reverse osmosis with varying water chemistry

reverse osmosis with varying water chemistry

Accepted Manuscript Removal of fluoride and natural organic matter removal from natural tropical brackish waters by nanofiltration/reverse osmosis wit...

2MB Sizes 3 Downloads 296 Views

Accepted Manuscript Removal of fluoride and natural organic matter removal from natural tropical brackish waters by nanofiltration/reverse osmosis with varying water chemistry Isaac Owusu-Agyeman, Michael Reinwald, Azam Jeihanipour, Andrea Iris Schäfer PII:

S0045-6535(18)31996-9

DOI:

https://doi.org/10.1016/j.chemosphere.2018.10.135

Reference:

CHEM 22397

To appear in:

ECSN

Received Date: 8 May 2018 Revised Date:

16 October 2018

Accepted Date: 18 October 2018

Please cite this article as: Owusu-Agyeman, I., Reinwald, M., Jeihanipour, A., Schäfer, A.I., Removal of fluoride and natural organic matter removal from natural tropical brackish waters by nanofiltration/ reverse osmosis with varying water chemistry, Chemosphere (2018), doi: https://doi.org/10.1016/ j.chemosphere.2018.10.135. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

1

ACCEPTED MANUSCRIPT Removal of fluoride and natural organic matter removal from natural tropical

2

brackish waters by nanofiltration/reverse osmosis with varying water chemistry

3 4

Isaac Owusu-Agyeman, Michael Reinwald, Azam Jeihanipour, Andrea Iris Schäfer*

RI PT

5 6

Membrane Technology Department, Institute of Functional Interfaces (IFG), Karlsruhe Institute of

7

Technology (KIT), Hermann-von-Helmholtz-Platz 1, 76344 Eggenstein-Leopoldshafen, Germany.

SC

8

*corresponding author: Andrea Iris Schäfer, +49 (0)721 6082 6906, [email protected]

10 11 12

submitted to Chemosphere

13

May 2018

M AN U

9

revision October 2018

15

TE D

14

Abstract

17

In the context of decentralised brackish water treatment in development applications, the

18

influence of water quality on membrane separation was investigated with real waters. High ionic

19

strength (low net driving pressure) on fluoride (F) retention by nanofiltration (NF) and reverse

20

osmosis (RO) was investigated over a wide pH range (2-12). Further, the influence of pH on the

21

permeation of natural organic matter (NOM) fractions, in particular low molecular weight (LMW)

22

neutrals, was elucidated. Natural and semi-natural waters from Tanzania with similar F

23

concentrations of about 50 mg L−1 but varying NOM and inorganic carbon (IC) concentration were

24

filtered with an NF and RO, namely NF270 and BW30.

AC C

EP

16

25

F retention by NF270 for the feed water with highest ionic strength and IC concentration

26

was lower and attributed to charge screening. This parameter further reduced at high pH due to co-

ACCEPTED MANUSCRIPT

ions (F− and CO32−) interactions and combined (synergistic) effect of high salt concentration and

28

pH on F. High NOM resulted in higher membrane zeta potential in comparison with low NOM

29

natural water. However, there was no significant difference in F retention due to the fact that F

30

retention enhancement was annulled by deposit formation on the membrane. The fraction of NOM

31

found in NF/RO permeates was dominated by LMW neutrals. This was attributed to their size and

32

uncharged nature, while their higher concentration at low pH remains unexplained. More humic

33

substances (HS) of higher molecularity and aromaticity permeated the NF270 when compared with

34

BW30, which can be explained with the different membrane molecular weight cut off (MWCO).

SC

RI PT

27

The study highlights the complexity of treating tropical natural waters with elevated F and

36

NOM concentrations. In order to develop appropriate membrane systems that will achieve optimal

37

F and NOM removal, the influence of water quality parameters such as pH, NOM content, ionic

38

strength and IC concentration requires understanding. Seasonal variation of water quality as well as

39

operational fluctuations, which occur in particular when such treatment processes are operated with

40

renewable energy, will require such challenges to be addressed. Further, given the high

41

permeability of low molecular weight (LMW) organics significant permeate side fouling may be

42

expected.

45 46 47

TE D

EP

44

Keywords: water quality; tropical water; carbonaceous water; humic substances; liquid

AC C

43

M AN U

35

chromatography organic carbon detection (LC-OCD)

2

48

ACCEPTED MANUSCRIPT

1. Introduction

Excessive fluoride (F) is a major drinking water quality issue in many global regions, and in

50

particular in the East African Rift Valley where some of the world's highest F concentrations occur

51

(Smedley et al., 2002). The occurrence of naturally high fluoride levels originates from the

52

interaction of water with fluoride-rich volcanic rocks in this region (Fawell et al., 2013). Abnormal

53

F levels occur through chemical weathering of fluorine bearing minerals and it are associated with

54

high CO2 pressure and low Ca content (Frencken, 1990; Gizaw, 1996). Subsequent evaporation in

55

waterways further increases the F concentration, alkalinity, and salinity of the water (Gizaw, 1996).

56

The high F concentration of surface- and ground waters in Northern Tanzania is associated

57

with sodium-bicarbonate (Na-HCO3) type of water and varying pH values (3−12) (unpublished

58

data, (Gizaw, 1996). This is evidence in the high Na+ (≈ 6600 mg L−1), HCO3− (≈ 4300 mg L−1) and

59

CO32− (≈ 7200 mg L−1) concentrations as well as high ionic strength accompanying fluoride-rich

60

soda lakes in this region (Nanyaro et al., 1984). Likewise, Magadi, a salt from soda lakes such as

61

Lake Natron in the region, which is used by locals for cooking, is found to have F content of up to

62

9 mg g−1 and consists mainly of trona (Na2CO3·NaHCO3·2H2O) (Nielsen, 1999; Kaseva, 2006).

TE D

M AN U

SC

RI PT

49

A moderate concentration of F in drinking water (1 mg L−1) is beneficial to humans by

64

preventing tooth decay (Fawell et al., 2006), while excessive F intake results in various forms of

65

dental and skeletal fluorosis (Fewtrell et al., 2006). The World Health Organization (WHO)

66

recommends guideline value of 1.5 mg L−1 of F in drinking water (Fawell et al., 2006). The high F

67

concentration in natural waters coupled with the unavailability of alternative water sources forced

68

the Tanzanian authorities to raise the F limit in drinking water to 8.0 mg L−1 in 2007. This was

69

reduced to 4.0 mg L−1 in 2010 (Government of Tanzania, 2007; African Development Bank, 2010).

70

Especially in tropical regions, natural waters with high F can have simultaneously elevated

71

concentration of natural organic matter (NOM). A typical example is a water from the Ngare

72

Nanyuki River, the main and most important river crossing the densely populated Ngarenanyuki

AC C

EP

63

3

ACCEPTED MANUSCRIPT

Ward in the Arumeru District of northern Tanzania (Istituto Oikos, 2011). The river water has been

74

reported to have F and NOM concentrations as high as 60 mg L−1 and 114 mg C L−1, respectively

75

(Owusu-Agyeman et al., 2018). NOM in natural waters can be aquagenic (originating from water)

76

or pedogenic (originating from soil). The origin of NOM determines its properties and/or

77

composition which inevitably affects treatability. It has been shown that NOM from pedogenic

78

origin has higher molecularity and aromaticity than aquagenic NOM (Filella, 2009). NOM can be

79

fractionated using liquid chromatograph organic carbon detection (LC-OCD), where fractions

80

include humic substances (HS), biopolymers, building blocks, low molecular weight (LMW)

81

organic acids and neutrals. LMW neutrals typically have a complex composition, and consist

82

mainly of alcohols, aldehydes, ketones, sugars, and amino acids (Huber et al., 2011). Pesticides,

83

pharmceutical and agro-chemicals with no charge can further contribute to LMW neutrals. The

84

catchment area of Ngare Nanyuki is known of using pesticides such as in the farmlands which can

85

potentially be tranportant to the river through runoff. Photobleaching can decrease the size of high-

86

molecular weight fractions of NOM to LMW constituents and change their ultravoilet absorbance

87

capacity (Brinkmann et al., 2003; Lou and Xie, 2006). Such organic compounds cannot always be

88

detected by UV-persulphate organic carbon detectors. Photobleaching is influenced by pH,

89

temperature and salinity (Song et al., 2017). Due to the complex nature of LMW neutrals, as well

90

as their regrowth potential, there is a critical need to investigate the removal of such materials in

91

water treatment.

AC C

EP

TE D

M AN U

SC

RI PT

73

92

NOM quality and quantity of water in tropical regions are different from non-tropical

93

waters (Findlay and Sinsabaugh, 2003; Johnson et al., 2011). This is evidenced by a spectroscopic

94

investigation that has indicated that NOM in tropical waters is mostly of pedogenic origin with the

95

biggest fraction being humic substances. This results in higher molecular weight and aromatic ring

96

content than NOM in waters from temperate regions (Oliveira et al., 2006). Tropical waters,

97

therefore, have higher specific ultraviolet absorbance (SUVA), where a value above 2 L mg−1 cm−1

4

ACCEPTED MANUSCRIPT

98

is an indication of high disinfection by-product formation potential (Environmental Protection

99

Agency Ireland, 2012). For waters from northern Tanzania, high aromatic contents with SUVA

100

values in the range of 2−6 L mg−1 m−1 have been reported (Shen and Schäfer, 2015; Aschermann et

101

al., 2016). Nanofiltration/reverse osmosis (NF/RO) are suitable for defluoridation and NOM removal

103

from water with high removal efficiency in comparison with other methods (Schäfer et al., 2004;

104

Ayoob et al., 2008; Song et al., 2011). High molecular weight (HMW) fractions of NOM are to a

105

large extent rejected by NF/RO and have been reported to be responsible for flux decline (Schäfer

106

et al., 2004; Fan et al., 2008). LMW organics remain a challenge even though the overall organics

107

rejection is high. While charged LMW molecules are retained to some extent by electrostatic

108

repulsion, LMW neutrals permeate more easily and may cause regrowth (Meylan et al., 2007).

M AN U

SC

RI PT

102

Fluoride removal by NF/RO is influenced by a number of factors including pH, initial F

110

concentration, solute-solute interaction, and ionic strength (Hu and Dickson, 2006; Shen and

111

Schäfer, 2015). A positive influence of NOM on F removal has been observed (Shen and Schäfer,

112

2015). In view of this beneficial observation and the simultaneous occurrence of F and IC; the

113

mechanisms of humic acid (HA) and IC impact on F retention by NF/RO have been explored using

114

synthetic waters (Owusu-Agyeman et al., 2017). However, little is known about the influence of

115

pH, IC speciation and ionic strength on NOM and F removal from real natural surface waters by

116

NF/RO membranes. This work contributes to the understanding of the complex and interlinked

117

processes of natural waters containing different quantities of NOM and IC at high F over a broad

118

pH range.

AC C

EP

TE D

109

119

The main research questions will be (i) what is the impact of extreme ionic strength (low

120

net driving pressure) on F retention at variable pH (and hence speciation) and (ii) how does pH

121

affect the permeation of NOM fractions, specifically LMW neutrals?

5

ACCEPTED MANUSCRIPT

122

2. Materials and Methods

123

2.1. Chemicals and Milli-Q water The pH of natural waters was adjusted with 1 M HCl (VWR Chemical, Germany) and 1 M

125

NaOH (Merck, EMD Millipore Corporation, Germany). Total ionic strength adjustment buffer

126

(TISAB) for F analysis was prepared using trans-1,2-Diaminocyclohexane tetraacetic acid (CDTA)

127

(Acros Organics, USA), NaCl (99.9%, VWR Chemicals, Germany) and glacial acetic acid (100%,

128

Merck, EMD Millipore Corporation, Germany). Milli-Q water having a conductivity of 0.05 µS

129

cm−1 (Milli-Q® Type 1 water, Merck Millipore, Germany) was used for the preparation of all

130

solutions.

131

2.2. Water sampling

M AN U

SC

RI PT

124

A semi-natural water containing Magadi (a salt containing trona and high F from soda

133

lakes) and two natural waters were studied. The sampling region of the waters was the Northern

134

region of Tanzania (see Figure 1). For the semi-natural water, two types of Magadi were purchased

135

from road side sellers at the Southwest of the Lake Natron (north of Engare Sero) during a field trip

136

in 2012. The two samples, different in composition, were mixed 1:1 (w/w) and dissolved in Milli-Q

137

water to prepare a solution with F concentration of around 50 mg L−1, which is similar to the two

138

natural waters.

AC C

EP

TE D

132

139 140

Figure 1. Pictures of A) Mdori water source, B) Magadi sellers at Lake Natron (north of

141

Engare Sero) and C) Ngare Nanyuki water source (© Schäfer).

142 6

ACCEPTED MANUSCRIPT

The two natural waters studied were from Mdori and Ngare Nanyuki (Olkungwado). The

144

water from Mdori was taken from a borehole in the location S03°47.273' E035°51.138' (Figure 2)

145

in December 2013. The Ngare Nanyuki water samples were collected from two locations

146

S03°10.929' E036°51.676' and S03°11.141' E036°51.738' in October 2013 and mixed 1:1 (v/v).

147

All natural water samples were collected in 5 L plastic containers. The containers were washed

148

thoroughly with the natural water before filling. Samples were sealed and stored at room

149

temperature before airlifting to Karlsruhe Institute of Technology (KIT), Germany. Upon arrival in

150

Germany, samples were stored at a temperature of 4 °C until used.

151 152

Figure 2. Geographical location of sampling sites (map adapted from Tracks4Africa)

154

AC C

153

EP

TE D

M AN U

SC

RI PT

143

2.3. NF/RO filtration system and protocol

155

A stainless steel stirred cell set up was used for the filtration process (Figure 3). The cell

156

had internal membrane diameter of 7 cm and effective membrane surface area of 38.5 cm2. The

157

internal volume of the cell is 990 mL and was equipped with a magnetic stirrer assembly

158

(Millipore, UK), set at a stirring speed of 400 rpm. Pressure and feed temperature were measured by

7

ACCEPTED MANUSCRIPT

159

a pressure transducer (PX209-300GV5) and a thermocouple (TJ2-CPSS-M60U-250-SB) purchased

160

from Omega Engineering, UK. Pressure regulator Pressure sensor Stirred cell setup 3

Stirred cell setup 2

P T

Data acquisition

Synthetic air Stirred cell

Magnetic stirrer

SC

Membrane Rpm

Stirred cell setup 1

Temperature sensor

Magnetic stirrer assembly

Support

CPU/ LabView

RI PT

Pressure regulator

Sample vials

25.0 g

162

M AN U

161

Balance

Figure 3. Schematic of stirred cell experimental apparatus with data acquisition

163

For each experiment, a new membrane coupon from the same membrane sheet was used.

165

Membranes were soaked in 10 mM NaCl for one hour prior to use. Pressure was set at 9.8 bar for

166

all experiments: compaction using Milli-Q water (1 hour), pure water flux determination (30

167

minutes), experiment with 400 mL of feed solutions, and post-filtration pure water flux

168

measurement (30 minutes). Normalized flux was determined as the ratio of permeate flux (J)

169

during filtration to initial pure water flux (JW0). Recovery was calculated as per equation (1) and

170

eight permeate samples of 25 mL each (total permeate volume of 200 mL) were collected, resulting

171

in a recovery of 50% for Mdori and Ngare Nanyuki waters. Filtration of Magadi waters was

172

performed at a recovery of 30% due to the high ionic strength resulting in high osmotic pressure.

AC C

EP

TE D

164

Recovery =

173 174

∑ VP VF

(1)

where: VP:

volume of permeate (mL) 8

VF:

175 176

ACCEPTED MANUSCRIPT

volume of feed (mL)

Osmotic pressure (П) of the feed water is calculated from the equation (2) (2)

П = RT·∑Ci (bar) 177

Where: R:

universal gas constant (L bar mol−1 K−1),

179

T:

absolute temperature (K),

180

∑Ci:

the sum of concentrations of solutes (mol L−1).

SC

RI PT

178

In the experiments pH was varied from pH 2 to 12, if possible. This is in such real waters

182

limited by the degassing at low pH, scaling at high pH and accompanied by a significant increase

183

of conductivity at pH values <4 and >10. The decision to vary pH of such natural water was taken

184

to investigate the impact of solute speciation on nanofiltration performance and to determine the

185

limitation to treat waters of similar nature but different pH. Significant variation of natural water

186

pH has been observed in prior work (Rossiter et al., 2010).

187

2.4. Membrane types and characteristics

TE D

M AN U

181

Two commercial membrane types, BW30 and NF270, were selected. Both membranes are

189

polyamide thin film composite (TFC) membrane types supplied as flat sheet samples by Dow

190

FilmTechTM (Minneapolis, MN, USA). BW30 is a fully aromatic polyamide based RO membrane

191

used in brackish water applications, whereas NF270 is a piperazine based polyamide NF membrane

192

with semi-aromatic, weakly acidic COO¯ groups (Khan et al., 2011). The main selection criteria

193

was to compare a tight (BW30) and a loose (NF270) membrane with very different salt retention

194

characteristics to understand mechanisms of inorganic and organic retention and to determine

195

limiting factors such as concentration polarization or fouling and scaling in treatment. The

196

incomplete fluoride retention displayed by the NF270 membrane allows to study the influence of

197

organic matter on fluoride retention better than if retention is high.

AC C

EP

188

9

198

ACCEPTED MANUSCRIPT

2.5. Streaming potential measurement

Streaming potential measurements were performed with an electrokinetic analyser

200

(SurPASS, Anton Paar GmbH, Austria) to determine the zeta potential of the membrane in Mdori

201

and Nagre Nanyuki waters. Streaming potential measurements were done by Anton Paar, Austria.

202

Membrane samples were mounted on two rectangular sample holders (2 cm by 1 cm). Sample

203

holders were placed in the adjustable gap cell at a gap height of approximately 100 µm. The system

204

was filled and rinsed with the natural water to be used as electrolyte before measurement.

205

2.6. Analytical methods

SC

RI PT

199

Electrical conductivity (EC) was measured using an InoLab Cond Level 2 meter with a

207

TetraCon® 325 electrode (WTW, Germany). The pH-measurement was conducted using a WTW

208

InoLab pH720 meter with a SenTix81 electrode (WTW, Germany). Fluoride concentration was

209

measured using an ion F800 selective electrode (ISE) and pH/ION 3310 meter (WTW, Germany).

210

TISAB buffer was used (sample to TISAB ratio of 1:1 (v/v)) for F measurements to mask chemical

211

interferences resulting from pH and EC and hence increase accuracy.

TE D

M AN U

206

Inorganic carbon (IC) and total organic carbon (TOC) were measured with a TOC analyser

213

(Sievers M9 portable TOC analyser with autosampler, GE Analytical Instruments, UK).

214

Absorbance was determined with a UV-Vis spectrometer (UV/VIS Spectrometer Lambda 25,

215

PerkinElmer, USA). All samples were measured against a Milli-Q water reference using a 1 cm

216

quartz cuvette. The UV absorbance at 254 nm (UV254) was used to calculate the specific ultraviolet

217

absorbance (SUVA) as in equation (3):

AC C

EP

212

SUVA =

UV254nm (L mg-1 m-1 ) DOC

(3)

218

Organic carbon in the Ngare Nanyuki water samples was characterised using liquid

219

chromatography – organic carbon detection (LC-OCD) (DOC Labor, Germany) equipped with UV

220

(UVD), organic carbon (OCD) and organic nitrogen (OND) detectors. This method characterizes

10

ACCEPTED MANUSCRIPT

221

the organic carbon fractions, separated by a size exclusion chromatography column (Huber et al.,

222

2011). Discrepancies between organic carbon concentration measured with the LC-OCD and TOC

223

analyser (Sievers M9) are discussed in the supporting information (SI) (Fig. SI 6). The aromaticity of NOM in the water samples and the molecularity of HS were determined

225

using LC-OCD. Aromaticity is defined as the ratio of spectral absorption coefficient (SAC) to the

226

concentration of HS in the dissolved organic carbon (DOC) (see equation (4)). Aromaticity is,

227

therefore, the specific UV absorption of the HS peak.

-1

Aromaticity (L mg m ) =

SAC (HS)(m-1 ) DOC (HS) (mg L-1 )

(4)

SC

-1

RI PT

224

Molecularity of HS was calculated on the basis of calibration of the column with reference

229

HS of the International Humic Substances Society (IHSS) (Huber et al., 2011). The molar mass

230

calibration of the column is derived from equation (5).

M = exp ( −

M AN U

228

tr − A ) B

232

where:

235 236

retention time (min)

A, B: matching coefficients calculated from calibration of the column with IHSS humic acid (HA) and fulvic acid (FA)

AC C

234

tr:

EP

233

TE D

231

(5)

The average molecular mass (Mn) of the HS fraction is then determined by equation (6).

Mn =

∑i(ni Mi ) ∑ i ni

(6)

237 238 239

where: n:

number of molecules

11

240

M:

ACCEPTED MANUSCRIPT

molar mass (g mol−1)

241

The radius (r) of HS fraction of NOM was calculated from equation (7) with the assumption

242

that the molecules are spherical. The equation was developed from the Stokes Einstein equation

243

(Worch, 1993). r = 2.037⋅ 10-9 ⋅ Mn0.53 (nm) 3. Results and discussion

245

3.1. Characteristics of water samples

RI PT

244

(7)

The characteristics of the three waters studied are presented in Table 1. Magadi is a salt

247

formed by natural evaporative processes from natural water streams flowing into the inland soda

248

lakes such as Lake Natron (see Figure 2). The inflowing water gradually evaporates and forms salt

249

crusts at its shore (Nielsen, 1999). Magadi consists of the carbonate mineral trona

250

(Na2CO3·NaHCO3·2H2O) mixed with minor contents of halite (NaCl) and high F concentration up

251

to 9 mg g−1 (Nielsen, 1999). Magadi from Lake Natron has been shown to have one of the highest F

252

concentrations in comparison with other East African Magadis (Nielsen, 1999). A semi-natural

253

water sample was prepared from the Magadi to have a final F concentration similar to the other

254

waters. This resulted in a pH 10, a high IC (1120 mg L−1), salt content (EC = 12430 µS cm−1) and F

255

(49 mg L−1), similar to typical water samples found in the Northern regions of Tanzania (Shen and

256

Schäfer, 2015).

AC C

EP

TE D

M AN U

SC

246

257

Mdori water is a hot borehole water and characterized by high F content (56.2 mg L−1) with

258

an IC concentration of 524 mg L−1. The high F concentration in the Mdori waters can be attributed

259

to dissolution of fluoride containing rocks. The conditions favourable for the high fluoride

260

concentration were high IC and Na+ and low Ca2+ concentrations (Rafique et al., 2015). The

261

detailed compositions of the water samples are provided in the supporting information (Table SI 1).

262

The salinity was low in comparison with the waters prepared from Magadi. The Mdori natural

263

water had NOM concentration of 11.4 mg L−1. The samples had a sulphuric smell indicating 12

ACCEPTED MANUSCRIPT

hydrogen sulphide content. This smell vanished with longer exposure to air. The Ngare Nanyuki

265

water shows similar characteristics in terms of F and IC as the Mdori water. In addition to the high

266

F levels, the Ngare Nanyuki water also had high NOM content (160 mg L−1) (Table 1). The

267

presence of high NOM results in a reddish brown colour, given the river its name Ngare Nanyuki

268

(meaning “red water” in the Maasai language). In a typical tropical region, the high NOM

269

concentration evolves mainly from the decomposition of plants. The alkaline nature (pH 8.6) of the

270

Ngare Nanyuki water can be a contributing factor for the high NOM due to increase in the

271

dissolution of NOM from peat soils (Shen et al., 2016).

272

Table 1. Characteristics of feed water samples Fluoride

EC

IC

TOC

Osmotic pressure

(mg F L− 1)

(µS cm−1)

(mg C L−1)

(mg C L−1)

(bar)

Magadi

10.0

49.2

12430

1120

1.5

6.7

Mdori

9.4

56.2

5110

524

11.4

3.0

Ngare Nanyuki

8.6

53.7

3990

437

160.0

2.3

273

3.2. Zeta potential measurement of membranes in Mdori and Ngare Nanyuki waters

TE D

274

SC

pH

M AN U

Sample

RI PT

264

Streaming potential results with Ngare Nanyuki and Mdori water as electrolyte solutions

276

were compared with 10 mM NaCl to determine the influence of the natural waters on the

277

membrane surface charge (Figure 4). BW30 membrane gave the highest negative zeta potential in

278

Ngare Nanyuki waters when compared with 10 mM NaCl and Mdori water and was attributed to

279

adsorption of NOM on the membrane surface (Childress and Deshmukh, 1998). However, for

280

NF270 which originally had a high zeta potential at high pH, the zeta potential was higher in Ngare

281

Nanyuki water than 10 mM NaCl only at pH <5. Unlike the 10 mM NaCl, the natural waters

282

contained some concentration of multivalent ions which may result in charge screening (Childress

283

and Elimelech, 1996). Mdori water showed somewhat abnormal behaviour for BW30. The zeta

284

potential remains slightly positive without showing an isoelectric point (IEP) in the pH range below

285

the intrinsic pH of this water source (Figure 4A). Since the result was different to all other

AC C

EP

275

13

ACCEPTED MANUSCRIPT

286

observations and especially different to the effect of Mdori on membrane NF270 (see Figure 4B),

287

the zeta potential analysis was repeated using a new sample of BW30. The repetition shows a

288

slightly negative zeta potential. This discrepancy (± 8 mV) between first and second measurement

289

was attributed to measurement error in such complex real samples. Further analysis of the membranes was required in order to exclude the influence of water

291

ionic strength on membrane zeta potential and to verify membrane surface modification by NOM

292

deposition. Zeta potential analysis was performed in the presence of a standard electrolyte NaCl at

293

a low but fixed ionic strength of 1 mM. For this purpose, the membrane was kept mounted in the

294

adjustable gap cell after each measurement with the Tanzanian natural waters and rinsed together

295

with the electrolyte circuit with 500 mL pure water. After exchanging the electrolyte solution to 1

296

mM NaCl, the zeta potential of the membrane was measured only in the acidic pH range 2−5.5

297

(Figure 4C and D). For both membranes the surface charge was higher after exposure to the high

298

NOM Ngare Nanyuki water than the Mdori water.

AC C

EP

TE D

M AN U

SC

RI PT

290

14

20

BW30

A

0 -20 10m NaCl Ngare Nanyuki Mdori Mdori rep

-40 -60 20 C

D

Mdori Ngare Nanyuki Clean

RI PT

Zeta potential (mV)

NF270

B

0 -20 -40 -60 2

6 pH

8

10 2

4

6 pH

8

10

M AN U

299

4

SC

Zeta potential (mV)

ACCEPTED MANUSCRIPT

300

Figure 4: Influence of Tanzania natural waters on membrane surface charge. Zeta potential

301

of (A) BW30 and (B) NF270 measured in the natural waters, and zeta potential of (C)

302

BW30 and (D) NF270 measured 1 in mM NaCl after exposure to natural Tanzania waters

303

305

3.3. NF/RO of fluoride-rich semi-natural water with high IC content and high ionic strength (Magadi)

TE D

304

F removal of the fluoride-rich semi natural water with high ionic strength was performed at

307

reduced recovery (30%) due to a high osmotic pressure built-up during the experiments (Figure 5).

308

pH did not have an observable effect on the normalized flux of the BW30 membrane. However, for

309

NF270, a flux decline was observed at higher pH (see Figure 5A), which was attributed to the

310

decrease in the net driving force at high pH due to the increase in ionic strength due to pH

311

adjustment (see Figure SI 4B). For the BW30 membrane no such flux decline was observed,

312

presumably due to the fact that there was no significant change in the net driving force with pH

313

(see Figure SI 4A). At pH >7, the concentration of IC in the feed was low due to degassing

314

(H2CO3/CO2). IC retention was low at pH 7−9 and increased at pH ≥10 (see Figure 5B), a trend

AC C

EP

306

15

ACCEPTED MANUSCRIPT

315

that has been explained with a shift in IC speciation from HCO3− to CO32− (Simpson et al., 1987;

316

Owusu-Agyeman et al., 2017). As expected F retention by the BW30 was higher than NF270 membrane. In comparison

318

with a similar study with less complex and low ionic strength (EC = 1500 µS cm−1) synthetic

319

waters, the F retention was lower (30−70 % at pH >6) for the more open NF270 membrane. F

320

retention of 80−90 % was achieved by NF270 at pH >6 in the previous study (Owusu-Agyeman et

321

al., 2017). The lower F retention in the more complex Magadi water can be attributed to the high

322

ionic strength which resulted in high osmotic pressure as well as membrane charge screening

323

(Bejaoui et al., 2014; Zaidi et al., 2015). Fluoride retention by NF270 increased initially with pH

324

but decreased towards more basic pH (Figure 5C). The initial increase in retention was as a result

325

of a shift in F speciation from HF to F− (Richards et al., 2009) and increase the membrane surface

326

charge (Teixeira et al., 2005; Mänttäri et al., 2006). The further decrease in F retention by NF270 at

327

higher pH can be explained by (i) co-ions (F− and CO32−) interaction (Owusu-Agyeman et al.,

328

2017) (ii) charge concentration polarisation (Verliefde et al., 2008) and a synergistic effect of

329

higher salt concentration and pH (Freger, 2004; Nilsson et al., 2008; Luo and Wan, 2013). The co-

330

ions interaction occurred between F− and CO32−. At high pH, IC existed as CO32− which is better

331

rejected than HCO3−. This created a charge deficit at the permeate side of the membrane, hence

332

forcing F− ions through the membranes to compensate for the shortage (Yaroshchuk et al., 2011).

333

Secondly, membrane negative charge increases with pH which in turns leads to an increase in the

334

attraction of positively charged ions, resulting in an increase in the membrane wall salt

335

concentration at high pH (Verliefde et al., 2008). The difference between bulk feed concentration

336

and wall concentration was higher for the NF270 than the BW30 membrane, especially at pH >5

337

(see Table SI 2). In consequence, the combined (synergistic) effect of salt and pH arose because of

338

the electrostatic screening by salt ions due to the stronger attraction between the negatively charged

339

membrane and the counter ions. This scenario resulted in the decrease in electrostatic repulsion

AC C

EP

TE D

M AN U

SC

RI PT

317

16

ACCEPTED MANUSCRIPT

340

between F− ions and membrane. However, the decrease in F retention at high pH was not observed

341

for the BW30 membrane, probably due to the fact that size exclusion is the main F retention

342

mechanism for this membrane (Richards et al., 2013; Shen and Schäfer, 2015). The overall EC retention by both BW30 and NF270 increased with pH (Figure 5D). The

344

increase in EC retention for both membranes was attributed to the shift in IC speciation from

345

HCO3− to CO32−. CO32− has a higher charge and hydrated radius (see Table SI 1) and consequently

346

it is more easily retained.

60

M AN U

J/JW0 (%)

80

40

80 60 40 20 0 0

0 100

4

6

8

10

pH

12

2

80 60 40 20 0

4

6

8 pH

10

12

EP

347

2

60

20

D

TE D

F retention (%)

0 100 C

80

40

Degassing

20

100 IC retention (%)

B

EC retention (%)

BW30 NF270

SC

100 A

RI PT

343

Figure 5. Influence of pH on (A) normalized flux (J/JW0), (B) IC retention, (C) F retention,

349

and (D) EC retention at 30% recovery of the semi-natural water (Magadi) with high IC

350

and high ionic strength (pressure 9.8 ± 0.1 bar and temperature 20 ± 2 oC)

351 352

AC C

348

3.4. NF/RO of natural surface water with low NOM and high fluoride content (Mdori)

353

The Mdori water was chosen due to the fact that it contained only a small amount of NOM

354

but high F concentration of around 50 mg L−1. Effects of NOM on the removal of F at different pH

17

ACCEPTED MANUSCRIPT

355

were therefore expected to be insignificant compared to the natural water with high NOM content

356

(Ngare Nanyuki). Normalized flux was relatively high compared with the Magadi waters especially for the

358

BW30 membrane (Figure 6A), and no significant change was observed as a function of pH. The

359

relatively high normalized flux (≈ 60%) was attributed to the lower osmotic pressure difference

360

between the bulk and permeate side of the membrane (Table 1). As a result of the relatively low

361

osmotic pressure, the net driving force of the Mdori water was higher than that of the Magadi water

362

(see Fig. SI 4). This was less pronounced for the NF270 due to the lower EC retention of this

363

membrane.

AC C

EP

TE D

M AN U

SC

RI PT

357

18

ACCEPTED MANUSCRIPT B

100 80

60

60

Degassing

40

40

20

20 0 100

D

C

80

60

60

40

40

0 100

4

2

0 2

4

6

8

10

12

2

4

6

8

10

80 60 40 20 0

12

pH

TE D

pH

M AN U

F

TOC retention (%)

0 E 6

SC

20

−1

−1

SUVA (L mg m )

EC retention (%)

80

20

364

IC retention (%)

BW30 NF270

80

0 100 F retention (%)

A

RI PT

J/JW0 (%)

100

Figure 6. (A) Normalized flux (J/JW0), (B) IC retention, (C) F retention, (D) EC retention,

366

(E) SUVA of permeates, (F) TOC retention as a function of pH at 50% recovery of the

367

Mdori water (pressure 9.8 ± 0.1 bar and temperature 20 ± 2 oC)

368

EP

365

IC and F retention (Figure 6B and C) of the Mdori were higher than that of the semi-natural

370

Magadi waters (Figure 5). At pH >4, F retention of ≈80% and >95% was achieved for NF270 and

371

BW30, respectively. Higher IC and F retention in comparison with that of the Magadi water can be

372

attributed to the relatively lower ionic strength and hence both a higher driving force and reduced

373

charge screening. EC retention increased with pH. The increase in EC with pH was explained with

374

an increase in membrane surface charge. Surface charge affects solute partitioning into the active

375

layer and in consequence an increase in rejection of ions through Donnan electrostatic exclusion

376

(Coronell et al., 2013; Wang et al., 2017). UVA values of both BW30 and NF270 permeates were <

AC C

369

19

ACCEPTED MANUSCRIPT

2 L mg C−1 m−1, which is an indication of low aromatic NOM content (Environmental Protection

378

Agency Ireland, 2012). Retention of NOM was greater than 95% over the entire pH range, which

379

confirms previous results (Shen and Schäfer, 2015), even though it must be noted that the NOM of

380

this water was very low.

381

3.5. NF/RO of natural surface water with high NOM and fluoride content (Ngare Nanyuki)

RI PT

377

The Ngare Nanyuki water was chosen to investigate the influence of high NOM content on

383

F removal. Given the fact that the ionic strength and IC content are comparable to those of the

384

Mdori water, this makes for a very interesting comparison.

SC

382

Figure 7A indicates a decrease in flux towards lower pH values. The decline in flux at low

386

pH (2−4) can be attributed to the increase in deposition of NOM molecules due to the reduction of

387

electrostatic repulsion between the membrane surface and NOM (Hong and Elimelech, 1997). Thus

388

at low pH, the membrane surface charge is no longer positively charged (see Fig. SI 5) and NOM

389

molecules are protonated. Further, NOM molecules have a more compact configuration at acidic

390

pH due to reduced charge density (Ghosh and Schnitzer, 1980), which results in increased

391

deposition.

TE D

M AN U

385

IC retention (Figure 7B) was similar to that of the Mdori waters (Figure 6), following the

393

trend of IC speciation as discussed in section 3.4. Fluoride retention increased with pH (Figure 7C).

394

This follows the increase in membrane charge and change in F speciation with pH. F retention by

395

BW30 increased sharply from 42% at pH 2 to 90% at pH 4. Thus F retention increased sharply

396

after a change in F speciation from HF to F− (pKa of HF is 3.2). However, for NF270, F retention

397

was about 20% at pH 2−4 and increased to 80% at pH 6. F retention for the Ngare Nanyuki waters

398

was significantly higher than that of the high ionic strength semi-natural Magadi waters, while only

399

a slight increase at higher pH when compared with Mdori (see Table 2) could be observed. EC

400

retention increased with pH due to the increase in the surface charge of the membrane (Figure 7D).

AC C

EP

392

20

ACCEPTED MANUSCRIPT BW30 NF270

B

100

60

60

Degassing

40

40

20

20 C

0 100

D

80

80

60

60

40

40

E

F

6

4

2

0 2

4

6

401

pH

8

10

M AN U

−1

−1

SUVA (L mg m )

0

20

SC

20

EC retention (%)

0 100 F retention (%)

80

RI PT

J/JW0 (%)

80

IC retention (%)

A

12

2

4

6

pH

8

0 100 80 60 40 20

TOC retention (%)

100

0 10

12

Figure 7. (A) Normalized flux (J/JW0), (B) IC retention, (C) fluoride retention, (D) EC

403

retention, (E) SUVA, (F) TOC retention as function of pH at a recovery of 50% of the

404

Ngare Nanyuki water (pressure 9.8 ± 0.1 bar and temperature 20 ± 2 oC)

EP

405

TE D

402

Based on previous observations it was expected that the high NOM Ngare Nanyuki water

407

has a higher F retention than the low NOM Mdori water. This is due to an increase in charge of

408

membrane surface by NOM (Shen and Schäfer, 2015). However, a previous study using synthetic

409

water has shown that F retention enhancement by NOM can be annulled by a high concentration of

410

NOM due to membrane deposit formation (Owusu-Agyeman et al., 2017). In view of this, there

411

was indeed no significant difference between F retention for Mdori and Ngare Nanyuki waters,

412

with the exception of NF270 where an increase of 4-5% was consistent (Table 2).

AC C

406

21

ACCEPTED MANUSCRIPT

Table 2. Fluoride retention as a function of pH for all waters and membranes used pH

Fluoride retention (%) Magadi

Mdori

BW30

10

95.1± 4

8.5 (Original)

BW30

NF270

98.9 ± 4

80.6 ± 3

98.8± 4

85.4 ± 4

64.6 ± 3

98.7 ± 4

87.2± 4

98.5± 4

91.0 ± 4

88.8 ± 4

71.2 ± 3

98.8 ± 4

85.8± 4

98.3± 4

89.6 ± 4

8

90.0 ± 4

73.3 ± 3

98. ± 4

86.9± 4

98.8± 4

91.0 ± 4

7

88.3 ± 4

78.3 ± 3

98.2 ± 4

89.1± 4

98.7± 4

87.7 ± 4

6

63.8 ± 3

59.7 ± 3

96.9 ± 4

81.0 ± 3

96.4± 4

85.8 ± 4

4

72.3 ± 5

8.7 ± 1

92.3 ± 6

20.0 ± 1

90.8± 6

29.6 ± 2

2

13.2± 1

12.9 ± 1

60.0± 6

47.5± 5

42.3± 4

29.6 ± 3

a

32.7 ± 1

NF270

pH 11

415

RI PT

93.2 ± 4

BW30 a

M AN U

414

12

NF270 a

Ngare Nanyuki

SC

413

Permeate SUVA values by both NF270 and BW30 were below 2 L mg−1 m−1 except for low

417

pH 2 and 4. This shows that both membranes retained the aromatic content of NOM better at pH >

418

4, while the somewhat higher SUVA values at pH 2 and 4 cannot be explained. TOC retention was

419

above 99 % over the entire pH range. The higher TOC retention observed for Ngare Nanyuki (high

420

NOM) waters is attributed to the fact that size exclusion is the main NOM retention mechanism

421

(Shen and Schäfer, 2015). Thus increase in feed NOM concentration did not significantly change

422

NOM concentration in the permeate (≈ 1 mg C L−1 as TOC) and hence retention remained high.

423

3.6. Characteristics of NOM in NF/RO permeate as a function of pH

AC C

EP

TE D

416

424

The NOM in the membrane permeates and original Ngare Nanyuki feed water were further

425

analysed using LC-OCD. This was done in order to understand the characteristics of the NOM

426

permeating the NF/RO membranes. Figure 8 and Figure 9 show the results of the LC-OCD analysis

427

of the original feed water as well as BW30 and NF270 permeates at different pH values.

22

BW30

ACCEPTED MANUSCRIPT LMWA LMWN BB HS BP TOC

−1

DOC (mg C L )

6 A 4

NF270

B

2 0 2

4

6

pH

8

10

12

2

4

6

pH

8

10

12

RI PT

428

Figure 8. Characteristics of (A) BW30 and (B) NF270 permeate NOM of Ngare Nanyuki

430

water as a function of pH (LMWA: LMW acids, LMWN: LMW neutrals, BB: building

431

blocks, HS: humic substances, BP: biopolymers measured with the LC-OCD and TOC:

432

TOC measured with Sievers M9 analyser)

SC

429

The main fraction of NOM in permeates of both membranes was LMW neutrals (Figure 8).

434

The LMW neutrals concentration of >4 mg C L−1 was recorded for the BW30 and NF270

435

permeates at pH 2−4 and pH 2, respectively. However, these values were not confirmed by TOC

436

analysis with Sievers M9 TOC analyser that measured TOC concentrations of ≈ 1 mg C L−1 over

437

the entire pH range (see Figure 8). This is a very curious discrepancy that triggered significant

438

further investigations and ongoing discussions.

TE D

M AN U

433

There is a clear indication from Figure 8 and Figure 9 that both BW30 and NF270

440

membranes show selectivity towards certain fractions of NOM. HS, building blocks, LMW acids,

441

and biopolymers are almost completely rejected by both NF270 and BW30 membranes over the

442

entire pH range. This is due to the fact these fractions consist of large and/or charged molecules

443

that can be rejected through either size exclusion or charge repulsion (Meylan et al., 2007). On the

444

other hand, LMW neutrals are smaller in size and uncharged, and hence could not be rejected by

445

charge interaction and steric hindrance. LMW neutrals are rarely found in natural waters and were

446

not expected in the feed and permeate. It was suspected (after thorough exclusion of the possibility

447

of sample contamination) that the LMW neutrals in the feed could be as a result of pesticides and

448

other agrochemicals used by farmers in the catchment of the Ngare Nanyuki River (Kihampa et al.,

449

2010). Pesticide analysis of the feed and one permeate (NF270, pH 2) showed that there are indeed

AC C

EP

439

23

ACCEPTED MANUSCRIPT

small concentrations of pesticides in µg/L range in the feed water which also permeated the

451

membrane, notable among them were Carbendazim, DEET, Diuron, Fenuron, and Monuron (Table

452

SI 3). However, the concentration of the pesticides analysed was orders of magnitude lower than

453

the LMW fraction concentrations. The possibility that the LMW neutrals in the feed could be as

454

results of contamination from agricultural activities in the catchment of the river remains. LMW

455

neutrals have been found in NF permeates by other studies and has been attributed to leaching from

456

the membrane and natural components of the NOM (Schäfer et al., 2004). Likewise, LMW neutral

457

peaks in natural waters were reported to permeate NF/RO membranes which was explained by the

458

uncharged nature and smaller size of LMW neutrals (Meylan et al., 2007). Furthermore, Figure 8

459

shows that the amount of LMW neutrals in permeate of BW30 was lower than NF270 permeates at

460

pH ≥ 6 due to fact that the molecular weight cut off of BW30 (98−100 Da) (Boleda et al., 2010;

461

Richards, 2012) is smaller than that of the NF270 membrane (155−180) (Boussu et al., 2006;

462

Richards, 2012). The LMW neutrals signals were higher at low pH 2−4 for both membranes and

463

the higher permeability of LMW neutrals at the acidic pH range could not be explained. Other

464

studies have recorded low retention of LMW organics by NF/RO membranes at low pH (below

465

pka) and attributed the low retention to change of structure of such solutes (Ozaki and Li, 2002;

466

Bellona and Drewes, 2005). As the compounds could not be identified, such claims remain

467

speculative and further investigation is required to identify processes such as photobleaching and

468

microbial degradation that possibly produce such LMW compounds in such tropical water. A

469

further field requiring more in depth study is the consequences of permeation of such organics that

470

are very biodegradable. In a membrane application this would no doubt contribute to permeate side

471

biofouling which is observed on occasion, where the ‘nutrients’ are delivered through the

472

membrane to the permeate channel.

AC C

EP

TE D

M AN U

SC

RI PT

450

473

Both the OCD and OND showed signals for the LMW neutrals (see Figure 9). This suggests

474

that LMW neutrals that permeated the NF/RO membranes included nitrogen bound organic

24

ACCEPTED MANUSCRIPT

compounds. However, the UVD of the LC-OCD did not show any signal for the LMW neutrals.

476

LMW neutrals are hydrophilic to amphiphilic and show a low or little response in UVD due to their

477

low UV absorptivity (Leenheer and Croué, 2003; Huber et al., 2011).

AC C

EP

TE D

M AN U

SC

RI PT

475

25

RI PT

LMW neutrals

SC

Relative signal

pH 12 pH 10

M AN U

pH 8.5 pH 8 pH 7 pH 6

TE D

pH 2 C

OCD OND

Permeates - BW30

pH 4

Permeates- NF270

pH 12

EP

pH 10 pH 8.5 pH 8 pH 7

AC C

Relative signal

LMW acids and HS

Biopolymer

B

Building blocks

HS

ACCEPTED MANUSCRIPT Feed UVD

Relative signal

A

pH 6 pH 4 pH 2

0 478

20 40 60 80 Retention time (min)

100

479

Figure 9. LC-OCD chromatograms of Ngare Nanyuki water (A) Feed and (B) BW30 and

480

(C) NF270 permeates as a function of pH

481 26

482

ACCEPTED MANUSCRIPT

3.7. Characteristics of HS in NF/RO permeates

The aromaticity and molecular weight of HS in the NF/RO permeate was plotted on the

484

humification diagram in order to compare the permeate HS with the feed and IHSS standards (see

485

Figure 10). It was observed that at low pH 2 and 4, the aromaticity of HS in the permeate was high

486

(13−20 L mg−1 m−1) for both membranes. Both the aromaticity and Mn of permeate HS for BW30

487

were lower than that of the NF270 at pH >6 (Figure 10A). At pH >6, the molecular weights of HS

488

that permeated the BW30 were 60−85% less than that of the feed, while those of the NF270 were

489

only 30−50% less. The lower Mn and aromaticity of BW30 than NF270 was attributed to the fact

490

that BW30 have smaller MWCO than NF270 (Cho et al., 2000). In consequence, the equivalent

491

radius of HS that permeated the BW30 membrane was lower than that of the NF270 (Figure 10B).

492

The estimated HS radius (>0.44 nm) shows that the permeating HS molecules were larger than the

493

reported pore radius of both membranes (0.30−32 nm for BW30 and 0.38−0.42 nm for NF270)

494

(Hilal et al., 2005; Richards, 2012; Simon et al., 2013; Dražević et al., 2014). This suggests that the

495

mode of transport of HS molecules through the membrane was through diffusion and charge

496

interaction. However, this suggestion is made with caution as the pore sizes are average effective

497

values and retention/transport is likely to be dominated by a small number of larger pores.

TE D

Low pH

16 12

498

BW30 NF270 Feed

1.0 0.8 0.6

Feed

NF270

8

B

0.4

IHSS-HA

BW30

IHSS-FA

4

0.2

Radius of HS (nm)

NF270 BW30 pH 2 pH 2 pH 4 pH 4 pH 6 pH 6 pH 7 pH 7 pH 8 pH 8 pH 8.5 pH 8.5 pH 10 pH 10 pH 12 pH 12

EP

20 A

AC C

−1

−1

SAC/DOC (HS) (L mg m )

M AN U

SC

RI PT

483

Marine-FA

0.0

0

400 600 800 1000 1200 1400 -1 Mn (gmol )

2

4

6

8

10

12

pH

499

Figure 10. Characteristics of permeate NOM of Ngare Nanyuki water (A) humification

500

diagram (B) molecular radius of HS of the permeate NOM as a function of pH

501

27

502

ACCEPTED MANUSCRIPT

4. Conclusions

Examining real and semi-natural tropical waters as a function of pH with an open (NF270)

504

and a dense (BW30) membrane allowed to elucidate the contribution of numerous factors to F and

505

NOM removal. Permeation of NOM fractions and specifically LMW neutrals was examined in

506

detail and a large contribution of difficult to detect LMW neutrals was identified.

RI PT

503

F retention by NF270 particularly for the semi natural waters with high EC and high IC

508

concentration was generally low (60−70%) and further decreased at high pH (down to 20%). The

509

decrease was attributed to co-ions (F− and CO32−) interaction and synergetic effect of high salt

510

concentration (low driving force) and pH. Relatively higher F retention was achieved for the waters

511

with lower EC and IC (Mdori and Ngare Nanyuki). Although high NOM concentration natural

512

water (Ngare Nanyuki) resulted in higher zeta potential in comparison with low NOM water

513

(Mdori), there was no significant difference between F retention with the exception of the loose

514

membrane were a small increase was observed. This was probably due to the fact that F retention

515

enhancement was annulled by deposit formation on the membrane.

TE D

M AN U

SC

507

About 60−90% of NOM fractions that permeated the NF/RO membranes were LWM

517

neutrals. There was higher permeability of LMW neutrals at low pH range 2−4 and was

518

unexplained as the specific compounds could not be identified. The HS of NOM that permeated

519

both NF/RO membranes at acidic pH (2−4) were more aromatic in nature. The fractions of HS that

520

permeated the BW30 membrane at pH ≥6, are of lower molecularity and aromaticity than those that

521

permeated the NF270 due to larger MWCO of NF270 than BW30. The permeation of

522

biodegradable organics poses challenges of controlling permeate side biofouling in full scale

523

application.

524

5. Acknowledgements

AC C

EP

516

525

The Deutscher Akademischer Austauschdienst (DAAD) is thanked for a Ph.D. stipend for

526

I.O.-A. Funding for IFG-MT was provided by the Helmholtz Association, Germany through the 28

ACCEPTED MANUSCRIPT

Rekrutierungsinitiative. The DOW Chemical Company (USA) kindly supplied the membranes

528

samples. Dr. Thomas Luxbacher and Sandra Zierler, Anton Paar GmbH, Austria performed the

529

streaming potential measurements. Elijah Paul and Lwitiko Pholds are thanked for sampling and

530

transporting Mdori water samples to KIT respectively. Marita Heinle (IFG-KIT) is thanked for

531

technical support with the TOC Analyzer and ICP-OES analysis. Dr. Stefan Huber (DOC-Labor,

532

Germany) performed the LC-OCD analysis and contributed extensive discussions. Dr. Frank

533

Schramm (IFG-KIT) is thanked for technical support with UV-Vis spectroscopy. Authors are

534

grateful to Prof. Dr. Heinz-Jürgen Brauch of the Technologiezentrum Wasser (TZW, Karlsruhe) for

535

pesticide analysis and discussion.

536

6. References

537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565

African Development Bank, 2010. Environmental and social management plan (ESMP) for Tanzania- rural water supply and sanitation programme (RWSSP) II [online]. Final Report, Tanzania (viewed on 17/01/2017) https://www.afdb.org/fileadmin/uploads/afdb/Documents/Project-andOperations/Tanzania%20%20Rural%20Water%20Supply%20and%20Sanitation%20Program%20II%20_AR_%20doc %20%2BMemo%5B1%5D.pdf. Aschermann, G., Jeihanipour, A., Shen, J., Mkongo, G., Dramas, L., Croue, J.-P., Schäfer, A., 2016. Seasonal variation of organic matter concentration and characteristics in the Maji ya Chai River (Tanzania): Impact on treatability by ultrafiltration. Water Research 101, 370381. Ayoob, S., Gupta, A.K., Bhat, V.T., 2008. A conceptual overview on sustainable technologies for the defluoridation of drinking water. Critical Reviews in Environmental Science and Technology 38, 401-470. 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. Separation Science and Technology 49, 1135-1145. Bellona, C., Drewes, J.E., 2005. The role of membrane surface charge and solute physico-chemical properties in the rejection of organic acids by NF membranes. Journal of Membrane Science 249, 227-234. Boleda, M.R., Majamaa, K., Aerts, P., Gómez, V., Galceran, M.T., Ventura, F., 2010. Removal of drugs of abuse from municipal wastewater using reverse osmosis membranes. Desalination and Water Treatment 21, 122-130. 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. Journal of Membrane Science 278, 418-427. Brinkmann, T., Sartorius, D., Frimmel, F.H., 2003. Photobleaching of humic rich dissolved organic matter. Aquatic Sciences 65, 415-424.

AC C

EP

TE D

M AN U

SC

RI PT

527

29

ACCEPTED MANUSCRIPT

EP

TE D

M AN U

SC

RI PT

Childress, A.E., Deshmukh, S.S., 1998. Effect of humic substances and anionic surfactants on the surface charge and performance of reverse osmosis membranes. Desalination 118, 167-174. Childress, A.E., Elimelech, M., 1996. Effect of solution chemistry on the surface charge of polymeric reverse osmosis and nanofiltration membranes. Journal of Membrane Science 119, 253-268. Cho, J., Amy, G., Pellegrino, J., 2000. Membrane filtration of natural organic matter: Factors and mechanisms affecting rejection and flux decline with charged ultrafiltration (UF) membrane. Journal of Membrane Science 164, 89-110. Coronell, O., Mi, B., Mariñas, B.J., Cahill, D.G., 2013. Modeling the effect of charge density in the active layers of reverse osmosis and nanofiltration membranes on the rejection of arsenic(III) and potassium iodide. Environmental Science & Technology 47, 420-428. Dražević, E., Košutić, K., Freger, V., 2014. Permeability and selectivity of reverse osmosis membranes: Correlation to swelling revisited. Water Research 49, 444-452. Environmental Protection Agency Ireland, 2012. EPA drinking water guidance on disinfection byproducts. Advice Note 4 Fan, L., Nguyen, T., Roddick, F.A., Harris, J.L., 2008. Low-pressure membrane filtration of secondary effluent in water reuse: Pre-treatment for fouling reduction. Journal of Membrane Science 320, 135-142. Fawell, J., Bailey, K., Chilton, J., Dahi, E., Fewtrell, L., Magara, Y., 2006. Fluoride in drinking water. World Health Organization, IWA Publishing London. Fawell, J., Bailey, K., Chilton, J., Dahi, E., Fewtrell, L., Magara, Y., 2013. Fluoride in drinkingwater. Water Intelligence Online 12. Fewtrell, L., Smith, S., Kay, D., Bartram, J., 2006. An attempt to estimate the global burden of disease due to fluoride in drinking water. Journal of Water and Health 4, 533-542. Filella, M., 2009. Freshwaters: which NOM matters? Environmental Chemistry Letters 7, 21-35. Findlay, S.E.G., Sinsabaugh, R.L., 2003. Aquatic ecosystems: interactivity of dissolved organic matter. Elsevier Science California,. Freger, V., 2004. Swelling and morphology of the skin layer of polyamide composite membranes:  An atomic force microscopy study. Environmental Science & Technology 38, 3168-3175. Frencken, J., 1990. Endemic fluorosis in developing countries: causes, effects and possible solutions. TNO Institute for Preventive Health Care, Leiden, The Netherlands. Ghosh, K., Schnitzer, M., 1980. Macromolecular structures of humic substances. Soil Science 129, 266-276. Gizaw, B., 1996. The origin of high bicarbonate and fluoride concentrations in waters of the Main Ethiopian Rift Valley, East African Rift system. Journal of African Earth Sciences 22, 391402. Government of Tanzania, 2007. Environmental management (water quality standards) regulations. Environmental Management Act (CAP 191), Dar es Salaam. (viewed on 17/01/2017) http://pim.maji.go.tz/tcomponents/regulation/CW0103a02/CW0103a02r03.pdf. Hilal, N., Al-Zoubi, H., Darwish, N.A., Mohammad, A.W., 2005. Characterisation of nanofiltration membranes using atomic force microscopy. Desalination 177, 187-199. Hong, S., Elimelech, M., 1997. Chemical and physical aspects of natural organic matter (NOM) fouling of nanofiltration membranes. Journal of Membrane Science 132, 159-181. Hu, K., Dickson, J.M., 2006. Nanofiltration membrane performance on fluoride removal from water. Journal of Membrane Science 279, 529-538. Huber, S.A., Balz, A., Abert, M., Pronk, W., 2011. Characterisation of aquatic humic and nonhumic matter with size-exclusion chromatography – organic carbon detection – organic nitrogen detection (LC-OCD-OND). Water Research 45, 879-885. Istituto Oikos, 2011. The Mount Meru challenge- Integrating conservation and development in Northern Tanzania. Ancora Libri, Milano (Italy). (viewed on 17/01/2017) http://www.istituto-

AC C

566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616

30

ACCEPTED MANUSCRIPT

EP

TE D

M AN U

SC

RI PT

oikos.org/files/download/2012/TheMountMeruChallenge.Integratingconservationanddevelop mentinNorthernTanzania.pdf. Johnson, M.S., Couto, E.G., Abdo, M., Lehmann, J., 2011. Fluorescence index as an indicator of dissolved organic carbon quality in hydrologic flowpaths of forested tropical watersheds. Biogeochemistry 105, 149-157. Kaseva, M.E., 2006. Contribution of trona (magadi) into excessive fluorosis-a case study in Maji ya Chai ward, northern Tanzania. Science of The Total Environment 366, 92-100. Khan, M.M., Stewart, P.S., Moll, D.J., Mickols, W.E., Nelson, S.E., Camper, A.K., 2011. Characterization and effect of biofouling on polyamide reverse osmosis and nanofiltration membrane surfaces. Biofouling 27, 173-183. Kihampa, C., Mato, R.R., Mohamed, H., 2010. Residues of organochlorinated pesticides in soil from tomato fields, Ngarenanyuki, Tanzania. Journal of Applied Sciences and Environmental Management 14. Leenheer, J.A., Croué, J.-P., 2003. Characterizing aquatic dissolved organic matter. Environmental Science & Technology 37, 18-26. Lou, T., Xie, H., 2006. Photochemical alteration of the molecular weight of dissolved organic matter. Chemosphere 65, 2333-2342. Luo, J., Wan, Y., 2013. Effects of pH and salt on nanofiltration - A critical review. Journal of Membrane Science 438, 18-28. Mänttäri, M., Pihlajamäki, A., Nyström, M., 2006. Effect of pH on hydrophilicity and charge and their effect on the filtration efficiency of NF membranes at different pH. Journal of Membrane Science 280, 311-320. Meylan, S., Hammes, F., Traber, J., Salhi, E., von Gunten, U., Pronk, W., 2007. Permeability of low molecular weight organics through nanofiltration membranes. Water Research 41, 39683976. Nanyaro, J.T., Aswathanarayana, U., Mungure, J.S., Lahermo, P.W., 1984. A geochemical model for the abnormal fluoride concentrations in waters in parts of northern Tanzania. Journal of African Earth Sciences 2, 129-140. Nielsen, J.M., 1999. East African magadi (trona): fluoride concentration and mineralogical composition. Journal of African Earth Sciences 29, 423-428. Nilsson, M., Trägårdh, G., Östergren, K., 2008. The influence of pH, salt and temperature on nanofiltration performance. Journal of Membrane Science 312, 97-106. Oliveira, J.L., Boroski, M., Azevedo, J.C.R., Nozaki, J., 2006. Spectroscopic investigation of humic substances in a tropical lake during a complete hydrological cycle. Acta hydrochimica et hydrobiologica 34, 608-617. Owusu-Agyeman, I., Jeihanipour, A., Luxbacher, T., Schäfer, A.I., 2017. Implications of humic acid, inorganic carbon and speciation on fluoride retention mechanisms in nanofiltration and reverse osmosis. Journal of Membrane Science 528, 82-94. Owusu-Agyeman, I., Shen, J., Schäfer, A.I., 2018. Renewable energy powered membrane technology: Impact of pH and ionic strength on fluoride and natural organic matter removal. Science of The Total Environment 621, 138-147. Ozaki, H., Li, H., 2002. Rejection of organic compounds by ultra-low pressure reverse osmosis membrane. Water Research 36, 123-130. Rafique, T., Naseem, S., Ozsvath, D., Hussain, R., Bhanger, M.I., Usmani, T.H., 2015. Geochemical controls of high Fluoride groundwater in Umarkot Sub-District, Thar Desert, Pakistan. Science of the Total Environment 530-531, 271-278. Richards, L.A., 2012. The removal of inorganic contaminants using nanofiltration and reverse osmosis. School of Engineering and Physical Sciences, Heriot-Watt University, Edinburgh PhD.

AC C

617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665

31

ACCEPTED MANUSCRIPT

EP

TE D

M AN U

SC

RI PT

Richards, L.A., Richards, B.S., Corry, B., Schäfer, A.I., 2013. Experimental energy barriers to anions transporting through nanofiltration membranes. Environmental Science & Technology 47, 1968-1976. Richards, L.A., Richards, B.S., Rossiter, H.M.A., Schäfer, A.I., 2009. Impact of speciation on fluoride, arsenic and magnesium retention by nanofiltration/reverse osmosis in remote Australian communities. Desalination 248, 177-183. Rossiter, H.M., Owusu, P.A., Awuah, E., MacDonald, A.M., Schäfer, A.I., 2010. Chemical drinking water quality in Ghana: Water costs and scope for advanced treatment. Science of the Total Environment 408, 2378-2386. 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. Journal of Membrane Science 242, 73-85. Shen, J., Gagliardi, S., McCoustra, M.R.S., Arrighi, V., 2016. Effect of humic substances aggregation on the determination of fluoride in water using an ion selective electrode. Chemosphere 159, 66-71. Shen, J., Schäfer, A.I., 2015. Factors affecting fluoride and natural organic matter (NOM) removal from natural waters in Tanzania by nanofiltration/reverse osmosis. Science of The Total Environment 527–528, 520-529. Simon, A., Price, W.E., Nghiem, L.D., 2013. Influence of formulated chemical cleaning reagents on the surface properties and separation efficiency of nanofiltration membranes. Journal of Membrane Science 432, 73-82. Simpson, A., Kerr, C., Buckley, C., 1987. The effect of pH on the nanofiltration of the carbonate system in solution. Desalination 64, 305-319. Smedley, P.L., Nkotagu, H., Pelig-Ba, K., MacDonald, A.M., Tyler-Whittle, R., Whitehead, E.J., Kinniburgh, D., 2002. Fluoride in groundwater from high-fluoride areas of Ghana and Tanzania. British Geological Survey, Nottingham. Song, G., Li, Y., Hu, S., Li, G., Zhao, R., Sun, X., Xie, H., 2017. Photobleaching of chromophoric dissolved organic matter (CDOM) in the Yangtze River estuary: kinetics and effects of temperature, pH, and salinity. Environmental Science: Processes & Impacts 19, 861-873. Song, H., Shao, J., He, Y., Hou, J., Chao, W., 2011. Natural organic matter removal and flux decline with charged ultrafiltration and nanofiltration membranes. Journal of Membrane Science 376, 179-187. Teixeira, M.R., Rosa, M.J., Nyström, M., 2005. The role of membrane charge on nanofiltration performance. Journal of Membrane Science 265, 160-166. Verliefde, A.R.D., Cornelissen, E.R., Heijman, S.G.J., Verberk, J.Q.J.C., Amy, G.L., Van der Bruggen, B., van Dijk, J.C., 2008. The role of electrostatic interactions on the rejection of organic solutes in aqueous solutions with nanofiltration. Journal of Membrane Science 322, 52-66. Wang, J., Kingsbury, R.S., Perry, L.A., Coronell, O., 2017. Partitioning of alkali metal salts and boric acid from aqueous phase into the polyamide active layers of reverse osmosis membranes. Environmental Science & Technology 51, 2295-2303. Worch, E., 1993. Eine neue Gleichung zur Berechnung von Diffusionskoeffizienten gelöster Stoffe. Vom Wasser 81. Yaroshchuk, A., Martínez-Lladó, X., Llenas, L., Rovira, M., de Pablo, J., 2011. Solution-diffusionfilm model for the description of pressure-driven trans-membrane transfer of electrolyte mixtures: One dominant salt and trace ions. Journal of Membrane Science 368, 192-201. Zaidi, S.M.J., Fadhillah, F., Khan, Z., Ismail, A.F., 2015. Salt and water transport in reverse osmosis thin film composite seawater desalination membranes. Desalination 368, 202-213.

AC C

666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714

32

ACCEPTED MANUSCRIPT Highlights • Real tropical waters with different IS, IC, NOM but similar F content treated by NF/RO • At high IS & pH, F retention decreases due to IC speciation, pH and salt effect • No enhancement effect of high NOM on F retention was observed • A discrepancy between permeate NOM measured by TOC analyser and LC−OCD was

RI PT

observed

AC C

EP

TE D

M AN U

SC

• According to LC−OCD results, LMW neutrals accounted for 60−90% of permeate NOM