Professional Documents
Culture Documents
June 2023
PUBLISHER
South African Institution of Civil Engineering
Block 19, Thornhill Office Park,
Bekker Road, Vorna Valley, Midrand, South Africa
Private Bag X200, Halfway House, 1685, South Africa
Tel +27 11 805 5947, Fax +27 11 805 5971
http://www.saice.org.za, civilinfo@saice.org.za
CONTENTS
EDITOR-IN-CHIEF
Dr Peter Day
Consultant, Jones & Wagener (Pty) Ltd
Extraordinary Professor of Geotechnical Engineering
Stellenbosch University
day@jaws.co.za
CO-EDITOR-IN-CHIEF
Prof Gerhard Heymann 2 Spalling of concrete in fire –
University of Pretoria
gerhard.heymann@up.ac.za an overview and local relevance
ADMINISTRATIVE EDITOR
Verelene de Koker J E van der Merwe
Tel +27 11 805 5947, Mobile +27 83 378 3996
verelene@saice.org.za
SUB-EDITORS
Prof EP Kearsley University of Pretoria
10 Design criteria for upstream raised
Prof CP Roth
Prof A Visser
University of Pretoria
University of Pretoria
tailings storage facilities
Prof J Wium Stellenbosch University
Prof M Zuidgeest University of Cape Town J A Wates
BALANCE OF THE JOURNAL EDITORIAL PANEL
Prof Y Ballim University of the Witwatersrand
Prof W Burdzik Retired from the University of Pretoria
Prof N de Koker Stellenbosch University
17 Could a one-size-fits‑all approach apply
Prof J du Plessis Stellenbosch University
Prof GC Fanourakis University of Johannesburg
to the extension of stage‑discharge
Dr T Gaspar
Prof M Gohnert
Sheffield University
University of the Witwatersrand
relationships at flow‑gauging weirs?
Dr C Herold Umfula Wempilo Consulting
Prof A Ilemobade University of the Witwatersrand
O J Gericke, V H Williams
Prof SW Jacobsz University of Pretoria
Dr C MacRobert Stellenbosch University
Dr O Nnene University of Cape Town
Prof S Nwaubani University of the Witwatersrand
28 Advanced numerical modelling of the nonlinear
Prof W Steyn University of Pretoria
Dr J van der Merwe University of Pretoria
mechanical behaviour of a laterally loaded
Mr M van Dijk
Prof G van Zijl
University of Pretoria
Stellenbosch University
pile embedded in stiff unsaturated clay
Prof C Venter University of Pretoria
Dr E Vorster Zutari K T Braun, N Bakas, G Markou, S W Jacobsz
Prof R Walls Stellenbosch University
Prof A Zingoni University of Cape Town
PEER REVIEWING / OPEN ACCESS
The Journal of the South African Institution of Civil Engineering,
39 Comparison of sinkhole size versus
which is distributed internationally, is a peer-reviewed,
open-access journal licensed under a Creative Commons
borehole hazard classification
Attribution Licence (CC BY-NC-ND). Readers may therefore
freely use and share the content as long as they credit the
J Bunk
original creator and publisher, do not change the material
in any way, and do not use it commercially.
Creative Commons
44 Guidelines for the preparation of
Attribution Licence papers and technical notes
DESIGN AND REPRODUCTION
Marketing Support Services, Ashlea Gardens, Pretoria
The South African Institution of Civil Engineering accepts no
responsibility for any statement made or opinion expressed
in this publication. Consequently, nobody connected with
the publication of this journal, in particular the proprietor,
the publisher and the editors, will be liable for any loss or
damage sustained by any reader as a result of his or her action
upon any statement or opinion published in this journal.
© South African Institution of Civil Engineering
and the authors 1
TECHNICAL PAPER
Spalling of concrete in fire –
an overview and
JOURNAL OF THE SOUTH AFRICAN
INSTITUTION OF CIVIL ENGINEERING
local relevance
ISSN 1021-2019 (print) | ISSN 2309-8775 (online)
Vol 65 No 2, June 2023, Pages 2–9, Paper 1410
(b)
Figure 1 E xamples of recent fire-induced spalling damage during a fire damage assessment by Recent observations of fire-induced
the author following the July 2021 riots in South Africa: (a) explosive spalling of the spalling of concrete in South Africa
steel-fibre-reinforced concrete floor, (b) spalling to the soffit and drop panel of an office The July 2021 riots in South Africa resulted
building, (c) spalling to the top of an exposed slab, and (d) corner spalling of a column in extensive fire damage to a warehouse
structure in Durban. In addition to par-
spalling is considered the most violent. Due Partial collapse of reinforced concrete tial collapse of the steel superstructure,
to the early, violent and often extensive building structures as a result of fire has extensive fire-induced spalling of concrete
nature of explosive spalling, it is considered been observed in the past. In many cases, elements resulted, as shown in Figure 1.
as the greatest hazard to reinforced con- however, reinforced concrete slabs are Severe explosive spalling of the steel-
crete structures if the extent is substantial. able to maintain structural integrity after fibre-reinforced concrete floor occurred in
Journal of the South African Institution of Civil Engineering Volume 65 Number 2 June 2023 3
isolated locations (Figure 1(a)) to depths of (e.g. Majorana et al 2010). General consensus due to increased thermal stress development
up to half of the floor thickness. Similarly, is that this is due to the compact microstruc- (Kodur & Phan 2007). Experimental obser-
extensive spalling resulted to slab soffits ture that is associated with greater compres- vations have also shown that moderate com-
of the adjacent office building, exposing sive strength. A strong correlation has been pressive stress increases spalling susceptibil-
bottom reinforcement (Figure 1(b)). Such shown to exist between the compressive ity (Carré et al 2013). Moderate compressive
spalling around columns has been shown strength of concrete and its intrinsic permea- stress adds to the overall internal stress
to drastically reduce punching shear bility (Van der Merwe 2022c), suggesting that state, but has also been shown to reduce the
resistance (Lu et al 2018). Localised fires higher concrete grades can be expected to be intrinsic permeability of concrete (Van der
in reinforced concrete structures inside associated with higher pore pressure build-up Merwe 2022b), which in turn can result in
the warehouse space showed substantial during fire exposure. Related to such a com- increased pore pressure development.
spalling of the top of slabs, as well as cases pact microstructure is the permeability of
of corner spalling of columns, exposing concrete, which governs pore pressure build-
reinforcement to fire (Figures 1(c) and 1(d)). up potential (Bažant & Kaplan 1996). Since PLAUSIBLE MECHANISMS
the permeability of concrete can be reduced OF SPALLING
without increasing compressive strength, Various mechanisms for explosive
PROPERTIES INFLUENCING Hertz (2003) concluded that permeability is a spalling have been proposed, and some
SPALLING more relevant indication of spalling risk than academic debate in this regard continues
Numerous parameters influence the sus- compressive strength. to date. Shorter and Harmathy (1961) and
ceptibility of concrete to spall in fire, rang- In addition to permeability, the moisture Harmathy (1965) proposed that pore pres-
ing from material production to construc- content determines pore pressure build-up sure build-up in the concrete matrix at ele-
tion and operational conditions. Properties during fire exposure, and is therefore consid- vated temperatures is the main mechanism
can be grouped into the following four ered by many as a crucial parameter related responsible for spalling. Once moisture
categories (Van der Merwe 2019): to spalling susceptibility (e.g. Hertz 2003). evaporation starts, the resulting pressure
In addition to its influence in pore pressure gradient forces part of the moisture to the
Concrete mixture composition build-up, increased moisture content can be heated surface, whilst the direction of heat
The aggregate type is said to influence spall- expected to increase the thermal stress gradi- flux forces the remainder to the cooler inner
ing susceptibility, with siliceous aggregate ent (Fib Bulletin 38: 2007). Latent heat during region. Here condensation is said to result in
concretes often reported as being more sus- the phase change from liquid water to vapour a region of increased moisture content. As
ceptible to spalling (Fib Bulletin 38: 2007). is associated with an increased specific heat heating continues, two distinct zones result
Substantial thermal damage can result due of concrete, with a subsequent slower increase – a dry region near the heated surface, and a
to the difference in the thermal expansion in temperature away from the heated surface saturated zone (moisture clog) deeper in the
coefficient of aggregate and the surrounding and a resulting increase in thermal stress. concrete section (Figure 2). The low perme-
cement paste. Quartz in siliceous aggregates Similarly, thermal properties that affect ther- ability in this moisture clog results in gradu-
is generally associated with a phase change mal stress development are also associated ally increasing pore pressure and finally in
at high temperatures that results in a rapid with spalling risk, including the specific heat spalling when the high-temperature tensile
increase in volume and subsequent thermal and thermal conductivity of concrete. strength of concrete is exceeded.
damage to concrete (Bažant & Kaplan 1996). Since fracture energy determines the Despite such a moisture clog having been
Although some contradictory results specific energy required for crack propaga- experimentally shown to develop, it is widely
have been reported in the literature, there tion, lower fracture energy is associated with accepted that pore pressure alone cannot be
appears to be general agreement that a increased spalling risk. Bažant and Prat (1988) responsible for spalling since the mechanism
larger coarse aggregate size reduces spalling concluded from experimental results that the does not account for increased spalling risk
susceptibility. Experimental results by Pan et fracture energy of concrete gradually decreas- due to confinement (Jansson & Boström
al (2012) support this, suggesting the cause es with increased temperature, thereby mak- 2012) and since resulting pore pressure is
to be related to an increased characteristic ing concrete gradually more susceptible to usually well below typical ranges of concrete
length of the fracture process zone that is spalling with increased temperature. tensile strength (Bažant & Thonguthai 1979).
associated with larger aggregates, although The high heating rate imposed during
increased permeability is likely an additional Structural properties fire exposure results in a high temperature
contributing factor (Van der Merwe 2019). Previous investigations have shown that the gradient over the depth of a concrete sec-
High cement and silica fume content is spalling susceptibility of concrete generally tion which, in turn, leads to the develop-
also said to increase spalling susceptibility, increases (i) with increased element size ment of thermal stress. The compressive
correlating with lower permeability and a (Boström et al 2007), (ii) where structural stress state that results near the heated
subsequent higher internal pressure build- geometry results in stress concentration at surface was proposed by Saito (1966) as
up potential (e.g. Klingsch 2014; Kodur & sharp corners (Khoury & Anderberg 2000), a mechanism of explosive spalling. It was
McGrath 2006). and (iii) where thermal restraint prevents suggested that such compressive stress
expansion of concrete during heating induces cracks that separate a surface
Material properties of (Majorana et al 2010). layer. Thermal buckling of this weakened
hardened concrete delaminated layer is the proposed cause of
It has often been stated in the literature that Loading spalling. Experimental observations have
greater compressive strength is often associ- It is generally accepted that an increased shown, however, that spalling can occur in
ated with increased spalling susceptibility heating rate increases spalling susceptibility some concretes at heating rates much lower
4 Volume 65 Number 2 June 2023 Journal of the South African Institution of Civil Engineering
Pressure difference Heat flux
Px
Tf
(1)
Ts (2)
σt (i)
(ii)
Patm Patm σℓ
Tx
(iii)
σp
f t2 = σ p2 + v ∙ σ p ∙ (σℓ + 2 ∙ σ t)(4)
0.8
f t,θ/f t,20 (–)
Journal of the South African Institution of Civil Engineering Volume 65 Number 2 June 2023 5
and the thickness of a concrete element
1.0 heated from two sides (Figure 6). An
adapted version of this nomogram was later
incorporated in Eurocode (EN 1995).
0.8
Numerical models
Degree of saturation (–)
20 35
Explosive spalling
Surface spalling
No spalling 30
15
25
Compressive stress (MPa)
(1)
20
σ (N/m2)
10
Spalling risk
15
(2)
10
5 Spalling
unlikely
5
0 0
0 40 80 120 160 0 50 100 150 200
(a) Specimen thickness (mm) (b) t (mm)
Figure 6 T he nomogram for spalling risk proposed by Sertmehemetoglu (adapted from Sertmehemetoglu 1977) proposed a means of assessing spalling
risk based on the thickness of a concrete element and the level of applied load; an adapted version of this nomogram was included in an earlier
version of Eurocode (adapted from EN 1995)
6 Volume 65 Number 2 June 2023 Journal of the South African Institution of Civil Engineering
Considering a 150 mm thick concrete slab high-temperature stress state in concrete into the surrounding porous cement matrix.
exposed to fire on one side, and with a refer- exceeds its tensile strength, steel fibres can The interconnected channels created in this
ence concrete tensile strength of 2.5 MPa, Lu be expected to increase the resistance to manner are said to increase permeability
(2015) proposed a boundary intrinsic perme- explosive spalling. Nevertheless, contra- and, subsequently, reduce pore pressure.
ability value of 2 × 10–17 m2. The correlation dicting experimental observations of its Another proposed mechanism suggests that
between concrete tensile strength and beneficial effect have been reported, with there exists a difference in thermal expan-
intrinsic permeability investigated by Van some tests showing improved spalling resist- sion between PP-fibres and the surrounding
der Merwe (2019) suggests that such intrinsic ance (Kodur et al 2003), whilst others notice concrete matrix, resulting in micro cracks
permeability values are, however, unlikely for no substantial improvement (Klingsch et al that increase overall permeability and,
the concrete tensile strength assumption. 2013). Moreover, the addition of steel fibres subsequently, reduce pore pressure. Zhang
has been noted to increase the violent nature et al (2018) investigated ultra-high perfor-
of explosive spalling in some cases due to the mance concrete with PP-fibres after heating
GENERAL METHODS release of a greater build-up of energy accom- to elevated temperatures at 5°C/min. A
OF PREVENTION modated by the increased tensile strength subsequent review of samples with a scan-
Experimental observations of explosive (Fib Bulletin 38: 2007). This supports recent ning electron microscope revealed micro
spalling have shown the occurrence of this observations of a steel-fibre-reinforced con- cracks in the concrete matrix between fibres
phenomenon to be complex and often errat- crete warehouse floor (Figure 1a). prior to melting, suggesting that the thermal
ic. Fib Bulletin 38 (2007) reports a test series It is widely agreed that the addition of expansion mismatch between fibres and the
of ten similar specimens where five samples polypropylene fibres (PP-fibres) reduces cement matrix is responsible for the reduced
spalled, whilst five did not. Elsewhere this the risk of spalling. Explosive spalling of spalling risk. Despite the variation in
observation is supported, mentioning that, concrete with compressive strength of up to conclusions, it is clear that PP-fibres reduce
for concrete samples from the same batch 150 MPa has been found to be avoided when the risk of explosive spalling by increasing
and under identical curing and testing adding 0.05 – 0.10% PP-fibres (Fib Bulletin permeability. It is argued here that the
conditions, some samples would spall whilst 38: 2007). Most conclusions in literature interconnectedness of increased porosity
others would not (Connolly 1995; Majorana seem to agree that the observed benefit of that is required to increase permeability, is
et al 2010). To this extent, Bisby et al (2014) adding PP-fibres is related to increased per- more likely to result from the formation of
recommend that “… attempts should be meability and improved moisture transport, microcracks than absorbed molten fibres.
made to avoid spalling … in practice, since thereby reducing pore pressure build-up (e.g.
spalling introduces uncertainties that are Heo et al 2012; Klingsch et al 2013; Maluk Protective barriers
very difficult to rationally account for in et al 2017). Increased entrained air during The aim of protective barriers, be they
design.” Where a risk of spalling is therefore the concrete mixing process, and poor panels, mortar or intumescent paint, is to
suspected, or where the consequence of adhesion between fibres and the surround- minimise heat transfer to concrete in the
explosive spalling is expected to have a sub- ing concrete matrix have been proposed event of a fire. No design criteria are cur-
stantial effect on structural integrity or the as potential mechanisms for the beneficial rently available for the use of these measures
safe evacuation of occupants, preventative effect of adding PP-fibres, but the formation to avoid explosive spalling. As such, care-
methods should be considered. of continuous channels after fibres melt, and ful experimental validation of protective
differential thermal expansion are the two barriers is typically required. Reviewing a
Addition of fibres most widely argued mechanisms. Following series of fire tests by Klingsch et al (2013),
In addition to improving ductility, the experimental tests of HSC with PP-fibres, Lu (2015) concluded that the provision
addition of steel fibres to concrete increases Kalifa et al (2001) studied test samples with of an insufficient protection layer can, in
its tensile strength. Since it is gener- a scanning electron microscope. They con- some cases, increase the violent nature
ally believed that spalling occurs when the cluded that molten PP-fibres are absorbed of explosive spalling. Figure 7 compares
(a) (b)
Figure 7 H
SC slabs after fire testing: (a) without a protective lining, explosive spalling started after 15 minutes and continued gradually to a depth
of 100 mm after stopping the test, (b) where a 10 mm protective lining was provided, spalling occurred with a single load explosion after
119 minutes to the same depth (Klingsch et al 2013 – ETH Zurich, open-access source for non-commercial use)
Journal of the South African Institution of Civil Engineering Volume 65 Number 2 June 2023 7
explosive spalling of a slab with and without occur if the moisture content is below
a protective lining (Klingsch et al 2013). (1) k% by weight, with 2.5 ≤ k ≤ 3.0% and a
Although the protective lining delayed the recommended value of 3%. Although the
occurrence of spalling, damage occurred to moisture content of concrete is undoubt-
the same depth. Moreover, compared to the (2) edly an important parameter affecting the
gradual, layer-by-layer spalling observed for propensity of concrete to spall during fire
the slab without a protective layer, spalling exposure, it is difficult to predict what the
occurred with a single loud explosion where moisture content of a concrete element will
the protective layer had been applied. be, and equally hard to influence its value.
Moreover, spalling has been observed at
Reinforcement details lower moisture content levels (e.g. Zheng et
A series of experimental and numerical al 2010). Results from various studies suggest
investigations of the behaviour of reinforced Figure 8 E xperimental observations by that the relevance of moisture content as a
concrete columns in fire were conducted by Kodur and McGrath (adapted from parameter affecting spalling susceptibility
Kodur and McGrath (2006), and Kodur and Kodur & McGrath 2007) suggest that depends on the type of aggregate, cross-
Phan (2007). They concluded that transverse improved confinement detailing sectional shape of the element, level of com-
confinement reinforcement details that of stirrups to columns improves pressive stress, concrete mixture and heating
are typically recommended in regions of the spalling resistance of columns: rate (Jansson 2013). It has been suggested that
seismic hazard are well suited to improve (1) ends of stirrups should be such a moisture content limit be replaced by
the explosive spalling resistance of columns. anchored into the confined core, and a list of concrete conditions known to result
The detail calls for (i) a reduced transverse (2) cross ties should be provided in a high moisture content and limited drying
reinforcement diameter with decreased (Jansson 2013). These can include concrete
vertical spacing instead of larger diameters data and the recommended reinforcement exposed to saturated conditions or with per-
at greater spacing, (ii) the provision of cross details are considered by the engineer. manent formwork that prevents drying.
ties, and (iii) anchorage of tie ends into con- HSC up to class C80/95 with a silica fume
fined concrete, i.e. a 135° hook instead of the content of less than 6% of the cement mass, is
more common 90° hook (Figure 8). treated as NSC. For higher concrete classes, CONCLUDING REMARKS
or where the silica fume content exceeds 6%, Although less severe types of fire-induced
four possible methods are recommended to spalling might affect normal concrete
CURRENT REGULATIONS prevent the occurrence of spalling: strength grades to a limited extent, the
AND GUIDELINES (A) providing mesh reinforcement near the susceptibility of concrete to spall explosively
Eurocode 2 (EN 2004) addresses the fire exposed concrete surface in fire is likely to become an increasingly
design of concrete structures and is per- (B) prescribing concrete that has been relevant risk to be considered by structural
haps the most suitable current reference, proven (experimentally or by experi- engineers as the use of denser concretes and
since the suite of SANS standards does ence) not to be susceptible to spalling higher concrete grades increase in South
not include similar clauses. This standard (C) providing a protective layer that has Africa. This paper highlighted some key
recommends that the occurrence of spalling been shown to prevent spalling considerations, but also showed that, to a
should either be prevented, or that the (D) including a minimum of 2 kg/m3 large extent, this phenomenon remains poorly
engineer should consider the consequences monofilament polypropylene fibres. understood. Much research is still needed
of spalling on the structural capacity of ele- However, even when adhering to regulations, in order to more accurately assess this risk.
ments. Spalling is addressed separately for spalling can still occur. A full-scale fire test To date, the only accurate method of assess-
normal strength concrete (NSC) and high was performed on a seven-storey reinforced ing the susceptibility of concrete to spall
strength concrete (HSC). It is suggested concrete building in Cardington in 2001 explosively in fire, is to perform fire tests of
that spalling of NSC is unlikely to occur if (Bailey 2002). Adhering to prescriptive rules, representative concrete elements. Since both
the initial free moisture content is less than Eurocode recommendations suggest that the size and geometry of a concrete element
3%. An earlier version of this standard (EN spalling should not occur and that its occur- influence the risk of spalling, small-scale
1995) mentioned that this moisture content rence can be ignored. Nevertheless, severe samples, such as standard concrete cubes
would most likely not be exceeded if a con- spalling did occur. The outcome highlights or cylinders, should not be considered as
crete element is subjected to internal expo- our poor understanding of the phenomenon representative (e.g. Boström et al 2007). Due
sure. Where this limit is exceeded, the risk of explosive spalling, and the limited means to the great cost of such fire tests, various
of spalling should be more carefully investi- available to practising engineers to assess the studies attempted to develop a numerical
gated, considering (i) the moisture content, spalling susceptibility of concrete. means of assessing spalling susceptibility.
(ii) aggregate type, (iii) concrete permeabil- Significant progress has been made towards
ity, and (iv) the expected (or design) heating such a numerical means of assessment (Lu
rate. It does, however, give no guidance as DEVELOPMENTS 2015; Lottman 2017), but fundamental input
to how these factors should be reviewed to Recent findings are set to influence the parameters for such models have not been the
assess spalling susceptibility. Nevertheless, evolution of Eurocode provisions pertain- focus of many studies (Van der Merwe 2019).
the standard recommends that no further ing to the susceptibility of concrete to Considering results reported to date, the most
in-depth spalling assessment is necessary spall explosively in fire. Eurocode (EN effective method of prevention is to include
if the tabulated concrete material property 2004) mentions that spalling should not polypropylene fibres to a concrete mixture.
8 Volume 65 Number 2 June 2023 Journal of the South African Institution of Civil Engineering
ACKNOWLEDGEMENTS assessment. Task Group 4.3. Lausanne, Switzerland: Proceedings, 10th International Conference on
The author wishes to thank Prof Dr Mario International Federation for Structural Concrete. Structures in Fire. Ulster University, Belfast, Ireland,
Fontana, retired professor at ETH Zurich, Gawin D, Majorana, C E & Schrefler, B A 1999. pp 77–84.
for his guidance during the author’s doctoral Numerical analysis of hygrothermal behaviour and Majorana, C E, Salomoni, V A, Mazzucco, G &
study focusing on material properties related damage of concrete at high temperature. Mechanics Khoury, G A 2010. An approach for modelling
to explosive spalling of concrete in fire. of Cohesive-Frictional Materials, 4: 37–74. concrete spalling in finite strains. Mathematics and
Harmathy, T Z 1965. Effect of moisture on the fire Computers in Simulation, 80: 1694–1712.
endurance of building elements. In: Moisture Maluk, C, Bisby, L & Terrasi, G P 2017. Effects of
REFERENCES in Materials in Relation to Fire Tests. West polypropylene fibre type and dose on the propensity
Albrektsson, J, Flansbjer, M, Lindqvist, J E & Jansson, R Conshohocken, PA: ASTM International, 74–95. for heat-induced concrete spalling. Engineering
2011. Assessment of concrete structures after fire. SP Heo, Y-S, Sanjayan, J G, Han, C-G & Han, M-C 2012. Structures, 141: 584–595.
Report 2011:19. SP Technical Research Institute of Limited effect of diameter of fibres on spalling Meyer-Ottens, C 1972. Zur Frage der Abplatzungen
Sweden. protection of concrete in fire. Materials and an Betonbauteilen aus Normalbeton bei
Bailey, C 2002. Holistic behaviour of concrete buildings Structures, 45: 325–335. Brandbeanspruchung. PhD Thesis. Braunschweig
in fire. Proceedings of the Institution of Civil Hertz, K D 2003. Limits of spalling of fire-exposed University of Technology, Germany.
Engineers, Structures and Buildings, 3: 199–212. concrete. Fire Safety Journal, 38: 103–116. Mindequia, J-C, Pimienta, P, Carré, H & La Borderie, C
Bažant, Z P & Thonguthai, W 1979. Pore pressure Jansson, R & Boström, L 2012. Determination of fire 2013. Experimental analysis of concrete spalling due
in heated concrete walls: Theoretical prediction. spalling of concrete: Relevance of different test to fire exposure. European Journal of Environmental
Magazine of Concrete Research, 31(107): 67–76. methods. Proceedings, 7th International Conference and Civil Engineering, 17(6): 453–466.
Bažant, Z P & Prat, P C 1988. Effect of temperature on Structures in Fire, Zurich, Switzerland, 6–8 June Pan, Z, Sanjayan, J & Kong, D 2012. Effect of aggregate
and humidity on fracture energy of concrete. ACI 2012, pp 581–587. size on spalling of geopolymer and Portland cement
Materials Journal, 85(4): 262–271. Jansson, R 2013. Fire spalling of concrete: Theoretical concretes subjected to elevated temperatures.
Bažant, Z P & Kaplan, M F 1996. Concrete at and experimental studies. PhD Thesis Stockholm, Construction and Building Materials, 36: 365–372.
High Temperatures: Material Properties and Sweden: KTH Royal Institute of Technology. Saito, H 1966. Explosive spalling of prestressed concrete
Mathematical Models. New York: Longman. Kalifa, P, Chéné, G & Gallé, C 2001. High-temperature in fire. Bulletin of the Fire Prevention Society of Japan,
Bisby, L, Mostafaei, H & Pimienta, P 2014. State-of- behaviour of HPC with polypropylene fibres: From 15(2): 23–30.
the-art on fire resistance of concrete structures: spalling to microstructure. Cement and Concrete Sertmehemetoglu, Y 1977. On a mechanism of spalling of
structure-fire model validation. Draft report Research, 31: 1487–1499. concrete under fire conditions. PhD Thesis. University
submitted to Applied Research Associates, Khoury, G A 2000. Effect of fire on concrete and of London, UK.
Albuquerque, NM. concrete structures. Proceedings of Structural Shorter, G W & Harmathy, T Z 1961. Discussion: “The
Boström, L, Wickström, U & Adl-Zarrabi, B 2007. Effect Engineering Materials, 2: 429–447. fire-resistance of prestressed concrete beams” by L A
of specimen size and loading conditions on spalling Khoury, G A & Anderberg, Y 2000. Concrete spalling Ashton & S C C Bate. Proceedings of the Institute of
of concrete. Fire and Materials, 31: 173–186. review. Fire Safety Design, 60: 5–12. Civil Engineers, 21(2): 313–315.
Buchanan, A H & Abu, A K 2017. Structural design for Klingsch, E, Frangi, A & Fontana, M 2013. Explosive Van der Merwe, J E 2019. Constitutive models towards
fire safety. Hoboken, NJ: Wiley. spalling of concrete in fire. IBK Test Report 352. the assessment of concrete spalling in fire. PhD Thesis.
Carré, H, Pimienta, P, La Borderie, C, Pereira, F & Zurich, Switzerland: ETH Institute of Structural ETH Zurich, Switzerland.
Mindeguia, J-C 2013. Effect of compressive loading Engineering (IBK) Van der Merwe, J E 2022a. Evaluation of concrete tensile
on the risk of spalling. MATEC Web of Conferences, Klingsch, E W H 2014. Explosive spalling of concrete in strength as a function of temperature. Construction
6: 01007. fire. PhD Thesis. ETH Zurich, Switzerland. and Building Materials, 329: 127179.
Connolly, R J 1995. The spalling of concrete in fires. PhD Kodur, V K R, Cheng, F-P, Wang, T-C & Sultan, M-A Van der Merwe, J E 2022b. Effect of compressive stress
Thesis. Birmingham, UK: Aston University. 2003. Effect of strength and fiber reinforcement on on the intrinsic permeability of concrete. Results in
Dwaikat, M B & Kodur, V 2009. Hydrothermal model fire resistance of high-strength concrete columns. Materials, 14: 100280.
for predicting fire-induced spalling in concrete Journal of Structural Engineering, 129(2): 253–259. Van der Merwe, J E 2022c. Towards the estimation of
structure systems. Fire Safety Journal, 44: 425–434. Kodur, V K R & McGrath, R 2006. Effect of silica fume concrete intrinsic permeability. Cement and Concrete
EN (European Standard) 1995. ENV 1992-1-2 (1995). and lateral confinement on fire endurance of high Research, 159: 106885.
Eurocode 2: Design of concrete structures. Part 1-2: strength concrete columns. Canadian Journal of Civil Zhang, D, Dasari, A & Tan, K H 2018. On the
General rules – Structural fire design. Brussels: Engineering, 33: 93–102. mechanism of prevention of explosive spalling in
European Committee for Standardization. Kodur, V K R & Phan, L 2007. Critical factors governing ultra-high performance concrete with polypropylene
EN 2004. EN 1992-1-2 (2004). Eurocode 2: Design the fire performance of high-strength concrete fibers. Cement and Concrete Research, 113: 169–177.
of concrete structures. Part 1-2: General rules – systems. Fire Safety Journal, 42: 482–488. Zheng, W Z, Hou, X M, Shi, D S & Xu, M X 2010.
Structural fire design. Brussels: European Committee Lottman, B B G 2017. The spalling mechanism of fire Experimental study on concrete spalling in
for Standardization. exposed concrete. PhD Thesis. Technical University prestressed slabs subjected to fire. Fire Safety Journal,
Fib Bulletin 38: 2007. State-of-the-art report. Fire design of Delft, Netherlands. 45: 283–297.
of concrete structures. Materials, structures and Lu, F2015. On the prediction of concrete spalling under Zhukov, V 1976. Reasons of explosive spalling of
modelling. Task Group 4.3-1. Lausanne, Switzerland: fire. PhD Thesis. ETH Zurich, Switzerland. concrete by fire. Beton I zhelezobeton (Concrete and
International Federation for Structural Concrete. Lu, F, Beartschi, R, Van der Merwe, J & Fontana, M Reinforced Concrete), Vol 3.
Fib Bulletin 46: 2008. State-of-the-art report. Fire design 2018. The effect of explosive spalling on punching
of concrete structures. Structural behaviour and shear strength of concrete slabs exposed to fire.
Journal of the South African Institution of Civil Engineering Volume 65 Number 2 June 2023 9
TECHNICAL PAPER
Design criteria for
upstream raised tailings
JOURNAL OF THE SOUTH AFRICAN
INSTITUTION OF CIVIL ENGINEERING
storage facilities
ISSN 1021-2019 (print) | ISSN 2309-8775 (online)
Vol 65 No 2, June 2023, Pages 10–16, Paper 1501
Journal of the South African Institution of Civil Engineering Volume 65 Number 2 June 2023 11
made of the natural climatic deficit to des-
iccate and stabilise the tailings to eliminate B
liquefaction potential.
Liquefiable
Non-liquefiable
No compelling reason to ban tailings prism
tailings
upstreaming in Southern Africa
The arguments for and against do not A D C
provide a clear direction, but it is evident
that there is no compelling reason to ban Figure 1 Illustration of supporting prism
upstreaming in Southern Africa. There
are, however, conditions for proceeding. desaturated condition, and to prevent secondly, in the role of independent tailings
Morgenstern (2018) refers to these condi- seepage from issuing from the face of review board member. This experience has
tions as the key principles that must be the tailings dam. shown that designers have begun to deviate
adhered to in the design, construction and 6. Design analyses must include both from the rules, often in minor ways, and
operation of tailings facilities. So, what undrained strength analysis and sometimes in significant and important
are these principles and how should we effective stress analysis, with design ways. For example, in the time before
be applying them? The following section controlled by the analysis type that best designs were prepared by professional engi-
presents the principles to which Professor represents the anticipated behaviour neers, facilities were limited to a maximum
Morgenstern was referring to, as ten rules. of the tailings. A wide range of factors, height of 30 m based on empirical rules.
including material type, degree of This limit made it relatively easy to achieve
consolidation, loading rate and stress a desaturated prism with width approxi-
TEN RULES FOR path, must be assessed in selecting the mately equal to the final height. However,
UPSTREAM RAISING appropriate analysis method. as the height of facilities has increased well
The following rules for upstreaming have 7. A high degree of regular performance above 30 m, the reason for the empirical
been derived based on experience and monitoring, reviews and ongoing rule has been forgotten and the required
with reference to the published work of involvement by the designer (or another desaturated prism width has not been
Lenhart (1950), Vick (1992) and Martin and engineer formally appointed to assume maintained. This has been particularly
McRoberts (1999). The rules have in some this role) is essential to confirm that prevalent and concerning for facilities
instances been modified and summarised design intent is being satisfied. raised with cyclones in the upstream mode.
by the author, and reference should there- 8. Conventional upstream dams cannot There is a need therefore always to refer to
fore be made to the original publications be considered for areas of moderate the ten rules, and to explicitly adopt them
for the full descriptions. seismicity without improved upstream in design.
1. A sufficiently wide beach, relative to construction, involving a combina-
the ultimate height of the dam, must tion of compaction or other means of The ten rules discussed
be maintained to form a strong, wide, densification (such as desiccation) of the
drained (unsaturated), and/or dilatant outer shell and good internal drainage. The outer shell or prism
(non-contractive) outer shell (prism). Upstream dams should not be consid- The dimensions of the outer shell, which
2. The prism upon which the stability ered in areas with high seismicity. must be maintained in a partially saturated
of the dam relies must not be com- 9. The design must be consistent in terms or dilatant state, need to be determined
promised by underlying tailings that of design requirements (e.g. minimum by analysis. As a starting point, a support-
are contractive or weak which, when beach width) versus operational ing prism that forms a triangle extending
subjected to excess strain or seismic requirements (e.g. pond size required downstream from the final crest will be
loading, can either weaken or liquefy to for clarification, storm storage and adequate in most cases (illustrated as
the point where the factor of safety is freeboard). The geotechnical design of triangle ABC in Figure 1). Optimisation
insufficient. upstream tailings dams should not be may reduce the required prism dimensions
3. The prism must be of sufficient width to carried out in ignorance of operating to the triangle ABD, including reposition-
retain “bursting pressures” (Casagrande constraints. ing of B to prevent topping failures. The
& McIver 1970) of the contractive tail- 10. Seepage conditions within the dam extended prism has implications for drain
ings that may be located upstream of must be well-defined, requiring a good design which go beyond the current guid-
the prism. understanding of hydraulic properties, ance as offered in ICOLD Bulletin 97
4. Where the tailings prism is relied on to pore pressure profiles and hydraulic (ICOLD 1994). This point also addresses
provide stability, the rate of raising of gradients. the second rule insofar as having a stable
the dam must be sufficiently slow, such Although these rules have not been pre- prism will preclude the presence of under-
that there is a sufficient dissipation of scribed by South African guidance and lying tailings that are contractive or weak.
excess pore pressures and desaturation regulations, they are generally understood
in the supporting prism. and followed. However, the author has Bursting pressure
5. There must be sufficient underdrain- been exposed to a large proportion of the The prism must be of sufficient width to
age (drainage blanket, finger drains, operational Southern African tailings retain “bursting pressures” which can build
etc) and/or a pervious foundation to facilities over the past 15 years – firstly, up behind upstream raises, as described
maintain the prism in a drained and as an operator with Fraser Alexander, and by Casagrande and McIver (1970) and
12 Volume 65 Number 2 June 2023 Journal of the South African Institution of Civil Engineering
Performance monitoring
Downstream facing of dam: Virtual (imaginary) plane on Water level on top of tailings It goes without saying that a high degree of
Berms of soil or rockfill to which bursting pressure acts
regular and rigorous performance monitor-
create outer wall of tailings
dam, placed on top of ing is required for upstream raised facilities
one another as the dam to provide assurance that the development
increases in height of the facility is in conformance with the
design requirements. Monitoring should
γw+t H be designed to measure performance for
all possible failure mechanisms and to
provide early warning of developments that
indicate a trend towards non-conformance.
Ground on which dam is built Monitoring cannot provide early warning
of undrained failure and therefore the
Tailings monitoring system should be designed to
The arrows represent bursting measure the parameters that must change
pressures acting on the virtual plane to initiate an undrained failure. Monitoring
of earthquakes is, however, not useful as an
Figure 2 Illustration of bursting pressures (adapted from Casagrande & McIver (1970)) observational tool.
illustrated in Figure 2. This mechanism is pond play an important role in determining Consistency between
seldom evaluated but needs specific atten- the extent and arrangement of drainage. design requirements and
tion, especially for existing upstream raises Drainage needs to be more extensive to operational outcomes
which are formed from earth or other bring about drawdown where the pool The design requirements must be explicitly
materials against which hydrostatic pres- is located close to the crest, and where stated by the designers so that performance
sure can develop. Casagrande and McIver layering, which leads to higher horizontal indicators can be set and monitored dur-
(1970) go on to say that “… the practice of than vertical permeability, may require ing operations. For example, a minimum
raising tailings dams by a series of dykes, the use of chimney or curtain drains that beach width is assumed for the drainage
each supported in part by the underlying interrupt the continuity of the horizontal design and stability analysis but is often
dyke and in part on soft or loose tailings, layers to intercept seepage moving laterally. only shown on cross-sections that appear
does not assure adequate safety. This To provide for the required drainage, seep- in the appendices of design reports. The
construction practice can provide safe dams age conditions must be well defined and minimum dry beach width therefore
only when the sand fractions of the efflu- understood. This requires comprehensive does not get to be established as a formal
ent … are separated and compacted to at analysis by specialised personnel. performance requirement. The process of
least the axis of the final dike of the dam.” crystalising design assumptions into opera-
The axis of the final dike referred to by Analysis tional performance criteria requires a good
Casagrande and McIver (1970) is illustrated The analyses must include both undrained understanding of the design models and
by line BC in Figure 1. and drained analysis as a starting point. of the upset conditions that could result
A drained analysis provides a measure in non-conformance. Senior oversight of
Rate of rise of the margin of safety against conven- design is particularly important for identi-
Rate of rise, tailings layer thickness and tional loading conditions in the absence of fying these requirements before operations
cycle management are determinants of the any events that might trigger undrained commence.
rate of desiccation of tailings in a particu- behaviour. It remains the most reliable
lar climatic setting. The rate of rise and method to establish the margin of safety.
the layer thickness and cycle time must An undrained analysis, on the other MEETING THE REQUIREMENTS
be established to achieve the required hand, indicates the margin of safety against IN SOUTHERN AFRICA
degree of desaturation to preclude lique- abnormal loading conditions causing a The ten rules can be met for new upstream
faction. The degree of desaturation that is failure. The most conservative approach designs without too much additional cost.
required to ensure that contraction during is to assume that the worst-case loading It is therefore strongly advocated that
shearing does not result in pore pressure conditions can arise, and to design accord- designers should fully internalise the rules
build up when subjected to straining or ingly. In many cases with existing facilities, and ensure that they are built into future
seismic loading is not well established. it is not practicable to assume the worst designs. Independent reviewers should also
Hence a maximum degree of satura- case, and it becomes necessary to examine focus on the rules and should be specifical-
tion of 85% is usually assumed, unless a the probability of occurrence of abnormal ly requested to provide opinion on whether
higher percentage can be demonstrated to loading conditions. Seismic loading is one the rules have been incorporated into the
suffice. of the loading conditions that can induce design and, if not, what the risks are.
undrained behaviour, and hence, where Many of the existing facilities in
Internal drainage the tailings can liquefy, upstreaming Southern Africa will not meet the require-
Sufficient underdrainage must be provided should not be considered in seismically ments, principally because they will fail on
to maintain the prism in a drained state. active areas without improved upstream rules 1 and 2. The prisms that have been
The facility geometry and location of the construction. designed to support the weak tailings in
Journal of the South African Institution of Civil Engineering Volume 65 Number 2 June 2023 13
the interior are often undersized owing to
incorrect positioning of the drains, failure
Urgent action including
of the drains or inadequate gradients. This
QQ Low FOS and
Drained consideration of
QQ High probability of failure
then leads to a situation where the critical cessation of operations
failure surface passes through tailings that
QQ Low FOS and Urgent remedy including
have inadequate peak and/or post-peak Undrained QQ High probability of failure consideration of
undrained shear strength and where the peak
for credible triggers cessation of operations
factor of safety criteria are not met.
The real risk is, however, somewhat High probability of liquefaction
Residual or
and loss of containment Timely remedy
different. Several tailings facilities exist in liquefied
due to extreme event
Southern Africa with undrained factors
of safety of less than 1 for peak undrained
strength with an implied probability of No action
14 Volume 65 Number 2 June 2023 Journal of the South African Institution of Civil Engineering
and Engineers of Record should not be
distracted by focusing blindly on meeting
Failure modes Define failure modes and consequences
the standards.
REFERENCES
Mitigate Develop mitigation measures for high risks
Boswell, J & Sobkowicz, J 2018. Leading versus lagging
indicators of tailings dam integrity. Proceedings,
Prepare costs and benefit for high risk and propose viable 22nd International Conference on Tailings and
Assess
mitigation to achieve ALARP
Mine Waste. 30 September – 2 October 2018,
Keystone, CO.
Figure 4 Simplified ALARP process Casagrande, L & McIver, B N 1970. Design and
construction of tailings dams. In Brawner, C O &
even further until the cost and other are applied, there is no reason why future Milligan, V (Eds). Proceedings, 1st International
impacts of additional risk reduction are upstream raised facilities cannot be Conference on Stability in Open Pit Mining,
grossly disproportionate to the benefit. designed and constructed to be as safe as American Institute of Mining, Metallurgical
The ALARP process must be carried centreline and downstream facilities. and Petroleum Engineers. New York, Vol 11,
out in a systematic manner and should be Eliminating upstreaming in Southern pp 181–203.
documented to provide an audit trail and Africa is not justified by the record of Cusano, G, Gonzalo, M R, Farrell, F & Remus, R
serve as the basis for justification to regula- failures of upstream facilities in the region. 2017. Best Available Techniques (BAT) reference
tors and stakeholders that the process has The cost of shutting down and replacing document for the non-ferrous metals industries.
been thorough, and that the outcome is existing facilities associated with operating Brussels: EIPPCB (European Integrated Pollution
defensible. The principles defining the mines would be ruinous for most mines, Prevention and Control Bureau). Available at:
process are not yet settled and there are while the risk has been shown by the https://eippcb.jrc.ec.europa.eu/reference/
many ways in which an outcome could be record to be reasonably low. non-ferrous-metals-industries-0.
achieved. Wates (2021) proposed a simpli- The continued use of existing upstream GTR (Global Tailings Review) 2020. Global
fied process that is illustrated in Figure 4. raised facilities can be motivated, provided industry standard on tailings management.
In the context of ALARP, tolerable risk, there is strong engineering backing and Available at: https://globaltailingsreview.org/
as defined in ICOLD Bulletin 130 (ICOLD comprehensive assessment of the probabil- global-industry-standard.
2005) and adapted from the United ity of the development of abnormal trigger ICOLD (International Commission on Large Dams)
Kingdom Health and Safety Executive mechanisms that could induce undrained 1994. Tailings dams design of drainage. Review and
definition, is “… a risk within a range that behaviour. The motivation for continued recommendations. Bulletin 97. Paris: ICOLD.
society can live with to secure certain net use of facilities that do not comply with ICOLD 2005. Risk assessment in dam safety
benefits. It is a range of risk that we do not the minimum factors of safety, but with management. A reconnaissance of benefits. methods
regard as negligible or as something we demonstrated acceptable margins of safety, and current applications. Bulletin 130. Paris:
might ignore, but rather as something we must be underpinned by a comprehensive ICOLD.
need to keep under review and reduce it still ALARP process that provides a defensi- Jennings, J E 1979. Mechanisms of failure and
further if we can.” ble argument for the mitigation that is associated design considerations: The failure of a
In any assessment as to whether risks implemented. slimes dam at Bafokeng. The Civil Engineer in South
have been reduced to ALARP, measures Obsession with the assumption that Africa, 21(16): 135–145.
to reduce risk can be ruled out only if the static liquefaction is the most significant Lenhart, W 1950. Control of tailings from washing
sacrifice involved in taking them would be risk for upstream tailings facilities intro- plants. Rock Products, 53(7): 72–80.
grossly disproportionate to the benefits of duces additional risk, since this distraction Martin, T E & McRoberts, E C 1999. Some
the risk reduction. could lead to the key mechanisms being considerations in the stability analysis of upstream
overlooked. It is critical to point out that a tailings dams. Proceedings, 13th International
design can be flawed (perhaps even fatally) Conference on Tailings and Mine Waste. Banff,
CONCLUSION if important, plausible failure mechanisms Canada, January 1999.
The call to eliminate or ban upstreaming is are not considered and that factors of McRoberts, E C T, MacGowan, R, Eenkooren, R &
not justified in regions with low seismicity. safety cannot fully account for gross omis- Pollock, G 2017. 50 years of successful learnings
Provided that the ten rules for upstreaming sions in design and analysis. Designers by experience: Suncor tailings geotechnical.
Journal of the South African Institution of Civil Engineering Volume 65 Number 2 June 2023 15
Proceedings, 21st International Conference Robertson, P K 2021a. Evaluation of flow liquefaction Vick, S G 1983. Planning, Design, and Analysis of
on Tailings and Mine Waste. Banff, Canada, and liquefied strength using the cone penetration Tailings Dams. Wiley Series in Geotechnical
pp 229–239. test: An update. Canadian Geotechnical Journal, Engineering. New York: Wiley.
Morgenstern, N R 2018. Geotechnical risk, regulation, 59(4). doi: 10.1139/cgj-2020-0657. Vick, S G 1992. Stability evaluation during staged
and public policy. Victor de Mello Lecture. Soils Robertson, P K 2021b. Can you safely construct a construction: A discussion. Journal of Geotechnical
and Rocks, 41(2): 107–129. tailings facility using upstream construction. 18th Engineering, 118: 1282–1289.
Roche, C, Thygesen, K & Baker, E (Eds) 2017. Mine Jennings Lecture. SAICE Geotechnical Division. Wagener, F 1997. The Merriespruit slimes dam failure.
tailings storage: Safety is no accident. A UNEP Available at: https://www.youtube.com/channel/ Overview and lessons learnt. Journal of the South
Rapid Response Assessment. United Nations UCZelzYmgXIXSsed_pXWvEUg. African Institution of Civil Engineering, 39(3): 11–15.
Environment Programme and GRID-Arendal, SANS (South African National Standard) 1998. SANS Wates, J A 2021. Practical application of ALARP.
Nairobi and Arendal. Available at: http://www. 10286: Mine Residue. Pretoria: SABS Standards Presentation to ICOLD Meeting. Santiago, Chile.
grida.no. Division. Available at: https://www.researchgate.net.
16 Volume 65 Number 2 June 2023 Journal of the South African Institution of Civil Engineering
Could a one-size-fits‑all TECHNICAL PAPER
approach apply to the JOURNAL OF THE SOUTH AFRICAN
INSTITUTION OF CIVIL ENGINEERING
extension of stage‑discharge ISSN 1021-2019 (print) | ISSN 2309-8775 (online)
Vol 65 No 2, June 2023, Pages 17–27, Paper 1568
flow‑gauging weirs?
of the Department of Civil Engineering at the
Central University of Technology, Free State
(CUT). He obtained his PhD Eng (Agriculture)
degree from the University of KwaZulu-Natal in
2016 and has more than 25 years’ professional
and academic experience, mainly in the fields of
O J Gericke, V H Williams
hydrology, water resources management, and
river hydraulics. He has published several papers in the design hydrology
field and is involved in research projects funded by the Water Research
At a flow-gauging weir, the stage or flow depth is normally measured continuously and Commission.
converted into discharge using a stage-discharge (SD) rating curve (RC). During flood Contact details:
events, the observed water levels often exceed the flow-gauging weir’s designed measuring Department of Civil Engineering
Faculty of Engineering, Built Environment and Information Technology
capacity or structural limit. Subsequently, the standard calibration of the flow-gauging weir
Central University of Technology, Free State
becomes irrelevant and the extension of the SD RC for above-structure-limit flow conditions Bloemfontein
is required. This paper attempts to identify a one-size-fits-all approach for the extension of South Africa
SD RCs by assessing seven indirect hydraulic extension methods and a one-dimensional HEC- T: +27 51 507 3516
E: jgericke@cut.ac.za
RAS modelling approach against direct SD measurements or extension methods at selected
flow-gauging sites in South Africa. In considering a ranking-based selection procedure and
VERNON WILLIAMS is a civil engineering
associated goodness-of-fit (GOF) criteria, the stepped backwater analysis, slope-area method, technologist and lecturer currently employed by
and 1-D HEC-RAS steady flow modelling proved to be the most appropriate. The other indirect the Tshwane University of Technology (TUT). He
extension methods resulted in larger statistical differences between the at-site benchmark and graduated with BTech Eng (Civil) and MEng
(Civil) degrees from the Central University of
modelled values. Given that the extension of RCs is significantly more affected by the site and
Technology, Free State (CUT), and has a particular
river reach geometry, initial hydraulic conditions, flow regimes, and level of submergence at interest in flood hydrology and river hydraulics.
high discharges than the actual extension method used, it was confirmed that a one-size-fits-all He currently has six years’ academic experience.
approach cannot be applied for the extension of SD RCs in South Africa. Contact details:
Department of Civil Engineering
Faculty of Engineering, Built Environment and Information Technology
Central University of Technology, Free State
INTRODUCTION limit (Petersen-Øverleir & Reitan 2009). Bloemfontein
The first flow-gauging weirs were con- Hence, the standard calibration of the South Africa
structed in 1904 in the Transvaal Province flow-gauging weir will no longer be a true T: +27 12 382 2458
E: williamsvh@tut.ac.za
to measure streamflow continuously in reflection of the actual discharges that
South Africa (Menné 1960). By the year had occurred during the flood events. The
2007, streamflow data had already been latter observed stage levels exceeding the
recorded continuously at 782 different maximum rated stage or flow depth (H)
Department of Water and Sanitation is a common problem in South Africa, as
(DWS) flow-gauging sites, comprising shown in Figure 1, where the RC exceed-
sharp-crested weirs (55%), crump weirs ances and missing data are expressed as
(35%), and the remaining 10% consisting a percentage of the annual maximum
of broad-crested weirs, dam spillways, series (AMS) at 806 DWS flow-gauging
and velocity-area gauging sites (Wessels & weirs with 20 or more years of record at
Rooseboom 2009). However, streamflow each. For example, a standard SD RC, as
is seldom directly measured; instead, the shown in Figure 2, must then be extended
stage or flow depth is measured at regular beyond H to reflect the high discharges (Q)
intervals and converted into discharge exceeding the designed measuring capacity.
using a stage-discharge (SD) rating curve In general, such extensions are normally
(RC) at a flow-gauging weir or specific limited to bankfull flow conditions using
river section. appropriate indirect methods, while high
During flood events, the observed water flow extensions above bankfull flow condi-
levels often exceed the flow-gauging weir’s tions should only be considered in cases
designed measuring capacity or structural where the existing SD relationship includes
The Journal of the South African Institution of Civil Engineering, which is distributed internationally, is a peer-reviewed, open-access journal licensed under a Creative
Commons Attribution Licence (CC BY-NC-ND). Readers may therefore freely use and share the content as long as they credit the original creator and publisher, do
not change the material in any way, and do not use it commercially. Copyright of this article remains with the authors. 17
floodplain flow on the full width of the
floodplain (Gericke & Smithers 2017). 35
Direct measurements, e.g. conventional
current gaugings, are not always possible 30 29.7
owing to various practical constraints asso-
ciated with high discharge events, e.g. high 25 24.6
18 Volume 65 Number 2 June 2023 Journal of the South African Institution of Civil Engineering
peakedness due to a steep rising limb in above-structure-limit, or conversely, for ii. Velocity extension hydraulic radius
relation to the hydrograph base length, and the lowest segment in the case of lower approach (VE-HRA): In using the
the relationship between the observed peak stages (EA UK 2002). The application VE-HRA, a plot of velocity against
discharge (QPxi) and associated direct run- of the SE is limited to uniform channel hydraulic radius (R) frequently
off (QDxi) values, should be used as addi- cross-sections and bankfull flow condi- presents a linear relationship; this
tional criteria to justify the HE extensions tions throughout the entire RC range can also be used to provide values of
up to 20%. Typically, in such cases, the (EA UK 2003). The SE is unsuitable velocity in the extended range of the
additional volume of direct runoff (QDE) for: (i) above bankfull flow conditions, RC. Once v and A have been derived,
due to the extension should be limited to and (ii) any transitions from in-channel Q can be determined using the
5%, i.e. QDE ≤ 0.05 QDxi. flow to overbank full flow conditions, conservation of mass (Ramsbottom
Given that the use of appropriate SD RC modular to submerged flow conditions & Whitlow 2003).
extension methods is warranted not only and vice versa, submerged flow to iii. Velocity extension Manning’s
to enhance the estimation of design flood complete channel control flow, or where approach (VE-MA): In using the
events, i.e. peak discharges, but also to any changes in vegetation and cross- VE-MA, the Manning’s equation is
impact on water resources management, sections occur. used by assuming that the square
this paper attempts to identify a one‑size- b. Logarithmic extension (LE): The LE root of the channel slope (So1/2) and
fits-all approach for the extension of has similarities to the SE; however, Manning’s roughness coefficient
SD RCs by assessing a selection of indirect the stage and discharge are plotted (n) are constant across the full
extension methods (e.g. hydraulic and one- on a logarithmic scale. The advantage range of discharges when compared
dimensional modelling approaches) against of the LE is that it can accommodate to AR2/3. The relationship of Q
direct extension (benchmark) methods changes in the slope of the rising limb versus AR2/3 is then extended to
(e.g. at-site conventional current gaugings, of an RC (EA UK 2002; Ramsbottom & provide the SD relationships for
hydrograph analyses, and level pool rout- Whitlow 2003). According to the British above-structure-limit conditions
ing techniques) at selected flow-gauging Standards Institution International (Ramsbottom & Whitlow 2003).
sites in South Africa. In this paper, it is Organization for Standardization (BSI d. Slope-area method (SAM): The
assumed that the SAM, and by extension ISO 1998), the LE is not recommended, SAM considers the hydraulic radius,
the divided channel method (DCM), are given that it can generate substantial cross-sectional area, channel bed
more appropriate and hydraulically cor- errors and is regarded as inferior to the (water surface) slope, and roughness
rect than the other SD extension methods SE in many ways. Essentially, the LE is coefficients (e.g. Manning’s n-values
available. However, the SBA is regarded a modified version of the SE limited to or Chézy’s ks-values) to determine the
as being more accurate than the SAM, a first-order polynomial function. The SD relationship at a given site. Two
while one-dimensional hydraulic models same hydraulic conditions, limitations, approaches are associated with SAM, i.e.
(e.g. HEC-RAS) better illustrate the actual and data requirements as applicable to the simple approach (SA) and the DCM
hydraulic conditions in a river reach than the SE, apply to the LE. (EA UK 2002). In using the SAM-SA,
simple hydraulic methods. c. Velocity extension (VE) methods: The the cross-sectional areas and water
VE methods can only accommodate surface slopes are determined for a range
limited changes in the channel section of discharges, while the mean velocity
MATERIALS AND METHODS geometry for above-structure-limit con- and discharge in the channel are calcu-
A review of the seven indirect hydraulic ditions, while transitions from modular lated using Manning’s or Chézy’s open-
extension methods and one-dimensional to submerged flow conditions are nor- channel flow equation (EA UK 2002).
HEC-RAS modelling is discussed in this mally poorly presented (EA UK 2003). The SAM-DCM requires surveyed cross-
section, followed by an overview of the The hydraulic conditions, which are sections that include the main channel
location and characteristics of the selected regarded as unsuitable for the SE, also and floodplains. The surveyed cross-
flow-gauging sites. Thereafter, the method- apply to the following VE methods: sections are divided into three panels
ologies adopted are expanded on in detail, i. Velocity extension simple approach using vertical division lines, i.e. the left
followed by the results, discussion and (VE-SA): The VE-SA requires the bank floodplain, main channel, and the
conclusions. computation of the cross-sectional right bank floodplain. The channel rat-
area and velocity at incremental river ing is extended above bankfull, assuming
Review of indirect hydraulic stages similar to the stages used to no friction on the vertical boundaries
extension methods derive the existing SD relationship. for above bankfull flow conditions.
It is essential that a hydrometric data, The RC extension is based on an Separate ratings are calculated for each
gauging, and rating review be conducted extension of velocity against stage. component using different roughness
prior to extending any SD relationship One cross-section is required at the coefficient values and summed to obtain
(Ramsbottom & Whitlow 2003). The fol- flow-gauging site to calculate the a rating for the section (Ramsbottom &
lowing hydraulic methods were considered change in cross-sectional area with Whitlow 2003). Although being regarded
and applied at each site: stage. Once the velocity (v) and cross- as a much more versatile method and
a. Simple extension (SE): The SE uses sectional area (A) have been derived, less sensitive than the SE, LE, and VE
the rating equation for the highest the discharge (Q) can be determined methods, similar hydraulic conditions,
limb of the RC to extend the SD using the conservation of mass limitations, and data requirements could
relationship to stages exceeding the (Ramsbottom & Whitlow 2003). apply to the SAM.
Journal of the South African Institution of Civil Engineering Volume 65 Number 2 June 2023 19
Figure 3 Location of the 10 flow-gauging sites in South Africa (Williams 2023)
e. Stepped backwater analysis (SBA): assessed by friction and contraction or b. Hydraulic conditions: Modular and/
The SBA is based on the standard expansion coefficients. The momentum or submerged flow conditions, variable
step method and applies to the equation applies to situations where submergence due to backwater effects
gradually varied flow (GVF) regime in the water surface profile varies rapidly, and vegetative growth, identification of
non-prismatic channels. Field survey e.g. hydraulic jumps, flow conditions possible hydraulic controls, sediment
data is required, e.g. the longitudinal at bridges, and river confluences transport, unsteady flow conditions,
river profile, cross-sections, and a (USACE 2016). and the influence of in-bank and out-of-
detailed description of the roughness bank flow paths.
coefficients applicable to each cross- Flow-gauging sites c. Geometric properties: Type of flow-
section (Chow 1959). The SBA calcula- Initially, all the SD extension reports avail- gauging weir, e.g. sharp or broad-crest-
tions are done on a trial-and-error able from DWS, i.e. the Discharge Table ed, crump, hydro-flumes, broad-crested
(iterative) basis, which can be easily Improvement Reports, were studied, given flank walls, flood or natural sections,
automated using solver functions in that a standard SD rating table is regarded overall river topography and layout,
Microsoft Excel. as the first step towards the extension river channel and flood plain geometry,
f. 1-D HEC-RAS modelling: The United of any SD relationship. Subsequently, a position of control points within the
States Army Corps of Engineers total of 10 flow-gauging sites, as shown in river system (especially at high flows),
(USACE) developed the Hydraulic Figure 3, were identified and selected. availability of survey data (e.g. cross-
Engineering Centre River Analysis In terms of the hydrometric and geo- sections, longitudinal sections, previous
System (HEC-RAS) software, with metric data requirements, the selection of flood surveys, and the previous or cur-
Version 1.0 having been released in the flow-gauging sites took the following rent allocation of site-specific roughness
August 1995 (Maeder 2015). The aspects into consideration: coefficients).
software can perform one-dimensional a. Streamflow data: Record length, Typically, the chosen flow-gauging sites in
steady flow, two-dimensional unsteady data quality, flow duration curves to Figure 3 are characterised by complex flow
flow, sediment transport, and water highlight the occurrence and frequency (hydraulic) conditions and are not neces-
temperature or quality modelling. The of minimum and maximum flow sarily sites that had gradual changes in
steady flow analysis in HEC-RAS is ranges, and the number of standard geometric properties. Ultimately, the sites
based on the GVF theory. In applying and extended SD relationship tables or chosen are inclusive of the following prob-
the energy equation, energy losses are curves available. lem areas: (i) submergence, (ii) estimation
20 Volume 65 Number 2 June 2023 Journal of the South African Institution of Civil Engineering
Table 1 Summary of flow-gauging site characteristics (Williams 2023)
Latitude Period River Type Structural Benchmark Indirect
Site
Longitude of record Province of structure limit (Q & H) method methods
Compound weir
24°04’58”S 1962/08/27 Mokolo Q = 115 m3/s Level pool
A4H005 5 Sharp crests SE, LE & VE
27°46’23”E to present Limpopo H = 1.4 m (back) routing
6 Broad crests
22°32’07”S 1995/02/02 Mogalakwena Compound weir Q = 153 m3/s Level pool
A6H035 SE, LE & VE
28°53’51”E to present Limpopo 4 Sharp crests H = 1.3 m (back) routing
C2H003
C2H003 28°49’11”S 1924/01/29 Vaal Compound weir Q = 481 m3/s Level pool
SE, LE & VE
(C2H130) 28°03’49”E to present Gauteng 8 Sharp crests H = 2.2 m (back) routing
C2H130 flood section (FS)
Compound weir SE, LE, VE,
29°02’33”S 1969/07/03 Riet Q = 90 m3/s
C5H014 Broad-crest SBA SAM, SBA &
24°35’57”E to present Free State H=1m
V-crump HEC-RAS
Compound weir SE, LE, VE,
33°53’52”S 1970/03/24 Bree Q = 81 m3/s SBA
H5H004 Hydro-flume SAM, SBA &
20°00’46”E to present Western Cape H = 1.4 m
2 Sharp crests HEC-RAS
Compound weir SE, LE, VE,
33°41’05”S 1982/04/07 Touws Q = 131 m3/s
J1H018 V-crump (low) SBA SAM, SBA &
21°08’46”E to present Western Cape H=2m
4 Crumps HEC-RAS
SE, LE, VE,
34°01’02”S 1964/05/01 Gourits Natural flood section (FS)
J4H002 N/A SBA SAM, SBA &
21°45’24”E to present Western Cape at bridge
HEC-RAS
34°01’40”S 1961/05/05 Groot Brak Compound weir Q = 55 m3/s Wolwedans
K2H002 SE, LE & VE
22°13’21”E to present Western Cape 4 Sharp crests H = 1.5 m Dam spillage
Compound weir
28°43’15”S 1967/08/01 Tugela Q = 98 m3/s Hydrograph
V1H026 4 Sharp crests SE, LE & VE
29°21’33”E to present KwaZulu-Natal H = 1.2 m analysis
Hydro-flume
Compound weir
SE, LE, VE,
24°46’08”S 1967/09/01 Sand Sharp crest Q = 76 m3/s
X3H008 SBA SAM, SBA &
31°23’24”E to present Mpumalanga Hydro-flume H = 1.95 m
HEC-RAS
Broad-crested flank walls
problems, (iii) complex geometry, and environment to enable the consistent fitting and subsequent extension of the
(iv) unsuitable sites (e.g. J4H002) for the application of the various hydraulic exten- SD relationships for above-structure-limit
construction of flow-gauging weirs. Given sion methods with limited manual user conditions.
the complex conditions and lack of suf- input. The semi-automated tools consist of
ficient geometric data available at some of different functions (as applicable to each SAM
these flow-gauging sites, not all the above- hydraulic extension method) to ultimately Essentially, the SAM is not regarded as a
listed indirect hydraulic extension methods provide peak discharges at specific stages simplified computation where Manning’s
could be applied. Table 1 contains a sum- above the structural limit. (MA) or Chézy’s (CHZ) equation for open-
mary of all the relevant characteristics channel flow is just used to determine
present at each flow-gauging site. SE and LE methods discharges at different stages. Hence, the
The analyses were conducted and auto- following steps are typically required
Processing of geometric data mated in two distinctive phases: (i) up to and were applied accordingly to deter-
The processing of geometric data, e.g. wet- structural limit, and (ii) above structural mine the: (i) upstream and downstream
ted perimeter, wetted area, and hydraulic limit. Rating curves, for the whole H-range conveyance from known geometric and
radius, was done using the Windows Cross- were generated automatically, while any hydrometric properties (using separate
Section Professional (WinXSPRO) (Hardy trend line evaluations were done manually worksheets for SAM-MA and SAM-CHZ),
et al 2005), which is essentially a channel to determine the best-fit polynomial func- (ii) average conveyance, (iii) energy slope
cross-section analyser. Prior to the use of tion based on the coefficient of determina- by assuming the velocity head equals
WinXSPRO, all the files were generated in tion (r 2) and the estimated discharge (QE). zero, and (iv) corresponding discharge
Microsoft Excel with Position-Elevation In each case, the best trend line equation and assumed discharge. These steps were
Free Form as the data format. In this paper, was used for the required extension. automated in Microsoft Excel using Visual
the outputs related to stage (H), area (A), Basic for Applications (VBA) and the
wetted perimeter (P), and hydraulic radius VE methods GOAL SEEK function.
(R), were used in the further computations. All the VE methods shared an input
worksheet, with inputs for H, A, R, and v up SBA
Development of a semi- to structural limit and separate inputs for The developed SBA worksheet is based on
automated extension tool above-structure-limit geometric proper- the methodology proposed by Chow (1959),
A semi-automated extension tool ties. Essentially, first-order polynomial i.e. the standard step method for natural
was developed in the Microsoft Excel equations were generated to enable the RC (non-prismatic) channels. The Froude
Journal of the South African Institution of Civil Engineering Volume 65 Number 2 June 2023 21
of the estimate (SEE, Equation 1), mean
absolute relative error (MARE, Equation
Input SD relationship and/or Input SD relationship and/or
geometry up to structural limit geometry above structural limit 2), root mean square error (RMSE,
Equation 3), coefficient of determination
(r², Equation 4), Nash-Sutcliffe coefficient
(NSE, Equation 5), and the z-test (Equation
Output: SE: Review best-fit LE: Logarithmic
6), were used as assessment criteria to
Standard RC trend line transformation
compare and assess the indirect extension
methods’ results to those of the at-site
direct extension (benchmark) methods.
Output: Extend RC using the Extend RC using a The z-test indicates how many standard
SE RC best-fit trend line polynomial function
deviations a sample scores above or below
the mean, and the results are classified as:
(i) similar (0 ≤ z ≤ 2), (ii) marginally differ-
Output: VE-SA: Output: ent (2 < z ≤ 2.5), (iii) significantly different
VE-SA RC Compute/plot v vs H LE RC (2.5 < z ≤ 3), and (iv) unacceptable (z > 3)
(Williams 2023).
Each indirect hydraulic extension
Output: VE-HRA: SAM-MA/CHZ: method was ranked against the different
VE-HRA RC Compute/plot v vs R Run procedure GOF criteria and summed and/or averaged
to provide the overall performance. Finally,
the overall rankings and associated relative
frequency of the various z-test ranking
Output: VE-MA:
VE-MA RC Plot Q vs AR2/3 clusters – as defined in (i) to (iv) above –
were used to establish the hierarchical
Output:
order of the various indirect hydraulic
SAM-MA/CHZ RC
extension methods.
⎫
Output: SBA-MA/CHZ:
⎪N
⎪
SBA-MA/CHZ RC Trial-and-error process
⎪ ∑ (QEi – QE)2 –
(N – 2) ⎪⎪i=1
1
SEE =
⎪
⎭
Figure 4 R
C extension flow diagram applicable to indirect hydraulic extension methods
(Williams 2023)
⎫N ⎫ ⎫
⎪ ∑ (QBi – QB)(QEi – QE)⎪ ⎪
2 0.5
⎭i=1 ⎭ ⎪
⎪
number (FR) for each cross-section was Typically, the known water surface eleva-
⎪
⎪
determined to define the flow regime, i.e. tions (up- and downstream) were used as (1)
⎪
N
⎭
super- and/or sub-critical flow. Essentially, the default boundary conditions at most of
∑ (QEi – QB)2
iterative trial-and-error calculations were the flow-gauging sites under consideration. i=1
used until channel profile convergence was In the absence of these known water surface
1 N |QEi – QBi|
reached, i.e. the profile approximates the elevations, the critical depth was entered
uniform depth profile and energy levels are as the downstream boundary condition,
MARE = 100
N
∑ QBi
(2)
i=1
in equilibrium. The procedure matrix or given that sub-critical flow conditions
flow diagram associated with all the indi- (FR < 1) were present at all the flow-gauging
N
rect hydraulic extension methods is shown sites under consideration. The simulated
∑ (QEi – QBi)2
in Figure 4. water surface levels were compared to the i=1
RMSE = (3)
observed water levels, while adjusting the N
1-D HEC-RAS steady flow modelling Manning’s n-values until similar water
The modelling in HEC-RAS was limited to surface levels were reached. N 2
steady flow analyses at the flood sections ∑ (QBi – QB)(QEi – QE)
i=1
only, given that the calibration data at the Assessment of results r² = (4)
N N
flow-gauging weirs is regarded as unsuitable A ranking-based selection procedure was
∑ (QBi – QB)2 ∑ (QEi – QE )2
for RC extensions at above-structure-limit developed to evaluate and select the best i=1 i=1
flow conditions. In addition, the calibration performing indirect hydraulic extension
data also lacks the information required methods at each flow-gauging site. In other N
to conduct an unsteady flow analysis words, the indirect extension method’s ∑ (QBi – QEi)2
i=1
for the full range of flows at the various estimation results (QE) were compared to NSE = 1 – (5)
N
flow-gauging sites under consideration. the at-site or benchmark extension results
∑ (QBi – QB)2
The desired flow regime, i.e. sub-critical, (QB) at each site (i) by considering a large i=1
supercritical, and/or mixed flow, was set of quantitative GOF criteria (Zhong &
defined prior to the actual simulation run Dutta 2015) to assess each method’s accu- QBi – QEi
z= (6)
by specifying the boundary conditions. racy and bias. Typically, the standard error √σBi + σEi
22 Volume 65 Number 2 June 2023 Journal of the South African Institution of Civil Engineering
Where:
SEE is the standard error of estimate
(m3/s)
MARE is the mean absolute relative error
(%)
RMSE is the root mean square error
r² is the coefficient of determination
NSE is the Nash-Sutcliffe coefficient
z is the z-test score
N is the sample size
QBi is the at-site benchmark discharge
(m3/s)
QB is the mean of the at-site bench-
mark discharge (m3/s)
QEi is the estimated (indirect) discharge
using extension methods (m3/s)
QE is the mean of the estimated (indi-
rect) discharge (m3/s)
σBi is the standard deviation of the at- Figure 5 Flow-gauging weir X3H008 in the Sand River (Rademeyer 2023)
site benchmark discharge (m3/s)
σEi is the standard deviation of the esti- was used as the preferred SD extension Given that the SBA-CHZ method is used
mated (indirect) discharge (m3/s). method. Hence, the SBA was also used as the benchmark method (default ranking
as a benchmark against which the other = 1) at this site, the SBA-MA demonstrated
indirect extension methods were compared the second-best overall ranking, followed
RESULTS AND DISCUSSION up to a stage of 5.4 m and corresponding by the HEC-RAS (FS) modelling in the
In this section, the overall ranking results discharge of 2 972 m3/s. The most recent 3rd position. By considering the individual
associated with each flow-gauging site are cross-sectional survey at the flow-gauging GOF statistics, the latter SAM-CHZ and
presented in a summative format. However, weir is inclusive of five flood sections (FSs). SAM-MA methods also proved to be
given that all the indirect hydraulic exten- The current (existing) SD relationship table equally accurate by being interchangeably
sion methods and HEC-RAS modelling (DT10) applicable to X3H008 is listed in ranked at either the 4th or 5th position.
were applied at flow-gauging site X3H008, Table 2. Overall, the LE is the most inappropriate
the latter gauging site is discussed in detail The SD RC extension results using all method with the poorest ranking for all the
as a representative example. Thereafter, the the indirect hydraulic extension methods GOF statistics under consideration.
overall results and performance achieved and HEC-RAS modelling are shown in Overall, the hydraulically correct
at the 10 flow-gauging sites are discussed Figure 6 (on page 24). The GOF statistics methods resulted in the lowest SE E ,
collectively. applicable to each method are listed in MARE, RMSE, and z-test values. All the
Table 3 (on page 24). methods demonstrated a high degree of
Flow-gauging weir X3H008 In Table 3, the various indirect SD association, with r2 values > 0.99; however,
Flow-gauging weir X3H008 is located in extension methods, except for the SE, this only highlights a high correlative
the Sand River, Mpumalanga Province. LE, and SAM-MA methods, provided trend between the estimated QEi values
The weir was constructed in 1967. As estimated discharge (QEi) values that as suggested by the various extension
shown in Figure 5, the structure consists exceeded the at-site benchmark discharges methods and the benchmark QBi values.
of a sharp-crest, hydro-flume, and broad- (QBi). The latter underestimations using Hence, the NSE results should also be
crested flank walls (Mathole 1999). The the SE and LE methods were evident for considered, whereas the VE-HRA and
structural limit is at a stage of 1.95 m, with stages (H) > 2 m (65% ≤ MARE ≤ 70%), VE-MA methods demonstrated the
an associated discharge of 76 m3/s. In the while the SAM-MA underestimations were lowest NSE values of ± 0.19 despite
latest DWS Calibration Report, the SBA much less significant, with MARE = 1.5%. their r2 values ≈ 0.99.
Table 2 Estimated discharge table 10 (DT10) at X3H008 based on the SBA (Nathanael et al 2018)
Stage increments (m) and discharge (m3/s)
Stage (m)
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
0 0.0 0.2 0.6 1.1 1.8 2.8 4.7 7.1 9.9 13.0
1 16.5 20.3 24.4 28.7 34.1 39.1 45.4 52.7 61.0 70.2
2 84.7 107.6 137.0 172.4 213.8 261.2 315.2 372.1 429.1 488.9
3 552.8 618.7 688.0 760.3 836 914 995 1 080 1 167 1 258
4 1 351 1 448 1 548 1 650 1 756 1 865 1 976 2 091 2 209 2 330
Journal of the South African Institution of Civil Engineering Volume 65 Number 2 June 2023 23
3 000
Benchmark
DT 8
DT 9
DT 10
2 500 SE
LE
VE-SA
VE-HRA
VE-MA
2 000 SAM-MA
SAM-CHZ
SBA-MA
Discharge (m3/s)
SBA-CHZ
HEC-RAS (FS)
1 500
1 000
500
0
0 1 2 3 4 5
Stage (m)
Figure 6 Indirect SD RC extensions in comparison to the benchmark rating at X3H008 (Williams 2023)
Average 319 271 1348 1809 1839 1217 1273 1235 1235 1236
Standard deviation 174 146 899 1436 1414 877 917 890 890 891
SEE (Eq (1); m3/s) 5.7 4.4 31.2 66.3 27.0 0.0 0.2 0.0 0.0 0.0
SEE ranking 7 6 9 10 8 3 5 2 1 4
MARE (Eq (2); %) 65.4 70.2 16.3 35.0 39.7 1.5 3.0 0.0 0.0 0.1
MARE ranking 9 10 6 7 8 4 5 2 1 3
RMSE (Eq (3)) 1156.5 1211.5 116.8 788.4 794.7 22.8 0.0 0.0 0.0 1.3
RMSE ranking 9 10 6 7 8 5 3 2 1 4
r 2 (Eq (4)) 0.999 0.999 0.999 0.998 1.000 1.000 1.000 1.000 1.000 1.000
r 2 ranking 8 7 9 10 6 4 5 2 1 3
NSE (Eq (5)) –0.74 –0.91 0.98 0.19 0.18 1.00 1.00 1.00 1.00 1.00
NSE ranking 9 10 6 7 8 5 3 2 1 4
z-test (Eq (6)) 28.1 30.0 2.7 11.9 12.6 0.4 0.9 0.0 0.0 0.0
z-test ranking 9 10 6 7 8 4 5 2 1 3
Sum of rankings 51 53 42 48 46 25 26 12 6 21
Overall ranking 9 10 6 8 7 4 5 2 1 3
*Benchmark method, MA (Manning’s approach), CHZ (Chézy’s approach), and FS (flood section)
24 Volume 65 Number 2 June 2023 Journal of the South African Institution of Civil Engineering
Table 4 Summary of the GOF-based rankings at the 10 flow-gauging sites (Williams 2023)
VE SAM SBA HEC-RAS
Gauging site SE LE
SA HRA MA MA CHZ MA CHZ (FS)
A4H005 4 5 3 1 2 - - - - -
A6H035 4 1 2 3 5 - - - - -
C2H003 3 1 5 4 2 - - - - -
C5H014 10 9 7 8 6 4 5 2 1 3
H5H004 10 9 7 8 6 4 5 2 1 3
J1H018 9 8 6 7 10 4 3 2 1 5
J4H002 7 9 5 8 10 6 4 2 1 3
K2H002 1 2 3 4 5 - - - - -
V1H026 1 2 3 5 4 - - - - -
X3H008 9 10 6 8 7 4 5 2 1 3
Avg rankings 5.8 5.6 4.7 5.6 5.7 4.4 4.4 2 1.2 3.4
Overall ranking 10 8 6 7 9 5 4 2 1 3
Summary of results the SE method at all the flow-gauging sites well with the SBA. By considering the
The overall ranking results, as listed in under consideration, except at A4H005, individual GOF statistics, the SAM-CHZ
Table 4, highlight that the SBA, SAM, J4H002, K2H002, V1H026, and X3H008. and SAM-MA methods also proved to
and 1-D HEC-RAS steady flow modelling However, this method is not recom- be equally accurate at flow-gauging sites
performed the best. The other indirect mended for significant extensions above C5H014, H5H004, and X3H008 by being
extension methods were characterised by the structural limit; typically, in these interchangeably ranked at either the 4th or
larger statistical differences between the above-structure-limit ranges, it proved 5th position.
at-site benchmark values (QBi) and the to result in estimated discharges (QEi)
modelled values (QEi). The VE-MA and SE significantly different from the benchmark SBA performance
methods are regarded as the least appropri- discharges (QBi). The SBA-CHZ is ranked overall in the 1st
ate methods, ranked respectively in the 9th position, while the SBA-MA is ranked in
and 10th positions. VE performance the 2nd position. However, the SBA was
The VE-SA (6th position) and VE-HRA also used as the preferred benchmark
SE performance (7th position) methods performed better method at 50% of the flow-gauging sites
The SE method is ranked overall in the than the VE-MA (9th position) method; under consideration. The latter preferential
10th position. The method tends to either however, the application of all these meth- use and associated rankings were also
over- or underestimate the at-site bench- ods was limited by data availability. The VE confirmed by Roux et al (2015): “From
mark discharges (QBi). At K2H002 and methods require a complete cross-section experience, DWS Flood Studies regard the
V1H026 it provided satisfactory results due of the weir at high flows, and such data SBA method as superior to the SAM when
to no changes in the geometry or hydraulic sets were not available at most of the sites flood peaks need to be estimated.” The
conditions above the structural limit. This under consideration, except at A4H005 latter statement is also supported in inter-
method is not suited to extend SD relation- where these methods were ranked as the national literature. The SBA is also less
ships at sites with complex hydraulics or top three methods. Hence, based on the affected by hydraulic conditions than the
geometry, and the use thereof should be promising results achieved at some of SAM, and is therefore regarded as a more
limited to flow-gauging sites with adequate the flow-gauging sites, the VE methods versatile method.
data and reasonably simple site charac- should be considered more often when SD The Chézy’s absolute roughness (ks)
teristics and hydraulics. It demonstrated relationships need to be extended. The version of the SBA, i.e. SBA-CHZ, is
the poorest performances at C5H014 and VE methods are also simpler to apply as preferred to the Manning’s n-value version
H5H004. opposed to the often-preferred SAM and (SBA-MA). This is ascribed to the fact that
SBA methods. Manning’s n-value is expressed in non-
LE performance measurable units, i.e. s/m1/3, and is con-
The LE method is ranked overall in the 8th SAM performance stant for a given discharge associated with
position. The method tends to either over- The SAM-CHZ is ranked overall in the a specific velocity and hydraulic radius.
or underestimate the at-site benchmark 4th position, while the SAM-MA is ranked The Chézy equation makes allowance
discharges (QBi). At A6H035 there are no in the 5th position. In comparison to the for changes in the roughness coefficient
apparent transitions beyond or above the SE, LE, and VE methods, it is hydraulically C (m0.5/s), as changes in the flow depth
structural limit; hence, the subsequent more correct. At J1H018 and J4H002, the and subsequently the hydraulic radius (R)
number one ranking at this site. The LE SAM-CHZ proved to be hydraulically more occur. As a result, channel profile conver-
method performed marginally better than correct (3rd to 4th position) and agreed gence is normally achieved sooner when
Journal of the South African Institution of Civil Engineering Volume 65 Number 2 June 2023 25
Chézy’s C-value (function of ks and R) is significantly. Typically, owing to insuf- conditions during high discharges at a
used in the SBA. Manning’s n-value does ficient data being available, no HEC-RAS flow-gauging site. In general, any extension
vary with changes in the hydraulic radius; modelling was possible at flow-gauging method must be hydraulically correct if
however, there is no explicit mathematical sites A4H005, A6H035, C2H003, K2H002 it is to be used as a robust approach to
procedure to determine this. Additionally, and V1H026. At the latter flow-gauging extend SD RCs beyond the structural limit.
it is also easier to visualise Chézy’s absolute sites, only calibration data up to the The extension of an RC is significantly
roughness (ks) as it represents the sizes structural limit was available, subsequently more affected by the site and river reach
of irregularities on the channel beds and resulting in discontinued flow at above- geometry, initial hydraulic conditions,
sides, expressed in metres. structure-limit conditions. In applying flow regimes, and level of submergence at
It is also important to note that the HEC-RAS at data deficient flow-gauging high discharges, than the actual extension
starting water surface level or elevation sites, overestimations will typically occur method used.
in the SBA is regarded as unknown or up to the structural limit, while above- Hence, there is no one-size-fits-all
indefinite; hence, several backwater profiles structure-limit flows will either be under- approach available for the extension of SD
based on different arbitrary water surface or overestimated. RCs in South Africa. For the interim, as rec-
levels should be computed until conver- ommended by Gericke and Smithers (2017),
gence occurs, i.e. the water levels cor- individual stage extensions (HE ≤ 1.2 H
respond with the uniform flow profile and CONCLUSIONS subjected to QDE ≤ 0.05 QDxi) should be
energy levels are in equilibrium. Therefore, The overall aim of this paper was to assess considered for bankfull or above bankfull
if convergence is not reached (probably a selection of indirect hydraulic exten- flow conditions in cases where site-specific
owing to insufficient cross-sectional survey sion methods against direct extension data is unavailable. It is also recommended
data), additional and more distant (up- and/ (benchmark) methods at selected flow- that the river geometry for a large sample of
or downstream) cross-sections would be gauging sites in South Africa. In terms of representative flow-gaugings in South Africa
required. At some of the flow-gauging sites the benchmark methods, the following be investigated to establish which simplified
under consideration, DWS had already approaches were used at the various flow- profile, e.g. trapezoidal cross-section with
established an arbitrary datum (reference gauging sites: (i) dam spillage (K2H002), 1:2 side slopes, could render acceptable dis-
level), and known water surface levels (e.g. (ii) hydrograph analysis (V1H026), (iii) level charges for bankfull and/or above bankfull
flood marks). Subsequently, defining an pool (back) routing (A4H005, A6H035, and flow conditions.
arbitrary point to reach profile convergence C2H130), and (iv) SBA (C5H014, H5H004,
was not required in these cases. J1H018, J4H002, and X3H008).
In assigning roughness parameters, The SBA-CHZ was ranked overall in ACKNOWLEDGEMENTS
i.e. Manning’s n-values, it should not be the 1st position, while the SBA-MA was Support for this research by the Central
limited to a cross-section, but it should ranked in the 2nd position. The SBA was University of Technology Free State (CUT)
rather be representative of the entire river used as the preferred benchmark method and the Water Research Commission
reach under consideration (George et al at 50% of the flow-gauging sites under (WRC) is gratefully acknowledged. We
1989). Observations made during this consideration, given that the SBA is less also wish to thank Mr Danie van der Spuy
study confirmed the difficulty in assigning affected by hydraulic conditions than the (former Specialist Engineer, Directorate
Manning’s n-values for a river reach and SAM and is therefore regarded as a more National Hydrological Services: Flood
the cross-sections under consideration. versatile method. The HEC-RAS steady Studies) for his expert and technical
Hence, the combined influence of various flow modelling was ranked overall in the advice pertaining to the extension of
input parameters (e.g. flow depth, turbu- 3rd position. In principle, the HEC-RAS stage-discharge relationships, as well as the
lence, bed forms, vegetation, sedimenta- steady flow modelling is also based on the anonymous reviewers for their constructive
tion and bed load, debris, obstructions, SBA; hence, the latter performance was review comments which have helped to
cross-section shape, geomorphological expected. However, the steady flow model- significantly improve the paper.
processes, and floodplain-main channel ling was in general limited by data suitabil-
interactions) needs to be considered when ity and availability. It was evident that the
assigning n-values during high flows HEC-RAS results improved as the number REFERENCES
(Lumbroso & Gaume 2012). Irrespective of of more detailed and distant cross-sections BSI ISO (British Standards Institution & International
the method used – e.g. SAM, SBA and/or increased. The SAM-CHZ and SAM-MA Organisation for Standardisation) 1998. BSI ISO
HEC-RAS – if representative and correct were ranked overall in the 4th and 5th 1100-2. Measurement of Liquid Flow in Open
n-values are assigned, comparable results position, respectively. In comparison to the Channels: Determination of the Stage-Discharge
will be achieved. SE, LE, and VE methods, the SAM proved Relation. London: BSI.
to be hydraulically more correct. However, Chow, V T 1959. Open-channel Hydraulics. New York:
HEC-RAS performance the performance of the SAM is highly McGraw-Hill.
It was evident that the HEC-RAS results dependent on detailed cross-sections, while Dymond, J R & Christian, R 1982. Accuracy of
improved as the number of more detailed roughness coefficients for between-reaches discharge determined from a rating curve.
and distant cross-sections increased. are also required. Hydrological Sciences Journal, 27(4): 493–504.
Therefore, by having a continuous surface, The SBA, SAM, and 1-D HEC-RAS doi: 10.1080/02626668209491128.
i.e. digital terrain model, from which high- steady flow modelling are therefore EA UK (Environmental Agency United Kingdom)
accuracy cross-sections can be extracted, regarded as the most appropriate indirect 2002. Stage 1: Extension of rating curves at gauging
the modelling results can be improved estimation methods to reflect the hydraulic stations. Report No 1. Bristol: EA UK.
26 Volume 65 Number 2 June 2023 Journal of the South African Institution of Civil Engineering
EA UK 2003. Stage 2: Extension of rating curves at Maeder, C 2015. The road to HEC-RAS. Middleton, Riet River at gauging weir C5H014. Report No
gauging stations using simple hydraulic techniques. WI: Civil GEO, US. Available at: https://www. C500-H014-2015.03. Pretoria: Department of Water
Report No 2. Bristol: EA UK. civilgeo.com/blog/the-road-to-HEC-RAS. and Sanitation.
George, J, Arcement, J R & Schneider, V R 1989. Guide Mathole, J 1999. Kalibrasieversoek (X3H008). Report Shao, Q, Dutta, D, Karim, F & Petheram, C 2018. A
for selecting Manning’s roughness coefficients for No DT5. Pretoria: Department of Water Affairs and method for extending stage-discharge relationships
natural and flood plains. Water Supply Paper 2339. Forestry. using a hydrodynamic model and quantifying
US Geological Survey. Denver. CO: Department of Menné, T C 1960. The role of hydrology in planning the the associated uncertainty. Journal of Hydrology,
the Interior. Available at: https://pubs.usgs.gov/ development of a country. Hydrological Research 556(2018): 154–172.
wsp/2339/report.pdf. Technical Report No 9. Pretoria: Department of Smithers, J C 2022. The initiation of a National Flood
Gericke, O J & Smithers, J C 2017. Direct estimation Water Affairs. Studies Programme for South Africa: Overview
of catchment response time parameters in medium Nathanael, J J 2016. Assessing the performance of and progress to date. Paper presented at the IAHS
to large catchments using observed streamflow regional flood frequency analysis methods in 2022 11th Scientific Assembly. Montpellier, France.
data. Hydrological Processes, 31(5):1125–1143. South Africa. MSc Dissertation. Pietermaritzburg: doi: 10.5194/iahs2022-229.
doi: 10.1002/hyp.11102. University of KwaZulu-Natal. USACE (United States Army Corps of Engineers)
Hardy, T, Panja, P & Mathias, P 2005. WinXSPRO: A Nathanael, J J, Rademeyer, P F & Van der Spuy, D 2018. 2016. HEC-RAS river analysis system user’s manual.
channel cross-section analyser. Lakewood, CO: US Discharge rating improvement of rating table and Davis, CA: Hydrologic Engineering Centre.
Department of Agriculture, Forest Service. high flow data Sand River at Exeter. Report No Wessels, P & Rooseboom, A 2009. Flow-gauging
Lang, M, Pobanz, K, Renard, B, Renouf, E & Sauquet, X3H008DT-2018.12. Pretoria: Department of Water structures in South African rivers. Part 1: An
E 2010. Extension of rating curves by hydraulic and Sanitation. overview. Water SA, 35(1): 1–10.
modelling with application to flood frequency analysis. Petersen-Øverleir, A & Reitan, T 2009. Accounting Williams, V H 2023. Assessment of indirect estimation
Hydrological Sciences Journal, 55(6): 883–898. https:// for rating curve imprecision in flood frequency methods to extend observed stage-discharge
doi.org/10.1080/02626667.2010.504186. analysis using likelihood-based methods. Journal of relationships for above-structure-limit conditions
Lindner, G A & Miller, A J 2012. Numerical Hydrology, 366(1–4): 89–100. at flow-gauging weirs. MEng Dissertation.
modelling of stage-discharge relationships in Rademeyer, P 2023. Personal communication. Pretoria: Bloemfontein: Central University of Technology,
urban streams Journal of Hydrological Engineering, Directorate: National Hydrological Services, Free State.
17: 590–596. Department of Water and Sanitation. Zhong, X & Dutta, U 2015. Engaging Nash-Sutcliffe
Lumbroso, D & Gaume, E 2012. Reducing the Ramsbottom, D M & Whitlow, C D 2003. Extension of efficiency and Model Efficiency Factor Indicators
uncertainty in indirect estimates of extreme rating curves at gauging stations: Best practice guidance in selecting and validating effective light rail
flash flood discharges. Journal of Hydrology, manual. Bristol, UK: Environmental Agency. system operation and maintenance cost models.
414–415: 16–30. https://doi.org/10.1016/j. Roux M, Rademeyer, P F & Van der Spuy, D 2015. Journal of Traffic & Transportation Engineering,
jhydrol.2011.08.048. Discharge rating improvement of rating table for 3(2015): 255–265.
Journal of the South African Institution of Civil Engineering Volume 65 Number 2 June 2023 27
TECHNICAL PAPER
Advanced numerical
JOURNAL OF THE SOUTH AFRICAN
INSTITUTION OF CIVIL ENGINEERING modelling of the nonlinear
mechanical behaviour
ISSN 1021-2019 (print) | ISSN 2309-8775 (online)
Vol 65 No 2, June 2023, Pages 28–38, Paper 1492
unsaturated clay
models of a singular pile embedded in soil,
where formulas using AI (artificial intelligence)
and machine learning algorithms will be able to
predict the failure force and horizontal displacement of a laterally loaded pile.
Contact details:
Department of Civil Engineering, University of Pretoria
Private Bag X20, Hatfield, Pretoria 0028, South Africa K T Braun, N Bakas, G Markou, S W Jacobsz
E: u17031215@tuks.co.za
Keywords: soil-structure interaction, reinforced concrete pile, soil Braun KT, Bakas N, Markou G, Jacobsz SW. Advanced numerical modelling of the nonlinear mechanical behaviour of a laterally loaded pile
embedded in stiff unsaturated clay. J. S. Afr. Inst. Civ. Eng. 2023:65(2), Art. #1492, 11 pages. http://dx.doi.org/10.17159/2309-8775/2023/v65n2a4
domain, mechanical behaviour, finite element method
The Journal of the South African Institution of Civil Engineering, which is distributed internationally, is a peer-reviewed, open-access journal licensed under a Creative
Commons Attribution Licence (CC BY-NC-ND). Readers may therefore freely use and share the content as long as they credit the original creator and publisher, do
28 not change the material in any way, and do not use it commercially. Copyright of this article remains with the authors.
σ’a σ’a
F Y
σ’f σ’y
G
σ’g
G
σ’g
Y F
σ’y σ’f
εp εe
–εe
εp
R B
0 εa 0 εa
εp εp
(a) (b)
Figure 1 S train hardening (a) and strain softening (b) (adapted from Atkinson 2017)
be combined into a single model known as homogeneous conditions are assumed. Soil reinforced concrete (RC) buildings was
the critical state model. Atkinson (2017) heterogeneity also induces stresses which investigated, which included SSI effects. In
proves that by considering the stress-strain are dependent on the soil variability, the their research work, a digitally generated
relationship of soils, they can typically soil-structure stiffness ratio, and the soil- dataset was used to derive a closed-form
behave in three different ways, namely: structure length ratio (soil fluctuation scale formula that was able to accurately predict
elastic-perfectly plastic, elastic-plastic and structural dimensions). the fundamental period of RC buildings
hardening, and elastic-plastic softening. A According to Kausel (2010), research with and without accounting for the SSI
perfectly plastic material continues to strain work considering the early history of SSI effect. A similar approach that will foresee
with constant stress, and therefore plastic describes the SSI phenomenon as some- the use of artificial intelligence (AI) and
deformation occurs during plastic deforma- what difficult to define. This was because machine learning algorithms will be used
tion only. Hence, it does not experience of the complex dynamic interactions due to by the authors after the successful delivery
elastic deformation under plastic strain. An the amplification of seismic waves present of the research work discussed herein.
elastic-plastic material that strain hardens within the soil layers. Another research The finite element method (FEM) is
experiences elastic strain during plastic work analysed the non-linear behaviour of one of the most commonly used numeri-
deformation, while an elastic-plastic mate- laterally loaded piles (Kavitha et al 2016), cal approaches in industry, as well as for
rial that strain softens releases elastic strain where they investigated the soil-pile inter- research purposes. This method aims
during plastic deformation (Atkinson 2017). action effect when loaded laterally. It was to predict the mechanical behaviour of
Figure 1 represents the stress-strain curves reported that the soil-pile interaction effect structures, allowing and assisting engineers
for an elastic-plastic strain hardening and depended on the soil properties, loading to numerically investigate structures.
softening material, respectively. type, ground slope, as well as the pile mate- For this reason, the FEM-based software
Jardine et al (1986) studied the influ- rial and diameter (Kavitha et al 2016). (ReconanFEA 2020) is used for the needs
ence of non-linear stress-strain experi- A study by Gravett and Markou (2021) of this research work. Furthermore, by
enced within soil-structure interaction performed a state-of-the-art investigation analysing numerical models under differ-
(SSI) problems. It was stated that from on wind turbine structures founded on soft ent loading conditions and using a variety
field and laboratory studies, at very small clay, where SSI was also considered. This of meshes, it can be determined if the
strains, most soils experience a non-linear work investigated the interaction of soil structural design is safe or not (Markou &
stress-strain behaviour. It was therefore and structural mechanics, and foresaw the Roeloffze 2021). Modelling of piles embed-
concluded that, despite the treatment of derivation of an optimum pile foundation ded in soil was performed by Gravett and
soil as a linear elastic material in SSI prob- geometry based on modal and ultimate Markou (2021), where experimental and
lems, which represents an effective tool for limit nonlinear pushover analyses. Gravett numerical results showed that ReconanFEA
expressing soil stiffness if non-linearities and Markou (2021) used ReconanFEA (2020) has the ability to objectively capture
are not considered, the results can be mis- (2020) to perform the analyses in a detailed the mechanical response of this complex
leading. Breysse et al (2005) investigated 3D approach to discretise, model and ana- problem. Therefore, the modelling that was
the effect of soil heterogeneity within SSI lyse numerous finite element (FE) models. conducted herein to investigate the behav-
problems. Breysse et al (2005) showed that They reported that the use of hexahedral iour of piles subjected to different loading
soil heterogeneity induces changes in the finite elements was ideal for this type of conditions follows a similar approach,
differential settlement, bending moments, SSI problem. It is noteworthy to refer to adopting detailed 3D modelling of solid
stresses, and possibly cracking within the the recent research work by Gravett et al elements to discretise both the concrete
SSI problem which cannot be predicted if (2021), where the fundamental period of and soil domains.
Journal of the South African Institution of Civil Engineering Volume 65 Number 2 June 2023 29
A research project which aimed to inves-
tigate the behaviour of piled foundations, σ2
particularly for wind turbines in partially
saturated clay soils, was recently undertaken
in South Africa. Expansive soils are located σy
in many sub-arid and sub-humid parts
across the African continent, and therefore
the design of superstructures within or on
expansive soils is critical for the future of
the socio-economy of Africa. Field tests
were carried out on a site near Vredefort
in South Africa. This site was underlain by
σ1
expansive clay which extends to depths in
excess of 7 m, with bedrock occurring at –σy σy
about 15 m. Two sets of lateral pile load tests
were conducted with the soil in a dry (in-
situ) state, as well as after saturating the soil.
The objective of this study was to use
advanced numerical models to analyse the
nonlinear mechanical behaviour of RC piles
–σy
embedded in clay by replicating in-situ lateral
load tests. Concrete, steel and soil properties
were modelled in the numerical analysis per-
formed herein according to the material tests
performed for the purposes of the physical Figure 2 V
on Mises 2D yielding criterion curve (adapted from Gravett & Markou 2021)
experiment. However, for unknown proper-
ties, specific engineering judgement had by Gravett and Markou (2021) describes The formulation that describes the
to be used to obtain representative values, how the corresponding hydrostatic and strength envelope of concrete (Markou et
especially in the case of the soil. Thereafter, deviatoric strains (ε 0 , γ 0) are used to pro- al 2018) is expressed as:
a sensitivity analysis was conducted on these duce the hydrostatic and deviatoric stress-
2 – τ 2 )cos θ + τ (2τ – τ )
2τ0c(τ0c
material properties to determine how sensi- es, by using a material model proposed by 0e 0c 0e 0c
2 – τ 2 )cos2θ + 5τ 2 – 4τ 2 τ 2
√4(τ0c
tive the assumptions of property values were Kotsovos and Pavlovic (1995) and extended 0e 0e 0c 0e
τ0u = (4)
in terms of the analysis. For this purpose, by Markou and Papadrakakis (2013). 2 – τ 2 )cos2θ + (2τ – τ )2
4(τ0c 0e 0e 0c
numerical software packages such as FEMAP
(Ouanta et al 2016) were used. By develop- σ 0 + σ id τ0 This expression represents a smooth con-
ε 0 = ε 0h + ε 0d = , γ0 = (1)
ing models through FEMAP and analysing 3KS 2GS vex curve, where θ is the rotational angle
them through the use of the ReconanFEA between the deviatoric stress vector and one
software package, while considering the input In Equation 1, σid expressed as a function of the projected stress principal axis located
parameters measured from the full-scale RC of (σ 0 , τ0 , fc) is an equivalent external stress, on the deviatoric plane. In addition, once the
pile load research project in Vredefort, South which accounts for the coupling effect failure criteria is satisfied, the smeared crack
Africa, correlations between the models between τ0 and ε 0d (Gravett & Markou approach is imitated where macro-cracking
and real-life experiments were performed. 2021). In addition, the bulk modulus KS (σ 0 , occurs (Gravett & Markou 2021).
Additionally, even though the physical τ0) and the shear modulus GS (σ 0 , τ0) are In terms of reinforcement modelling,
experiment did not proceed to ultimate obtained by ignoring σid. By considering the Menegotto-Pinto method, which
failure, it is shown herein that the numerical the relationship in Equation 1, the strains in also considers the Bauschinger effect,
simulations provided much missing informa- global coordinates are expressed as follows: is integrated within the ReconanFEA
tion, complementing the physical experiment software (Menegotto 1973). The stress-
by answering questions such as revealing the σ ij + σ idδij 3vS strain relationship for the steel material is
εij = – (σ 0 + σ id)δij(2)
failure mechanism, as well as the ultimate 2GS ES presented as:
lateral load at which the RC pile failed.
Where: δij is the Kronecker delta, σ ij and εij (1 – b)ε*
σ* = bε* + (5)
are the total stresses and strains, respec- (1 + ε*R)1/R
MODELLING OF CONCRETE tively. ES (σ 0 , τ 0 , fc) and νS (σ 0 , τ 0 , fc) are the
AND SOIL MATERIAL secant Young modulus and Poisson’s ratio, Where:
Modelling in 3D considers the effect of respectively, which are derived from KS and
out-of-plane stresses, where the stress- GS (Markou & Roeloffze 2021). ES and νS ε – εr
ε* = ,(6)
strain relationship for each stress-state is are expressed as: ε 0 – εr
expressed by the hydrostatic (σ 0) and devia-
toric components (τ 0) (Markou & Roeloffze 9KSGS 3KS – 2GS σ – σr
ES = , vS = (3) σ* = ,(7)
2021). The recent research work presented 3KS + GS 6KS + 2GS σ0 – σr
30 Volume 65 Number 2 June 2023 Journal of the South African Institution of Civil Engineering
discretised with 8-noded isoparametric
Concrete pile hexahedral elements. This first model (see
Soil domain Rebars and stirrups
Applied load
Figure 3) assumed a full bond between
the concrete-soil interface. The reinforce-
ment of the pile (12Ø25 longitudinal and
Ø10@150 stirrups) was also discretised and
analysed in this study through the use of
embedded rod elements that can be seen in
Figure 3.
Given that the experiment foresaw the
application of static loads, the soil domain
was fixed at the bottom, and the nodes
along the perimeter of the soil domain
Figure 3 F inite element mesh of model 1 (left), load application (middle) and reinforcement were restrained from displacing horizon-
details (right) tally, while allowed to undergo settlement
(z-global axis). Figure 3 also shows the
point of load application that was chosen
according to the field experiment (0.5 m
above the soil surface). The total applied
load within the model was divided into
3m
Journal of the South African Institution of Civil Engineering Volume 65 Number 2 June 2023 31
Table 2 C
oncrete (grey section) and steel
(green section) material input 800
parameters as determined from
700
standard material testing
600
Parameter Value Units
Poisson’s ratio 0.3 – Figure 5 “ Gap” model with the best fit – comparison between the numerical and experimental curves
Load (kN) Mesh deformation Solid Von Mises strains Load (kN) Mesh deformation Solid Von Mises strains
0.0005
0.000469
0.000438
0.0005
0.000406
0.000469
0.000375
0.000438
0.000344
0.000406
0.000313
0.000375
0.000281
0.000344
100 0.00025
0.000313
0.000219
0.000281
0.000188
0.000156 500 0.00025
0.000125 0.000219
0.0000938 0.000188
0.0000625 0.000156
0.0000313 0.000125
0. 0.0000938
0.0000625
0.0000313
0.
0.0005
0.000469
0.000438
0.000406
0.000375
0.000344
0.000313
0.000281
200 0.00025 0.0005
0.000469
0.000219
0.000188 0.000438
0.000156 0.000406
0.000125 0.000375
0.0000938 0.000344
0.0000625 0.000313
0.0000313 0.000281
0. 600 0.00025
0.000219
0.000188
0.000156
0.0005 0.000125
0.000469 0.0000938
0.000438 0.0000625
0.000406 0.0000313
0.
0.000375
0.000344
0.000313
0.000281
300 0.00025
0.000219
0.000188
0.000156
0.000125
0.0000938 0.0005
0.0000625 0.000469
0.0000313 0.000438
0. 0.000406
700 Plastic hinge
formation
0.000375
0.000344
(load 0.000313
0.000281
0.0005
0.000469 increment 0.00025
0.000219
0.000438
0.000406 prior to 0.000188
0.000375 0.000156
0.000344 failure) 0.000125
0.000313 0.0000938
0.000281 0.0000625
400 0.00025 0.0000313
0.
0.000219
0.000188
0.000156
0.000125
0.0000938
0.0000625
0.0000313
0.
Figure 6 “ Gap” model with the best fit – deformed shapes and Von
Mises strain contours prior to failure
32 Volume 65 Number 2 June 2023 Journal of the South African Institution of Civil Engineering
600 600
500 500
Horizontal load (kN)
300 300
200 200
100 100
0 0
0 10 20 30 40 50 60 0 10 20 30 40 50 60
(a) Horizontal displacement (mm) (d) Horizontal displacement (mm)
Experimental E =50 MPa, CS = 89 kPa E =50 MPa, CS = 100 kPa E =50 MPa, CS = 110 kPa
E =50 MPa, CS = 120 kPa E =50 MPa, CS = 130 kPa E =50 MPa, CS = 140 kPa E =50 MPa, CS = 150 kPa
600 600
500 500
Horizontal load (kN)
300 300
200 200
100 100
0 0
0 10 20 30 40 50 60 0 10 20 30 40 50 60
(b) Horizontal displacement (mm) (e) Horizontal displacement (mm)
Experimental E =100 MPa, CS = 89 kPa E =100 MPa, CS = 100 kPa E =100 MPa, CS = 110 kPa
E =100 MPa, CS = 120 kPa E =100 MPa, CS = 130 kPa E =100 MPa, CS = 140 kPa E =100 MPa, CS = 150 kPa
600 600
500 500
Horizontal load (kN)
400 400
300 300
200 200
100 100
0 0
0 10 20 30 40 50 60 0 10 20 30 40 50 60
(c) Horizontal displacement (mm) (f) Horizontal displacement (mm)
Experimental E =150 MPa, CS = 89 kPa E =150 MPa, CS = 100 kPa E =150 MPa, CS = 110 kPa
E =150 MPa, CS = 120 kPa E =150 MPa, CS = 130 kPa E =150 MPa, CS = 140 kPa E =150 MPa, CS = 150 kPa
Figure 7 H
orizontal load versus horizontal displacement for different soil material combinations: (a–c) “No Gap” and (d–f) “Gap” models
to 500 MPa, an average value derived from program. Both the “No Gap” and “Gap” terminated. Consequently, neither the RC
experimental tests. models were analysed under the same pile nor the soil reached a full failure state,
loading conditions, while the nonlinear even though the deformation was larger
analysis algorithm and the corresponding than the allowable limit according to inter-
NUMERICAL INVESTIGATION parameters were kept the same. national design codes. For this reason, the
For the purpose of this study, static non- During the physical experiment, the numerical models engaged in this research
linear analyses were performed by applying maximum imposed load was 500 kN, study were also used to predict the physi-
a lateral load until failure. The numerical which was the load level at which the cal failure mode of the specimen and the
results obtained were compared to the data hydraulic jack reached its maximum exten- ultimate capacity that the embedded RC
obtained from the field experimentation sion; therefore, the experiment had to be pile could develop. From the numerical
Journal of the South African Institution of Civil Engineering Volume 65 Number 2 June 2023 33
results derived using the two numerical
models (“No Gap” and “Gap”), it was found
that the latter resulted in the best match to
the experimental response, i.e. where the
numerically derived lateral load-displace-
ment curve best matched the experimental
data (Figure 5 on p 32). The “Gap” model Z
100
in front of the RC pile specimen, indicating
a 3D compressive failure of the clay mate- 80
rial due to excessive compressive stresses.
60
This led to large horizontal deformations
that caused the failure of the RC pile’s lon- 40
gitudinal reinforcement, which was located 20
in the tensile region of the circular section.
0
As mentioned above, the physical experi- 0 20 40 60 80 100 120 140 160 180 200
ment was terminated at a total horizontal Strain 10 –6 (microstrain)
applied load of 500 kN. The numerical E =100 MPa, CS = 89 kPa E =100 MPa, CS = 120 kPa E =100 MPa, CS = 150 kPa
model predicted failure to occur at a hori- E =100 MPa, CS = 100 kPa E =100 MPa, CS = 130 kPa
zontally applied load of 700 kN, thus, the E =100 MPa, CS = 110 kPa E =100 MPa, CS = 140 kPa
model allowed a failure mechanism to be
observed. Figure 9 S tress-strain curves of the soil element in front of the pile
In order to further investigate the
numerical response of the models devel- stiffer response and hence less horizontal (yellow elements for the soil and red for the
oped, a sensitivity analysis was performed deformation. It can also be observed that pile). These elements are located in front
where the Young’s modulus (ES) for the the “No Gap” model gives a stiffer overall of the pile and represent the most critically
soil material was kept constant at values response attributed to the introduction of loaded elements in the numerical analysis.
of 50 MPa, 100 MPa and 150 MPa respec- additional stiffness at the interface of the Figures 9 and 10 represent the stress-strain
tively, while the compressive strength RC pile and the soil domain. Finally, it is plots for the respective soil and concrete
(CS) of the soil was varied from 89 kPa to evident that ReconanFEA (2020) was able elements determined from the nonlin-
150 kPa (see Table 1). In Figure 7 the total to successfully run all of the developed ear analysis. For these plots, the chosen
imposed horizontal load versus the hori- models without numerical instabilities; combination of a soil Young’s modulus of
zontal displacement of the pile can be seen therefore providing the required insight in 100 MPa and a compressive strength rang-
as derived from the nonlinear analyses. A relation to the local and global nonlinear ing from 89 kPa to 150 kPa was used for
comparison with the experimental curve is mechanical behaviour of the RC pile and the soil.
also depicted. the soil domain. Figure 9 represents the stress-strain
As can be seen from Figure 7 (on Soil and concrete stress and strain curve for the soil element in front of the
p 33), the stiffness of the pile and the soil curves for the most critically loaded ele- RC pile. It can be observed that each curve
structural system decreases with increasing ments in the model were developed to get reaches its compressive strength value at a
horizontal load, which is attributed to the a better understanding of the material different strain. The slope (E) of each curve
damage development in the two domains. response during load application. Figure 8 is relatively similar, whereafter plastic
In addition, an increase in compressive indicates the elements where the solid Von deformation occurs once each material has
strength of the soil material results in a Mises stress-strain plots were generated reached its ultimate compressive strength.
34 Volume 65 Number 2 June 2023 Journal of the South African Institution of Civil Engineering
Since neither strain hardening nor strain
10 000 softening occurs within the soil element,
9 000 the overall mechanical response of the soil
material is classified as an elastic-perfectly
8 000
plastic material.
7 000 Figure 10 represents the stress-strain
curves of the concrete element in front of
6 000
Stress (kPa)
20 PARAMETRIC INVESTIGATION
A sensitivity analysis was conducted on
the parameters which could not be explic-
itly measured in the physical experiment.
10 These parameters were the Young’s modu-
lus (E) and the compressive strength (CS)
of the soil. For this sensitivity analysis,
the “Gap” model with the same geometry,
0
materials, constraints, loading conditions
and nonlinear analysis was used. The best
fit between the numerical result and the
field observation resulted from using an
E value of 100 MPa and a compressive
–10
strength of 110 kPa. These values were
adopted as the base combination for the
sensitivity analysis. This combination
was used to determine how sensitive the
–20
critical parameters were to the horizontal
displacement of the model for different
load increments, which would also provide
insight on how these parameters affect
–30
–0.4 –0.3 –0.2 –0.1 0 0.1 0.2 0.3 0.4
the overall mechanical behaviour of
Change in compressive strength (CS) the model.
100 kN (Change CS) 400 kN (Change CS) 700 kN (Change CS) In order to illustrate the sensitivity of
200 kN (Change CS) 500 kN (Change CS) these parameters, the Young’s modulus
300 kN (Change CS) 600 kN (Change CS) was kept constant, while the compressive
strength was increased and decreased
Figure 11 S ensitivity of the model due to soil compressive strength in terms of horizontal displacement by 30% accordingly. Figure 11 shows the
Journal of the South African Institution of Civil Engineering Volume 65 Number 2 June 2023 35
sensitivity of the model in terms of com-
pressive soil strength and how the lateral 25
displacement of the pile varied under dif-
ferent lateral load increments, ranging from
20
100 kN to 700 kN. Next, the compressive
strength of the soil was kept constant,
36 Volume 65 Number 2 June 2023 Journal of the South African Institution of Civil Engineering
40
30
20
Change in horizontal displacement (%)
10
–10
–20
–30
–40 –30 –20 –10 0 10 20 30 40
Change in parameters (%)
100 kN (Change E) 500 kN (Change E) 100 kN (Change CS) 500 kN (Change CS)
200 kN (Change E) 600 kN (Change E) 200 kN (Change CS) 600 kN (Change CS)
300 kN (Change E) 700 kN (Change E) 300 kN (Change CS) 700 kN (Change CS)
400 kN (Change E) 400 kN (Change CS)
Figure 13 S ensitivity of the E and CS on the change in horizontal displacement of the model
machine hosted at the HPC National Gravett, D Z & Markou, G 2021. State-of-the-art Karthigeyan, S, Ramakrishna, V & Rajagopal, K 2006.
Competence Center of Cyprus at the investigation of wind turbine structures founded Influence of vertical load on the lateral response of piles
Cyprus Institute (Project pro21b103). on soft clay by considering the soil-foundation- in sand. Computers and Geotechnics, 33(2): 121–131.
structure interaction phenomenon: Optimization Kausel, E 2010. Early history of soil-structure
of battered RC piles. Engineering Structures, 235: interaction. Soil Dynamics and Earthquake
REFERENCES 112013. Engineering, 30(9): 822–832.
Atkinson, J 2017. The Mechanics of Soils and Gravett, D Z, Mourlas, C, Taljaard, V-L, Bakas, N, Kavitha, P, Beena, K & Narayanan, K 2016. A review on
Foundations. Boca Raton, FL: CRC Press. Markou, G & Papadrakakis, M 2021. New soil-structure interaction analysis of laterally loaded
Breysse, D, Niandou, H, Elachachi, S & Houy, L 2005. fundamental period formulae for soil-reinforced piles. Innovative Infrastructure Solutions, 1(1): 1–15.
A generic approach to soil-structure interaction concrete structures interaction using machine Kotsovos, M D & Pavlovic, M N 1995. Structural
considering the effects of soil heterogeneity. learning algorithms and ANNs. Soil Dynamics and Concrete: Finite-element Analysis for Limit-state
Geotechnique, 55(2): 143–150. Earthquake Engineering, 144: 106656. Design. London: Thomas Telford.
Červenka, J & Papanikolaou, V K 2008. Three- Jardine, R, Potts, D, Fourie, A & Burland, J 1986. Markou, G, Alhamaydeh, M & Saadi, D 2018. Effects
dimensional combined fracture-plastic material Studies of the influence of non-linear stress- of the soil-structure-interaction phenomenon on
model for concrete. International Journal of strain characteristics in soil-structure interaction. RC structures with pile foundations. Proceedings,
Plasticity, 24(12): 2192–2220. Geotechnique, 36(3): 377–396. 9th Greek Association of Computational Mechanics
Journal of the South African Institution of Civil Engineering Volume 65 Number 2 June 2023 37
(GRACM) International Congress on Computational Menegotto, M 1973. Method of analysis for cyclically M, Panait, C T, Tamas, R, Caruntu, G & Cristea,
Mechanics. 4–6 June 2018, Chania, Greece. loaded RC plane frames including changes in I (Eds). Advanced Topics in Optoelectronics,
Markou, G & Papadrakakis, M 2013. Computationally geometry and non-elastic behavior of elements Microelectronics, and Nanotechnologies
efficient 3D finite element modeling of RC structures. under combined normal force and bending. (Conference). Vol 10010. 5–28 August 2016,
Computers and Concrete, 12(4): 443–498. Proceedings, International Association for Bridge Constanta, Romania. Bellingham, WA: SPIE
Markou, G & Roeloffze, W 2021. Finite element and Structural Engineering (IABSE) Symposium on Publications, p 1001020.
modelling of plain and reinforced concrete Resistance and Ultimate Deformability of Structures ReconanFEA v2.00. User’s Manual, 2020.
specimens with the Kotsovos and Pavlovic material Acted on by Well Defined Repeated Loads. Lisbon, Available at: https://www.researchgate.net/
model, smeared crack approach and fine meshes. Portugal, pp 15–22. publication/342361609_ReConAn_v200_Finite_
International Journal of Damage Mechanics, 30(6): Oanta, E M, Panait, C & Raicu, A 2016. Original data Element_Analysis_Software_User%27s_Manual.
845–871. doi: 10.1177/1056789520986601. preprocessor for Femap/Nastran. In Vladescu,
38 Volume 65 Number 2 June 2023 Journal of the South African Institution of Civil Engineering
TECHNICAL NOTE
Comparison of sinkhole
size versus borehole
JOURNAL OF THE SOUTH AFRICAN
INSTITUTION OF CIVIL ENGINEERING
hazard classification
ISSN 1021-2019 (print) | ISSN 2309-8775 (online)
Vol 65 No 2, June 2023, Pages 39–43, Note 1578
Bunk J. Comparison of sinkhole size versus borehole hazard classification. Keywords: dolomite, sinkhole, hazard comparison, borehole
J. S. Afr. Inst. Civ. Eng. 2023:65(2), Art. #1578, 5 pages. http://dx.doi.org/10.17159/2309-8775/2023/v65n2a5
The Journal of the South African Institution of Civil Engineering, which is distributed internationally, is a peer-reviewed, open-access journal licensed under a Creative
Commons Attribution Licence (CC BY-NC-ND). Readers may therefore freely use and share the content as long as they credit the original creator and publisher, do
not change the material in any way, and do not use it commercially. Copyright of this article remains with the authors. 39
The manifestation of a sinkhole or sub- Table 1 Definition of the eight inherent hazard classes (after Buttrick et al 2001)
sidence at ground surface does not always Inherent
define the depth and lateral extent of an hazard Characterisation of area
instability (Kleinhans & Van Rooy 2016). class (IHC)
The receptacle at depth that accommodates Areas characterised as reflecting a low inherent hazard of sinkhole and subsidence
Class 1
the eroded material may not be directly formation (all sizes) with respect to ingress of water
below the area of visual impact on the sur-
Areas characterised as reflecting up to a medium inherent hazard of small-sized
face, but can be located at distances more Class 2
sinkhole (< 2 m) and subsidence formation with respect to ingress of water
than 20 m from the sinkhole or subsidence
Areas characterised as reflecting up to a medium inherent hazard of medium-sized
manifestation at surface. The evaluation of Class 3
sinkhole (2–5 m) and subsidence formation with respect to ingress of water
an event should therefore consider covering
a surface area of between 20 m to 50 m Areas characterised as reflecting up to a medium inherent hazard of large-sized
Class 4
sinkhole (5–15 m) and subsidence formation with respect to ingress of water
around the sinkhole or subsidence area to
ensure that the area of influence has been Areas characterised as reflecting up to a high inherent hazard of small-sized sinkhole
Class 5
covered (Kleinhans & Van Rooy 2016). (< 2 m) and subsidence (all sizes) formation with respect to ingress of water
18%
Tshwane Ekurhuleni
267 sinkholes 31% 106 sinkholes
42%
Overestimated Overestimated
57% 24% Correct Correct
Underestimated Underestimated
27%
Figure 1 C
omparison of IHC values of boreholes versus IHC values of actual sinkhole volumes, i.e. borehole IHC is either overestimating or
underestimating actual sinkhole volume
40 Volume 65 Number 2 June 2023 Journal of the South African Institution of Civil Engineering
Tshwane, 267 sinkholes Ekurhuleni, 106 sinkholes
70 35
60 30
Number of sinkholes
Number of sinkholes
50 25
40 20
30 15
20 10
10 5
0 0
–3 –2 –1 0 1 2 3 4 5 6 –3 –2 –1 0 1 2 3 4 5 6 7
IHC sinkhole compared to IHC borehole IHC sinkhole compared to IHC borehole
Number of sinkholes
5
25
4
20
3
15
2
10
5 1
0 0
–3 –2 –1 0 1 2 3 4 5 6 7 –2 –1 0 1 2 3
IHC sinkhole compared to IHC borehole IHC sinkhole compared to IHC borehole
Figure 2 D
ifference between IHC calculated from sinkhole volume and classified from borehole log (thus IHC sinkhole value minus IHC borehole value)
5 171 sinkhole entries in the database, but assigned IHC value based on the volume was made to 25% of the classifications of
only 546 have been used in this study. The differed less from the IHC of the borehole the boreholes in the database. For the rest
limiting factors were dimensions that were than the assigned IHC based on the diam- of the classifications, only the borehole log
not available for all entries, and the fact eter. For this study, only the IHC based on was taken into account and not the general
that a borehole had to be present within the volume of the sinkhole was taken into geological conditions of the area.
50 m from the sinkhole. consideration. It was further assumed that The groundwater table was incorporated
The sinkhole database includes all the area around the sinkhole reflected a in the classifications as per the method pro-
sinkholes in South Africa and is subdivided high inherent hazard, thus IHC values of 5 posed by Oosthuizen and Van Rooy (2015).
into different municipalities or regions, to 8 were assigned to the sinkholes. The comparison in this paper considered the
namely the Far West Rand region, Tshwane, The comparison of the IHC values was ingress scenario of the boreholes and did not
Ekurhuleni, Johannesburg and Sedibeng thus undertaken from two perspectives: take the dewatering scenario into account.
Municipalities and the Mpumalanga, North (a) to determine whether the hazard has
West and Northern Cape Provinces. Four been accurately predicted (thus IHC values
regions were analysed: the Far West Rand, of 5 to 8 and not IHC values less than 5), COMPARISON
Tshwane and Ekurhuleni Municipalities as and (b) whether the size of the sinkhole
separate areas, while the rest of the data was corresponds to the IHC value of the bore- Direct correlation of the IHC values
combined to form a fourth group. For this hole. The percentages of the overestimated, The inherent hazard classifications (IHCs)
comparison, sinkholes were selected that correct and underestimated classifications of the individual boreholes and the IHCs of
were up to 50 m away from a borehole. In of the boreholes versus the sinkholes are the corresponding sinkholes are compared
a couple of cases, the hazard classification presented in Figure 1. in Figure 1. Only approximately one quarter
of boreholes in the vicinity was used if the The classifications of the boreholes is deemed classified correctly (23%), while
information on the borehole was insuf- were verified individually. Nine percent of half of the IHCs of the boreholes are gener-
ficient or not representative of the area. the classifications were corrected, implying ally underestimating the size and risk of the
To compare the sinkholes with the that a higher hazard was allocated than the sinkhole, with 26% of the boreholes overesti-
borehole classification, an IHC value was initial classification, while four percent were mating the sinkhole size and/or hazard.
assigned to each sinkhole, based on the borderline IHC 3/4 or IHC 6, in which case Although only a small representation
diameter of the sinkhole and the volume the latter was chosen as the inherent hazard is classified correctly, 61% of the classifica-
of the sinkhole. The maximum volume for classification. Twelve percent of the classifi- tions in the Tshwane Municipality area dif-
a small, medium and large sinkhole was cations were correct, but were changed to a fered by one value, i.e. one class up or down
calculated from the dimensions as specified higher hazard classification due to the clas- from the sinkhole hazard classification.
by Buttrick et al (2001). These values are sification of the surrounding boreholes; thus This represents 67% for the Ekurhuleni
tabulated in Table 2, and a minimum depth the geological conditions of the boreholes Municipality, 56% for the Far West Rand
of 1 m was used to calculate the volume in the area were worse than the borehole region and 65% in the other provinces, as
of the size sinkhole. It was found that the originally classified. In total, an adjustment illustrated in Figure 2. This indicates that
Journal of the South African Institution of Civil Engineering Volume 65 Number 2 June 2023 41
Figure 3 V
isual inherent hazard classification for the Waterkloof Airbase, Tshwane Municipality
there is a relatively good IHC correlation it suggests that the size of the sinkhole has problematic, since the risk has been under-
between the hazard classification of the been underestimated. The blue, green and estimated from the drilling.
boreholes and the sinkhole sizes. yellow dots indicate that the hazard has The areas where discrepancies occurred
been undervalued as either low or medium have been assessed in terms of the subsur-
Classifications with too low risk risk for sinkhole formation, thus IHC 1 to 4 face conditions and hazard classifications
As discussed previously, the areas where classifications. of the boreholes in the immediate vicinity.
sinkholes occur should be classified in During the correction of the classifications
terms of high risk. This suggests that IHC of the boreholes, the following factors were
values from 1 to 4 are not deemed satisfac- CONCLUSION noted:
tory in an area where sinkholes occur. In Although a very high percentage has been QQ Many boreholes were too shallow to
the Tshwane Municipal area, 22% of the underestimated (ranging from 42% to classify appropriately (30 m and less in
boreholes were classified with IHC values 59% for the different regions), there is a depth).
of 4 and lower, suggesting a low to medium relatively good IHC correlation between QQ Classification changed when bore-
hazard for sinkhole formation. For the the hazard classification of the boreholes holes in the area had a higher hazard
Ekurhuleni Municipality, this figure is 17%, and the sinkholes. The number of classifi- classification.
in the Far West Rand region 24% and for cations that are correct and differ by one QQ Deep competent overburden profiles,
the other provinces 29%. value, is 61% in the Tshwane Municipality but underlain by cavities and/or weak
The correlation between the clas- area, 67% for the Ekurhuleni Municipality, soil.
sification of the borehole and those of the 56% for the Far West Rand region and QQ Deep competent profiles, especially in
sinkhole is visually indicated in Figure 3. 65% in the other provinces. The current the Far West Rand, but still sinkholes
The larger circle is colour-coded for the methodology to classify the boreholes has occurred in the area.
IHC of the sinkholes, while the smaller not proved to be one hundred percent QQ Shallow dolomite in an area, but the
circle represents the IHC of the borehole. accurate to predict the size of the sinkhole hazard not classified as high.
Where the colours are the same, it suggests for the area. QQ Boreholes classified as borderline
a good correlation for hazard classification The other fact is that an IHC of 1 to 4 between high and medium hazard.
between the sinkhole and the borehole in (low to medium risk) was assigned to the It is important to take note that when a
proximity thereof. However, if a lighter col- boreholes in 17% to 29% of the cases for profile has cavities or weak soil in the pro-
our is present within the sinkhole marking, the different areas. These areas are more file, even if a thick overburden is present,
42 Volume 65 Number 2 June 2023 Journal of the South African Institution of Civil Engineering
sinkholes can occur. These boreholes in the borehole or other laboratory tests. on generic geological models of a dolomite
should rather be classified with IHC 6 Further research needs to be carried out environment on the East Rand, South Africa.
or higher to incorporate the high risk of around the discrepancies between the IHC Journal of African Earth Sciences, 117: 86–101.
sinkhole occurrence. In a shallow dolomitic of boreholes compared to actual sinkholes. Oosthuizen, T & Van Rooy, J L 2015. Hazard of
area, the risk for small sinkholes to form is sinkhole formation in the Centurion CBD using the
high, and the classification of the boreholes simplified method of scenario supposition. Journal
should reflect that. REFERENCES of the South African Institution of Civil Engineering,
It is important to take cognisance of the Buttrick, D B 1992. Characterization and appropriate 7(2): 69–75.
boreholes in the vicinity when assessing development on sites on dolomite. PhD Thesis. SANS (South African Standard) 2012. SANS.
a site. Boreholes with a high hazard can University of Pretoria. 1936-2: Development of Dolomitic Land. Part 2:
contribute to the instability of the area. In Buttrick, D B & Van Schalkwyk, A 1995. The method Geotechnical Investigations and Determinations.
an area with shallow dolomite bedrock, the of scenario supposition for stability evaluation of Pretoria: SABS Standards Division.
risk for the occurrence of small sinkholes sites on dolomitic land in South Africa. Journal of Sartain, N, Mian, J, O’Riordan, N & Storry, R 2011.
is higher than generally anticipated. Deep the South African Institution of Civil Engineering, Case study on the assessment of sinkhole risk for
competent profiles have a profound influ- 37(4): 4–14 the development of infrastructure over karstic
ence on the instability of the area, and Buttrick, D B, Van Schalkwyk, A, Kleywegt, R J & ground. In Vogt, N, Schuppener, B, Straub, D
alternative ways to determine the stability of Watermeyer, R B 2001. Proposed method for & Bräu, G (Eds). Proceedings, 3rd International
these areas should be investigated. Another dolomite land hazard and risk assessment in South Symposium on Geotechnical Safety and Risk (ISGSR
factor that might be included in this process, Africa. Journal of the South African Institution of 2011). 2–3 June 2011, Munich, Germany.
is to determine whether the type of over- Civil Engineering, 43(2): 27–36. Wagener, F von M 1982. Engineering construction on
burden has an influence on the formation of Kleinhans, I & Van Rooy, J L 2016. Guidelines for dolomites. PhD Thesis. Pietermaritzburg: University
sinkholes and whether it can be recognised sinkhole and subsidence rehabilitation based of KwaZulu-Natal.
Journal of the South African Institution of Civil Engineering Volume 65 Number 2 June 2023 43
Guidelines for the preparation of
papers and technical notes
INTRODUCTION submitted material has not been published previously, is not
The Journal of the South African Institution of Civil Engineering, under consideration for publication elsewhere and will not be
which is a peer-reviewed, open-access journal, is published quar- submitted elsewhere while under consideration by the SAICE
terly in March, June, September and December. Articles submitted Journal Editorial Panel. It is the responsibility of the authors
for publication are reviewed by a panel of referees under the guid- to ensure that publication of any paper in the journal will not
ance of the SAICE Journal Editorial Panel. The journal publishes constitute a breach of any agreement or the transgression of
research papers covering all the disciplines of civil engineering any law. The corresponding author should confirm that all
(structural, geotechnical, railway, coastal/marine, water, construc- co-authors have read and approved the manuscript and accept
tion, environmental, municipal, transportation) and associated these conditions. Authors are responsible for obtaining permis-
topics that are relevant to the civil engineering profession, and sion to publish experimental data and other information that
that preferably have particular relevance to civil engineering in may be confidential or sensitive. Authors are also responsible
Southern Africa and the African continent. When preparing arti- for obtaining written permission from copyright owners (usu-
cles for publication, authors should please take note of, and comply ally the publishers) when reproducing material that had been
with, the guidelines as set out below. published elsewhere. If such material had been published under
a Creative Commons Licence, appropriate recognition must be
included in the submitted manuscript.
CLASSIFICATION OF ARTICLES
CONSIDERED FOR PUBLICATION
QQ Technical papers are well-researched, in-depth, fully refer- SUBMISSION PROCEDURES AND
enced technical articles not exceeding 6 000 words in length REQUIRED FORMAT
(excluding tables, illustrations and the list of references). QQ Online submission: Manuscripts must be uploaded as PDF
Related papers that deal with ‘softer sciences’ (e.g. education, files (http://journal.saice.org.za). Individual file sizes may
social upliftment, etc) are accepted if they are of a techni- not exceed 10 MB. Should you experience problems when
cal nature and of particular interest to the civil engineering uploading your paper, please contact the administrative editor
profession. The latter type of paper will be subject not only (verelene@saice.org.za).
to peer-review by civil engineers, but also to review by non- QQ Format: Manuscripts should be prepared in MS Word and
engineering specialists in the field covered by the paper. presented in double-line spacing, single-column layout with
QQ Technical notes are short, fully referenced technical articles 25 mm wide margins. Line numbers must be applied to the
that do not exceed 2 000 words. A typical technical note will whole document. All pages should bear the authors’ names and
have limited scope often dealing with a single technical issue be numbered at the bottom of the page. With the exception of
of particular importance to civil engineering. tables and figures (see below) the document should be typed in
QQ Review papers are considered for publication as either techni- Times New Roman 12 pt font. Contributions should be accom-
cal papers or technical notes on condition that they are the panied by an abstract of not more than 200 words.
original work of the author and will assist the reader with the QQ First page: The first page of the manuscript should include
understanding, interpreting, or applying of the subject under the title of the paper, the number of words of the main text
review. A review paper must contain criteria by which the work (i.e. excluding figures, tables and the list of references), the
under review was evaluated and must contribute by synthesis- initials, titles and surnames of the authors, professional status
ing the information and drawing new conclusions from the (if applicable), SAICE affiliation (Member, Fellow, Visitor,
dissemination of the previously published work. etc), telephone numbers (landline and mobile), and e-mail
QQ Discussions on published articles are welcomed up to six and postal addresses. The name of the corresponding author
months after publication. The length of discussion contribu- should be underlined. Five keywords should be suggested.
tions is limited to 1 500 words. Where appropriate, discussion QQ Figures, tables, photos and illustrations:
contributions will be subject to the normal reviewing process QQ These should preferably be submitted in colour, as the jour-
and will be forwarded to the authors of the original article nal is a full-colour publication.
for reply. QQ Their positions should be clearly marked in the text as fol-
POLICY REGARDING LANGUAGE AND be numbered consecutively and should appear in the text
ORIGINALITY OF SUBMITTED ARTICLES directly after they have been referred to for the first time.
QQ Language: Manuscripts should be presented in English, as the QQ Illustrations must be accompanied by appropriate captions.
journal is distributed internationally. Captions for tables should appear above the table. All other
QQ Original work: Papers and technical notes must be original captions should appear below the illustration (figures,
contributions. Authors of JSAICE articles must affirm that graphs, photos).
44 Volume 65 Number 2 June 2023 Journal of the South African Institution of Civil Engineering
QQ Only those figures and photographs essential to the under- should be in chronological order. The following are examples of
standing of the text should be included. All illustrations a journal article, a book and a conference paper:
should be referred to in the text. QQ Donald, A W 1954. Pore pressure coefficients.
QQ Figures should be produced using computer graphics. Hand- Géotechnique, 4(4): 143–147.
drafted figures will not be accepted. Lettering on figures QQ Terzaghi, K 1943. Practical soil mechanics. New York: Wiley.
should be equivalent to a Times New Roman 9 pt font, or QQ Smith, A W, Black, C & White, A 1999. Factors affecting the
slightly larger (up to 12 pt) if desired. Lettering smaller than strength of soils. Proceedings, ASCE Conference on Shear
9 pt is not acceptable. Strength of Soils, Colorado, pp 503–532.
QQ Tables should be typed in Times New Roman 9 pt font. They Papers published previously in the Journal of the South African
should not duplicate information already given in the text, Institution of Civil Engineering should be cited if applicable.
nor contain material that would be better presented graphi- QQ Footnotes, trade names, acronyms, abbreviations: These
cally. Tabular matter should be as simple as possible, with should be avoided. If acronyms are used, they should be defined
brief column headings and a minimum number of columns. when they first appear in the text. Do not use full stops after
QQ Mathematical expressions and presentation of symbols: abbreviations or acronyms.
QQ Equations should be presented in a clear form which can QQ Return of amended papers: Papers requiring amendments will
easily be read by non-mathematicians. Each equation be accepted up to six months after the referee reports had been
should appear on a separate line and should be numbered sent to authors, after which time the paper will be withdrawn
consecutively. from the system.
QQ Symbols should preferably reflect those used in Microsoft
Journal of the South African Institution of Civil Engineering Volume 65 Number 2 June 2023 45