Long-term trial of a community-scale decentralized point-of-use drinking water treatment system

Two billion people lack access to safely managed drinking water services, many of these are in low/middle income countries where centralised systems are impractical. Decentra-lised point-of-use drinking water treatment systems offer alternative solutions in remote or resource constrained settings. The main aim of this study was to assess the long-term (3 year) operation and performance of a point-of-use drinking water treatment system (POU-DWTS). A biologically contaminated urban drainage pond was used as a water source and the quality of the produced drinking water was assessed over two independent trials. The decentralised POU-DWTS combined ultrafiltration membranes with disinfection from electrochemically generated hypochlorous acid (HOCl). The operational parameters, such as flow rate, free available chlorine and transmembrane pressure, were monitored in real-time and recorded via a remote monitoring system. Water quality from the source and treated water was assessed over two trial periods within the 3-year operational trial: an 11-week period at the start and a 22-week trial at the end. All water samples were assessed for a range of basic, chemical, microbiological and metal water quality parameters. The results demonstrate that the decentralised POU-DWTS is capable of continuously producing high quality drinking water when HOCl is continuously used to dose water prior to entering the ultrafiltration [UF] membranes. Over the 3-year operational study, the continuous dosing of HOCl pre-UF membranes resulted in stable permeability, indicating no occurrences of irreversible biofouling within the UF membranes and that good membrane ‘health’ was maintained


Introduction
It was reported in 2020 that 5.8 billion people (74% of the global population) had access to safely managed drinking water (located on premises, free from contamination and available when needed), partly due to the work undertaken towards achieving the Millennium and Sustainable Development Goals [SDGs] [1,2].However, this leaves almost 2 billion people without access to safely managed water services, with 122 million people relying solely on surface waters [2,3].Consumption of biologically contaminated water can potentially lead to diarrhoeal and other waterborne diseases, but this can be prevnted through relatively simple drinking water treatment interventions such as filtration (e.g.sand or ultrafiltration) and disinfection (e.g.chlorine or ultraviolet [UV]) [4,5].However, low and low-middle income countries (LIC and LMICs) are often disproportionately affected by a lack of access to safely managed drinking water services.Whereby, only 29% of people in LICs and 58% of people in LMICs have access [2,[6][7][8], in part due to the uneven global distribution of freshwater [9].
The two main challenges concerning the production of potable water are quality (biological and chemical) and quantity; managing both is required to ensure "availability and sustainable management of water for all" in line with United Nations SDG 6 [10].In high/middle income countries, access to biologically safe drinking water typically comes from centralised treatment systems that supply households through vast distribution networks [11], or through the use of household/individual disinfection or filtration processes [12,13].There is a current technology gap for decentralised treatment options at a community scale.Hence, research and development of off-grid or decentralised water treatment systems for the provision of safe drinking water has increased in recent years [14][15][16][17][18].These systems represent alternatives to centralised water treatment, which are often unfeasible in remote, rural or temporary community locations [14][15][16][17]19].The development of these decentralised systems are necessary as currently 3.4 billion people live in rural settings [20], of which 40% do not have access to safely manged drinking water services [8].
Several review articles have summarised studies that use point-of-use (POU) hybrid treatment solutions (e.g.filtration and disinfection) with the intention of applications in LICs or LMICs [6,7,17].Many of these decentralised POU treatment systems that have been developed utilise hybrid treatment processes and vary in complexity and capacity [17,21].Filtration techniques physically remove particulates through size exclusion to help reduce water turbidity, reduce concentrations of organic matter, and reduce microbial load, thus reducing the presence of potentially pathogenic microorganisms [17].However, filtration alone does not provide biologically safe water for consumption [18].Minimising organic matter prior to disinfection processes through filtration will increase disinfection capability, as well as reduce the potential formation of unwanted disinfection by-products [DBPs] [22].Integrated disinfection processes that occur after filtration allow for the production and maintenance of biologically safe water [23].Decentralised hybrid drinking water treatment systems often excludes fine membrane filtration, such as ultrafiltration, due to the increased propensity for biofouling to occur, resulting in non-operational periods [24,25].In addition, membrane manufacturers advise against continuous dosing of some disinfectants (e.g.OCl -or Cl 2 ) due to their oxidative nature that can negatively affect the membranes' material surface [25].
Electrochemical generation of chlorine disinfectants, such as hypochlorous acid (HOCl), has been widely employed in healthcare settings [26] and the food production industry [27,28] as it allows high quality disinfectants to be produced on demand and in-situ, with minimal resources required (NaCl, water and electricity).Electrochemically generated chlorine solutions, frequently referred to as electrolysed water (EW), electrochemically activated solutions (ECAS) and electrolysed oxidising water (EOW), can be generated through the electrolysis of a weak saline solution [29].Utilising such technology in a decentralised drinking water treatment system minimises the need to store and transport hazardous chemicals.
This study builds upon a feasibility investigation of a small-scale proof-of-concept decentralised drinking water production system (POC-DWPS) previously developed that combines multi-step filtration and disinfection processes [18].The system described previously has been scaled up, improved and manufactured to enable an increase in the volume of treated water produced, from 0.12 m 3 hr -1 to a maximum output of 0.75 m 3 hr -1 .Therefore, the main aim of this study was to assess the long-term (3 years) operation and performance of this point-of-use drinking water treatment system (POU-DWTS) when supplied with source water from a heavily biologically contaminated urban drainage pond, coupled with two focussed field trials to assess the quality of produced water.

Operational and water quality trial periods
The POU-DWTS was trialled between 26 th November 2019 and 27 th September 2022 (a total of 34 months), to assess the operational capability and treatment efficiency.This study builds upon a previous feasibility investigation of a proof-of-concept drinking water treatment system [18].Throughout this operational period two water quality trials were conducted: Field trial 1 (FT1) was conducted between 26 th November 2019 and 7 th February 2020 (11 weeks); and Field trial 2 (FT2) was conducted between 3 rd May and 27 th September 2022 (22 weeks).No permits were required for the operational or water quality trials as the POU-DWTS and water source were located on university premises (University of the West of England, Bristol).

Point-of-use drinking water treatment system (POU-DWTS)
A schematic of the POU-DWTS is shown in Fig 1 with all pumps, digital controllers and inline sensors supplied by Portsmouth Aqua Ltd.Raw water was drawn from an urban drainage holding pond, (coordinates-N51˚29 0 56@ (51.498888),W2˚32 0 39@ (-2.544166), see S1 Fig) located at the University of the West of England, Bristol, (UWE, Bristol), UK, which has an area of approximately 1,300 m 2 .Water from this pond was pumped into two settle tanks (1000-L intermediate bulk containers), before passing through 5 μm prefilter-ionisers (Turbidex, PolletWater Group, Belgium), into a header tank (1000-L intermediate bulk containers).Water from the header tank was then drawn through a 50 μm reverse flushing filter and dosed with HOCl using a commercial generator (see section Generation and use of Hypochlorous acid for further details).Dosing values were set to a maximum concentration of 2.0 mg L -1 of free available chlorine and an oxidation reduction potential (ORP) of 720 mV.HOCl dosing was performed using a SMART S-DDA, digital diaphragm dosing pump (Grundfos), in which dosing volumes were automatically adjusted in-line, with treated water, via feedback control from in situ free available chlorine and ORP sensors; (see Fig 1).The resultant disinfected water was then pumped through the ultrafiltration (UF) membranes (0.02 μm pore size; Dizzer 4040-4.0Inge, Lenntech, The Netherlands) and the treated water stored in a treated water tank (1000-L intermediate bulk containers) that, when full, was allowed to "trickle-return" back into the source water body via an extended PVCu drainage pipe located in an established reed bed to minimise the effect treated water has on the raw water source.Throughout this study all microbial, chemical, and metallic 'contaminants' are naturally occurring within the raw water (urban drainage pond), that collects surface runoff from across the southern area of the UWE Bristol's Frenchay campus.

Telemetry system-Operational parameters
In total, 44 POU-DWTS operational parameters were continuously monitored via the Grundfos Remote Manager interface (Grundfos, Denmark).Specific parameters included, flow rate, transmembrane pressure (TMP), oxidation reduction potential (ORP) and free available chlorine concentrations (s::can, Messtechnik GmbH, Austria) of the treated water, post the UF membranes.All data was collected over a 5-minute period provide a rolling 5-minute average.The 'health' of UF membranes was determined by calculating the permeability using Eqs 1-3 [18]:

Generation and use of hypochlorous acid [HOCl]
Hypochlorous acid (HOCl) was generated at a rate of 90 L hr -1 through the electrochemical activation of a saturated NaCl solution and softened water (CalSoft Non-Electrical, CalMag Ltd, UK) using a PaquaLyte ELA-900 generator (Envirolyte, Estonia) and stored in a 100-L tank (Fig 1).HOCl was generated to produce a free available chlorine concentration of 365 ± 15 mg L -1 , an ORP of 800 ± 15 mV and pH of 5.65 ± 0.15.Between May 2021 and April 2022 HOCl was generated at a free available chlorine concentration of 441 ± 30 mg L -1 (ORP of 1110 ± 7 mV and pH of 6.6 ± 0.04) to investigate whether a reduced volume dosing rate would be more resource efficient, whilst maintaining treated water quality and UF membrane permeability.HOCl was used as a disinfectant step after the pre-filters, but before water enters the UF membrane columns within the POU-DWTS (see Fig 1).The POU-DWTS initiated an ) and consisted of UF membranes being soaked for 1 hour with HOCl, subsequently followed by a backwash with HOCl before the treatment system restarted.This sequential increase in the TMP enhanced clean threshold was manually increased twice throughout the long-term trial, as the operational limits of the POU-DWTS were better understood.

Field trial 1
Field trial 1 (FT1) was conducted over an 11-week period between November 2019 and February 2020, to test the quality of water produced at two different flow rates.A flow rate of 0.3 m 3 hr -1 (0.299 ± 0.02 m 3 hr -1 ) was tested between weeks 1-7, before being increased to 0.5 m 3 hr -1 (0.488 ± 0.03 m 3 hr -1 ), for the remaining 4-weeks.These flow rates were chosen to test the POU-DWTS's capability of producing biologically safe water at a community scale (e.g.> 400 people) [30].
Field trial 1-Water sampling and analysis.Raw water samples were taken directly from the urban drainage pond, whilst POU-DWTS water samples were taken from a sample tap immediately before the treated water tank (Fig 1).Samples were collected weekly in sterile vessels, supplied by Wessex Water Scientific Laboratories (ISO 17025 accredited), except over weeks 5 and 6 when the accredited laboratory was closed.An overview of the parameters tested is described in S1 Table .Individual (bromoform, bromodichloromethane, chloroform and dibromochloromethane) and total THMs (sum of 4 THMs) were also quantified by Wessex Water Laboratories in the week 1 and week 11 of the trial.Water samples were transported to Wessex Scientific Laboratory (Saltford, UK) within one hour of sampling for independent analysis.

Field trial 2
Field trial 2 (FT2) was conducted over a 22-week period between May and September 2022 (samples collected weekly), to assess the water quality towards the end of the three-year POU-DWTS operational period, particularly during the summer months when the microbial load of the source water was known to be at its highest.As per FT1, raw water samples were taken directly from the urban drainage pond, whilst POU-DWTS water samples were taken from the sample tap located immediately before the treated water tank (Fig 1).Basic, microbiological, chemical, and metallic quality of the raw and treated water were assessed within inhouse laboratories on a weekly basis according to the methods detailed below.The FT2 flow rate was set to 0.3 m 3 hr -1 (0.22 ± 0.1 m 3 hr -1 ).
FT2 -Basic water quality.The pH was measured using a benchtop meter (FiveEasy, Mettler Toledo, UK), and conductivity was measured using a handheld meter (Acument AP75, Fisher Scientific, UK).Free available chlorine (FAC) was measured using the N, N-diethyl-pphenylenediamine sulphate (DPD) no. 1 Palintest method [31].Hardness (Hardicol), alkalinity (Alkaphot), colour and turbidity were all measured according to Palintest methods and quantified using a Palintest Photometer 7500 (Gateshead, UK).All measurements were taken in triplicate.Total organic carbon [TOC] was quantified using an EnviroTOC (Elementar, Langenselbold, Germany).All TOC samples were filtered through 0.45 μm sterile cellulose filters within 30 minutes of sample collection and stored at 4˚C prior to analysis.Samples were then decanted into 30 mL glass universals that had been prepared by soaking in 0.5 M nitric acid for 24 hours, rinsed with ultrapure water (18.2MΩ) and dried.The filled glass universals were placed into an autosampler and covered with aluminium foil before auto injection sampling in triplicate.

FT2 -Microbiological water quality. All microbiological media was purchased from
Thermo Fisher Scientific, UK.Total coliforms and presumptive Escherichia coli were quantified using the membrane filtration method [32] on membrane lactose glucuronide agar (Oxoid CM1031B) and incubated at 30˚C for 4 hours, then at 37˚C for 18-20 hours.Presumptive Enterococci were quantified using the membrane filtration method [33] on Slanetz and Bartley agar (Oxoid CM0377B) and incubated at 37˚C for 4 hours then at 44˚C for a further 40-44 hours.Clostridium perfringens were quantified by pour plate method using Perfingens Tryptose Sulphite Cycloserine agar base (Oxoid CM0587B) supplemented with Tryptose Sulphite Cycloserine selective supplement (Oxoid SR0088E) and egg yolk emulsion (Oxoid SR0047).Set pour plates were then incubated under anaerobic conditions at 35˚C for 24 hours with a humidity of > 80%.Total viable bacteria were quantified by spread plate methods on R2A agar (Oxoid CM0906B), with samples of raw and treated waters being incubated at both 22˚C and 30˚C [34].All raw and treated water samples were plated in triplicate and resultant colonies were counted and expressed as CFU mL -1 or CFU 100 mL -1 .
FT2 -Chemical and metallic water quality.Where necessary ultrapure water with a resistivity output of 18.2 MΩ was used for the preparation of suppressant, mobile phase, and standard solutions.The concentrations of chemical ions present in the raw and treated water samples were determined using ion chromatography (850 Professional IC, Metrohm, UK) compete with autosampler.Anions quantified were chloride (Cl -), nitrate (NO 3 ) and sulphate (SO 4
Metals present in raw and treated water samples were determined by Inductively Coupled Plasma Optical Emission spectroscopy (ICP-OES; model 5110, Agilent, UK) with autosampler.A 0.25 M nitric acid solution was used as a rinse prior to and in between sample analysis and the following metals were quantified: aluminium (Al), magnesium (Mg), sodium (Na), silica (Si) and zinc (Zn).All standards for IC and ICP-OES analysis were certified (TraceCert) and purchased from Fisher Scientific, UK.

Data analysis
POU-DMTS telemetry data was downloaded from the Grundfos Remote Management page into Microsoft Excel, before graphs were produced in RStudio (2022.07.2, Build 576) and edited in Microsoft PowerPoint.Code used to produce Fig 2 can be found in Supplementary Information (S1 Text).For FT1 and FT2 statistical analysis and graph construction (FT2 only) was performed using GraphPad Prism version 9.00 for Windows (GraphPad Software, San Diego, CA).Significant differences between raw water and treated water samples were determined using an unpaired two-tailed t-test with a confidence interval of 95% for each parameter listed in S1 Table, whereby a p value of < 0.05 was considered significant.

Long-term operation and performance of the POU-DWTS
The POU-DWTS was operated for a 34-month period from November 2019 to September 2022.The key POU-DWTS operational parameters of flow rate, permeability, free available chlorine and ORP were continuously monitored and are shown in Fig 2. The sporadic data available between April and August 2020 is a result of intermittent telemetry recording caused by a faulty SIM card, whereby sensor data was unable to be uploaded from the POU-DWTS to the Grundfos Remote Management Page.Installation of a new SIM card rectified this issue.
The flow rate was set at 0.3 m 3 hr -1 between November 2019 and January 2020 (Fig 2A ), to assess initial system performance immediately after installation and commissioning (October 2019).The flow rate was then increased to 0.5 m 3 hr -1 between January and April 2020 to investigate the impact on both system performance and output water quality (see Field Trial 1).From April 2020 to August 2020 the flow rate was set at 0.3 m 3 hr -1 , and then increased to 0.4 m 3 hr -1 from August to November 2020, before briefly being increased to 0.5 m 3 hr -1 in December 2020.In January 2021 the flow rate was reduced to 0.3 m 3 hr -1 before the flow rate was then gradually increased from 0.4 m 3 hr -1 to 0.75 m 3 hr -1 between January 2021 and March 2021 to understand the potential maximum throughput of the POU-DWTS when supplied with water from the raw water source (Fig 2A).The maximum consistent flow rate achieved was 0.75 m 3 hr -1 in March 2021.However, this resulted in frequent instances where the POU-DWTS entered automatic enhanced HOCl cleans, as the transmembrane pressure exceeded the maximum set point of 1.5 bar.Between 1 st March and 7 th March 2021, the POU-DWTS entered into automatic enhanced HOCl cleans on 81 occasions.Therefore, although system operation was maintained, the average operational time per day was reduced.For example, on the 7 th March 2021 there were 15 instances of automatic enhanced HOCl cleans, resulting in just under 9 hours of operation, yet there were no instances of automatic enhanced cleans at a flow rate of 0.3 m 3 hr -1 .As a result, the set flow rate of the POU-DWTS was set to 0.3 m 3 hr -1 throughout April 2021, and no automatic enhanced cleans occurred during this period, or for the remaining operational period (May 2021 -September 2022).During May and August 2021, the flow rate was increased to 0.4 m 3 hr -1 but was reduced back to 0.3 m 3 hr -1 in August 2021 until September 2022 (trial end) to minimise non-operational periods.
The permeability of the UF membranes within the POU-DWTS (calculated from filtration flux and the transmembrane pressure, Eqs 1 -3) was also continuously monitored over the 3-year operational period (Fig 2B).Permeability is an industry standard to infer the 'health' of UF membranes, whereby decreasing permeability indicates the occurrence of biofouling, due to the build-up of biofilm within the UF membrane pores resulting in blockage [24].The permeability correlates with flow rate fluxes throughout the three-year operational period demonstrating that there was no reduction in the performance of the POU-DWTS.In addition, the UF membranes were not removed or replaced at any time throughout the study, nor was any recommended strong acid/alkaline chemical clean undertaken at any point.The use of in-line HOCl disinfection dosing, in combination with HOCl backwashes and automatic enhanced HOCl cleans alone are demonstrated to maintain operational health of the UF membranes.
The level of free available chlorine (FAC) and ORP of the treated water was continuously monitored over the 3-year operational period within the POU-DWTS to enable real-time water quality control monitoring (Fig 2C and 2D).As part of the remote management, this data was used to control and maintain the correct concentrations of HOCl solutions being for in-line dosing, proportional the variable incoming flow of source water which was also continuously monitored (Fig 2A).Between November 2019 and April 2021, HOCl was generated and stored in a 100-L disinfection tank (see schematic; Fig 1A ) to a set FAC concentration of 365 ± 15 mg L -1 and ORP of 800 ± 15 mV (validated by periodic spot sampling).The operational POU-DWTS was set to dose in-line (pre-UF membrane) at a concentration of 2.0 mg L - 1 of FAC.This resulted in an average FAC concentration of 1.15 mg L -1 over the entire 34-month period, which varied by ± 2.5 mg L -1 (n = 209,581) caused by the inherent dynamic nature of the treatment system (e.g.backwash periods, source water quality variability).Throughout the operational period, the measured FAC concentrations were consistently below 5 mg L -1 and the ORP between 700-800 mV, as is recommended by international drinking water guidelines.From May 2021 to April 2022 (blue box; Fig 2 ), the HOCl was generated and stored to a FAC concentration of 442 ± 30 mg L -1 (ORP of 1110 ± 7 mV).The concentration of HOCl was increased to enable a reduced volumes of HOCl (at higher concentrations) to be dosed in-line within the POU-DWTS.This increases system efficiency in terms of power and resource management but also results in greater variability of the in-line FAC concentrations observed in the treated water.On Numerous occasions instances FAC concentrations exceeded 5 mg L -1 (Fig 2C).Consequently, there was also a notable increase in the variability of the recorded ORP within the range of ~750 mV-950 mV (Fig 2D).Although this increase in FAC and ORP did not negatively impact the performance of the UF membranes (as permeability remained stable; Fig 2B), however due to the instability of these water quality parameters within the treated water FAC concentrations were reset and returned to concentrations of 365 ± 15 mg L -1 and ORP of 800 ± 15 mV (validated by periodic spot sampling) for the remainder of the operational period, inclusive of FT2 (April 2022 -September 2022).
It should also be noted that during the operational period when a higher concentration HOCl solution was dosed in-line within the POU-DWTS (albeit at a lower volume), there was one month (July 2021) when the HOCl dosing location was changed from pre-UF membranes (shown as "6" in Fig 1 ) to pre-50 μm filter (shown as "5" in S2 Fig) .Given the success of HOCl dosing in managing the UF membranes it was proposed to attempt to move the HOCl dosing point upstream to also help manage the 50 μm filter (although no actual issues had been observed).The dotted blue boxes in Fig 2 represents this period, whereby although there was no negative effect on the UF membranes (permeability remained unchanged; Fig 2B ), it resulted in no FAC detected by the in-line sensors and a substantial reduction in ORP of the treated water.Therefore, the HOCl dosing point reverted to the original location (pre the UF membranes) in August 2021, whereby the concentrations of FAC and ORP immediately returned to the values recorded before the dosing location had been altered.However, it should be noted that POU treatment systems do not require the same need for residual chlorine to be present in treated final water as it does not travel throughout a distribution network, like treated water produced in centralised drinking water treatment facilities.

Water quality Field Trial 1 (FT1)
A field trial (FT1) to assess water quality was undertaken during the first 11 weeks of the POU-DWTS operational period (see Fig 2).The main aim of this was to determine whether the treated water produced adhered to international drinking water quality standards [35,36].During FT1, a total of 18 samples were collected, 9 samples from the raw water source and 9 samples of POU-DWTS treated water (one sample per water source per week).These were assessed for biological, physicochemical (basic), chemical and metallic water quality parameters by an ISO 17025 accredited laboratory.Table 1 shows the average values for both raw and treated water samples for each of these water quality parameters (i.e., averaged across the full trial period, n = 9), highlighting where significant differences were observed.
The microbial load of the raw water exceeded permissible thresholds set out by the Drinking Water Inspectorate [36] for samples taken throughout FT1, with the confirmed presence of Clostridium perfringens, coliforms (including thermotolerant), Enterococci and Escherichia coli (including thermotolerant).All treated water samples met Drinking Water Inspectorate standards whereby there were no detectable potential pathogens (Table 1), demonstrating that the treated water was bacterially safe for consumption.There were some low levels of detectable heterotrophic bacteria (HPCs) present in the treated water, although this was significantly reduced compared to the raw water (p<0.01)and there is no maximum permissible threshold for HPCs present in drinking water.
Most basic water quality parameters (alkalinity, colour, conductivity, non-carbonate hardness, pH, total organic carbon and total hardness) of the raw water source were within Drinking Water Inspectorate limits, except for turbidity, where an average of 13 ± 4 nephelometric turbidity units [NTU] was significantly above the maximum Drinking Water Inspectorate limit of 4 NTU.All basic water quality parameters for treated water samples were within Drinking Water Inspectorate limits (see Table 1), including a significant reduction in colour and turbidity compared to the raw water (p<0.01).For the chemical analysis undertaken, all raw water and treated water samples were within the Drinking Water Inspectorate limits throughout FT1 (Table 1), although there was a significant reduction in ammonium within the treated water (p<0.0001),albeit from a low baseline.Interestingly, chloride was the only parameter to slightly (although non-significantly, p > 0.05) increase from 39 ± 10 mg L -1 within the raw water source, to 57 ± 34 mg L -1 within the treated water, as a result of the HOCl being dosed within the POU-DWTS (which is produced from the electrolysis of a low salt (NaCl) solution).The concentration of metals measured within the treated water samples were all within Drinking Water Inspectorate limits (Table 1).Whereas, for the raw water samples, aluminium (0.4 ± 0.1 mg L -1 ) and iron concentrations (0.37 ± 0.12 mg L -1 ) exceeded the acceptable 0.2 mg L -1 concentration set by the Drinking Water Inspectorate.When directly comparing the metal concentration of raw and treated water samples, the POU-DWTS significantly reduced the concentrations of aluminium, iron, lead, and manganese (p<0.0001).Interestingly, as with chloride, sodium concentrations slightly (although non-significantly) increased from 26 ± 6 mg L -1 within the raw water source, to 60 ± 83 mg L -1 within the treated water, again as a result of the HOCl being dosed within the POU-DWTS.
Six weeks into FT1, the set flow rate was increased from 0.3 m 3 hr -1 to 0.5 m 3 hr -1 , to investigate the impact on both system performance and output water quality.However, it is clear Table 1.Microbiological, basic, chemical and metal parameter results for raw water and treated water samples taken from a POU-DWTS during an 11 week field trial (FT1; Nov 2019 -Jan 2020).Results shown are the calculated mean from the independent ISO 17025 accredited laboratory reports (n = 9 ±SD).Significant difference (Sig.diff) was calculated for the raw and treated water, this was undertaken through an unpaired two-tailed t-test, with a confidence interval of 95% (**** = p<0.0001;*** = p<0.001;** = p<0.01;* = p<0.05;NS = not significant).Figures in bold represent values above accepted Drinking Water Inspectorate limits, and where the is no specified Drinking Water Inspectorate limit, this is represented by NA. from the laboratory results that this had no impact on the quality of the water produced, with all FT1 treated water samples met Drinking Water Inspectorate drinking water standards.In weeks 1 and 11 of FT1, individual (chloroform, bromodichloromethane, dibromochloromethane and bromoform) and total trihalomethanes (THMs; sum of individual concentrations) were analysed within the treated water.In week 1 the concentration of individual THM species were: 4.92 μg L -1 , 1.43 μg L -1 , 1.0 μg L -1 , and < 0.66 μg L -1 for chloroform, bromodichloromethane, dibromochloromethane, and bromoform, respectively.In week 11 the concentration of individual THM species were: 2.89 μg L -1 , 1.31 μg L -1 , 1.38 μg L -1 , and <0.66 μg L -1 for chloroform, bromodichloromethane, dibromochloromethane, and bromoform, respectively.The dominant THM species present was chloroform (~60%), which has been frequently observed throughout chlorine-based drinking water disinfection processes [29,37].In week 1, with a flow rate of 0.3 m 3 hr -1 , total THMs were 7.35 μg L -1 , and then reduced to 5.58 μg L -1 in week 11 of the trial, after the flow rate had been increased to 0.5 m 3 hr -1 .These values are within commonly set total THM legal limits of 80 μg L -1 [38] or 100 μg L -1 [36,39].At the set flow rate of 0.3 m 3 hr -1 the total organic carbon [TOC] was 1.96 ± 0.37 mg L -1 (week 1), compared to 1.81 ± 0.13 mg L -1 at the increased set point of 0.5 m 3 hr -1 (week 11).

Water quality Field Trial 2 (FT2)
A second water quality field trial (FT2) was conducted towards the end of the three-year POU-DWTS operational period (May-Sep 2022; see Fig 2), whereby the quality of the treated water was assessed over 22-weeks.This included spring/summer seasons when microbial numbers in raw waters are generally much higher compared to autumn/winter months.During FT2, 44 samples were collected in total, 22 samples from the raw water source and 22 samples of POU-DWTS treated water.Again, these were assessed for biological, physicochemical (basic), chemical and metallic water quality parameters, although this time analysed in the University's research laboratories.
Fig 3 shows the recovered potential pathogens from raw and treated water samples over the 22-week FT2 period.There is a significant reduction (p < 0.0001) in total coliforms, presumptive E. coli, Clostridium perfringens and Enterococci when comparing the raw and treated water samples.In fact, zero pathogens are recovered from treated water samples except in weeks 4 and 20, when low levels were recovered.However, during these two weeks there was a temporary technical issue with the electrochemical HOCl generator, whereby it to an accumulation of salt within the cell, it had entered standby mode and stopped HOCl production.This was easily rectified, though the performance of a descaling protocol, in-line with manufacturers guidelines, to remove salt built up in the electrochemical cell.This accumulation of excess salt in the cell was attributed to the production of high concentration HOCl solutions at 442 ± 30 mg L -1 .Therefore, to avoid this as the field trial progresses, production concentration of HOCl was reset to 365 ± 15 mg L -1 .This demonstrates the importance of maintaining the in-line HOCl dosing regimen to ensure healthy membranes are maintained within the POU-DWTS.For all subsequent FT2 data analysis presented here, data relating to the occasions where the HOCl generator was temporarily in standby mode (weeks 4 & 20), is not shown.
Table 2 shows the average biological, physicochemical (basic), chemical and metallic water quality parameter values for both raw and treated water samples (i.e., averaged across the full trial period; except weeks 4 & 20), highlighting where significant differences were observed.As per FT1, there was a high microbial load within the raw water source, whereby significant numbers of total coliforms, presumptive E. coli, Clostridium perfringens and Enterococci were recovered from all samples, hence this water was unsafe for human consumption according to Drinking Water Inspectorate and WHO standards [35,36].In comparison, no potential pathogens were recovered from any of the treated water samples (when the system was fully operational), therefore demonstrating that the treated water was bacterially safe for consumption.Similarly to FT1, there was a significant reduction in HPCs between the raw water source and treated water samples (p<0.0001;Table 2), although a low level (<1 log 10 CFU mL -1 ) was still present within the treated water samples.
The basic water quality parameters of raw water samples (alkalinity, conductivity, FAC, pH, TOC and total hardness) were within drinking water quality guidelines (where applicable), except for colour (258 ± 67 mg L -1 Pt/Co) and turbidity (44 ± 13 NTU).For the treated water samples, there was a significant reduction in alkalinity, colour, TOC, total hardness and turbidity (p < 0.0001), whereby all treated water samples were within Drinking Water Inspectorate limits.For the chemical and metal analysis undertaken, all samples were within drinking water guidelines.Manganese was the only metal to significantly reduce from the raw water to the treated water samples (p < 0.01).The increase in chloride and sodium concentrations from raw water to treated water samples were also observed in FT1 due to in-situ HOCl dosing.The increase in sodium concentration observed in raw and treated water samples between FT1 (raw 26 ± 6 mg L -1 , treated: 60 ± 83 mg L -1 ) and FT2 (raw: 126 ± 32 mg L -1 , treated: 153± 35 mg L -1 ) could be due to low level accumulation in the raw water source, as rainfall is often reduced in summer months, thus reducing any dilution effect.Significant increases were observed between the raw water and the treated water within FT2 for the following chemical parameters nitrate (p = 0.0067), phosphate (p = 0.0018) and sulphate (p = 0.0133), and significant increases were also seen for magnesium (p = 0.0133) and potassium (p = 0.0067).However, the increases in these parameters did not exceed drinking water guidelines.

Discussion
This study investigated the long-term operation of a novel POU-DWTS when supplied with source water from a heavily biologically contaminated urban drainage pond.In addition, the quality of the produced drinking water was monitored and assessed through two field trials.Throughout the 3-year POU-DWTS operational period the combination of continuous dosing Table 2. Analytical results for the of microbiological, basic, chemical and metallic parameters of the raw water and treated water samples taken throughout Field Trial 2; the 22-week MVP DWTS trial.Results shown are the calculated mean from in-house analysis.Significant difference (Sig.diff) was calculated for the raw and treated water, this was undertaken through an unpaired two-tailed t-test, with a confidence interval of 95% (**** = p<0.0001;*** = p<0.001;** = p<0.01;* = p<0.05;NS = not significant).Figures in bold represent values above accepted Drinking Water Inspectorate limits, and where the is no specified Drinking Water Inspectorate limit, this is represented by NA. of HOCl, programmed backwashes and automatic enhanced cleans, resulted in continuous operation (without the requirement for the UF membranes to be removed to undergo strong acid-alkaline chemical cleans.This is a significant and key finding as the limitation for implementing UF membranes in rural or resource constrained areas is the affinity for biofouling to occur due to biofilm formation on membrane surfaces, causing pores to become blocked [24,25].Biofouling in the membranes leads to an increase in the transmembrane pressure, and a reduction in membrane permeability.When this occurs UF membranes are usually removed and subject to a strong acid-alkaline chemical clean (e.g., NaOH, HCl, H 2 SO 4 or NaOCl) to remove the organic material, present in the form of extracellular polymeric substances [40].This results in operational down time, an increase in the operational maintenance requirements and costs.Throughout the three-year operational period, the flow rate of the POU-DWTS was adjusted to determine the optimal flow rate from the water source under investigation (an urban drainage holding pond).The optimal flow rate was established to be approximately 0.3 m 3 hr -1 , the maximum flow rate trialled (0.75 m 3 hr -1 ) resulted in an increase in biofouling followed by an increase in transmembrane pressure which initiated automatic enhanced HOCl cleans once the set threshold was reached (see methods).During these cleaning events drinking water is not produced.When the flow rate of the POU-DWTS was set to 0.75 m 3 hr -1 flow rate, there were 15 instances of automatic enhanced HOCl cleans on the first day of operation, resulting in an operational period of just 9 hours.During this time, there was approximately 6.75 m 3 of drinking water produced over the 24-hour period.This level of production would provide enough drinking water for approximately 45 people (for high income countries, the recommended guidelines are ~150 L per person per day.In emergency, humanitarian or disaster relief environments, this is amplified to 450 people, where ~15 L per person per day is recommended [30].However, running the POU-DWTS with a flow rate of 0.3 m 3 hr -1 over a 24-hour period would provide a comparable volume of drinking water (7.2 m 3 ) but without any non-operational periods.

Parameter
The two water quality field trials (November 2019 -January 2021 and May 2022 -September 2022) demonstrated that the source water under investigation was not safe for human consumption due to the consistent presence of potential pathogenic bacteria (total coliforms, E. coli, Enterococci and Clostridium perfringens), which exceeded internationally recognised drinking water guidelines [35].Conversely, the POU-DWTS was successful in producing safe drinking water that adhered to WHO drinking water guidelines [35].Interestingly, there were two instances when the produced water failed in terms of microbiological quality, and these were found to directly correlate with specific issues relating to interruptions in HOCl production during operation.Excessive salt build-up in the electrolytic cells causes the HOCl generator to enter standby mode until an automated descale or clean is performed.This occurred when HOCl solution at high concentrations were produced (i.e., 442 ± 30 mg L -1 ).The optimum operational window for the concentration of HOCl was found to be 365 ± 15 mg L -1 when using an input flow rate of 0.3 m 3 hr -1 .This reinforces the need for continuous HOCl dosing and monitoring to ensure the production of biologically safe drinking water during POU-DWTS operation.
For several water quality parameters there was a consistent observed reduction in concentrations, between the raw water and the treated water samples.For example, in both FT1 and FT2 the colour, turbidity and TOC was reduced in the treated water by the POU-DWTS filtration processes (which includes ultrafiltration).A reduction in TOC, reduces the potential of FAC to react with organic matter, this is a known precursor to THM formation and other production of chlorine-based disinfection by-products [41].The presence of iron, lead, manganese and zinc in drinking water is often associated with corrosion issues in the pipes of supply networks, resulting in unwanted organoleptic properties [42] as well as potential exceedances of drinking water quality guidelines [35,36].Throughout FT1 and FT2 these parameters were consistently reduced throughout the treatment process, likely as a result of the 5 μm prefilterionisers.However, for several parameters in FT2 (Cl -, Na + , NO 3 -, SO 4 -, Mg 2+ and conductivity) there was an observed increases between the raw water source and the treated water.The increases in Cl -and Na + are likely associated with HOCl dosing (see "Generation and use of hypochlorous acid [HOCl]").The reactive nature of HOCl could lead to the oxidation of anions e.g.ammonium to nitrite to nitrate [43], and a release of metal ions from chelates.Further research to elucidate reaction mechanisms is needed.In all instances the drinking water produced from the POU-DWTS were within international drinking water quality guidelines.
It is clear that the differences in the chemical and metal composition observed in the raw water varied between FT1 and FT2, which led to variation in chemical and metal concentrations in treated water.This is likely a result of the different seasons when the trials took place, with FT1 taking place throughout winter (November 2019 -February 2020), whereas FT2 took place throughout spring/summer (May-September 2022).For example, the average concentration of sodium in the raw water in FT1 was approximately five times lower than throughout FT2.This could be a result of accumulation in the raw water source due to reduced rainfall and, therefore, a reduced dilution effect or from road run-off.Interestingly, literature has shown that higher concentrations of both sodium and chloride over the winter months in freshwater and drinking water due to leaching of salt (NaCl) from road de-icing and road runoff [44].
There are several benefits for utilising electrochemically generated HOCl in POU-DWTS.Firstly, the antimicrobial efficacy of HOCl has been shown to be up 100 times greater, in comparison to OCl -solutions such as NaOCl and Ca(OCl) 2 [45][46][47] which are frequently used in drinking water disinfection processes [48].The increased antimicrobial efficacy is thought to be a result of the sequestration of electrons from the bacterial cell surface via the neutral HOCl molecule and increased oxidative stress caused by higher ORP [49], in addition to presence of FAC.Secondly, electrochemically generated HOCl is known to be fast-acting [50,51].This is important in decentralised or POU-DWTS, where residual (chlorine) disinfection is not required nor necessary to maintain water quality throughout a distribution network, which in turn reduces chlorine demand.To better understand the HOCl demand of the POU-DWTS, a simple mass-balance model for HOCl generated FAC was utilised, whereby: Where FT2 is used as an example, assuming a HOCl FAC dose of 2mg L -1 , and a final measured FAC residual in the treated water of 0.52 mg L -1 (see Table 2), the FAC demand will be 1.48 mg L -1 .Consequently, the hourly HOCl FAC demand for the system when operating at 0.3m 3 hr -1 would be 444 mg hr -1 .Finally, electrochemically generated HOCl can be produced on-site, on-demand and requires only a saturated saline (NaCl) solution as the raw material.This greatly simplifies the supply chain, reducing the need to purchase, transport and store disinfectants.Such technology could provide communities or facilities with the means of enhanced improved resilience against supply chain issues.In addition, it potentially enables HOCl solutions to be used more widely as a general disinfectant to clean surfaces [26], rinse fresh produce [52,53] or to clean wounds [54].
The POU-DWTS was powered by a 230 V alternating current, 32-amp power supply, resulting in an average power draw of 0.96 W L -1 and a maximum power draw of 2.9 kWh.The power required to constantly run the MVP DWTS could conceivably be achieved through renewable power sources, such as photovoltaic panels.However, the potential for such units to be implemented in LICs/LMICs would require further field trials to ensure the system is capable of optimal performance whilst experiencing different environmental conditions, such as variations in temperature and humidity.
The previous proof-of-concept DWTS developed by the authors focussed on the production of biologically safe drinking water, both with and without pre-/post-UF membrane disinfection, employing an acidic electrochemically activated solution [ECAS] (pH 3.3 ± 0.16) [18,55].The proof-of-concept trial determined that ECAS dosing was required to produce biologically safe water (zero coliforms present in treated water) and to also maintain stable permeability across UF membranes [55].The previous proof-of-concept DWTS produced 2.88 m 3 per day, 60% less than the 7.2 m 3 produced by the current POU-DWTS which could provide drinking water for up to 192 people per day in humanitarian settings (15 L per person) [30], or approximately 19 people per day for all domestic purposes (150 L per person).The previous proof-of-concept (feasibility) trial also reinforced the need for in-situ disinfection as part of drinking water treatment to inactivate microorganisms, even where ultrafiltration is deployed [17].Continuous in-situ dosing of chlorine solutions is not advised in-line with UF membranes due to negative impacts the oxidative solutions have on the material surfaces [24,25].However, this study has demonstrated that the use of HOCl does not appear to affect the UF membrane performance within the POU-DWTS.Once the POU-DWTS is decommissioned at UWE Bristol's Frenchay Campus, the UF membranes will be assessed for material surface damage by comparison with new UF membrane columns, to investigate any oxidative damage as a consequence of continuous HOCl dosing.The decommissioned UF membranes will also be assessed for any adverse effects biofilms had on the material, and quantify biofilm density, structure and diversity.
Future trials should investigate and optimise flow rate and maximum drinking water output with minimal automated enhanced HOCl cleans on a variety of water sources, environments and climates.These trials should elucidate whether additional processes would be required to reduce key contaminants in some situations such as arsenic [56], lead [57], fluoride [58], or per-/polyfluoroalkyl substances [PFAS] [59] that can be detected in certain global source waters.It would also be worthwhile to assess whether the POU DWTS acts as a source/sink for microplastics, as recent studies have shown that UF membranes are capable of relatively high removal efficiencies for microplastics [60,61].In addition, future trials should investigate the feasibility and practicalities around long-term installation of a POU-DWTS when powered solely from renewable energy sources.

Conclusion
This is the first time that HOCl has been utilised to produce internationally standard drinking water, whilst also managing biofilms within ultrafiltration membranes.The purpose of this study was to demonstrate the operational viability of the POU-DWTS over a long-term sustained period (3 years).Throughout the operational period the POU-DWTS produced drinking water that adhered to international drinking water quality guidelines, when using electrochemically generated HOCl to continuously dose the system.Throughout the 3-year operational period HOCl was continuously dosed resulting in stable permeability, indicating good membrane 'health'.Whereby the UF membranes were not replaced and did not have to undergo acid/alkaline chemical cleans at any point.This trial also demonstrated that the POU-DWTS continuously produced high quality drinking water, that met international water quality guidelines, at a community scale within the location studied.The POU-DWTS has recently been trialled at the Doha Expo 2023 by Portsmouth Aqua, but future studies would seek to trial the POU-DWTS in varying environmental and climatic conditions, specifically in LICs.

Fig 1 .
Fig 1.Schematic of the point-of-use drinking water treatment system (POU-DWTS).Raw water is drawn from the environmental water body (1) into two 1000-L settle tanks (2), prior to entering 5 μm prefilter-ionisers (3), before passing into a 1000-L header tank (4).Header tank water then passes through a 50 μm reverse flushing filter (5), dosed with HOCl [A] (controlled by a dosing instrumentation digital [DID] controller [B]) generated from the PaquaLyte generator and then passes through 0.2 μm ultrafiltration membranes (UF) (6).Filtered and disinfected water is stored in a 1000-L treated water tank (7).Free available chlorine and oxidation reduction potential [ORP] of treated water is monitored through in-line s::can sensors [C], which informs the DID controller enabling dosing to be automatically adjusted.Pressure gauges pre-and post-UF membranes are labelled as [P].Water sample collection points (raw water and treated water tank) are indicated by red Asterix [*].Treated overflow waters are returned back to the environmental water body and represented by a dashed line (-).https://doi.org/10.1371/journal.pwat.0000187.g001