In this study high performance size exclusion chromatography (HPSEC) was used to compare an ultrafiltration (UF) membrane and alum coagulation for their capacity to remove different fractions of natural organic matter (NOM) from water. At the same time, the removal of disinfection by-product (DBP) precursors, as measured by trihalomethane formation potential (THMFP) and haloacetic acid formation potential (HAAFP), was also detected. The results show that the UF membrane mainly removed the aliphatic biopolymer fraction, while alum coagulation mainly removed the humic substances fraction. The results of DBP precursor analysis show that more THMFP was removed by the UF membrane than HAAFP, while the reverse was true for alum coagulation. It is conjectured that the aliphatic biopolymer fraction is the major precursor for trihalomethanes (THMs), while the humic substances fraction is the major precursor for haloacetic acids (HAAs).

INTRODUCTION

Natural organic matter (NOM) is a complex mixture of organic materials present in natural waters, and can cause problems with regard to undesirable color, taste and odor. NOM has a significant impact on many aspects of water treatment. For example, it may increase the chemical dosage in coagulation, interfere with the iron and manganese removal process, and cause fouling in the membrane process (Jacangelo et al. 1995). The residual NOM in the treated water may induce biological regrowth in the distribution system, and cause a deterioration in water quality (Matilainen et al. 2011). Moreover, NOM has been found to be the major contributor to disinfection by-product (DBP) formation, such as trihalomethanes (THMs) and haloacetic acids (HAAs), when chlorine is used as a disinfectant (Kim & Yu 2005; Hua & Reckhow 2007). The problem of DBPs has been of increasing concern due to their adverse health effects (Singer 2006), and thus any NOM that exists in source water must be characterized to achieve effective DBP control in water treatment. However, NOM has a broad range of characteristics. It may be carried into the water body by runoff from the catchment (allochthonous organic matter), and thus include humic substances. It may also be produced by the biological activity that occurs inside the water body itself (autochthonous organic matter), which may include polysaccharides, proteins, and so on (Amy 2008). The organic molecules of the former are thought to contain more aromatic structures and are more hydrophobic than those of the latter.

NOM has been reported as the principal precursor in the formation of THMs and HAAs (Christman et al. 1983). The characteristics of NOM, including carboxylic/phenolic acidity, aromatic/aliphatic content, and ultraviolet absorbance, have been shown to affect the formation of DBPs. Hydrophobic NOM, which generally has a higher aromatic content, phenolic acidity, and ultraviolet absorbance, also has greater halogenated organics formation potential (Chen et al. 2008). Further, the relative distribution of HAAs and THMs is thought to be influenced by the hydrophobic/hydrophilic distribution of NOM in the water being chlorinated (Liang & Singer 2003).

However, there is conflicting information regarding which NOM fractions are dominant precursors for THMs and HAAs. Some researchers propose that the hydrophilic fraction is a more significant precursor of THMs than HAAs, based on the observation that the former were generally present at higher concentrations than the latter in finished drinking water, which usually contains a greater hydrophilic than hydrophobic fraction. However, the opposite has been shown from the chlorination of humic substances (Croue et al. 2000). Liang & Singer (2003) studied the influence of various water quality and treatment characteristics on the formation and distribution of nine haloacetic acids and four trihalomethanes in drinking water under controlled chlorination conditions. The NOM was fractionated by XAD-8 resin. The results also showed that haloacetic acid precursors had a higher aromatic content than trihalomethane precursors. However, other studies reached different conclusions. For example: Kim & Yu (2005) fractionated the NOM in raw and processed water from a conventional water treatment plant into hydrophobic and hydrophilic fractions by XAD-resin, and compared DBP formation of three isolated fractions. They concluded that the formation of THMs was highly influenced by the hydrophobic fraction, while the formation of HAAs depended more on the hydrophilic fraction. Lamsal et al. (2012) studied the characteristics of NOM from surface water through resin fractionation and evaluated the various fractions in terms of their DBP formation potential. They found that the removal of the hydrophobic acid fraction by a treatment process resulted in a greater reduction in the trihalomethane formation potential (THMFP) than of the haloacetic acid formation potential (HAAFP).

A number of characterization techniques have been employed to obtain a better understanding of the types of NOM present in source water, and their removal or transformation through the water treatment process train (Matilainen et al. 2011). High performance size exclusion chromatography (HPSEC), especially with an organic carbon detector (OCD), has been shown to be an attractive option for NOM characterization, due to its ease of operation, ability to detect any type of organic carbon bonded species, and also its simplicity with regard to sample preparation and requirement for minimal sample volume (Allpike et al. 2005). In this study HPSEC was used to compare the UF membrane filtration and alum coagulation process for their capacity to remove different fractions of NOM from water. At the same time, the removal of DBP precursors by the two units was also examined. Finally, the correlation between organic fractions and precursors for THMs and HAAs was evaluated.

MATERIALS AND METHODS

Source water

The source water used in this study was collected from the effluent of the slow sand filters in a drinking water treatment plant in Taiwan. The raw water of the plant was drawn from a eutrophic lake. The treatment train before slow sand filtration included coagulation, dissolved air flotation, and rapid sand filtration. Two batches of source water were collected. The first one, obtained in March 2014, was used for the ultrafiltration (UF) study. The second one, collected in November 2014, was used for the coagulation study. As this study focused on the removal of dissolved organic matter, the collected source water was first filtered through 1.2/0.5 and 0.45 μm disposable capsule filters (both from Millipore Corporation) to remove any suspended solids before further usage.

UF membrane filtration

Membrane filtration was conducted with a bench-scale dead-end membrane testing system (Figure 1) under room temperature (23 ± 2 °C). The central element of the system was a single hollow fiber UF membrane with a length of 12 cm, which was put inside a Teflon tube (I.D. 4 mm). Two T-type connectors (polypropylene material) were plugged into both ends of the tube, and epoxy resin was used to seal and fix the hollow fiber inside it, as shown in Figure 2.
Figure 1

Schematic diagram of the bench-scale UF membrane filtration system.

Figure 1

Schematic diagram of the bench-scale UF membrane filtration system.

Figure 2

Schematic diagram of the single hollow fiber UF membrane.

Figure 2

Schematic diagram of the single hollow fiber UF membrane.

Feed water was first put into a reservoir (RC-800 Mini-Reservoir, Amicon, USA), and was mixed using a magnetic stirrer at 25 rpm. It was then pushed into the hollow fiber membrane in an inside-out flow pattern. Constant-pressure filtration (0.7 bar) was maintained by gas pressure regulated from a nitrogen cylinder, and monitored with a digital pressure gauge (Model P-100 PSIG-D, Alicat Scientific, USA). Four kinds of UF membrane were used, namely PVC (polyvinyl chloride), CA (cellulose acetate), PVDF (polyvinylidene fluoride) and PS (phosphatidylserine), and the membrane characteristics are shown in Table 1.

Table 1

The membrane characteristics

MembraneNominal pore size (nm)Contact angle (deg)MWCO (kDa)
PVDF 30 66 100 
CA 20 46 100 
PVC 10 87 50 
PS 20 72 200 
MembraneNominal pore size (nm)Contact angle (deg)MWCO (kDa)
PVDF 30 66 100 
CA 20 46 100 
PVC 10 87 50 
PS 20 72 200 

Coagulation with jar tests

Alum (Al2(SO4)3.18H2O, Merck, Germany) was used as the coagulant with various dosages of 0.5, 1, 2, 4, 8, and 16 mg/L (as Al). A six paddle gang stirrer (Jar Test Apparatus, Phipps & Bird, Richmond, Virginia, USA) was used at variable speed. The filtered source water samples (1 L) were placed on the gang stirrer with six samples tested at a time and the coagulant was added while stirring at 100 rpm. After 3 min of rapid mixing at 100 rpm, the speed was reduced to 35 rpm for 15 min, then allowed to settle for 15 min. The supernatants were collected, and filtered through a 0.45 μm filter (cellulose acetate, Toyo Roshi, Japan) before further analysis.

Water quality analysis

To investigate the composition of dissolved organic matter as a function of apparent molecular weight (AMW), high performance liquid chromatography (HPLC, LC-20 ATV, Shimadzu, Japan) size exclusion chromatography (SEC) was conducted with sequential on-line detectors consisting of a UVD (254 nm, SPD-20A, UV-Vis detector, Shimadzu) and OCD (modified Sievers Total Organic Carbon Analyzer 900 Turbo, GE Water & Process Technologies). The chromatographic column (20 (I.D.) × 200 (L) mm TSK HW-50S, Toso, Japan) was a weak cation exchange resin (Toyopearl HW-50S, 30 μm) based on polymethacrylate. The eluent system consisted of phosphate buffer (2.4 mM NaH2PO4 + 1.6 mM Na2HPO4; pH 6.8), containing sodium sulphate (25 mM), to achieve an ionic strength of 100 mM. The eluent flow rate was 0.5 mL/min, and the sample volume was 2 mL. The PEG standard was used to determine apparent molecular weight. Prior to chromatographic separation, the samples were made particle-free by passing through a 0.45 μm filter (cellulose acetate, Advantec, Japan).

Chromatograms were analyzed by using the peak-fitting technique to resolve the overlapping peaks and to determine the area under each peak. PeakFit (Version 4.12, Systat Software Inc., USA, CA), a commercially available software package, was used for this analysis. The procedure was adapted from the method described by Chow et al. (2008).

For the DBP precursor analysis, an adequate amount of sodium hypochlorite solution was first injected into the samples to ensure that at least 1 mg/L free residual chlorine existed at the end of a 7-day (25 °C) incubation period. Residual chlorine was measured by the DPD ferrous method (APHA, AWWA & WEF 1998). The samples were then dechlorinated by adding sodium thiosulfate solution. Four species of trihalomethanes, namely chloroform, dichlorobromomethane, dibromochloromethane, and bromoform, were measured by the purge and trap packed-column gas chromatographic method using a gas chromatograph (Model 3400 GC, Varian, USA) equipped with a purge and trap module (Model LCS-2000, Tekmar, USA) and an electron capture detector. The measurement of nine species of haloacetic acids, namely monochloroacetic acid, monobromoacetic acid, dichloroacetic acid, dibromoacetic acid, trichloroacetic acid, tribromoacetic acid, bromochloroacetic acid, dichlorobromoacetic acid, and dibromochloroacetic acid, basically involved liquid-liquid extraction with methyl tertiary butyl ether (MTBE) and esterification with diazomethane prior to gas chromatography–electron capture detector (GC-ECD) analysis.

RESULTS AND DISCUSSION

Characteristics of source water

Two batches of effluent from the slow sand filter were collected and used as the source water for this study. The first batch was collected during March 2014 (specified as source water A, SW-A), while the second one was collected during November 2014 (specified as source water B, SW-B). The general water quality of these source waters is shown in Table 2. It is noted that the two batches have comparable pH and alkalinity values. However, SW-A has higher NPDOC (non-purgeable dissolved organic carbon) and lower SUVA (specific UV absorbance) values than those of SW-B.

Table 2

Characteristics of source water prior to UF membrane filtration and coagulation

 Water quality
SamplepHNPDOC (mg/L)UV254 (cm−1)SUVA (L/mg-m)Alkalinity (mg/L as CaCO3)
Source water A (SW-A) 7.2 6.8 0.070 1.00 85 
Source water B (SW-B) 7.5 4.2 0.058 1.38 35 
 Water quality
SamplepHNPDOC (mg/L)UV254 (cm−1)SUVA (L/mg-m)Alkalinity (mg/L as CaCO3)
Source water A (SW-A) 7.2 6.8 0.070 1.00 85 
Source water B (SW-B) 7.5 4.2 0.058 1.38 35 

Figure 3 shows the variations in the dissolved organic content with AMW of the source water, as obtained using HPSEC chromatography equipped with OCD, as well as the results from peak-fitting. First, these reveal that the shape of the chromatograms of SW-A and SW-B is similar. This means that the AMW distribution of the dissolved organic content in both batches of source water is similar, even though the heights of the SW-A peaks are higher than the corresponding peaks of SW-B. The major dissolved organic groups of the source water include Peaks A, B, C and D, with AMW of about 50,000 Da, 1,650 Da, 1,300 Da and 630 Da, respectively.
Figure 3

Distribution of molecular weight of organic fractions as detected by HPSEC-OCD in (a) SW-A and (b) SW-B.

Figure 3

Distribution of molecular weight of organic fractions as detected by HPSEC-OCD in (a) SW-A and (b) SW-B.

Figure 4 shows the overlapping figures of the HPSEC-OCD and UVD chromatograms. It indicates that Peak A of HPSEC-OCD has no corresponding peak in the UVD chromatograms, while Peaks B, C and D do. As OCD can detect all NOM containing carbon, while only NOM with conjugated or aromatic organic constituents absorbs ultraviolet light, Peak A is probably contributed by biopolymers, such as polysaccharides or amino sugars, which are hydrophilic. The other peaks can be associated with humic substances (HS) (Peak B), building blocks (Peak C) which represent HS-like materials of lower molecular weight, and low molecular weight acids (Peak D) (Chow et al. 2008; Huber et al. 2011).
Figure 4

Characteristics of organic fractions as detected by HPSEC-OCD and HPSEC-UVD in (a) SW-A and (b) SW-B.

Figure 4

Characteristics of organic fractions as detected by HPSEC-OCD and HPSEC-UVD in (a) SW-A and (b) SW-B.

The peak area and percentage distribution of the resolved peaks of the HPSEC-OCD chromatograms in the source water by peak-fitting are shown in Table 3. First, the total area of SW-A is larger than that of SW-B, and this is consistent with the higher NPDOC value of the former. Second, the percentage contributed by Peak A in SW-A (33.6%) is much larger than that of SW-B (6.5%). This means that the NOM in SW-A contains a greater hydrophilic organic fraction than that in SW-B. This is also consistent with the lower SUVA value of SW-A than that of SW-B, as shown in Table 2.

Table 3

The peak area and percentage distribution of the resolved peaks of the HPSEC-OCD chromatograms of source water with peak-fitting

SamplesSW-A
SW-B
Organic fractionsPeak APeak BPeak CPeak DTotalPeak APeak BPeak CPeak DTotal
Area (a.u.) 309.7 243.4 181.9 186.9 921.9 31.0 182.4 106.0 159.9 479.3 
Percentage (%) 33.6 26.4 19.7 20.3 100.0 6.5 38.0 22.1 33.4 100.0 
SamplesSW-A
SW-B
Organic fractionsPeak APeak BPeak CPeak DTotalPeak APeak BPeak CPeak DTotal
Area (a.u.) 309.7 243.4 181.9 186.9 921.9 31.0 182.4 106.0 159.9 479.3 
Percentage (%) 33.6 26.4 19.7 20.3 100.0 6.5 38.0 22.1 33.4 100.0 

The percentages contributed by Peaks B, C and D of SW-B are larger than those of SW-A. These peaks contain mainly humic substances, which feature molecules with aromatic structures and conjugated C = C double bonds. The organic fraction absorbs more UV light per unit concentration of DOC, i.e. have a higher SUVA value, than the non-humic substances (Edzwald 1993). The source water of the treatment plant is from a eutrophic lake. March is in the middle of the dry season in Taiwan, while November is in the later period of the wet season. It is thus speculated that the high NPDOC and high biopolymer organic content of SW-A are due to the biological activity in the lake, e.g., algae respiration and decay processes. The low NPDOC but high humic substances content in the SW-B are due to the watershed runoff, as the DOC associated with allochthonous origin is composed of aquatic humic matter (Edzwald & Tobiason 2011).

The NOM organic fraction removal by UF membrane filtration and alum coagulation

Four types of UF membrane made from different materials, namely PVDF, CA, PVC, and PS, are used to treat the SW-A. The areas of the resolved peaks of the HPSEC-OCD chromatograms by peak-fitting of both feed water and permeate are shown in Table 4. The percentage reduction in total area and separated peaks between feed water and permeate are also listed. The results show that the total area reduction of all membranes was about 20%. This is within the range of DOC removal, 14–49%, reported in the literature for surface water treatment by loose UF membranes (MWCO ≥ 60 kDa) (Sillanpää et al. 2014).

Table 4

The organic fraction removal by various UF membranes using HPSEC-OCD combined with peak-fitting

Permeate
Organic fractionsFeed water (SW-A)PVDFCAPVCPS
Peak A 309.7 143.0 (18.1§118.1 (20.8§108.6 (21.8§165.0 (15.7§
Peak B 243.4 220.0 (2.5§240.2 (0.3§242.7 (0.1§219.0 (2.6§
Peak C 181.9 170.0 (1.3§180.4 (0.2§173.3 (0.9§171.0 (1.2§
Peak D 186.9 187.0 (0.0§181.5 (0.6§188.9 ( − 0.2§183.0 (0.4§
Total 921.9 720.0 (21.9*) 720.2 (21.9*) 713.5 (22.6*) 738.0 (19.9*) 
Permeate
Organic fractionsFeed water (SW-A)PVDFCAPVCPS
Peak A 309.7 143.0 (18.1§118.1 (20.8§108.6 (21.8§165.0 (15.7§
Peak B 243.4 220.0 (2.5§240.2 (0.3§242.7 (0.1§219.0 (2.6§
Peak C 181.9 170.0 (1.3§180.4 (0.2§173.3 (0.9§171.0 (1.2§
Peak D 186.9 187.0 (0.0§181.5 (0.6§188.9 ( − 0.2§183.0 (0.4§
Total 921.9 720.0 (21.9*) 720.2 (21.9*) 713.5 (22.6*) 738.0 (19.9*) 

*Percentage reduction in total peak area between feed water and permeate.

§Percentage reduction in each peak = .

Further, comparing the area reduction between different peaks, Table 4 clearly shows that Peak A, which represents biopolymers, such as polysaccharides and protein, has a much larger percentage reduction than the other peaks. The organic fraction was probably adsorbed by the membranes, consistent with earlier reports in the literature. For example, Kennedy et al. (2005) studied the interaction between the fractional components of NOM in surface water and a hydrophilic PES/PVC hollow fiber ultrafiltration membrane. The results from LC (liquid chromatography) OCD analysis of the UF feed and permeate showed that among all the organic fractions, only polysaccharides were rejected by the UF membrane. Halle et al. (2009) studied fouling control when using a hollow fiber PVDF UF membrane to treat surface water for drinking water production. Based on LC-OCD analysis of the UF membrane influent and effluent, they reported that only the biopolymer fraction was substantially rejected by the membrane. The biopolymer fraction was also reported to be the most significantly reduced organic fraction by a UF membrane in a number of wastewater reclamation studies (Henderson et al. 2011; Zheng et al. 2012). It can also be noticed that PS has the lowest reduction, while PVC has the highest. This is consistent with the largest and the lowest MWCO, i.e., 200 kDa and 50 kDa for PS and PVC, respectively. In general, NOM removal is reported to increase as the MWCO, or pore size, decreases (Cho et al. 2000).

Table 5 lists the areas of Peaks A–D (from peak-fitting) and their sums of the source water and the supernatants from coagulation under various alum dosages, namely 0.5, 1, 2, 4, 8, and 16 mg/L (as Al), and the percentage reduction in the area of each peak, as compared to that of the source water. Firstly, it can be seen that the areas of all peaks decrease with increasing coagulant dosage. Secondly, Table 5 reveals that Peak B, which represents humic substances, has a much higher reduction than all other peaks. For example, with an alum dosage of 16 mg/L, as Al, the Peak B reduction was 21.4%, while the reduction for the other three peaks was only around 4.5%. Humic substances feature molecules with an aromatic structure and conjugated C = C double bonds, and thus they have a higher SUVA value and are more hydrophobic than the non-humic substances. It has been well-established in the literature that humic substances are more amenable to removal by coagulation than non-humic substances (Edzwald 1993; Chow et al. 2009).

Table 5

The organic fraction removal by coagulation under various alum dosages using HPSEC-OCD combined with peak-fitting

Treated water (alum dosage, mg/L as Al)
Organic fractionsSource water (SW-B)0.5124816
Peak A 31.0 29.5 (0.3§28.3 (0.6§21.3 (2.0§14.7 (3.4§10.1 (4.4§7.3 (4.9§
Peak B 182.4 177.0 (1.1§173.4 (1.9§151.7 (6.4§121.2 (12.8§84.7 (20.4§79.8 (21.4§
Peak C 106.0 104.2 (0.4§103.6 (0.5§100.0 (1.3§88.7 (3.6§89.8 (3.4§84.3 (4.5§
Peak D 159.9 145.0 (3.1§142.8 (3.6§141.6 (3.8§133.6 (5.5§129.5 (6.3§139.8 (4.2§
Total 479.3 455.7 (4.9*) 448.1 (6.5*) 414.6 (13.5*) 358.2 (25.3*) 314.1 (34.5*) 311.2 (35.1*) 
Treated water (alum dosage, mg/L as Al)
Organic fractionsSource water (SW-B)0.5124816
Peak A 31.0 29.5 (0.3§28.3 (0.6§21.3 (2.0§14.7 (3.4§10.1 (4.4§7.3 (4.9§
Peak B 182.4 177.0 (1.1§173.4 (1.9§151.7 (6.4§121.2 (12.8§84.7 (20.4§79.8 (21.4§
Peak C 106.0 104.2 (0.4§103.6 (0.5§100.0 (1.3§88.7 (3.6§89.8 (3.4§84.3 (4.5§
Peak D 159.9 145.0 (3.1§142.8 (3.6§141.6 (3.8§133.6 (5.5§129.5 (6.3§139.8 (4.2§
Total 479.3 455.7 (4.9*) 448.1 (6.5*) 414.6 (13.5*) 358.2 (25.3*) 314.1 (34.5*) 311.2 (35.1*) 

*Percentage reduction in total peak area between source water and treated water.

§Percentage reduction in each peak = .

Figure 5 shows a comparison of NOM fraction removal by UF membrane and alum coagulation. The horizontal axis is the percentage removal of the area of Peak A, while the vertical axis is the percentage removal of the sum of the areas of Peaks B, C, and D. The solid line drawn at a 1:1 slope represents an equal percentage removal of these two NOM groups. The figure clearly illustrates that Peak A or the biopolymer fraction was preferably removed by the UF membrane over other fractions, while the reverse was true for alum coagulation.
Figure 5

Comparison of NOM fraction removal by UF membrane and coagulation.

Figure 5

Comparison of NOM fraction removal by UF membrane and coagulation.

The correlation between DBP precursors and NOM fractions based on HPSEC-OCD

Table 6 shows the concentration of DBP precursors, expressed as THMFP and HAAFP, in the feed water (SW-A) and the permeate of the four UF membranes. The percentage reductions between feed water and permeate are also listed. These show that PVC has the largest reduction in both THMFP and HAAFP, PS has the lowest, while PVDF and CA are in between. This is consistent with the Peak A area reduction based on HPSEC-OCD chromatograms combined with peak-fitting, as shown in Table 4. Further, Table 6 also shows that, except for PVDF, the three other UF membranes have greater THMFP reductions than that found with HAAFP.

Table 6

The removal of disinfection by-products precursors by various UF membranes

Permeate
DBPFP (μg/L)Feed water (SW-A)PVDFCAPVCPS
THMFP 313.7 253.6 (19.2*) 164.2 (47.7*) 147.6 (52.9*) 274.4 (12.5*) 
HAAFP 299.6 241.6 (19.4*) 236.3 (21.1*) 219.0 (26.9*) 272.2 (9.1*) 
Permeate
DBPFP (μg/L)Feed water (SW-A)PVDFCAPVCPS
THMFP 313.7 253.6 (19.2*) 164.2 (47.7*) 147.6 (52.9*) 274.4 (12.5*) 
HAAFP 299.6 241.6 (19.4*) 236.3 (21.1*) 219.0 (26.9*) 272.2 (9.1*) 

*Percentage removal.

Table 7 shows the THMFP and HAAFP of the source water (SW-B) and the supernatants from coagulation under various alum dosages, as well as the percentage reduction between the source and treated water. Firstly, the results show that the percentage reduction in both DBP precursors increased with increasing alum dosage. Secondly, under a specific dosage, HAAFP has a greater reduction than THMFP.

Table 7

The removal of disinfection by-product precursors by coagulation under various alum dosages

Treated water (alum dosage, mg/L as Al)
DBPFP (μg/L)Source water (SW-B)0.5124816
THMFP 911.6 826.4 (9.3*) 779.7 (14.5*) 720.2 (21.0*) 672.6 (26.2*) 517.5 (43.2*) 420.2 (53.9*) 
HAAFP 257.8 205.4 (20.3*) 183.0 (29.0*) 169.1 (34.4*) 148.6 (42.4*) 107.5 (58.3*) 94.8 (63.2*) 
Treated water (alum dosage, mg/L as Al)
DBPFP (μg/L)Source water (SW-B)0.5124816
THMFP 911.6 826.4 (9.3*) 779.7 (14.5*) 720.2 (21.0*) 672.6 (26.2*) 517.5 (43.2*) 420.2 (53.9*) 
HAAFP 257.8 205.4 (20.3*) 183.0 (29.0*) 169.1 (34.4*) 148.6 (42.4*) 107.5 (58.3*) 94.8 (63.2*) 

*Percentage removal.

Figure 6 shows a comparison of THMFP and HAAFP removal by UF membrane and alum coagulation. The solid line drawn at a 1:1 slope represents an equal percentage removal of the two parameters. The figure shows that more THMFP was removed by the UF membrane than HAAFP, while the reverse was true for alum coagulation. This can be understood in combination with Figure 5, which indicates that Peak A was preferably removed by the UF membrane over other fractions, while the reverse is true for alum coagulation. It is thus conjectured that NOM in Peak A, which mainly consists of hydrophilic biopolymers, is the major precursor for THMs, while the NOM in Peak B, which represents hydrophobic humic substances, is the main precursor for HAAs.
Figure 6

Comparison of THMFP and HAAFP removal by UF membrane and coagulation.

Figure 6

Comparison of THMFP and HAAFP removal by UF membrane and coagulation.

CONCLUSIONS

Based on the results from the HPSEC-OCD chromatograms of the source water and treated water from UF membrane filtration and alum coagulation, it is shown that, among all NOM fractions, the UF membrane mainly removed the aliphatic biopolymer fraction, while alum coagulation mainly removed the humic substances fraction. The DBP precursors analysis shows that more THMFP was removed by the UF membrane than HAAFP, while the reverse was true for alum coagulation. It is thus conjectured that the aliphatic biopolymer fraction is the major precursor for THMs, while the humic substances fraction is the major precursor for HAAs.

ACKNOWLEDGEMENTS

The financial support provided for this study by the Ministry of Science and Technology, Taiwan (Contract No. 101-2221-E-006-156-MY3) is greatly appreciated.

REFERENCES

REFERENCES
Allpike
B. P.
Heitz
A.
Joll
C. A.
Kagi
R. I.
Abbt-Braun
G.
Frimmel
F. H.
Brinkmann
T.
Her
N.
Amy
G.
2005
Size exclusion chromatography to characterize DOC removal in drinking water treatment
.
Environmental Science and Technology
39
(
7
),
2334
2342
.
APHA, AWWA & WEF
1998
Standard Methods for the Examination of Water and Wastewater
, 20th edn.
APHA
,
Washington, DC
,
USA
.
Chen
C.
Zhang
X. J.
Zhu
L. X.
Liu
J.
He
W. J.
2008
Disinfection by-products and their precursors in a water treatment plant in North China: seasonal changes and fraction analysis
.
Science of the Total Environment
397
(
1
),
140
147
.
Chow
C. W.
Fabris
R.
van Leeuwen
J.
Wang
D.
Drikas
M.
2008
Assessing natural organic matter treatability using high performance size exclusion chromatography
.
Environmental Science and Technology
42
(
17
),
6683
6689
.
Chow
C. W.
van Leeuwen
J. A.
Fabris
R.
Drikas
M.
2009
Optimised coagulation using aluminium sulfate for the removal of dissolved organic carbon
.
Desalination
245
(
1
),
120
134
.
Christman
R. F.
Norwood
D. L.
Millington
D. S.
Johnson
J. D.
Stevens
A. A.
1983
Identity and yields of major halogenated products of aquatic fulvic acid chlorination
.
Environmental Science and Technology
17
(
10
),
625
628
.
Croue
J. P.
Korshin
G. V.
Benjamin
M. M.
2000
Characterization of Natural Organic Matter in Drinking Water
.
American Water Works Association
,
Denver, CO,
USA
.
Edzwald
J. K.
1993
Coagulation in drinking water treatment: particles, organics and coagulants
.
Water Science and Technology
27
(
11
),
21
35
.
Edzwald
J. K.
Tobiason
J. E.
2011
Chemical principles, source water composition, and watershed protection
. In:
Water Quality & Treatment: A Handbook on Drinking Water
(
Edzwald
J. K.
, ed.), 6th edn.
AWWA, McGraw-Hill
,
New York
.
Halle
C.
Huck
P. M.
Peldszus
S.
Haberkamp
J.
Jekel
M.
2009
Assessing the performance of biological filtration as pretreatment to low pressure membranes for drinking water
.
Environmental Science and Technology
43
(
10
),
3878
3884
.
Henderson
R. K.
Subhi
N.
Antony
A.
Khan
S. J.
Murphy
K. R.
Leslie
G. L.
Chen
V.
Stuetz
R. M.
Le-Clech
P.
2011
Evaluation of effluent organic matter fouling in ultrafiltration treatment using advanced organic characterisation techniques
.
Journal of Membrane Science
382
(
1
),
50
59
.
Jacangelo
J. G.
Denarco
J.
Douglas
M.
1995
Selected processes for removing NOM: an overview
.
Journal American Water Works Association
87
(
1
),
64
77
.
Kennedy
M. D.
Chun
H. K.
Quintanilla Yangali
V. A.
Heijman
B. G. J.
Schippers
J. C.
2005
Natural organic matter (NOM) fouling of ultrafiltration membranes: fractionation of NOM in surface water and characterisation by LC-OCD
.
Desalination
178
(
1–3
),
73
83
.
Lamsal
R.
Montreuil
K. R.
Kent
F. C.
Walsh
M. E.
Gagnon
G. A.
2012
Characterization and removal of natural organic matter by an integrated membrane system
.
Desalination
303
,
12
16
.
Matilainen
A.
Gjessing
E. T.
Lahtinen
T.
Hed
L.
Bhatnagar
A.
Sillanpää
M.
2011
An overview of the methods used in the characterisation of natural organic matter (NOM) in relation to drinking water treatment
.
Chemosphere
83
(
11
),
1431
1442
.
Sillanpää
M.
Metsämuuronen
S.
Mänttäri
M.
2014
Membranes
. In:
Natural Organic Matter in Water: Characterization and Treatment Method
(
Silaanpää
M.
, ed.).
Butterworth-Heinemann
,
Oxford, UK
.
Singer
P. C.
2006
DBPs in drinking water: additional scientific and policy considerations for public health protection
.
Journal American Water Works Association
98
(
10
),
73
80
.