A variety of surface waters used for drinking water sources were collected from different parts of Japan to investigate a correlation between the concentration of hydrophilic biopolymer (e.g. proteins and polysaccharides) in the feed water and membrane fouling in microfiltration. Hollow-fiber membranes made from polyvinylidene fluoride with a nominal pore size of 0.1 μm were used in the series of experiments involving the constant-flow mode of operation with automatic periodical backwashing. Easily available indexes of water quality, such as dissolved organic carbon, UV absorbance, Ca concentration and turbidity could not explain the degree of fouling encountered in the filtration tests. In contrast, a clear correlation between the concentrations of biopolymers determined by liquid chromatography with organic carbon detection (LC-OCD) and membrane fouling was confirmed in this study. The concentrations of humics exhibited a weak correlation. The impacts of seasonal variation of feed water and coagulant dosage on membrane fouling were also explained well by biopolymer concentrations. Concentrations of biopolymers can be a useful indicator of the fouling potential of feed water in microfiltration.

INTRODUCTION

The application of low-pressure membranes (microfiltration (MF) and ultrafiltration (UF) membranes) in water treatment remains hindered by membrane fouling, that is, the deterioration of permeabilities of membranes. Natural organic matter (NOM) plays an important role in the evolution of membrane fouling. However, NOM comprises a variety of organic compounds including humic substances, proteins and polysaccharides. In addition, compositions of NOM vary from time-to-time and place-to-place. Early studies dealing with membrane fouling by NOM suggested that the hydrophobic fraction of NOM (i.e. humic substances) was important (Jucker & Clark 1994; Yuan & Zydney 1999,, 2000). In contrast, a significant contribution of hydrophilic NOM (i.e. polysaccharides and proteins) in membrane fouling was indicated in later studies (Kimura et al. 2004; Lee et al. 2004). It is still not clear which part of NOM is eventually important in the evolution of membrane fouling. There were several weak points in previous studies that addressed membrane fouling in low-pressure membranes. Some studies examined limited numbers of waters, whereas, others conducted their experiments with apparatuses that were completely different from full-scale membrane plants (e.g. a flat-sheet membrane cell operated under a constant pressure versus hollow-fiber membranes operated with a constant flow rate). As a result, there is no available consensus regarding the important fraction of NOM in the evolution of membrane fouling in low-pressure membranes.

Recently, analysis of aquatic organic matter by liquid chromatography with organic carbon detection (LC-OCD) has become popular and has been used for the investigation of membrane fouling in MF/UF membranes (Hallé et al. 2009; Myat et al. 2012; Tian et al. 2013). The presence of biopolymers, which have large molecular weights and less ultraviolet sensitivity, can be examined by LC-OCD (Huber et al. 2011). Several studies suggested significant correlations between membrane fouling in MF/UF membranes and biopolymer concentrations (Hallé et al. 2009; Tian et al. 2013); however, those studies also had the drawbacks (i.e. limited numbers of waters examined and/or inappropriate configuration of the membranes). In our previous study, the correlation between membrane fouling and biopolymer concentrations in the feed water was demonstrated on the basis of experimental results using multiple surface waters and hollow-fiber membranes operated under a constant flow rate with routine backwashing (Kimura et al. 2014). The principal aim of this study was to confirm the correlation by examining a number of different types of surface waters. Samples were collected from different parts of Japan to examine waters with ample diversities. In addition, the influence of the seasonal variations of the water quality and the coagulant dose on membrane fouling was investigated in terms of the biopolymer concentrations.

MATERIAL AND METHODS

Sample collections

Six anonymous surface waters (designated River A, B, C, D, E and F hereafter) were collected from different parts of Japan. All of the waters studied are used as drinking water sources. Regarding Rivers A, B and C, the samples were repeatedly collected in different seasons. Table 1 summarizes the quality of waters examined in this study. Approximately 200 L of each of the samples was shipped to the university laboratory under cooling. Upon arrival, sodium azide (2 mg/L) was added to the samples to suppress the activity of microorganisms. The samples were stored in a refrigerator until the experiments were performed. Prior to the experiments, samples were placed in the laboratory for 24 h so that the experiments were performed under a controlled room temperature (20 °C).

Table 1

Quality of the feed water examined in this study

 DOC (mg/L)Turbidity (TU)SUVA (L/mg/m)Ca (mg/L)
River A (spring) 0.7 7.9 4.0 3.6 
River A (summer) 0.8 0.1 3.7 7.5 
River A (autumn) 0.9 1.0 3.5 7.9 
River A (winter) 0.9 0.4 4.0 8.7 
River B (spring) 2.2 23.4 4.2 7.8 
River B (summer) 2.0 15.6 4.6 9.5 
River B (autumn) 1.9 1.4 3.7 7.5 
River B (winter) 1.8 1.5 4.0 11.8 
River C (summer) 1.9 1.8 3.1 8.0 
River C (autumn) 2.2 3.0 3.6 8.1 
River C (winter) 1.4 3.1 2.3 12 
River D 1.4 7.2 2.8 4.3 
River E 1.3 3.4 3.2 12.3 
River F 2.0 1.6 2.7 34.3 
 DOC (mg/L)Turbidity (TU)SUVA (L/mg/m)Ca (mg/L)
River A (spring) 0.7 7.9 4.0 3.6 
River A (summer) 0.8 0.1 3.7 7.5 
River A (autumn) 0.9 1.0 3.5 7.9 
River A (winter) 0.9 0.4 4.0 8.7 
River B (spring) 2.2 23.4 4.2 7.8 
River B (summer) 2.0 15.6 4.6 9.5 
River B (autumn) 1.9 1.4 3.7 7.5 
River B (winter) 1.8 1.5 4.0 11.8 
River C (summer) 1.9 1.8 3.1 8.0 
River C (autumn) 2.2 3.0 3.6 8.1 
River C (winter) 1.4 3.1 2.3 12 
River D 1.4 7.2 2.8 4.3 
River E 1.3 3.4 3.2 12.3 
River F 2.0 1.6 2.7 34.3 

DOC: dissolved organic carbon; SUVA: specific ultraviolet absorbance.

Membrane filtration apparatus

Mini-size hollow-fiber MF membrane modules (surface area: 0.0128 m2) were fabricated in the laboratory and used in the series of experiments. The membranes used in this study were manufactured by Asahi Kasei (Tokyo, Japan). The material and nominal pore size of the membranes were polyvinylidene fluoride and 0.1 μm, respectively. A membrane module was immersed in a filtration tank (volume: 160 cm3) and the constant-flow mode of filtration was performed by using a peristaltic pump. The membrane flux was fixed at 63 LMH, which is in the range of typical membrane fluxes set in full-scale membrane units, in all experiments. The backwash was automatically performed at 94 LMH of flux every 30 min of operation. Each backwash was continued for 30 s and accompanied with air scouring. The trans-membrane pressure (TMP) in each experiment was monitored by a digital pressure meter (Nagano Keiki, Tokyo, Japan). In this study, new membranes were used in each test.

For all samples, the membrane filtration experiments were performed with pre-coagulated waters as well as raw waters. Pre-coagulation was performed with polyaluminum chlorides. The dose of the coagulant was fixed at 2 mg-Al/L unless stated. The mixing of waters and coagulant was performed under G value of 100 s−1 for 2 min. The coagulated waters were introduced to the filtration tank after brief sedimentation (40 min).

Analytical methods

The concentrations of dissolved organic carbon (DOC), biopolymers and humic substances were determined using an LC-OCD system (Model 8, DOC-LABOR, Karlsruhe, Germany). UV absorbance at 254 nm was measured by a spectrophotometer (UV-1800, Shimadzu, Kyoto, Japan). Before UV absorbance and DOC measurements were performed, the samples were filtered using 0.45-μm PTFE membranes. The turbidity measurements were performed using a turbidity meter (SEP-PT-706D, Mitsubishi Chemical Analytech, Mie, Japan). The concentrations of calcium and aluminum were determined by inductively coupled plasma-atomic emission spectroscopy (ICPS-7500, Shimadzu, Kyoto, Japan).

RESULTS AND DISCUSSION

Seasonal variation of membrane fouling caused by surface waters

Figure 1 shows the TMP increases in the experiments examining waters collected from River A. As stated in the previous section, samples were collected from River A in different seasons to investigate the seasonal variation of membrane fouling. The data shown in Figure 1 represent experiments conducted without pre-coagulation. In all experiments, the TMP increases were nearly linear. As demonstrated in Figure 1, the rates of the TMP increases were significantly different, depending on the seasons in which the samples were collected. For River A, membrane fouling was the most severe in summer and was the least in spring. Seasonal variations of membrane fouling caused by surface waters were also examined in Rivers B and C. The membrane fouling behaviors caused by the samples collected from Rivers B and C were also significantly different depending on the seasons (data not shown), although different seasons were accompanied with severe fouling. The sample collected in spring caused the most severe membrane fouling with the sample from River B, whereas for River C the sample collected in autumn caused the most severe membrane fouling. The water quality indices that are quickly available (summarized in Table 1) do not explain the seasonal variations in the membrane fouling caused by surface waters. Our previous study also demonstrated the poor ability of the conventional water quality indices, such as DOC for prediction of membrane fouling in MF (Kimura et al. 2014). This study confirmed that the conventional water quality indices were not useful for the prediction of membrane fouling in MF.

Figure 1

TMP increases in the experiments with samples collected from River A (without pre-coagulation).

Figure 1

TMP increases in the experiments with samples collected from River A (without pre-coagulation).

Analysis of extracted foulants by LC-OCD

Figure 2 compares the LC-OCD chromatograms obtained for the sample collected from River C (winter) and the foulants extracted from the fouled membranes at the termination of the corresponding operation. The extraction of foulants was performed with sodium hydroxide (pH 12) for 24 h at the termination of the continuous filtration. A peak of humics (retention time of approximately 45 min) was dominant in the feed water, whereas a peak of biopolymers (retention time of approximately 30 min) was significant in the foulants. This trend was observed for all samples and foulants (data not shown). The dominance of hydrophilic organic compounds in foulants in MF/UF membranes treating surface waters was also reported in our previous studies, in which powerful analytical tools (e.g. 13C-NMR) were used (Yamamura et al. 2007).

Figure 2

LC-OCD chromatograms of (a) the sample collected from River C (winter) and (b) the extracted foulants from the fouled membrane filtering that water.

Figure 2

LC-OCD chromatograms of (a) the sample collected from River C (winter) and (b) the extracted foulants from the fouled membrane filtering that water.

Mitigation of membrane fouling through pre-coagulation

In this study, all of the samples were filtered with the hollow-fiber MF membranes after pre-coagulation as well. The extents of mitigation of membrane fouling through pre-coagulation considerably differed depending on the samples. Residual aluminum did not exhibit a clear correlation with membrane fouling (data not shown). Owing to the limitation of the space available, only clear examples of poor mitigation (River B (winter)) and substantial mitigation (River C (winter)) are shown in Figure 3. In the case of the River B (winter) sample, rather limited mitigation of membrane fouling was observed with pre-coagulation. The increase of TMP was considerably suppressed with pre-coagulation in the case of the River C (winter), although that water exhibited almost an identical TMP increase as that of River B (winter) water without pre-coagulation. Pre-coagulation substantially reduced the concentration of biopolymers in the sample of River C (winter). In the samples with which pre-coagulation substantially reduced concentrations of biopolymers, the effect of pre-coagulation on mitigation of membrane fouling became significant.

Figure 3

TMP increases in the experiments where the samples collected from River B (winter) and River C (winter) were filtered. The hollow marks represent the TMP increases in the experiments conducted without pre-coagulation, whereas the solid marks represent the TMP increases in the experiments conducted with pre-coagulation.

Figure 3

TMP increases in the experiments where the samples collected from River B (winter) and River C (winter) were filtered. The hollow marks represent the TMP increases in the experiments conducted without pre-coagulation, whereas the solid marks represent the TMP increases in the experiments conducted with pre-coagulation.

In this study, the effect of coagulant dose on the mitigation of membrane fouling was also investigated. Figure 4 shows the result of experiments with the samples collected from River B (autumn) and River D, in which the coagulant dose was varied. In the case of the River B (autumn) sample, the increase of TMP was slowed with the increase of coagulant dose. Biopolymer concentrations were also reduced with the gradual increase of coagulant. In contrast, an increase of coagulant dose did not exhibit mitigation of membrane fouling in the case of River D. In that experiment, biopolymer concentrations did not change, regardless of the coagulant dose. These results also imply that biopolymers were the major contributor to membrane fouling in this study.

Figure 4

TMP increases in the experiments for which the coagulant dose was varied. The left panel represents the data obtained with the sample collected from River B (autumn) and the right panel represents the data obtained with the sample collected from River D.

Figure 4

TMP increases in the experiments for which the coagulant dose was varied. The left panel represents the data obtained with the sample collected from River B (autumn) and the right panel represents the data obtained with the sample collected from River D.

Correlations between organic compound fractions and membrane fouling

Figures 5 and 6 show the correlations between the rate of TMP increase in each experiment (i.e. membrane fouling) and the concentrations of humics and biopolymer in the feed water, respectively. The data shown in Figures 5 and 6 include experiments both with and without pre-coagulation. The concentrations of humics exhibited a weak correlation with membrane fouling. Analyses of the distribution of membrane fouling (reversible versus irreversible fouling) demonstrated that the membrane fouling observed in this study was mostly irreversible fouling: wiping the membrane surface with a sponge at the termination of the operations slightly restored the permeability. The physical cleaning used in this study (i.e. routine backwashing) was found to be quite effective. The weak correlation between the concentrations of humics and membrane fouling shown in Figure 5 implies that humics were not the major player in the evolution of irreversible fouling in this study; however, it dominated in the raw surface waters (see Figure 2). Early studies suggested that humics were the major players in membrane fouling (Jucker & Clark 1994; Yuan & Zydney 1999, 2000). In those studies, reversible fouling caused by humics might have been dominant: the efficiency of physical cleaning in those studies might have been low.

Figure 5

Correlation between humics concentrations in the feed and TMP increase.

Figure 5

Correlation between humics concentrations in the feed and TMP increase.

Figure 6

Correlation between biopolymer concentrations in the feed and TMP increase.

Figure 6

Correlation between biopolymer concentrations in the feed and TMP increase.

As stated before, effectiveness of pre-coagulation significantly varied depending on the samples. However, the concentrations of biopolymer exhibited a very clear correlation with membrane fouling, regardless of the effectiveness of pre-coagulation. The biopolymer concentration could explain the fouling for coagulated water as well. In addition, the data with coagulation and the data without coagulation seem to exhibit a single trend. This result strongly suggests that biopolymer concentration determined by LC-OCD is a good indicator for the prediction of membrane fouling in MF of surface waters.

CONCLUSIONS

In this study, a variety of surface waters used as drinking water sources were collected from different parts of Japan to investigate membrane fouling in MF. Conventional water quality indices, such as DOC, could not explain the degree of membrane fouling. The biopolymer concentration determined by LC-OCD exhibited a very good correlation (R2 = 0.82) with membrane fouling. The influence of seasonal variation of water quality in surface waters and pre-coagulation on membrane fouling could also be explained well by the biopolymer concentrations. In contrast, the correlation between humics that had been suggested as major foulants and fouling was weak (R2 = 0.45). Concentrations of biopolymers determined by LC-OCD can be a useful indicator of the fouling potential of feed water in MF.

REFERENCES

REFERENCES
Hallé
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
.
Kimura
K.
Hane
Y.
Watanabe
Y.
Amy
G.
Ohkuma
N.
2004
Irreversible membrane fouling during ultrafiltration of surface water
.
Water Research
38
(
14–15
),
3431
3441
.
Yuan
W.
Zydney
A. L.
1999
Humic acid fouling during microfiltration
.
Journal of Membrane Science
157
,
1
12
.
Yuan
W.
Zydney
A. L.
2000
Humic acid fouling during ultrafiltration
.
Environmental Science and Technology
34
(
23
),
5043
5050
.