Removal of waterborne bacteria from surface water and groundwater by cost-effective household water treatment systems ( HWTS ) : A sustainable solution for improving water quality in rural communities of Africa

In this study 5 household water-treatment devices/systems (HWTS) were constructed using inexpensive local materials (sand, gravel, zeolites and clays). They included the silver-impregnated porous pot filter (SIPP), the ceramic candle filter (CCF), the conventional biosand filter (BSF-S), a modified biosand filter with zeolites (BSF-Z), and a bucket filter (BF). Their ability to remove turbidity and pathogenic bacteria (Vibrio cholerae, Salmonella typhimurium and Shigella dysenteriae) from synthetic sterile water, groundwater and surface-water sources was evaluated. The flow rates ranged from 0.05 l·h-1 to 2.49 l·h-1 for SIPP; 1 l·h-1 to 4 l·h-1 for CCF; 0.81 l·h-1 to 6.84 l·h-1 for BSF-S; 1.74 l·h-1 to 19.2 l·h-1 for BSF-Z; and from 106.5 l·h-1 to 160.5 l·h-1 for BF. The highest (64% to 98% (0.74 to 1.08 NTU)) and lowest (14% to 76% (2.91 to 7.19 NTU)) average percentage turbidity removals were noted for SIPP and BF, respectively. The SIPP was the only device that consistently removed 100% of all target pathogens throughout the study. Its performance was found to be significantly superior (p<0.05) compared to that of the other four devices. Sixty (60%) to 100% bacterial removals were observed for BSF-S; 90% to 100% for BSF-Z; 90% to 100% for CCF; and 40% to 99.9% for BF. Based on the findings of this study the SIPP can be recommended for use by rural communities as it consistently produced high-quality water that complied with the SANS 241 turbidity and microbiological limits for drinking water.


INTRODUCTION
In 2008, the World Health Organisation reported the percentage of rural populations within some African countries still using unimproved water sources.These were reported to be 72% for the Democratic Republic of Congo, 71% for Madagascar and Mozambique, 62% for Angola, 55% for Tanzania, 54% for Zambia, 28% for Zimbabwe, 23% for Malawi, 19% for Lesotho, 14% for Swaziland, 12% for Namibia and 10% for Botswana (WHO/UNICEF, 2010).In South Africa, the supply of safe drinking water to rural and urban populations has improved from 59% in 1994 to 97% in 2010.According to statistics released in 2010, about 1.65 million out of 49 million people in South Africa do not have access to a safe water supply.The Department of Water Affairs is therefore seeking simple and appropriate water treatment technology options by means of which the 3% of the population in need of safe drinking water can treat the water sources that are available to them (DWA, 2010).
A number of studies have suggested that the key to reducing or even eradicating the burden of waterborne disease is through appropriate sanitation facilities and piped water systems.These could take decades to be established, especially in impoverished rural communities of African countries.Various point-of-use (POU) water treatment methods, which include biosand and ceramic filtration, appropriate chemical disinfection (e.g. the use of disinfectants such as chlorine and iodine), solar disinfection and natural water purifiers (e.g.Moringa oleifera) have been reported to improve the microbial quality of drinking water as well as to decrease the incidence of endemic diarrhoea caused by waterborne pathogens (Murcott, 2006).Household water treatment may provide African governments with a quick short-term solution to ensuring that all their people, especially those in rural areas, have access to safe drinking water (Sobsey, 2002).Due to the low cost of manufacturing the filters using locally available materials and the simplicity and ease of construction and maintenance, POU water-treatment systems enable users to have potable water available almost immediately after installation (Mol, 2001).
By using a list of selection and evaluation criteria, 5 household water treatment systems (HWTS) were chosen for this study from a wide range of water treatment filters.The selected devices were the silver-impregnated porous pot filter (SIPP), the ceramic candle filter (CCF), the conventional biosand filter (BSF-S), a modified biosand filter with zeolites (BSF-Z), and a bucket filter (BF).The main idea in constructing these filters was to evaluate the effect of flow rate and the presence of the biological layer and clay materials with or without silver on the reduction of microbial contaminants and turbidity.The BSF and the BSF-Z are both slow sand filters while the BF is a fast or rapid sand filter, but all three use sand as the filtering media.While both BSF-S and BSF-Z have a biological layer, which is absent in the BF, the BSF-Z has an additional zeolite layer.The CCF and the SIPP devices differed from these three filters since they have clay material and not 446 sand as filter medium.This study therefore focused on comparing these selected filters for the removal of turbidity and pathogenic bacteria (S. dysenteriae, S. typhimurium and V. cholerae) also taking into account the flow rate of each filter.

Selection criteria
An extensive study of the literature was undertaken and a list of more than 20 devices was proposed for evaluation.A number of selection and evaluation criteria were used to condense that list to 5 HWTS selected for this study (Mwabi et al., 2012).The purpose of these criteria was to ensure that the filters: (i) produce water that meets the South African National Standard -SANS 241 (SABS, 2011) microbiological drinking water specification; (ii) are easy to construct, operate and maintain; (iii) are cost-effective by being constructed from locally available material and not running on electricity, and (iv) are able to produce a minimum volume of 25 ℓ•person -1 •d -1 of water as recommended by the South African Department of Water Affairs (DWAF, 2002).

Construction of the household water-treatment systems/ devices (HWTS)
All materials used in the manufacturing process of the selected devices are locally available in South Africa.These were natural resources (gravel, sand, clinoptilolite zeolite and clay) that may also be found in the immediate environment of rural communities.The use of locally available materials helps to reduce a community's dependence on outside sources and often reduces the manufacturing costs of HWTS (Murphy et al., 2009).The construction methods for these modified filters have been published by Mahlangu et al. (2011;2012) and Mwabi et al. (2011;2012) and manufacturing costs were found to be dependent on the availability and accessibility of materials locally.

Silver-impregnated porous-pot filter
The silver-impregnated porous pot filter (SIPP) was developed as part of a previous project commissioned by the Water Research Commission of South Africa (WRC Project No. K8/810;Momba et al., 2008).The SIPP (Fig. 1) was manufactured according to a method described by Momba et al. (2008).
The SIPP is similar to the ceramic silver-impregnated pot filter (CSF) described by Van Halem et al. (2009), but it differs from the CSF in that it was fired after impregnating the silver nitrate instead of being coated with colloidal silver after firing as was done in the construction of the CSF.The silver was impregnated in the clay and sawdust mixture before shaping and firing the pot, to reduce leaching of the silver into treated water.
The total price for a complete SIPP filter is between ZAR240.56 and ZAR290.56.

Ceramic candle filter
A Just Water Ceramic Candle Filter was obtained from Headstream Water Holdings (South Africa).It consists of a material cover with a pore size of 5 µm, placed over a hollow, dome-shaped ceramic candle (10 cm x 10 cm; height and width) which contains pores that filter particles down to 0.5 µm.The ceramic shell contains an activated charcoal interior medium that covers a carbon-fibre blanket at the base of the filter with pore sizes of 0.2 µm.The candle filter was wedged between two 25 ℓ buckets by inserting and attaching the candle filter through the base of the upper bucket and through the lid of the bottom bucket, as well as inserting a spigot 5 cm from the base of the lower bucket, as illustrated in Fig. 2a-b.The total manufacturing cost amounted to ZAR501.15.

Biosand filter
Two types of plastic biosand filters were constructed in the workshop of the Tshwane University of Technology, Pretoria, South Africa, with some modifications to the biosand filter design developed by David Manz at the University of Calgary, Canada, in the early 1900s (Buzunis, 1995).Briefly, two modifications were made to the original design, (i) an adjustment to the height and width of the filter (41 cm high, 32 cm wide); and (ii) the replacement of the fine sand with clinoptilolite zeolite to serve as filter media (BSF-Z).The filter size was scaled down and a 25 ℓ plastic bucket was used to ensure that  (Misaelides, 2011).This modification aimed to determine whether the added zeolite would enhance the performance of a biosand filter.The formation of a biolayer is the key component of the biosand filter for removal of pathogens.CAWST (2010) recommends up to 30 days for the biolayer to fully form.The construction of these two biosand filters has been described by Mahlangu et al., (2011) and illustrated in Fig. 3.The total manufacturing cost of the two biological filters amounted to ZAR133.16 for BSF-S and ZAR164.23 for the BSF-Z filter.

The bucket filter
The bucket filter was also constructed in the workshop of the Tshwane University of Technology (Mahlangu et al., 2012).It was made of two 25 ℓ plastic buckets mounted on top of each other (Fig. 4).The top bucket contained the filter media while the lower bucket collected and temporarily stored the filtered water.The total manufacturing cost amounted to ZAR149.18.

Evaluation of the performance of the devices
The performance of the selected home water treatment devices in removing indicator bacteria (E.coli; faecal coliforms) and pathogenic bacteria was conducted simultaneously.The findings for removal of indicator bacteria have been published elsewhere (Mwabi et al., 2012).

Preparation of synthetic water sample
Bacterial strains of Salmonella typhi murium ATCC 14028 were obtained from the American Type Culture Collection (Quantum Biotechnologies, RSA) and those of Shigella dysenteriae and Vibrio cholerae from the stock cultures of the CSIR in Pretoria, South Africa.These strains were confirmed by cultural tests using selective agar media, according to the methods prescribed by Environment Agency (2002;2006) andAPHA (2005).For each test bacterium, 1 mℓ of an overnight culture was serially diluted in 9 mℓ sterile physiological water (0.9% w/v NaCl) and spreadplated onto selective agar plates.Plates were incubated at 36°C ± 1°C for 24 h and the resulting colonies were counted to express the initial bacterial concentrations as CFU•mℓ -1 .The aliquots of the overnight cultures corresponding to approximately 10 6 CFU•mℓ -1 were inoculated into a 20 ℓ final volume of sterile normal saline water (0.9% w/v).The spiked water samples were shaken vigorously several times before being passed through the filtering devices.

Environmental water sources
Source water samples were collected 6 times from each of 4 different sites in Gauteng Province, South Africa, between 27 September 2010 and 18 March 2011.Surface water samples of low (SWL) and high (SWH) turbidity were collected from the Apies River (Pretoria) and from Hartbeespoort Dam (Hartbeespoort, North West Province), respectively.Groundwater samples with low (GWL) and high (GWH) turbidity were collected from boreholes in Delmas (on the border of the Mpumalanga and Gauteng Provinces) and Wallmannsthal (Gauteng Province), respectively.In this study the raw water was classified according to the turbidity level and was defined as follows: 2 NTU to 18 NTU = SWL, 10 NTU to 40 NTU = SWH, 2 NTU to 10 NTU = GWL and 2 NTU to 15 NTU = GWH.The water samples were collected in sterile plastic containers and transported to the laboratory.During the study period, S. typhimurium, V. cholerae and S. dysenteriae were not detected in groundwater samples and the reference bacterial strains were spiked into these water sources following the same procedure mentioned in the preparation of synthetic water samples.

Operation of the devices
Source-water samples were filtered through each device in the laboratory as follows: 5 ℓ•d -1 for SIPP and 20 ℓ•d -1 for each of the remaining HWTS devices.Different volumes of filtrates were

448
collected at 1 h intervals over the 3 h period of filtration, with the assumption that enough purified water would have been produced in this period for drinking and cooking.One unit of each type of filter was used for the duration of this study.

Flow-rate analysis
The flow rates of the SIPP, CCF and BSF-S were measured by recording the volume of water collected in 1 h, over a period of 3 h, to obtain a triplicate reading.For the BSF-Z, the flow rate was measured by recording the volume of water collected in 1 min, immediately after the water had been poured into the filter; this was done at hourly intervals over 3 h.For the BF, it was measured by recording the time it took to filter 20 ℓ of water.Flow rate was recorded as litres per hour (ℓ•h -1 ) in this study.

Turbidity removal by HWTS devices
A turbidity meter (Eutech, RSA) was used to determine the level of turbidity in water samples before and after filtration.Readings for each sample were taken in triplicate.The percentage turbidity reduction achieved by each of the filter devices was calculated using the following equation:

Performance of the HWTS in removing waterborne bacteria
The presence/absence of the target organisms was detected before and after treatment of the environmental water samples.With some modifications, standard methods (Environmental Agency (2002;2006), which included pre-enrichment steps and streak plate techniques, were used for the isolation and detection of these organisms.For Salmonella spp.and Shigella spp., filtration of 500 mℓ of water sample through sterile 47 mm diameter membranes of 0.45 µm pore size (Millipore) was followed by the immersion of these membranes in 50 mℓ sterile brain-heart infusion broth (Merck, SA) and incubation at 36°C ± 1°C for 6-8 h.For Vibrio spp., a similar pre-enrichment method was used with the exception that the membranes were immersed in 100 mℓ double-strength alkaline peptone water (pH 8.5).Serial dilution of the pre-enriched suspensions and selective culture media ((xylose lysine deoxycholate agar (XLD) and/or Salmonella-Shigella agar (SS) and thiosulfate citrate bile sucrose (TCBS) agar (Merck, SA)) were used for the detection the target organisms as described by standard methods.
Bacterial counts in spiked water samples were calculated and expressed in log 10 units.The log 10 bacterial reduction and percentage removal efficiencies (Brözel and Cloete, 1991) were calculated using the following formula: Following the isolation and detection of organisms, 5 characteristic colonies for presumptive Salmonella spp., Shigella spp.and Vibrio spp.were randomly selected from different plates for each water sample, transferred onto the selective media by the streak-plate method and incubated at 36°C ± 1°C for 24 h.The colonies were further purified by the same methods at least 3 times using nutrient agar (Biolab), and submitted to Gram staining, oxidase test and the API 20E identification systems (bioMérieux, Marcy-l'Etoile, France).Colonies of each target bacterium presumptively identified and confirmed to be positive by the API 20E identification system were sub-cultured onto their respective selective media 3 times before being used for molecular identification.
Two hundred microliter (200 μℓ) of an overnight culture of each target bacterial strain was suspended in sterile Milli-Q water and the genomic DNA was extracted using the ZR Fungal/ Bacterial DNA Kit (ZYMO Research, Pretoria, South Africa) according to the procedures provided by the manufacturer.Species-specific primers were used for the amplification of a specific target gene associated with each target pathogenic bacterium.The ipaH gene encoding for the invasion plasmid antigen H was used to detect S. dysenteriae.The ipaB gene encoding for the invasion plasmid antigen B was used to detect S. typhimurium and the EspM gene encoding for a component of the cytoplasmic membrane protein of V. cholerae was used to identify Vibrio spp.Primers used in this study were synthesised by Inqaba Biotechnical Industries (Pty) Ltd, Pretoria, South Africa.
The PCR amplification of the target bacterial DNA was carried out in a thermal cycler (MJ Mini TM Personal Thermal Cycler, Biorad) using 200 µℓ PCR tubes and a reaction mixture volume of 50 µℓ containing 10 ng to 20 ng of template DNA, 25 µℓ 2 X Dream Taq TM PCR master mix (10X Dream Taq TM buffer, 2 µM dnTP mix and 1.25 u Dream Taq TM polymerase) and a 10 µM concentration of each PCR primer and then made up to 50 µℓ with ultra-pure nuclease-free water.The cycling conditions used by previous investigators were followed after standardisation in our laboratory.For the amplification of target genes of S. dysenteriae, S. typhimurium and V. cholerae the cycling parameters described by Thiem et al. ( 2004), Kong et al. (2002) and Momba et al. ( 2006) were followed, respectively.An aliquot of 10 µℓ of PCR product was electrophoresed through a 1.5% agarose (w/v) gel (Merck, RSA) in 1 U TAE buffer (40 mM Tris-HCl, 20mM Na-acetate, 1 mM EDTA, pH 8.5, Biorad) and stained with 0.5 μg•mℓ -1 ethidium bromide (EtBr, Merck).The amplified product was visualised under UV light in an InGenius L Gel documentation system (Syngene, Vacutec RSA).A 100-bp ladder (Fermentas, supplied by InqabaBiotec SA) was included on each gel as a molecular size standard.The electrophoresis was carried out at 80V for 30 min.A negative control consisted of all PCR reagents except for the template DNA and a positive control having genomic DNA of the reference strains was also included in each PCR run.

Leaching of silver ions in water filtered by SIPP
The preliminary experimental studies to determine the amount of silver ions leached out by the SIPP during filtration of water were conducted by Momba and co-authors (2010b).The authors conducted a series of analytical and mechanical characterisation tests on the SIPP, which included the XRF analysis that confirmed the presence of Ag, the breaking strength and the porosity of the filter.It was found that the silver leached from the SIPP filter at concentrations ranging between 0.5 mg•ℓ -1 and 0.6 mg•ℓ -1 .The Ag elution was greater in the early stages (within the first 5 ℓ) but appeared to begin to stabilise after filtering a total volume of 10 ℓ intake water.
For the present study, the amount of silver leached by SIPP was measured at the end of the study after a total volume of 305 ℓ had been filtered, to determine whether leaching of silver  (2010b).Briefly, the SIPP filter was soaked in 20 ℓ of deionised water overnight prior to use.The concentration of the silver in the filtered water was monitored at 1 h intervals over a 3 h period.The first, second and third filter runs were performed with deionised water, groundwater and surface water, respectively.The Spectro Acros ICP spectrometer (Spectro, RSA) was used to detect and determine the concentration of silver in each water sample.A silver standard was prepared from a solution of silver nitrate (Merck, RSA,) according to the manufacturers' instructions.

Statistical analysis
Data used to assess for water quality before and after treatment through each selected device were subjected to oneway analysis of variance (ANOVA) to compare more than two groups.Comparisons were made between the treatment means of each device per water source to determine if there were significant differences between treatments.Where differences were observed, pair-wise comparisons or post hoc tests were performed and the Wilcoxon Rank-Sum Test was used to compare the two groups.The interpretation was performed at the 95% confidence limit.

RESULTS AND DISCUSSION
The quantity of water produced by household water-treatment systems is important, as each of these systems must be able to produce the minimum quantity of potable water of 25 ℓ•person -1 •d -1 for basic human activities, as prescribed by the Regulations under Section 9 of the Water Services Act (No. 108 of 1997) of South Africa; Norms and Standards for Quality Water Services (DWAF, 2002).The HWTS used in this study could all produce the requisite 25 ℓ•person -1 •d -1 (Fig. 5a-e).The average flow rates obtained ranged from 0.05 ℓ•h -1 to 2.49 ℓ•h -1 for SIPP; from 1 ℓ•h -1 to 4 ℓ•h -1 for CCF; from 0.81 ℓ•h -1 to 6.84 ℓ•h -1 for BSF-S; from 1.74 ℓ•h -1 to 19.2 ℓ•h -1 for BSF-Z; and from 106.5 ℓ•h -1 to 160.5 ℓ•h -1 for BF (Fig. 5).The flow rates of the selected devices were similar to those reported for commonly distributed HWTS in developing countries, which ranged between 10 ℓ•h -1 and 60 ℓ•h -1 for biosand filters (Elliott  , 2008); between 1 ℓ•h -1 and 11 ℓ•h -1 for ceramic candle filters (Brown, 2007); and between 1 ℓ•h -1 and 3 ℓ•h -1 for clay-pot filters (Van Halem et al., 2009).
During the study period, the devices were cleaned twice, when flow rates declined to a point where the minimum required volume of 25 ℓ•p -1 •d -1 was no longer being produced (Fig. 5).The decline in flow rates of the HWTS resulted from the accumulation of dirt and particles on the surface of the fine sand layer of the biosand filters and bucket filter (Ngai et al., 2007) or in the micro-pores of the SIPP and CCF filters (Clasen and Boisson, 2006).The first cleaning was done before filtering groundwater samples of high turbidity (GWH) and the second cleaning was done before filtering surface-water samples of high turbidity (SWH).The biosand filters were cleaned by removing the top 1 cm to 2 cm layer of fine sand, thoroughly washing it with deionised water and thereafter replacing it in the BSF-S and BSF-Z buckets (Lea, 2008).The ceramic candle filter and silver-impregnated porous-pot filter were cleaned by scrubbing the ceramic shell and the inside of the clay pot, respectively, with a Scotch-Brite™ scrub pad and clean water.The cleaning of the filters assisted in regaining the flow rate as shown in Fig. 5a-d.These findings substantiate studies done by Ngai and co-authors (2007) and Low (2002), who reported that flow rates of the biosand filters and ceramic filters increased after cleaning.
Turbidity relates to the degree of microbiological and organic/inorganic chemical content as well as colloidal content of water.The turbidity level of unfiltered water from the four water sources was found to be unacceptable (Table 1, 2) as none complied with the SANS 241 (SABS, 2011) turbidity limit for drinking water in South Africa, which is <1 NTU.After filtration of the various water sources, the turbidity ranged between 0.74 NTU and 1 NTU (64% to 98% reduction) for SIPP; between 0.76 NTU and 7.56 NTU (58% to 94% reduction) for CCF; between 0.79 NTU and 8.2 NTU (11% to 93% reduction) for BSF-S; between 1.07 NTU and 4.6 NTU (45% to 95% reduction) for BSF-Z; and between 0.87 NTU and 11 NTU (14% to 76%) for BF (Table 1).Some of these results are similar to figures reported in the literature, as Low (2002) reported the percentage turbidity removal efficiencies to range from 83% to 99% for ceramic silver-coated water filters.Percentage turbidity removals ranging from 88% to 99% have been also reported for ceramic candle filters (Franz, 2004), while percentage turbidity reductions ranging from 90% to 95% have been reported for biosand filters (Ngai et al., 2004).The BF showed the lowest (14% to 76%) turbidity removal efficiency throughout the study, which was attributed to its high flow rate.Previous studies have shown that high flow rates do not allow enough retention/contact time between the contaminated water and the filter media, consequently reducing the efficiency of the sand media to trap particles (Ngai et al., 2004).It is important to ensure that each filter removes the maximum level of turbidity so that it may provide high-quality drinking water that complies with the SANS 241 (SABS, 2011) turbidity limit.
The main objective of this study was to remove and/or inactivate specific diarrhoeagenic bacteria from water samples.A mentioned above, no presumptive Shigella, Salmonella and Vibrio spp.were detected in groundwater source samples.

451
Consequently, the reference strains of the target organisms were spiked into the water samples (Table 2).For surface water sources, the analysis of the water samples resulted in the isolation of presumptive V. cholerae, S. typhimurium and S. dysen teriae by culture-based methods using selective media.These results indicated that 100% of water samples collected from both surface water of low turbidity (SWL) and surface water of high turbidity (SWH) were contaminated with the target organisms (Table 2).Results obtained from molecular studies confirmed that the selected colonies of presumptive Shigella and Salmonella found in surface water samples were definitely S. dysenteriae and S. typhimurium (Table 3) and those of presumptive Vibrio spp.colonies obtained from the Apies River (SWL) and Hartbeespoort Dam (SWH) water samples, were V. cholerae (Table 3).The synthetic water (sterile saline solution) samples were spiked with pure laboratory cultures of each target organism to an average concentration of 6 log 10 for synthetic water, while groundwater samples were spiked with an average concentration ranging between 3 and 5 log 10 CFU•mℓ -1 .
Although the bacterial quality of the intake water samples varied during each trial, there was a considerable reduction in bacterial counts after filtering the water samples.In general, the performance of each device depended on the type of organism.No target organisms were detected in both types of surface water sources after filtration by the SIPP filter (Table 4).The removal of maximum concentrations (6 log 10 to 7 log 10 units; >99.99% removal) of target pathogenic bacteria from synthetic water and groundwater (0.6 log 10 to 5 log 10 units; 99% to 100% removal) was observed after treating this water source with SIPP (Table 4).These findings are similar to the results of a study done by Van Halem (2006), where the author attempted to determine the highest possible reduction of target pathogenic bacteria by a silver-impregnated clay-pot filter.This author spiked extremely high concentrations of E. coli K12, which resulted in a 7 log 10 unit reduction.Furthermore, it was noted that, of the five selected devices, SIPP was the only device that achieved total removal of pathogenic bacteria from both groundwater and surface-water samples.This is further supported by published results of faecal coliform and E. coli by these filters, which showed total removal in SIPP (Mwabi et al., 2012).This HWTS therefore consistently produced high-quality drinking water throughout the study (Table 4).The mechanism by which the SIPP successfully removed pathogenic bacteria was by filtering them out in the fine 0.2 µm to 0.5 µm micropores created in the clay pot when the clay is kiln-fired during the manufacturing process, burning off the sawdust (Van Halem et al., 2006;Momba et al., 2010b).It is also possible that the silver nano-particles, which are embedded within the clay during manufacturing, contribute to high pathogen-removal efficiency (Michen et al., 2011).Previous studies have reported the effect of Ag in a water-purification application, irrespective of substrate and have revealed that Ag ions have antiviral and bacteriostatic properties (Nangmenyi et al., 2009;Michen et al., 2012).A study by Momba and co-authors (2010b) revealed that only the Ag-impregnated pot was significantly more effective in removing E. coli, compared to the control pot that had not been impregnated with AgNO 3 .This superior performance was attributed to the Ag nanoparticles embedded in the micropores.
Prior to filtration, 0.13 mg•ℓ -1 and 0.07 mg•ℓ -1 of silver were detected in groundwater and surface water samples, respectively.The amount of silver that leached from the SIPP filter ranged between 0.54 mg•ℓ -1 and 0.98 mg•ℓ -1 , between 0.22 mg•ℓ -1 and 0.28 mg•ℓ -1 and between 0.24 mg•ℓ -1 and 0.28 mg•ℓ -1 during filtration of deionised water, groundwater and surface water, respectively.The results obtained from groundwater and surface water showed that the amount of silver leached by SIPP decreased over time (Fig. 6), as the results were much lower than those reported by Momba and co-authors (2010b), who found that 0.5 mg•ℓ -1 to 0.6 mg•ℓ -1 of silver was initially leached from the SIPP device (Fig. 6).These results exceeded the WHO (2011) MCL for silver (0.1 mg•ℓ -1 ).To date, the only known health consequence of excessive silver intake is a condition known as argyria, which may develop due to improper

452
exposure to chemical forms of the element silver, in which skin and hair become discoloured by silver accumulation.This has been linked to excessive consumption of medications containing silver and not to the use of silver in drinking-water disinfection devices (Lantagne, 2001).
In this study, the CCF removed S. dysenteriae, S. typhimu rium and V. cholerae, ranging from 1.3 log 10 to 2.6 log 10 units (93% to 99.6%), from synthetic water, and from 1 to 3.9 log 10 units (90% to 99.9%) from groundwater samples.No V. cholerae and S. typhimurium were detected in any of the surface water sources after filtration, while S. dysenteriae was detected in 1/18 (5%) of the water samples collected from the CCF (Table 4).The findings for synthetic water and groundwater samples spiked with target organisms substantiate the findings reported by Clasen and Boisson (2006) that showed that ceramic candle filters can remove up to 99.99% or 4 log 10 units of faecal bacteria.It has been reported that the CCF removes bacteria from water by surface filtration and depth filtration.Some contaminants are trapped on the surface of the ceramic candle, as they are too large to pass through the fine pores, while smaller particles may be trapped within the pore channels of the ceramic candle (CAWST, 2010).These findings are further supported by results that indicated that the CCF removed 1.8 to 3.2 log 10 (98% to 99.9%) of E. coli (Table 4) (Mwabi et al., 2012).
In general, the average removal of pathogenic bacteria by BSF-S ranged from 3.7 log 10 to 4.8 log 10 units (>99.99 % removal) for synthetic water and from 0.6 log 10 to 3.9 log 10 units (60% to 100% removal) for spiked groundwater samples.Bacterial removal by BSF-Z ranged from 3.3 log 10 to 4.6 log 10 units (>99.9 to 100% removal) for synthetic water and from 1 log 10  † The log and percentage bacterial removals were calculated from average values of triplicate samples within 1 h, 2 h and 3 h after filtration.to 4.1 log 10 units (90-100% removal) for spiked groundwater samples (Table 4).No V. cholerae and S. typhimurium were detected in any of the surface water sources after filtration in both the BSF-S and BSF-Z, while S. dysenteriae was detected in 2/18 (11%) of the water samples collected from the BSF-S (Table 4).The bacterial removal efficiencies of the biosand filter (BSF) have been reported to range between 60% and 100% (Ngai et al., 2007 andDevi et al., 2008).The higher performance of the BSF-S can be attributed to the enhanced straining efficiency through the fine pores formed in the filter media of 0.15 mm grain size, in combination with the development of the biological layer over time.Elliott and co-authors (2008) have reported that the removal of bacteria by biosand filters at the initial stage occurs by sedimentation and straining.With frequent use of these filters, the removal efficiencies increased, sometimes up to 100%.This is due to the maturation of the biological layer (schmutzdecke) which consists of bacteria, algae, protozoa and invertebrates that enhance bacterial removal.The constant resting water level in the biosand filters was a major factor for the development of the schmutzdecke (Sobsey et al., 2008).It is possible that the natural zeolite used as the filter media in the BSF-Z contributed to the improved performance of this filter.Kallo and Ming (2001) and Misaelides (2011) have previously reported that natural zeolites can remove bacterial pathogens as well as viruses in wastewater.Removal occurs by attachment of microorganisms to the large crystalline surface of the zeolite.
The experimental studies using surface water of high turbidity (SWH) showed that the performance of the BSF-Z was higher than that of the BSF-S in removing pathogens.The BSF-S showed average bacterial removal efficiencies from spiked GWL and GWH of 2.5 log 10 units (99%) and 0.6 log 10 units for S. dysenteriae, 2.2 log 10 units (99.2%) and 3.4 log 10 units (>99%) for S. typhimurium, and 1.3 log 10 units (94.2%) and 3.9 log 10 units (100%) for V. cholerae, respectively (Table 4).The BSF-Z showed average removals of bacteria from spiked GWL and GWH of 2.7 log 10 units (99.7%) and 1.0 log 10 units (90%), 2.1 log 10 units (99%) and 4.1 log 10 units (100%), and 3.9 log 10 units (100%) and 2.4 log 10 units (99.4%) for the abovementioned pathogens, respectively (Table 4).These findings vindicate reports by Ricke and co-workers (1995), who found that zeolites are able to inhibit the number of viable S. typhimurium in water as natural zeolites have antimicrobial properties in soil and water (Uchida et al., 1995).It has also been reported that zeolites can adsorb cholera toxins (Ravin et al., 1997).The bucket filter (BF) exhibited the lowest reductions of the target pathogenic bacteria for all water sources tested, compared to the other HWTS devices, and the growth detected in the enrichment cultures of the target organism was an indicator of this filter's poor performance.Removal of pathogens from synthetic water ranged from 1.7 log 10 to 3.6 log 10 units (97% to 99.9%), while removal efficiency from spiked groundwater ranged from 0.4 log 10 to 3.7 log 10 units (40% to 99.99%).No V.cholerae and S. dysenteriae were detected in SWL after filtration with the BF but S. typhimurium was detected in 1/18 (5%) of the water samples collect from the BF.After filtration of SWH by the BF, both V.cholerae and S. dysenteriae were detected in 2/18 (11%) and 3/18 (17%), respectively, of the samples collected from this filter (Table 4).The poor bacterial removal efficiency was a consequence of the rapid flow rate of this device.The high flow rate reduced the retention time between filter media and contaminants within the water (Campos et al., 2002).This further explained why turbidity reductions achieved by this filter were poor (14% to 76%) compared to those of the other filters (Table 1).As mentioned earlier, high turbidity levels are associated with high microbiological contamination.Another factor that could be linked to the poor performance of this filter was that there was no resting water level in this device.Hence the filter media always dried out before the next filter run.This

454
means that no biological layer could be established in this filter to enhance its performance.Therefore, bacterial removal in the BF occurred through mechanical trapping only (Elliott et al., 2008;Sobsey et al., 2008).
Statistical analysis of the data showed that the flow rate of the SIPP was significantly lower (Table 5) compared to the flow rates of the remaining four devices (p<0.05).The pores that formed within the SIPP, of 0.2 µm to 0.5 µm (Momba et al., 2010b), were much finer than the pores formed in the other filters and hence they reduced the flow rate of the SIPP.The flow rate of the CCF was found to be similar to that of the SIPP when filtering SWH (p>0.05).This is due to the fact that the flow rates of these two devices had declined to <1 ℓ•h -1 at that stage during the treatment of water samples (Fig. 6a-b).
In general, the SIPP turbidity removal efficiency was found to be significantly higher (Table 6) compared to those of the CCF, BSF-S and BF (p<0.05).Statistical data also showed that the SIPP and the BSF-Z performed similarly in reducing turbidity (p>0.05).Based on the fact that the SIPP filter was the only device that consistently produced water that contained 0 CFU•100 mℓ -1 of bacterial contaminants and had higher turbidity removal efficiency, its performance was compared to those of the other four HWTS (Table 5).Statistical comparison of the devices showed that the SIPP filter had the highest bacterial removal efficiency for the water samples tested.This was expected as 100% bacterial removal efficiency was consistently achieved by the SIPP (p >0.05), whereas water produced by the remaining four devices still contained some bacterial contaminants (Table 4).Statistical analysis also revealed that the BSF-Z had a significantly superior performance (p <0.005) in removing Salmonella and Shigella spp.from SWH compared to the BSF-S, as well as a higher performance in removing Vibrio spp.from SWL (Table 5).The removal of pathogens by BSF-Z could be occurring via particle straining and it is possible that the zeolite could be enhancing the inactivation of S. typhimurium and V. cholerae (Uchida et al., 1995;Ravin et al., 1997).It is of great importance that the use of POU household water treatment devices be coupled with safe water storage practices to prevent recontamination of treated water, and rural communities should be advised to keep storage containers clean at all times (Potgieter et al., 2009).

CONCLUSIONS AND RECOMMENDATIONS
The aim of this study was to find a sustainable and affordable solution for improving drinking water quality in rural areas by evaluating 5 household water-treatment devices that are affordable to rural communities, easy to manufacture and operate, and, most importantly, remove turbidity and pathogens effectively from contaminated water.Based on the findings of this study the use of the silver-impregnated porous pot can be recommended, as it was the only filter that produced safe drinking water, with <1 NTU for turbidity level and 0 CFU•100 mℓ -1 for pathogenic bacteria.To prolong the lifespan of the filter, it is recommended that highly turbid water is pre-treated by settling or straining before filtering in the SIPP.The CCF and biosand filters can also be used due to their high flow rates, but would require additional pre-treatment steps to further disinfect the water produced.The bucket filter (BF) that consistently produced >100 ℓ•h -1 for the duration of the study showed the poorest performance in terms of removing turbidity and bacterial pathogens.This device is not recommended for household water-treatment purposes, but could be used as a pre-treatment filter for high-turbidity (>50 NTU) water samples.A future investigation can include a study whereby the turbidity and bacterial removal efficiency of the SIPP filter are compared to a non-silver clay pot to confirm whether silver is crucial for superior performance.Further studies should also be conducted on the biosand filter containing zeolite, as it is novel and has shown promising results in removing turbidity and bacterial pathogens.A social acceptance study coupled with a workshop to introduce the devices to rural communities is another approach that can add value to these findings.

Figure 2
Figure 2 (a) Schematic diagram; and (b) photograph of the ceramic candle filter (CCF)