Downscaling reverse osmosis for single-household wastewater reuse: towards low-cost decentralised sanitation through a batch open-loop configuration

Recycling water from municipal wastewater is critical for water-scarce regions and is reliant upon reverse osmosis (RO) for producing high-purity recycled water (Bartels et al. 2005; Malaeb & Ayoub 2011). To date, water recycling has been implemented through large-scale, centralised infrastructure within high-income countries. However, there is a significant demand for water recycling in low-income countries, where infrastructure is primarily decentralised, with an emphasis on pit latrines that serve as intermediate storage facilities and are prone to flooding and leaching into local groundwater resources (Mercer et al. 2021). Due to the significant costs associated with pit emptying, unregulated discharge of untreated wastewater is prevalent, resulting in increased enteric pathogen contamination of local water resources and high morbidity rates worldwide (WHO/UNICEF Joint Water Supply, and Sanitation Monitoring Programme 2015). Enabling water recycling from wastewater at the single-household scale in low-income countries, therefore, provides the following three benefits: (i) improving local water resources through reducing enteric pathogen contamination; (ii) reducing pit latrine emptying costs, by reducing frequency and volume of pit latrine emptying; and (iii) creating a higher quality water product than is afforded from local sources of unknown provenance (Kamranvand et al. 2018).

RO has been successfully scaled down to a single-household application for drinking water, where the feed pressure provided by the mains supply typically offsets the need for a feed pump, while the relative low salinity of the feedwater imparts a negligible osmotic pressure, resulting in high-quality water production, and a good recovery ratio (RR) from a single element configuration. The blackwater produced prior to discharge into a pit latrine is typically concentrated due to the use of low flush volumes, or the adoption of dry toilet systems, which is increasingly popular due to water scarcity (Mkhize et al. 2017) and ethical considerations. Consequently, urine is the dominant liquid fraction, creating a saline wastewater matrix, comprising a broad range of inorganic and organic salts, low-MW organic compounds such as urea, and higher MW organics including secreted proteins and sugars (Kamranvand et al. 2020). The separation mechanisms imposed by the dense semi-permeable RO membrane (e.g., size-exclusion, charge exclusion, and other physico-chemical interactions) have been demonstrated to reject a range of nutrients and micropollutants present in household wastewater (Van Voorthuizen et al. 2005; Yoon & Lueptow 2005; Garcia et al. 2013) and urine (Ek et al. 2006), however, concentrated blackwater (or faecally contaminated urine) increases complexity as water quality can vary temporally, through dynamic hydrolysis reactions (Kamranvand et al. 2020). Concentrated blackwater also raises the osmotic pressure, where a transmembrane pressure (TMP) of 50 bar has been estimated to achieve an 80% RR from urine (Maurer et al. 2006). The membrane element configuration, and pump duty required to deliver high water recoveries from blackwater at a competitive cost are therefore considerably more challenging than for drinking water at a single-household scale.

Large-scale RO water recovery facilities are designed in a ‘single-pass’ plug flow arrangement (Figure 1), where a stage, comprising multiple RO elements, each with a conversion efficiency of ∼15%, are arranged in series, while multiple stages with intermediate pumping are used to increase water recovery from the concentrate (Elimelech & Phillip 2011). However, several practical limitations exist for this configuration to recover water at a single-household scale: (i) significant variability in feed composition, will introduce a considerable inconsistency in product recovery; (ii) high RRs in single-pass will expose specific elements to high solute concentrations, augmenting fouling; and (iii) the cross-flow and geometric restrictions of commercial RO element design yield around 9% product recovery per 100 cm element length (40″) (Singh 2015), making system scalability and cost practically unfeasible. A batch-configured RO instead comprises a single stage, as the required RR is achieved by recirculating the concentrate (Figure 1), which is advantageous in handling variation in feedwater composition, while the shorter element contact time with the final high-recovery concentrate has been shown to limit fouling, particularly when high RRs are sought within a low-volume process (Warsinger et al. 2016). A single-stage design can also improve the cost of an RO sanitation solution, which must meet an affordability index of US$0.03–0.05 person⁻¹ day⁻¹ to be considered economically viable (Tobias et al. 2017). However, repressurisation of the recirculating concentrate is energetically costly, and while a few process designs have been proposed to recover or maintain pressure in the recirculation loop (e.g., closed circuit RO) (Gray 2002; Qiu & Davies 2012; Warsinger et al. 2016; Werber et al. 2017), such interventions necessarily increase capital cost for pumping which are already sensitive due to the osmotic pressure of blackwater and the scale of implementation.

This study, therefore, seeks to explore the technical and economic feasibility of a batch open-loop RO configuration for single-household scale water recycling from blackwater. The system analysis is conducted experimentally on incoming flow rates comparable to a full-scale system (approximately six people) with a critical focus on three key objectives: (i) evaluate a single-stage design to minimise depressurisation/repressurisation during recycling to reduce energy losses by increasing product recovery per pass through multi-element configuration and hydrodynamic conditions; (ii) characterise permeate to establish the influence of concentrated blackwater on water quality by operating in open-loop batch recycle at high RRs; and (iii) conduct cost analyses of the finalised batch open-loop configuration to determine key cost drivers, and feasibility of RO as a cost-effective sanitation solution for water recycling from blackwater at a single-household scale.

The batch open-loop RO system (Figure 2) comprised 1812 membrane elements between 1 and 3 (Table 1) configured in series or in parallel to evaluate product recovery and pressure drop within a single-stage design. A 19 L stainless steel vessel was used to retain the feed, with the concentrate recycled using a high-pressure diaphragm pump (M-03S, Hydracell, USA) controlled through a variable frequency drive (Control Techniques M200, Sterlitech, USA). The test setup was complemented with a pressure relief valve (200–1,500 psi, Swagelok, UK), pressure gauges (0–100 barg, Swagelok, UK), pressure regulators (Sterlitech, USA), and a variable area flow meter (0.5–7 L min⁻¹, Sterlitech, USA). A heater/chiller (Grant TC120-R4, Wolflabs, UK) connected to a stainless steel recirculation coil placed in the feed vessel maintained the feed temperature at 25 °C. Before first use, the 1812 elements were flushed with deionised (DI) water at 3 L min⁻¹ for 30 min to remove the preservative solution and conditioned at 30 bar for 1 h with DI water. For a standard experiment, 10 L of feed solution was added to the feed vessel and allowed to equilibrate at 25 °C before initiating testing.

Chemicals were of reagent grade (Fisher, UK; Sigma Aldrich, UK). Human urine and faeces were collected from anonymous donors with informed consent through a process approved by the Cranfield University Research Ethics System (CURES: 2310/207; 2407/2017). The composition and flow of blackwater were estimated based on the mass balance of a single-household wastewater treatment system for six users. Average urine and faeces mass flows of 2.2 L person⁻¹ day⁻¹ and 0.026 kg person⁻¹ day⁻¹, respectively were assumed (Rose et al. 2015). The faeces mass flow was below that nominally assumed per capita (0.128 kg capita⁻¹ day⁻¹Rose et al. 2015) to account for upstream solid/liquid separation by the toilet interface, where 80% total suspended solids removal has been demonstrated with source separation (Kamranvand et al. 2021). Tap water was used to simulate flush water comprising 2.7 L person⁻¹ day⁻¹, equivalent to the flow rate from a low flush toilet. Before use, these solutions were combined and mixed with a magnetic stirrer at 800 rpm for 30 min, to produce a homogenised blackwater. A two-stage ultrafiltration (UF)-RO unit was employed, comparable to that implemented for single-household drinking water production. The blackwater was filtered through a UF membrane (Figure A1, MP 1018-102, Koch Separation Solutions, USA) with a nominal pore size of 0.03 μm (area, 0.07 m²) at a constant flux of 6 L m⁻² h⁻¹, the cross-flow velocity (CFV) of 10 cm s⁻¹, and an average TMP of <50 mbar. To increase data production, synthetic blackwater was prepared based on the chemical analysis of urine in the initial screening phase for set conditions (Table 2) (Putnam 1971). Solutions of NaCl with equivalent conductivities to the estimated wastewater were used for the initial identification of boundary conditions (urine, 0.2 M NaCl; equivalent toilet wastewater 0.056 M NaCl) (Mercer et al. 2019a), to decouple dynamic fouling effects from the examination of osmotic pressure (and gross pressure) requirements. Chemical oxygen demand (COD), total ammoniacal nitrogen (NH₄-N), total nitrogen (TN), and total phosphorous (TP) were determined using standard photometric test kits (Spectroquant^®; COD 114541; NH4⁺-N 114559; TN 114763; TP 114763; Merck, UK). Solution pH was determined with a pH probe (4330, Jenway, Stone, UK) calibrated to standard buffers (4, 7, and 11) before measurement. Conductivity was recorded using a conductivity meter (CDH SD1, Omega, UK).

The impact of feed pressure on product recovery and rejection in open-loop batch RO with a single element was studied using a synthetic saline feed, with a conductivity equivalent to a real concentrated blackwater (Figure 3). Increasing feed pressure improved water flux to achieve higher water productivity per pass, and reduced processing time from 10,317 to 2,190 s to achieve a RR exceeding 85%. Processing time and number of passes are important criteria for conversion, as this reduces the normalised pumping energy demand. Consequently, despite the higher feed pressure, operating at 30 bar reduced pumping energy from 21 to 12.5 kWh m⁻³ (Figure 3). Concentration polarisation, membrane resistance, and residence time can account for the significant difference between the theoretical separation energy (red dashed line, Supplementary Material, Appendix A) and actual energy demand. Higher operating feed pressures in open-loop batch RO configuration also improved rejection, particularly as RR increases (Figure 3), where the higher selectivity can be explained by the relative dependency of water permeability on pressure, whereas salt permeability is largely independent of TMP (Shrivastava et al. 2015). The loss in selectivity is exacerbated at lower pressures, which risks compliance failure on product quality (ISO 30500 2018) and as operating TMP approaches the osmotic pressure of the solution, permeability loss becomes significant.

Increasing the CFV slightly improved RO rejection at the initial stage of the batch cycle through diminishing concentration polarisation (Jang et al. 2019) but the salt rejection became comparable as feed concentration increased at higher RRs (Figure 4). Considerably higher product recovery per pass was identified by reducing the CFV, which extended element residence time, resulting in a 77% reduction in energy demand from 21 to 4.8 kWh m⁻³. However, an increased probability for longer term fouling/biofouling of the RO membrane from real blackwater will likely necessitate sustaining higher CFV, and so the cost of extending residence time at higher CFV through increasing number of elements per stage or accepting higher pumping costs for a single element approach must be considered.

The impact of stage design was investigated by increasing the number of elements in the open-loop batch RO system to improve product conversion per pass, and reduce the energy required for water production. Membrane elements in parallel were also investigated to simulate wider modules that improve gross product conversion per pass with a linearly scalable pressure drop (Figure 5). The time required to achieve >85% product recovery, decreased from 2,832 to 1,500 s and 960 s as elements were added in parallel. The velocity was fixed in each element; thus, it is the increase in effective cross-sectional area from 5.94 to 17.8 cm² which improves conversion through increasing flow rate. As such, scaling-up element width did not improve product conversion efficiency per pass, but instead reduced the number of passes required to achieve the desired concentration. Consequently, the normalised energy requirement for water recovery does not change (Figure 5), but the potential for concentration polarisation should reduce through the shorter path length (Maskan et al. 2000). However, the salt rejection was comparable to multiple elements in series, indicating that the specified CFV of 0.028 m s⁻¹ was sufficient to limit polarisation (Jang et al. 2019), while the difference in polarisation effects between configurations may be less evident in open-loop batch RO. Increasing the number of elements in series, therefore provided comparable processing times to achieve product recovery >85%, but increased product recovery per pass, in doing so reducing the energy required for product water recovery from 4.9 to 1.6 kWh m⁻³ for a one- and three-element stage, respectively (Figure 5).

An initial water flux of 54 kg m⁻² h⁻¹ for blackwater was higher than for the synthetic (47 kg m⁻² h⁻¹), but flux decline was similar for both feedwaters during the open-loop batch RO cycle, indicating that the reduction in flux with increasing RR was ostensibly due to the increase in osmotic pressure (Figure 6). Importantly, this flux establishes that a single 1812 element (0.23 m²) can enable water recovery from a full-scale system under these conditions (water capacity for six-person system 33 L day⁻¹), operating only 3 h every 24 h. Therefore, whereas a multi-element stage can improve energy efficiency in water recovery, capital cost will favour a single element design. Product water quality was compared against reference values set out within ISO30500 on non-sewered sanitation systems (American National Standards Institute (ANSI) 2016). Rejection for total phosphorus, pH, and COD met the cited reference values for discharge and reuse (Table 4). Escherichia coli was below the limit of detection in the RO permeate sample. This compares to an E. coli concentration in the UF feed of 3.44 log₁₀ CFU mL⁻¹, corresponding to a total log reduction value (LRV) of 5.44. A COD reduction of 96% was sufficient to achieve Category B status (discharge). The residual permeate sCOD is associated with high permeability low-MW uncharged organic molecules (e.g., urea) and volatile organic compounds (e.g., p-Cresol) (Mercer et al. 2019b), which may require a further polishing stage (e.g., granular activated carbon (GAC)) to achieve Category A status for reuse (American National Standards Institute (ANSI) 2016). While TN and COD rejection met ISO30500 reference values at the outset of batch cycling, both determinands fell below this specification following progressive product recovery. This can be accounted for by the decline in selectivity imposed by the reduction in flux following the progressive increase in osmotic pressure (Shrivastava et al. 2015), as evidenced by the increase in permeate ammonia coincident with the reduction in flux having achieved a RR >70%. This increase in transmission of free ammonia can also be attributed to the hydrolytic transformation of urea to ammonium, coupled with the progressive increase in pH towards the pKa, shifting the equilibrium of ammonium towards ammonia, which was observed in this study (Kamranvand et al. 2020). Rejection may be therefore improved by increasing feed pressure to rebalance selectivity but may also be controlled by reducing fluid residence time (storage) to constrain the hydrolysis of urea. System certification using ISO30500 will characterise rejection at a systems level (i.e., across both UF and RO). While this does not improve the net rejection for determinands which are poorly rejected by UF (e.g., ammoniacal nitrogen), this multi-barrier approach will improve overall conformance for wider determinands. Importantly, this study evidences that a single-stage batch open-loop RO can recover high-quality water from concentrated blackwater, creating an enabling sanitation solution.

The economic viability of the batch open-loop RO for low-income single-household application was characterised using annual cash flows stimulated by the sanitation affordability index (US$0.05 person⁻¹ day⁻¹) for a six-person system (US$109.5 y⁻¹), where a positive NPV (discount rate 6%) indicates an economically viable solution (Table 3). Critical capital cost items were identified and their significance to valuation weighted through sensitivity analysis. At a single-household scale, the dominant cost items are the feed pump and membrane (unit cost and replacement frequency over 15-year lifetime). While economies of scale are achieved for pumping in large-scale high salinity RO applications with multistage centrifugal pumps, pump selection for high-pressure low flow applications is considerably more difficult, deferring instead to metering pumps with a minimum unit price of ∼US$1,000 (Figure 7). Plunger pumps match the system curve (Figure 7) which are convenient for RO as the produced flow is independent of the head imposed (Green & Perry 2007), and have been specified for small-scale RO systems, but forms the largest contribution to system cost (Table 3). To achieve a break-even NPV at a fixed element cost of US$100, pump cost must be reduced to ∼US$250 (Figure 8), through material selection, pump simplification, or economies of scale, where material selection should consider corrosion and pressure limitations. Based on a pump cost of US$250, the significance of membrane replacement and unit cost have been evaluated (Figure 8). While RO element lifetime is important, break-even on NPV is more strongly weighted towards unit cost than replacement frequency. Due to the lower usage of 1812 RO membranes, unit cost is presently around US$150–200, which equates to ∼US$650 m⁻². However, due to widespread application and roll-to-roll processing, large-scale RO membrane costs are closer to ∼US$35 m⁻² (Lu et al. 2007). With a 3-year membrane lifetime and 1812 element cost of US$25 (∼US$109 m⁻²), a positive NPV of US$267 indicates that pump costs could be increased to more than US$500 while maintaining a break-even NPV. The additional investment made available to pumping, can therefore increase financial viability. A single 1812 element can also process the daily blackwater load from a single house in 3 h, suggesting that a lower membrane area element could be constructed to reduce cost, provided this does not risk robustness to sealing and the high operating pressures required.

In this study, a batch open-loop RO system has been investigated to enable clean water production from blackwater produced at decentralised scale. For single-stage batch processing, a multi-element configuration was most advantageous in reducing specific energy demand (kWh m⁻³) towards the theoretical minimum energy requirement for separation. However, limited flush water inclusion can reduce daily fluid volume to between 1.5 and 5 L capita⁻¹ day⁻¹. Consequently, the total daily volume is sufficiently low at a single-household scale (∼30 L day⁻¹ for six people) to impart a negligible gross energy demand, in which case the decision as to which configuration is best will be determined by cost. This low flow rate was complemented by high fluxes in real blackwater, suggesting a single 1812 element, commonly used for domestic drinking water production from tap water, is also sufficient for single-household blackwater treatment. The 1812 elements for low pressure RO (i.e., tap water) are US$30, around five times less than the high-pressure 1812 element cost. While slight modifications in the spiral wound element are required for mechanical strength (e.g., spacer selection and outer wrap), together with consideration of material compatibility for high concentration of free ammonia and other potentially corrosive components, it is conceivable that a similar price point can be achieved at comparable economies of scale of production. This price point falls within the sanitation affordability index and reduces cost sensitivity to element replacement frequency. However, the system cost is presently dominated by the high-pressure pump, requiring a reduction in the unit cost of around 50% to ∼US$500 to permit use at a household scale. Reciprocating pumps with comparable pressures and flows (100 bar, 8 L min⁻¹) have been designed for pressure washers by simplifying material and design (and market penetration) to achieve a unit cost of US$100, which when complemented with investment to reduce RO element cost, can deliver an affordable sanitation solution. For the batch open-loop RO, permeate quality sufficient for discharge is achieved. However, the osmotic retardation of flux, coupled with the dynamic transformation of nitrogen species at higher RRs (and extended residence times) were observed to slightly hinder water quality. Permeate quality did not achieve the reuse standard due to the transmission of sCOD. As with nitrogen, this may be improved by increasing feed pressure to enhance selectivity or incorporating a polishing step. Importantly, high-quality water was recovered from blackwater and at a cost near to the affordability index for sanitation. While the value of this water maybe in its reclamation, the direct discharge of this water provides for a safer environment, and through promoting a large reduction in stored volume within pit latrines, creates an indirect economic benefit in reducing pit emptying costs (US$35–95 household⁻¹ annum⁻¹). The focus of this research was to examine whether RO could offer a viable solution for decentralised water reclamation. Having now confirmed the potential, further research is required to evidence long-term robustness, particularly with reference to scaling and fouling mechanisms, to identify suitable long-term operational strategies. While the feed matrix is complex, the redox environment, elevated pH and high feed ammonia concentrations have demonstrated inhibition in a wide range of microorganisms, which may reduce biofilm growth ordinarily observed in classical applications.

This research was funded by the Bill & Melinda Gates Foundation via Georgia Tech Research Corporation (Funding ID: INV-002230). The findings and conclusions contained within are those of the authors and do not necessarily reflect positions or policies of the funders.

All relevant data are available from an online repository or repositories. Data underlying this paper can be accessed from: http://dx.doi.org/10.17862/cranfield.rd.16674337.