(NOM) removal from natural waters in Tanzania by

Science of the Total Environment 527–528 (2015) 520–529

Contents lists available at ScienceDirect

Science of the Total Environment journal homepage: www.elsevier.com/locate/scitotenv

Factors affecting fluoride and natural organic matter (NOM) removal from natural waters in Tanzania by nanofiltration/reverse osmosis Junjie Shen a,b, Andrea I. Schäfer c,⁎ a b c

School of Engineering and Physical Sciences, Heriot-Watt University, Edinburgh EH14 4AS, United Kingdom The Nelson Mandela African Institute of Science and Technology, P.O. Box 447, Tengeru, Arusha, Tanzania Institute of Functional Interfaces (IFG), Karlsruhe Institute of Technology (KIT), Hermann-von-Helmholtz-Platz 1, 76344 Eggenstein-Leopoldshafen, Germany

H I G H L I G H T S • Natural water sources sampled from northern Tanzania indicated high fluoride, inorganic carbon and NOM content. • Effects of operating conditions and water compositions on F and NOM removal by NF/RO were examined for many natural waters. • A positive effect of NOM on F retention is reported for these natural waters.

a r t i c l e

i n f o

Article history: Received 10 February 2015 Received in revised form 11 April 2015 Accepted 11 April 2015 Available online 23 May 2015 Editor: D. Barcelo Keywords: Fluoride NOM Nanofiltration Reverse osmosis Natural water

a b s t r a c t This study examined the feasibility of nanofiltration (NF) and reverse osmosis (RO) in treating challenging natural tropical waters containing high fluoride and natural organic matter (NOM). A total of 166 water samples were collected from 120 sources within northern Tanzania over a period of 16 months. Chemical analysis showed that 81% of the samples have fluoride levels exceeding the WHO drinking guideline of 1.5 mg/L. The highest fluoride levels were detected in waters characterized by high ionic strength, high inorganic carbon and on some occasions high total organic carbon (TOC) concentrations. Bench-scale experiments with 22 representative waters (selected based on fluoride concentration, salinity, origin and in some instances organic matter) and 6 NF/RO membranes revealed that ionic strength and recovery affected fluoride retention and permeate flux. This is predominantly due to osmotic pressure and hence the variation of diffusion/convection contributes to fluoride transport. Different membranes had distinct fluoride removal capacities, showing different raw water concentration treatability limits regarding the WHO guideline compliance. BW30, BW30-LE and NF90 membranes had a feed concentration limit of 30–40 mg/L at 50% recovery. NOM retention was independent of water matrices but is governed predominantly by size exclusion. NOM was observed to have a positive impact on fluoride removal. Several mechanisms could contribute but further studies are required before a conclusion could be drawn. In summary, NF/RO membranes were proved to remove both fluoride and NOM reliably even from the most challenging Tanzanian waters, increasing the available drinking water sources. © 2015 Elsevier B.V. All rights reserved.

1. Introduction Globally, high fluoride concentration in drinking water has been found to cause severe health risks to humans (Bhatnagar et al., 2011; Fawell et al., 2006). The World Health Organization (WHO) recommends a guideline value of fluoride to be 1.5 mg/L at which the harmful effect should be minimal. However, drinking water with fluoride concentrations above the guideline value is consumed by more than 200 million people in over 20 developed and developing countries (Amini ⁎ Corresponding author at: The Nelson Mandela African Institute of Science and Technology, P.O. Box 447, Tengeru, Arusha, Tanzania. E-mail address: [email protected] (A.I. Schäfer).

http://dx.doi.org/10.1016/j.scitotenv.2015.04.037 0048-9697/© 2015 Elsevier B.V. All rights reserved.

et al., 2008; Fawell et al., 2006). In the East African Rift Valley which is a naturally high fluoride zone, over 80 million people exhibit varying degrees of fluorosis symptoms (Frencken, 1990; Smedley et al., 2002). Tanzania, located in east Africa, is projected to face serious water stress (defined as average water resources below 1500 m3 per capita per year) by 2025 due to population growth (World Bank, 2006). The increased fluoride guideline value of 4 mg/L in Tanzania reflects the difficult situation of fluoride contamination that is worsened by water scarcity (Tanzania Bureau of Standards, 2008). Nanofiltration (NF) and reverse osmosis (RO) are very promising defluoridation methods due to their high fluoride retention compared to conventional methods such as adsorption and precipitation (Ayoob et al., 2008). In NF/RO processes, fluoride transport through the

J. Shen, A.I. Schäfer / Science of the Total Environment 527–528 (2015) 520–529

membrane consists of three mechanisms: diffusion, convention, and electromigration (Bowen and Mohammad, 1998; Bowen and Mukhtar, 1996; Mohammad, 2002). Fluoride retention by NF/RO is generally governed by steric effects (size exclusion) and electrical effects (charge repulsion, Donnan exclusion and dielectric exclusion) (Shen and Schäfer, 2014). A variety of parameters, including feed compositions, membrane characteristics and operational conditions, lead conditions to be more conducive to the dominant effect of one mechanism over the other (Bejaoui et al., 2011; Mnif et al., 2010; Richards et al., 2010; Shen and Schäfer, 2014). Natural organic matter (NOM) is another important contaminant in drinking water treatment as it leads to undesirable color, taste and odor, and reacts with common disinfectants to produce a variety of toxic disinfection by-products including trihalomethanes (THM) (Sadiq and Rodriguez, 2004). Large NOM compounds like polysaccharides and humic substances can be effectively retained by NF/RO, during which size exclusion, charge repulsion, and hydrophobic interactions are the well-known removal mechanisms (Jarusutthirak et al., 2007; Schäfer et al., 2004). But low molecular weight (LMW) fractions are likely to permeate the membranes (especially the NF membranes), which produces the problem of bacterial regrowth in the drinking water distribution system (Meylan et al., 2007). Besides, NOM may cause severe membrane fouling and hence reduce system productivity and lifetime (Hong and Elimelech, 1997). Both reversible and irreversible fouling mechanisms have been understood through numerous studies (Schäfer et al., 1998; Zularisam et al., 2006). NOM characterization, solution chemistry, membrane properties and operational conditions have been reported to influence the fouling mechanisms to different extents (Tang et al., 2007). The simultaneous occurrence of fluoride and NOM in natural tropical waters, as those commonly found in Tanzania, sets a significant challenge for water engineers. NOM has been found to compete with fluoride ions in adsorptive processes, particularly adsorption onto activated carbon (Newcombe et al., 1997) and bone char (Brunson and Sabatini, 2014). However, the importance of NOM in fluoride removal by NF/RO is not understood to date. The present study fills this gap by treating a number of natural water samples containing both high fluoride and NOM with a selection of NF/RO membranes. The effect of NOM and other parameters such as salinity on fluoride retention by NF/RO will be investigated systematically by (i) examining the chemical composition of high fluoride waters in Tanzania; (ii) determining the mechanisms by which fluoride and tropical NOM are removed by NF/RO, and (iii) exploring the impact of NOM on fluoride removal by NF/RO.

2. Materials and methods 2.1. Sampling regions The northern regions of Tanzania, as part of the major East African Rift system, have the most extensive areas of fluoride-rich waters and possibly the highest reported fluoride concentrations in the world. For this reason this area is an excellent choice for this study. In order to identify local water quality, and in consequence the most interesting waters for treatment, samples were collected from surface water (rivers, springs, freshwater lakes), groundwater (boreholes and wells), and soda lakes. Soda lakes are not suitable for drinking applications, but were studied as an important part of the local geology and the local water cycle. At some locations repeat samples were taken for studies of seasonal variation (results not included in this paper). The precise location of sites was determined in the field using a hand-held GPS device (Garmin GPSmap 60csx, USA).

521

The sampling sites are shown in Fig. 1. The sites were selected in a manner aimed to identify the variation of water quality over different water types with the overall aim to identify worst case water supply scenarios and subsequent treatment challenges. Samples were collected in 5 L or 12 L plastic containers (washed thoroughly with the sample water prior to collection) and stored at room temperature in the laboratory. 2.2. Chemical analysis The pH and electrical conductivity (EC) were measured in situ by a pH/conductivity meter (WTW Multi 340i, Germany). Fluoride ion (F−) was determined by an ion-selective electrode in conjunction with an Ag/AgCl reference electrode connected to a pH meter (826 pH Mobile Meter, Metrohm, UK). A total ionic strength adjustment buffer (TISAB) solution of pH 5–5.5 was used to reduce interferences resulting from pH and EC. The TISAB buffer consisted of analytical grade glacial acetic acid, 1,2-diaminocyclohexanetetraacetic acid (CDTA), sodium chloride and sodium hydroxide (Fisher Scientific, UK). Inorganic carbon (IC), in− cluding carbonate (CO2− 3 ) and bicarbonate (HCO3 ), was measured by a TOC analyzer (Sievers 900 Portable TOC Analyzer with Autosampler, GE Analytical Instruments, UK). NOM was measured as total organic carbon (TOC) by the same TOC analyzer but coupled with the inorganic carbon remover (ICR) unit. Turbidity was measured by a turbidity meter (TN100, Eutech, USA). UV absorbance was measured at 254 nm wavelength by a UV spectrophotometer (UV-2800, UNICO, USA), with Mt Meru bottled water used as the reference (blank). This water is local surface water treated by reverse osmosis, ozonation, UV treatment. It was selected due to its superior chemical quality compared to all other available waters (pH 7.2, EC b 18 μS/cm, TOC b 0.01 mgC/L, fluoride b 0.01 mg/L). 2.3. NF/RO system and filtration protocol A stainless steel stirred cell (described by Neale, 2009) was used for the bench-scale experiments (see schematic in Fig. 2). The volume of the cell is 990 mL and the internal exposed membrane diameter is 70 mm, giving a membrane surface area of 38.48 cm2. The cell contains a magnetic stirrer assembly (Millipore, UK) operated at 300 rpm using a magnetic stirrer plate (SM1, Stuart Instrument, UK). The cell was pressurized with compressed medical air (TOL, Tanzania) with a maximum pressure of 10 bar. Internal pressure and temperature were measured by a pressure transducer (PX219-30V85G5V, Omega Engineering, UK) and a thermocouple (TJ2-CPSS-M60U-250-SB, Omega Engineering, UK). Permeate mass was measured by an electronic balance (Adventurer Pro 2102, Ohaus, Switzerland). All the data were collected by a laptop with a data acquisition module (DAQ-54, Omega Engineering, UK). From a total of 166 water samples collected from 120 sites over 16 months, 22 samples were selected for NF/RO experiments in this study. The selection criteria included: (1) fluoride concentration above WHO guideline of 1.5 mg/L; (2) osmotic pressure lower than the applied pressure of 10 bar; (3) representative samples from surface water, groundwater and soda lakes and where applicable the organic matter content. Osmotic pressure was calculated by ionic concentrations using the van't Hoff formula (van't Hoff, 1887). Only ions with concentrations N 1 mg/L were taken account in the calculation. A summary of water quality of the selected samples is provided in Table 3. While all waters were analyzed with IC and ICP to determine the full ion balance, this data is not shown in this paper. Prior to each experiment, a new membrane was rinsed with clean water and then compacted for one hour at 10 bar. Pure water flux was subsequently determined at 10 bar for 30 min. For each batch experiment, the feed solution volume was 400 mL. Eight consecutive permeate samples (25 mL each for a total filtrate of 200 mL) were collected. After the experiment, pure water flux was measured again. Some key terms used in the experiment are defined as follows. Recovery is the ratio of permeate to feed volume. Maximum recovery in

522


Fig. 1. Map of northern Tanzania with sampling sites. Adapted from Tracks4Africa.

these bench-scale experiments is 50%, which is within the practical range of full-scale systems (30–90%) (DiGiano et al., 2000). Retention is the percentage of solute removed from the feed and to calculate retention in the stirred cell the cell concentration (CR) is required. CR is calculated as a function of permeate volume from mass balance, where:

CR ¼

X C FV F− CPi VPi VR

ð1Þ

CF and CPi are the initial feed and permeate concentration (mg/L), VF and VPi are the feed and permeate volume (mL), while VR is the retentate volume remaining in the cell (mL). Observed retention (R) is thus calculated as per Eq. (2) and was not corrected for real retention based on boundary layer concentration (Jonsson, 1986). R¼

C 1− P 100% CR

ð2Þ

2.4. Membrane characteristics Six commercial membranes were used in the study. BW30 and BW30-LE are two commonly used brackish water RO membranes while NF90, NF270, TFC-SR2 and TFC-SR3 are NF membranes. All these membranes are thin film composite (TFC) membranes with a polyamide-based active layer (Freger et al., 2002; Pontié et al., 2008; Tang et al., 2009). The characteristics of the membranes are summarized in Table 1. The molecular weight cut-off (MWCO) and pore radius are particularly important because they imply which solutes may be sterically retained by the membrane. As Table 1 shows, NaCl retention generally decreases with increasing MWCO or pore radius, indicating the importance of size exclusion. There is a trade-off between solute retention and water permeability. Membranes with larger MWCO/pore radius (e.g., NF270, TFC-SR2) have lower salt retention and higher water permeability, and vice versa. This is attributed to the convective transport of solutes that are carried with the water through the membrane.

Fig. 2. Schematic of stirred cell experimental apparatus.


523

Table 1 Characterization of the selected membranes, data adapted from Boussu et al. (2006), Richards (2012). Membrane type

MWCO (Da)

Pore radius (nm)

NaCl retentiona (%)

Water permeability (L/h·m2·bar)

Contact angle (°)

Zeta potentialb (mV)

BW30 BW30-LE NF90 NF270 TFC-SR2 TFC-SR3

98 98 100 180 460 167

0.32 NA 0.34 0.38 0.52 0.38

99.8 99.5 83.7 42.4 22.7 40.8

4.4 5.8 9.7 16.2 14.1 5.7

40.3 ± 1.1 NA 47.9 ± 1.7 29.1 ± 1.6 61.5 ± 2.6 48.5 ± 1.4

−12 −35c −26 −14 −16 −20

a b c

0.1 M NaCl, at 10 bar. at pH 8, in 20 mM NaCl and 1 mM NaHCO3. At pH 8, in 1 mM KCl.

Contact angle is regarded as the indicator for the hydrophilicity or hydrophobicity of a membrane. Hydrophilic membranes interact more with water and have a smaller contact angle. Hydrophilicity further contributes to high permeability due to the higher solubility of water in the membrane and diffusive transport of water. As Table 1 shows, the most hydrophilic membrane NF270 has indeed the highest pure water permeability. Membrane surface charge is another important parameter because it affects charge repulsion. The membrane zeta potential was calculated from the streaming potential which was measured using an electrokinetic analyzer (EKA, Anton Paar KG, Austria) (Richards, 2012). At pH 8 all the membranes are negatively charged (note the Tanzanian water samples are generally of more alkaline nature). Since the pH and ionic strength of real water samples are quite different from those at zeta potential measurement, the zeta potential values of different membranes should be considered relative to each other.

3. Results and discussion 3.1. Water quality In order to evaluate the water quality comprehensively, a classification scheme was proposed in the present work in Tanzanian context (Table 2). The scheme is based on the concept raised by Sargaonkar and Deshpande (2003). Parameters including turbidity, EC, fluoride, TOC and specific UV absorbance (SUVA) are important indicators for this classification. SUVA indicates the aromaticity and hydrophobicity of NOM and hence the potential to form THMs (Chowdhury and Champagne, 2008; Weishaar et al., 2003). The classification basis can be found in the Supporting information (SI). The overall water quality is characterized by alkaline pH and high ionic strength, which is associated with the local geology. Typically, alkaline lava and ash contain high contents of silicate minerals and glass. Natural water sources, particularly those containing dissolved CO2, react readily with alkaline silicate to release sodium and bicarbonate ions (Ghiglieri et al., 2012). As will be shown later, the alkaline environment as well as the calcium content exerts a positive influence over fluoride and NOM concentration. The major parameters of the 22 NF/RO treated samples are provided in Table 3. Fig. 3 summarizes the results and shows that fluoride and TOC are the parameters responsible for most of the heavy contamination. In the subsequent section each water quality parameter is analyzed.

In terms of fluoride concentration, the majority of samples (81%) contain fluoride exceeding the WHO limit (1.5 mg/L), whereas 70% are above the Tanzanian limit (4 mg/L) (Fig. 3). Soda lakes have much higher fluoride concentrations than surface water and groundwater, with the highest occurring in Lake Big Momella (1391 mg/L). Fluoride up to 88 mg/L was observed in a surface river (Pagasi River) which is flowing into Lake Natron. As for groundwater, the highest fluoride concentration (62 mg/L) was found at a borehole near Lake Manyara. Fluorides are naturally released into water by the dissolution of fluoride-containing rocks and soils (Edmunds and Smedley, 2005). The dissolution process is affected by various factors including rock composition, groundwater age, residence time, and pathway conditions (Kim and Jeong, 2005). The volcanic rocks of East Africa are richer in fluoride than analogous rocks in other parts of the world (Gizaw, 1996). Fluoride concentration in water is heavily controlled by the solubility of minerals, especially calcium fluorite (CaF2) which has the lowest solubility (15 mg/L at 18 °C) (Kwasnik, 1963). In bicarbonate-alkaline water, Ca2 + concentration is limited by CaCO3 precipitation. Consequently there is not enough Ca2 + to fix F−, resulting in high fluoride concentration in water. The fluoride results confirm previous findings that fluoride concentration was positively correlated with salinity, Na+, and HCO− 3 (Rango et al., 2012). Evaporation processes further favors fluoride accumulation, which explains the extraordinarily high salinity and fluoride concentration in those closed-basin soda lakes. In terms of NOM, TOC results suggest that NOM exists as a major component in some of the Tanzanian waters. SUVA results indicate that humic substances are the dominant NOM fraction in these samples. Over 75% of the water samples have NOM concentrations exceeding the EPA guideline (2 mgC/L TOC, details in SI). In some cases very high concentrations were observed, for example up to 260 mgC/L in Ngare Nanyuki Swamp (which means “red water” in Masai) and 35 mgC/L in Maji ya Chai River (“tea river” in Swahili). Such tropical black water owes the black color to NOM, which was also reported elsewhere, for example, in the Siak river in Indonesia (Rixen et al., 2010) and the Rio Negro in Brazil (Barbosa et al., 2003). However, for some alkaline soda lakes, it was claimed that ferrous sulfide is the major reason for the blackness and the abundance of NOM and relatively high temperature account for the high rate of sulfide production (Stahl, 1979). Excessive iron and sulfate concentrations were also found in some samples of this study (data not shown). Natural waters receive NOM from two major sources: (1) the leaching of surrounding soils (pedogenic origin) and (2) the metabolic activities of microorganisms (aquagenic origin) (Schwarzenbach et al.,

Table 2 Proposed water classification in Tanzanian context. Classification

Excellent

Acceptable

Slightly contaminated

Moderately contaminated

Heavily contaminated

pH Turbidity (NTU) EC (μS/cm) F− (mg/L) TOC (mgC/L) SUVA

6.5–8.5 b0.5 b500 b0.5 b1 b0.5

8.5–9.5 0.5–2 500–1000 0.5–1.5 1–2 0.5–1

9.5–10.5 2–4 1000–2500 1.5–4 2–4 1–2

10.5–12.5 4–10 2500–10,000 4–10 4–10 2–4

N12.5 N10 N10,000 N10 N10 N4

524


Table 3 Water quality of the 22 NF/RO treated samples. ID

pH

EC (μS/cm)

Turbidity (NTU)

TOC (mgC/L)

IC (mgC/L)

F− (mg/L)

GPS

1 3 6 8 9 10 11 12 13 15 20 21 23 25 26 28 30 37 40 81 101 120

9.46 9.53 8.92 8.29 8.76 8.75 8.35 8.33 8.00 8.65 9.62 8.38 8.99 9.05 8.34 8.43 8.67 9.60 7.53 8.78 9.44 8.92

9230 8710 1545 2270 922 908 314 548 504 1097 3020 4720 16,150 1535 472 907 609 14,800 443 3930 5310 3340

5.31 14.23 12.84 9.20 3.73 2.15 5.79 0.12 0.90 0.65 3.91 21.75 38.96 497.8 10.13 0.31 0.40 69.20 0.45 0.04 0.78 1.03

32.8 39.8 3.9 243.0 2.3 3.5 3.5 0.9 2.4 0.8 1.7 270.0 58.2 25.7 7.2 1.3 2.1 124.0 10.1 11.4 12.5 114.0

749.0 736.0 123.0 245.0 83.2 79.4 32.3 52.5 47.0 81.1 160.0 463.0 908.0 119.0 50.3 80.3 53.5 856.0 35.6 413.5 441.0 353.0

212.0 239.9 19.4 42.4 22.8 10.2 3.3 10.9 12.8 4.8 22.9 33.1 56.6 3.6 2.6 20.2 17.1 40.7 13.4 50.5 57.6 59.7

S03°12.753′ E036°53.688′ S03°13.322′ E036°53.493′ S03°14.974′ E036°50.913′ S03°17.947′ E036°52.831′ S03°18.179′ E036°52.955′ S03°14.620′ E036°50.350′ S03°13.527′ E036°53.118′ S03°10.225′ E036°40.745′ S03°09.484′ E036°40.887′ S02°32.534′ E035°53.458′ S03°37.004′ E035°44.268′ S03°37.195′ E035°44.382′ S03°27.684′ E035°54.571′ S03°25.509′ E035°53.055′ S03°22.951′ E036°47.478′ S03°19.493′ E036°53.067′ S03°18.937′ E036°52.647′ S03°46.484′ E035°48.063′ S03°10.135′ E036°41.779′ S03°21.190′ E037°08.166′ S03°47.273′ E035°51.138′ S03°10.929′ E036°51.676′

2005; Zumstein and Buffle, 1989). Some researchers have suggested that pedogenic NOM is more hydrophobic and aromatic than aquagenic NOM (Huber et al., 2011; Zumstein and Buffle, 1989). Previous studies using LC-OCD showed that NOM can be characterized into five fractions, namely biopolymers (polysaccharides, proteins and colloidal organics) (N20,000 Da), humic substances (approximately 1000 Da), building blocks (300–500 Da), LMW acids (b 350 Da) and LMW neutrals (b350 Da) (Fujioka et al., 2013; Haberkamp et al., 2008; Huber et al., 2011). 3.2. Impact of membrane characteristics and recovery on fluoride and NOM retention Membrane behavior in terms of fluoride and NOM retention was examined for a natural water (sample 8) containing both high fluoride and high NOM. As seen from Fig. 4a, fluoride retention decreased steadily with increasing MWCO, indicating the importance of size exclusion. As expected the MWCO is directly correlated with permeability (Table 1). The NF270 and TFC-SR3 membranes have very similar pore sizes (MWCO 180 and 167 Da, respectively) but different surface charge densities (Table 1). The more negatively charged TFC-SR3 membrane

had higher fluoride retention than the NF270 (Fig. 4b) despite the slightly higher MWCO. Although fluoride is a small ion with an ionic radius of 0.13 nm, it is highly hydrated in water and thus has a relatively large hydrated radius of 0.34 nm (Richards et al., 2012a). When comparing fluoride retention with the reported NaCl retention of these membranes it should be noted that the fluoride retention is in the same order, albeit significantly higher than NaCl (tested at different conditions). In contrast, NOM retention is consistently high independent of membrane type (Fig. 4b). This is because the molecular size of NOM is significantly larger compared to the membrane pore size (Yoon et al., 2005). Only small amounts of LMW fractions are able to permeate the membrane (Schäfer et al., 2004). These differences can be explained with the conceptual model (Fig. 5) of three key regimes for fluoride with regard to membrane pore versus ion size (Richards et al., 2012b): (1) the hydrated ion fits completely inside the pore; (2) the pore size is between the size of the bare ion and the hydrated ion; (3) the bare ion does not fit inside the pore.

Fig. 3. Percentage of contamination levels of major contaminants in Tanzanian samples.


525

Fig. 4. Fluoride and NOM retention as a function of membrane type with increasing (a) MWCO and (b) zeta potential at 50% recovery (sample 8, water compositions see Table 3).

Of the membranes used in this study, three membranes (NF270, TFC-SR2, TFC-SR3) apply in Regime 1, where charge repulsion becomes dominant instead of size exclusion. The other three membranes (BW30LE, BW30 and NF90) belong to Regime 2. Their high fluoride retention is a result of size exclusion due to hydration, and energy barriers caused by partial dehydration (Richards et al., 2013). The results confirm that the hydration of fluoride cannot be ignored in NF/RO when the membrane pore size is similar to the dimension of the hydrated ion. Recovery is an important hydrodynamic operating parameter in NF/ RO process. Fig. 6 shows the impact of recovery on permeate fluoride concentration and fluoride retention at varying recoveries from 6.25% to 50%. With increasing recovery, permeate fluoride concentration increases as a result of increased concentration in the boundary layer. Regime 1 membranes TFC-SR2 and NF270 have the largest decline in retention. This is presumably due to the fact that higher permeability facilitates concentration polarization at the membrane surface and thus the diffusion rate of fluoride (Bowen and Jenner, 1995; Van de Lisdonk et al., 2001). In contrast, Regime 2 membranes BW30 and BW30-LE have the smallest retention decline, and can effectively remove fluoride to below the WHO guideline almost independent of concentration, which highlights a size exclusion mechanism. Unlike fluoride, permeate NOM concentration decreases with increasing recovery, and reaches a stable value at high recoveries (Fig. 7). This can be explained by size exclusion and the convective transport of NOM which involves LMW fractions being carried in water stream through the membrane (Schäfer et al., 2004). Results show that permeate water flux decreases gradually with increasing recovery (Fig. S), which is attributed to the increase of osmotic pressure in both bulk solution and the boundary layer (van den Berg and Smolders, 1989; Wijmans et al., 1985). The convective transport of NOM thus changes with the permeate water flux. The increased NOM retention

is the result of decreased permeate concentration and increased feed concentration. 3.3. Impact of raw water characteristics on fluoride and NOM retention The influence of ionic strength on fluoride and NOM retention was studied by treating a variety of natural waters with EC ranging from 0.3 to 16.1 mS/cm by the Regime 2 membrane BW30. The net driving pressure refers to the difference between the applied pressure and the osmotic pressure. It should be noted that the osmotic pressure used here was calculated from feed ionic concentrations and during the experiments increases with recovery as a function of retention. This increase was not considered. At the constant applied pressure of 10 bar the net driving pressure decreases proportionally with initial EC (Fig. 8a), resulting the permeate water flux decreases. The extremely low permeate flux at high initial EC (e.g., soda lake samples) is attributed to concentration polarization and osmotic pressure effects, which then due to diffusion result in the very low fluoride retention of about 70% (Fig. 8b). The overall effect of initial EC on fluoride retention is also contributed by charge screening, particularly for Regime 1 membranes where charge repulsion is the dominant mechanism (Bejaoui et al., 2014; Nasr et al., 2013; Paugam et al., 2004). Indeed at high ionic strength, cations such as sodium and calcium neutralize partially the negative charges of the membrane, which consequently decreases the charge repulsion between fluoride and the membrane (Szymczyk and Fievet, 2006). In contrast, the retention of NOM is independent of initial EC. As mentioned above, only LMW fractions can permeate the membrane while other fractions are rejected by size exclusion. Initial EC affects the permeate water flux but its effect on NOM retention is not measurable most likely because the LMW organic fraction is only a small

Fig. 5. Schematic of the three regimes for fluoride with regard to membrane pore size (1) the hydrated ion fits completely inside the pore; (2) the pore size is between the size of the bare ion and the hydrated ion; (3) the bare ion does not fit inside the pore.


25 Net driving pressure Permeate flux

10

20

8

15

6

10

4 5

2 0 100

2

a

Permeate flux (L/m h)

12

Net driving pressure (bar)

526

0

b

Retention (%)

80 60 40 20

Fluoride NOM

0 0

2

4

6

8

10 12 14 16

Initial EC (mS/cm) Fig. 6. (a) Permeate fluoride and (b) fluoride retention by different membranes as a function of recovery (sample 8, Table 3).

Fig. 7. (a) Permeate NOM and (b) NOM retention by different membranes as a function of recovery (sample 8, Table 3).

Fig. 8. (a) Net driving pressure and permeate flux and (b) fluoride and NOM retention as a function of initial EC at 50% recovery (BW30, water compositions see Table 3).

fraction of the TOC and hence of low concentration. A few scatters in NOM retention at low EC ranges are in fact due to the experimental errors from the low initial NOM contents in those samples (for water quality details see Table 3). To understand the dependence on initial fluoride and NOM concentration, the results of the natural water samples tested above with the BW30 membrane were plotted. Selected water samples were further treated with the other membranes to better understand retention mechanisms. Fig. 9 shows that increasing initial fluoride concentration leads to increased permeate concentration and decreased retention, which confirms the diffusive transport of fluoride (Mohammad, 2002). The trend lines of permeate fluoride concentration have a good linear relationships for all membranes, though some scatter is inevitable as each data point represents a very different water composition. The permeate fluoride concentrations of soda lakes are exceptionally high as a result of the extreme ionic strength. These results are included to highlight the complexity and variation of these natural waters. The intersection of the trend line and the WHO guideline indicates the treatability limit in terms of raw water fluoride concentration for a certain membrane. At 50% recovery, the limit for Regime 2 membranes BW30-LE, BW30 and NF90 is 30–40 mg/L (except for soda lakes); for Regime 1 membranes TFC-SR3 it is about 20 mg/L, NF270 and TFC-SR2 it is 5– 10 mg/L. The treatability limit figures offer valuable information for system design and membrane selection. The impact of initial NOM concentration on NOM retention is demonstrated in Fig. 10. All membranes remove NOM consistently below the EPA guideline of 2 mg/L for all water samples except for two soda lakes where low fluxes cause irrational behavior combined with high experimental errors at very low NOM concentrations. The permeate NOM concentration is independent of initial NOM for the very different

Fluoride retention (%) Permeate fluoride (mg/L)


30 27 24 21 18 15 12 9 6 3 0 100

527

a

WHO guideline 1.5 mg/L

b

80 60 40

BW30 BW30-LE NF90 NF270 TFC-SR2 TFC-SR3

20 0 0

50

100

150

200

250

300

Initial NOM (mgC/L) Fig. 9. (a) Permeate fluoride and (b) fluoride retention by different membranes as a function of initial fluoride concentration at 50% recovery.

Fig. 11. (a) Permeate fluoride and (b) fluoride retention by different membranes as a function of initial NOM concentration at 50% recovery.

waters tested due to size exclusion. The NOM retention thus increases with initial NOM concentration. Interestingly, the presence of NOM affects fluoride retention positively (Fig. 11). This trend is particularly evident for Regime 1 membranes. A number of mechanisms could contribute to this observation: (1) increase of membrane surface charge due to the adsorption of negatively charged NOM on the membrane (Childress and Elimelech, 1996; Shim et al., 2002); (2) charge repulsion between fluoride ions and NOM (Kilduff et al., 2004); (3) membrane pore size restriction due to NOM fouling (Comerton et al., 2009; Jarusutthirak et al., 2007); (4) territorial binding of fluoride to NOM structure (Hayes et al., 1995); (5) variation in Donnan exclusion due to the multivalent ions associated with NOM (Seidel et al., 2001; Yaroshchuk, 2001).

Indeed, more studies on membrane characterization and solute–solute interactions are needed before any of the above mentioned mechanisms can be confirmed. With natural water being used as the feed solution, caution must be applied as the effect of NOM on fluoride might be interfered with other parameters, such as ionic strength. This work aims in the first instance to show the influence of NOM on fluoride retention in such natural waters where NOM and fluoride occur in sometimes very high concentrations. Further work is in progress to elucidate the contribution of the possible mechanisms to the observed behavior. 4. Conclusions Fig. 10. (a) Permeate NOM and (b) NOM retention by different membranes as a function of initial NOM concentration at 50% recovery.

In northern Tanzania, the dominant water type is bicarbonatealkaline water. Excessive fluoride concentration is widely detected in

528


such waters as a result of chemical weathering and dissolution. NOM exists as another major constituent, often in combination with fluoride. Such water are common in areas where swamps and tropical black waters occur in fluoride-rich areas. Bench-scale experiments indicate that fluoride transfer through NF/ RO predominantly by diffusion while NOM is retained due to size exclusion or convection. Fluoride retention is significantly influenced by recovery and raw water composition, particularly ionic strength and fluoride and to some degree NOM concentration. High ionic strength results in reduction of the effective driving force and thus enhanced fluoride permeation through the membrane. Initial fluoride concentration determines the treatability of the raw water in terms of the WHO guideline. NOM concentration was determined to have a distinguishable positive effect on fluoride removal, although the mechanism is difficult to elucidate due to the variation of other water parameters in these natural waters. More research with both comparable natural waters and synthetic mixtures is in progress to determine the association between NOM and fluoride and its effect on membrane performance. Acknowledgments The authors would like to thank Leverhulme Royal Society Africa Award SADWAT-Tanzania for project funding. PhD studentship for Junjie Shen was provided by Energy Technology Partnership (ETP) Scholarship with the Drinking Water Quality Regulator for Scotland (DWQR, UK) being the industrial sponsor. A very special thank you goes to Prof Bryce Richards (KIT, Germany) for his financial, technical and material support to this project. We also thank Prof Anthony Szymczyk (Institut des Sciences Chimiques de Rennes, France) for streaming potential measurement of the BW30-LE membrane; Mr Matthew Bower (DWQR, UK) for providing valuable comments on the water classification scheme; Mr John Tobin (Heriot-Watt University, UK) for careful proofreading of the manuscript. Membrane samples were kindly supplied by Dow Chemicals and Koch Membrane Systems. Appendix A. Supplementary data Supplementary data to this article can be found online at http://dx. doi.org/10.1016/j.scitotenv.2015.04.037. References Amini, M., Mueller, K., Abbaspour, K.C., Rosenberg, T., Afyuni, M., Møller, K.N., Sarr, M., Johnson, C.A., 2008. Statistical modeling of global geogenic fluoride contamination in groundwaters. Environ. Sci. Technol. 42, 3662–3668. Ayoob, S., Gupta, A.K., Bhat, V.T., 2008. A conceptual overview on sustainable technologies for the defluoridation of drinking water. Crit. Rev. Environ. Sci. Technol. 38, 401–470. Barbosa, A.C., de Souza, J., Dorea, J.G., Jardim, W.F., Fadini, P.S., 2003. Mercury biomagnification in a tropical black water, Rio Negro, Brazil. Arch. Environ. Contam. Toxicol. 45, 235–246. Bejaoui, I., Mnif, A., Hamrouni, B., 2011. Influence of operating conditions on the retention of fluoride from water by nanofiltration. Desalin. Water Treat. 29, 39–46. Bejaoui, I., Mnif, A., Hamrouni, B., 2014. Performance of reverse osmosis and nanofiltration in the removal of fluoride from model water and metal packaging industrial effluent. Sep. Sci. Technol. 49, 1135–1145. Bhatnagar, A., Kumar, E., Sillanpää, M., 2011. Fluoride removal from water by adsorption—a review. Chem. Eng. J. 171, 811–840. Boussu, K., Zhang, Y., Cocquyt, J., Van der Meeren, P., Volodin, A., Van Haesendonck, C., Martens, J.A., Van der Bruggen, B., 2006. Characterization of polymeric nanofiltration membranes for systematic analysis of membrane performance. J. Membr. Sci. 278, 418–427. Bowen, W.R., Jenner, F., 1995. Theoretical descriptions of membrane filtration of colloids and fine particles: an assessment and review. Adv. Colloid Interf. Sci. 56, 141–200. Bowen, W.R., Mohammad, A.W., 1998. Characterization and prediction of nanofiltration membrane performance—a general assessment. Chem. Eng. Res. Des. 76, 885–893. Bowen, W.R., Mukhtar, H., 1996. Characterisation and prediction of separation performance of nanofiltration membranes. J. Membr. Sci. 112, 263–274. Brunson, L.R., Sabatini, D.A., 2014. Practical considerations, column studies and natural organic material competition for fluoride removal with bone char and aluminum amended materials in the Main Ethiopian Rift Valley. Sci. Total Environ. 488–489, 580–587.

Childress, A.E., Elimelech, M., 1996. Effect of solution chemistry on the surface charge of polymeric reverse osmosis and nanofiltration membranes. J. Membr. Sci. 119, 253–268. Chowdhury, S., Champagne, P., 2008. An investigation on parameters for modeling THMs formation. Glob. NEST J. 10, 80–91. Comerton, A.M., Andrews, R.C., Bagley, D.M., 2009. The influence of natural organic matter and cations on fouled nanofiltration membrane effective molecular weight cut-off. J. Membr. Sci. 327, 155–163. DiGiano, F.A., Arweiler, S., Riddick, J.A., 2000. Alternative tests for evaluating NF fouling. J. Am. Water Works Assoc. 92, 103–115. Edmunds, W.M., Smedley, P., 2005. Fluoride in natural waters. In: Selinus, O., Alloway, B., Centeno, J.A., Finkelman, R.B., Fuge, R., Lindh, U., et al. (Eds.), Essentials of Medical Geology: Impacts of The Natural Environment on Public Health. Academic Press, US, pp. 301–329. Fawell, J., Bailey, K., Chilton, J., Dahi, E., Fewtrell, L., Magara, Y., 2006. Fluoride in Drinking Water. London World Health Organization. Freger, V., Gilron, J., Belfer, S., 2002. TFC polyamide membranes modified by grafting of hydrophilic polymers: a FT-IR/AFM/TEM study. J. Membr. Sci. 209, 283–292. Frencken, J.E., 1990. Endemic Fluorosis in Developing Countries: Causes, Effects and Possible Solutions. NIPG-TNO, TNO Institute for Preventive Health Care. Fujioka, T., Khan, S.J., McDonald, J.A., Henderson, R.K., Poussade, Y., Drewes, J.E., Nghiem, L.D., 2013. Effects of membrane fouling on N-nitrosamine rejection by nanofiltration and reverse osmosis membranes. J. Membr. Sci. 427, 311–319. Ghiglieri, G., Pittalis, D., Cerri, G., Oggiano, G., 2012. Hydrogeology and hydrogeochemistry of an alkaline volcanic area: the NE Mt. Meru slope (East African Rift — Northern Tanzania). Hydrol. Earth Syst. Sci. Discuss. 16, 529–541. Gizaw, B., 1996. The origin of high bicarbonate and fluoride concentrations in waters of the Main Ethiopian Rift Valley, East African Rift system. J. Afr. Earth Sci. 22, 391–402. Haberkamp, J., Ernst, M., Böckelmann, U., Szewzyk, U., Jekel, M., 2008. Complexity of ultrafiltration membrane fouling caused by macromolecular dissolved organic compounds in secondary effluents. Water Res. 42, 3153–3161. Hayes, D., Carter, J., Manning, T., 1995. Fluoride binding to humic acid. J. Radioanal. Nucl. Chem. 201, 135–141. Hong, S., Elimelech, M., 1997. Chemical and physical aspects of natural organic matter (NOM) fouling of nanofiltration membranes. J. Membr. Sci. 132, 159–181. Huber, S.A., Balz, A., Abert, M., Pronk, W., 2011. Characterisation of aquatic humic and non-humic matter with size-exclusion chromatography–organic carbon detection– organic nitrogen detection (LC-OCD-OND). Water Res. 45, 879–885. Jarusutthirak, C., Mattaraj, S., Jiraratananon, R., 2007. Factors affecting nanofiltration performances in natural organic matter rejection and flux decline. Sep. Purif. Technol. 58, 68–75. Jonsson, G., 1986. Transport phenomena in ultrafiltration: membrane selectivity and boundary layer phenomena. Pure Appl. Chem. 58, 1647. Kilduff, J.E., Mattaraj, S., Belfort, G., 2004. Flux decline during nanofiltration of naturallyoccurring dissolved organic matter: effects of osmotic pressure, membrane permeability, and cake formation. J. Membr. Sci. 239, 39–53. Kim, K., Jeong, G.Y., 2005. Factors influencing natural occurrence of fluoride-rich groundwaters: a case study in the southeastern part of the Korean Peninsula. Chemosphere 58, 1399–1408. Kwasnik, W., 1963. Fluorine compounds. In: Brauer, G. (Ed.), Handbook of Preparative Inorganic Chemistry 1. Academic Press, New York, US. Meylan, S., Hammes, F., Traber, J., Salhi, E., von Gunten, U., Pronk, W., 2007. Permeability of low molecular weight organics through nanofiltration membranes. Water Res. 41, 3968–3976. Mnif, A., Ben Sik Ali, M., Hamrouni, B., 2010. Effect of some physical and chemical parameters on fluoride removal by nanofiltration. Ionics 16, 245–253. Mohammad, A.W., 2002. A modified Donnan–steric-pore model for predicting flux and rejection of dye/NaCl mixture in nanofiltration membranes. Sep. Sci. Technol. 37, 1009–1029. Nasr, A.B., Charcosset, C., Amar, R.B., Walha, K., 2013. Defluoridation of water by nanofiltration. J. Fluor. Chem. 150, 92–97. Neale, P.A., 2009. Influence of solute–solute interactions on membrane filtration. School of Engineering, PhD Thesis. University of Edinburgh, Edinburgh, UK. Newcombe, G., Drikas, M., Assemi, S., Beckett, R., 1997. Influence of characterised natural organic material on activated carbon adsorption: I. Characterisation of concentrated reservoir water. Water Res. 31, 965–972. Paugam, L., Taha, S., Dorange, G., Jaouen, P., Quéméneur, F., 2004. Mechanism of nitrate ions transfer in nanofiltration depending on pressure, pH, concentration and medium composition. J. Membr. Sci. 231, 37–46. Pontié, M., Dach, H., Leparc, J., Hafsi, M., Lhassani, A., 2008. Novel approach combining physico-chemical characterizations and mass transfer modelling of nanofiltration and low pressure reverse osmosis membranes for brackish water desalination intensification. Desalination 221, 174–191. Rango, T., Kravchenko, J., Atlaw, B., McCornick, P.G., Jeuland, M., Merola, B., Vengosh, A., 2012. Groundwater quality and its health impact: an assessment of dental fluorosis in rural inhabitants of the Main Ethiopian Rift. Environ. Int. 43, 37–47. Richards, L.A., 2012. The removal of inorganic contaminants using nanofiltration and reverse osmosis. School of Engineering and Physical Sciences, PhD thesis. Heriot-Watt University, Edinburgh, UK. Richards, L.A., Vuachère, M., Schäfer, A.I., 2010. Impact of pH on the removal of fluoride, nitrate and boron by nanofiltration/reverse osmosis. Desalination 261, 331–337. Richards, L.A., Schäfer, A.I., Richards, B.S., Corry, B., 2012a. The importance of dehydration in determining ion transport in narrow pores. Small 8, 1701–1709. Richards, L.A., Schafer, A.I., Richards, B.S., Corry, B., 2012b. Quantifying barriers to monovalent anion transport in narrow non-polar pores. Phys. Chem. Chem. Phys. 14, 11633–11638.

J. Shen, A.I. Schäfer / Science of the Total Environment 527–528 (2015) 520–529 Richards, L.A., Richards, B.S., Corry, B., Schäfer, A.I., 2013. Experimental energy barriers to anions transporting through nanofiltration membranes. Environ. Sci. Technol. 47, 1968–1976. Rixen, T., Baum, A., Sepryani, H., Pohlmann, T., Jose, C., Samiaji, J., 2010. Dissolved oxygen and its response to eutrophication in a tropical black water river. J. Environ. Manag. 91, 1730–1737. Sadiq, R., Rodriguez, M.J., 2004. Disinfection by-products (DBPs) in drinking water and predictive models for their occurrence: a review. Sci. Total Environ. 321, 21–46. Sargaonkar, A., Deshpande, V., 2003. Development of an overall index of pollution for surface water based on a general classification scheme in Indian context. Environ. Monit. Assess. 89, 43–67. Schäfer, A.I., Fane, A.G., Waite, T.D., 1998. Nanofiltration of natural organic matter: removal, fouling and the influence of multivalent ions. Desalination 118, 109–122. Schäfer, A.I., Pihlajamäki, A., Fane, A.G., Waite, T.D., Nyström, M., 2004. Natural organic matter removal by nanofiltration: effects of solution chemistry on retention of low molar mass acids versus bulk organic matter. J. Membr. Sci. 242, 73–85. Schwarzenbach, R.P., Gschwend, P.M., Imboden, D.M., 2005. An introduction to environmental organic chemicals. Environmental Organic Chemistry. John Wiley & Sons, Inc., pp. 13–54. Seidel, A., Waypa, J.J., Elimelech, M., 2001. Role of charge (Donnan) exclusion in removal of arsenic from water by a negatively charged porous nanofiltration membrane. Environ. Eng. Sci. 18, 105–113. Shen, J., Schäfer, A., 2014. Removal of fluoride and uranium by nanofiltration and reverse osmosis: a review. Chemosphere 117, 679–691. Shim, Y., Lee, H.-J., Lee, S., Moon, S.-H., Cho, J., 2002. Effects of natural organic matter and ionic species on membrane surface charge. Environ. Sci. Technol. 36, 3864–3871. Smedley, P.L., Nkotagu, H., Pelig-Ba, K., MacDonald, A.M., Tyler-Whittle, R., Whitehead, E.J., Kinniburgh, D.G., 2002. Fluoride in groundwater from high-fluoride areas of Ghana and Tanzania. British Geological Survey, Nottingham. Stahl, J.B., 1979. Black water and two peculiar types of stratification in an organically loaded strip-mine lake. Water Res. 13, 467–471. Szymczyk, A., Fievet, P., 2006. Ion transport through nanofiltration membranes: the steric, electric and dielectric exclusion model. Desalination 200, 122–124.

529

Tang, C.Y., Kwon, Y.-N., Leckie, J.O., 2007. Fouling of reverse osmosis and nanofiltration membranes by humic acid—Effects of solution composition and hydrodynamic conditions. J. Membr. Sci. 290, 86–94. Tang, C.Y., Kwon, Y.-N., Leckie, J.O., 2009. Effect of membrane chemistry and coating layer on physiochemical properties of thin film composite polyamide RO and NF membranes: I. FTIR and XPS characterization of polyamide and coating layer chemistry. Desalination 242, 149–167. Tanzania Bureau of Standards, 2008. Tanzania Water Quality Standards. 2013 (Dar es Salaam, Tanzania). Van de Lisdonk, C.A.C., Rietman, B.M., Heijman, S.G.J., Sterk, G.R., Schippers, J.C., 2001. Prediction of supersaturation and monitoring of scaling in reverse osmosis and nanofiltration membrane systems. Desalination 138, 259–270. van den Berg, G.B., Smolders, C.A., 1989. Concentration polarization phenomena during dead-end ultrafiltration of protein mixtures. The influence of solute–solute interactions. J. Membr. Sci. 47, 1–24. van't Hoff, J.H., 1887. The function of osmotic pressure in the analogy between solutions and gases. Proc. Phys. Soc. Lond. 9, 307. Weishaar, J.L., Aiken, G.R., Bergamaschi, B.A., Fram, M.S., Fujii, R., Mopper, K., 2003. Evaluation of specific ultraviolet absorbance as an indicator of the chemical composition and reactivity of dissolved organic carbon. Environ. Sci. Technol. 37, 4702–4708. Wijmans, J.G., Nakao, S., van den Berg, J.W.A., Troelstra, F.R., Smolders, C.A., 1985. Hydrodynamic resistance of concentration polarization boundary layers in ultrafiltration. J. Membr. Sci. 22, 117–135. World Bank, 2006. Water Resources in Tanzania (Dar es Salaam). Yaroshchuk, A.E., 2001. Non-steric mechanisms of nanofiltration: superposition of Donnan and dielectric exclusion. Sep. Purif. Technol. 22–23, 143–158. Yoon, Y., Amy, G., Cho, J., Her, N., 2005. Effects of retained natural organic matter (NOM) on NOM rejection and membrane flux decline with nanofiltration and ultrafiltration. Desalination 173, 209–221. Zularisam, A.W., Ismail, A.F., Salim, R., 2006. Behaviours of natural organic matter in membrane filtration for surface water treatment—a review. Desalination 194, 211–231. Zumstein, J., Buffle, J., 1989. Circulation of pedogenic and aquagenic organic matter in an eutrophic lake. Water Res. 23, 229–239.