Enhancing the quality of arsenic-contaminated groundwater using a bio-sand filter with iron-mixed clay pellets

The direct consumption of arsenic (As) contaminated drinking water, at any high concentration, can lead to the development of many incurable or fatal diseases. More than 150 million people worldwide are at risk from As contamination in drinking water, the majority living in countries like Bangladesh, Cambodia, China, India, Laos, Myanmar, Nepal, Pakistan, Taiwan and Vietnam (Singh et al. 2015). In Cambodia, large numbers of households, particularly in rural areas, still rely on groundwater from boreholes for drinking because it is considered relatively free of pathogens. In many regions, however, such as Kratie, Kandal, and areas south and southeast of Phnom Penh, elevated As concentrations, exceeding the WHO recommended guideline level of 10 μg/L, have been reported (Luu et al. 2009; Karagas et al. 2015). This may arise from natural enrichment by geothermal activity in the upper Mekong basin (Luu et al. 2009).

The concentration of As in groundwater in Kandal province has been reported at up to 1,543 μg/L, with As concentrations up to 6,000 μg/L elsewhere (Luu et al. 2009; Kang et al. 2014). Arsenic-poisoning (arsenicosis) is a major concern to millions in Cambodia as water consumption from tube wells is high (Kang et al. 2014). Thus, treating As-contaminated groundwater to provide potable supplies at household-scale is very necessary.

Numerous conventional, household-scale, water treatment systems are available, mostly for enhancing the microbial quality of water, and include filtration, flocculation, chlorination and solar disinfection (Ahammed & Davra 2011). Bio-sand filters (BSFs) have been gaining more attention for household water treatment since their initial design and development in the early 1990s, because of their high pollutant removal efficiency, technical simplicity, cost-effectiveness, low maintenance needs, ability to produce large volumes of treated water, and use of local materials (Baig et al. 2011; Mahmood et al. 2011; Stauber et al. 2012). The BSF is a small-scale, intermittently operated, slow sand filter with concrete filter bodies for household use (Ahammed & Davra 2011; Stauber et al. 2012). Concrete BSFs use several types of material, take lots of time to build and are heavy, making them difficult to transport. The recently introduced plastic BSF may overcome those drawbacks.

The conventional BSF was initially designed to remove suspended solids and microbes using sand media, considered a non-reactive material in filtration (Noubactep & Caré 2010; Baig et al. 2011). To improve its As-removal performance, more affinitive and active media must be used. The strong affinity between inorganic As species and iron is well known (Kang et al. 2014). Many iron-based adsorbents have shown promise in As removal, e.g., montmorillonite-supported nanoscale zero-valent iron, porous ceramic adsorbent, iron-mixed ceramic pellets, and iron-mixed porous pellet adsorbent (Chen et al. 2010; Shafiquzzam et al. 2013; Bhowmick et al. 2014; Te et al. 2017). However, those studies were conducted in batch mode and no study has been made of the use of iron-mixed porous adsorbent in a BSF to treat a real As-contaminated groundwater.

In this study, a plastic BSF was used, with an iron-mixed porous pellet adsorbent as the active medium between the sand layers, to remove As from contaminated groundwater. Preferential flow rate, filtration rate, As adsorption efficiency, pH variation, turbidity removal, iron and organic carbon leaching, and dissolved oxygen (DO) were investigated.

The iron-mixed porous pellet adsorbent was prepared by binding iron oxide (19.22%), iron powder (28.63%) and rice bran (15%) with natural clay (52.15%). Iron oxide powder (Fe₂O₃) and iron powder (Fe⁰) were purchased from Italmar chemical supply company (Thailand) and a local supply store, respectively. The natural clay used was collected from the field in Dankwian District, Nakhon Ratchasima 30000, Thailand. Deionized water was added slowly to the mixture to produce a homogeneous paste, which was stirred strongly by hand for 5 to 10 minutes and dried at 104 ± 1 °C for 24 hours, before further heating at 600 °C for 2 to 3 hours in a muffle furnace to carbonize the rice bran. After cooling, the product was sieved to produce the desired particle size of 0.6 to 1.18 mm. The resulting adsorbent had a specific surface (BET) of 19.393 m²/g, mean pore size 20.169 nm, particle density 1.986 g/cm³, skeletal density 2.465 g/cm³, and point of zero charge 7.50.

The filter was built with 5 mm thick polyvinyl chloride pipe bought from the local market. It was 1 m tall with internal diameter 20 cm. It is referred to herein as the lightweight modified bio-sand filter (LMBSF) and illustrated in Figure 1. The filter consisted of 5 cm of gravel (2 to 4 cm), 5 cm of coarse sand (1.5 to 3 mm), and 60 cm of fine sand (0.5 to 1.5 mm), the latter including a 10 cm iron-mixed porous pellet adsorbent layer in the middle. To minimize air space development and short circuiting, the media were placed in the filter with water present. A plastic diffusion plate, 5 cm above the retaining water, helped distribution of the daily influent water. The filter's maximum holding capacity was 31.4 L and it was operated intermittently using real As-contaminated groundwater collected from the field twice a day. The second charge was done 6 hours after the first and each charge comprised 15 L. The experiments were conducted at room temperature – i.e., 25 to 35 °C. The raw water never stood, after collection, for more than 12 hours prior to being treated.

The influent water was collected from a tube well at Dei Eith Primary School, Kien Svay district, Kandal province, Cambodia, which is in a well-known As contamination zone. The feed water characteristics are presented in Table 1, and it is noted that the water's As concentration is very high compared to the maxima recommended by both WHO (10 μg/L) and the Cambodian authorities (50 μg/L) (Kang et al. 2014).

The tracer test was conducted to evaluate the hydraulic performance of the filter, particularly with respect to preferential flow paths, if any. The test was run at the start of treatment before any As-contaminated water had been put through the system. The filter was fed with 2.5 L of 300 mg-NaCl/L solution (the tracer) after the retaining water was decanted, after which distilled water was poured in continuously to push the tracer further down in the media. Effluent samples were collected at pre-set time intervals and their electrical conductivity (EC) measured is a proxy for NaCl concentration.

Influent and effluent samples were collected at the start of feeding and at the end of the last raw water feed, respectively. Sample portions were filtered immediately through a 0.22 μm syringe filter, acidified with concentrated HNO₃, and kept at 4 ± 1 °C for As determination. The remaining samples were analyzed for parameters including turbidity, pH, DO, iron, and total organic carbon (TOC). The pH, DO and EC were measured using digital meters, and turbidity was measured with a Hach turbidimeter. Iron was analyzed by colorimetric spectrophotometry, and TOC was analyzed using the Walkley-Black titration method. The As concentration was determined by inductively coupled plasma-optimal emission spectrometry (ICP-OES, Optima 8000, PerkinElmer, USA).

The EC values from the tracer test produced a bell-shaped curve, indicating that the tracer concentration rose rapidly in the treated effluent and then dropped sharply to its initial concentration while being washed out (Figure 2). From the cumulative fraction data, the times for 10 and 90% tracer recovery from the filter were about 19 and 35 minutes, respectively. Using the technique, demonstrated by Tchobanoglous et al. (2003), the Morrill Dispersion Index (MDI) and volumetric efficiency were 1.82 and approximately 55%, respectively. The United States Environmental Protection Agency (USEPA) classifies flow with an MDI below 2.0 as effective plug flow, while the MDI of an ideal plug flow reactor is 1.0 (Bradley et al. 2011). Thus, preferential flow could be established for aqueous solutions through the media in LMBSF.

The filtration rate was observed by measuring the flow rates at the start (immediately after charging), and 5 and 10 minutes after introduction of 5 L of influent. The results are presented in Figure 3. The starting flow rate declined progressively from 678 to 166 mL/min after 30 days of testing. The 5 and 10 minute flow rates fell gradually from 267 to 140 mL/min and 122 to 114.5 mL/min, respectively. The fall in hydraulic flow rate over time was caused by the drop in water level in the headspace. The fall observed over the test period is attributed to maturation and particle accumulation in the filter media. Head loss accumulation has a significant effect on the time needed to provide the desired effluent volume. The production of 15 L of As-treated groundwater was approximately 44 minutes for the first 3 operating days, but rose to about 3 hours on the last day.

The influent and effluent As concentrations, and the filter's As-removal efficiency are illustrated in Figure 4. The results show that, with the groundwater As-concentration in the range 355 to 587 μg/L, the filter achieved removal efficiencies of between about 97 and 99.5%, during the 30-day study period. For the first 24 days, As-removal efficiency varied little, fluctuating slightly at over 99%, with the effluent arsenic concentration within the WHO drinking water guideline. The removal efficiency decreased, with significant fluctuation, after 24 days, falling to about 97% on day 30. The effluent As-concentration on day 30 was about 16.5 μg/L, exceeding the WHO guideline, and the filter was stopped.

As-removal performance decreased with time, suggesting surface saturation in the filter media. The presence of Fe in groundwater can lead to arsenic removal through co-precipitation, when dissolved Fe is oxidized to form iron (hydro)oxide precipitates (Roberts et al. 2004; Smith et al. 2017). However, the contribution of Fe to arsenic removal can be altered by the presence of phosphate in groundwater, as the latter is well known as a competitor with arsenic for adsorption sites on iron oxide (Tyruvola et al. 2006; Chiew et al. 2009). Some previous researchers have attempted to introduce modified BSF for arsenic removal but the efficiency of their systems was low compared to that of LMBSF. Ngai et al. (2007) introduced brick chips and iron nails on the diffuser in the BSF to treat groundwater (>50 μg-As/L) and achieved arsenic removal from 88 to 95%. Similarly, Chiew et al. (2009) found that a BSF containing iron nails could achieve As uptake efficiency between 39 and 75% from groundwater with As concentration exceeding 146 μg/L, phosphorus >0.91 mg-P/L and iron <5 mg-Fe/L. Smith et al. (2017) report that nails embedded in the top sand layer treated influent As (226 to 240 μg/L) to achieve a maximum removal rate of 81%. Shah et al. (2013) used Pinus bark and brick powder in the filter's middle sand layer to remove influent arsenic ranging from 50 to 350 μg/L, without involving co-existing ions, and obtained As-removal of between 80 and 100%. Baig et al. (2013) used iron-coated, honeycomb briquette cinders in the middle sand layer and demonstrated that the As-removal efficiency from an influent containing 200 μg/L of arsenic dropped to 60% after 24 days.

The pH of the influent and effluent waters changed within the ranges 7.0 to 7.4 and 7.4 to 7.7, respectively (Figure 5). Increased effluent pH levels up to 9.0, from the influent level of 7.5, were observed in a sand filter containing iron filings and powder (Biterna et al. 2007). In an arsenic filter in Kanchan-Nepal, an increase of 0.37 pH units was reported after filtration (Ngai et al. 2007). Baig et al. (2013) reported that, using their sand filter containing iron-coated honeycomb briquette cinders, the effluent pH was in the range 7.5 to 8.1 when the influent pH was 7.1. The increases in pH after filtration could arise from carbonate mineral dissolution from the sand particles, and/or the release of OH⁻ groups as a result of ligand exchange during adsorption onto the adsorbent surface (Biterna et al. 2007; Ngai et al. 2007; Tiwari & Lee 2012).

The influent and effluent turbidities ranged from 42 to 75 NTU and 0.2 to 1.6 NTU, respectively, representing removal efficiency of 97 to 99% (Figure 6). The filter produced effluent turbidity below 5 NTU throughout the treatment period, perhaps because of pollutant-ion attachment on the filter media surface as well as the presence of precipitates from the influent. Turbidity removal by LMBSF exceeded that from some previous studies, e.g., 93% for Kanchan (Ngai et al. 2007), up to 96% for a BSF modified with iron oxide-coated sand (Ahammed & Davra 2011), and up to 91.5% for a brass/zero valent iron filter (Yildiz 2016).

The influent iron concentration varied between 4.1 and 6.6 mg/L. The effluent iron concentration, however, was between 0.01 and 0.03 mg/L, below the maximum recommended level (0.3 mg/L) for drinking water – see Figure 7(a). Iron removal efficiency exceeded 99% throughout the treatment period. The dissolved iron in the influent groundwater could be oxidized and precipitated, and then trapped in the filter media layers. Low iron concentrations (<0.3 mg/L) in the treated water improve its appearance and taste, and make its use socially acceptable (Ngai et al. 2007).

Figure 7(b) shows the TOC concentrations of the influent and effluent during the trial. The influent TOC varied between 31.2 and 74.1 mg/L – the limit of detection was 2 mg/L and TOC was never detected in the effluent. The filter removed TOC efficiently. Even using rice bran as the porosifier, there was no leaching of organic carbon to the treated water.

The effluent DO concentration is shown in Figure 8. It appears that the oxygen concentration in water passing through the filter decreased during the 30-day treatment. Initially, the effluent DO was 8.1 mg/L, but it had fallen to 6.9 mg/L on the last day. DO concentrations exceeding 1 mg/L could imply aerobic conditions in the filter bed (Yildiz 2016) and this might contribute to improved As-removal. The predominant arsenic species found in groundwater is trivalent but it oxidizes quickly to pentavalent arsenic, which binds more strongly to iron hydroxides (Baig et al. 2013). However, the results obtained may not represent aerobic conditions within the filter bed, but rather interference from atmospheric oxygen at the outlet during sample collection. Further research is needed to develop suitable methods for measuring DO concentrations within the filter bed.

Conventional BSFs could be modified using lightweight materials and more active adsorbents as filter media, to improve the quality of As-contaminated groundwater. An iron-mixed porous pellet adsorbent embedded in the middle sand layer of the LMBSF has been shown to remove arsenic efficiently. With a daily influent charge of 30 L, the filter's effluent arsenic concentration was below the WHO guideline for 29 days of the trial. The breakthrough concentration would be later than this if the Cambodian standard for arsenic were applied. The filter also removed high turbidity efficiently so that, throughout the 30-day period, the effluent turbidity was below the recommended maximum for drinking water (5 NTU). Using iron oxide, iron powder and rice bran to develop the adsorbent used in the filter does not give rise to any health risk in the filtered water. Introducing a surface-amended adsorbent layer into the LMBSF should be sufficiently safe and provide arsenic-safe drinking water at household-scale in arsenic contaminated areas.

This study was supported by the Center for Scientific and Technological Equipment, and School of Environmental Engineering, Institute of Engineering, Suranaree University of Technology, Nakhon Ratchasima 30000, Thailand.