Polluted runoff from informal settlements in developing countries poses a growing challenge due to the elevated and variable nature of contaminants, particularly nutrients and pathogens, introduced to the environment. Cost-effective and scalable treatment systems with the ability to reduce nutrient and other pollutant concentrations in contaminated runoff are desirable. Biofilters are passive water treatment systems that have the potential to achieve this. The Franschhoek Water Hub, a research site for nature-based solutions, features six large biofiltration cells designed to remediate runoff from an informal settlement. Due to their large size, understanding hydraulic behaviour and validating the design proves challenging. To address this, a scaled-down version of the Water Hub's biofilters was constructed to inform design criteria for purpose-built filters. The pilot-scale subsurface flow biofilter, filled with 8–11 mm aggregate stone, had an available volume of 225 L. Pulse tracer studies conducted at various flow rates demonstrated that the system approximated plug flow behaviour. Lower flow rates resulted in deeper tracer infiltration, which is crucial for maximising the distribution of nutrients within the filter bed. This research contributes to the effective design and operation of biofiltration systems, which hold promise for addressing surface water contamination issues in resource-constrained regions.

  • Pulse tracer studies at various flow rates on a pilot-scale subsurface flow biofilter, mimicking the design of a field-scale system, investigated fluid flow behaviour in a field-scale system.

  • The extent of plug flow behaviour and degree of mixing in the radial direction increased with decreasing flow rate.

  • Low operational flow rates ensure greater utilisation of the active filter bed.

Surface water runoff from densely populated and poorly serviced informal settlements presents a growing problem in South Africa and other regions of the Global South, resulting in health risks and environmental degradation due to elevated nutrient levels (nitrogen and phosphorus), bacteria, and pathogen levels (Taing et al. 2013; Malulu 2016; SERI 2018). This study forms part of the efforts to remediate the poor water quality of the Stiebeuel River, which is located near the town of Franschhoek in the Western Cape Province of South Africa. An informal settlement, Langrug, is located near the source of the river. The water quality of the Stiebeuel River varies throughout the year due to natural factors such as seasonal variations in runoff volume and weather conditions, as well as human influences, particularly stemming from the informal settlement (Fell 2017). This poses a challenge to the treatment of contaminated surface water, as is found in many such locations across southern Africa.

A cost-effective, scalable, and easy-to-maintain treatment system is preferable in this context, especially due to the limited availability of resources, particularly capital and skilled personnel, in peri-urban areas of South Africa. Suitable treatment options include engineered semi-passive nature-based solutions (NBS) such as biofilters and constructed wetlands (Ramírez-Agudelo et al. 2020). When properly designed, biofilters and constructed wetlands have been shown to lower nutrient and major pollutant concentrations in contaminated surface water runoff considerably. These treatment systems have been proposed as effective solutions for the remediation of polluted runoff (Bratieres et al. 2008; Chandrasena et al. 2017; Nivala et al. 2017; Winter et al. 2023). These systems are also desirable as they provide access to clean water with the potential for reuse for multiple purposes (Armitage et al. 2013; Zhang et al. 2016).

The Franschhoek Water Hub, which is in the Stiebeuel River catchment, provides a test site for technologies appropriate for low-cost and distributed operation towards improved water quality in such environments. Field-scale biofiltration cells are one of the technologies housed at the Franschhoek Water Hub. Proof of concept of these biofiltration cells has been achieved, but the large scale of the system makes studying the processes that govern the system and optimisation of the system a challenge (Winter et al. 2023). Therefore, a pilot-scale biofiltration system, based on the design of the field-scale biofilters at the Water Hub, was used to conduct detailed studies on the factors affecting system performance and is the focus of this study.

Hydraulic residence time has been identified as a key factor affecting the performance of horizontally orientated subsurface flow biofilters (Bodin et al. 2013). Design choices such as the location of inlet and outlet ports, vessel orientation, and type of packing media have a notable impact on fluid flow behaviour in the system (Narayanan & Narayan 2019; Zhong et al. 2020; Wang et al. 2021). A poor design may result in dead zones or short-circuiting within the system, which reduces the treatment efficiency of the system (Bonner et al. 2018). A sufficient hydraulic retention time is necessary for the degradation of pollutants in a biofiltration system (Ferreira et al. 2021). Hydraulic retention time is dependent on the operational flow rate as well as the fluid pathway. Fluid flow in a system is geared towards the path of least resistance; system design is critical in ensuring minimal dead zones and short-circuiting (Bonner et al. 2018).

The hydraulics of biofilters and constructed wetlands often approximate hydraulic behaviour seen in packed bed reactors, demonstrating approximate plug flow behaviour (Narayanan & Narayan 2019). As such, minimal mixing in these systems is expected while pollutant concentrations vary along the length of the biofilter (concentration gradient in the axial direction). Empirical data from analyses of horizontally orientated subsurface flow fixed film biofiltration systems show that fluid flow in the axial direction is dominant and axial dispersion is minimal (Wang et al. 2014; Bonner et al. 2018; Gajewska & Skrzypiec 2018). These systems favour pollutant degradation where the reaction rate is dependent on reactant concentration due to the low degree of mixing in the system (Gajewska & Skrzypiec 2018).

The aim of this study is to use a pilot biofilter that is representative of the geometry of the large-scale biofilters in place at the Water Hub in Franschhoek, South Africa (Winter et al. 2023) to provide insight into the hydraulic characteristics of these biofilters using tracer studies at feasible operating flow rates. This informs optimisation of their operation and improved design of further biofilters for the treatment of polluted runoff from informal settlements. This study is valuable in the context of addressing poor water quality from informal settlements because while the benefits of such systems are acknowledged, there is limited availability of empirical evidence showcasing the practical implementation of nature-based treatment systems for the decentralised remediation of surface waters contaminated by informal settlements.

The pilot biofilter

The pilot biofilter was designed to mimic the design and operation of the Water Hub biofilters so that the experiments performed on the pilot system provide insight into the design and operation of the Water Hub biofilters. A schematic of the aerial and end views of the pilot biofilter can be seen in Figures 1 and 2. The length and width of the pilot biofilter were scaled down from the length and width of the Water Hub biofilters by a factor of 8. The depth, equivalent to bed height, was kept the same as the Water Hub biofilters to encourage a similar microbiological environment. Therefore, a 225-L pilot-scale horizontally orientated biofilter with a length of 2 m, interior width of 0.44 m, exterior width of 0.48 m, and depth of 0.7 m was designed and constructed for the purpose of this study. A 25-mm inlet distribution pipe spanned the width of the biofilter and contained four 5-mm holes spaced 100 mm apart to allow an even flow distribution at the inlet of the pilot biofilter (this was modelled to be identical to the inlet distribution pipes of the Water Hub biofilters). River stone of diameter 8–11 mm was used as the filter media in the system as this system was modelled after the stone biofilters at the Water Hub. The filter media depth was 65 cm with the inlet distribution pipe situated directly above the stone media. The measured bed voidage of the pilot-scale system was 0.42.
Figure 1

End views of pilot biofilter (dimensions in cm).

Figure 1

End views of pilot biofilter (dimensions in cm).

Close modal
Figure 2

Aerial schematic of pilot biofilter (dimensions in cm): (a) aerial schematic of unpacked pilot biofilter (drawn to scale) showing 25 mm inlet pipe (D), inlet distributor with four inlet points (E), and outlets A, B, and C; and (b) aerial view of pilot biofilter (not to scale) with position of sampling ports 1–10 and outlets (A, B, and C) labelled in red.

Figure 2

Aerial schematic of pilot biofilter (dimensions in cm): (a) aerial schematic of unpacked pilot biofilter (drawn to scale) showing 25 mm inlet pipe (D), inlet distributor with four inlet points (E), and outlets A, B, and C; and (b) aerial view of pilot biofilter (not to scale) with position of sampling ports 1–10 and outlets (A, B, and C) labelled in red.

Close modal

Figure 1 shows the end views of the pilot biofilter when facing the inlet and outlets. The two side outlets (outlet A and outlet C) are positioned 2 cm above the bottom of the pilot biofilter. These outlets may experience obstructed flow in the event of a sediment build-up at the bottom of the biofilter. The middle outlet (outlet B) is therefore positioned 3 cm above outlets A and C to ensure that effluent can always exit the pilot biofilter unimpeded.

Figure 2 is an aerial view of the pilot biofilter detailing the inlets and outlets as well as the design of the inlet distribution pipe. The casing of the pilot biofilter was constructed out of plywood. The exterior edges and corners were waterproofed using silicon sealant. The interior edges and corners of the biofilter were waterproofed with a fibreglass layer. A layer of plastic sheeting was placed over the fibreglass as a further waterproofing measure. Ten sampling ports (SP) constructed as stainless-steel mesh wire cylinders with a diameter of 20 mm were placed in the pilot biofilter to enable sampling at different depths. This was particularly useful in the tracer studies where tracer concentrations were taken from 5 cm below the surface as well as 35 cm below the surface. Figure 2 is an aerial view of the pilot biofilter with the SP (1–10) and outlets (A, B, and C) labelled. The distances of the SP to the filter walls can be noted in Figure 2(a).

Pulse tracer studies

Pulse tracer studies at various potential operating flow rates were conducted to generate concentration–time data to determine the residence time distributions of the pilot biofilter at the respective flow rates. In a pulse tracer study, a fixed volume of the tracer of known concentration is injected into the system inlet as a single pulse, and the concentration of the tracer is monitored at the outlet.

Allura Red AC (C18H14N2Na2O8S) dye was chosen as the tracer for the pulse tracer studies as it contains two sulphonic acid groups, which minimise adsorption onto the filtration media, thus reducing tracer loss due to adsorption. Preliminary batch beaker studies showed that the adsorption of the Allura Red dye onto the stone media was minimal, with 3.02% of the dye being adsorbed onto the stone media after 48 h.

Empirically, the maximum flow capacity in the pilot biofilter was 9 L/min (avoiding splashing and spillage), while the minimum flow capacity was 0.5 L/min. The pulse tracer studies were conducted at 3, 2, 1.5, 1, 0.75, and 0.5 L/min. The low flow rates were chosen because the system was designed to treat contaminated runoff, which requires prolonged contact between the fluid and filtration medium in order to remove pollutants. It must also be noted that laminar flow is required for tracer studies since laminar flow is reversible. The resulting residence time distributions can be normalised with time, thereby allowing comparisons between residence time distributions of different flow rates (Fogler 2011). The Reynolds number for laminar flow in a packed bed reactor is less than 10 and is greater than 2,000 for turbulent flow. A flow rate of less than 10.5 L/min results in laminar flow in the pilot biofilter, while a flow rate of greater than 2,150 L/min results in turbulent flow. The chosen operational flow rates are well within the laminar flow region.

The tracer studies were conducted at a constant inlet flow rate at the desired flow rate. The liquid level in the biofilters was monitored (using digital flow meters) until liquid could be seen below the surface of the stones. The middle outlet valve of the pilot biofilter (outlet B) was then adjusted such that the outlet flow rate was the same as the inlet flow rate (steady state). Once a steady state was achieved, 50 mL of Allura Red dye at a concentration of 5 g/L was injected into the tracer inlet of the pilot biofilter – this was considered time 0 for the tracer studies. To ensure consistency of results, data collection was conducted in triplicate. Samples were taken at 5 and 35 cm below the surface at each flow rate. The sampling regime for the tracer studies is presented in Table 1. The tracer concentration was measured as absorbance at 504 nm within 24 h.

Table 1

Sampling regime for pulse tracer studies at all six experimental flow rates

 
 

Modelling hydraulic performance using residence time distribution

The hydraulic characteristics of the system were calculated using the methods outlined in Fogler (2011). The theoretical residence time, , of the pilot biofilter system was calculated using Equation (1). The theoretical residence time describes the hydraulic residence time that would be observed if the pilot biofilter operated in an ideal manner (i.e., no dead zones or channelling; perfect plug flow behaviour) (Fogler 2011).
formula
(1)
where V is the available volume of the pilot biofilter (L), and Q is the experimental flow rate (L/min).
The residence time distribution function, E(t), is derived from the concentration–time data observed in the pulse tracer studies using the following equation.
formula
(2)
The mean residence time distribution is calculated using Equation (3). This is a representation of the actual hydraulic residence time in the system.
formula
(3)
The variance is the centred second moment of the residence time distribution. It is used to describe the spread of the data and is calculated using the following equation.
formula
(4)
The reversible flow properties of laminar flow allow for the residence time distributions to be normalised with respect to time. This is valuable as it enables comparisons between hydraulic data generated at different flow rates. Equations (5) and (6) show the normalisation of the hydraulic data into dimensionless variables.
formula
(5)
formula
(6)
where θ represents the number of reactor volumes that flowed through the system.
The effective volume utilisation, e, describes the portion of the reactor volume utilised. The un-utilised portion of the reactor is assumed to be dead space. The effective volume utilisation can be calculated using the following equation.
formula
(7)
The tanks-in-series model is used to determine the number of equally sized, ideal continuous stirred tank reactors (CSTRs) in series that approximate the hydraulic performance of the system, giving an indication of approximation of plug flow compared with a fully mixed system. The number of tanks in series is determined using the following equation.
formula
(8)

The fluid dynamics of the system approach mixed flow behaviour when N → 1, while the fluid dynamics approach ideal plug flow behaviour when N → ∞.

The hydraulic efficiency is used to determine the hydraulic performance of non-ideal reactors, with ideal plug flow systems being 100% efficient. Hydraulic efficiency can be calculated using the following equation.
formula
(9)
The Peclet number (Pe) can be used to determine whether advective flow or dispersive mass transfer is dominant in the system. This is relevant as fluid flow patterns have an impact on the manner in which nutrients are distributed in the system, affecting their availability for biological metabolisation. Pe is calculated using the relationship described in the following equation.
formula
(10)

Advective flow is dominant, and dispersive mass transfer is negligible if Pe ≫ 1. The system approaches ideal plug flow behaviour as Pe increases with mixing occurring in the radial direction. The dispersion number (D) is the inverse of Pe and informs the overall extent of axial dispersion in the system. When D is 0, the system approaches ideal plug flow behaviour with negligible axial dispersion.

Tracer recovery (%) is calculated using the following equation.
formula
(11)
where M0 is the mass of tracer injected into the system at time 0.

Concentration profiles

Figures 3 and 4 show the concentration profiles along the length of the pilot biofilter at 5 and 35 cm below the surface, respectively. Figure 3 demonstrates that the pilot biofilter showed an even flow distribution at a depth of 5 cm across all flow rates with a peak concentration of 6 mg/L. The sampling locations on the left and right at the same distance along the biofilter yielded the same concentration profiles.
Figure 3

Concentration profiles 5 cm below surface for pulse tracer studies: (a) 3, (b) 2, (c) 1.5, (d) 1, (e) 0.75, and (f) 0.5 L/min.

Figure 3

Concentration profiles 5 cm below surface for pulse tracer studies: (a) 3, (b) 2, (c) 1.5, (d) 1, (e) 0.75, and (f) 0.5 L/min.

Close modal
Figure 4

Concentration profiles 35 cm below surface for pulse tracer studies: (a) 3, (b) 2, (c) 1.5, (d) 1, (e) 0.75, and (f) 0.5 L/min.

Figure 4

Concentration profiles 35 cm below surface for pulse tracer studies: (a) 3, (b) 2, (c) 1.5, (d) 1, (e) 0.75, and (f) 0.5 L/min.

Close modal

A correlation is observed between the decrease in dye concentration at 5 cm below the surface and the increase in dye concentration at 35 cm below the surface along the length of the pilot biofilter. As the operational flow rate decreases, the dye infiltrates deeper into the bed of the pilot biofilter while also being evenly distributed radially due to the increased hydraulic retention time in the system. This increase in tracer concentration along the length of the pilot biofilter 35 cm below the surface at an experimental flow rate of 0.5 L/min is greater than those observed at higher flow rates. A study by McKie et al. (2019) found that most pollutants were removed within the top 45% of the filter, making sufficient fluid distribution in the top half of the biofilter imperative.

The extent of infiltration achieved is critical when determining operational flow rates for the treatment of contaminated water from the Stiebeuel River in the pilot-scale system. Lower flow rates should yield a greater extent of pollutant removal due to the deeper infiltration observed, which implies that more of the bed is active. An operational flow rate of 0.5 L/min allows for the greatest depth of fluid penetration, allowing nutrients and microorganisms present in the contaminated water deeper into the pilot biofilter. This should result in increased biological activity in the system and is imperative in the formation of aerobic, semi-aerobic, and anaerobic zones in the system.

Normalised residence time distribution

Figure 5 shows the normalised residence time distribution of the pulse tracer studies. Individual residence time distributions were obtained by applying Equation (2) to the concentration–time data collected from Outlet B. The normalisation was then carried out as per Equations (5) and (6). Θ represents the number of reactor volumes of fluid that have passed through the system, based on entrance conditions, in time, t.
Figure 5

Normalised residence time distribution at all experimental flow rates.

Figure 5

Normalised residence time distribution at all experimental flow rates.

Close modal

The sharp peaks observed in Figure 5 are indicative of plug flow behaviour in the system across all experimental flow rates. As the flow rate increases, the peaks of the residence time distributions are shorter and broader, meaning that there is a greater degree of mixing in the axial direction at higher flow rates. The normalised residence time distributions at the higher flow rates (1.5, 2, and 3 L/min) have long tails, which are indicative of non-ideal fluid behaviour such as dead zones, hold-up, and channelling. The peaks at 1.5, 2, and 3 L/min are not as well defined as their counterparts at the lower flow rates. The peaks at these flow rates also do not occur at θ = 1, which is a deviation from ideal plug flow behaviour and is characteristic of non-ideal fluid flow behaviour. The normalised residence time distribution at 0.5 L/min approximates the mixing characteristics of plug flow more so than the residence time distributions at the higher flow rates. The sharp, tall peak at θ = 1 and lack of a long tail is evidence of plug flow with minimal short-circuiting or dead zones.

Hydraulic characteristics

Table 2 shows the hydraulic characteristics of the pilot biofilter system, defined at the six experimental flow rates. The mean tracer recovery across all the pulse tracer studies that were conducted is 97.1%, which is acceptable. The mean tracer loss of 2.9% can be attributed to adsorption onto the surface of the stones.

Table 2

Hydraulic characteristics of the pilot biofilter at various flow rates

3 L/min2 L/min1.5 L/min1 L/min0.75 L/min0.5 L/min
Tracer recovery (%) 95.1 98.7 98.9 96.3 96.9 96.4 
(min) 75 112.5 150 225 300 450 
(min) 68.1 103 136 208 280 434 
(min239.4 133 174 188 201 175 
e (%) 90.8 91.6 90.7 92.5 93.4 96.4 
Λ (%) 90.1 90.5 89.8 92.1 93.2 96.3 
N 118 93 106 230 391 1,070 
Pe 215 143 205 401 584 1,410 
D 4.65 × 103 6.99 × 103 4.88 × 103 2.49 × 103 1.71 × 103 7.09 × 104 
3 L/min2 L/min1.5 L/min1 L/min0.75 L/min0.5 L/min
Tracer recovery (%) 95.1 98.7 98.9 96.3 96.9 96.4 
(min) 75 112.5 150 225 300 450 
(min) 68.1 103 136 208 280 434 
(min239.4 133 174 188 201 175 
e (%) 90.8 91.6 90.7 92.5 93.4 96.4 
Λ (%) 90.1 90.5 89.8 92.1 93.2 96.3 
N 118 93 106 230 391 1,070 
Pe 215 143 205 401 584 1,410 
D 4.65 × 103 6.99 × 103 4.88 × 103 2.49 × 103 1.71 × 103 7.09 × 104 

Note: is theoretical residence time, is the mean residence time, is the variance, e is the effective volume utilisation, Λ is the hydraulic efficiency, N is the number of tanks in series, Pe is the Peclet number, and D is the dispersion number.

In each instance, the mean residence time in the system, , is lower than the theoretical residence time, . This is proof of non-ideal flow behaviour in the pilot biofilter system as for ideal flow behaviour. In general, the effective volume utilisation, e, and hydraulic efficiency, Λ, are greater at the lower flow rates. The greatest effective volume utilisation of 96.4% and greatest hydraulic efficiency of 96.3% are seen at 0.5 L/min.

The number of tanks in series, N, is greatest at a flow rate of 0.5 L/min, with the number of tanks in series being 1,070. This implies that the extent of plug flow behaviour and degree of mixing in the radial direction increases with a decrease in flow rate. This is supported by the fact that the number of tanks in a series is lower at the higher flow rates.

The Peclet number, Pe, is an indicator of whether the flow is governed by radial advection or dispersion in the axial direction. The large Pe indicates that there is a majority of radial advection occurring in the system. Pe is greatest at a flow rate of 0.5 L/min and considerably lower at the higher flow rates, implying that Pe → ∞ as the flow rate decreases. This shows that the extent of dispersive mass transfer decreases as the flow rates decrease, even though the flow is governed by advection at all the flow rates.

Furthermore, the linear regression in Figure 6 displays a negative correlation between the flow rate and Pe, supported by the observation that lower flow rates result in higher Pe while higher flow rates result in lower Pe. The R2 value of 0.478 and R value of −0.69 indicate a moderate negative correlation between Pe and flow rate.
Figure 6

Pe vs flow rate for all experimental runs.

Figure 6

Pe vs flow rate for all experimental runs.

Close modal

The dispersion number (D) for all experimental flow rates, as seen in Table 2, is relatively low and approaches zero as the flow rate decreases, indicating that axial dispersion is minimal in the system (Fogler 2011). The D values presented in Table 2 are approximately 10 times lower in magnitude than the D values reported by von Sperling et al. (2023) for full-scale constructed wetlands (>2.5 m2). The lower dispersion numbers observed in this study can be attributed to the smaller scale of the system.

It is expected that fluid velocity profiles should be similar across each differential ‘slice’ of an ideal plug flow system; however, this is not the case for the pilot-scale biofilter. The dye concentration profiles at 5 cm below the surface (Figure 3) and at 35 cm below the surface (Figure 4) show differing concentration profiles, indicating potential deviation in fluid velocity across the cross-sectional area of the pilot-scale biofilter. This may speak to the influent infiltration system design being insufficient. However, this could be a result of insufficient dye injection into the system and a consequent dilution effect, which is more likely. An example of an inadequately designed infiltration and drainage system is illustrated in the findings of a study by Bonner et al. (2017), where a pilot-scale non-vegetated constructed wetland filled with gravel displayed mixed flow characteristics and a hull-shaped flow profile.

While the dye concentrations differ across the depth of the filter in this study, the fluid flow characteristics of the pilot biofilter system approximate plug flow with the extent of plug flow behaviour increasing as the system flow rate decreases. This is comparable with a study on an adequately designed pilot-scale horizontal subsurface flow constructed wetland filled with coarse gravel by Dittrich et al. (2021), which found that the flow pattern in the system approximated plug flow behaviour. Additionally, the findings of this study align with studies by Wang et al. (2014), and Okhravi et al. (2017) demonstrated that lower inlet flow rates could reduce the short-circuiting in non-vegetated subsurface flow constructed wetlands, decreasing the volume of dead zones and increasing the active area.

A maximum dead space of 9.3% is seen at a flow rate of 1.5 L/min, while a minimum dead space of 3.6% is seen at a flow rate of 0.5 L/min in Table 2. It is expected that a flow rate of 0.5 L/min will result in more effective pollutant degradation in future pollutant removal studies due to the depth of infiltration achieved and the mixing characteristics observed in the system. It can be assumed that the field-scale biofilters at the Water Hub have similar hydraulic characteristics at the same operational flow rates. It is, therefore, recommended that the field-scale biofilters at the Water Hub biofilters be operated at a low flow rate to avoid inefficiency in the system. However, this will severely limit the treatment capacity of the Water Hub biofilters due to the low throughput rate.

The pilot biofilter approximates plug flow behaviour at all six experimental flow rates with an effective volume utilisation of at least 90.8%. The effective volume utilisation and hydraulic efficiency of the system increase with a decrease in flow rate, with the highest hydraulic efficiency of 96.3% being observed at 0.5 L/min. Dye concentrations across the depth in the pilot biofilter deviate from expected uniformity, with concentrations at 35 cm below the surface being lower than those observed at 5 cm below the surface, potentially influenced by the activity of the infiltration pipe at the inlet or dilution to the lower bed.

The pulse tracer studies offer valuable insights for treating contaminated surface water from the Stiebeuel River using Water Hub's field-scale biofilters. Concentration profiles at 5 and 35 cm depths reveal deeper tracer infiltration at lower flow rates, suggesting a preference for such rates to maximise nutrient dissemination and microbial activity. It is recommended to replace the single infiltration inlet with multiple inlets at various depths in the Water Hub biofilters for enhanced treatment efficiency. Despite determining optimal flow rates, addressing highly variable water quality in informal settlements remains challenging. The impact of hydraulic characteristics on treating Stiebeuel River water, especially regarding nitrogen compounds and phosphate removal, requires further investigation. In the broader context of tackling water contamination in informal settlements, this research contributes essential knowledge for practical, efficient, and cost-effective solutions.

All relevant data are available from an online repository at https://figshare.com/s/f64b77047dacf7064b01.

The authors declare there is no conflict.

Armitage
N.
,
Vice
M.
,
Fisher-Jeffes
L.
,
Winter
K.
,
Spiegel
A.
&
Dunstan
J.
2013
Alternative technology for stormwater management: Sustainable drainage systems
.
Water Research Commission
,
76
86
.
Bodin
H.
,
Persson
J.
,
Englund
J. E.
&
Milberg
P.
2013
Influence of residence time analyses on estimates of wetland hydraulics and pollutant removal
.
Journal of Hydrology
501
,
1
12
.
https://doi.org/10.1016/j.jhydrol.2013.07.022.
Bonner
R.
,
Aylward
L.
,
Kappelmeyer
U.
&
Sheridan
C.
2017
A comparison of three different residence time distribution modelling methodologies for horizontal subsurface flow constructed wetlands
.
Ecological Engineering
99
,
99
113
.
https://doi.org/10.1016/j.ecoleng.2016.11.024.
Bonner
R.
,
Aylward
L.
,
Kappelmeyer
U.
&
Sheridan
C.
2018
Combining tracer studies and biomimetic design principles to investigate clogging in constructed wetlands
.
Water SA
44
(
4
),
764
770
.
https://doi.org/10.4314/wsa.v44i4.24
.
Bratieres
K.
,
Fletcher
T. D.
,
Deletic
A.
&
Zinger
Y.
2008
Nutrient and sediment removal by stormwater biofilters: A large-scale design optimisation study
.
Water Research
42
(
14
),
3930
3940
.
https://doi.org/10.1016/j.watres.2008.06.009.
Chandrasena
G. I.
,
Shirdashtzadeh
M.
,
Li
Y. L.
,
Deletic
A.
,
Hathaway
J. M.
&
McCarthy
D. T.
2017
Retention and survival of E. coli in stormwater biofilters: Role of vegetation, rhizosphere microorganisms and antimicrobial filter media
.
Ecological Engineering
102
,
166
177
.
Dittrich
E.
,
Klincsik
M.
,
Somfai
D.
,
Dolgos-Kovács
A.
,
Kiss
T.
&
Szekeres
A.
2021
Analysis of conservative tracer measurement results inside a planted horizontal subsurface flow constructed wetland filled with coarse gravel using Frechet distribution
.
Environmental Science and Pollution Research
28
,
5180
5240
.
https://doi.org/10.1007/s11356-020-10246-9/Published.
Fell
J.
2017
An Analysis of Surface Water from an Informal Settlement, Langrug, Franschhoek: Down a Slippery Slope
.
MSc Dissertation
,
Department of Environmental and Geographical Science, University of Cape Town
.
Ferreira
A. G.
,
Borges
A. C.
&
Rosa
A. P.
2021
Comparison of first-order kinetic models for sewage treatment in horizontal subsurface-flow constructed wetlands
.
Environmental Technology (United Kingdom)
42
(
28
),
4511
4518
.
https://doi.org/10.1080/09593330.2020.1769741
.
Fogler
H. S.
2011
Essentials of Chemical Reaction Engineering (International)
.
Pearson Education
,
Boston, MA
.
Gajewska
M.
&
Skrzypiec
K.
2018
Kinetics of nitrogen removal processes in constructed wetlands
.
E3S Web of Conferences
26
.
https://doi.org/10.1051/e3sconf/20182600001
.
Malulu
I. C.
2016
Opportunities for Integrating Sustainable Urban Drainage Systems (SuDS) in Informal Settlements as Part of Stormwater Management
.
Available from
: https://scholar.sun.ac.za.
McKie
M. J.
,
Bertoia
C.
,
Taylor-Edmonds
L.
,
Andrews
S. A.
&
Andrews
R. C.
2019
Pilot-scale comparison of cyclically and continuously operated drinking water biofilters: Evaluation of biomass, biological activity and treated water quality
.
Water Research
149
,
488
495
.
https://doi.org/10.1016/j.watres.2018.11.033
.
Narayanan
C. M.
&
Narayan
V.
2019
Biological wastewater treatment and bioreactor design: A review
. In:
Sustainable Environment Research
, Vol.
1
(
1
).
BioMed Central Ltd
.
https://doi.org/10.1186/s42834-019-0036-1
.
Nivala
J.
,
Puigagut
J.
,
Von Sperling
M.
,
Dotro
G.
,
Stein
O.
,
Molle
P.
&
Langergraber
G.
2017
Treatment Wetlands
, 1st edn, Vol.
7
.
IWA Publishing
,
London, UK
.
Okhravi
S.
,
Eslamian
S.
&
Fathianpour
N.
2017
Assessing the effects of flow distribution on the internal hydraulic behavior of a constructed horizontal subsurface flow wetland using a numerical model and a tracer study
.
Ecohydrology & Hydrobiology
17
(
4
),
264
273
.
https://doi.org/10.1016/j.ecohyd.2017.07.002.
Ramírez-Agudelo
N. A.
,
Anento
R. P.
,
Villares
M.
&
Roca
E.
2020
Nature-based solutions for water management in peri-urban areas: Barriers and lessons learned from implementation experiences
. In:
Sustainability
, Vol.
12
(
23
).
MDPI AG
, pp.
1
36
.
https://doi.org/10.3390/su12239799
.
SERI
2018
Informal Settlements and Human Rights in South Africa
.
Available from
: https://www.blacksash.org.za/.
Taing
L.
,
Armitage
N.
,
Ashipala
N.
,
Spiegel
A.
,
2013
In:
Sanitation Services in Informal Settlements – Sewering Lessons From Western Cape
(
Naidoo
V.
&
Mwale
J.
eds.).
Water Research Commission
.
https://doi.org/WRCReport No.TT557/13
.
von Sperling
M.
,
Wallace
S. D.
&
Nivala
J.
2023
Representing performance of horizontal flow treatment wetlands: The Tanks In Series (TIS) and the plug flow with dispersion (PFD) approaches and their application to design
.
Science of the Total Environment
859
,
160259
.
https://doi.org/10.1016/j.scitotenv.2022.160259.
Wang
Y.
,
Song
X.
,
Liao
W.
,
Niu
R.
,
Wang
W.
,
Ding
Y.
,
Wang
Y.
&
Yan
D.
2014
Impacts of inlet–outlet configuration, flow rate and filter size on hydraulic behavior of quasi-2-dimensional horizontal constructed wetland: NaCl and dye tracer test
.
Ecological Engineering
69
,
177
185
.
https://doi.org/10.1016/j.ecoleng.2014.03.071.
Wang
R.
,
Xu
L.
,
Xu
X.
,
Xu
Z.
,
Zhang
X.
,
Cong
X.
&
Tong
K.
2021
Hydraulic characteristics of small-scale constructed wetland based on residence time distribution
.
Environmental Technology (United Kingdom)
https://doi.org/10.1080/09593330.2021.1994018.
Winter
K.
,
Mgese
S.
,
Nicklin
E.
&
Maraj
K.
2023
Treating and reusing polluted runoff from an informal settlement
.
South Africa. Water Practice and Technology
18
(
4
),
796
809
.
https://doi.org/10.2166/wpt.2023.045
.
Zhang
J.
,
Wu
H.
,
Xu
J.
,
Fan
J.
,
Liu
H.
,
Liang
S.
,
2016
Green technologies for sustainable water management
. In:
Green Technologies for Sustainable Water Management (First)
(
Ngo
H.
,
Guo
W.
,
Surampalli
R.
&
Zhang
T.
eds).
American Society of Civil Engineers
,
Reston, VA
.
Zhong
H.
,
Li
P.
,
Guo
X.
&
Zhang
C.
2020
The influence of particle sizes and inlet-Outlet configuration on the hydraulic characteristics of horizontal subsurface constructed wetlands
.
IOP Conference Series: Materials Science and Engineering
735
(
1
).
https://doi.org/10.1088/1757-899X/735/1/012082
.
This is an Open Access article distributed under the terms of the Creative Commons Attribution Licence (CC BY 4.0), which permits copying, adaptation and redistribution, provided the original work is properly cited (http://creativecommons.org/licenses/by/4.0/).