The work commences in the upper region of The Gambia River with UF. Along the river, water becomes more saline thus requiring NF/RO. At the end of the journey water reuse of the Kotu WWTP is considered.
Treatment of the freshwater region of The Gambia River with ultrafiltration
To evaluate to what extent UF can treat turbidity and OM from the freshwater region (Kuntaur TDS 37 mg/L and Kudang TDS 155 mg/L) of The Gambia River, different MWCO UF membranes were investigated (Fig. 1). Kuntaur and Kudang samples, which can be found around 230–250 km up from the river mouth, were selected as these are situated in the permanent freshwater region 10. Flow rate and temperature data are summarised in Figure S6.
UF of all MWCO (10-1000 kDa) has generally achieved turbidity below 0.5 NTU and reduced the OM from Kuntaur and Kudang samples to a TOC of around 1.2 mg/L which is the median worldwide value for groundwater. At this TOC concentration, a minimum of 0.3 mg/L of free chlorine is required to avoid bacterial regrowth 83. Free chlorine reacts with OM to produce DBPs 84, which can be controlled with enhanced OM removal. Chemical addition can be challenging to implement in such tropical or remote regions. Alternatively, solar-based disinfection technologies can be used in countries that receive high solar irradiation, and ultimately high UV irradiation that drives the disinfection process 85. This is the case of The Gambia which has a global horizontal irradiation of 5.7 kWh/m2 86 and is located in the region of very high to extreme UV radiation index (8–11+) 87.
To characterise the OM fractions in Kudang and Kuntaur and their removal by UF, feed and permeate samples were analysed by liquid chromatography–organic carbon detection (LC-OCD). Specific UV absorbance (SUVA), which is the ratio of UV absorbance at 254 nm to TOC value, was determined to evaluate the effect on the aromaticity 88 (Fig. 2). In the context of treatability, characterising OM fractions and aromaticity was essential to elucidate the potential DBPs that can be formed when chemical disinfectants are added 89.
The dominant OM fraction in Kuntaur and Kudang samples was humic substances (HS) (4.5–30.4 kDa 90), followed by low molecular weight neutrals (LMWn) (< 350 Da 35) and building blocks (BB) (around 380 Da 35) (Fig. 2A, B). This was in agreement with the previous findings on the fractions of the OM in The Gambia River, including Kudang and Kuntaur 10, which is typical for rivers in tropical environments 91. Regarding removal, an insignificant decrease in the concentration of the different OM fractions from the feed to the permeate was observed for both samples regardless of the MWCO of the UF membrane. SUVA increased significantly in the UF permeate compared to the feed (Fig. 2C, D). This could be attributed to the removal of the fractions having low aromaticity by the UF. The presence of aromatic fractions with SUVA > 2 L/mg.m can form unknown DBPs 27.
UF, which is a well-established technology that can achieve up to 6-log bacteria removal depending on the pore size 25, can be adopted for drinking water supply to the communities located in the permanent freshwater regions of The Gambia River. This is technically possible through decentralised systems, where the treated water can be used for immediate consumption after UF directly without post-disinfection or with UV/solar-based disinfection to mitigate microbial regrowth during storage. For the implementation of such decentralised water treatment systems, acceptance may extend beyond just technology performance and reliability, and instead depend on the perceptions of the technology users 92. At first, the reliability of the UF process, particularly in terms of membrane fouling, must be evaluated.
During the treatment of Kuntaur and Kudang samples, characterised by high turbidity (98–140 NTU), UF fouling resulting from deposits was expected. This was monitored through transmembrane pressure (TMP) variations during UF operation at a constant flux (260 L/m2.h) (Fig. 3).
At the constant flux operation, TMP increased as membrane deposits formed. This is typical fouling behaviour of constant flow dead-end filtration 93. TMP increased for both samples (Kudang and Kuntaur) with very similar trends, even though the membrane deposits had a different appearance (images in Fig. 3). This difference in appearance could result from different size distributions and charges of the particles in both samples 36, 94. Hence, different interactions with the membrane and the OM present in the samples may have occurred, resulting in a particular combination of size and stickiness 95. The size and charge of particles were not measured. These proprieties are likely to change with the transport of the water samples due to aggregation.
Before the collected permeate volume of 50 mL for Kuntaur and 250 mL for Kudang, the TMP observed for the different UF membranes followed the MWCO (highest TMP with 10 kDa UF and the lowest with 500–1000 kDa UF). However, this correlation did not hold for the 100–1000 kDa UF membranes after the specified permeate volumes were collected. This may indicate different fouling mechanisms occurring. In constant flux dead-end operation, as in this case, small particles can cause standard pore blocking (internal blocking), while bigger ones can lead to complete/intermediate pore blocking or the formation of a cake layer 36. Deposition of a particle is complex and may be enhanced in the presence of OM that can act as a bridge between particles 96. The OM remaining in the concentrate for Kudang was lower than that of Kuntaur, even though the feed TOC was similar. This indicates that more deposit was formed which is confirmed by mass balance (Figure S7).
To evaluate the deposits and their chemical composition found on the UF membrane surface, scanning electron microscope-energy dispersive X-ray (SEM-EDX) analysis was carried out for the 100 kDa UF membrane used for filtrating Kuntaur water (Fig. 4).
The deposits observed on the UF, which formed a layer with about 5 µm thickness, were due to the turbidity of Kuntaur water sample. Turbidity in surface water is typically attributed to suspended matter such as clay particles, OM, and some microorganisms 97. However, on SEM images of the 100 kDa UF, no microorganisms (e.g. bacteria) could be detected, while the deposits were a mixture of particles and OM.
The chemical composition of the fouled UF membrane revealed high intensity in aluminium (Al; 1.5 keV) and silicon (Si; 1.7 keV). Al and Si are the main constituents of feldspar-type minerals which are one of the most common and abundant classes of minerals on Earth 98. These are commonly found in the sediments of tropical rivers, which originate from coarser materials 99. Al and Si were indeed found in the Kedougou region (in Senegal), which is part of The Gambia River catchment 100. Al, in particular, has been found in Kuntaur with a concentration of 51 µg/L 10. Sulfur (S; 0.2 keV), carbon (C; 0.3 keV), and oxygen (O; 0.5 keV) probably originate from the membrane material (polyethersulfone) and the deposited OM. In addition to OM, other nutrients, such as phosphorus (P; 2.1 keV) can be responsible for biofilm formation and growth on the membrane surface 101.
Drinking water supply from the freshwater region of The Gambia River is feasible as UF experiments have revealed. Desalination of the tidal and saline regions (with salinity up to 21 g/L) using NF/RO was subsequently investigated.
Desalination along The Gambia River by nanofiltration/reverse osmosis
Further downstream in the tidal saline region, tighter membranes are required for salinity removal to reach the recommended guideline by WHO (TDS < 1 g/L 23). Water samples, from freshwater and tidal saline regions of The Gambia River with a salinity range of 25 mg/L to 21 g/L and a turbidity range of 3 to 198 NTU, were treated with NF/RO membranes. Salinity and turbidity removal, as well as TMP and water permeability exhibited by the NF/RO membranes are shown in Fig. 5. Concentration polarisation (CP) parameters are shown in Figure S9, while filtration data are summarised in Figures S10-S14.
At controlled flux (77 L/m2.h for ≤ 0.8 g/L TDS, 38 L/m2.h for 1.7-4.0 g/L TDS, 77 L/m2.h for 10–21 g/L TDS; Table S6), the TMP increased with increasing TDS (Fig. 5A). For example, when the TDS of the Gambia River increased from ≤ 0.1 to 0.8 g/L, the TMP at 77 L/m2.h increased for all membranes. This was also the case for the TDS ranges 1.7–4.0 g/L (38 L/m2 h) and 10–21 g/L (19 L/m2 h), except for NF270, which showed similar TMP. Water permeability (Lp) of the NF/RO membranes decreased with salinity (Fig. 5B). This was due CP layer that induced a higher osmotic pressure near the membrane (Table S6), and ultimately a higher driving force was required to keep the controlled water flux through the NF/RO membranes. The CP layer acts as an additional resistance that hinders the movement of the water molecules to permeate through the NF/RO membranes 102. NF/RO fouling that occurred during the experiments (Figure S8), may have induced further resistance to the water permeability in the NF/RO membranes.
Turbidity removal of ≥ 98% was achieved with NF/RO for the water samples located in the TDS range ≤ 0.1-4.0 g/L (Fig. 5F). Downstream at the TDS range of 10–21 g/L, the turbidity removal was 90–95%. It should be noted that the limit of detection (LoD likely below 0.1 NTU; Figure S2) of the turbidity meter may have overestimated the permeate turbidity, which can explain the low turbidity removal with NF/RO. The recommended WHO guideline of 0.5 NTU was achieved even in turbid water (with up to 200 NTU) (Fig. 5D).
The WHO recommendation for TDS in drinking water (< 1 g/L corresponding to an EC < 1.6 mS/cm for The Gambia River; Figure S1) was achieved with the RO membrane designed for brackish water desalination (BW30) from a feed TDS ≤ 4 g/L and with the low permeability NF membrane (NF90) with a feed TDS ≤ 3 g/L, while the high permeability NF membrane (NF270) could only treat salinities up to 1.7 g/L (Fig. 5C). As expected RO can handle a higher salinity compared to NF. NF is attractive for the low salinity range where the target water quality can be achieved with lower energy consumption compared to RO 103. Powering NF/RO technology with solar energy is an appealing choice for water supply in The Gambia, particularly in rural locations where a reliable electricity grid is lacking. The implementation of these processes may be slow because deploying technology and achieving long-term acceptance in a community requires considering social and economic factors, as well as addressing technical challenges related to the operation, maintenance, and adaptation of the technology to local conditions 104.
The above NF/RO experiments were carried out without pretreatment of turbidity, which inevitably resulted in membrane fouling (Figure S8). To identify the nature and characteristics of the deposits on the NF/RO membranes, SEM-EDX analysis of the membrane surface was performed (Fig. 6).
Interestingly, the deposits look different from those of UF (Fig. 4). The membranes used for the high TDS water from James Island (Fig. 6A) showed a thick crust of flaky deposits on the membrane surface, potentially minerals and diatoms. In particular, the structure of the deposits observed on NF90 was similar to bacteria with oval and spin structures attached to the membrane surface. This is similar to the early stage of biofouling that occurs within 1–24 h of contact 105. Biofouling growth can be further enhanced with the presence of nutrients 101.
The elemental surface characterisation indicated the presence of Al and Si on the NF/RO membranes, which further confirms that the turbidity of the river was characterised by particles of feldspar-type minerals as previously observed in the UF deposits. However, Al and Si signals decreased when moving downstream (from Farafenni to James Island) and correlated with the decrease in turbidity. The reduction of turbidity when moving downstream can be explained by the charge screening of particles at higher salinities which enhance aggregation and sedimentation 106. Desalination of such turbid waters with NF/RO requires a pretreatment step to enhance the reliability of NF/RO, avoid water permeability decline due to fouling, increase the lifetime of the NF/RO membranes, and reduce membrane cleaning frequency 82. This can be achieved by UF, where a silt density index (SDI) < 2, a fouling index used to estimate the rate of colloidal and particle deposition, and turbidity < 0.1 NTU can be achieved 24.
Water reuse using NF/RO, focusing on MP removal in the river mouth, will subsequently be investigated.
Water reuse with nanofiltration/reverse osmosis
Towards the coast, at the river mouth, The Gambia River water is as salty as seawater, and treatment is very energy intense. For example, a large-scale state-of-the-art RO system exhibits exhibit a specific energy consumption from 3.5 to 4.5 kWh/m3 to desalinate seawater (35 g/L TDS) at 50% recovery 52. An alternative water supply is the treated effluent of the WWTP NAWEC Gambia Kotu ponds (Kotu-outlet), or the nearby stream where the WWTP effluent is discharged (Kotu-stream). In the previous work on the water quality of The Gambia River, pesticides, pharmaceuticals, steroid hormones, EDC, and PFAS were found in Kotu-stream, which receives the treated wastewater effluent 10. Subsequently, further samples were collected to assess the contribution of the WWTP to MP discharge. Higher MP contamination was expected in the WWTP effluent (Kotu-outlet), which was interesting for NF/RO treatment for reuse purposes. To evaluate to what extent NF/RO technology can remove MP, Kotu-stream (8.7 g/L) and Kotu-outlet (1.4 g/L) were treated with NF/RO membranes. Results are shown in Fig. 7, filtration data in Figure S15, and OM removal can be found in Figure S17.
As expected, the concentrations of MP in Kotu-outlet were higher compared to Kotu-stream with similarities in MP, while the majority of the MP types were detected in both samples. The exceptions were 3 PFAS (perfluorheptanoate (PFHpA), perfluoroctanoate (PFOA), perfluorbutansulfonate (PFBS)) detected in Kotu-stream with a concentration up to 6 ng/L and not in Kotu-outlet. Concentrations for individual pesticides were higher than 0.1 µg/L, corresponding to The Gambia (NEMA) and EU standards for both surface water and drinking water quality, except for glyphosate in Kotu-stream. Most importantly, DEET, which was initially suspected to result from insect repellent used by the sampling team during the previous field trip 10, was present in both Kotu samples (25 µg/L in Kotu-outlet and 11 µg/L in Kotu-stream), when no insect repellent was used during the sampling. This could originate from the wastewater piped from the Kotu resort. DEET indeed commonly reaches surface water through wastewater discharge and can further accumulate in sediments 110. Most importantly, the total concentration of pesticides in Kotu-stream (17.7 µg/L) was higher than NEMA and EU environmental quality standards for surface water (0.5 µg/L for total pesticides) 107, 108. For hormones, concentrations were higher than the EU environmental quality standards for surface water, particularly estrone with concentrations (1.6–0.7 µg/L) which is around 4 orders of magnitude higher than the standard quality of surface water (3.6·10− 5 µg/L 108). Similarly, the concentrations of the EDC, bisphenol A, and nonylphenol, in both Kotu samples exceeded the EU environmental quality standards for surface water. Bisphenol A is known as a xenoestrogen that mimics the effects of estrogen 111, and is also an indicator of plastic waste 112.
MP removal with NF/RO, BW30, and NF90 membranes reduced the concentrations of most pesticides to below 100 ng/L, except for DEET in Kotu-outlet for all membranes and in Kotu-stream for NF90. This low removal of DEET from Kotu-outlet could be due to the high DEET concentrations (11–25 µg/L). NF270 membrane, with the highest water permeability and the smallest pore radius (Table S3), was only effective for the herbicide glyphosate and its derivative AMPA, as well as the fungicide metalaxyl. This was because of the low MW of these MP (170–279 Da) compared to benzotriazole (119 Da). NF270 membrane was not effective for pharmaceuticals removal, particularly ibuprofen and diclofenac, while NF90 and BW30 membranes were able to reduce the pharmaceuticals significantly in both Kotu feed samples. An exception when treating Kotu-outlet was paracetamol, where the concentration in the permeate of NF270 and NF90 membranes was higher than in the feed sample, and only a slight reduction was observed with BW30. No conclusive explanation could be provided for such observation since this was not the case with Kotu-stream, where NF90 reduced the concentration to 1.3 µg/L (from 8.9 µg/L with 85% removal of paracetamol), and BW30 to below the limit of quantification of the analytical method for paracetamol (LoQ 50 ng/L; Table S10). While cross-contamination during sampling or analysis could be one reason for the case of Kotu-outlet, the removal of paracetamol with NF90 and BW30 in Kotu-stream was within the range reported by Licona et al. 113. The removal of paracetamol in NF, primarily governed by size exclusion 113, can be reduced significantly in mixtures containing high concentrations of MP 114, such as in the case of Kotu-outlet.
While NF/RO membranes could achieve the drinking water standards set by WHO and EU regarding MP concentrations, it is not advisable to treat Kotu-outlet and Kotu-stream for direct potable water reuse, before thorough investigations on public acceptance and the risks related to process failures. A more realistic proposal could be onsite water reuse for toilet flushing and green space watering in the nearby Kotu resort, for which the Kotu WWTP was made. Water reuse from WWTP effluent for toilet flushing and watering the green space has been implemented for various resorts worldwide, such as in South Africa, Australia, Singapore, USA, and some European countries, such as Spain, Greece, Cyprus, France, and Italy 115. The implementation of a water reuse system for such purposes, however, requires encouragement from local regulatory policies as well as social and cultural acceptance of water reuse 116.
In addition to water quality and purposes, energy consumption is a crucial factor in the implementation of desalination and water reuse processes. Desalination of the tidal and saline regions and water reuse were subsequently compared in terms of energy requirements.
Energy requirements for desalination and water reuse opportunities in The Gambia
To further elucidate the supply purpose and the energy requirement of desalination and water reuse, TMP, permeate quality, measured as EC (a salinity surrogate), and the minimum (or theoretical) SEC (SECmin) were investigated for the NF/RO experiments. The experiments involved waters from the tidal/saline region (1.7–21 g/L TDS), as well as Kotu-outlet (1.4 g/L TDS) and Kotu-stream (8.7 g/L TDS), as shown in Fig. 8. Experimental SEC is not a reliable figure-of-merit for determining the energy requirements of NF/RO operations at the lab-scale cross-flow system, because of the powerful (oversized) pump for a very small membrane area which caused excessive energy consumption (Figure S14).
At the constant flux of 19 L/m2.h, treatment of Kotu-outlet (1.4 g/L TDS) and Kotu-stream (8.7 g/L TDS) with NF/RO membranes exhibited high TMP (16–21 bar) regardless of the water salinity (Fig. 8A). A similar pressure range was required to desalinate the tidal/saline region of The Gambia River with 15–21 g/L TDS using the low water permeability NF membrane (NF90) and the RO membrane (BW30) (Fig. 8A). The osmotic pressure at the membrane surface (πm) was around 1.5 bar when treating the Kotu-outlet, which intuitively would require low driving force to overcome πm (Fig. 8B). This high TMP requirement for Kotu samples was due to the significant reduction in water permeability (Figure S16) caused by fouling (Figure S18). Similarly to desalination, water reuse with NF/RO would necessitate a UF pretreatment, which generally requires ~ 0.3 Wh/L 117 and can avoid pressure increase for the NF/RO process, in the case of a constant flux scenario, caused by fouling.
The theoretical SEC (or SECmin) for desalination with NF90 and BW30 increased from 0.1 to up to 1.6 Wh/L with TDS, due to the increase in the pressure requirement to overcome the osmotic pressure of the feed solution (Fig. 8D). This was not the case for NF270, in which the SECmin did not reveal measurable variations with TDS, which could be due to the low rejection characteristics of this membrane. A SECmin of around ≤ 0.1 Wh/L was required for water reuse using NF/RO (Fig. 8D). The actual SEC for sea- and brackish water desalination with RO is typically ranges from 0.4 to 4 Wh/L 118, while municipal water reuse with RO is 0.3–2.4 Wh/L 119. These reported SEC for desalination and water reuse are higher than the obtained SECmin, simply because the SECmin is calculated based on the thermodynamic limit for salt-water separation assuming an ideal process (e.g. 100% pump efficiency) 120.
If water reuse was considered for watering the green space of the Kotu resort, the permeate quality, in terms of salinity, of NF90 and BW30 membranes met the FAO recommendation for irrigation without restriction (< 0.7 mS/cm 121) (Fig. 8C). This water quality could not be achieved when treating the Kotu-stream, nor the FAO recommendation with the moderate restriction that applies to non-food crop irrigation (< 0.7-3.0 mS/cm 121), such as urban green space watering.
Increasing water availability in The Gambia
The previous findings on harnessing the freshwater region of the Gambia River for drinking water supply, and exploring desalination and water reuse options are summarised in Fig. 9. The results provide opportunities to harness the freshwater region of The Gambia River, a major water resource of The Gambia, to augment water availability. Treating the pristine freshwater with UF can improve the water quality by removing turbidity and achieving partial organic matter removal. Fouling was observed during UF experiments at the scale, therefore a regular backwash strategy would need to be implemented in pilot-scale trials. Desalination of the saline/tidal region of The Gambia River, where salinity increases towards the river mouth, with NF/RO can enhance freshwater availability in the country. The results revealed that NF90 and BW30 membranes can achieve the recommended drinking water guideline of TDS (< 1 g/L TDS) when the feed water salinity is as high as 4 g/L, while NF270 can only treat waters with salinities below 2 g/L. The effluent of the WWTP (Kotu ponds) and the nearby contaminated stream connected to The Gambia River in the urban region near the river mouth provide a non-potable water reuse opportunity, in which NF/RO membranes were able to remove the majority of the detected MP, thus protecting the environment from discharge, while supplying low salinity water that can be used for non-food crop irrigation, such as watering the green space of Kotu resort. To implement such an option, future work on piloting on-site with community interactions is envisaged.
Scaling up the findings: Bridging local insights to global solutions
While providing valuable insights into integrating desalination and water reuse for water supply in The Gambia, this research has the potential to be extended beyond the local context and help address similar challenges in other countries worldwide.
The insights into the desalination of the tidal/saline region of The Gambia River with NF/RO are valuable to countries where rivers are exposed to seawater intrusion and freshwater resources are limited. These conditions are already found in rivers from neighboring countries of West Africa, such as the Senegal River (Senegal), and extended to coastal aquifers 122. Other worldwide large rivers such as Mekong (Asia), Valdivia (South America), and Euphrates (Middle East) are facing the same issue of salinity intrusion with less freshwater flowing downstream resulting from climate change and hydroclimatic extremes 123. While one can argue that desalination is not a cheap practice for lower-middle-income countries, principally due to the energy costs, decentralised desalination systems autonomously powered with renewable energy were extensively tested 53. Perhaps, more engagement of stakeholders is needed to promote the water-energy nexus 124.
On a worldwide level, regions experiencing significant urbanisation with limited to moderate access to wastewater treatment facilities, like the Banjul area, face quality issues with their water resources, especially surface water, with an anticipated 30% rise in pollution from urban wastewater by 2050 125, 126. In particular, Sub-Saharan Africa, where The Gambia is located, is increasingly becoming the dominant region of surface water pollution relative to current wastewater management practices 127 . In Sub-Saharan Africa, discharging untreated wastewater to water bodies, which serve as a breeding ground for infectious pathogens, is the major cause of waterborne disease outbreaks 128 and significant environmental degradation. Wastewater reuse is the most promising approach to increase water availability to meet the growing water demand 129 , while conserving the environment 130 . Potable water reuse and non-potable water reuse, particularly irrigation, have been implemented in some Sub-Saharan African countries, such as Namibia, South Africa, and Nigeria 131 , using membrane processes such as UF and RO 132 . NF/RO technology can enhance the quality of treated wastewater effluents, through desalination and MP removal. Such technology can be used in decentralised approaches for small communities, while water reuse is promoted towards a circular economy 133 . For large-scale systems, the cost of water reuse was estimated at around 0.3–0.8$/m3 134, compared to 0.6-1$/m3 for sea- and brackish water desalination 135 , while the cost increases with decreasing scale and depends on local circumstances.