lable at ScienceDirect

Chemosphere 168 (2017) 155e162

Contents lists avai

Chemosphere

journal homepage: www.elsevier .com/locate/chemosphere

Comparison of sand-based water filters for point-of-use arsenicremoval in China

Kate Smith a, b, Zhenyu Li b, Bohan Chen c, Honggang Liang d, Xinyi Zhang a, b, Ruifei Xu a, e,Zhilin Li f, Huanfang Dai a, b, Caijie Wei a, b, Shuming Liu b, *

a RISE, Tsinghua University, Beijing 100084, Chinab School of Environment, Tsinghua University, Beijing 100084, Chinac College of Physics and Optoelectronics, Taiyuan University of Technology, Taiyuan 030024, Chinad College of Environmental Science and Engineering, Taiyuan University of Technology, Taiyuan 030024, Chinae School of Social Sciences, Tsinghua University, Beijing 100084, Chinaf College of Computer Science and Technology, Taiyuan University of Technology, Taiyuan 030024, China

h i g h l i g h t s

* Corresponding author.E-mail address: shumingliu@tsinghua.edu.cn (S. L

http://dx.doi.org/10.1016/j.chemosphere.2016.10.0210045-6535/© 2016 Elsevier Ltd. All rights reserved.

g r a p h i c a l a b s t r a c t

� 5 filters were compared to find themost effective for an As-affectedvillage.

� Iron-based filters can remove As, butremoval is affected by filter design.

� Nails placed in biosand filter sandwere more effective than thoseplaced above.

� The arsenic biosand filter (nails abovesand) rarely removed arsenic to<50 mg/L.

� The biosand filter with embeddednails removed arsenic to <50 mg/L forsix months.

a r t i c l e i n f o

Article history:Received 20 March 2016Received in revised form26 September 2016Accepted 7 October 2016

Handling Editor: X. Cao

Keywords:Biosand filterContaminationDrinking waterIronSONO filter

a b s t r a c t

Contamination of groundwater wells by arsenic is a major problem in China. This study compared arsenicremoval efficiency of five sand-based point-of-use filters with the aim of selecting the most effectivefilter for use in a village in Shanxi province, where the main groundwater source had arsenic concen-tration >200 mg/L. A biosand filter, two arsenic biosand filters, a SONO-style filter and a version of thebiosand filter with nails embedded in the sand were tested. The biosand filter with embedded nails wasthe most consistent and effective under the study conditions, likely due to increased contact time be-tweenwater and nails and sustained corrosion. Effluent arsenic was below China's standard of 50 mg/L formore than six months after construction. The removal rate averaged 92% and was never below 86%. Incomparison, arsenic removal for the nail-free biosand filter was never higher than 53% and declined withtime. The arsenic biosand filter, in which nails sit in a diffuser basin above the sand, performed better buteffluent arsenic almost always exceeded the standard. This highlights the positive impact on arsenicremoval of embedding nails within the top layer of biosand filter sand and the promise of this low-costfiltration method for rural areas affected by arsenic contamination.

© 2016 Elsevier Ltd. All rights reserved.

iu).

K. Smith et al. / Chemosphere 168 (2017) 155e162156

1. Introduction

Presence of unsafe levels of arsenic in drinking water is a majorproblem in China. Arsenic is a carcinogen so excessive exposurethrough drinking water leads to increased risk of cancer of the skin,lungs, bladder and kidney (Kapaj et al., 2006). To avoid this risk, theWorld Health Organisation (WHO) recommends arsenic levels indrinking water be lower than 10 mg L�1 (WHO, 2011). It has beenestimated that around 20 million people in China are exposed towater containing arsenic higher than this level (Rodriguez-Ladoet al., 2013).

Shanxi province is one of the top three provinces in China forexposure to arsenic in drinking water (Yu et al., 2007). A studyconducted between 2001 and 2005 found that over 12% of wells inShanxi had arsenic levels over 50 mg L�1 (Yu et al., 2007), which isChina's national standard for small-scale distributed water supply(PRC, 2007). Datong and Taiyuan basins are the two areas mostaffected by arsenic poisoning (Zhang et al., 2013).

Arsenic occurs naturally in the Earth's crust and can leach intogroundwater if conditions are favorable. Exposure to arsenic indrinking water in China is most common where water is supplieduntreated from groundwater wells, which tends to be in rural areas.The Chinese government has been working to improve coverage ofcentralized water supply networks (Xinhua, 2013), but many vil-lages still rely on water sources with arsenic over China's nationalstandard. In these cases, it is important for households to haveaccess to a point-of-use filtration method that removes arsenic.This method should ideally be low cost and simple to install. Itshould also consistently reduce arsenic to below 50 mg L�1.

Arsenic can be removed using a number of different methodsand materials, which are categorized under oxidation, coagulation/flocculation and adsorption and described in detail by Kowalski(2014). Iron-water systems generally remove arsenic throughadsorption and co-precipitation (Ghauch, 2015) and are popular foruse by households because they remove contaminants and havegood reactivity under natural conditions, while also being lowenergy, low cost and lowmaintenance (Tepong-Tsinde et al., 2015).Iron is not only used for removal or arsenic or limited to householdscale. It was applied at water treatment plant scale by Kowalski andSøgaard (2014) with an aeration unit separating the iron and sandcomponents and has been used for removal of uranium fromdrinking water wells (Gottinger et al., 2013) and a number ofpositively charged species (Noubactep, 2015). Issues do exist withthe use of iron in filters. One of the major flaws is loss of perme-ability (Noubactep, 2014). For this reason, household filters like theSONO filter are described as using the next generation of filtermaterials (Rahman et al., 2013). This filter uses porous iron com-posite to overcome the rapid decrease in water flow rate thatcharacterized earlier generations of iron-based household filters(e.g. 3-Kolshi filter) (Noubactep, 2009; Noubactep et al., 2009).Since the creation of the SONO filter, other iron-based porousmaterials have been tested for use in filters (Rahman et al., 2013).Nails can still be used as the active iron material (as shown inChaudhari et al. (2014)), with periodic agitation of the iron nails todislodge the build-up of iron oxides.

This study tested the ability of a number of sand and iron-basedhousehold filtration devices to remove arsenic from influentgroundwater, with the aim of identifying the most effective filterfor use in a village in Shanxi province. The location of the controlledfield study was Liangjiabu village in Taiyuan basin, where the mainvillage well had an arsenic concentration of over 200 mg L�1. Thefollowing sand-based filterswere chosen for this study: the biosandfilter, the arsenic biosand filter, a close adaptation of the SONOfilter, and a modified version of the biosand filter with nailsembedded in the sand. These low-cost filters were operated by

households for approximately five months. As iron is known toassist arsenic removal, the biosand filter with no added iron actedas a control. All other filters contained 5 kg of iron.

The biosand filter (control filter) is an adaptation of the tradi-tional slow sand filter that removes microbial contamination fromwater, as described in Ngai et al. (2007) with the most significantelement being the biologically active biofilm that forms on the toplayer of sand (Chiew et al., 2009). The arsenic biosand filter (aconventional filter also known as the Kanchan arsenic filter) em-ploys the same design, with the addition of a diffuser basin filledwith rusting nails placed in the top of the filter. Arsenic in waterflowing through the nails is adsorbed by rust and precipitates, withremoval being achieved when precipitate is trapped in the sand(Ngai et al., 2007). The original SONO filter is a successful innova-tion in the area of sand-iron household filters. It consists of twobuckets. The bottom one contains layers of sand and activatedcarbon or charcoal and the top one contains layers of sand and aporous layer of loose rusted iron filings (also called the compositeiron matrix (CIM)), which are prepared through a process of wet-ting and drying (Hussam, 2010; Neumann et al., 2013). In the SONOfilter, water is poured into the top bucket and filters down throughthe iron filings. This causes corrosion of the iron and the formationof iron phases that remove arsenic from thewater and trap it withinthe iron matrix (Neumann et al., 2013). The SONO-style filter usedin the current study was an adaptation of the SONO filter con-structed by following the SONO filter patent (Hussam, 2010) asclosely as possible. The final filter used for comparison in this studywas an innovative modification of the original biosand filter. In thissystem, nails were embedded just under the top layer of sand in thebiosand filter so as to investigate the impact of increasing contacttime between the iron and water as suggested by Chiew et al.(2009) and Noubactep et al. (2009). A total of five filters wereinstalled in five different households using the same influent water.These included two identical arsenic biosand filters, with onehousehold instructed to filter all water twice through to increasecontact time between the nails and water.

The arsenic biosand filter, SONO-style filter and the use of iron insand filters to remove arsenic have separately been the subject ofprevious studies, both in the field and the laboratory (Leupin andHug, 2005; Hussam and Munir, 2007; Ngai et al., 2007; Chiewet al., 2009; Neumann et al., 2013; Singh et al., 2014; Wenk et al.,2014). A number of theory-based studies are available that intro-duce the concept of embedding iron nails in the biosand filter(Noubactep et al., 2010, 2012; Tepong-Tsinde et al., 2015) and astudy by Bradley et al. (2011) applies this construction for virusremoval. To the knowledge of the authors, this is the first study tocompare and provide experimental results on arsenic removal ef-ficiency of the biosand filter, biosand filter with embedded nails,arsenic biosand filter and a SONO-style filter for influent wateroriginating from the same high arsenic groundwater source.

2. Materials and methods

2.1. Filter construction

The biosand filter and two arsenic biosand filters (also known asKanchan Arsenic Filters) used in this experiment were built ac-cording to Ngai et al. (2007) Each filter was made using a largewashed plastic bucket of height 70 cm and estimated capacity 80 L.Groundwater was added and the following layers were loaded intoeach bucket in order and leveled: 7 cm of large washed gravel withdiameters 5e13 mm; 3 cm of small washed gravel with diameters3e5mm; 30 cm of washed sand sieved through a 2mm screen; and5 cm of unwashed sand (<2 mm) as the final layer. The outlet pipewas cut so that a 5 cm layer of standing water remained above the

K. Smith et al. / Chemosphere 168 (2017) 155e162 157

top layer of sand. A circular piece of plastic was placed on thesurface of the sand to prevent flushing. No more modificationswere made to the filter designated as the biosand filter. For thearsenic biosand filters, a shallow plastic diffuser basinwas preparedby drilling 3e5mmholes with even spacing in the bottom. This wasloaded with 5 kg of iron nails, each approximately 2 cm long, andplaced in the top of the bucket. The nails were entirely coveredwithlarge stones (3e6 cm in diameter) to limit displacement of nailsduring filter use. A fourth filter, named the NIS filter, was con-structed using the same method as the biosand filter, with oneexception: a layer (1e2 cm thick) of 5 kg of nails was embedded3 cm below the top surface of the sand. The layers used in the NISfilter are shown in Fig. 1.

Five kilograms of nails were used so that comparison could bemade between this study and previous studies, namely Ngai et al.(2007) and Chiew et al. (2009), which assessed the performanceof the arsenic biosand filter containing 5 kg of nails. Within thecurrent study, all filters were built using the same weight of iron sothat the active contents of the filter would be comparable. Thereactivity of the nails was not measured using laboratory analysis,but basic observations practical in a field situation were used todetermine that the nails were relatively reactive. Firstly, the nailswere very thin, meaning high surface area for reaction. Secondly,when the nails were soaked before being added to the filter, thewater quickly became reddish-brown, indicating rapid rusting. Thenails also rapidly showed evidence of rusting within the filter in theform of reddish-brown streaks on the filter wall. Thirdly, the nailswere the cheapest available for the size, suggesting anti-corrosionmeasures taken during the manufacturing process would havebeen minimal.

A SONO-style filter was built on the basis of instructions pro-vided in Hussam and Munir (2007), Hussam (2010) and Neumannet al. (2013). The filter was constructed using two buckets ofheight 47 cm and average diameter 38 cm. Small holes were drilledaround the top rim of each bucket at even spacing. The bucketswere placed on a stand constructed to have two levels. Waterpoured into the top bucket flowed through a tap and pipe to thebottom bucket and exited from this bucket via another tap. Theinlet to each tap was covered with brick chips of 1e3 cm diameter.Three layers of media were then placed in the bottom bucket (frombottom to top): 9 kg of wet fine sand sieved using a 1e1.5 mmscreen on the bottom; 1 kg of small grain activated carbon; net with

Fig. 1. The NIS filter. The original filter consisted of one large bucket (on the right). A secondof filter media.

pore diameter 2 mm; a top layer of 10 kg of wet coarse sand sievedthrough a screen of 3e4 mm. Three layers of material were placedin the top bucket with each layer separated by a thin porous net:10 kg of wet coarse sand on the bottom; net with pore diameter2 mm; 5 kg of rusted iron filings; 2 mm net; 10 kg wet coarse sandon top. Schematic representations of the SONO filter can be foundin Neumann et al. (2013) and Hussam and Munir (2007) and aphoto of the filter used in this study is shown in Supporting In-formation (SI) Fig. SI1. The iron component of the filter was pre-pared using iron filings. Filings were mostly <20 mm, with themajority 1e10 mm. The filings were put through a process ofwashing, wetting and drying to promote rusting, with the aim ofcreating an active surface for adsorption and immobilization ofarsenic present in water (Hussam and Munir, 2007). The process isbased on Hussam (2010) and is detailed in SI Methods.

2.2. Site selection

Liangjiabu village in Pingyao county, Shanxi province, waschosen as the location of this field experiment. The village wasconsidered appropriate for the study because the main watersource is a groundwater well with arsenic concentration of over200 mg L�1. This is more than four times the national standard of50 mg L�1 and more than 20 times the WHO guideline of 10 mg L�1

(PRC, 2007; WHO, 2011). Other water quality parameters for thisgroundwater are given in Table SI1. This water is piped untreated tovillage households. The village was selected for the study aftersurveys of 20 households revealed aesthetic problems (e.g. debris,odor) affecting the well water meant households were receptive toa low-cost filter.

2.3. Filter operation

Each filter was built in a different house in the village andoperated by the household. Households were chosen on the basisthat they were reliable, willing participants well known to theauthors' main contact in the village. This enabled testing andfeedback on filters under actual field conditions. The biosand filter,arsenic biosand filters and NIS filter were constructed on March 15,2015, and were in use until at least late August 2015. The SONO-style filter was constructed two weeks later on March 29, 2015,and was still in use in late August.

smaller bucket (on the left) was added after Week 10. H ¼ height of layer; D ¼ diameter

Fig. 2. Arsenic levels in biosand filter (BSF) influent and effluent. The week number isthe number of weeks following construction. Low arsenic results for week 19 werefrom tests conducted at another testing centre.

K. Smith et al. / Chemosphere 168 (2017) 155e162158

Each household collected influent water from household tapsconnected to the centralized pipe network and stored it in openwater containers. Daily water diaries were kept by householdsbetween weeks ending April 25 and June 7. These are shown inTable SI2 and revealed that an average of 235 L of water was addedweekly to each filter. The household operating one of the arsenicbiosand filters was instructed to filter water twice before using. Allother households were instructed to filter water once.

Households with the biosand filter, arsenic biosand filters andNIS filter performed maintenance on the filters when flow becameslow enough to make the filter inconvenient to use. Maintenanceinvolved addingwater to the filter and gently agitating the top layerof sand. The water was then scooped out and households wereinstructed to dispose of it on the clay dirt roads outside theboundary of their yards. This procedure was performed one ormore times, depending on the flow rate. Regular maintenance ofSONO filters is generally not necessary for low-iron groundwater(Hussam and Munir, 2007).

2.4. Sample collection and testing

Influent and effluent samples for arsenic testing were initiallycollected once aweek from all filters. The first sample was collectedwithin an hour of construction for the SONO-style filter and twoweeks after construction for all other filters. From week 10, sam-pling was conducted fortnightly or monthly. Standing water sam-ples were also collected from the arsenic biosand filters on 10occasions. Standing water refers to water that has passed throughthe nails but not the sand. Samples were collected in clean plastic15 mL bottles. Influent samples were collected from the watercontainer used to store water from the tap linked to the village'scentralized water supply system. Effluent samples were collectedfrom the container used to catch filtered water while the filter wasin use. Some effluent samples collected earlier in the experimentwere collected directly from the filter outlet. Each bottle waswashed thoroughly using the water to be sampled before beingfilled with this water.

All effluent samples were analysed for total arsenic (As). InfluentAs for all filters was 226e240 mg L�1 in the first week. This indicatedall households were supplied by the same well. From then on, fiveinfluent samples were collected at each sampling, but only twowere tested for As. The average was used to represent all houses.Other influent samples were stored in a covered, dry place. Thesewere tested at a later date to confirm As concentration. All totalarsenic testing was done using inductively coupled plasma massspectrometry (ICP-MS) with a detection limit of 0.01 mg L�1. Detailson ICP-MS method and accuracy testing and testing of other waterquality parameters is described in SI Methods.

3. Results and discussion

3.1. Arsenic removal by the biosand filter

In the literature on household sand filters, arsenic removal ef-ficiency of the biosand filter can be unclear. Studies on another typeof iron-free sand filter showed that As removal can be excellent forgroundwater with high Fe (Berg et al., 2006; Nitzsche et al., 2015),but was below 70% for Fe < 3.7 mg L�1 (Berg et al., 2006). Fig. 2shows that the biosand filter reduced As but never to below thenational standard of 50 mg L�1. Maximum removal was just over50% but this decreased to 13% after five months of use.

Presence of Fe in groundwater can lead to arsenic removal whenthis dissolved Fe is oxidized through exposure to atmospheric ox-ygen to form Fe(III) (hydr)oxides (Roberts et al., 2004; Leupin et al.,2005; Berg et al., 2006). When this precipitates, co-precipitation

with As can occur (Roberts et al., 2004; Berg et al., 2006). Testingof filter influent samples for iron (Fe) revealed low to moderateconcentrations (<0.2 mg L�1) on average. Fe(II) is oxidized morequickly at pH 8 than pH 7 (Leupin et al., 2005) and average pH ofinfluent water was 8.0 ± 0.1 when tested on March 29 and similarwhen tested two months later. This suggests that pH conditionswere suitable for the oxidation of small amounts of soluble Fenaturally present in village well water, which led to co-precipitation with As and removal by the biosand filter. Giventhat Fe concentrationwas low, this process was insufficient to meetthe standard. Total As concentration in water filtered by blank (i.e.no iron) sand filters tends to increase with total volume of waterfiltered (Hussam and Munir, 2007), which explains the downwardtrend in removal rate for the biosand filter. Small amounts of ironoxide present in the sand used for filter construction may also havecontributed to arsenic removal by the biosand filter.

3.2. Arsenic removal by the arsenic biosand filter

The arsenic biosand filter was originally found to have anaverage arsenic removal rate of 88e95% according to a field study ofaround 1000 filters by Ngai et al. (2007) The maximum removalrate of the two arsenic biosand filters tested in the current studywas 81%. This declined to around 50% by the end of the five-monthstudy, excluding extreme values. Fig. 3 shows that effluent As wasbelow 50 mg L�1 on only one occasion. This concurs with findings byChiew et al. (2009) and Singh et al. (2014) from studies on arsenicbiosand filters in Cambodia and Nepal using real groundwater.Singh et al. (2014) found that only 54% of filters used by householdsfor over six months were able to reduce influent As of�50 mg L�1 to<50 mg L�1. Our study found that insufficient As removal was also aproblem for arsenic biosand filters used by households for fivemonths or less when iron nails were still fresh. Comparison byChiew et al. (2009) of three arsenic biosand filters found thataverage As removal was between 39% and 75%. These filters werefed groundwaters with an average As �146 mg L�1 and phosphorus(P) �0.91 mg L�1 (Chiew et al., 2009). Presence of phosphorus(assumed to be in the form of phosphate) in groundwater can havea negative effect on filter performance because phosphorus com-petes with arsenic for adsorption sites on iron oxides (Dixit andHering, 2003; Tyrovola et al., 2006). Influent water used in thecurrent study had lower average P concentration (0.45mg L�1) thanthat used by Chiew et al. (2009). This should positively influence Asremoval but removal rates were still mostly �76%, as shown inFig. 3.

Both arsenic biosand filters had lower effluent As and higher

Fig. 3. Arsenic levels in influent and effluent from the arsenic biosand filter with waterfiltered once through (ABF1) and the arsenic biosand filter with water filtered twicethrough (ABF2).

K. Smith et al. / Chemosphere 168 (2017) 155e162 159

removal rate than the biosand filter. This indicated the addition of abasin of nails indeed assisted As removal. Exposure of iron nails inthe arsenic biosand filter to both water and oxygen (both in the airand dissolved oxygen in the water) causes rusting (Chiew et al.,2009). This corrosion is a source of Fe(II), which is oxidizedthrough aeration to form Fe(III) (hydr)oxides that precipitate withabsorbed arsenic (Roberts et al., 2004). Chiew et al. (2009) testedseveral arsenic biosand filter standing water samples and found Asconcentration to be equal to influent As, leading to the conclusionthat nails in the filter had minimal effect on As removal perfor-mance. Chiew et al. (2009) suggested limited residence in thediffuser provided little time for iron in the nails to oxidize to solubleFe(II) or insoluble iron oxides, meaning As removal by the filter waslikely mostly the result of dissolved iron naturally present ingroundwater. Results of the current study agree that contact timebetween water and nails in the arsenic biosand filter and filterconditions are not enough to lead to sufficient As removal. How-ever, the performance of both arsenic biosand filters was superiorto the nail-free biosand filter when using the same influent water.This suggests that the nails increased formation of iron oxide-arsenic precipitate that could then be trapped in filter media.Standing water arsenic levels for the arsenic biosand filter (waterfiltered once) generally supported the conclusion that the presenceof nails improves arsenic removal efficiency. Arsenic in standingwater was lower than influent As for this filter in 8 out of 10 cases,with reduction in As between 12 and 60%.

Fig. 3 shows an overall decrease in removal efficiency over timein the arsenic biosand filters, which supports findings by Singhet al. (2014) Previous testing of arsenic biosand filters by the au-thors of the current study support this finding. A batch of arsenic

biosand filters were constructed in Shanxi province in July 2014 forhouseholds using well water with As >10 mg L�1. Eight filterstested just after construction using a Palintest Arsenator had Asremoval rates of between 63% and 95%. Influent As was11e67 mg L�1 and all effluent As was below 10 mg L�1 (see Fig. SI2).A further eight filters from the same area were tested using aPalintest Arsenator in July 2015. These filters were in use for atleast 10 months, with most built in the same batch as filters testedin July 2014. The maximum removal rate was 56% for influent As of11e97 mg L�1 and only 4 of the 8 filters were capable of removingany arsenic (shown in Fig. SI3). For the four other filters, effluentAs was comparable or slightly higher than influent As. These re-sults suggested the average removal efficiency of the arsenic bio-sand filter was high shortly after construction but decreasedsignificantly within a year.

3.3. Effect of attempts to increase contact time in arsenic biosandfilter

In light of unsatisfactory performance by the arsenic biosandfilter, Chiew et al. (2009) suggested the filter should be modified ina way that increased contact time between the water and the nails.In the current study, one of the two households using an arsenicbiosand filter was instructed to pass each bucket of water throughthe filter twice before drinking. The aim in filtering the same watertwice was to increase intermittent exposure of the nails to air topromote rusting and double the contact time between the nails andwater to promote As removal. Testing results and feedback from thehousehold suggested this method of operating the arsenic biosandfilter was unable to ensure sufficient As removal andmade the filterless easy to use. Effluent samples taken from this filter (ABF2 inFig. 3) had higher arsenic levels than those taken from the otherarsenic biosand filter (ABF1 in Fig. 3), except on one occasion.However, comparison between standing water and effluent for thisfilter (ABF2) indicated that water could contain lower arsenic whenfiltered through twice. In four cases, standing water As was lowerthan effluent As. This led to the conclusion that standing watercontained water that had already passed through the filter once. Intwo of these cases, the As in the standing water was just below thestandard of 50 mg L�1. At other times, standing water As was higherthan effluent As. This inconsistency showed that it was difficult todetermine whether water exiting the filter had been filtered onceor twice when taking samples. It was also difficult for the house-hold to determine whether water had been filtered twice, whichmeant the likelihood of drinking water filtered only once was high.Furthermore, the flow rate for this filter decreased significantlyevery fewweeks, whichmeant the household had to clean the filtermore regularly. This was likely due to greater build up of rustcaused by increased filtration of water. When filter flow decreasesrapidly and cleaning becomes more frequent, this makes the filterless convenient for the user, which can affect uptake.

In an attempt to improve the performance of the arsenic biosandfilter throughwhich water was filtered twice, the diffuser basinwasreplaced at week 10 by a basin with half the number of evenlyspaced holes. The aim in reducing the number of holes was to slowthe speed at which water passed through the basin, thus increasingcontact time with the nails. This did not improve removal rate andeffluent As exceeded 100 mg L�1 on two occasions after the changein basin.

3.4. Arsenic removal by SONO-style filter

Arsenic was below the standard in two effluent samples takenfrom the SONO-style filter within a week of construction but abovethe standard in all others. Influent and effluent samples taken

K. Smith et al. / Chemosphere 168 (2017) 155e162160

within an hour of constructing the SONO-style filter showed areduction in As of 97% to 8 mg L�1 from over 200 mg L�1. After thefirst two tests, removal rate for the SONO-style filter varied be-tween 45% and 77%, as shown in Fig. 4. Effluent As was sometimeshigher for the SONO-style filter than for the arsenic biosand filters.

SONO filter technology has previously been tested thoroughly infield and laboratory studies using various influent arsenic levels(Hussam and Munir, 2007; Neumann et al., 2013). It passed theEnvironmental Technology Verification Arsenic Mitigation(ETVAM) program jointly developed by the Canadian andBangladesh governments (Hussam and Munir, 2007; OCETA, n.d.).and is known to have a number of benefits, including a long life-span, safe disposal, limited maintenance and effectiveness atremoving a number of contaminants including arsenic (Hussam,2010; Neumann et al., 2013). The initial effectiveness of theSONO-style filter used in the current study suggests that thepreparation of the iron filings was successful. It is likely that themain reason for the drop in performance of the SONO-style filterover the course of the study was the flow rate at which it wasoperated.

The recommended flow rate for the SONO filter is 20e30 L h�1

(Hussam and Munir, 2007). When tested for flow rates of40e60 L h�1, effluent As concentration remains below 50 mg L�1 butshows a distinct increasing trend as flow rate increases (Hussamand Munir, 2007). Higher flow through the top bucket of theSONO-style filter limits contact time between the water and theiron filings, which can decrease As removal. The top bucket of theSONO-style filter used in the current study was fitted with a metaltap bought from a local hardware store, as suggested by Hussamand Munir (2007) When fully open, this tap had a maximumflow rate of around 140 L h�1, which was well over the recom-mended rate. The household using the filter tended to fully openthis tap when filtering water and otherwise kept it closed. Efforts tosolve this problem part way through the study were only moder-ately successful. On one occasion, the tap was adjusted to stay openat the correct flow but later moved during use. It was later re-adjusted to the correct flow rate in week 17 and was still in placea month later, at which stage testing showed an increase in Asremoval rate from 65% (week 17) to 75% (week 21). By week 27,however, effluent arsenic had increased to around 100 mg L�1. Thisindicates that filter operation patterns should be carefully consid-ered when choosing the SONO-style filter, as this can have a sig-nificant effect on As removal performance. It is suggested thatusing a valve would allow users freedom to turn taps on and offwhile permanently ensuring flow through the tap does not exceed20e30 L h�1.

Fig. 4. Arsenic levels in influent and effluent for the SONO-style filter.

3.5. Arsenic removal by a biosand filter with nails embedded in thesand

The NIS filter was constructed in the same way as the biosandfilter except that 5 kg of nails were added 3 cm below the uppersurface of the sand. Results from arsenic testing over 29 weeks ofuse are shown in Fig. 5. A selection of results for other influent andeffluent water quality parameters for this filter are provided inTable SI3. Arsenic in effluent from the NIS filter was always belowthe standard of 50 mg L�1 and generally below 30 mg L�1. The arsenicremoval rate was between 86% and 95% and averaged 92%. Thearsenic removal performance of the NIS filter exceeded all otherfilters in this study. The consistency and effectiveness of the NISfilter when compared to the arsenic biosand filters was likely due tothe placement of the nails in the sand. Chiew et al. (2009) foundthat flow through the arsenic biosand filter diffuser basinwasmuchfaster than flow through the sand. The actual residence time ofwater in the diffuser was much less than the predicted residencetime calculated based on average (overall) flow rates for the entiresystem (Chiew et al., 2009). Thus, if nails are placed in the sand,water flows past them at a much slower rate, leaving more time foriron from the nails to dissolve and form iron oxides and for arsenicto be adsorbed to the iron oxides and precipitate from the water.Continued exposure of the nails to awet or damp environment mayalso have reduced the problem of encrustation of iron oxides on thenails, as mentioned by Chiew et al. (2009). Nails used in the normalarsenic biosand filter can become coated with iron oxides within amonth of construction due to exposure to air and this decreases theability of the nails to oxidize more and adsorb arsenic (Chiew et al.,2009). In the NIS filter, the nails were submerged in the top layer ofsand. It may be that this delayed the passivation of the nails byslowing the formation of oxide scale and allowing sustainedcorrosion (Bostick et al., 2009; Noubactep, 2014). The diameter ofthe NIS bucket was larger than the diameter of the diffuser basin,which meant nails were also less likely to cake together.

Nails in the NIS filter were less exposed to open air because theywere placed under sand and standing water but their exposure todissolved oxygen in the water was sufficient to cause rusting. Fluffyreddish-brown particles were found to exit the filter in the effluent.A sample containing large numbers of these particles was collectedfrom the bottom of a bucket of NIS effluent. These particles weredissolved into the water sample and the sample was tested for Feusing inductively coupled plasma atomic emission spectroscopy(ICP-AES). Iron concentration was 13.34 mg L�1, much higher thanthe Fe of <0.2 mg L�1 generally found in the influent groundwater.Thus, it was deduced these particles were rust from the nails so asecond smaller filter bucket with no nails was built in front of theoriginal larger filter bucket. This second bucket was 45 cm � 28 cmx 28 cm. The dimensions of layers andmedia are given in Fig. 1. Thisarrangement proved effective at capturing the rust particles fromwater that flowed out of the big bucket and into the small bucket.

Fig. 5. Arsenic levels in NIS filter influent and effluent.

K. Smith et al. / Chemosphere 168 (2017) 155e162 161

Starting week 10, effluent samples were collected from the smallerbucket. Fig. SI4 shows the complete NIS filter used in this study.

3.6. Reducing the size and total cost of the NIS filter

The addition of the second bucket increased the cost of the NISfilter and the time required to wash sand and stones, potentiallymaking it a less attractive option for households. To overcome thisissue, we decided to test the effect of reducing the size of the firstbucket. The aim was to see whether a smaller filter constructedaccording to the same method (i.e. same weight of iron nails placeda few centimetres below the top layer of sand) would also becapable of reducing As to below the standard. This set of two smallbuckets e one with sand, stones and nails and the other just withsand and stones e can be referred to as the mini-NIS filter. Filterdimensions are given in SI Methods and the filter is shown inFig. SI5. Influent and effluent As are shown in Fig. SI6. The filter wasbuilt 10 weeks after the original NIS filter and operated usingdifferent influent water than the other five filters in the study. Thiswater originated entirely or partially from the village's secondarywell and had arsenic levels that varied between 62 and 154 mg L�1.Six tests of effluent water during the 22 weeks after construction ofthis filter showed that effluent As never exceeded the standard of50 mg L�1. The removal rate varied between 68% and 97%, partly dueto the differing influent arsenic levels.

The main problem affecting the mini-NIS filter was severeclogging of the filter due to caking (i.e. cementation) of the nails inthe sand. The effect of this clogging on filter flow rate meant usershad to perform more regular filter maintenance that involvedremoving the top layer of sand and splitting up the nails. Cemen-tation of iron in iron-sand filters due to corrosion products has beenreported by Westerhoff and James (2003). This is known todecrease filter permeability and lead to the need for increasinglyhigh influent water pressure. Noubactep and Care (2010) andNoubactep et al. (2010) suggest mixing filter sand with iron tosustain permeability and extend filter life. For filters using ironnails, spreading nails through a layer of sand could cause water tobypass active material and compromise effective contaminantremoval (Bradley et al., 2011). Reducing the weight of iron nailsused (i.e. <5 kg) is another possibility but it is important to haveenough nails so that active iron is not depleted too quickly, as thiswould make the filter unable to effectively remove arsenic(Noubactep et al., 2010). The original full-size NIS filter in thisexperiment did not have clogging problems over six months of use,likely because 5 kg of nails formed a much thinner iron layer in thisfilter. It may be possible to modify the mini-NIS filter so that thethickness of the iron nails is reducedwithout compromising arsenicremoval. Reducing filter clogging andmonitoring long-term arsenicremoval of this filter should be the primary focus of furtherresearch.

4. Conclusions

The biosand filter with embedded nails (otherwise known as theNIS filter) had the best performance of the four types of filterstested under field conditions using the same influent.

� The NIS filter consistently removed arsenic from village wellwater (>200 mg L�1) to below the standard of 50 mg L�1 over asix-month period, with an average removal rate of 92%.

� The modification of the NIS filter to include a smaller bucket aswell as a larger bucket prevented rust particles from exiting inthe effluent. The complete filter can be entirely constructed forless than $40e50 using materials purchased locally in Pingyaocounty, Shanxi.

� The presence of iron in the arsenic biosand filter and SONO-stylefilter meant these filters were more effective than the biosandfilter.

� The NIS filter was the only filter consistently capable of meetingthe As standard, likely due to the increased contact betweeninfluent water and nails embedded in the filter sand.

� High flow rate likely compromised As removal by the SONO-style filter.

Acknowledgements

The authors would like to acknowledge financial support andsupport in kind provided by Shanxi Youth Federation, Palintest,Program for Changjiang Scholars and Innovative ResearchTeams in Universities (IRT 1261) and Major Water Program(2014ZX07406003), as well as guidance from the Shanxi Depart-ment of Environmental Protection. This study benefited from theassistance of Yongfu Liang from Liangjiabu village and students ofTaiyuan University of Technology Science and Technology Associ-ation and RISE who helped with filter construction, sampling andobtaining results. We are also grateful to Xi Chen who helped withlaboratory testing, to previous RISE presidents who helped estab-lish the project and to academics working in the area of low-costarsenic removal who shared their knowledge with the authors.

Appendix A. Supplementary data

Supplementary data related to this article can be found at http://dx.doi.org/10.1016/j.chemosphere.2016.10.021.

References

Berg, M., Luzi, S., Trang, P.T.K., Viet, P.H., Giger, W., Stuben, D., 2006. Arsenic removalfrom groundwater by household sand filters: comparative field study, modelcalculations, and health benefits. Environ. Sci. Technol. 40, 5567e5573.

Bostick, B.C., Chiew, H., Sampson, M.L., 2009. Response to comment on “Effect ofgroundwater iron and phosphate on the efficacy of arsenic removal by iron-amended BioSand Filters”. Environ. Sci. Technol. 43, 8696e8697.

Bradley, I., Straub, A., Maraccini, P., Markazi, S., Nguyen, T.H., 2011. Iron oxideamended biosand filters for virus removal. Water Res. 45, 4501e4510.

Chaudhari, S., Banerji, T., Kumar, P.R., 2014. 7-Domestic- and Community-scaleArsenic Removal Technologies Suitable for Developing Countries A2-ahuja,Satinder. Water Reclamation and Sustainability. Elsevier, Boston, pp. 155e182.

Chiew, H., Sampson, M.L., Huch, S., Ken, S., Bostick, B.C., 2009. Effect of groundwateriron and phosphate on the efficacy of arsenic removal by iron-amended Bio-Sand filters. Environ. Sci. Technol. 43, 6295e6300.

Dixit, S., Hering, J.G., 2003. Comparison of arsenic(V) and arsenic(III) sorption ontoiron oxide minerals: implications for arsenic mobility. Environ. Sci. Technol. 37,4182e4189.

Ghauch, A., 2015. Iron-based Metallic Systems: an Excellent Choice for SustainableWater Treatment. TU Bergakad.

Gottinger, A.M., McMartin, D.W., Wild, D.J., Moldovan, B., 2013. Integration of zerovalent iron sand beds into biological treatment systems for uranium removalfrom drinking water wells in rural Canada. Can. J. Civ. Eng. 40, 945e950.

Hussam, A., 2010. Iron composition based water filtration system for the removal ofchemical species containing arsenic and other metal cations and anions. in:Publication, U.S.P.A. (Ed.). George Mason Intellectual Properties, United States.

Hussam, A., Munir, A.K.M., 2007. A simple and effective arsenic filter based oncomposite iron matrix: development and deployment studies for groundwaterof Bangladesh. J. Environ. Sci. Heal A 42, 1869e1878.

Kapaj, S., Peterson, H., Liber, K., Bhattacharya, P., 2006. Human health effects fromchronic arsenic poisoning- A review. J. Environ. Sci. Heal A 41, 2399e2428.

Kowalski, K., 2014. New Method for Arsenic Compounds Elimination from NaturallyContaminated Drinking Water Systems. Faculty of Engineering and Science.Aalborg University, Denmark.

Kowalski, K.P., Søgaard, E.G., 2014. Implementation of zero-valent iron (ZVI) intodrinking water supply e role of the ZVI and biological processes. Chemosphere117, 108e114.

Leupin, O.X., Hug, S.J., 2005. Oxidation and removal of arsenic (III) from aeratedgroundwater by filtration through sand and zerovalent iron. Water Res. 39,1729e1740.

Leupin, O.X., Hug, S.J., Badruzzaman, A.B.M., 2005. Arsenic removal fromBangladesh tube well water with filter columns containing zerovalent ironfilings and sand. Environ. Sci. Technol. 39, 8032e8037.

Neumann, A., Kaegi, R., Voegelin, A., Hussam, A., Munir, A.K.M., Hug, S.J., 2013.

K. Smith et al. / Chemosphere 168 (2017) 155e162162

Arsenic removal with composite iron matrix filters in Bangladesh: a field andlaboratory study. Environ. Sci. Technol. 47, 4544e4554.

Ngai, T.K.K., Shrestha, R.R., Dangol, B., Maharjan, M., Murcott, S.E., 2007. Design forsustainable development - household drinking water filter for arsenic andpathogen treatment in Nepal. J. Environ. Sci. Heal A 42, 1879e1888.

Nitzsche, K.S., Lan, V.M., Trang, P.T.K., Viet, P.H., Berg, M., Voegelin, A., Planer-Friedrich, B., Zahoransky, J., Muller, S.K., Byrne, J.M., Schroder, C., Behrens, S.,Kappler, A., 2015. Arsenic removal from drinking water by a household sandfilter in Vietnam - effect of filter usage practices on arsenic removal efficiencyand microbiological water quality. Sci. Total Environ. 502, 526e536.

Noubactep, C., 2009. Comment on “Effect of groundwater iron and phosphate onthe efficacy of arsenic removal by iron-amended BioSand Filters”. Environ. Sci.Technol. 43, 8698.

Noubactep, C., 2014. Flaws in the design of Fe(0)-based filtration systems? Che-mosphere 117, 104e107.

Noubactep, C., 2015. Metallic iron for environmental remediation: a review of re-views. Water Res. 85, 114e123.

Noubactep, C., Care, S., 2010. Enhancing sustainability of household water filters bymixing metallic iron with porous materials. Chem. Eng. J. 162, 635e642.

Noubactep, C., Care, S., Togue-Kamga, F., Schoner, A., Woafo, P., 2010. Extendingservice life of household water filters by mixing metallic iron with sand. Clean-Soil Air Water 38, 951e959.

Noubactep, C., Schoner, A., Woafo, P., 2009. Metallic iron filters for universal accessto safe drinking water. Clean-Soil Air Water 37, 930e937.

Noubactep, C., Temgoua, E., Rahman, M.A., 2012. Designing iron-amended biosandfilters for decentralized safe drinking water provision. Clean-Soil Air Water 40,798e807.

OCETA, n.d. Assessment of Five Technologies for Mitigating Arsenic in BangladeshWell Water. Environmental Technology Verification-Arsenic Mitigation. OntarioCentre for Environmental Technology Advancement.

PRC, 2007. Standards for Drinking Water Quality GB5749-2006 (People's Republic ofChina, Beijing).

Rahman, M.A., Karmakar, S., Salama, H., Gactha-Bandjun, N., Btatkeu, K.B.D.,

Noubactep, C., 2013. Optimising the design of Fe0-based filtration systems forwater treatment: the suitability of porous iron composites. J. Appl. Solut. Chem.Model. 2, 165e177.

Roberts, L.C., Hug, S.J., Ruettimann, T., Billah, M., Khan, A.W., Rahman, M.T., 2004.Arsenic removal with iron(II) and iron(III) in waters with high silicate andphosphate concentrations. Environ. Sci. Technol. 38, 307e315.

Rodriguez-Lado, L., Sun, G.F., Berg, M., Zhang, Q., Xue, H.B., Zheng, Q.M.,Johnson, C.A., 2013. Groundwater arsenic contamination throughout China.Science 341, 866e868.

Singh, A., Smith, L.S., Shrestha, S., Maden, N., 2014. Efficacy of arsenic filtration byKanchan Arsenic Filter in Nepal. J. Water Health 12, 596e599.

Tepong-Tsinde, R., Crane, R., Noubactep, C., Nassi, A., Ruppert, H., 2015. Testingmetallic iron filtration systems for decentralized water treatment at pilot scale.Water-Sui 7, 868e897.

Tyrovola, K., Nikolaidis, N.P., Veranis, N., Kallithrakas-Kontos, N., Koulouridakis, P.E.,2006. Arsenic removal from geothermal waters with zero-valent iron - effect oftemperature, phosphate and nitrate. Water Res. 40, 2375e2386.

Wenk, C.B., Kaegi, R., Hug, S.J., 2014. Factors affecting arsenic and uranium removalwith zero-valent iron: laboratory tests with Kanchan-type iron nail filter col-umns with different groundwaters. Environ. Chem. 11, 547e557.

Westerhoff, P., James, J., 2003. Nitrate removal in zero-valent iron packed columns.Water Res. 37, 1818e1830.

WHO, 2011. Guidelines for Drinking-water Quality, fourth ed. World Health Orga-nisation, Geneva.

Xinhua, 2013. China pledges Safe Rural Drinking Water by 2015. China Daily, Beijing.Yu, G.Q., Sun, D.J., Zheng, Y., 2007. Health effects of exposure to natural arsenic in

groundwater and coal in China: an overview of occurrence. Environ. HealthPerspect. 115, 636e642.

Zhang, Q., Rodriguez-Lado, L., Liu, J., Johnson, C.A., Zheng, Q.M., Sun, G.F., 2013.Coupling predicted model of arsenic in groundwater with endemic arsenismoccurrence in Shanxi Province, Northern China. J. Hazard Mater. 262,1147e1153.