ARTICLE IN PRESS WAT E R R E S E A R C H 42 (2008) 714– 722 Available at www.sciencedirect.com journal homepage: www.elsevier.com/locate/watres Removal of natural organic matter and THM formation potential by ultra- and nanofiltration of surface water Ángeles de la Rubia, Manuel Rodrı́guez, Vı́ctor M. León, Daniel Prats Water and Environmental Science Institute, University of Alicante, PO Box 99, 03080 Alicante, Spain ar t ic l e i n f o abs tra ct Article history: Natural organic matter (NOM) and trihalomethane formation potential (THMFP) removal Received 31 January 2007 were evaluated by ultrafiltration (UF) and nanofiltration (NF). Ten different raw water Received in revised form sources in Alicante province (SE Spain) were analysed. 27 July 2007 Accepted 31 July 2007 Five types of membranes of different materials were tested with a dead-end-type stirred UF cell. Additional measurements, such as dissolved organic carbon, ultraviolet absorbance (254 nm), THMFP, ion concentration, pH, conductivity, etc. were made on raw water, Keywords: Natural organic matter Size exclusion chromatography Surface water Trihalomethane formation potential Ultrafiltration/nanofiltration permeates and concentrates. The SUVA value was used to determine the hydrophobicity of the water analysed. The elimination of NOM and THMFP is correlated with the molecular weight (MW) of NOM determined by size exclusion chromatography (SEC). The flux decline trends were correlated with cation concentration. NOM removal by UF is low, which correlates with the average MW determined by SEC with an average value of 922 g/mol (between 833 and 1031 g/mol). However, the NOM removal obtained with the NF90 and NF270 NF membranes for all water sources is almost complete (90%). THMFP removal is related to hydrophobicity and permeability of membrane. The NFT50 membrane removes almost 100% of the THMFP of more hydrophobic waters. & 2007 Elsevier Ltd. All rights reserved. 1. Introduction Natural organic matter (NOM) is a complex matrix of organic compounds present in natural surface water sources. Not only does it affect the odour, colour and taste of water but it also affects several processes in drinking water treatment (Aoustin et al., 2001). It is a precursor of trihalomethanes (THMs) after chlorination of drinking water (Park et al., 2005; Zhang and Minear, 2006). Chlorination has been the main means for disinfecting municipal drinking water in many countries, including Spain, for many decades and it will continue to be the most common disinfection process. The added chlorine reacts with naturally occurring organic matter to form a wide range of undesired halogenated organic compounds, often referred to as disinfection by-products (DBPs). Among the most widely occurring by-products are THMs, haloacetic acids, haloacetonitriles and haloketones. Corresponding author. Tel.: +34 965903400x2920; fax: +34 965903826. E-mail address: angeles.rubia@ua.es (A. de la Rubia). 0043-1354/$ - see front matter & 2007 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2007.07.049 Several treatment processes or their combinations are capable of removing NOM from water. Coagulation has often been used as a pretreatment for the removal of NOM to meet water quality requirements and in some cases where microfiltration alone is inadequate (Lahoussine-Turcaud et al., 1990; Wiesner et al., 1992; Vickers et al., 1995). The application of advanced oxidation processes for the removal of organics from water is gaining importance in water treatment (Murray and Parsons, 2004). Matilainen et al. (2006) evaluated the utilisation of active carbon to adsorb NOM in the final stages of surface water treatment. NOM can be effectively rejected during filtration by ultrafiltration (UF) and nanofiltration (NF) membranes, requiring relatively low pressure (Yoon et al., 2005). In fact, during the last decade the removal of NOM from ground and surface water for drinking water production was increasingly carried out by NF (Gorenflo et al., 2002). UF membranes remove NOM ARTICLE IN PRESS WAT E R R E S E A R C H 715 42 (2008) 714 – 722 by a sieving mechanism, and a previous coagulation treatment could improve NOM removal and decrease the resistance of membranes (Lahoussine-Turcaud et al., 1990). NF membranes are able to effectively remove NOM through a combination of size exclusion and physical–chemical interactions such as electrostatic repulsion and adsorption (Teixeira and Rosa, 2006). This process combines the advantages of a continuous production of quality water: almost total NOM rejection (490%) achieved with NF membranes having a 300 g/mol molecular weight cut-off (MWCO), high recovery (85–90%) and a low demand for chemicals. NF membrane water treatment processes not only remove NOM from surface water but also remove multivalent ions and small hazardous microcontaminants (e.g., pesticides, toxins, endocrine disruptors, etc.) (Gorenflo et al., 2002). Therefore, a membrane is a physical barrier that rejects macromolecules larger than the membrane pore size. This explains only the steric exclusion mechanism. However, NOM comprises a wide variety of macromolecules having several functional groups influencing NOM charge and hydrophobicity. Thus, to optimise the operation of NF and UF membranes, for the minimisation of the aforementioned adverse aspects, not only the quantitative amount of NOM removed by the membranes but also the qualitative changes in the NOM characteristics (molecular weight (MW) and hydrophobicity) from the feed to the permeate water should be rigorously considered (Lee et al., 2005; Yoon et al., 2005). MW and MW distributions are important factors in NOM characterisation because they relate to DBP formation potential (Amy et al., 1987). This paper focuses on the filtration of NOM from natural water using UF and NF membranes. The objective of this study was to investigate the UF and NF of natural surface water with different physicochemical properties. The work focused on characterising the MW distribution of surface waters in the province of Alicante, the flux evolution, the rejection of NOM and ions (Ca2+, Mg2+ and others), the quality of the NOM in the permeate and the reactivity of THM formation potential (THMFP). 2. Materials and methods 2.1. Source water Ten sources of water, including water reservoirs (Crevillente, Guadalest, La Pedrera, Tibi and Villajoyosa), different water channels (Taibilla, Mayayo, Reguerón and Reina) and the Segura River, were used to perform dead-end stirred-cell membrane tests. The samples were collected during the winter of 2006. All membrane filtration tests were performed with source water that was prefiltered using a 0.45 mm filter, corresponding to the specification for dissolved organic carbon (DOC). Several analytical methods were utilised to quantify NOM content: DOC, ultraviolet absorbance (UVA) at 254 nm (UVA254) and specific UVA (SUVA ¼ UVA254/DOC). The UVA of NOM is attributed exclusively to aromatic chromophores. In a drinking water treatment area, UVA254 is widely used as an indicator of the overall NOM concentration (Chin et al., 1994). In addition, the SUVA has been widely used as an index representing the relatively aromatic content of the colloidal carbon and therefore the NOM (Yoon et al., 2005). For this study, DOC, UVA254, SUVA, pH and conductivity were measured for each water source. The DOC (mg/L), SUVA (L/mg m), pH and ionic strength (mol/L) values of each water source are shown in Fig. 1. In addition, the concentration of the following anions: chloride, nitrate, sulphate and bicarbonate, and cations: calcium, sodium, magnesium and potassium, were measured. These data, together with conductivity, are tabulated in Table 1. All the water samples used in these filtration measurements were stored at 4 1C prior to their use after they had been prefiltered by 0.45 mm filtration. The raw water analysis (except for ion concentration) was always repeated at the same time the filtration measurements were taken. No significant change in composition over the period of this DOC pH SUVA Ionic strength 0.1 0.08 6 0.06 4.5 0.04 3 0.02 1.5 0 a ua da le st G jo yo s vi lla dr er a Pe ib illa La Ta bi Ti e vi lle nt ra re C ío R R eg Se ue gu ró o n 0 M ay ay R Ionic strength (mol/L) 7.5 ei na DOC (mg/l), SUVA (L/mg·m), pH 9 Fig. 1 – Raw water values of DOC (mg/L), SUVA (L/mg m), ionic strength (mol/L) and pH. ARTICLE IN PRESS 716 WA T E R R E S E A R C H 42 (2008) 714– 722 Table 1 – Ion concentration and conductivity of the raw water Surface waters K+ Ca2+ Na+ Mg2+ HCO 3 Reservoirs Crevillente Guadalest La Pedrera Tibi Villajoyosa 7.7 1.9 4.7 16.9 4.3 143.6 72.7 128.8 243.8 157.7 149.7 10.1 75.6 300.1 65.2 108.4 12.1 77.4 82.8 43.9 16.0 17.8 18.6 29.8 19.0 Water channels Mayayo Reguerón Reina Taibilla 20.1 22.6 23.2 4.1 416.7 373.1 399.2 149.7 769.0 547.7 713.9 69.8 327.0 246.7 308.1 76.7 River Segura 23.3 209.2 325.1 132.2 NO 3 SO2 4 Conductivity (mS/cm) 192.9 10.8 102.1 397.7 83.4 2.9 3.1 1.0 0.8 5.8 501.9 31.6 381.1 594.6 297.5 1675 370 1180 2470 1040 40.2 24.5 40.6 19.9 1047.7 694.0 1000.8 93.5 79.8 77.8 69.9 2.4 1522.3 1257.8 1353.6 394.8 5730 4430 5400 1200 26.1 412.7 22.4 753.7 4070 Cl Table 2 – Properties of ultra- and nanofiltration membranes tested Manufacturer Millipore Millipore Dow-Filmtec Dow-Filmtec Alfa laval a b c d e f g h Membrane type Material, filtration type MWCO (Da) Clean water flux (mL/min cm2) YM1 YC05 NF270 NF90 NFT50 Regenerated cellulose acetate Cellulose acetate Polypiperazine-based Polyamide thin-film composite Polypiperazine 1000a 500a 150b 200c 430d 200f 300g 150h 0.03–0.04a 0.02–0.04a 0.167e 0.006f 0.01h According to supplier, Millipore. Long et al. (2005). Braeken et al. (2006). Park et al. (2005). Mänttäri et al. (2004). Krieg et al. (2004). Reardon et al. (2005). Teixeira and Rosa (2006). study was observed and since the feed water was prefiltered by nominal 0.45 mm microfilters, biological organisms were not considered to be a significant part of the membrane challenge. 2.2. Experimental set-up All UF and NF experiments were performed in an Amicon dead-end-type stirred UF cell; the membrane diameter was 63.5 mm (nominal 62 mm) and the feed volume used was 200 mL. All tests were performed running the unit at a constant stirring speed of 200 rpm. The applied pressure (DP) was regulated using N2 gas. After installing in the test cell one of the many membranes utilised, the membrane was precompacted: pure water was permeated through it at 4  105 Pa for about 120 min and then the pure water flux at 3  105 Pa was measured. After precompaction, the water to be analysed was passed through the membrane to obtain 130 mL of permeate. The permeate flux was measured continuously. After the water sample had been passed through, the membrane was washed thoroughly with deionised water and its pure water permeability was measured in order to determine the difference before and after the assay. The permeate flux was measured gravimetrically using a Cobos CB Complet laboratory scale that was connected to and monitored via a computer. 2.3. Membranes The NOM rejection and flux-decline measurements were performed on three NF and two UF membranes. General information about the membranes is given in Table 2, including the manufacturer, material, MWCO and the clean water flux. 2.3.1. Ultrafiltration membranes YM1 is a regenerated cellulose acetate membrane and YC05 is a cellulose acetate membrane; both are hydrophilic membranes (Millipore; Kekki et al., 1997). These UF membranes exhibit a high removal of low-concentration non-specific proteins. Furthermore, the YC05 membrane yields a high solute rejection of NaCl. ARTICLE IN PRESS WAT E R R E S E A R C H 2.3.2. r ¼ Mw =Mn , Nanofiltration membranes  The FILMTECTM NF270 is a membrane designed to remove   42 (2008) 714 – 722 high percentages of total organic carbon (TOC) and THM precursors while having a medium to high salt passage and medium hardness passage (supplier). This membrane is relatively hydrophilic. The NF270 membrane thus shows an exceptionally high retention of uncharged compounds (e.g., glucose), and at the same time permits a relatively high flux. The membrane surface is negatively charged. This suggests the possibility of electrostatic repulsion of negatively charged NOM components by the membrane. The pure water permeability of the NF270 is especially high compared with other NF membranes (Mänttäri et al., 2004). The FILMTECTM NF90 is a NF membrane designed to remove a high percentage of salts, nitrates, iron and organic compounds such as pesticides, herbicides and THM precursors. The membrane surface charge is slightly negative (Reardon et al., 2005). The DSS NFT-50 (Alfa Laval) NF membrane is a polymeric three-layer thin-film membrane with an active layer of aromatic/aliphatic polyamide. The surface of this membrane is negatively charged in the 4.4–8.3 pH range and its negative charge increases with the pH (Teixeira et al., 2005). A Shimazdu high-performance liquid chromatography system consisting of an injector (SIL-IOA), a pump (LC-7A) and diode-array detector (SPD-M6A) has been used for HPSEC analysis. Separation by size exclusion was performed using a G3000SW column (7.5 mm  30 cm, Tosoh Bioscience Gmbh, 10 mm silica). A precolumn of the same packing material (7.5 mm  7.5 cm, Tosoh Bioscience Gmbh, 10 mm silica) was used after the injection valve. Samples (40 mL) were injected into the HPSEC column at a flow-rate of 1 mL/min (Chin et al., 1994). The mobile phase was a phosphate buffer at pH 6.8 and 0.1 N NaCl that was previously filtered (0.45 mm) and degassed with ultrasound. The detection was performed measuring absorbances at 224 and 254 nm simultaneously. The response at 224 nm is higher than at 254 nm. Therefore, these results were used in this study. Sodium polystyrene sulfonates (MWs of 210, 1400, 4300, 13,000 and 32,000 g/mol) were used as standards. A linear calibration curve (r240.99) was used to calculate MWs. The baseline of the chromatograms was changed due to tailing and was set as 0 at 2% of the maximum chromatogram height, based on the approach of Zhou et al. (2000). Mn, Mw and r (number-averaged MW, weight-averaged MW and polydispersivity, respectively) were determined using the following equations: Pn i¼1 ðhi Mi Þ , (1) Mw ¼ P n i¼1 hi Pn h Mn ¼ Pn i¼1 i , i¼1 ðhi =Mi Þ (2) (3) where hi and Mi are the height of the HPSEC chromatogram and the MW at eluted volume i, respectively. NOM MW estimation by HPSEC can be influenced by numerous factors such as calibration standards, undesirable column packing/resin interactions, unsuitable data handling of chromatograms and detection methods (Zhou et al., 2000). Although HPSEC with UVA detection is not the most adequate detector for quantitative analysis of MW estimation (Her et al., 2002), it is useful for qualitative analysis and for determining MWs using a proper mobile phase and calibration standard (Zhou et al., 2000). 2.5. Analytical methods Samples were analysed for DOC (Shimadzu TOC 5000A analyser), UV254 nm absorbance (Shimadzu UV-1601 UV–VIS spectrophotometer), pH (Crison 2000/6657 pH meter) and conductivity (Crison Micro CM 2200 conductivity meter) these analyses and THMFP, were performed according to standard methods (APHA, AWWA, WPCF, 1995). DOC, UVA254, pH and conductivity in raw water, concentrates and permeates were characterised. DOC and UVA rejection are defined as the observed rejection: R¼ 2.4. High-performance size exclusion chromatography (HPSEC) 717 cb  cp 100%, cb where cb and cp are the concentrations in the bulk solution (concentrate) and the permeate, respectively. Metal content was determined by atomic absorption spectrometry (Perkin-Elmer 2100 Flame AAS and HGA-700 graphite furnace). Anion concentrations were measured with an IC chromatographic system (DIONEX DX5OO). THMFP was determined using a gas chromatograph with a mass spectroscopy detector (HP/AGILENT TECHNOLOGIES 6890N) with purge and trap injection (TEKMAR DOHRMANN), 3100 Sample Concentrator. 3. Results and discussion 3.1. NOM and ion removal by NF and UF The efficiencies of the NF and UF membranes in removing NOM, with respect to UVA254 and DOC for the 10 water samples analysed, are summarised in Table 3. As listed in Table 3, the NOM was removed more efficiently by the NF than by the UF membranes. In fact, the NF90 and the NF270 membranes yield almost total DOC rejection (90%). Although the NFT50 membrane has lower MWCO than the rest of the NF membranes, it is less efficient for low DOC and ionic strength samples (Villajoyosa, Taibilla and Guadalest). However, the NFT50 membrane shows higher removal efficiency for higher DOC samples. This membrane is  also less efficient in the removal of anions (Cl, NO 3 , HCO3 ) + + and cations (K , Na ) than the rest of NF membranes (Table 4). The NF90 membrane is the most efficient in the removal of ions, being higher than 60% in the majority of cases except for 2+ or Mg2+. The nitrates, and is close to 100% for SO2 4 , Ca ARTICLE IN PRESS 718 WA T E R R E S E A R C H 42 (2008) 714– 722 Table 3 – NOM removal by NF and UF membranes using DOC and UVA254 nm measurements for all tested water Surface water Raw water DOC (mg/L) Removal (%) YM1 Crevillente Guadalest La Pedrera Tibi Villajoyosa Mayayo Reguerón Reina Segura Taibilla 6.21 1.44 2.62 5.41 1.88 6.50 6.32 7.06 6.28 1.56 YC05 NF270 NF90 NFT50 DOC UVA DOC UVA DOC UVA DOC UVA DOC UVA 62.85 22.01 45.50 35.97 50.90 58.05 38.48 65.47 67.52 32.05 65.93 64.85 64.86 72.27 77.97 69.42 65.34 65.07 63.25 67.90 85.35 56.67 66.18 89.80 70.74 84.55 65.24 93.16 87.32 73.90 88.43 74.06 85.62 93.19 79.71 89.86 83.81 86.71 83.41 72.15 95.72 68.61 85.50 90.68 76.54 87.25 83.50 92.03 99.16 84.84 91.28 83.26 88.82 92.44 90.14 91.43 89.70 90.98 93.15 93.10 94.73 70.00 74.92 99.09 93.62 88.51 88.96 90.33 94.41 84.20 90.33 76.15 85.30 93.61 90.14 89.86 91.66 84.48 89.64 88.86 87.15 67.89 88.63 84.29 34.84 93.20 95.87 100 98.49 73.64 92.23 58.58 87.54 92.94 63.48 85.83 94.60 93.69 94.56 76.39 Table 4 – Ion percentage removal for each membrane tested Ion Cl NO 3 SO2 4 HCO 3 K+ Na+ Ca2+ Mg2+ Removal (%) YM1 YC05 NF270 NF90 NFT50 6.3473.39 15.6676.04 22.19710.10 11.3979.86 3.6474.20 5.8373.12 11.17710.34 16.91712.66 20.19710.52 32.34723.39 96.2971.32 37.37716.63 12.1976.18 16.4077.79 82.49711.40 86.7079.60 12.2178.13 26.97729.41 93.0471.81 37.04714.68 31.24716.52 35.12713.97 56.6979.84 57.26726.65 73.57718.13 46.13728.08 98.4670.79 88.5077.79 62.44715.79 68.41715.16 97.3971.00 93.65711.97 8.5476.04 10.86712.44 81.60730.44 24.12719.74 27.5179.02 30.0178.42 50.05718.86 52.78724.39 higher efficiency of this membrane is probably due to their specific material filtration (slightly negative charge) that permits to improve the results obtained with other NF membranes. The NOM removal efficiencies, as determined by UVA254, were higher than those obtained by DOC (Table 3), except in the case of the Reina irrigation water channel and Segura River. This indicates that aromatic/hydrophobic compounds (UV absorbance at 254 nm is attributed mainly to the absorption of these compounds) can be preferentially removed over the entire range of membrane pore size (Schäfer et al., 2000; Lee et al., 2005). The SUVA values of the Reina irrigation water channel and Segura River are about 3 L/mg m (Fig. 1). The DOC of these water sources is probably composed largely of humic substances and relatively hydrophobic and aromatic compounds (Edzwald and Van Benschoten, 1990). 3.2. Changes in NOM size from feed to permeate The MW and distribution of four NOM types: Guadalest (water reservoir), Segura (river), Taibilla (water channel) and Mayayo (irrigation water channel), included in the feed and permeates are compared in Fig. 2. The weight average (Mw) and the number average (Mn) of all the feed water and permeates obtained from the membranes are compared in Table 5. The MW of the NOM was lower in the permeates than in the feed water, as can be seen in Table 5 and Fig. 2. This result indicates that the membranes reject relatively large molecules more, showing that the hydrophilic NOM is preferentially transmitted through the membrane pores, as its molecular size is relatively small compared with that of its hydrophobic counterpart. However, in some samples an increase of the lower MW NOM in permeates of UF (Segura River, Reina and Reguerón) and NF (Mayayo, Reina, Reguerón and Segura River) is observed. This could be due to a fractionisation of some high MW compounds that reduces the molecular volume and permits them to cross membranes and consequently an increase in the low MW portion is detected in permeates. The small r (Mw:Mn ratio) indicates low NOM heterogeneity in the 10 raw water samples analysed. The MWs of the NOM analysed are of the same order of magnitude; the Mw varied from 833 for Reguerón to 1031 for Crevillente, and because of this the NOM was removed efficiently. It was found that the Mw of the NOM in permeates was also of similar value; ARTICLE IN PRESS WAT E R R E S E A R C H Response UVA224nm a MAYAYO 6.E+05 4.E+05 SEGURA RIVER GUADALEST 2.E+05 b Response UVA224nm FEED WATER TAIBILLA 0.E+00 2000 1500 1000 500 Molecular Weight (daltons) MAYAYO 6.E+05 0 PERMEATE YM1 SEGURA RIVER TAIBILLA 4.E+05 GUADALEST 2.E+05 0.E+00 2000 c Response UVA224nm 719 42 (2008) 714 – 722 6.E+05 1500 1000 500 Molecular Weight (daltons) 0 PERMEATE NF90 MAYAYO SEGURA RIVER TAIBILLA 4.E+05 GUADALEST 2.E+05 0.E+00 2000 1500 1000 500 Molecular Weight (daltons) 0 Fig. 2 – Molecular weight distribution of the NOM contained in the feed (a), UF permeate (b) and NF permeate (c) of Mayayo irrigation water channel, Segura river, Taibilla water channel and Guadalest water reservoir. electrostatic repulsion rather than size exclusion is the more dominant mechanism for the removal of NOM by all membranes. In fact, the NFT50 membrane, with the lower theoretical MWCO, reduces the Mw of samples with low DOC and ionic strength (e.g., Guadalest and Taibilla) less than the rest of the NF membranes. Nonetheless, as can be seen in Fig. 2, the MW distribution of Guadalest evolves in such a way that the MW decreases, as is apparent from the shifting of the maximum of the Guadalest curve to lower MW values. The efficiency of the NF90 membrane was better than that of the other membranes tested for the rest of the surface water samples. In general, all the source water analysed has the same or smaller Mw than MWCO; therefore, all the NOM does not pass through the membrane pores, but not only average MW and MWCO should be considered: NOM rejection was governed by electrostatic repulsion and hydrophobic adsorption for the negatively charged membranes. 3.3. DBP reactivity of the NOM included in each feed and the permeates The DBP reactivity in terms of THM ( ¼ THMFP/DOC (mg/mg)) for all permeates, is listed in Table 6, compared with those of all NOM in the feed water. THMFP removal is known to be proportional to the NOM removed (Lee et al., 2005), resulting in significantly reduced THMFP for the product water obtained from the NF, and even from the UF. However, Table 6 clearly indicates that the DBP reactivity reduction was not as high as the NOM removal, suggesting that the reactive NOM for THM formation was still included in both the NF and the UF permeates. These trends were more obvious for YC05 and NF270 product water than for the other permeates. Based on these results, it can be suggested that even though the membrane processes must take much of the credit for the NOM removal (in terms of THMFP reduction), the selection of appropriate membranes and operating conditions for DBP minimisation are still rather important, as ARTICLE IN PRESS 720 766 570 615 125 39 753 902 697 736 677 532 776 363 618 720 688 749 565 909 762 769 776 779 770 849 762 758 758 758 758 623 697 716 712 695 689 755 712 750 750 690 716 Mn Mw Mn Mw Mn Mw Mn 42 (2008) 714– 722 small amounts of the NOM included in the produced water may be problematic due its high DBP reactivity. The NOM present in irrigation water channels and rivers has more hydrophobic character (greater SUVA) and is therefore eliminated more easily. Hence, the THMFP reactivity for these types of water is always low, but water with a lower SUVA value (Fig. 1), or rather, water types with organic matter that has a more hydrophilic character, show a greater THMFP reactivity. Moreover, the YC05 and NF270 membranes with MWCO at around 500 Da permits the passage of the more reactive organic matter present in this type of water, resulting in THM reactivity values greater than that of the feed water; the high flow-rate that the NF270 membrane permits could also contribute to the greater reactivity observed (Lee et al., 2005). The NF90 and NFT50 membranes with MWCO at around 200 Da, retains the hydrophilic organic matter, which diminishes the THM reactivity substantially. The NFT50 does this particularly well, with nearly 100% removal achieved for water with greater SUVA values. 833 765 760 754 756 755 586 627 656 421 654 746 870 754 798 883 770 754 206 NA 606 644 679 670 888 NA 708 725 755 744 326 59 NA 563 746 143 Mn Mw Mn Mw Mn Mw 3.4. 982 633 NA 945 846 796 914 588 NA 641 642 686 1016 720 NA 710 749 770 841 659 263 385 535 801 NA: not available. 927 728 662 663 659 804 950 750 653 735 740 NA 1031 798 741 748 783 NA Feed YM1 YC05 NF270 NF90 NFT50 Mn Mw Mw Mn Mw Mn Mw Tibi La Pedrera Guadalest Crevillente Table 5 – Measured molecular weight of NOM in each feed and permeate Villajoyosa Mayayo Reguerón Reina Segura river Taibilla WA T E R R E S E A R C H Effect of ion concentration on flow-rate It is important to note that ions interact strongly with the membrane, especially with NF membranes. The NF270 membrane exhibited the highest permeability values of all the membranes tested, followed by the NF90 membrane. The NFT50 membrane shows no significant flux decline over the running time with any of the types of water studied. This could be so for two possible reasons: firstly, a low pressure is applied, and secondly, minimal concentration polarisation occurs. Thus, this membrane removes both hydrophilic and hydrophobic NOM (Teixeira and Rosa, 2006). The presence of dissolved Na+, Mg2+ or/and Ca2+ in the waters of Reina, Mayayo, Reguerón (irrigation water channel) and Segura River has a strong influence on NF membrane performance, and the flux is lower than for water with lower ionic strength. The Ca2+ ions can bind with the acidic functional groups of the NOM, elevating the degree of hydrophobicity of the NOM molecules and developing a dense thick fouling layer on the membrane surface (Teixeira and Rosa, 2006). As can be seen in Fig. 3, the water flux for the NF90 and NF270 membranes shows a decrease in permeability with time (Nilsson et al., 2006). In water with a lower ionic strength, high pH and lower concentration of divalent ions such as the waters of Taibilla, La Pedrera, Villajoyosa and Guadalest, a loose thin fouling layer develops on the membrane surface and a lower flux decline is obtained. The SUVA value of and Ca2+ concentration in water from Reina, Mayayo, Reguerón and Segura River is high; therefore the flux declines rapidly. The SUVA value of Taibilla is also high, but the Ca2+ concentration is lower than that for the others, and in this case a lower flux decline is observed. Thus, the presence of calcium ions is much more important for flux than the type of NOM (Teixeira and Rosa, 2006). 4. Conclusions All membranes tested were well suited to eliminate NOM, typically for raw water sources. NF removed most of the ARTICLE IN PRESS WAT E R R E S E A R C H 721 42 (2008) 714 – 722 Table 6 – THM reactivities of NOM in each feed and permeate THM reactivity (mg/mg) Feed water YM1 YC05 NF270 NF90 NFT50 53.46 108.03 85.17 69.09 103.70 133.26 116.74 57.29 90.65 161.14 40.68 151.10 74.77 48.37 190.85 51.06 51.34 26.17 59.94 115.04 55.95 164.33 142.85 34.34 390.71 35.82 62.83 30.72 38.06 84.83 48.65 131.80 146.99 63.61 116.37 17.98 23.92 18.66 44.64 211.69 34.60 94.45 37.15 25.49 34.32 27.12 38.20 17.03 12.42 108.31 29.35 94.16 44.28 23.25 42.84 ND 8.56 ND 1.19 31.17 Crevillente Guadalest La Pedrera Tibi Villajoyosa Mayayo Reguerón Reina Segura River Taibilla ND: not detected. 1.E-07 YM1 YC05 NF270 NF90 NFT50 VILLAJOYOSA 1.E-07 GUADALEST 8.E-08 Flux (m/s) 8.E-08 6.E-08 4.E-08 2.E-08 6.E-08 4.E-08 2.E-08 0.E+00 0.E+00 0 2000 4000 1.E-07 6000 0 8000 2000 4000 time, s 9.E-08 LA PEDRERA 6000 8000 TAIBILLA 8.E-08 6.E-08 6.E-08 4.E-08 3.E-08 2.E-08 0.E+00 0.E+00 0 2000 4000 9.E-08 6000 0 8000 2000 4000 8.E-08 TIBI 6000 8000 CREVILLENTE 6.E-08 6.E-08 4.E-08 3.E-08 2.E-08 0.E+00 0.E+00 0 2000 4000 8.E-08 6000 0 8000 2000 4000 SEGURA RIVER 6000 8000 REGUERÓN 6.E-08 6.E-08 4.E-08 4.E-08 2.E-08 2.E-08 0.E+00 0.E+00 0 2000 4000 6.E-08 6000 0 8000 2000 4000 6000 6.E-08 MAYAYO 4.E-08 4.E-08 2.E-08 2.E-08 0.E+00 8000 REINA 0.E+00 0 2000 4000 6000 8000 0 2000 4000 Fig. 3 – Flux evolution of all the water and membranes tested. 6000 8000 ARTICLE IN PRESS 722 WA T E R R E S E A R C H organic matter (70–95%) whereas UF rejected mainly the high molecular refractory part (30–85%). The higher NOM removal efficiencies were obtained with the NF90 and NF270 membranes for all the tested water; however, the NFT50 was less efficient for low DOC and ionic strength water. The NF90 membrane was also the most efficient at anion and cation 2+ 2+  removal, especially for SO2 4 , Ca , Mg and HCO3 at the testing conditions. NF led to a high rejection of NOM and a high reduction of THM precursors for all the water sources, while the rejection of NOM and THM precursors by UF was clearly lower. The NFT50 membrane was the most efficient at reducing the THM reactivity for all the tested water at the testing conditions. However, the obtained permeates from the NF270 and UF membranes showed higher THM reactivities, especially in low DOC and ionic strength water. Flux decreases in the presence of NOM and decreases further in the presence of both NOM and calcium ions. The estimation of MW is an important factor that aids in the understanding of the physical and chemical properties of NOM and the determination of appropriate water treatment process selection, design and operation. Acknowledgements This study was financially supported by the Comisión Interministerial de Ciencia y Tecnologı́a (CICYT) of the Spanish Government: Plan Nacional de I+D 2004-2007 (ref. CTM2004-03056) and by a grant from the Consejerı́a de Innovación, Tecnologı́a y Empresa de la Junta de Andalucı́a para el Perfeccionamiento de Investigadores. R E F E R E N C E S Amy, G.L., Collins, M.R., Kuo, C.J., King, P.H., 1987. Comparing gel permeation chromatography and ultrafiltration for the molecular weight characterization of aquatic organic matter. J. Am. Water Works Assoc. 79 (1), 43–49. Aoustin, E., Schäfer, A.I., Fane, A.G., Waite, T.D., 2001. Ultrafiltration of natural organic matter. Sep. Purif. Technol. 22–23, 63–78. APHA, AWWA, WPCF, 1995. Standard Methods for the Examinations of Water and Wastewater, 19th ed. American Public Health Association, Washington, DC. Braeken, L., Bettens, B., Boussu, K., Van der Meeren, P., Cocquyt, J., Vermant, J., Van der Bruggen, B., 2006. Transport mechanisms of dissolved organic compounds in aqueous solution during nanofiltration. J. Membr. Sci. 279, 311–319. Chin, Y., Aiken, G., O’Loughlin, E., 1994. Molecular weight, polydispersity and spectroscopic properties of aquatic humic substances. Environ. Sci. Technol. 28, 1853–1858. Edzwald, J.K., Van Benschoten, J.B., 1990. Aluminium coagulation of natural organic matter. In: Hahn, H.H., Klute, R. (Eds.), Chemical Water and Wastewater Treatment, vol. 40. Springer, Berlin, pp. 341–359. Gorenflo, A., Velázquez-Padrón, D., Frimmel, F.H., 2002. Nanofiltration of a German groundwater of high hardness and NOM content: performance and costs. Desalination 15 (1), 253–265. Her, N., Amy, G., Foss, D., Cho, J., 2002. Variations of molecular weight estimation by HP-size exclusion chromatography with 42 (2008) 714– 722 UVA versus online DOC detection. Environ. Sci. Technol. 36, 3393–3399. Kekki, T., Rosenberg, R.J., Jaakkola, T., 1997. Physico-chemical forms of radiostrontium in simulated freshwaters. J. Radioanal. Nucl. Chem. 224 (1–2), 77–81. Krieg, H.M., Modise, S.J., Keizer, K., Neomagus, H.W.J.P., 2004. Salt rejection in nanofiltration for single and binary salt mixtures in view of sulphate removal. Desalination 171, 205–215. Lahoussine-Turcaud, V., Wiesner, M.R., Bottero, J.-Y., Mallevaille, J., 1990. Coagulation pretreatment for ultrafiltration of a surface water. J. Am. Water Works Assoc. 82 (12), 76–81. Lee, S., Kwon, B., Sun, M., Cho, J., 2005. Characterizations of NOM included NF and UF membrane permeates. Desalination 173, 131–142. Long, F., Zhu, A., Wang, X.L., Zhu, W.P., 2005. Membrane flux and CaCO3 crystallization in the unstirred dead-end nanofiltration of magnetic solution. Desalination 186, 243–254. Mänttäri, M., Pekuri, T., Nyström, M., 2004. NF270, a new membrane having promising characteristics and being suitable for treatment of dilute effluents from the paper industry. J. Membr. Sci. 242, 107–116. Matilainen, A., Vieno, N., Tuhkanen, T., 2006. Efficiency of the activated carbon filtration in the natural organic matter removal. Environ. Int. 32 (3), 324–331. Murray, C.A., Parsons, S.A., 2004. Removal of NOM from drinking water: Fenton’s and photo-Fenton’s processes. Chemosphere 54 (7), 1017–1023. Nilsson, M., Trägardh, G., Östergren, K., 2006. The influence of sodium chloride on mass transfer in a polyamide nanofiltration membrane at elevated temperatures. J. Membr. Sci. 280, 928–936. Park, N., Kwon, B., Sun, M., Ahn, H., Kim, C., Kwoak, C., Lee, D., Chae, S., Hyung, H., Cho, J., 2005. Application of various membranes to remove NOM typically occurring in Korea with respect to DBP, AOC and transport parameters. Desalination 178, 161–169. Reardon, R., Treadway, J., Buckley, B., Hobbs, C., 2005. Is ultrafiltration better than microfiltration as pre-treatment for reverse osmosis? Pilot scale results. In: Water Environment Federation’s Technical Exhibition and Conference (WEFTEC), November, Washington, DC, USA, pp. 3530–3546. Schäfer, A.I., Fane, A.G., Waite, T.D., 2000. Fouling effects on rejection in the membrane filtration of natural waters. Desalination 131, 215–224. Teixeira, M.R., Rosa, M.J., 2006. The impact of the water background inorganic matrix on the natural organic matter removal by nanofiltration. J. Membr. Sci. 279, 513–520. Teixeira, M.R., Rosa, M.J., Nyström, M., 2005. The role of membrane charge on nanofiltration performance. J. Membr. Sci. 265, 160–166. Vickers, J.C., Thompson, M.A., Kelkar, U.G., 1995. The use of membrane filtration with coagulation processes for improved NOM removal. Desalination 102, 57–61. Wiesner, M.R., Veerapaneni, S., Brejchová, D., 1992. Improvements in membrane microfiltration using coagulation pre-treatment. In: Hahn, H.H., Klute, R. (Eds.), Chemical Water and Wastewater Treatment. Springer, Berlin. Yoon, Y., Amy, G., Cho, J., Her, N., 2005. Effects of retained natural organic matter (NOM) on NOM rejection and membrane flux decline with nanofiltration and ultrafiltration. Desalination 173, 209–221. Zhang, X., Minear, R.A., 2006. Removal of low-molecular weight DBPs and inorganic ions for characterization of high-molecular weight DBPs in drinking water. Water Res. 40, 1043–1051. Zhou, Q., Cabaniss, S.E., Maurice, P.A., 2000. Considerations in the use of high-pressure size exclusion chromatography (HPSEC) for determining molecular weights of aquatic humic substances. Water Res. 34 (14), 3505–3514.