Electrodialysis and Nanofiltration of Surface Water for Subsequent Use

ARTICLE IN PRESS

Water Research 37 (2003) 3867–3874

Electrodialysis and nanoﬁltration of surface water for subsequent use as inﬁltration water B. Van der Bruggena,*, R. Milisa, C. Vandecasteelea, P. Bielenb, E. Van Sanb, K. Huysmanb

a Laboratory for Environmental Technology, Department of Chemical Engineering, University of Leuven, W. de Croylaan 46, Heverlee B 3001, Belgium b Pidpa Water Production—Puriﬁcation and Pilot Plant Department, Desguinlei 246, Antwerpen B 2018, Belgium

Received 18 September 2002; accepted 6 May 2003

Abstract

In order to achieve stable groundwater levels, an equilibrium between the use of groundwater for drinking water production and natural or artificial groundwater recharge by infiltration is needed. Local governments usually require that the composition of the water used for artificial recharge is similar to the surface water that is naturally present in the specific recharge area. In this paper, electrodialysis (ED) and nanofiltration were evaluated as possible treatment technologies for surface water from a canal in Flanders, the North of Belgium, in view of infiltration at critical places on heathlands. Both methods were evaluated on the basis of a comparison between the water composition after treatment and the composition of local surface waters. The treatment generally consists of a tuning of pH and the removal of contaminants originating from industrial and agricultural activity, e.g., nitrates and pesticides. Further evaluation of the influence of the composition of the water on the characteristics of the artificial recharge, however, was not 2 envisaged. In a case study of water from the canal Schoten-Dessel, satisfactory concentration reductions of Cl ,SO4 ; + 2+ + 2+ NO3 ; HCO3 ; Na ,Mg ,K and Ca were obtained by ultrafiltration pretreatment followed by ED. Nanofiltration with UTC-20, N30F, Desal 51 HL, UTC-60 and Desal 5 DL membranes resulted in an insufficient removal level, especially for the monovalent ions. r 2003 Elsevier Ltd. All rights reserved.

Keywords: Electrodialysis; Nanoﬁltration; Membrane treatment; Surface water; Groundwater recharge

1. Introduction million m3/year, of which 59.4% (171 million m3) was effectively used in 1998 [1]. The effect of intensive use of At present, approximately 50% of the water volume groundwater resources (on a relatively small surface produced by the Flemish drinking water companies area of 13522 km2) on a longer time scale may eventually originates from deep groundwater layers, corresponding lead to depletion of groundwater layers in certain to a volume of 170 million m3/year in 1998. The Flemish regions. As a rule of thumb, water management has to industry is allowed to extract an additional 288 be considered an important element in the national economy when the use of water exceeds 30% of the water reserves; Flanders falls into this category. There- *Corresponding author. Tel.: +32-16-32-23-40; fax: +32-16- fore, the Flemish Government promotes the use of 32-29-91. surface water for industrial purposes and the saving of E-mail addresses: [email protected] groundwater resources for activities where the use of (B. Van der Bruggen), [email protected] (E. Van San), groundwater has a high additional value [2]. Currently, [email protected] (K. Huysman). there is a decrease of the fraction of groundwater used

0043-1354/03/$ - see front matter r 2003 Elsevier Ltd. All rights reserved. doi:10.1016/S0043-1354(03)00296-3 ARTICLE IN PRESS 3868 B. Van der Bruggen et al. / Water Research 37 (2003) 3867–3874 for drinking water production (from 53.1% in 1990 to characteristics are not studied; the legal requirement of 48.5% in 2000 [3]), which has prompted a re-evaluation using water with similar composition to the natural of industrial authorizations for the use of groundwater. surface waters is accepted as a practical guideline for The latter policies and a rising taxation on wastewater drinking water companies. However, it is acknowledged discharge resulted in an increased interest of different that further studies on the influence of the water (and industries in water reuse [4–6]. However, in spite of these soil) composition will be useful for the evaluation of the actions, local problems may still occur and other options criteria for a safe sustainable groundwater management are being explored. One of these options is the artificial by using artificial recharge methods. recharge of groundwater layers with alternative water RO can be used to eliminate small organic com- sources. The water source may be treated wastewater [7], pounds, including pharmaceuticals, and ions [18,19].In as is the case in the municipal wastewater reuse project view of protecting the quality of existing groundwater in Tel Aviv, Israel [8], and in the drinking water sources, the composition of the infiltration water should production process of the IWVA in Koksijde, Flanders be as close as possible to the groundwater composition [9]. At IWVA, domestic wastewater is treated by or to the infiltrating surface water that was already microfiltration followed by reverse osmosis (RO) and present, and no additional contaminants should be infiltrated in the dunes. The natural environment ensures introduced. Another aspect of the water composition a quality of the ‘‘new’’ groundwater comparable to that that should be investigated carefully is the danger of of the ‘‘original’’ groundwater in the dunes. Hygienic blocking the infiltration system due to the precipitation aspects were taken into consideration and appear to be of solid products. For example, certain solubility under control. Another well-known project was carried products can be exceeded locally when infiltration water out in California, where potable use of groundwater contacts soil components. from aquifers recharged with treated sewage effluent RO may result in too low permeate concentrations for seemed to be the technological answer to water scarcity. ions. The RO permeate is usually very corrosive and However, the protests against the link between sewage may extract ions from soil minerals. After infiltration, water and drinking water, even after soil infiltration, concentrations increase upon contact with the soil. The prove that this issue is still not cleared out [10]. treated water eventually has the same ‘‘natural’’ quality Conventional groundwater recharge is usually as the groundwater, without causing contamination. achieved by putting surface water in basins, furrows, However, due to interference with geochemical trans- ditches, or other facilities where it infiltrates into the soil formations [20], the composition of the groundwater and moves downward to recharge aquifers [11]. Direct may change due to the impact of groundwater recharge recharge is achieved with injection wells in the aquifer [21]. Moreover, RO requires a large energy input and [11]. A large number of hydrogeological studies were produces a brine that can create disposal problems. carried out to estimate the impact of recharge on the This paper compares two treatment technologies, groundwater level [12–14]. Stable groundwater levels are electrodialysis (ED) and nanofiltration (NF), as possible desirable, and may require the use of complex models, alternatives to RO in a case study where only low especially where the locations of groundwater recharge concentrations of synthetic organics and pesticides are and of groundwater withdrawal are different. However, present. Surface water from a canal was treated by both this paper only describes surface methods prior to methods after an adequate pretreatment; the removal artificial recharge that should provide a water quality levels with both processes were compared to the acceptable for subsequent use in recharge basins on the concentrations found in natural surface waters at the basis of legal requirements; discussions on the risk of location where groundwater recharge is envisaged. contamination if groundwater recharge is carried out with contaminated water can be found elsewhere [15,16]. Drinking water companies working on sustainable 2. Materials and methods groundwater management are usually faced with legal requirements concerning the quality of the water used 2.1. Feed water for recharge: the composition should be similar to the composition of the water naturally occurring in the area. The raw water used in this study originates from the Aspects of controlling groundwater quality during canal Mol-Dessel in Flanders, Belgium. Pidpa, the groundwater recharge are not always fully documented drinking water company in the region, carried out the in literature. A study on the environmental fate of pretreatment of the raw water using a 300 mm prefilter pharmaceutical components [17] indicated a high followed by ultrafiltration (UF). The UF membrane was persistence of pharmaceuticals under aerobic and manufactured by X-Flow, The Netherlands, and had a anaerobic groundwater conditions, with elimination length of 1.5 m, a membrane surface 35 m2; the levels below 20%. In this article, effects of the water capillaries had a diameter of 0.8 mm. FeCl3 at a and soil composition on the groundwater and soil concentration around 3 ppm was used as a coagulant ARTICLE IN PRESS B. Van der Bruggen et al. / Water Research 37 (2003) 3867–3874 3869

Table 1 prevent migration of chloride ions to the anode rinsing Composition of the UF permeate used as feed for electro- solution, which would lead to Cl2 production at the dialysis and nanofiltration anode. The dimensions of the membranes are 2 2 Component Concentration (mg/l) 100 100 mm with an active surface of 58 cm . The distance between two membranes is 0.5 mm. The spacers HCO3 88.3 2+ are made of silicone and polyester. Ca 58.7 The volume of the diluate, the concentrate and the SO2 31.0 4 electrode rinsing is 2.5 l. Sampling was done from the Cl 27.3 Na+ 20.9 recirculation line through each vessel; pH and conductivity were measured continuously. The feed velocity NO3 15.6 Mg2+ 6.3 in the stack ranged from 7.5 to 10 cm/s. The applied K+ 5.2 electric potential ranged from 5 to 20 V (maximal electric potential in the stack is 2 V per membrane or 24 V). The electrode rinsing solution was a synthetic 5 wt% Na2SO4 solution in all experiments. during UF, so that suspended solids are absent in the UF permeate. The UF permeate was then further 2.4. Nanofiltration experiments treated with UV irradiation for disinfection and to avoid biological growth in the permeate. A cross-flow NF setup (Amafilter, Test Rig PSS1TZ) The ED and NF experiments were carried out in with 90 mm flat sheet membranes was used for the parallel with the UF permeate. experiments. The active membrane surface area is The composition of the UF permeate is given in 0.0044 m2; the channel length in the module (TZA 944) Table 1. Chemical oxygen demand and biological is 293 mm. The hydraulic diameter in the feed channel is oxygen demand values were below the detection limit. 4.2 mm. The temperature was 25C and the transmem- Nitrites appeared to be absent; concentrations of iron, brane pressure was 10 bar. The cross-flow velocity was aluminum, manganese, nickel, copper, zinc, lead and maintained at 6 m/s. barium were below 35 mg/l and are not further taken into Five NF membranes were used, with molecular weight consideration. cut-off (MWC) between 150 and 400 Da: UTC-20 (MWCB200) and UTC-60 (MWCB180), manufac- 2.2. Analytical methods tured by Toray Ind. Inc. (Japan); Desal 51 HL (MWC 150-300) and Desal 5 DL (MWC 150-300), manufac- Cations were determined by inductively coupled tured by Osmonics (Vista, CA, USA), and N30F plasma—mass spectrometry using a Plasma Quad (MWCB400), manufactured by Nadir (Wiesbaden, PQ2+ (VG-Elemental). Anions were determined by Germany). capillary zone electrophoresis using a Waters (Milfors, USA) Quanta 4000 system with online UV detection at 254 nm. Conductivity was measured with an Orion 3. Results and discussion Model 160 conductivity meter. Measurement of pH was carried out with an Orion Model 420A pH meter. Concentrations of anions and cations during ED at 20 V are presented in Fig. 1. All ions appear to be 2.3. Electrodialysis experiments removed to the same extent; no significant differences between monovalent and divalent ions were observed. ED experiments were carried out with a Berghof BEL- This was expected given the fact that the ED membranes 500 ED setup. Three separated circuits, each with a used are not selective. Half-life times are given in centrifugal pump, are present for the diluate, for the Table 2; all values are in the same order of magnitude. 2 concentrate and for the electrode rinsing. The membrane SO4 appears to be an exception because it is removed stack is connected to a DC electric potential through more slowly. This may be caused by the fact that TiO2-coated titanium electrodes. The stack consisted of generally lower diffusion coefficients are found for seven Selemions CMV cationic exchange membranes bivalent ions through ion exchange membranes [22]. and five Selemions AMV anionic exchange membranes. The pH evolution in the diluate fraction during the Two consecutive CMV cationic exchange membranes experiment is given in Fig. 2. The starting pH is 7.5; at were used at the cathodic side of the stack to prevent this pH, the CaCO3 concentration is close to saturation. contact between the diluate and the cathode rinsing Because the temperature increases in the experiment, solution (H2 production at the cathode may occur by oversaturation and precipitation of CaCO3 on the reduction) and a single CMV cationic exchange mem- membrane surface may occur. The oversaturation may brane was used at the anodic side of the stack in order to be even more likely when concentration polarization ARTICLE IN PRESS 3870 B. Van der Bruggen et al. / Water Research 37 (2003) 3867–3874

Anions 100 90 SO4 80 NO3 HCO3 70 Cl 60 50 40 30 20 relative concentration (%) 10 0 0 5 10 15 20 25 30 35 40 45 50 time (min)

Cations 100 90 Na 80 Mg Ca 70 K 60 50 40 30 20 relative concentration (%) 10 0 0 5 10 15 20 25 30 35 40 45 50 time (min) Fig. 1. Relative concentrations of anions and cations during ED at 20 V.

Table 2 17.5 min at 5 V. The variation of half-life times for other Half-life time of anions and cations during electrodialysis components is similar (Fig. 3). The ion concentrations in the permeate with the Component Half-life time (min) different nanofiltration membranes are summarized in HCO3 8.0 Table 3. It can be seen that there is a large difference Ca2+ 9.6 between monovalent and divalent ions, which is the 2 SO4 22.0 most pronounced for anions: sulfate is efficiently Cl 8.9 removed, whereas the elimination of chloride, nitrate Na+ 13.8 and bicarbonate is relatively low. This effect is caused by NO3 9.5 Mg2+ 9.7 the assumed negative surface charge of the membranes, K+ 7.5 which results in the formation of a potential (the Donnan potential) at the membrane surface. For multivalent anions, charge interactions are sufficient to obtain a (nearly) complete rejection; for monovalent ions, only a partial rejection can be obtained. Cations are also occurs, causing a local pH increase near the membrane rejected because of their size (the hydrated ions are surface. After a slow linear pH decrease from 7.5 to 6, a generally larger than for monovalent ions), and because fast pH decrease down to 4.5 was observed. At this pH, permeation of cations is coupled to permeation of carbonates are converted to carbon dioxide, which was anions (due to the electroneutrality of the solution). indeed confirmed in the experiments by the generation of Multivalent cations are more efficiently rejected than gas bubbles in the diluate tank. The low pH may have monovalent cations. This leads to relatively low rejec- + + affected the decrease of the HCO3 concentration in the tions for Na and K , and intermediate rejections for right part of the concentration profile (Fig. 1). Ca2+ and Mg2+. The applied electric voltage influenced the rate of ion Differences between the five membranes can be removal in the direction of lower removal rates at lower explained by the differences in membrane characteristics voltages. The half-life time of HCO3 increases from such as surface charge and pore size. Ion size may play a 8.0 min at 20 V to 9.6 min at 15 V, 12.6 min at 10 V and role in the rejection for membranes with low MWC such ARTICLE IN PRESS B. Van der Bruggen et al. / Water Research 37 (2003) 3867–3874 3871

6 pH

3 0 5 10 15 20 25 30 35 40 45 50 time (min) Fig. 2. pH evolution during ED of the UF permeate (20 V).

100 90

80 Ca 70 Na 60 Mg K 50 Cl 40 SO4 30 NO3 Half-life time (min) 20 HCO3 10 0 0 5 10 15 20 25 Electric potential (V) Fig. 3. Half-life times for the removal of anions and cations with ED as a function of the applied electric potential.

Table 3 Ion concentrations obtained in the permeate with a number of nanoﬁltration membranes (feed: UF permeate; pressure: 10 bar; temperature: 25C)

2 2+ 2+ + + Membrane Water Conductivity pH Cl SO4 (ppm) NO3 HCO3 Ca Mg Na K ﬂux (ms/cm) (ppm) (ppm) (ppm) (ppm) (ppm) (ppm) (ppm) (l/m2h)

UF permeate 505 7.50 27.3 31.0 15.6 88.3 58.7 6.3 20.9 5.2 N30F 30.2 293 7.74 23 4 15 54 33.0 3.7 14.5 3.4 Desal 51 HL 181.2 151 7.30 14 o0.5 14 15 13.1 1.1 10.1 2.3 UTC-60 241.3 164 7.53 16 o0.5 14 20 13.6 1.0 11.4 2.7 UTC-20 148.2 119 7.30 7 3 4 23 9.8 1.1 8.7 2.1 Desal 5 DL 94.3 121 7.15 13 o0.5 8 17 7.5 0.4 12.6 2.5

as UTC-60 and UTC-20 [23,24], because an effect of results in a significant Cl rejection. Comparable effects sterical hindrance is added to the charge effect. This can can be seen for the other ions. be seen from the comparison between, e.g., Cl rejection The concentrations in the ED diluate or the nanofil- with N30F and with the four other membranes. N30F is tration permeate should be compared with the water a membrane with a higher MWC than the other quality in the target region for infiltration. The major membranes so that it can be assumed that only charge difficulty for the comparison is that the water quality is interactions are present, leading to a low Cl rejection. different in every sampling point. An overview of three For UTC-60, UTC-20, Desal 5 DL and Desal 51 HL, a sampling points in the basins, two sampling points in combination of charge interactions and size interactions side canals, and the canal itself is given in Table 4. ARTICLE IN PRESS 3872 B. Van der Bruggen et al. / Water Research 37 (2003) 3867–3874

Table 4 Average water quality during the year at six different sampling points

Parameter Basin Side canals Main canal

1 2 3 456

Al (mg/l) 769 467 358 468 958 317–428

NH4 (mg/l) 1.6–4.1 1–1.6 0.6–1.2 0.2 0.3 0.06–0.26 Atrazine (mg/l) — — — — — 0.02–0.09

HCO3 (mg/l) 15 3 26 20.5 25 174 BOD (mg/l) — — — — — 1.4–3.7 Ca (mg/l) 3–7 2 10 11 13 48–63 COD (mg/l) — — — — — 9–13 Cl (mg/l) 5–7 4–8 10–13 13 20 22–54 Diuron (mg/l) — — — — — 0.01–0.2 Conductivity (mS/cm) 60–84 47–50 120–127 130 158 364–490 Pesticides (mg/l) — — — — — 0.06–0.4 Fe (mg/l) 0.96 0.76 0.85 0.78 1.97 0.11–0.75 Isoproturon (mg/l) — — — — — 0.01–0.3 K (mg/l) 0.6–5.3 1.4 7.3–8.3 6 11 4–6 Mg (mg/l) 0.6–1.1 0.6 2.4 2.2 2.1 6.3–8.3 Mn (mg/l) 244 83 188 238 183 11–58 Na (mg/l) 2.4–3.3 2.8 6.1 6.5 8.1 21–31

NO3 (mg/l) 0.5–1.9 0.5–0.7 0.5–2 1.1 3.3 9.4–19 DOC (mg/l) 26.4 19.3 18.8 19.7 23.9 2.8–6.4 Dissolved oxygen (mg/l) 3.3 4.9 9.4 8.7 6.8 7.5–11

P2O5 (mg/l) 76–707 26–71 100–200 120 554 263–585 Simazine (mg/l) — — — — — 0.01–0.04

SO4 (mg/l) 12–14 10 20–23 22 21 35–50 Total hardness (F) 2.3 0.7 3.4 3.7 4 17.3–19.6 PH 4.3–5.4 4.5–4.9 6.1–6.3 6 5.7 7.6–7.9 Suspended solids (mg/l) 9.4 30.4–160 5.9–10.2 8.2 13.8 5.8–21.1

The order of the columns represents the proximity to the canal Mol-Dessel (left column is the most far from the canal).

Concentrations of ions and trace contaminants are Table 5 summarizes the maximum concentrations generally lower when moving away from the canal and allowed in the purified water, together with the to the basins where infiltration into the soil takes place. concentrations obtained via ED after 34 and 43 min Exceptions are the dissolved organic carbon concentra- (times depend on the equipment and are not representa- þ tion and the NH4 concentration, which are higher in the tive for other installations, but the relative concentra- basins because of (residues of) biological activity. tions are correct), and the concentrations obtained with Concentrations of Fe3+,Al3+ and suspended solids the two most performant NF membranes, i.e., UTC-20 are comparable; all other concentrations are lower in the and Desal 5 HL. It can be seen that the removal with basins than in the canal water. The basin water generally nanofiltration is insufficient; the concentrations of has low ion concentrations and a low pH. Pesticides monovalent anions (Cl , HCO3 and NO3 ) are too high (atrazine, diuron, isoproturon, simazine and total for both membranes, and the concentration of all pesticide concentration) are not detectable in the basins cations (Ca2+,Mg2+,Na+,K+) is also too high for and side canals, and are very low in the main canal. This both membranes (with the exception of Mg2+ for Desal is because there are no intense agricultural activities 5 DL). The ED diluate, on the other hand, meets all nearby the canal Mol-Dessel. However, in order to requirements after 43 min. Concentrations of all ions are avoid accumulation of pesticides in the basins, all below the suggested standards, although some concentra- pesticides should be removed from the canal water. This tions (Cl ,HCO3 ) may seem even too low. Therefore, it effect may also be present for heavy metals, but since may be better to use the standards rather as guidelines heavy metals are already present in the basins (although than as maximal allowable concentrations. If the ED in low concentrations), this parameter is not considered process is stopped earlier (34 min in the experiment), 2+ + to be problematic. Therefore, the ion concentrations are concentrations of Ca ,Na and NO3 exceed slightly the most important evaluation parameters for the ED the respective standards, but in general, all concentrations diluate and the NF permeate. come closer to the suggested standards. ARTICLE IN PRESS B. Van der Bruggen et al. / Water Research 37 (2003) 3867–3874 3873

Table 5 Maximal allowable concentration in the puriﬁed water

Parameter Maximal ED diluate. ED diluate. NF-permeate. NF-permeate. concentration 34 min (mg/l) 43 min (mg/l) Desal 5 DL UTC-20 (mg/l) (mg/l) (mg/l)

Cl 5 1.3 0.8 13 7 2 SO4 12 8.7 6.5 o0.5 3 NO3 0.5 0.6 0.4 8 4 HCO3 15 4.7 o2.0 17 23 Na+ 2.4 3.6 2.4 12.6 8.7 Mg2+ 0.6 0.5 0.3 0.4 1.1 K+ 0.6 0.2 0.1 2.5 2.1 Ca2+ 3 4.0 2.1 7.5 9.8

Concentrations obtained in the electrodialysis diluate at two different times and concentrations in the nanoﬁltration permeate for the two most performant membranes UTC-20 and Desal 5 DL.

4. Conclusions [4] Van der Bruggen B, De Vreese I, Vandecasteele C. Water reclamation in the textile industry: nanofiltration of dye The ED diluate can be made fit for soil infiltration baths for wool dyeing. Ind Eng Chem Res through the basins studied. The concentrations in the 2001;40(18):3973–8. basins are seen as guidelines; concentrations exactly [5] Tang C, Chen V. Nanofiltration of textile wastewater for matching with the target concentrations cannot be water reuse. Desalination 2002;143(1):11–20. [6] Mavrov V, Chmiel H, Belieres E. Spent process water obtained for all components because (a) concentrations desalination and organic removal by membranes for water differ from sampling point to sampling point, and (b) reuse in the food industry. Desalination 2001;138(1–3): the removal efficiency in ED is not adjustable for every 65–74. component. Pesticides, not considered in this study [7] Tsagarakis KP, Tsoumanis P, Chartzoulakis K, Angelakis because concentrations in the canal water were low, are AN. Water resources status including wastewater treat- not removed by ED, as they are uncharged. This may be ment and reuse in Greece—related problems and prospec- a disadvantage when the risk of pesticide accumulation tives. Water Int 2001;26(2):252–8. is considered. [8] Kanarek A, Michail M. Groundwater recharge with The nanofiltration permeate is not fit for soil municipal effluent: Dan region reclamation project, Israel. infiltration because of high concentrations of cations Water Sci Technol 1996;34(11):227–33. and monovalent anions. On the other hand, pesticides [9] Dewettinck T, Van Houtte E, Geenens D, Van Hege K, can be removed by NF [25,26]. A further option is to Verstraete W. HACCP (hazard analysis and critical study RO membranes, because these retain monovalent control points) to guarantee safe water reuse and drinking water production—a case study. Water Sci Technol as well as divalent ions, and organic components such as 2001;43(12):31–8. pesticides. However, concentrations in the permeate will [10] Oleszkiewicz JA, Sullivan DE. Water recycling: state-of- then be much lower than the target concentration the-art and some challenges. Proceedings of the Vth obtained from the concentrations in the basins. International Scientific and Technical Conference on Water Supply and Water Quality, Gdansk, Poland, 26– 28 July 2002. Acknowledgements [11] Bouwer H. Artificial recharge of groundwater: hydrogeology and engineering. Hydrogeol J 2002;10(1): This study was financed with a post-doctoral grant 121–42. [12] Chapman TG, Malone RW. Comparison of models for from FWO-Vlaanderen (B. Van der Bruggen). estimation of groundwater recharge, using data from a deep weighing lysimeter. Math Comput Simul 2002; 59(1–3):3–17. References [13] Scanlon BR, Healy RW, Cook PG. Choosing appropriate techniques for quantifying groundwater recharge. Hydro- [1] Vlaams Integraal Wateroverleg Comite! (VIWC), 2002, geol J 2002;10(1):18–39. http://viwc.lin.vlaanderen.be. [14] Sanford W. Recharge and groundwater models: an over- [2] VRIND indicatoren Leefmilieu, 2001, http://aps.vlaander- view. Hydrogeol J 2002;10(1):110–20. en.be. [15] Heberer T. Tracking persistent pharmaceutical residues [3] FRED—metadatabank, Milieu en natuur, 2002, http:// from municipal sewage to drinking water. J. Hydrol aps.vlaanderen.be. 2002;266(3–4):175–89. ARTICLE IN PRESS 3874 B. Van der Bruggen et al. / Water Research 37 (2003) 3867–3874

[16] Bouwer H. Groundwater problems caused by irrigation reclaimed wastewater and recharged groundwater. Water with sewage effluent. J Environ Health 2000;63(3):17–20. Environ Res 1996;68(5):867–76. [17] Preuss G, Willme U, Zullei-Seibert N. Behaviour of some [22] Miyoshi H. Diffusion coefficients of ions through ion- pharmaceuticals during artificial groundwater recharge— exchange membranes for Donnan dialysis using ions of the elimination and effects on microbiology. Acta Hydrochim same valence. Chem Eng Sci 1997;52(7):1087–96. Hydrobiol 2002;29(5):269. [23] Schaep J, Van der Bruggen B, Vandecasteele C, Wilms D. [18] Drewes JE, Fox P, Jekel M. Occurrence of iodinated X-ray Influence of ion size and charge in nanofiltration. Separ contrast media in domestic effluents and their fate during Purification Technol 1998;14:155–62. indirect potable reuse. J Environ Sci Health A—Toxic/ [24] Van der Bruggen B, Schaep J, Wilms D, Vandecasteele C. Hazard Substances Environ Eng 2001;36(9):1633–45. Influence of molecular size, polarity and charge on the [19] Levine B, Reich K, Sheilds P, Suffet IH, Lazarova V. retention of organic molecules by nanofiltration. J Membr Water quality assessment for indirect potable reuse: a new Sci 1999;156:29–41. methodology for controlling trace organic compounds at [25] Kosutic K, Kunst B. Removal of organics from aqueous the west basin water recycling plant (California, USA). solutions by commercial RO and NF membranes of Water Sci Technol 2001;43(10):249–57. characterized porosities. Desalination 2002;142(1): [20] Johnson JS, Baker LA, Fox P. Geochemical transforma- 47–56. tions during artificial groundwater recharge: soil–water [26] Van der Bruggen B, Everaert K, Wilms D, Vandecasteele interactions of inorganic constituents. Water Res C. Application of nanofiltration for the removal of 1999;33(1):196–206. pesticides and hardness from ground water: rejection [21] Fujita Y, Ding WH, Reinhard M. Identification of properties and economic evaluation. J Membr Sci wastewater dissolved organic carbon characteristics in 2001;193(2):239–48.