Water SA 50(4) 365–383 / Oct 2024 Research paper

https://doi.org/10.17159/wsa/2024.v50.i4.4015

Improving the water quality in the Zandvlei Estuary, Cape Town, by retrofitting
sustainable drainage systems in the Diep River catchment
Geordie Thewlis1 and Neil Armitage1

Department of Civil Engineering, University of Cape Town, Private Bag X3, Cape Town 7701, South Africa
1

CORRESPONDENCE
The Zandvlei Estuary is the only functioning estuary along the False Bay coastline of Cape Town and is Neil Armitage
therefore of extreme local ecological importance. The most significant problems are eutrophication
and siltation caused by the increased total inorganic nitrogen (TIN) and soluble ready phosphorus (SRP) EMAIL
levels due to urban development and the associated increased impervious surface area in the catchment [email protected]
that drains into it. In South Africa, stormwater drainage systems conventionally channel everything they
collect into receiving water bodies without significant treatment. Sustainable drainage systems (SuDS) DATES
provide an alternative approach to managing stormwater runoff. They are designed to manage both Received: 30 August 2022
stormwater quality and quantity while potentially improving biodiversity and amenity. This project Accepted: 11 October 2024
modelled the potential improvement in the quality of the water entering Zandvlei Estuary resulting
KEYWORDS
from the implementation of selected SuDS control measures in Zandvlei’s Diep River catchment using
sustainable drainage systems
the software program, PCSWMM. SRP, TIN, total phosphorus (TP) and total suspended solids (TSS) were stormwater
selected as pollutant indicators. Treatment trains that included a large, constructed wetland at the bottom eutrophication
of the catchment will likely provide the greatest improvements to the water quality entering Zandvlei – PCSWMM
potentially reducing SRP, TIN, TP and TSS by approximately 59%, 53%, 53%, and 66%, respectively – as well treatment train
as potentially reducing the runoff by 48%. wetlands

COPYRIGHT
INTRODUCTION © The Author(s)
Published under a Creative
The Zandvlei Estuary, located in the Southern Suburbs of Cape Town, South Africa, provides 80% of Commons Attribution 4.0
the estuarine area in False Bay, making it by far the largest of the eight estuaries found along the False International Licence
Bay coastline (Brown and Magoba, 2009). It is bordered by the suburbs of Lakeside, Muizenberg, (CC BY 4.0)
Marina da Gama and Steenberg. Its three main catchments – Diep, Keysers and Westlake (Fig. 1)
– support multiple land uses with the Diep River catchment being the most urbanised, including
commercial and industrial zones (Coastal & Environmental Consulting, 2010).
Litter, hydrocarbons, heavy metals, excess nutrients, and sediments associated with urban
development have been allowed to flow directly into the estuary. As the degree of urbanisation has
increased, so too have the loads received by the estuary, resulting in eutrophication, loss of habitat,
and excess sedimentation to the detriment of its functionality (Thornton et al., 1995). Potamogeton
pectinatus, commonly known as pondweed, and the accompanying epiphytic algae, Cladophora/
Enteromorpha spp., are commonly observed. The National Biodiversity Assessment (Van Niekerk
and Turpie, 2012) assigned Zandvlei a ‘D’ Present Ecological State rating in 2011, which was
confirmed in 2018 (Van Niekerk et al., 2018). Meantime, Zandvlei Estuary has also been given an
‘Important’ Biodiversity Importance Rating with recommendations that it be re-established to a
more functional state. However, the enormous impact of urban development makes this challenging
(Thornton et al., 1995).
In South Africa, most stormwater drainage systems contribute to the physical degradation and
ecological destruction of rivers and receiving water bodies through a singular focus on removing
stormwater runoff as quickly as possible through concrete pipes and channels with little to no regard
for the runoff quality. Pollutants and contaminants are swept from impermeable surfaces such as
roofs, roads, and parking areas and deposited into downstream receiving water bodies without
significant intervention to remove harmful substances. Nutrients are washed from fields and gardens.
Raised flood peaks cause erosion and subsequent deposition (Armitage et al., 2013).
Sustainable drainage systems (SuDS) provide a different approach to stormwater drainage. They are
designed to manage both stormwater quality and quantity while potentially improving biodiversity
and amenity (Armitage et al., 2013). There is a growing awareness of their potential in South Africa
(Nyawo and Tanyimboh, 2018). This project thus investigated how selected SuDS treatment trains
may improve Zandvlei Estuary’s water quality through the development and use of a coupled
hydraulic/water quality model in PCSWMM – a customised version of the freely available EPA
SWMM software (CHI, 2020).

METHOD
Several stormwater modelling software packages were investigated and PCSWMM (CHI, 2020) was
selected based on its availability, functionality and applicability, and to maintain continuity with
similar investigations elsewhere. The research framework is presented in Fig. 2.

ISSN (online) 1816-7950 365

Available on website https://www.watersa.net
Figure 1. Zandvlei Estuary locality map showing principal catchments – adapted from Wikimedia Maps (Wikimedia, 2021)

Figure 2. Research framework

Water SA 50(4) 365–383 / Oct 2024 366

https://doi.org/10.17159/wsa/2024.v50.i4.4015
The following steps were followed: gauges, 229 sub-catchments and a single outfall at the end
of the stormwater conveyance network (Fig. 3).
• A coupled hydraulic/hydrological model was set up to
• A water quality model was developed to simulate 4
represent the Diep catchment in its current state. The
stormwater constituent indicators: soluble ready phosphorus
principal stormwater conveyance network was based on
(SRP), total inorganic nitrogen (TIN), total phosphorus
shapefiles obtained from the City of Cape Town (CCT)
(TP), and total suspended solids (TSS). SRP and TIN were
that described the open watercourses, stormwater pipes,
modelled as they are the primary causes of eutrophication
manholes and catchpits.
in water bodies (DWAF, 1996), while TP and TSS were
• A digital elevation model (DEM) from the UCT Geographic
included as they are good indicators of pollution and the
Information Systems (GIS) Unit was used to model surface
CCT requires their loads to be reduced when new SuDS
elevations and delineate the Diep catchment into sub-
developments are implemented (CCT, 2009). The indicators
catchments. Various parameters such as land use, soil types
were simulated using event mean concentrations (EMC)
and infiltration and runoff properties were assigned to each
that are widely used for modelling stormwater constituents.
sub-catchment.
Published data were used to provide preliminary EMC
• Five rain gauges were linked to the PCSWMM model
values which were then adjusted using water quality data
(Fig. 3), each with a unique time series. The rainfall records
from the CCT for the Diep catchment.
were assessed to ensure they met various requirements
• The calibrated and verified hydraulic and water quality
pertaining to record duration, data consistency, and data
model that broadly represented the Diep catchment in its
reliability. All records suitable for use in the model were
current state was then used as the baseline for:
measured at daily intervals and had to be disaggregated to
15-min intervals. • A pre-development model scenario that was created
• The model was calibrated and verified using observed data to give an indication of the situation before urban
from CCT flow gauges. The final calibrated and verified development began.
model included 3 stormwater bodies, 291 junctions, 294 • Five SuDS scenarios that were created to test various
channels (open watercourses, conduits, and pipes), 5 rain treatment train designs.

Figure 3. Model network showing rain gauge sub-catchments, conduits, storages, and outfall – adapted from Wikimedia Maps (Wikimedia, 2021)

Water SA 50(4) 365–383 / Oct 2024 367

https://doi.org/10.17159/wsa/2024.v50.i4.4015
Constructing the model Brabec et al. (2002) reported on impervious surface percentages
from several literature sources, and these values were then refined
Land-use and drainage properties
by overlaying the land-use zones on satellite imagery of the Diep
Land use played a critical role in the development of the catchment in QGIS (QGIS, 2022) and estimating impervious
hydrological model as it directly impacts runoff volumes, runoff percentages.
rates, and indicator build-up and wash-off throughout the
PCSWMM calculates the average velocity of overland flow using
catchment.
the Manning Equation. A Manning’s roughness coefficient, n, was
The CCT (2018) provided a GIS shapefile of the land use in the city determined for both the pervious and impervious sections of each
and its surrounding areas. Some land uses had to be corrected, and land use.
duplicate entries removed. The land uses were verified through
A CCT (2018) GIS shapefile provided the upper soil types for the
site visits and satellite imagery (Fig. 4 and Table 1).
entire city, and this was used to identify the dominant soils within
The impervious percentage of an area affects the volume and the Diep catchment, which are sand, loam, and sandy loam. The
speed of surface runoff with more impervious surfaces generally Green-Ampt infiltration method was used for this research as the soil
resulting in larger runoff volumes and flow rates. (Li et al., 2021). parameters required for this method are widely available in literature.

Figure 4. Adjusted land use map – adapted from Wikimedia Maps (Wikimedia, 2021)

Table 1. Land-use categories used in this research

Land uses
Agricultural Public open space Rural
Commercial Residential – high density School grounds
Environmental conservation Residential – medium density Sports fields
Industrial Residential – low density Roadways
Institutional Residential – very low density Train line

Water SA 50(4) 365–383 / Oct 2024 368

https://doi.org/10.17159/wsa/2024.v50.i4.4015
Streamflow data were collected from the CCT, South African Weather Service,
Department of Water and Sanitation (DWS), UCT’s Climate
Continuous flow data were used to calibrate the hydrological
System Analysis Group, and private citizens. The 21 records
model. The CCT has several streamflow monitoring stations
were then checked for duration, consistency, and data reliability.
throughout the city, three of which are located within the Diep
Mitchell et al. (2008) recommend that rainfall time series should
catchment (Fig. 5): DIEP05CS (Doordrift Road), LPVL05AS
ideally have a minimum duration of 10 years if they are to be used
(Little Princess Vlei) and WYNB05BS (Maynardville Park).
in a continuous stormwater system simulation, as this allows the
However, only DIEP05CS provided reliable data – and then
capture of both intra- and inter-annual variation. Those records
only from the middle of 2013. Sadly, this is situated in the upper
with substantially less than the recommended 10 years, with
reaches of the catchment area and thus could only be used to
large data gaps or unreliable data, were thus excluded – leaving
calibrate those portions of the catchment.
12 records. A suitable timeframe for the hydrological model
The CCT streamflow sensors measure streamflow as water depth simulations was then determined by searching for the period
(m). The depth readings had to be converted to flow rates (m3/s) exceeding 10 years that encompassed the largest number of rainfall
for the calibration process. This was achieved using a rating curve station records. 16 January 2003 – 6 December 2015 was chosen
devised by Rohrer (2017) based on a calibration table received as it was covered by the maximum number of 5 stations (Fig. 5).
from the CCT.
The modelling time-step was also an important consideration.
Daily time-step intervals significantly underestimate stormwater
Rainfall data
runoff volumes (Coombes and Barry, 2007). The ideal time-step
Cape Town has a Mediterranean climate and experiences mild, wet for continuous rainfall model simulations is often considered to be
winters and warm, dry summers. However, due to the mountain 5 min to account for the response time of small sub-catchments.
ranges in and to the east of the city, there are numerous micro- However, given the relatively large sub-catchment areas, the
climates that cause significant areal variation in rainfall (World lack of sufficient data sets with 5-min intervals, and the limited
Weather Online, 2021). The Diep catchment, located at the foot disaggregation abilities of PCSWMM, 15-min time-steps were
of the Peninsula Mountain Chain experiences mean annual deemed to be a reasonable compromise. The five selected rainfall
precipitations ranging from 800 to 1 400 mm/year. In a bid to gauges all recorded at daily intervals and the data thus had to be
capture this variation, 21 rainfall records for the catchment disaggregated to 15-min intervals.

Figure 5. Streamflow monitors and rainfall gauges – adapted from Wikimedia Maps (Wikimedia, 2021)

Water SA 50(4) 365–383 / Oct 2024 369

https://doi.org/10.17159/wsa/2024.v50.i4.4015
The disaggregation process consisted of 2 steps. Firstly, Three existing large stormwater ponds were included in the model:
NetSTORM was used to transform the rainfall records from Little Princess Vlei, Langevlei, and the Maynardville Park Pond.
daily to hourly intervals using rainfall data with hourly or sub-
hourly interval data from nearby stations. Secondly, PCSWMM Calibration and validation
was used to disaggregate the hourly data into 15-min data. The
Calibration and validation of the model were undertaken to
disaggregation process is based on sampling event distributions
reduce the uncertainty of crucial estimated parameters within
from nearby high-resolution rainfall records within the same
the model (James, 2005). PCSWMM’s Sensitivity Radio Tuning
climatic region. While the disaggregation process cannot recreate
Calibration (SRTC) tool was utilised to calibrate the model. The
the actual rainfall events it generates stochastic rainfall data
parameters calibrated included: the sub-catchment properties,
with the same underlying statistics. No sensitivity analysis was
the Manning’s coefficients, the depression storage depths, the
conducted. The rain gauge data were assigned to 5 separate sub-
percentage of impervious areas with no depression storage, the
catchments according to their proximity (Fig. 3).
percentage of runoff routed to pervious areas, and the Green
and Ampt parameters. The sub-catchment parameters were a
Sub-catchment delineation and development
particular focus of the calibration process as they significantly
Sub-catchments provide the base computational unit for the impacted the model output, but published data were initially used
hydrological processes in PCSWMM with each having a single for most as they could not be measured on site.
outlet point determined by their topography (Rossman, 2015).
PCSWMM calibrates and validates models on storm events: 26
PCSWMM offers several tools to aid in model development. storm events were identified from the observed rainfall data. They
The Watershed Delineation tool was used to delineate the entire were split roughly 2:1, with 17 events used to calibrate the model
catchment into sub-catchments using a 1 x 1 m DEM developed and the remaining 9 used to validate the calibrated parameters
by the UCT GIS Unit using light detection and ranging (LiDAR) following methods used in similar studies (Mancipe-Munoz et al.,
data obtained from the CCT. Small errors that might have 2014). DIEP05CS was used for the flow data. Although DIEP05CS
been introduced into the model resulting from the automated only accounted for the upstream portions of the catchment, the
delineation process not being able to identify local deviations calibrated parameters were adjusted equally in both the gauged
in surface level were accounted for during the calibration and and ungauged sections on the assumption that the behaviour of
verification processes. each would be similar. At the end of the calibration and verification
PCSWMM’s Area Weighting tool was used to create weighted process, the model was deemed an acceptable representation of
the physical catchment.
averages for runoff and infiltration properties for each sub-
catchment based on the land uses and soil types present within Table 2 presents the model errors after calibration and validation.
the sub-catchment. Each sub-catchment was defined in terms of Values were determined for: the total flow volume, max flow
3 separate subareas (pervious, impervious, and impervious with rates, and comparison with measured flow hydrographs; and
no depression storage) that, by default, are drained independently 3 error functions were used: integral square error rating (ISE),
by the sub-catchment outlet. PCSWMM allows users the option Nash-Sutcliffe efficiency (NSE), and coefficient of determination
of routing a percentage of the runoff to a separate pervious (R2). Although the minimisation of model error is of the utmost
subarea to model potential infiltration. The sub-catchment outlet importance, there is no generally accepted standard for what
then drains any surplus runoff. This approach was used for the might be considered acceptable (James, 2005). Moriasi et al.
agricultural, public open space, environmental conservation, and (2007) and Golmohammadi et al. (2014) suggest an NSE value
sports field land uses. between 0 and 1 and an R2 of greater than 0.5, while Santhi et
al. (2001) recommend an NSE value greater than 0.5 and an R2
Stormwater conveyance network value greater than 0.6. The calibration was considered acceptable
for this study as both NSE and R2 were significantly greater than
The Diep catchment stormwater conveyance network (Fig. 3) 0.5 for all parameters measured.
comprises natural channels, closed conduits, and open constructed
channels. It was initially modelled using the GIS shapefiles from Figure 6 presents storm event hydrographs of observed and
the CCT open data portal (CCT, 2018) but, given the large modelled data for a typical storm event (27–30 August 2013)
extent of the study site and the absence of detailed data on the after the completion of the calibration and validation processes.
smaller diameter conduits, it was then decided to exclude closed As the rainfall data had to be disaggregated through a stochastic
conduits with a diameter of 675 mm or less. Sub-catchments process, the modelled runoff does not visually match the observed
that would have been drained by the removed conduits had their hydrograph particularly well; however, the error measurements
outlets assigned to the points where they would have connected nevertheless indicate that the storm is calibrated to an acceptable
to included conduits (in Fig. 3 these sub-catchments do not standard as both the NSE and R2 values exceeded 0.5.
appear to be connected to the system as the removed conduits
are not indicated). Inaccuracies introduced by this approach were
accounted for through the calibration and verification of the Table 2. Model errors
model. Parameter Error function Calibrated Validated
Many larger conduits were missing diameter, invert, and/or slope Total flow ISE rating Good Fair
data. These were estimated using standard design procedures such volume NSE 0.883 0.842
as those found in the Neighbourhood Planning and Design Guide R2 0.943 0.935
(CSIR, 2019) on the assumption that this is likely how they were Max flow rates ISE rating Fair Fair
designed.
NSE 0.701 0.707
The open watercourse shapefile includes all the natural and R2 0.748 0.747
altered waterways interconnecting the conveyance network. Hydrograph ISE rating Fair Fair
Natural channel sections were incorporated into the model using
NSE 0.685 0.64
the DEM and PCSWMM’s Transect Creator tool. Constructed
channels with regular sections were measured on site. R2 0.724 0.721

Water SA 50(4) 365–383 / Oct 2024 370

https://doi.org/10.17159/wsa/2024.v50.i4.4015
Figure 6. Modelled and observed hydrographs of a typical storm event (27–30 August 2013)

Figure 7. Water quality sampling locations

Development of the water quality model

The CCT Scientific Services Branch has been monitoring the After analysing the data, it became apparent, however, that many
water quality of many of Cape Town’s rivers for decades. Monthly of the sampling points were inappropriate for this project due
grab samples are taken and dissolved oxygen, temperature, to missing stormwater water quality measurements and a low
salinity, pH, suspended solids, conductivity, total phosphates, number of data entries, leaving only 6 suitable sampling points
orthophosphates, nitrites and nitrates, ammonia, and E. coli (Fig. 7): CR16 (Fig. 8), Little Princessvlei North (LPVN) (Fig. 9),
measured. These data were obtained from the CCT for the Little Princessvlei South (LPVS), Langevlei Inflow (LVI) (Fig. 10),
period from 2000 to 2020 for 12 locations within the catchment. Langevlei Outflow (LVO) (Fig. 11) and CR21 (Fig. 12).

Water SA 50(4) 365–383 / Oct 2024 371

https://doi.org/10.17159/wsa/2024.v50.i4.4015
Figure 8. CR16 sampling point (outflow of Alphen Drive culvert) Figure 9. Little Princessvlei North (LPVN) sampling point

Figure 10. Langevlei Inflow (LVI) sampling point Figure 11. Langevlei Outflow (LVO) sampling point

Figure 12. CR21 sampling point

While the CCT grab samples were useful for locating areas of vary with location (Tuomela et al., 2019). The published values
high pollutant concentrations and guiding SuDS deployment, thus had to be adjusted for the Cape Town context. The values that
they were not directly usable in PCSWMM. Instead, EMCs were were selected were for SRP and TIN as they are the primary causes
used to model stormwater pollutant wash-off and transportation of eutrophication, and TP and TSS as the CCT Management
as EMC values are readily available in literature and PCSWMM of Urban Stormwater Impacts Policy (CCT, 2009) requires
can easily accommodate them (Lin, 2004). developers to achieve TP and TSS reductions of 45% and 80%,
respectively. The current situation (As-is) model was run using
Preliminary EMC values were developed for the land uses in the preliminary EMC values as input and the modelled indicator
the catchment using published data (District Department of the concentrations as outputs at the CCT sampling points obtained.
Environment, 2014; Järveläinen et al., 2017; Kayhanian et al., The EMC input values were then manually adjusted for each land
2007; Mitchell, 2005; Nordeidet et al., 2004; Song et al., 2019; use by comparing the modelled output concentrations to those
Tuomela et al., 2019; USEPA, 1983; Wicke et al., 2021). However, measured by the CCT until a reasonable match was achieved. The
these were all European or American studies and EMC values final input EMC values used in the model are presented in Table 3.

Water SA 50(4) 365–383 / Oct 2024 372

https://doi.org/10.17159/wsa/2024.v50.i4.4015
Table 3. EMC input values used in the model simulations
Land use SRP TIN TP TSS
mg/L
Agricultural 0.300 1.284 1.05 30.0
Commercial 0.150 1.500 0.15 23.0
Environmental conservation 0.080 0.420 0.05 12.0
Industrial 0.110 0.804 0.10 6.5
Institutional 0.110 0.918 0.165 8.96
Public open space 0.080 1.188 0.20 10.0
Residential – high density 0.140 1.500 0.45 7.0
Residential – medium density 0.042 0.400 0.25 10.0
Residential – low density 0.041 1.000 0.05 6.0
Residential – very low density 0.160 1.600 0.40 22.0
Rural 0.029 0.675 0.10 8.0
School grounds 0.123 0.600 0.25 7.0
Sports fields 0.022 0.500 0.165 7.17
Roadways 0.011 1.446 0.10 8.5
Train line 0.022 1.200 0.05 7.39

Modelling SuDS in PCSWMM Table 4. SuDS readily available in PCSWMM (CHI, 2019)

SuDS may be implemented in PCSWMM using the ‘LID Control’ Available in the ‘LID Control’ tools Excluded regional
control SuDS
tools. The tools offer 9 SuDS Stormwater Control Measures Source control Local control
(SCM) (Table 4). Pollutant removal in them is closely associated Rain gardens Bio-retention cells Detention ponds
with stormwater removal through infiltration. Unfortunately, no
Green roofs Infiltration trenches Retention ponds
‘regional controls’, i.e., ponds and wetlands (Armitage et al., 2013),
are included among the tools; however, it is possible to model Rain barrels Vegetative swales Constructed wetlands
regional controls by inserting conduits, junctions, and storage Rooftop disconnection
units. New sub-catchments were created in the PCSWMM model Permeable pavements
to represent individual SuDS SCMs (Fig. 13). This allowed for the
creation of treatment trains as the outflow of one SuDS SCM can
be directed into others downstream.
The pollutant removal abilities of regional controls may be
modelled in PCSWMM by assigning treatment functions for each
pollutant. This could be a fixed percentage removal or a decay
function that indicates the pollutant removal by the SuDS SCM
over time (CHI, 2021). In this project, first-order decay functions
were derived for ponds and wetlands with hydraulic retention
time (HRT) as the independent variable based on published
experimental data. Decay functions ideally require data specific
to each intervention site. Climatic factors, such as temperature,
have a significant impact on the treatment ability of SuDS, with
higher temperatures generally correlating with higher removal
efficiencies (Akratos and Tsihrintzis, 2007). Therefore, published
data on the performance of wetlands and ponds from systems
with similar climates to that of Cape Town were prioritised. Since
PCSWMM requires that treatment equations be defined in terms Figure 13. SuDS (LID) placement approach (Computational Hydraulics
of fractional removal the decay function curves were used in the Inc., 2024; used with permission)
form:
R = 1 – e –k x HRT (1)
Table 5. Treatment equations for retention ponds and wetlands –
where: R = removal fraction of the target pollutant; k = decay derived from experimental data collected by Abbassi et al. (2011);
coefficient associated with the target pollutant; HRT = hydraulic Akratos and Tsihrintzis (2007) and Kabenge et al. (2018)
retention time (h).
Indicator Treatment equations
The inevitable variability in wetland and retention pond High-level removal Low-level removal
efficiencies caused by environmental and hydrological factors was
catered for by establishing both high- and low-level treatment SRP R=1–e −0.016 x HRT
R = 1 – e−0.0048 x HRT
equations for each indicator to provide potential treatment TIN R = 1 – e −0.012 x HRT R = 1 – e−0.004 x HRT
ranges. The removal efficiency curves for SRP, TIN, TP and TSS TP R = 1 – e−0.007 x HRT R = 1 – e−0.0043 x HRT
for retention ponds and wetlands are presented in Figs 14 to 17
and the equations listed in Table 5. TSS R=1–e −0.03 x HRT
R = 1 – e−0.013 x HRT

Water SA 50(4) 365–383 / Oct 2024 373

https://doi.org/10.17159/wsa/2024.v50.i4.4015
Figure 14. SRP removal efficiency curves for retention ponds and wetlands

Figure 15. TIN removal efficiency curves for retention ponds and wetlands

Figure 16. TP removal efficiency curves for retention ponds and wetlands

Figure 17. TSS removal efficiency curves for retention ponds and wetlands

Water SA 50(4) 365–383 / Oct 2024 374

https://doi.org/10.17159/wsa/2024.v50.i4.4015
Model scenarios PCSWMM’s subarea routing function was used to route runoff
from impervious to pervious areas in each sub-catchment. The
As-is
treatment range was established using high and low sub-area
The As-is Scenario was developed to represent the Diep catchment routing percentages for each land use (Table 6) producing high
in its current state. It included the topography, current land and low reduction potentials, respectively.
uses and stormwater drainage networks of the catchment. The
The agricultural areas (A1, A2, A3 and A4) on either side of the
calibrated and validated model was then used as a baseline for all
M3 Freeway produce crops that often require fertilisers, pesticides
subsequent models.
and other products that may result in poor quality runoff. Large
SuDS are not viable in these areas as they would reduce the
Pre-development productive agricultural area, thus swales were deemed the most
The Pre-development Scenario was developed to estimate the effective interventions. Four large swales were placed along the
natural pollutant indicator loads and flow rates that probably contours to reduce the slope, and as close to boundaries as possible
best represent sustainable conditions in the Diep catchment. The to reduce the intrusion into productive agricultural land (Fig. 18).
following changes were made to the As-is Scenario: all land-uses
were set to ‘Environmental conservation’, the pipe network was
removed, constructed conduits were altered to represent more Table 6. Subarea routing percentages for Scenario 1
natural channels, and culverts under roads were replaced with Land use Subarea Land use Subarea
open channels. The scenario could not be calibrated as no flow routing (%) routing (%)
data were available for any period prior to urbanisation. Agricultural 60–70 Residential 75–85
– low
Scenario 1 – source controls Commercial 0 Residential 90–100
– very low
Scenario 1 assumed various source controls – SuDS SCMs that Environmental 90–100 Rural 0
manage stormwater runoff at or near the source – to reduce conservation
runoff volumes from sites and reduce the pollutant loads received
Industrial 0 School 50–60
downstream. It illustrated the effect these may have on reducing grounds
pollutants in areas with open pervious spaces. Since designing
Institutional 40–50 Sports fields 90–100
and modelling these systems for every site in PCSWMM would
Public open space 60–70 Roadways 0
have been highly intensive and would have introduced additional
uncertainty into the model through the need for additional Residential – high 0 Train line 0
parameter estimation, an alternative approach was used. Residential – medium 60–70

Figure 18. Scenario 1 swale locations – adapted from Wikimedia Maps (Wikimedia, 2021)

Water SA 50(4) 365–383 / Oct 2024 375

https://doi.org/10.17159/wsa/2024.v50.i4.4015
A single swale was modelled along the lower boundary of the Scenario 2 thus reincorporated four small existing wetlands and
Plumstead Cemetery (C1). This area produces large volumes of an existing overflow retention pond into the formal stormwater
runoff due to its size. It is neighboured by a sports club that may drainage system. An existing detention pond was converted into
utilise nutrient-based fertilisers to maintain the playing fields and a retention pond. These were all modelled as storage units in
the cemetery swale also collects the runoff from this site. Runoff PCSWMM (Fig. 19). Practically, reincorporation of the wetlands
exiting the swale may be directed towards the existing stormwater and the existing overflow pond would be accomplished by
network, or a small irrigation dam may be constructed to capture removing the berms/walls that prevent direct flow into them.
runoff and allow the sports club and the cemetery to water their
One example of the wetlands that could be re-introduced into the
extensive open areas during dry periods. main river channel is at Ian Taylor Road. It was disconnected from
the Diep River when the river section was channelised (Figs 20
Scenario 2 – historic wetlands and ponds and 21). The concrete walls of the channel currently do not allow
Zandvlei and its catchments have experienced a long history of stormwater to enter the wetland, while an adjacent berm further
modification (Jack, 2006). The relatively flat topography of the separates the two systems.
Diep catchment, especially in the lower reaches, used to support As the area utilised by the wetland is small, the flow into the
an extensive floodplain wetland but urban development has system should be limited to protect the wetland from damage
reduced the extent of the large floodplain wetland to small, isolated and reduce the risk of flooding to the medium-density residential
wetland areas scattered around the catchment (Obree, 2004). areas surrounding the site. However, low flows from smaller, more
Some of these provide attenuation storage during large storm frequent storms are the primary target for SuDS as they transport
events, while others are entirely disconnected from the system. the bulk of the contaminant load.

Figure 19. Existing ponds and wetlands reincorporated into the drainage system

Water SA 50(4) 365–383 / Oct 2024 376

https://doi.org/10.17159/wsa/2024.v50.i4.4015
Figure 20. Existing overgrown marsh/wetland area on Ian Taylor Road

Figure 21. Proposed layout of the Ian Taylor Road wetland

Scenario 3 – new confluence wetland sediment and litter, a sediment basin and litter trap should be
installed immediately upstream.
A single, large-scale regional control at the confluence of the
Sand River and the Langevlei Canal was developed for Scenario
Scenario 4 – source controls and confluence wetland
3. Currently, this is a large unused open area bounded by the
two rivers, making access for recreational activities difficult. The The fourth scenario combined Scenarios 1 and 3 to create a more
small triangular marsh area currently does not receive any runoff holistic treatment train. It thus included the source controls of
from the concrete-lined canals. A constructed wetland that would Scenario 1 with the proposed large wetland at the confluence of
receive all the runoff from both river systems was thus included the Langevlei Canal and the Sand River.
for this 0.5 km2 area (Fig. 22).
Scenario 5 – existing and confluence wetlands
Owing to height differences, the entire site would need to be
excavated for the wetland. As all the runoff from the entire The final scenario combined two existing wetlands and a retention
catchment would be channelled towards this system, an pond (only) from Scenario 2 with the proposed new wetland at the
emergency overflow would be required to protect it during high confluence of the two rivers (Scenario 3). The regional controls
flows. As there is considerable open apace on either side, there would not target specific areas of high indicator inflow but rather
is also a significant opportunity to expand the wetland or utilise treat the entire system at locations where large areas are available.
this space. Additionally, as the runoff collects large volumes of This scenario would likely have a large impact on the water quality

Water SA 50(4) 365–383 / Oct 2024 377

https://doi.org/10.17159/wsa/2024.v50.i4.4015
Figure 22. New confluence wetland location – adapted from Bing Maps (Microsoft Bing, 2021)

Figure 23. As-is Scenario and Pre-development Scenario outflow rates during a 6-month storm event (16–17 August 2005)

Table 7. As-is Scenario indicator quantities and runoff volumes (12 years and 11 months)
Runoff volume (106 m3) Total SRP load (tonnes) Total TIN load (tonnes) Total TP load (tonnes) Total TSS load (tonnes)
61.8 4.4 49.9 11.1 541

received by Zandvlei but may leave isolated areas in the catchment Pre-development Scenario
with poor water quality. The wetlands and ponds incorporated in
The outflow rates experienced by the As-is and Pre-development
this scenario include:
Scenarios are compared in Fig. 23 for a 6-month return period
• Existing: storm event (16–17 August 2005). This figure illustrates the
• Ian Taylor Road wetland extent to which urban development has reduced the infiltration
• Sunbury Road wetland ability of the catchment and significantly increased the outflow
• M3 Freeway retention pond rates – and thus volumes – experienced in the lower reaches.
• New confluence wetland
SuDS load reductions
RESULTS AND DISCUSSION The outputs of the SuDS scenarios are compared to those
obtained from the As-is Scenario in the form of percentage
As-is Scenario
reductions (Fig. 24), as this is required by the CCT (2009)
Table 7 provides the cumulative quantities produced by the As-is stormwater impact policy. The indicator loads deposited into
Scenario at the outfall to Zandvlei between 16 January 2003 and Zandvlei from each of the SuDS scenarios are presented in
6 December 2015. Figs 25 to 28.

Water SA 50(4) 365–383 / Oct 2024 378

https://doi.org/10.17159/wsa/2024.v50.i4.4015
Figure 24. Indicative removal percentages compared with current levels as modelled in the five SuDS scenarios (the area covered by the bars
indicates the range of uncertainty)

Figure 25. SRP load ranges from SuDS scenarios (16 January 2003 to 6 December 2015); orange bars represent the range

Figure 26. TIN load ranges from SuDS scenarios (16 January 2003 to 6 December 2015); orange bars represent the range

Figure 27. TP load ranges from SuDS scenarios (16 January 2003 to 6 December 2015); orange bars represent the range

Water SA 50(4) 365–383 / Oct 2024 379

https://doi.org/10.17159/wsa/2024.v50.i4.4015
Figure 28. TSS load ranges from SuDS scenarios (16 January 2003 to 6 December 2015); orange bars represent the range

Scenario 1 (source controls) provided some level of improvement; the As-is Scenario would likely be reduced in four of the five SuDS
however, this scenario presented the lowest removal percentages scenarios (Fig. 29) with Scenario 2 (reincorporation of historic
for all 4 indicators. This is likely due to the land uses targeted by ponds and wetlands) and Scenario 5 providing concentrations
the SuDS; as this scenario required permeable areas for the source well below those of the As-is Scenario. The TIN concentrations
controls, land uses with minimal pervious areas were largely are all well below the eutrophic range (2.5–10 mg/L).
untreated. Unfortunately, these are often the sites that produce the
Scenario 1 (source controls) presented a slightly increased mean
highest wash-off concentrations. The reductions obtained from
SRP outflow concentration. This is likely due to the areas targeted
Scenario 1 did not meet the TP and TSS reductions of 45% and
by the SuDS; the targeted pervious areas are associated with lower
80%, respectively, required by the CCT.
mean wash-off concentrations than the impervious areas. The
Scenario 2 (reincorporation of historic ponds and wetlands) large impervious areas in the middle and lower reaches of the
provided larger indicator reductions than Scenario 1, albeit catchment were not significantly affected in Scenario 1, thus the
with the greater degree of uncertainty associated with the wide runoff from these areas continued to flow into the river networks
range of treatment potential associated with ponds and wetlands. without treatment. Scenario 3’s large confluence wetland reduced
This scenario also fell short of the CCT recommendations, with both the SRP load and runoff volume by approximately 45%,
maximum reductions of TP and TSS of 34% and 56%, respectively. resulting in a mean SRP outflow concentration similar to that of
the As-is Scenario.
Scenario 3 (new confluence wetland) provided bigger reductions
than Scenarios 1 and 2 for all pollutant indicators except TSS. The
SuDS runoff
CCT 45% reduction requirement for TP was met for by the upper
– but not the lower – treatment potential. Unfortunately, Scenario Urban development has significantly increased the impervious
3 did not meet the TSS reduction requirement. surfaces within catchments resulting in hugely increased runoff
volumes. Furthermore, the channelisation of the river network
As expected, Scenarios 4 (source controls and confluence
has increased runoff flow rates. The likely reduction in runoff
wetland) and 5 (existing and confluence wetlands) provided the
volumes and flow rates due to the implementation of SuDS was
most significant reductions. These scenarios incorporated SuDS
thus assessed. Runoff reductions are provided as percentage
from the first three scenarios to develop treatment trains and
decreases from the As-is runoff volume (Fig. 31). Additionally,
create more robust systems. Both Scenario 4 and 5 completely
the runoff flow rates experienced at the outfall of each scenario
met the CCT TP requirement, with Scenario 5 providing the
during a typical storm event (18–21 April 2010) are presented in
greatest reduction. Unfortunately, neither Scenarios 4 nor 5 met
Fig. 32.
the targeted TSS reduction of 80%. As the TSS targets were not
met in any of the five SuDS scenarios, the rate of siltation within Table 8 presents the predicted total runoff from the Diep River
the estuary will not be satisfactorily decreased. catchment over the 12 years and 11 months modelling period.
The likely sustainable indicator loads from the Pre-development Scenario 1 (source controls) predicted a moderate runoff
Scenario were not obtained in any SuDS scenario (Figs 25 to 28). reduction with a range of 21–26%. As this scenario targeted land
Scenario 5, providing the lowest loads for each indicator, presented uses with pervious areas, the sites with large impervious areas
the closest results to that of the Pre-development Scenario. providing large runoff volumes were not impacted.
Scenario 2 (reincorporation of historic ponds and wetlands)
SuDS outflow concentrations
produced a small runoff reduction of 11%. The reintroduced
Excessive SRP and TIN concentrations are responsible for the systems targeted low flows from smaller storms. The larger flows
overgrowth of plants and cause eutrophication in water bodies. bypass the new systems and continue down the existing channels
The concentrations of these entering Zandvlei must therefore until they reach Zandvlei.
be reduced. Figures 29 and 30 present the mean outflow
The large confluence wetland system at the discharge point into
concentrations at the model outlet from each SuDS scenario and
the estuary receives the entirety of the flows from the Diep/Sand
the concentration range in which eutrophication may occur, as
River and Langevlei Canal systems and provides considerable
specified by DWAF (1996).
attenuation storage that slows runoff and allows infiltration. As
The SRP outflow concentrations are still within the eutrophic range expected, the scenarios that included this wetland (Scenarios
for each SuDS scenario (0.025–0.25 mg/L); thus, eutrophication 3, 4, and 5) produced the most significant drops in runoff
will likely continue. The mean SRP concentration obtained from volumes with all three producing runoff decreases of over 40%.

Water SA 50(4) 365–383 / Oct 2024 380

https://doi.org/10.17159/wsa/2024.v50.i4.4015
Figure 29. SRP model outflow concentrations; grey band represents the eutrophic range (0.025–0.25 mg/L); orange bars represent the
concentration range due to high and low treatment potentials

Figure 30. TIN model outflow concentrations; eutrophic range of 2.5–10 mg/L; orange bars represent the concentration range due to high and low
treatment potentials

Figure 31. Runoff volume reduction percentages; orange bars represent the range

Figure 32. Outfall flow rates during a typical storm event (18–21 April 2010)

Water SA 50(4) 365–383 / Oct 2024 381

https://doi.org/10.17159/wsa/2024.v50.i4.4015
Table 8. Total runoff volumes (12 years and 11 months) for different scenarios
Scenario As-is Pre-development Scenario 1 Scenario 2 Scenario 3 Scenario 4 Scenario 5
Total runoff volume (106 m3) 61.8 6.2 46.0–48.9 54.9 34.0 30.3–31.2 32.3

As Scenario 4 (source controls and confluence wetland) includes the Pre-development Scenario. Therefore, the overall objective
elements from Scenario 1, this resulted in two separate runoff of improving the water quality within Zandvlei Estuary using
values. Table 8 presents the runoff volumes received by Zandvlei SuDS in the Diep catchment is achievable, but improving the
over the entire simulation period. water quality to the sustainable conditions observed in the Pre-
development model would require additional interventions.
The Pre-development Scenario suggests a total runoff volume
decrease of approximately 90% compared with the As-is Scenario.
AUTHOR CONTRIBUTIONS
The magnitude of this reduction is to be expected as the catchment
would have had significantly less impervious area allowing much Geordie Thewlis was responsible for the collection of the data, the
more infiltration. The reduction is over 35% more than the best construction and running of the various models, the analysis of the
performing SuDS scenario, Scenario 4. model outputs, and the draft paper. Neil Armitage was responsible
for the conceptualisation of the project, critical intellectual input
CONCLUSIONS during the research, and the final editing of the paper.
The City of Cape Town (2009) Management of Urban Stormwater
REFERENCES
Impacts Policy requires that new developments should decrease
TP and TSS loads by at least 45% and 80%, respectively. ABBASSI B, AL-ZBOON K, RADAIDEH J and WAHBEH A (2011)
Additionally, the South African Water Quality Guidelines for Using constructed wetlands to improve drainage water quality
Aquatic Ecosystems (DWAF, 1996) state that SRP and TIN are from hydroponics farms. Irrig. Drain. 60 (3) 370–380. https://doi.
responsible for the eutrophic states in estuaries, a significant org/10.1002/IRD.580
problem for Zandvlei (Thornton et al., 1995). Therefore, the AKRATOS C and TSIHRINTZIS V (2007) Effect of temperature, HRT,
Diep catchment – the most urbanised contributor to the flow vegetation and porous media on removal efficiency of pilot-scale
horizontal subsurface flow constructed wetlands. Ecol.Eng. 29 (2)
into Zandvlei – was modelled to see the likely impact of various
173–191. https://doi.org/10.1016/j.ecoleng.2006.06.013
SuDS scenarios using SRP, TIN, TP and TSS as the stormwater
ARMITAGE N, VICE M, FISHER-JEFFES L, WINTER K, SPIEGEL
constituent indicators.
A and DUN (2013) The South African guidelines for sustainable
Various SuDS scenarios were tested under conditions of both high drainage systems (Issue March). WRC Report No. TT 558/13. Water
and low treatment to determine the likely pollutant reduction Research Commission, Pretoria.
ranges for SRP, TIN, TP and TSS. The results were compared BRABEC E, SCHULTE S and RICHARDS P (2002) Impervious
to those obtained from As-is and Pre-development Scenarios. surfaces and water quality: A review of current literature and its
Scenarios 1 (source controls) and 2 (reincorporation of historic implications for watershed planning. J. Plan. Lit. 16 499–514. https://
ponds and wetlands) did not meet either CCT target. Scenario doi.org/10.1177/088541202400903563
BROWN C and MAGOBA R (2009) Rivers and Wetlands of Cape Town
3 (new confluence wetland) partially met the CCT TP reduction
– Caring for our Rich Aquatic Heritage. WRC Report No. TT 378/08.
target at the top end. Scenarios 4 (source controls and confluence
Water Research Commission, Pretoria.
wetland) and 5 (existing and confluence wetlands) fully met the
CCT (City of Cape Town) (2009) Management of Urban Stormwater
TP reduction target. None of the five SuDS scenarios met the 80% Impacts Policy. Catchment, Stormwater and River Management
TSS target. As a result, sedimentation may occur in Zandvlei at Branch. City of Cape Town, Cape Town.
a rate faster than that experienced before urbanisation began; CCT (City of Cape Town) (2018) City of Cape Town Open Data
however, the rate will be lower than that currently experienced. Portal. URL: https://web1.capetown.gov.za/web1/OpenDataPortal/
None of the scenarios resulted in mean outfall SRP concentrations (Accessed 4 May 2020).
below the eutrophic range, thus eutrophication may still occur in COASTAL & ENVIRONMENTAL CONSULTING (2010) Situation
Assessment for the Zandvlei Estuary. Coastal & Environmental
Zandvlei. However, Scenarios 2, 3, 4 and 5 will all likely result in
Consulting, Cape Town.
lower mean SRP concentrations than the As-is Scenario.
CHI (Computational Hydraulics International) (2019) Introduction
The Pre-development Scenario indicated a 90% decrease in runoff to modelling LIDs in PCSWMM and SWMM5. Computation
volume compared with the As-is Scenario and was over 35% better Hydraulics International, Cincinnati.
than the best SuDS scenario. The modelled runoff received by CHI (Computational Hydraulics International) (2020) CHI – PCSWMM.
Zandvlei from the Diep catchment was reduced between 10% and URL: https://www.pcswmm.com/ (Accessed 4 May 2020).
55% in the various SuDS scenarios, with Scenario 2 and Scenario CHI (Computational Hydraulics International) (2021) PCSWMM
4 providing the lowest and highest reductions, respectively. support. URL: https://support.chiwater.com/ (Accessed 14 May 2020).
CHI (Computational Hydraulics Inc.) (2024) PCSWMM/SWMM5
Based on the results obtained from the various scenarios, the Training: Introduction to modeling LIDs in PCSWMM and SWMM5.
implementation of Scenario 5 would provide the most significant Presentation No: K052. Shared with permission, 29 October 2024.
improvement to Zandvlei’s water quality. However, while this COOMBES P and BARRY M (2007) The effect of selection of time steps
scenario would likely surpass the CCT 45% TP reduction and average assumptions on the continuous simulation of rainwater
target it would not meet the 80% TSS reduction target. Many harvesting strategies. Water Sci. Technol. 55 (4) 125–133. https://doi.
of the wetlands and ponds utilised in the scenario already exist; org/10.2166/wst.2007.102
however, rehabilitation, remediation and maintenance of these CSIR (Council for Scientific and Industrial Research) (2019) Section L:
sites are required before they may be used effectively. Currently, Stormwater. In CSIR (ed.) Red Book: The Neighbourhood Planning
the location proposed for the large, confluence wetland is unused. and Design Guide. Department of Human Settlements, Pretoria.
DWAF (Department of Water Affairs and Forestry, South Africa)
Unfortunately, the indicator loads from Scenario 5 were (1996) South African Water Quality Guidelines Volume 7: Aquatic
substantially larger than the likely sustainable results obtained by Ecosystems (1st edn). DWAF, Pretoria.

Water SA 50(4) 365–383 / Oct 2024 382

https://doi.org/10.17159/wsa/2024.v50.i4.4015
DISTRICT DEPARTMENT OF THE ENVIRONMENT (2014) OBREE M (2004) Catchment, stormwater and river management in
Technical Memorandum – Selection of event mean concentrations Cape Town, South Africa. J. Water Manage. Model. https://doi.
(EMCs). LimnoTech, Washington DC. org/10.14796/JWMM.R220-28
GOLMOHAMMADI G, PRASHER S, MADANI A and RUDRA R QGIS (2022) QGIS User Guide. https://docs.qgis.org/3.22/en/docs/
(2014) Evaluating three hydrological distributed watershed models: user_manual/ (Accessed 26 April 2020).
MIKE-SHE, APEX, SWAT. Hydrology 1 (1) 20–39. https://doi.org/ ROHRER A (2017) The viability of using the stormwater ponds on the
10.3390/hydrology1010020 Diep River in the Constantia Valley for stormwater harvesting.
JACK S (2006) Changing land use/land cover around an urban estuary: Thesis, Faculty of Engineering & the Built Environment, Urban
Implications for ecosystem functioning. BSc Honours in Botany Water Management, University of Cape Town. URL: http://hdl.
dissertation, Department of Biological Sciences, University of Cape handle.net/11427/24918
Town. ROSSMAN L (2015) Storm Water Management Model User Manual
JAMES W (2005) Rules for Responsible Modeling (4th edn). (5.1, Issue September). USEPA, Cincinnati OH.
Computational Hydraulics International (CHI), Guelph. SANTHI C, ARNOLD J, WILLIAMS J, DUGAS W, SRINIVASAN R
JÄRVELÄINEN J, SILLANPÄÄ N and KOIVUSALO H (2017) Land- and HAUCK L (2001) Validation of the SWAT model on a large river
use based stormwater pollutant load estimation and monitoring basin with point and nonpoint sources. J. Am. Water Resour. Ass.
system design. Urb. Water J. 14 (3) 223–236. https://doi.org/10.108 37 (5) 1169–1188. https://doi.org/10.1111/j.1752-1688.2001.tb03630.x
0/1573062X.2015.1086005 SONG H, QIN T, WANG J and WONG T (2019) Characteristics of
KABENGE I, OUMA G, ABOAGYE D and BANADDA N (2018) stormwater quality in Singapore catchments in 9 different types of
Performance of a constructed wetland as an upstream intervention land use. Water 11 (1089) 1–10. https://doi.org/10.3390/w11051089
for stormwater runoff quality management. Environ. Sci. Pollut. Res. THORNTON J, BEEKMAN H, BODDINGTON G, DICK R,
25 36765–36774. https://doi.org/10.1007/s11356-018-3580-z HARDING W, LIEF M, MORRISON I and QUICK A (1995) The
KAYHANIAN M, SUVERKROPP C, RUBY A and TSAY K (2007) ecology and management of Zandvlei (Cape Province, South Africa),
Characterization and prediction of highway runoff constituent an enriched shallow African estuary. In McComb A: Eutrophic,
event mean concentration. J. Environ. Manage. 85 279–295. https:// Shallow Estuaries and Lagoons. Vol. 1. CRC Press, Florida. 109–128.
doi.org/10.1016/j.jenvman.2006.09.024 TUOMELA C, SILLANPÄÄ N and KOIVUSALO H (2019) Assessment
LI F, LI E, ZHANG C, SAMAT A, LIU W, LI C and ATKINSON P of stormwater pollutant loads and source area contributions with
(2021) Estimating artificial impervious surface percentage in Asia stormwater management model (SWMM). J. Environ. Manage. 233
by fusing multi-temporal MODIS and VIIRS nighttime light data. 719–727. https://doi.org/10.1016/j.jenvman.2018.12.061
Remote Sens. 13 (212) 1–23. https://doi.org/10.3390/rs13020212 USEPA (United States Environmental Protection Agency) (1983)
LIN J (2004) Review of published export coefficient and event mean Results of the Nationwide Urban Runoff Program. USEPA,
concentration (EMC) data. URL: http://www.epa.sov/cei-bin/clarit Washington DC.
gw?op-Displav&document= (Accessed 14 June 2021). VAN NIEKERK L, ADAMS J, JAMES N, LAMBERTH S, MACKAY C,
MANCIPE-MUNOZ N, BUCHBERGER S, SUIDAN M and LU T TURPIE J, RAJKARAN A, WEERTS S and WHITFIELD A (2018)
(2014) Calibration of rainfall-runoff model in urban watersheds National Biodiversity Assessment 2018 (Vol. 3). URL: http://hdl.
for stormwater management assessment. J. Water Resour. Plan. handle.net/20.500.12143/6373 (Accessed 24 March 2020).
Manage. 140 (6). VAN NIEKERK L and TURPIE J (2012) National Biodiversity
MICROSOFT BING (2021) Bing Maps. URL: https://www.bing.com/ Assessment 2011: Technical Report: Vol. 3: Estuary. CSIR Report
maps/ (Accessed 22 November 2021). Number CSIR/NRE/ECOS/ER/2011/0045/B. Council for Scientific
MITCHELL G (2005) Mapping hazard from urban non-point and Industrial Research, Stellenbosch.
pollution: A screening model to support sustainable urban drainage WICKE D, MATZINGER A, SONNENBERG H, CARADOT N,
planning. J. Environ. Manage. 74 (1) 1–9. https://doi.org/10.1016/ SCHUBERT R-L, DICK R, HEINZMANN B, DÜNNBIER U, VON
j.jenvman.2004.08.002 SEGGERN D and ROUAULT P (2021) Micropollutants in urban
MITCHELL V, MCCARTHY D, DELETIC A and FLETCHER T stormwater runoff of different land uses. Water 13 (9) 1312. https://
(2008) Urban stormwater harvesting - sensitivity of a storage doi.org/10.3390/w13091312
behaviour model. Environ. Model. Softw. 23 (6) 782–793. https://doi. WIKIMEDIA (2021) Wikimedia Maps. URL: https://maps.wikimedia.
org/10.1016/j.envsoft.2007.09.006 org/#12/-34.0545/18.4924 (Accessed 22 November 2021).
MORIASI D, ARNOLD J, VAN LIEW M, BINGNER R, HARMEL R WORLD WEATHER ONLINE (2021) Cape Town, Western Cape, South
and VEITH T (2007) Model evaluation guidelines for systematic Africa weather averages. URL: https://www.worldweatheronline.
quantification of accuracy in watershed simulations. Trans. ASABE com/cape-town-weather-averages/western-cape/za.aspx (Accessed
50 (3) 885–900. https://doi.org/10.13031/2013.23153 16 September 2021).
NORDEIDET B, NORDEIDE T, ÅSTEBØL SO and HVITVED-
JACOBSEN T (2004) Prioritising and planning of urban stormwater
treatment in the Alna watercourse in Oslo. Sci. Total Environ.
334–335 231–238. https://doi.org/10.1016/j.scitotenv.2004.04.040
NYAWO R and TANYIMBOH T (2018) Conventional versus
sustainable drainage systems: evaluation of stormwater management
in an urban residential complex. In: International Conference
on Sustainable Sanitation, Waste and Water Management. URL:
https://www.researchgate.net/publication/329192034_Conventional_
Versus_Sustainable_Drainage_Systems_Evaluation_of_Stormwater_
Management_in_an_Urban_Residential_Complex/link/5c21499fa6f
dccfc7067019d/download (Accessed 14 June 2021).

Water SA 50(4) 365–383 / Oct 2024 383

https://doi.org/10.17159/wsa/2024.v50.i4.4015