3. in the nitrate concentrations in the test et al 2010).
Ammonia is oxidised into nitrites and then water. From Day 9 to Day 12 the biofilm Many researchers have demonstrated
further oxidised into nitrates. Experiment cover average fell from 13% to 3.9%, while the importance of phosphorous for bac
A showed that monochloramine is a the nitrate concentrations fell from 3.22 terial growth (Lehtola et al 2001; Smith &
more stable disinfectant than chlorine. mg/ℓ on Day 9 to 0.7 mg/ℓ on Day 12. Prairie 2004). Figure 4 shows the rela-
However, when monochloramine decom- Other researchers have also shown similar tionships between biofilm growth and
poses, ammonia is a by-product (Dvorak relationships between bacteria and nitrate phosphate concentrations. As shown in
& Skipton 2013). As monochloramine acts (Feitag et al 1987; Chu et al 2005). Figure 4, the phosphate concentrations
as a source of ammonia, the ammonia will The biofilms on the substrate and the are really low and are almost constant in
ultimately be converted to nitrates that will nutrients present are co-dependent. Higher relation to inorganic nitrogen components.
promote biofilm growth, thus limiting its levels of nutrients stimulate more biofilm However, it can be seen that, as the biofilm
covers increased between Day 6 and Day 9, CONCLUSION AND disinfectant decay, nutrient limitations and
the phosphorous levels decreased from 0.06 RECOMMENDATIONS biofilm growth do not occur in isolation
mg/ℓ to 0.03 mg/ℓ, showing that biofilms For ‘dead zones’ in pipe networks, chlorine of one another. The results of this will be
also use phosphorous as a ‘food’ source. is slightly better at inhibiting biofilm able to relate disinfectant decay to nutrient
After carbon and nitrogen, phosphorous formation in the short term (Day 0 to availability. Organic nitrogen should also
is the most important nutrient for biofilm Day 7); however, monochloramine is a be monitored during the course of the
growth (LeChevallier et al 1991). more persistent disinfectant and is better experiment to relate the biofilm growth to
able to penetrate bacteria present in the the total nitrogen present and to the total
Comparison between water. Chloramine’s ability to penetrate Kjehdahl nitrogen (TKN).
Experiment A and Experiment B and inactivate biofilm over a longer period
Using Experiment B as a control, it was of time means that monochloramine has a
observed that the trend for biofilm growth better biofilm-inhibiting ability. Because of ACKNOWLEDGEMENTS
in the control (no disinfectant) and chlo- chloramine’s inactivation and penetrating This work was supported by the Deutscher
rine is similar. Both show an increase in ability, even after chloramine has been Akademischer Austausch Dienst (DAAD),
biofilm cover from the Day 0 until there is completely decayed, lower biofilm covers the National Research Foundation (NRF),
a peak biofilm cover, followed by a biofilm will be observed, compared to chlorine. For uMgeni Water and WRC South Africa.
‘crash.’ However, the rate at which the ‘dead zones’ in water networks, the risk of
biofilms grew were different. From Day 0 complete disinfectant decay is much higher
to Day 6, the average biofilm cover on than in zones which experience continu- REFERENCES
Experiment B coupons grew from 0% cover ous flow. For future research, a combined Babcock, R, Walton, R 2008. Understanding the
to 4.6% cover. From Day 0 to Day 7, the chlorine-monochloramine disinfection Impact of Biofilm Growth on Pipe Roughness. Paper
average biofilm cover on chlorine coupons could be set up to see the combined effect presented at the World Environmental and Water
from Experiment A grew from 0% cover to of a fast-reacting chlorine disinfectant and Resources Congress, American Society of Civil
13% cover. This shows the effectiveness of a more stable persistent monochloramine Engineers, 12–16 May, Honolulu, Hawaii.
chlorine in the short term (7 days) when disinfectant.. Baibeau, H 2011. Microbiology and isolation of
compared to the biofilm cover observed in Statistical analysis showed that there nitrifying bacteria. Fundamentals and control
Experiment B. was a strong relationship between the of nitrification in chloraminated drinking water
The peak on both the chlorine coupons nitrates and the biofilm, hence affirming distribution system. AWWA Manual of Water
from Experiment A and the coupons from the fact that nitrogen is a primary nutri- Supply Practices M56, 69–73.
Experiment B were very similar (13% – ent. Also, it was found that there was a Chambers, V K, Creasey, J D & Joy, J S 1995. Modeling
14.5%). This shows that the ‘carrying carrying capacity of the closed system, free and total chlorine decay in potable water
capacity’ of the water samples was 13% – beyond which biofilm populations present distribution-systems. Journal of Water Supply
14.5%. Beyond this, the nutrients were ‘crashed’. Research and Technology-Aqua, 44(2): 60–69.
used up, and this led to a ‘crash’ in the Biofilms should be identified through Chen, X & Stewart, P S 1996. Chlorine penetration into
biofilm cover present. The trend observed methods that are currently being used artificial biofilm is limited by a reaction-diffusion
in the monochloramine coupon was in biology/microbiology, such as DNA interaction. Environmental Science and Technology,
significantly different to the Experiment extraction and analysis, to see if there are 30(6): 2078–2083.
B coupons and the chlorine coupons. On different strains that grow under different Chu, C, Lu, C & Lee, C 2005. Effects of inorganic
the monochloramine coupons, the biofilm disinfectant conditions. Other physico- nutrients on the regrowth of heterotrophic bacteria
cover increased continuously from Day chemical factors that contribute to biofilm in drinking water distribution systems. Journal of
0 to Day 7. Over the seven-day period, growth, such as biological oxygen demand Environmental Management, 74(3): 255–263.
biofilm cover grew from 0.0% to 5%. From (BOD), chemical oxygen demand (COD), Coniglio, M A, Stefano, M & Mohamed, Y H 2015.
Day 7 onwards to Day 12, the biofilm etc, should be investigated as well. Monochloramine for controlling legionella in
cover remained at around 7%, even though Together with SEM, atomic force biofilms. How much we know? Journal of Nature
the chloramine disinfectant had decayed microscopy (AFM) should be used for and Science, 1(2): e44.
and was below detection limits (below biofilm analysis to provide a roughness Conovera, R J & Corner, E D S 1968. Respiration and
0.05 mg/ℓ). This shows the effectiveness of profile of the specimen. Roughness influ- nitrogen excretion by some marine zooplankton in
the persistence and inactivation potential ences the head loss in pipelines, and a relation to their life cycles. Journal of the Marine
of monochloramine when compared to the roughness profile will be able to provide Biological Association of the United Kingdom, 48(1):
biofilm cover observed in Experiment B. thicknesses of biofilms, especially when 49–75.
Also, at low residuals, chlorine has a much biofilms are ‘stacked’ on top of one another. Dvorak, B I & Skipton, S O 2013. Chloramines water
diminished effect on the biofilm growth, Inoculation for the water samples should disinfection: Omaha Metropolitan Utilities District
as the trend observed with biofilm growth be done using known cultures, to have as and Lincoln Water System. Lincoln, NE: University
on chlorine disinfectant was very similar much uniformity as possible, as biological of Nebraska.
to Experiment B (control) where nitrates activities in pond water are dynamic and McVay, R D 2009. Chlorine production and monitoring
were the driving force of biofilm. Visually, vary according to temperature, UV light, in Florida’s water supply systems. Florida Water
both Experiment A and Experiment B depth and season. The testing should Resources Journal, 16–26.
had very similar biofilm structures. Both combine disinfectant decay and nutrient Freitag, A, Rudert, M & Bock, E 1987. Growth of
had a gel-like EPS which adhered to the concentrations instead of running each nitrobacter by dissimilatoric nitrate reduction.
pipe coupons. experiment individually, as the effects of FEMS Microbiology Letters, 48(2): 105–109.
Hua, F, West, J R, Barker, R A & Forster, A C F 1999. Manuel, C M, Nunes, O C & Melo, L F 2010. Unsteady Smith, E M & Prairie, Y T 2004. Bacterial metabolism
Modeling of chlorine decay in municipal water state flow and stagnation in distribution systems and growth efficiency in lakes: the importance
supplies. Water Research, 33(12): 2735–2746. affect the biological stability of drinking water. of phosphorus availability. Limnology and
Jang, H J, Choi, Y J & Ka, J O 2011. Effects of diverse Biofouling, 26(2): 129–39. Oceanography, 49(1): 139–147.
water pipe materials on bacterial communities and Momba, M, Cloete, T, Venter, S N & Kfir, R 1998. Turetgen, I 2007. Comparison of the efficacy of free
water quality in the annular reactor. Journal of Evaluation of the impact of disinfection processes residual chlorine and monochloramine against
Microbiology and Biotechnology, 21(2): 115–123. on the formation of biofilms in potable surface biofilms in model and full scale cooling towers.
Kerr, C J, Osborn, K S, Rickard, A H, Robson, G D & water distribution systems. Water Science and Biofouling, 20(2): 81–85.
Handley, P S 2003. Biofilms in water distribution Technology, 38(8–9): 283–289. US EPA (United States Environmental Protection
systems. In: Mara, D & Horan, N J (Eds). Handbook Momba, M, Kfir, R, Venter, S & Cloete, T 2000. An Agency) 1992. Control of biofilm growth in
of Water and Wastewater Microbiology. San Diego, overview of biofilm formation in distribution drinking water distribution systems. Report No
CA: Elsevier, 757–775. systems and its impact on the deterioration of water EPA/625/R-92/001. Washington, DC: Office of
Lars, R D & Douglas, B W 2011. Physiochemical quality. Water SA, 26(1): 59–66. Research and Development.
regulation of biofilm formation. MRS Bulletin/ Momba, M N B & Kaleni, P 2002. Regrowth and Usher, K M, Kaksonena, A H, Cole, I & Marney,
Materials Research Bulletin, 36(5): 347–355. survival of indicator microorganisms on the D 2014. Critical review: microbially influenced
LeChevallier, M W, Lowry, C D & Lee, R G 1990. surfaces of household containers used for the corrosion of buried carban steel pipes. International
Disinfection of biofilms in a model distribution storage of drinking water in rural communities of Biodeterioration and Biodegradation, 93: 84–106.
system. Journal of the American Water Works South Africa. Water Research, 36: 3023–3028. Van Zyl, J E 2014. Introduction to operation and
Association, 82(7): 87–99. Momba, M, Ndaliso, S, Binda, M & Makala, N 2002. maintenance of water distribution systems. WRC
LeChevallier, M, Schulz, W & Lee, R G 1991. Bacterial Inhibition of biofilm regrowth in potable water Report No TT 600/14. Pretoria: Water Research
nutrients in drinking water. Applied Environmental systems. Pretoria: Water Research Commision. Commission, 59–74.
Microbiology, 57(3): 857–862. Morvay, A A, Decun, M, Scurtu, M, Sala, C, Morar, Winstanley, H F, Chapwanya, M, McGuinness, M J
LeChevallier, M, Schulz, W & Lee, R G 1997. Bacterial A & Sarandan, M 2011. Biofilm formation on & Fowler, A C 2010. A polymer-solvent model of
nutrients in drinking water. Applied Environmental materials commonly used in household drinking biofilm growth. Proceedings of the Royal Society A,
Microbiology, 57(3): 857–862. water systems. Water Science and Technology: Water 467(2129): 1449–1467.
Lehtola, M J, Miettinen, I T, Vartiainen, T & Supply, 11(2): 252–257. WHO (World Health Organization) 2000.
Myllykangas, T 2001. Microbially available organic Park, S K & Kim, Y K 2008. Effect of chlorine Disinfectants and disinfectant by-products. In:
carbon, phosphorus, and microbial growth in concentration on biofilm maintenance on pipe Environmental Health Criteria 216. Geneva:
ozonated drinking water. Water Research, 35(7): surfaces exposed to nutrient-limited drinking water. WHO, 529.
1635–1640. Water SA, 34(3): 373–380. WHO (World Health Organization) 2011. Microbial
Lipphaus, P, Hammes, F, Kotzsch, S, Green, J, Gillespie, Pressman, J G, Lee, W H, Bishop, P L & Wahman, D G fact sheets. In: Guidelines for Drinking-Water
S & Nocker, A 2014. Microbiological tap water 2012. Effect of free ammonia concentration on Quality, 4th ed. Valleta, Malta: Gutenberg, 231–305.
profile of a medium-sized building and effect monochloramine penetration within a nitrifying Zhang, W & DiGiano, F A 2002. Comparison of
of water stagnation. Environmental Technology, biofilm and its effect on activity, viability and bacterial regrowth in distribution systems using
35(5–8): 620–628. recovery. Water Research, 46(3): 882–894. free chlorine and chloramine: a statistical study of
Manuel, C M, Nunes, O C & Melo, L F 2007. Dynamics Simoes, L C & Simoes, M 2013. Biofilms in drinking causative factors. Water Research, 36(6): 1469–1482.
of drinking water biofilm in flow/non-flow water: problems and solutions. Royal Society of
conditions. Water Research, 41: 551–562. Chemistry Advances, 3: 2520–2533.
Catchment response time TECHNICAL PAPER
and design rainfall: the Journal of the South African
key input parameters for ISSN 1021-2019
design flood estimation in DR JACO GERICKE Pr Eng graduated with
ungauged catchments
B Tech Eng (Civil) and M Tech Eng (Civil)
degrees from the Central University of
Technology, Free State (CUT). He was also
awarded the BSc (Hons) Appl Sci Water
Resources Eng degree by the University of
Pretoria, the MSc Eng (Civil) degree by
O J Gericke
Stellenbosch University and a
PhD Eng (Agriculture) degree by the University of KwaZulu-Natal. He has
20 years of professional and academic experience, and has published a
Catchment response time and design rainfall are regarded as fundamental input to all design number of papers in the design hydrology field. He is currently a Senior
Lecturer at the CUT.
flood estimation methods in ungauged catchments, while errors in estimated catchment
response time and design rainfall directly impact on estimated peak discharges. This paper Contact details:
Unit for Sustainable Water and Environment
presents the independent testing and comparison of the latest catchment response time
Department of Civil Engineering
and design rainfall estimation methodologies with current well-known and simplified Central University of Technology, Free State
methodologies used in South Africa to ultimately highlight the impact thereof on design flood Private Bag X20539
estimation. The results confirmed that catchment response time, design rainfall, and to some Bloemfontein 9300
South Africa
lesser extent runoff coefficients, are the key input parameters for design flood estimation in T: +27 51 507 3516 / +27 73 582 6812
ungauged catchments and have a significant impact on the design of hydraulic structures. E: jgericke@cut.ac.za
It is recommended that the current well-known and simplified catchment response time
(USBR TC equation) and design rainfall (modified Hershfield/TR102 DDF approach) estimation
methodologies should be replaced with the empirical G&S TC equations and the RLMA&SI
DDF approach when deterministic design floods are estimated in ungauged catchments in
South Africa.
INTRODUCTION Single-event deterministic design flood

Design flood estimation is necessary for the estimation methods are the most com-
planning, design and operation of hydraulic monly used by practitioners in ungauged
structures, e.g. culverts, bridges and/or spill- catchments (Van Vuuren et al 2012).
ways at a particular site in a specific region In the application of these single-event
(Pegram & Parak 2004). In South Africa, deterministic methods, all complex,
three basic approaches to design flood esti- heterogeneous catchment processes are
mation are available, namely the probabil- lumped into a single process to enable the
istic, deterministic and empirical methods estimation of the expected output (design
(Parak & Pegram 2006; Smithers 2012; Van flood) from causative input (design rainfall)
der Spuy & Rademeyer 2016). In gauged in a simple and robust manner (Gericke &
catchments, despite uncertainties and errors Du Plessis 2013). Design rainfall comprises
in measurement, observed peak discharges a depth and duration (directly proportional
are regarded as the best estimate of the to the catchment response time) associated
true peak discharge (Gericke & Smithers with a given annual exceedance probability
2016b). In terms of design flood estimation (AEP) or return period (T).
in gauged catchments, probabilistic methods The catchment response time is nor-
that are adequate in both length and qual- mally expressed as a single time parameter,
ity of data are normally used to conduct a e.g. time of concentration (TC), lag time
frequency analysis of observed flood peak (TL) and/or time to peak (TP). In other
data from a flow-gauging site (Smithers words, estimates of peak discharge are
2012). In ungauged catchments, practition- based on a single representative catchment
ers are required to estimate design floods response time parameter (e.g. TP, TC and/
using either deterministic and/or empirical or TL), while the catchment is at an ‘average
methods, although regional probabilistic condition’ and the hazard or risk associ-
methods or continuous simulation models ated with a specific event is reflected by
could also be used to transfer design values the joint-probability of the 1:T-year design Key words: catchment response time, design rainfall,
from gauged to ungauged sites. rainfall and 1:T-year design flood events design flood estimation, ungauged catchments
Gericke OJ. Catchment response time and design rainfall: the key input parameters for design flood estimation in ungauged
catchments. J. S. Afr. Inst. Civ. Eng. 2018:60(4), Art. #0327, 17 pages. http://dx.doi.org/10.17159/2309-8775/2018/v60n4a6 51
(Rahman et al 2002; SANRAL 2013). This development, without using any local cor- the methodologies involved in meeting
assumption considers the probabilistic rection factors. In a recent study, Gericke the objectives are detailed, followed by the
nature of rainfall, but the probabilistic and Smithers (2016b; 2017) used observed results, discussion and conclusions.
behaviour of other inputs and parameters catchment response time parameters
is ignored. Taking into consideration the to derive new empirical time parameter
vast complexity and spatial and temporal equations for medium to large catchments STUDY OBJECTIVES AND
variability of catchment processes and in South Africa. These derived catchment- ASSUMPTIONS
their driving forces, as well as the probable specific/regional empirical time parameter The overall objective of this study is to
significant bias introduced by ignoring equations (referred to as G&S TC equation independently test and compare the latest
the joint-probability of rainfall and runoff, in this paper) resulted in improved peak catchment response time and design rain-
it is not surprising that only relatively discharge estimates when compared to the fall estimation methodologies with current
simple deterministic methods represent- USBR TC equation in 60 of the 74 catch- well-known and simplified methodologies
ing the real world processes are recog- ments considered. used in South Africa to ultimately high-
nised and used in design flood practice In terms of design rainfall, Gericke and light the impact thereof on design flood
(Smithers 2012). Du Plessis (2011) evaluated five depth- estimation. The specific objectives are
Catchment response time and design duration-frequency (DDF) approaches to: (i) conduct at-site probabilistic flood
rainfall are therefore regarded as funda- commonly used in South Africa to frequency analyses in gauged catchments,
mental input to all design flood estimation estimate design rainfall depths. The DDF (ii) compare and evaluate the combined use
methods in ungauged catchments, while approaches that were evaluated included of the recommended USBR TC equation
errors in estimated catchment response those based on: (i) Log-Extreme Value (USBR 1973) and modified Hershfield DDF
time and design rainfall will directly Type I (LEV1) distributions (Midgley approach and/or TR102 design rainfall
impact on estimated peak discharges. & Pitman 1978), (ii) Technical Report information (Adamson 1981) to the com-
Bondelid et al (1982) indicated that as 102 (TR102) daily design rainfall infor- bined use of the empirical G&S TC equa-
much as 75% of the total error in design mation (Adamson 1981), (iii) Regional tions (Gericke & Smithers 2016b; 2017) and
peak discharge estimates in ungauged Linear Moment Algorithm South African the RLMA&SI DDF approach (Smithers &
catchments could be ascribed to errors Weather Services (RLMA-SAWS) Schulze 2003; 2004), (iii) translate the time
in the estimation of catchment response n-day design point rainfall information parameter and design rainfall estimation
time parameters. Grimaldi et al (2012) (Smithers & Schulze 2000b), (iv) modified results to design peak discharges using
highlighted that estimates of catchment Hershfield equation (Alexander 2001), and an appropriate single-event deterministic
response time, using different equations, (v) Regional Linear Moment Algorithm design flood estimation method, (iv) verify
may differ from each other by up to and Scale Invariance (RLMA&SI) approach and test the consistency, robustness and
500%. Gericke and Smithers (2014) also (Smithers & Schulze 2003; 2004). It was accuracy of the deterministic design esti-
showed that the underestimation of time recommended that the M&P/LEV1 and mates (QT) by comparing these design esti-
parameters by 80% or more could result modified Hershfield DDF relationships mates with the at-site probabilistic flood
in slightly lower design rainfall depths, should be seen as conservative estimates, frequency analyses (QP), and (v) highlight
although of much higher intensities, hence and their use should be limited to the impact of these over- or underestima-
ultimately resulting in the overestimation small (TC ≤ 6 hours) and medium-sized tions on prospective hydraulic designs,
of design peak discharges of up to 200%. (6 < TC ≤ 24 hours) catchments, while the while attempting to identify the influence
Empirical time parameter estimation RMLA&SI approach should be regarded of possible source(s) that might contribute
methods are widely used in South Africa, as the standard DDF relationship for all to the differences in the estimation results.
with only the TL methods proposed by catchment response times and catchment The Standard Design Flood (SDF)
Pullen (1969) and Schmidt and Schulze sizes under consideration. method (Alexander 2002; Gericke & Du
(1984) being developed locally. In terms of Potential future improvements in peak Plessis 2012; SANRAL 2013) was selected
TC estimation, the empirical Kerby (1959) discharge estimation using event-based as the most suitable single-event deter-
and United States Bureau of Reclamation design flood estimation methods will not ministic method to estimate the design
(USBR 1973) equations are recommended be realised if practitioners continue to use peak discharges, since it is: (i) a regionally
for general use in South Africa for overland inappropriate time parameter and design calibrated version of the Rational method
and channel flow conditions respectively rainfall estimation methods. Not only will and is not subject to user-biasedness in
(SANRAL 2013). However, both these the accuracy of design flood estimation terms of the selection of site-specific runoff
equations were developed and calibrated methods be limited, but it will also have coefficients, (ii) deterministic-probabilistic
in the United States of America (USA) for an indirect impact on hydraulic designs, in nature, and (iii) applicable to catchment
catchment areas less than 4 ha and 45 ha i.e. underestimated time parameter values areas up to 40 000 km², which coincide
respectively (McCuen et al 1984). Gericke and higher design rainfall intensities will with the catchment area ranges considered
and Smithers (2014; 2016a; 2016b) high- result in overdesigned hydraulic structures, in this study, e.g. 28 km² to 31 283 km².
lighted the inherent limitations and incon- and the overestimation of time parameters The use of the SDF method is further
sistencies introduced when these TC equa- associated with lower design rainfall inten- justified given that the primary focus of
tions, which are currently recommended sities will result in underdesigns. this paper is on the impact of catchment
for general practice in South Africa, are The study objectives and assumptions response time and design rainfall estimates
applied outside their bounds, both in terms are discussed in the next section, followed on peak discharge, and not on the design
of areal extent and their original regions of by a summary of the study area. Thereafter, flood estimation method itself.
20°E 25°E 30°E
25°S
25°S
Legend
● Flow-gauging station
NR
CR
30°S
SWCR
ESCR
30°S
Primary drainage regions
Projected coordinate system:

WGS-1984
35°S Projection:
Africa Albers Equal-Area
Scale:
Not defined
35°S Source:
DWAF (1995)
20°E 25°E 30°E
Figure 1 Location of the 48 gauged catchments in the four climatological regions
This study is based on the following frequency analyses that need to be con- precipitation (MAP), rainfall type, distribu-
assumptions: ducted to convert observed daily point tion and rainfall seasonality), catchment
■■ The conceptual TC equals T P. The rainfall to a design rainfall depth associ- geomorphology, channel geomorphology,
conceptual TC is normally defined ated with the catchment response time, geographical location, and altitude above
as the time required for the entire as well as the uncertainty of the relative mean sea level found in South Africa.
catchment to contribute runoff at the applicability thereof and whether the The catchment areas range between
catchment outlet, while TP is defined rainfall magnitude-frequency relation- 28 km² and 31 283 km² and are regarded
as the time interval between the start ships will be satisfactorily accommo- as ‘gauged’, since Department of Water and
of effective rainfall and the peak dis- dated in these alternatives. Sanitation (DWS) flow-gauging stations
charge of a single-peaked hydrograph are located at the outlet of each catchment.
(McCuen et al 1984; McCuen 2005; Table 1 contains a summary of the SDF
USDA NRCS 2010; SANRAL 2013). STUDY AREA basin numbers and main geomorphological
However, this definition of TP is also South Africa is located on the southern- catchment properties, e.g. MAP, catchment
regarded as the conceptual definition of most tip of Africa and is demarcated into area (A), hydraulic length (LH), centroid
TC (McCuen et al 1984; Seybert 2006) 22 primary drainage regions (A to X) as distance (LC), average catchment slope (S)
and Gericke and Smithers (2014) also shown in Figure 1. These primary drain- and main river slope (SCH), for each catch-
showed that TC ≈ TP. age regions are further delineated into ment under consideration.
■■ Channel flow dominates the catch- 148 secondary drainage regions, i.e. A1, A2 The influences of each variable or
ment response time in medium to large to X4 (Midgley et al 1994). The 48 gauged parameter listed in Table 1 are highlighted,
catchments and is representative of the catchments in this study are located in where applicable, in the subsequent sec-
total travel time; hence, the current 23 of these secondary drainage regions tions. The DWS station numbers are also
common practice to divide the principal comprising SDF basins 1–3, 9, 17, 18 and used as catchment descriptors for easy
flow path into segments of overland 23–26, which are located in four distinctive reference in all the subsequent tables
flow and channel flow to estimate the climatological regions of South Africa, i.e. and figures.
total travel time was not applied. the Northern Region (NR), Central Region
■■ Practitioners tend to use only well- (CR), Southern Winter Coastal Region
known and simplified DDF relationships (SWCR) and Eastern Summer Coastal METHODOLOGY
to estimate design rainfall depths, Region (ESCR) (Gericke and Smithers This section provides the detailed meth-
irrespective of whether numerical or 2016b; 2017). The four climatological odology applied in each of the 48 gauged
graphical methods are used. This is regions are representative of the broad catchments. The following procedures
probably due to the probabilistic rainfall variations in climate (e.g. mean annual were performed: (i) at-site probabilistic
Table 1 M
ain geomorphological properties of the 48 catchments in the four climatological flood frequency analyses, (ii) estimation of
regions (after Gericke & Smithers 2016b; 2017) catchment response time, (iii) estimation
Catchment characteristics of design rainfall, and (iv) estimation of
Catchment deterministic design floods.
Region SDF MAP A LH LC S SCH
descriptor
basin (mm) (km²) (km) (km) (%) (%)
At-site probabilistic flood
A2H012 1 690 2 555 57.4 22.1 5.30 0.69
frequency analyses
A2H013 1 672 1 161 64.2 37.2 7.03 0.52 Seventy-nine gauged catchments were
A2H019 1 670 6 120 132.2 72.8 5.78 0.36 initially considered for possible inclusion in
A2H021 1 611 7 483 215.5 69.9 2.85 0.19 this study, but only 48 catchments met the
Northern
A5H004 2 623 636 68.4 37.4 8.73 0.71 following screening criteria: (i) streamflow
A6H006 2 633 180 25.3 9.4 6.32 1.10 record lengths (N) ≥ 25 years, and (ii) the
use of standard DWS discharge rating
A9H001 3 830 914 82.1 44.2 10.17 0.50
tables within the maximum rated flood
A9H002 3 1 128 103 37.7 19.0 17.47 2.01
level (H). However, in some cases where
A9H003 3 967 61 16.3 10.7 15.87 1.16
the observed flood levels exceeded H, the
C5H007 9 495 346 40.8 17.4 1.75 0.34 extrapolation of the rating curves up to
C5H014 9 433 31 283 326.2 207.2 2.13 0.10 or beyond bankfull flow conditions were
C5H015 9 519 5 939 160.5 81.0 2.77 0.14 considered. The high flow extensions above
C5H022 9 654 39 8.0 2.7 10.29 1.70 bankfull flow conditions were only con-
C5H023 9 648 185 29.2 17.4 7.09 0.58 sidered in cases where the existing DWS
Central
discharge rating table included floodplain

C5H039 9 516 6 331 187.1 102.7 2.65 0.13
flow on the full width of the floodplain.
C5R001 9 488 922 86.4 53.2 3.05 0.23
In essence, the individual stage extrapola-
C5R002 9 420 10 260 201.7 125.1 4.37 0.13
tions (HE), whether for bankfull or above
C5R003 9 549 937 53.8 31.1 5.04 0.27 bankfull flow conditions, were limited to a
C5R004 9 518 6 331 186.7 106.4 4.19 0.13 maximum of 30%, i.e. HE ≤ 1.3 H. Only 40
C5R005 9 660 116 16.2 7.9 5.50 0.90 events (1.5%) of the total of 2 665 annual
G1H007 17 899 724 55.5 29.0 26.21 0.46 maximum series (AMS) events analysed
G1H008 17 558 394 25.8 5.8 18.89 1.61
were subjected to such HE extrapolations,
i.e. 17 events with HE ≤ 1.1 H, 7 events
Southern Winter Coastal
G2H008 17 1 345 22 6.2 2.6 51.76 5.53

with 1.1 H < HE ≤ 1.15 H, 8 events with
G4H005 18 1 065 146 29.6 14.4 20.71 1.58
1.15 H < HE ≤ 1.20 H, and 8 events with
H1H018 18 666 109 22.8 9.3 41.61 3.20 1.20 H < HE ≤ 1.30 H.
H2H003 18 267 743 62.0 19.7 37.06 1.54 At-site probabilistic flood frequency
H4H006 18 450 2 878 109.9 26.9 29.21 0.47 analysis of the AMS was conducted at the
H6H003 18 859 500 38.6 13.6 25.56 0.97 48 flow-gauging stations to summarise the
H7H003 18 526 458 47.9 23.4 23.13 0.94 observed flood peaks, estimate parameters
H7H004 18 566 28 15.7 7.5 31.28 4.54
and select appropriate theoretical probabil-
ity distributions. The observed flood peaks
T1H004 23 897 4 923 204.5 99.1 13.39 0.50
were summarised by ranking the AMS in
T3H005 23 877 2 565 160.2 86.7 21.42 0.45
a descending order of magnitude, and the
T3H006 23 853 4 282 197.0 112.9 16.76 0.34 Cunnane plotting position (Equation 1;
T4H001 24 881 723 68.0 31.8 16.59 0.95 SANRAL 2013) was used to assign AEP or
T5H001 25 960 3 639 199.6 85.3 17.75 0.61 T values to the plotted values.
T5H004 25 1 060 537 67.4 23.9 22.66 0.77
N + 0.20
Eastern Summer Coastal
U2H005 25 979 2 523 175.0 69.8 12.71 0.68 T= (1)

m – 0.40
U2H006 25 1 130 338 49.0 22.8 12.77 0.67
U2H011 25 1 013 176 35.5 18.0 14.60 1.28
Where:
U2H012 25 953 431 57.3 24.6 11.15 0.68
T = return period (years)
U2H013 25 985 296 50.6 29.0 14.91 1.78 m = number, in descending order, of the
V1H009 26 813 195 28.1 15.3 8.71 0.58 ranked AMS events
V2H001 26 901 1 951 188.5 87.2 12.47 0.40 N = record length (years)
V2H002 26 993 945 104.8 48.0 12.80 0.41
V3H005 26 895 677 86.2 50.3 11.75 0.25 The Method of Moments (MM) and Linear
Moments (LM) were used to estimate
V3H007 26 898 129 24.9 16.9 15.73 0.93
parameters to ultimately enable the fitting
V5H002 26 841 28 893 505.0 287.2 13.52 0.27
of theoretical probability distributions to
V6H002 26 839 12 854 312.3 118.5 14.09 0.24
the AMS values. Statistical properties (e.g.
Table 2 Regional calibration coefficients applicable to Equation 3 (Gericke & Smithers 2016b) RLMA-SAWS database) within the
catchment boundary. Both the MAP
Regional calibration coefficients (* 10 –2)
Region and average design point rainfall
x1 x2 x3 x4 x5 depths for storm durations of 1 to 7
Northern 100.280 99.993 99.865 101.612 91.344 days were estimated.
■■ Step 2: A single rainfall station located
Central 100.313 99.984 106.106 98.608 98.081
approximately at the geographical
Southern Winter Coastal 100.174 99.931 101.805 104.310 99.648 centre of each catchment, and which
is representative of the average meteo
Eastern Summer Coastal 100.297 99.991 99.594 101.177 97.529
rological conditions as estimated in
Step 1, was then selected from those
mean, standard deviation, skewness and PT1 = 1.13(0.41 + 0.64ln T) rainfall stations used in Step 1 as the
coefficient of variation) of each AMS (normal (–0.11 + 0.27ln(60TC1)) base station to estimate the RLMA&SI
and log10-transformed) and visual inspection (0.79M0.69R0.20)(4) gridded design point rainfall values.
of the plotted values were used to select the ■■ Step 3: With the single rainfall station
most suitable theoretical probability distribu- Where: as selected in Step 2, the appropriate
tion in each catchment. The Log-Normal PT1 = design point rainfall depth (mm) critical storm durations (e.g. 5 minutes
(LN) distribution was only considered where M = 2-year mean of the annual daily to 7 days), return periods (e.g. 2 to
the logarithms of the AMS have a near sym- maxima rainfall (mm) 200 years) and block size (e.g. spatial
metrical distribution or where the skewness R = average number of days per year resolution of 1΄ × 1΄ grid points), were
coefficients were close to zero. In all other on which thunder was heard (days/ selected. The block size was specified
asymmetrical data sets, the Log-Pearson year) in such a way that the whole extent of
Type 3 (LP3), General Extreme Value (GEV) T = return period (years) each catchment under consideration
and/or General Logistic (GLO) distributions TC1 = time of concentration estimated is covered with grid points. The latter
were considered. using Equation 2 (hours) block of grid points was then extracted
using the Clip tool available from
Estimation of catchment Equation 4 is only applicable to TC values the Extract toolset contained in the
response time less than six hours. For TC values exceed- Analysis Tools toolbox to include only
The catchment response time was esti- ing six hours and less than 24 hours, the grid points within the boundary of
mated using both the USBR (1973) equa- linear interpolation was applied between each catchment.
tion (Equation 2), which is currently widely Equation 4 and the one-day design rainfall ■■ Step 4: Lastly, the gridded point values
used in South Africa, and the new regional depths from TR102. In cases where TC for the catchment-specific critical storm
G&S equation (Equation 3) derived by exceeded 24 hours, linear interpolation durations (TC1 and TC2) and return
Gericke and Smithers (2016b; 2017). between the n-day design rainfall depth periods under consideration were con-
TC1 = ⎛⎜
0.87LH ⎛
values was used (SANRAL 2013). verted to an average catchment value
⎜
2 0.385
⎝ 10SCH ⎝
The second set of design point rainfall using the arithmetic mean and linear
(2)
depths and intensities was based on the interpolation, respectively.
RLMA&SI approach (PT2) and associated Areal reduction factors (ARFs) were esti-
TC2 = x1 MAPx2 Ax3LC x4 LH x5S(3) critical storm durations (TC2) estimated mated using Equation 5 (Alexander 2001;
using Equation 3. The RLMA&SI approach SANRAL 2013) in order to convert the
Where: is automated and is included in the soft- average design point rainfall depths or
TC1, 2 = time of concentration (hours) ware program, Design Rainfall Estimation intensities to average areal design rainfall
A = catchment area (km²) in South Africa (Smithers & Schulze 2003; depths or intensities.
LC = centroid distance (km) 2004), which facilitates the estimation of
LH = hydraulic length (km) design rainfall depths at a spatial resolu- ARF = [90 000 – 12 800ln A
MAP = mean annual precipitation (mm) tion of 1-arc minute, for any location in + 9 830ln(60TC1,2)]0.4 (5)
S = average catchment slope (%) South Africa, for durations ranging from
SCH = average main river slope (%) 5 minutes to 7 days, and for return periods Where:
x1 to x5 = regional calibration coefficients of 2 to 200 years. The RLMA&SI gridded ARF = areal reduction factor (%)
as listed in Table 2 design point and average catchment design A = catchment area (km²)
point rainfall values were estimated by TC1, 2 = time of concentration estimated
Estimation of design rainfall making use of the following steps in the using either Equations 2 or 3
The design rainfall information was esti- ArcGIS™ environment: (hours)
mated using two different DDF approaches. ■■ Step 1: The average catchment design
The first set of design point rainfall depths point rainfall representative of the Estimation of deterministic
and intensities was based on the modified average meteorological conditions design floods
Hershfield equation (Equation 4; Alexander in each catchment was estimated The time parameter and design rainfall
2001) and/or TR102 design rainfall infor- by applying the Thiessen polygon results based on the combined use of
mation with the associated critical storm method (Wilson 1990) to all the daily Equations 2, 4 and 5, and Equations 3, 5
durations (TC1) estimated using Equation 2. design rainfall stations (from the and the RLMA&SI approach, served as
Table 3 At-site probabilistic design flood estimation results
Catchment N HE ratio Probability Design peak discharge (QP , m3/s)

Region
descriptor (years) (n events) distribution Q2 Q5 Q10 Q20 Q50 Q100 Q200
A2H012 95 – GEV/MM 126 247 346 458 625 770 933
A2H013 73 – GEV/MM 30 110 214 326 519 706 933
A2H019 66 1.09 (1) LP3/MM 58 210 405 689 1 240 1 825 2 588
A2H021 62 – LP3/MM 41 117 187 265 378 469 563
Northern
A5H004 61 1.08 (2) LP3/MM 23 91 172 258 399 528 678

A6H006 64 – GEV/MM 12 28 38 49 63 73 84
A9H001 51 – LP3/MM 18 64 127 229 450 713 1 094
A9H002 38 – LP3/MM 19 60 107 169 278 384 512
A9H003 59 – LP3/MM 19 60 97 139 197 243 289
C5H007 58 1.09 (4) GLO/LM 19 49 84 138 257 407 642
C5H014 37 1.06 (3) GEV/MM 104 1 637 4 146 5 432 7 146 8 468 9 818
C5H015 34 – LP3/MM 289 756 1 088 1 380 1 702 1 898 2 060
C5H022 36 1.06 (1) LP3/MM 9 24 39 56 81 102 125
C5H023 25 – GLO/LM 19 41 63 92 146 205 286
Central
C5H039 45 1.23 (3) LP3/MM 66 166 280 438 744 1 072 1 513
C5R001 82 – LP3/MM 31 94 169 276 482 701 992
C5R002 95 – GEV/MM 218 686 1 201 1 704 2 506 3 242 4 115
C5R003 89 – GEV/MM 75 225 440 656 812 974 1 146
C5R004 60 – GLO/LM 290 647 937 1 256 1 808 2 368 3 078
C5R005 27 – LP3/MM 35 86 139 198 287 368 461
G1H007 25 1.13 (3) GEV/MM 336 465 528 575 623 650 672
G1H008 63 1.03 (1) GLO/LM 107 187 249 321 435 541 669
G2H008 49 – GEV/MM 27 32 35 36 38 39 40
G4H005 58 1.13 (4) GEV/MM 50 73 89 104 124 139 155
H1H018 48 – GEV/MM 349 494 576 646 725 777 823
H2H003 37 – GEV/MM 90 158 202 243 295 334 371
H4H006 37 1.07 (5) GEV/MM 525 834 1 031 1 215 1 444 1 611 1 772
H6H003 34 1.18 (5) GEV/MM 123 202 255 306 373 423 474
H7H003 27 – LP3/MM 72 146 208 277 380 467 562
H7H004 62 – GEV/MM 17 37 51 65 85 101 117
T1H004 26 – GEV/MM 329 658 893 1 132 1 463 1 728 2 008
T3H005 65 – GLO/LM 267 523 740 999 1 440 1 872 2 417
T3H006 61 1.18 (3) GEV/MM 392 682 869 1 044 1 265 1 427 1 584
T4H001 66 – LP3/MM 89 219 358 540 867 1 195 1 609
T5H001 48 – GLO/LM 497 863 1 232 1 734 2 715 3 814 5 369
T5H004 68 – GLO/LM 83 134 183 247 365 491 661
U2H005 62 – GEV/MM 102 199 269 343 446 530 620

U2H006 63 – LP3/MM 22 51 89 149 287 461 735
U2H011 58 – LP3/MM 46 107 168 246 383 517 683
U2H012 56 1.26 (4) GEV/MM 47 115 178 253 374 485 616
U2H013 57 – GEV/MM 39 85 121 162 225 281 345
V1H009 63 – LP3/MM 132 242 303 351 399 426 447
V2H001 46 – GLO/LM 120 258 384 544 832 1 131 1 527
V2H002 67 – LP3/MM 61 121 190 289 491 723 1055
V3H005 47 1.29 (1) GEV/MM 67 114 142 169 201 224 247
V3H007 69 – LP3/MM 34 69 97 125 165 197 230
V5H002 56 – GEV/MM 1 451 2 431 3 071 3 678 4 454 5 028 5 594
V6H002 90 – GEV/MM 874 1 311 1 585 1 838 2 151 2 375 2 589
1 000
Probabilistic peak discharge (m3/s)
100
AMS (A2H012) GEV/MM (A2H012)

10
AMS (A2H019) LP3/MM (A2H019)
AMS (A2H021) LP3/MM (A2H021)
AMS (A5H004) LP3/MM (A5H004)
AMS (A9H001) LP3/MM (A9H001)
AMS (A9H002) LP3/MM (A9H002)
AMS (A9H003) LP3/MM (A9H003)
1
1 10 100
Return period (years)
Figure 2 P
robabilistic plots (1 ≤ T ≤ 1 000-year) based on the ranked AMS and Cunnane plotting position (Equation 1) at a catchment level in the
Northern Region
input to the SDF method (Equation 6; RESULTS AND DISCUSSION a catchment level in the four climatological
Alexander 2002) to ultimately estimate the The results from the application of the regions are shown in Figures 2 to 5(b).
deterministic design floods. above methodology in the 48 catchments It is evident from Figures 2 to 5(b) that
+ ⎛⎜ T ⎛⎜⎛⎜ 100 – 2 ⎛⎜
are presented in this section. the dispersion about the mean (standard
C2 Y C C
100 ⎝ 2.33 ⎝⎝ 100 100 ⎝
deviation) is relatively high in most of the
QT1, 2 = 0 .278
At-site probabilistic flood catchments, while, due to the asymmetrical
IT1,2 A(6)
frequency analyses nature of each AMS, the lower tails of the
The at-site probabilistic design flood probability distribution curves proved to be
Where: estimation results (QP) are presented in generally longer than the upper tails. At a
QT1 = design peak discharge (m3/s) Table 3. regional level, the probabilistic curve fitting
estimated using the standard SDF The average AMS record length of all was dominated by LP3/MM distribution in
method the catchments listed in Table 3 is 56 years, the NR (Figure 2) and CR (Figure 3), while
QT2 = design peak discharge (m3/s) while only 40 events (1.5%) of the 2 665 the GEV/MM distribution was the most
estimated using the new SDF AMS events were being subjected to the appropriate in both the SWCR (Figure 4)
procedure HE extrapolations, as discussed in the and ESCR (Figures 5(a) and (b)). The use
A = catchment area (km²) Methodology section above. The statistical of the GLO/LM distribution was limited
C2 = 2-year return period runoff properties of each AMS dataset also con- to the CR (three catchments), ESCR (four
coefficient firmed the asymmetrical nature thereof, i.e. catchments), and SWCR (one catchment).
C100 = 100-year return period runoff a high degree of variability and skewness. Overall, the above selection and use of
coefficient Consequently, the GEV/MM and LP3/MM theoretical probability distributions also
IT1 = average areal design rainfall probability distributions were regarded as proved to be in agreement with the general
intensity (mm/h) estimated using the most suitable distributions in 46% and recommendations for at-site probabilistic
Equations 2, 4 and Equation 5 38% of all the catchments, respectively. The flood frequency analyses in South Africa.
IT2 = average areal design rainfall GLO/LM probability distribution proved For example, Alexander (2001) recommends
intensity (mm/h) estimated using to be the most appropriate distribution in only the LP3 distribution, Görgens (2007)
Equations 3, 5 and the RLMA&SI only 16% of all the catchments. The proba- recommends both the LP3 and GEV distribu-
approach bilistic plots based on the ranked AMS and tions, while Van der Spuy and Rademeyer
Y T = return period factor Cunnane plotting position (Equation 1) at (2016) extend their recommendation by
10 000
1 000
100
AMS (C5H007) GLO/LM (C5H007)

AMS (C5H014) GEV/MM (C5H014)
AMS (C5H015) LP3/MM (C5H015)
AMS (C5H022) LP3/MM (C5H022)
AMS (C5H023) GLO/LM (C5H023)
10 AMS (C5H039) LP3/MM (C5H039)
AMS (C5R001) LP3/MM (C5R001)
AMS (C5R002) GEV/MM (C5R002)
AMS (C5R003) GEV/MM (C5R003)
AMS (C5R004) GLO/LM (C5R004)
AMS (C5R005) LP3/MM (C5R005)
1
1 10 100
Figure 3 P
Central Region
1 000
100
AMS (G1H007) GEV/MM (G1H007)

AMS (G1H008) GLO/LM (G1H008)
10 AMS (G2H008) GEV/MM (G2H008)
AMS (G4H005) GEV/MM (G4H005)
AMS (H1H018) GEV/MM (H1H018)
AMS (H7H003) LP3/MM (H7H003)
1
1 10 100
Figure 4 P
SWC Region
10 000
1 000
100
AMS (T1H004) GEV/MM (T1H004)

AMS (T3H005) GLO/LM (T3H005)
10 AMS (T4H001) LP3/MM (T4H001)
AMS (U2H005) GEV/MM (U2H005)
AMS (U2H006) LP3/MM (U2H006)
AMS (U2H011) LP3/MM (U2H011)
1
1 10 100
Figure 5(a) P
ESC Region
10 000
1 000
100

10
AMS (T4H001) LP3/MM (T4H001)
AMS (U2H005) GEV/MM (U2H005)
AMS (U2H006) LP3/MM (U2H006)
AMS (U2H011) LP3/MM (U2H011)
1
1 10 100
Figure 5(b) P
ESC Region
including the LN distribution as well. The
80 GLO/LM probability distribution is used
TC1 = 0.76TC2 – 1.31
extensively internationally as a standard
r 2 = 0.68
procedure for flood frequency analysis, while
70 LM parameter estimators could be used for
the screening of discordant data and test-
ing clusters for homogeneity (Smithers &
Time of concentration (TC1, Equation 2 , hours)
60
Schulze 2000a). However, Alexander (2001)
cautioned that LM parameter estimators are
too robust against outliers, and emphasised
50
that both low and high outliers are important
characteristics of the flood peak maxima.
40 The suppression of the effect of outliers could
result in unrealistic estimates of higher return
period values.
30
Estimation of catchment
response time
20 A scatter plot of the catchment response
times estimated using Equation 2 (USBR)
and Equation 3 (G&S) is shown in Figure 6.
10 As shown in Figure 6, the r² value of 0.68
confirms the moderate degree of association
between the catchment response times esti-
0
0 10 20 30 40 50 60 70 80 mated using Equations 2 and 3, respectively.
Time of concentration (TC2, Equation 3 , hours) The USBR method’s (Equation 2) slope
Time of concentration 1:1 Line Linear trendline (0.76), less than unity and negative y-inter-
cept (-1.31), highlight that this method has
Figure 6 Scatter plot of the TC1 (Equation 2) and TC2 (Equation 3) values at a catchment level an overall tendency to underestimate the TC
values in comparison to the G&S method
(Equation 3). On average, Equation 2 under-
350 estimated the time of concentration with
1:2-year (r 2 =0.34)
46% in 38 catchments when compared to
1:5-year (r 2 =0.34)
1:10-year (r 2 =0.33)
Equation 3, while an average overestima-
300 1:20-year (r 2 =0.31) tion of 36% is evident in the 10 remaining
1:50-year (r 2 =0.29) catchments. Such average differences in the
Hershfield/TR102 design rainfall (PT1, Equation 4, mm)
1:100-year (r 2 =0.27) catchment response time must be clearly

1:200-year (r 2 =0.25) understood in the context of the actual
250
response time associated with the size of a
particular catchment, as the impact thereof
might be critical in a small catchment, while
200 being less significant in a larger catchment.
However, irrespective of the catchment
size and/or differences in response time,
150
these estimated TC values will have a direct
impact on both the estimates of design
rainfall and peak discharge, all of which are
elaborated on in the subsequent sections.
100
Estimation of design rainfall

A scatter plot of the design rainfall depths
50 (PT1 and PT2) associated with the TC1 and
TC2 values in each catchment are shown in
Figure 7.
It is evident from Figure 7 that the
0
0 50 100 150 200 250 300 350 degree of association between the two DDF
RLMA&SI design rainfall (PT2, mm) approaches decreases with an increasing
return period, with the r² values rang-
Figure 7 S catter plot of the modified Hershfield/TR102 and RLMA&SI design rainfall depths at a ing between 0.34 (T = 2-year) and 0.25
catchment level (T = 200-year). The modified Hershfield/
Table 4 Deterministic design flood estimation results
Catchment Standard SDF method (Equation 6; QT1, m3/s) New SDF procedure (Equation 6; QT2, m3/s)
Region
descriptor Q2 Q5 Q10 Q20 Q50 Q100 Q200 Q2 Q5 Q10 Q20 Q50 Q100 Q200
A2H012 218 728 1 193 1 719 2 522 3 214 3 962 271 774 1 183 1630 2 305 2 892 3 543
A2H013 94 307 499 717 1 053 1 346 1 666 144 408 615 837 1 158 1 427 1 716
A2H019 294 875 1 379 1 949 2 874 3 729 4 691 375 1 065 1 615 2 209 3 088 3 836 4 652
A2H021 245 752 1 201 1 721 2 561 3 336 4 241 167 475 721 987 1 384 1 724 2 095
Northern
A5H004 28 125 214 315 473 611 761 49 185 294 412 587 735 894
A6H006 20 93 163 243 364 466 575 18 70 111 156 222 277 338
A9H001 37 205 360 542 831 1 089 1 383 57 284 466 661 943 1 174 1 420
A9H002 10 61 110 167 252 325 402 18 88 146 210 307 391 482
A9H003 9 52 93 141 213 275 340 11 57 93 132 188 234 283
C5H007 45 150 245 354 519 662 812 67 188 282 381 523 640 762
C5H014 674 2 072 3 304 4 769 7 114 9 278 11 746 828 2 352 3 536 4 781 6 562 8 038 9 575
C5H015 242 726 1 154 1 646 2 435 3 172 4 001 396 1 115 1 672 2 261 3 109 3 815 4 556
C5H022 24 85 141 205 300 380 465 11 31 46 63 88 109 131
C5H023 36 126 209 304 445 563 689 27 77 115 156 217 267 321
Central
C5H039 234 704 1 125 1 605 2 383 3 105 3 935 215 605 907 1 226 1 686 2 069 2 470
C5R001 66 199 313 448 656 851 1 054 68 192 290 395 552 685 830
C5R002 351 1 059 1 692 2 418 3 591 4 679 5 931 324 922 1 398 1 907 2 658 3 290 3 975
C5R003 90 291 471 678 993 1 273 1 567 111 312 469 637 881 1 088 1 307
C5R004 236 710 1 135 1 620 2 404 3 132 3 969 221 621 937 1 275 1 775 2 200 2 661
C5R005 38 132 220 320 468 593 726 27 77 116 159 223 278 339
G1H007 149 298 421 555 752 912 1 086 135 242 321 402 514 604 700
G1H008 136 312 464 631 872 1 069 1 275 172 301 392 481 600 691 782
G2H008 28 64 95 129 178 218 261 18 34 45 57 74 88 103

G4H005 59 136 202 274 379 465 554 38 70 95 120 155 183 214
H1H018 64 147 219 297 411 504 601 96 171 226 281 358 420 485
H2H003 188 410 600 806 1 105 1 352 1 611 154 277 367 457 582 681 784
H4H006 498 916 1 252 1 607 2 130 2 580 3 066 369 657 866 1 079 1 373 1 609 1 853
H6H003 142 321 475 643 886 1 085 1 294 160 289 386 485 623 735 852
H7H003 121 264 386 518 711 869 1 036 89 178 251 331 455 564 686
H7H004 24 55 83 112 155 190 227 15 30 42 55 76 94 114
T1H004 227 1 069 1 745 2 486 3 644 4 647 5 765 138 673 1 108 1 583 2 299 2 922 3 612
T3H005 137 644 1 051 1 500 2 202 2 807 3 487 100 477 780 1 107 1 598 2 020 2 478
T3H006 186 876 1 429 2 032 2 975 3 794 4 702 147 703 1 152 1 641 2 379 3 021 3 724
T4H001 133 555 949 1 409 2 121 2 750 3 450 92 334 539 772 1 145 1 491 1 891
T5H001 168 768 1 238 1 739 2 503 3 150 3 884 122 585 970 1 405 2 094 2 721 3 443
T5H004 53 291 507 752 1 122 1 436 1 776 27 127 207 293 422 531 649
U2H005 132 599 963 1 348 1 937 2 433 2 997 71 353 594 868 1 304 1 707 2 172
U2H006 40 226 399 596 895 1 148 1 421 20 97 162 238 359 473 605
U2H011 33 193 348 526 795 1 024 1 268 18 88 146 212 318 414 526
U2H012 46 255 447 664 994 1 274 1 576 31 156 266 393 601 800 1 033
U2H013 47 278 498 752 1 136 1 462 1 811 21 103 169 240 347 440 541
V1H009 41 126 204 292 422 531 647 45 110 161 215 294 360 433
V2H001 145 367 553 760 1 086 1 381 1 712 99 245 362 488 681 848 1 033
V2H002 95 249 380 526 755 960 1 187 72 178 264 358 503 629 771
V3H005 69 178 271 375 537 683 846 66 160 231 306 413 501 593
V3H007 34 106 171 245 354 446 542 29 70 102 135 183 222 264
V5H002 927 2 379 3 553 4 899 6 908 8 715 10 749 904 2 248 3 295 4 406 6 019 7 376 8 839
V6H002 560 1 429 2 150 2 960 4 194 5 306 6 563 477 1 176 1 714 2 278 3 080 3 741 4 443
TR102 design rainfall depths were generally estimated 24-hour event values estimated estimated using Equation 7 are shown in
lower than the RLMA&SI design rainfall using Equation 4 (modified Hershfield Figures 8(a) to 8(e).
QT-ratio = ⎛⎜
QT1,2 ⎛
depths for T ≤ 20-year. The frequency of equation) and the TR102 1-day design
⎜ – 1
⎝ QP ⎝
these underestimations decreased with rainfall information, on which the equation
(7)
an increase in return period (e.g. 85% of is based, were also evident. According to
the events at T = 2-year versus 50% of the Smithers and Schulze (2003), the latter
events at T = 20-year), while the magni- inconsistencies are ascribed to the fact that Where:
tude thereof remained relatively constant the functional relationship of Equation 4 QT-ratio = peak discharge ratio (positive =
and varied between –21% and –25%. For does not accommodate the curvilinear overestimation and negative =
T > 20-year, the opposite trend is evident. relationship between design rainfall depth underestimation)
The modified Hershfield/TR102 design and log-transformed duration as applicable QT1 = design peak discharge (m3/s)
rainfall depths were generally higher than to most rainfall stations. estimated using the standard
the RLMA&SI design rainfall depths, while The above differences in design rain- SDF method
the frequency of these overestimations fall depths using the two different DDF QT2 = design peak discharge (m3/s)
increased with an increase in return period approaches are truly appreciated when estimated using the new SDF
(e.g. 50% of the events at T = 20-year versus converted into design rainfall intensities, procedure
61% of the events at T = 200-year). The i.e. underestimated time parameters would QP = at-site probabilistic design peak
latter overestimations remained relatively result in higher design rainfall intensities, discharge (m3/s)
constant and varied between 11% and 19%. while the overestimation of time parameters
The above differences, evident between is associated with lower design rainfall A summary of the goodness-of-fit (GOF)
the two DDF approaches, are most likely intensities. Both these scenarios would have statistics based on the comparison between
attributed to: (i) the longer record lengths a direct impact on the estimation of design the at-site probabilistic design floods
and stringent data quality control proce- floods, as detailed in the next section. (QP ; Table 3) and the SDF design floods
dures used in the RLMA&SI approach, (Table 4) is listed in Table 5. The root
and (ii) the different approaches to design Estimation of deterministic mean square error (RMSE) is specifically
rainfall estimation used, i.e. a single site design floods included in Table 5 to ensure that the
approach (Adamson 1981) versus a regional The SDF design flood estimation results accumulated over- and/or underestima-
approach (Smithers & Schulze 2003; 2004). (Equation 6; QT1 and QT2) are presented tions are accounted for, i.e. to highlight the
Furthermore, inconsistencies between the in Table 4, while the peak discharge ratios actual size (not source or type) of errors
6.5
2-year 2-year
6.0 5-year 5-year
10-year 10-year
5.5 20-year 20-year
50-year 50-year
5.0
100-year 100-year
4.5 200-year 200-year
Peak discharge ratio (Equation 7)
4.0
3.5
3.0
2.5
2.0
1.5
1.0
0.5
–0.5
SDF Basin 1 SDF Basin 2 SDF Basin 3
–1.0
A2H012 A2H013 A2H019 A2H021 A5H004 A6H006 A9H001 A9H002 A9H003
Region
Figure 8(a) Peak discharge ratios (Equation 7) at a catchment level in the Northern Region; light fill = standard SDF method (QT1) and dark fill = new
SDF procedure (QT2)
3.5
2-year 2-year
5-year 5-year
10-year 10-year
3.0
20-year 20-year
50-year 50-year
100-year 100-year
2.5
200-year 200-year
2.0
1.5
1.0
0.5
SDF Basin 9
–0.5
C5H007 C5H014 C5H015 C5H022 C5H023 C5H039 C5R001 C5R002 C5R003 C5R004 C5R005
Region
Figure 8(b) Peak discharge ratios (Equation 7) at a catchment level in the Central Region; light fill = standard SDF method (QT1) and dark fill = new
SDF procedure (QT2)
5.5
2-year 2-year
5.0 5-year 5-year
10-year 10-year
4.5 20-year 20-year
50-year 50-year
100-year 100-year
4.0
200-year 200-year
3.5
3.0
2.5
2.0
1.5
1.0
0.5
–0.5
SDF Basin 17 SDF Basin 18
–1.0
G1H007 G1H008 G2H008 G4H005 H1H018 H2H003 H4H006 H6H003 H7H003 H7H004
Region
Figure 8(c) Peak discharge ratios (Equation 7) at a catchment level in the SWC Region; light fill = standard SDF method (QT1) and dark fill = new SDF
procedure (QT2)
4.0
2-year 2-year
5-year 5-year
3.5 10-year 10-year
20-year 20-year
3.0 50-year 50-year
100-year 100-year
200-year 200-year
2.5
2.0
1.5
1.0
0.5
–0.5
SDF Basin 23 SDF Basin 24 SDF Basin 25

–1.0
T1H004 T3H005 T3H006 T4H001 T5H001 T5H004 U2H005 U2H006 U2H011
Region
Figure 8(d) Peak discharge ratios (Equation 7) at a catchment level in the ESC Region; light fill = standard SDF method (QT1) and dark fill = new SDF
procedure (QT2)
4.5
2-year 2-year
5-year 5-year
4.0 10-year 10-year
20-year 20-year
3.5 50-year 50-year
100-year 100-year
200-year 200-year
3.0
2.5
2.0
1.5
1.0
0.5
–0.5
SDF Basin 25 SDF Basin 26

–1.0
U2H012 U2H013 V1H009 V2H001 V2H002 V3H005 V3H007 V5H002 V6H002
Region
Figure 8(e) Peak discharge ratios (Equation 7) at a catchment level in the ESC Region; light fill = standard SDF method (QT1) and dark fill = new SDF
procedure (QT2)
Table 5 Average QT-ratios (Equation 7) and GOF statistics at a regional level
Standard SDF method (Equation 7) New SDF procedure (Equation 7)
Region GOF
Q2 Q5 Q10 Q20 Q50 Q100 Q200 Q2 Q5 Q10 Q20 Q50 Q100 Q200
QT-ratio (+) 1.96 2.17 2.12 2.11 2.14 2.18 2.22 2.45 2.31 1.97 1.75 1.53 1.42 1.54
QT-ratio (–) –0.50 –0.13 –0.04 –0.01 –0.09 –0.15 –0.22 –0.24 –0.05 –0.04 –0.05 –0.05 –0.04 –0.04
NR
r² value 0.48 0.74 0.73 0.67 0.56 0.49 0.43 0.55 0.85 0.93 0.93 0.86 0.79 0.72
RMSE 332 1 068 1 688 2 364 3 400 4 305 5 292 390 1 134 1 658 2 166 2 821 3 307 3 775
QT-ratio (+) 1.55 1.27 1.21 1.14 1.04 0.88 0.81 1.65 1.03 0.89 0.70 0.50 0.43 0.39
QT-ratio (–) –0.17 –0.04 –0.20 –0.12 – – – –0.23 –0.07 –0.10 –0.12 –0.11 –0.10 –0.12
CR
r² value 0.21 0.92 0.91 0.93 0.94 0.95 0.96 0.22 0.95 0.94 0.94 0.93 0.93 0.92
RMSE 615 812 1 339 1 635 2 250 3 081 4 244 760 963 1 097 1 393 1 822 2 228 2 735
QT-ratio (+) 0.40 0.76 1.05 1.31 1.47 1.69 1.91 0.46 0.41 0.41 0.48 0.57 0.64 0.54
QT-ratio (–) –0.48 –0.53 –0.41 –0.29 –0.43 –0.35 –0.27 –0.39 –0.32 –0.33 –0.28 –0.21 –0.15 –0.22
SWCR
r² value 0.54 0.60 0.63 0.66 0.69 0.72 0.74 0.58 0.66 0.71 0.75 0.79 0.82 0.84
RMSE 361 517 700 949 1 376 1 773 2 218 373 470 509 537 575 619 686
QT-ratio (+) 0.37 1.08 1.23 1.35 1.43 1.47 1.50 0.11 0.37 0.39 0.46 0.52 0.62 0.81
QT-ratio (–) –0.37 –0.20 –0.33 –0.17 –0.08 –0.17 –0.28 –0.43 –0.18 –0.22 –0.20 –0.21 –0.20 –0.20
ESCR
r² value 0.95 0.94 0.92 0.90 0.84 0.76 0.63 0.94 0.98 0.97 0.95 0.90 0.82 0.71
RMSE 752 828 1750 2 965 4 991 6 908 9 180 858 443 672 1 347 2 576 3 788 5 293
produced by the two SDF procedures, with ■■ Central Region (SDF basin 9, (QT1, Equation 6) when compared to the
the objective function to minimise the Figure 8(b)): The standard SDF flood at-site probabilistic flood estimates in
RMSE to zero. peaks (QT1) exceeded the at-site all the catchments, except in catchment
The results contained in Tables 4 and 5, probabilistic flood peaks (QP) in all the G1H007 where both methods had a
as well as Figures 8(a) to 8(e), are indicative catchments, except for the lower return tendency to underestimate the at-site
of several trends associated with specific periods (T ≤ 20-year) in catchments probabilistic flood peaks with between
catchments and return periods, which are C5H014 and C5R004. In more than 80% 20% and 60%. In all other catchments
highlighted below: of all the catchments considered, the and corresponding return periods, the
■■ Northern Region (SDF basins 1–3, new SDF procedure (QT2, Equation 6) new SDF procedure (QT2, Equation 6)
Figure 8(a)): The SDF flood peaks (QT1 resulted in improved estimates in com- demonstrated the best average results
and QT2) exceeded the at-site probabi parison to the standard SDF method (41%–64% overestimation, 15%–39%
listic flood peaks (QP) in all the catch- (QT1, Equation 6) when compared to the underestimation, 0.58 ≤ r² ≤ 0.84, and
ments, except for catchments A9H002 at-site probabilistic flood estimates. The 373 ≤ RMSE ≤ 686).
and A9H003 in SDF basin 3. In 56% of all standard SDF method (QT1, Equation 6) ■■ Eastern Summer Coastal Region (SDF
the catchments considered, the new SDF overestimated the at-site probabilistic basins 23–26, Figures 8(d)–8(e)): The
procedure (QT2, Equation 6) resulted flood peaks with between 3% and 323%, new SDF procedure (QT2, Equation 6)
in improved estimates in comparison whereas the new SDF procedure’s resulted in improved estimates in com-
to the standard SDF method (QT1, overestimations are limited to 264%. parison to the standard SDF method
Equation 6) when compared to the at-site However, in catchment C5H014, both (QT1, Equation 6) when compared to
probabilistic flood estimates, especially the standard SDF method and new SDF the at-site probabilistic flood estimates
in catchments A2H021 and A6H006. In procedure overestimated the at-site in all the catchments, except for the
the latter catchments, the standard SDF probabilistic flood peaks by a factor two-year return period. The new SDF
method (QT1, Equation 6) overestimated > 5 for T = 2-year. For all other return procedure (QT2, Equation 6) demon-
the at-site probabilistic flood peaks with periods, the new SDF procedure (QT2, strated better average results (11%–81%
between 62% and 653%, whereas the new Equation 6) demonstrated the best aver- overestimation, 18%–43% underestima-
SDF procedure’s overestimations are age results (39%–165% overestimation, tion, 0.71 ≤ r² ≤ 0.98, and 443 ≤ RMSE
limited to 300%. On average, the new ≤ 12% underestimation, 0.92 ≤ r² ≤ 0.95, ≤ 5 293) than the standard SDF method,
SDF procedure (QT2, Equation 6) demon- and 963 ≤ RMSE ≤ 2 735). i.e. 37%–150% overestimation, 8%–37%
strated the best results, especially for the ■■ Southern Winter Coastal Region (SDF underestimation, 0.63 ≤ r² ≤ 0.95, and
higher return periods (e.g. T = 10–200- basins 17 and 18, Figure 8(c)): The 752 ≤ RMSE ≤ 9 180.
year; 142%–175% overestimation, ≤ 5% new SDF procedure (QT2, Equation 6) Overall, the new SDF procedure (QT2,
underestimation, 0.72 ≤ r² ≤ 0.93, and resulted in improved estimates in com- Equation 6) resulted in improved estimates
1 658 ≤ RMSE ≤ 3 775). parison to the standard SDF method in comparison to the standard SDF method
(QT1, Equation 6) when compared to the modifications, e.g. modified runoff coef- coefficients, are key input parameters for
at-site probabilistic flood estimates in more ficients (Pegram 2003) and probabilistic design flood estimation in ungauged catch-
than 80% of all the catchments in the four approaches (Alexander 2002; Calitz & ments, and have a significant impact on the
climatological regions. Such improvements Smithers 2016), were suggested locally and design of hydraulic structures. Typically,
in design flood estimation also confirm abroad (Pilgrim & Cordery 1993) to deviate high runoff coefficients, underestimated
that catchment response time and design from a deterministic to a more probabilis- time parameters and associated lower
rainfall are fundamental inputs to design tic-deterministic approach. The SDF meth- design rainfall depths, although of much
flood estimation in ungauged catchments. od is a typical example thereof, but due higher intensities, would result in over
However, despite the improvement in design to several design limitations – i.e. region- designed hydraulic structures, while low
flood estimation achieved in this study, the alisation scheme adopted, lack of testing runoff coefficients and overestimated time
high over- and/or underestimations are still for homogeneity, outdated design rainfall parameters would result in underdesigns.
regarded as unacceptable and indicative information (TR102), etc – the method Not only will hydraulic structures be over-
that neither design rainfall nor catchment generally proved to be too conservative or underdesigned, but associated socio-
response time in these catchments could be (Gericke & Du Plessis 2012). However, by economic implications might render some
regarded as the only fundamental input to using the more appropriate design rainfall projects as not being feasible, while any loss
design flood estimation. In essence, catch- information and catchment response times of life due to excessive flood damages and
ment response time should be regarded as as input to the SDF method (this study), insufficient infrastructure is not excluded.
enigmatic, since, although it is assumed the results improved accordingly. In doing It is recommended that the current well-
to be an independent time parameter, it is so, the variation of runoff coefficients with known and simplified catchment response
actually dependent on the peak discharge, return period is also incorporated. Thus, time (USBR TC equation) and design rainfall
which in turn is also dependent on the as the intensity and volume of rainfall (modified Hershfield/TR102 DDF approach)
design rainfall. Therefore, the latter over- increases, the effect of the internal storage estimation methodologies should be
and/or underestimations could also be of catchments decreases, which leads to an replaced with the empirical G&S TC equa-
ascribed to the regional SDF runoff coef- increase in the runoff coefficients. tions and the RLMA&SI DDF approach
ficients not being representative of the aver- Typically, the large proportional dif- when deterministic design floods are
age catchment conditions and/or physical ferences between the C2 and C100 runoff estimated in ungauged catchments in South
regional descriptors. coefficients (Equation 6), highlight that the Africa. However, since the G&S TC equa-
The term ‘runoff coefficient’ is com- SDF method assumes that a larger proportions (Equation 3) are limited to only four
monly used in flood hydrology (Young tion of rainfall would contribute to the climatological regions in South Africa, the
et al 2009; SANRAL 2013; Van der Spuy & flood peaks and acknowledge that the ante further refinement thereof in terms of cali-
Rademeyer 2016) to represent the percent- cedent soil moisture status of a catchment bration, verification and possible regionali
age of effective rainfall that is transformed introduces additional variability into the sation in other regions, is acknowledged.
to direct runoff. Runoff coefficients vary rainfall-runoff process. However, variability The proposed new SDF procedures are
substantially with the time scale of aggre- increases with an increase in catchment recommended for the estimation of flood
gation, i.e. in small catchments (< 15 km²) size; hence, the difficulty to successfully peaks with return periods in excess of and
runoff coefficients represent an overall cut- establish a relationship between regional/ including 10 years (T ≥10 years). In order to
off threshold separating effective rainfall catchment descriptors and runoff coef- improve the depth of hydrological runoff
from total rainfall and are readily obtain- ficients in larger catchments. Hydrological data in South Africa, it is recommended that
able from lookup tables, whereas in larger literature, e.g. Pilgrim and Cordery (1993), flow records be obtained from the DWS
rural catchments, the runoff coefficients Parak and Pegram (2006), and Gericke and and verified. The verified data can then be
are normally associated with land use, Du Plessis (2012) also confirmed the latter utilised to improve on the findings in this
soils and catchment slopes (Efstratiadis et and concluded that runoff coefficients are study and to refine methods to be used for
al 2014). In both cases, these runoff coef- essentially functions of the return period the purpose of design flood estimations,
ficients are regarded as constant; however, and catchment response time. including those for T < 10 years.
it is obvious that its value depends both on Furthermore, the current research ini-
the antecedent soil moisture conditions tiative of Calitz and Smithers (2016), which
and on the rainfall intensity. To overcome CONCLUSIONS focuses on the development and assess-
this shortcoming, larger runoff coefficients The overall objective of this study was to ment of regional runoff coefficients to be
are normally assigned to higher return independently test and compare the latest incorporated in the Probabilistic Rational
periods, i.e. runoff coefficients increase catchment response time and design rain- Method (PRM) for South Africa, should be
as the return period increases, but such fall estimation methodologies with current supported and welcomed by all research
recommendations are not based on sys- well-known and simplified methodolo- academic institutions and engineering
tematic investigations and favour arbitrary gies used in South Africa to ultimately practitioners involved in modern flood
choices (Efstratiadis et al 2014). highlight the impact thereof on design hydrology practice.
Despite the simplicity of estimating flood estimation.
runoff coefficients, it is evident that runoff Building upon the critical assessment of
coefficients play a secondary role in the available definitions, estimation procedures ACKNOWLEDGEMENTS
overall predictive capacity of most deter- and the results from this study, it is evident Support for this research by the National
ministic design flood estimation methods that catchment response time, design Research Foundation (NRF), University
in ungauged catchments. Hence, several rainfall, and to some lesser extent runoff of KwaZulu-Natal (UKZN) and Central
University of Technology, Free State (CUT) used in current flood hydrology practice in flood frequency curves from rainfall. Journal of
is gratefully acknowledged. I also wish to South Africa? Journal of the South African Hydrology, 256: 196–210.
thank the anonymous reviewers of this Institution of Civil Engineering, 58(1): 2–15. SANRAL (South African National Roads Agency
paper for their constructive review com- DOI: 10.17159/2309-8775/2016/v58n1a1. Limited) 2013. Drainage Manual, 6th ed. Pretoria:
ments, which have helped to significantly Gericke, O J & Smithers, J C 2016b. Derivation and SANRAL.
improve the paper. verification of empirical catchment response time Schmidt, E J & Schulze, R E 1984. Improved estimation
equations for medium to large catchments in South of peak flow rates using modified SCS lag equations.
Africa. Hydrological Processes, 30(23): 4384–4404. ACRU Report No 17. Pietermaritzburg: University of
REFERENCES DOI: 10.1002/hyp.10922. Natal, Department of Agricultural Engineering.
Adamson, P T 1981. Southern African storm rainfall. Gericke, O J & Smithers, J C 2017. Direct estimation Seybert, T A 2006. Stormwater management for land
Technical Report TR102. Pretoria: Department of of catchment response time parameters in medium development: methods and calculations for quantity
Environmental Affairs and Tourism. to large catchments using observed streamflow control. Hoboken, NJ: Wiley.
Alexander, W J R 2001. Flood risk reduction measures: data. Hydrological Processes, 31(5): 1125–1143. Smithers, J C 2012. Review: Methods for design
incorporating flood hydrology for Southern Africa. DOI: 10.1002/hyp.11102. flood estimation in South Africa. Water SA, 38(4):
Pretoria: University of Pretoria, Department of Civil Grimaldi, S, Petroselli, A, Tuaro, F & Porfiri, M 2012. 633–646.
and Biosystems Engineering. Time of concentration: a paradox in modern Smithers, J C & Schulze, R E 2000a. Development
Alexander, W J R 2002. The standard design flood. hydrology. Hydrological Sciences Journal, 57(2): 217– and evsaluation of techniques for estimating short
Journal of the South African Institution of Civil 228. DOI: 10.1080/02626667.2011.644244. duration design rainfall in South Africa. WRC
Engineering, 44(1): 26–30. Görgens, A H M 2007. Joint Peak-Volume (JPV) design Report No 681/1/00. Pretoria: Water Research
Bondelid, T R, McCuen, R H & Jackson, T J 1982. flood hydrographs for South Africa. WRC Report No Commission.
Sensitivity of SCS models to curve number 1420/3/07. Pretoria: Water Research Commission. Smithers, J C & Schulze, R E 2000b. Long duration
variation. Water Resources Bulletin, 20(2): 337–349. Kerby, W S 1959. Time of concentration for overland design rainfall estimates for South Africa. WRC
Calitz J P & Smithers J C 2016. Development and flow. Civil Engineering, 29(3): 174. Report No 811/1/00. Pretoria: Water Research
assesment of a Probabilistic Rational Method for McCuen, R H 2005. Hydrologic Analysis and Design, Commission.
design flood estimation in South Africa. WRC 3rd ed. Upper Saddle River, NJ: Prentice-Hall. Smithers, J C & Schulze, R E 2003. WRC Report No
Inception Report No K5/2748. Pretoria: Water McCuen, R H, Wong, S L & Rawls, W J 1984. 1060/01/03. Design rainfall and flood estimation in
Research Commission. Estimating urban time of concentration. Journal of South Africa. Pretoria: Water Research Commission.
DWAF (Department of Water Affairs and Forestry) Hydraulic Engineering, 110(7): 887–904. Smithers, J C & Schulze, R E 2004. The estimation of
1995. GIS data: drainage regions of South Africa. Midgley, D C & Pitman, W V 1978. A Depth-Duration- design rainfall for South Africa using a regional
Pretoria: DWAF. Frequency diagram for point rainfall in Southern scale invariant approach. Water SA, 30(4): 435–444.
Efstratiadis, A, Koussis, A D & Mamassis, N 2014. Africa. HRU Report No 2/78. Johannesburg: USBR (United States Bureau of Reclamation) 1973.
Flood design recipes vs. reality: can predictions University of the Witwatersrand, Hydrological Design of Small Dams, 2nd ed. Washington, DC:
for ungauged basins be trusted? Natural Hazards Research Unit. Water Resources Technical Publications.
and Earth System Sciences, (14): 1417–1428. Midgley, D C, Pitman, W V & Middleton, B J 1994. USDA NRCS (United States Department of
DOI: 10.5194/nhess-14-1417-2014. Surface water resources of South Africa. WRC Agriculture, Natural Resources Conservation
Gericke, O J & Du Plessis, J A 2011. Evaluation of Report No 298/2/94. Pretoria: Water Research Service) 2010. Time of concentration. In:
critical storm duration rainfall estimates used in Commission Woodward, D E (Ed.) National Engineering
flood hydrology in South Africa. Water SA, 37(4): Parak, M & Pegram G G S 2006. Rational formula from Handbook. Washington, DC: USDA NRCS, Ch 15,
453–470. DOI: 10.4314/wsa.v37i4.4. Runhydrograph. Water SA, 32(2): 163–180. Section 4, Part 630, 1–18.
Gericke, O J & Du Plessis, J A 2012. Evaluation of the Pegram, G G S 2003. Rainfall, Rational Formula and Van der Spuy, D & Rademeyer, P F 2016. Flood
standard design flood method in selected basins Regional Maximum Flood – Some scaling links. frequency estimation methods as applied in the
in South Africa. Journal of the South African Australian Journal of Water Resources, 7(1): 29–39. Department of Water and Sanitation. Pretoria:
Institution of Civil Engineering, 54(2): 2–14. Pegram, G G S & Parak, M 2004. A review of the Department of Water and Sanitation.
Gericke, O J & Du Plessis, J A 2013. Development of a Regional Maximum Flood and Rational Formula Van Vuuren, S J, Van Dijk, M & Coetzee, G L 2012.
customised design flood estimation tool to estimate using geomorphological information and observed Status review and requirements of overhauling
floods in gauged and ungauged catchments. Water floods. Water SA, 30(3): 377–392. flood determination methods in South Africa. WRC
SA, 39(1): 67–94. DOI: 10.4314/wsa.v39i1.9. Pilgrim, D H & Cordery, I 1993. Flood runoff. In: Report No K8/994/1. Pretoria: Water Research
Gericke, O J & Smithers, J C 2014. Review of methods Maidment, D R (Ed.). Handbook of Hydrology. New Commission.
used to estimate catchment response time York: McGraw-Hill, Ch 9, 1-42. Wilson, E M 1990. Engineering hydrology, 4th ed.
for the purpose of peak discharge estimation. Pullen, R A 1969. Synthetic unitgraphs for South Africa. London: Macmillan.
Hydrological Sciences Journal, 59(11): 1935–1971. HRU Report No 3/69. Johannesburg: University of Young, C B, McEnroe, B M & Rome, A C 2009.
DOI: 10.1080/02626667.2013.866712. the Witwatersrand, Hydrological Research Unit. Empirical determination of rational method runoff
Gericke, O J & Smithers, J C 2016a. Are estimates Rahman, A, Weinmann, P E, Hoang, T M T & coefficients. Journal of Hydrologic Engineering, (14):
of catchment response time inconsistent as Laurenson, E M 2002. Monte Carlo simulation of 1283–1289.
DISCUSSION
Journal of the South African Publishing particulars of
the paper under discussion
ISSN 1021-2019
Vol 60 (1) 2018, Pages 44–51, Paper 1410:

Comparative analysis of methods
of pile-bearing capacity evaluation
using CPT logs from tropical soils
(http://dx.doi.org/10.17159/2309-8775/2018/v60n1a5)
I N Obeta, M E Onyia, D A Obiekwe
COMMENT Reference
The authors really did a great job and pre- Nottingham, L C 1975. Use of quasi-static friction
sented a very interesting article. However, cone penetrometer data to predict load capacity
we would like to draw attention to a few of displacement piles. PhD thesis. Gainesville,
inaccuracies. FL: University of Florida, Department of Civil
The authors analysed the obtained Engineering, p 553.
results and observed that the results
from the pile capacity, according to
Schmertmann’s method, received excep- RESPONSE FROM AUTHORS
tionally high values. Since CPT test charts, We appreciate the remarks from D Sližytė
soil descriptions, types of piles and pile and R Mackeviĉius, but would like to state
diameters are not included in the article, it the following:
is difficult to say why the authors obtained 1. Equation 1, which is presented as
such results. However, we would like to qc1 + qc2
qt = , and which can be seen
point out that in the article Equation 1, qc1 2
and qc2 should not be “cone tip resistance”, in Nottingham (1975) and is further
but as Nottingham (1975) stated, quoted by illustrated by Schmertmann (1978), is
you, should be smaller: correct.
“If the mechanical penetrometer is used 2. Schmertmann (1978) in his illustration
in clays, the computed qc value should be stated that “qc1 = average qc values over
multiplied by 0.60 to account for the pos- a distance of xB below the pile tip (path
sible increase in qc resulting from friction a-b-c). Sum qc values in both the down-
on the tip mantle. If the design is to be ward (path a-b) and upward (path b-c)
based on yield capacity criteria, multiply directions. Use actual qc values along
the computed ultimate tip resistance by path a-b and the minimum path rule
0.73.” along path b-c. Compute qc1 for x-values
Also, after the analysis of graphics it from 0.7 to 3.75 and use the minimum
was noticed that the pile capacities, accord- qc1 value obtained.”
ing to De Ruiter, are equal to 0 (in Figures 3. In respect of qc2 (which is obtained as
1–4). This should not be the case even with the average qc over a distance 8D above
very weak soil. the pile tip) the following: Although
not stated in the body of work, the
D Sližytė authors agree with the observations of
danute.slizyte@vgtu.lt D Sližytė and R Mackevičius that this
R Mackeviĉius value, when sought for in clay, should
rimantas.mackevicius@vgtu.lt be multiplied by 0.6 to account for the
Vilnius Gediminas Technical University possible increase in qc values resulting
Lithuania from friction on the tip mantle, and
Obeta IN, Onyia ME, Obiekwe DA. Comparative analysis of methods of pile-bearing capacity evaluation using CPT logs from tropical soils.
this the authors did. The authors only were carried out on clays (Tumay & FL: University of Florida, Department of Civil
did not state this categorically in the Fakhroo 1982). Engineering, p 553.
work, because we assume it is common Schmertmann, J H 1978. Guidelines for cone
knowledge, as it is also presented in I N Obeta penetration test (Performance and design). Report
many other standard texts. ifeanyi.obeta@unn.edu.ng No. FHWA-TS-78-209. Washington, DC: U.S.
4. A careful observation of Figures 1 to Dr M E Onyia Department of Transportation, p 145.
4 will show that the zero pile capacity ebieonyia64@yahoo.com Tumay, M T & Fakhroo, M 1982. Friction pile capacity
values are not those of De Ruiter, but D A Obiekwe prediction in cohesive soils using electric quasi-
on the contrary they are pile capacities damianobiekwe@gmail.com static penetration tests. Interim Research Report
attributed to the Tumay and Fakhroo University of Nigeria, Nsukka, Nigeria No. 1. Baton Rouge, LA: Louisiana Department of
method. They were assigned zero values Transportation and Development, Research and
by the authors, because Figures 1 to 4 Development Section.
show the pile capacities obtained in References
sands (refer to page 51, Figure 10 of the Nottingham, L C 1975. Use of quasi-static friction
article), and the Tumay and Fakhroo cone penetrometer data to predict load capacity
methods were generated while tests of displacement piles. PhD thesis. Gainesville,

Journal Vol 60 4 2018 December

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Journal Vol 60 4 2018 December

Uploaded by

Copyright:

Available Formats

Volume 60 Number 4

of the South African Institution of Civil Engineering December 2018

■■  tudy of the Durban Bight shoreline evolution under

Design and reproduction

© South African Institution of Civil Engineering 1

B1 shoreline model, employing 200 to 400

Annual rate (m3)

applied for the simulations. Beach

SHORELINE MODEL CALIBRATION 2 000 000

Offset of +2 m CD contour (m)

20 Measured offset data 20 Measured offset data

Beacon 13 (RMSEP = 12 m) Beacon D (RMSEP = 14 m)

Offset of +2 m CD contour (m)

20 Measured offset data 20 Measured offset data

Beacon DN8 (RMSEP = 10 m) Beacon DN12 (RMSEP = 11 m)

Offset of +2 m CD contour (m)

20 Measured offset data 20 Measured offset data

in the measured data can, to a large extent,

Beacon 12 (RMSEP = 8 m) Beacon 13 (RMSEP = 11 m)

Offset of +2 m CD contour (m)

20 Measured shoreline position 20 Measured shoreline position

Beacon 15 (RMSEP = 18 m) Beacon 18 (RMSEP = 11 m)

Offset of +2 m CD contour (m)

1.960 The first and most probable future scenario

1.948 the proximity of Durban, there might

1.96 10 (Dairy Beach Pier) present over time. A decreasing steepness in

Distance north from Vetch's Reef (m)

2.04 DN9 Accretion 2.04 DN9

E (Walter Gilbert Road) E (Walter Gilbert Road)

Scenario D (b) Scenario D (c)

Distance north from Vetch's Reef (m)

2.02 Erosion 2.02 Erosion

1.96 10 (Dairy Beach Pier)

Scenario E (c) decreasing Scenario E (a) increasing

Distance north from Vetch's Reef (m)

2.02 Erosion 2.02 Erosion

1.96 12 (West Street Jetty) 1.96 10 (Dairy Beach Pier)

Distance north from Vetch's Reef (m)

E (Walter Gilbert Road) 1.98 G

Scenario E (a) decreasing Scenario E (b) decreasing)

Distance north from Vetch's Reef (m)

2.02 Erosion 2.02 Erosion

1.96 10 (Dairy Beach Pier)

Scenario E (b) increasing

DR NICO DE KOKER, Associate Member of

De Koker N, Elvin AA. Risk-based member reliability in structural design.

xopt Marginal benefit of increasing

slope ∝ marginal benefit

Σ pfx ⎛⎜di + Σ Γij cj xjm ⎛⎜(17)

where xj = x for j ∈ u and xj = 1 otherwise.

Gbuckle ∝ I/L2 ∝ A2/L2(15)

At maximum expected benefit, x = xopt 5.3

popt = 1 + erf ⎛⎜– opt ⎛⎜ 

exp ⎛⎜– opt ⎛⎜ 

If d/c ≫ 1, the relation becomes:

if di ≫ ΣjΓijcj the coefficient of p΄opt

3 510 – 2 050 2 270 165.1 × 4.5

The bridge is designed assuming a 5 5 R 0.5k R 2.0m 4.00 × 103

member 1.2 3 0.50

4.0 1.1 0.25

Starting from the assumption that the Member-level reliability parameters

3.8 1.1 0.5 beam

government infrastructure FATE T MUZONDO holds a BSc in Agricultural

■■ tudy of the Durban Bight shoreline evolution under

Σ pfx ⎛⎜di + Σ Γij cj xjm ⎛⎜(17)

Gbuckle ∝ I/L2 ∝ A2/L2(15)

popt = 1 + erf ⎛⎜– opt ⎛⎜

exp ⎛⎜– opt ⎛⎜

U2H005 25 979 2 523 175.0 69.8 12.71 0.68 T= (1)