Environ. Sci. Technol. 2001, 35, 3988-3999
Disinfection Byproduct Relationships and Speciation in Chlorinated Nanofiltered Waters S H A N K A R A R A M A N C H E L L A M * ,† A N D STUART W. KRASNER‡ Department of Civil and Environmental Engineering, University of Houston, Houston, Texas 77204-4003, and Water Quality Division, Metropolitan Water District of Southern California, 700 Moreno Avenue, La Verne, California 91750-3399
The formation and speciation of disinfection byproducts (DBPs) resulting from chlorination of nanofilter permeates obtained from various source water locations and membrane types are examined. Specific ultraviolet absorbance and bromide utilization are shown to decrease following nanofiltration. Both dissolved organic carbon (DOC) concentration and ultraviolet absorbance at 254 nm were found to correlate strongly with trihalomethane (THM), haloacetic acid (HAA), and total organic halide (TOX) concentrations in chlorinated nanofilter permeates, suggesting that they can be employed as surrogates for DBPs in nanofiltered waters. Because smooth curves were obtained for individual THM and HAA species as well as bromine and chlorine incorporation into THMs and HAAs as a function of Br-/DOC molar ratio, it is likely that mole fractions of these DBPs are more strongly influenced by chlorination conditions, Br-, and DOC concentrations than NOM source and membrane type. Mole fractions of mono-, di-, and trihalogenated HAAs were found to be independent of Br-/DOC. Even at a very low Br-/DOC of 2.9 µM/mM, the mixed bromochloro- and tribromoacetic acids constituted 20% of total HAAs on a molar basis. This increased to ∼50% as Br-/DOC increased to ∼25 µM/ mM or more, proving that a large fraction of HAAs may not be covered under existing federal regulations. Total THM and HAA9 concentrations decreased in permeate waters with increasing Br-/DOC suggesting that nanofilter permeates are limited with respect to DBP precursors.
Introduction Chlorination of drinking waters containing organic carbon and bromide ion results in the formation of potentially carcinogenic, teratogenic, and mutagenic byproducts, including trihalomethanes (THMs) and haloacetic acids (HAAs). Because stable analytical standards and methods for bromodichloroacetic acid (BrCl2AA), dibromochloroacetic acid (Br2ClAA), and tribromoacetic acid (Br3AA) were only recently developed, most studies have not measured all nine HAAs containing bromine and chlorine. Consequently, only five HAAs: monochloro-, dichloro-, trichloro-, monobromo-, and * Corresponding author phone: (713)743-4265; fax: (713)743-4260; e-mail:
[email protected]. † University of Houston. ‡ Metropolitan Water District of Southern California. 3988
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 35, NO. 19, 2001
dibromoacetic acid (ClAA, Cl2AA, Cl3AA, BrAA, and Br2AA, respectively) are currently regulated under the Disinfectants/ Disinfection By-Products (D/DBP) Rule (1). A sixth HAA (viz. bromochloroacetic acid, BrClAA) was included in the Information Collection Rule (ICR) (2). An aggregate water quality parameter, total organic halide (TOX) quantifies the total concentration of halogenated organic compounds in water and is a measure of both DBPs and synthetic organic contaminants. Even though the mixed bromochloro HAA species, Br3AA, and TOX are not currently regulated, it is crucial to examine their occurrence and treatability in drinking water because of their potential adverse human health effects. The very limited number of studies that have reported all nine HAAs containing Br and Cl show that the nonregulated HAAs (BrClAA, Br2ClAA, BrCl2AA, Br3AA) can comprise a substantial fraction of the total molar concentration of HAAs (3-6). Almost all prior laboratory-scale investigations of HAA9 speciation have been conducted by holding the dissolved organic carbon (DOC) concentration constant in raw water or commercial or extracted humic acids and artificially spiking various amounts of the bromide ion (3-5). In contrast, nanofiltration (NF), which is one of the most promising methods for DBP control, inherently alters the Br-/DOC ratio between feed and permeate waters, thereby changing DBP speciation upon chlorination (6). Most NF studies for DBP control have been site-specific and have typically reported only reductions in total mass concentrations of specific DBPs (e.g. refs 7-10). Even though previous research has demonstrated that NF achieves high removals of DBP precursors (6-8), changes in reactivity to chlorine and specific yield have not yet been assessed. A better understanding of the commonalities in DBP speciation in permeate waters may reduce the need for and extent of site-specific studies. Additionally, it may provide a mechanistic basis for DBP formation leading to improved control strategies as more NF plants are installed. Further, under the current D/DBP rule, HAAs are regulated at lower levels than THMs. However, if free chlorine is used as the primary disinfectant at low bromide ion concentrations and pH, HAA mass concentrations have been shown to be higher than THM mass concentrations (11). Thus, a generalized method of interpreting DBP speciation may allow a more equitable regulation of these undesirable byproducts of water chlorination health effects withstanding. The occurrence of and correlations between various classes of DBPs have been reported predominantly for conventionally treated waters (11-14). Ultraviolet (UV) absorbance at 254 nm (UV254) and 1-cm path length and DOC have been reported to be good surrogates for THM and HAA5 formation during conventional treatment (12, 15, 16). However, surrogates for THMs and HAA9 in NF permeate waters have not yet been evaluated. If simple surrogates for DBPs in nanofilter permeates are identified, the need for expensive and time-consuming DBP analyses could potentially be limited to regulatory compliance and more frequent monitoring of these surrogates undertaken. This will allow better process control thereby reducing potentially adverse health effects of various DBPs. The objectives of this paper are to investigate possible changes in reactivity of natural organic matter (NOM) following NF. Relationships between DOC, UV254, chlorine consumption, THMs, HAAs, and TOX concentrations in chlorinated NF feed and permeate waters are evaluated. Universalities in THM and HAA formation and speciation are also explored with particular emphasis on the extent of 10.1021/es010775n CCC: $20.00
2001 American Chemical Society Published on Web 09/05/2001
TABLE 1. Summary of Membrane Characteristics as Specified by Respective Manufacturers designation
manufacturer
model number
composition
I II III IV
Hydranautics, San Diego, CA Koch Fluid Systems, San Diego, CA Dow FilmTec, Midland, MI Dow FilmTec, Midland, MI
NTR 7450 TFC-SR NF 200B NF 45
modified polysulfone polyamide polypiperazine polyamide
a
Molecular weight cut-off.
b
MWCOa (Da) hydrophobicity 600-800 300 200-400 400
hydrophobic not available hydrophilic hydrophobic
surface chargeb negative slightly negative highly negative negative
Reported at near neutral pH.
TABLE 2. Summary of NF Feed Water Quality and SDS Conditions designation
source water
SUVA (L mg-1 m-1)
pH
Br- (µM)
membranes employed
target SDS conditiona
A B C D E F G
Biscayne Aquifer, FL Biscayne Aquifer, FL Biscayne Aquifer, FL Biscayne Aquifer, FL Lake Meade, VA Caloosahatchee River, FL Rio Grande River, TX
3.17-3.95 3.96-4.98 2.48-3.16 3.10-4.13 1.23-1.89 1.89-2.57 1.97-2.46
7.83-8.30 7.85-8.34 7.63-8.05 7.57-8.48 6.58-7.13 7.19-7.85 7.91-8.25
1.63-2.25 1.75-1.88 0.78-2.38 0.78-1.06 0.48-0.63 1.38-3.82 2.38-7.88
I, II I, II, III I, II I, II I, IV I, II, III II, III
pH ) 8.0, 24 °C, 24 h pH ) 9.0, 24 °C, 24 h pH ) 9.0, 24 °C, 24 h pH ) 8.0, 24 °C, 24 h pH ) 7.3, 10-28 °C, 60-72 h pH ) 8.0, 24 °C, 24-72 h pH ) 8.2, 24 °C, 24-72 h
a Quarterly simulated distribution system conditions in terms of pH, temperature, and holding time. In all cases, the target free chlorine residual at the end of incubation was 1 mg/L.
bromide utilization and formation of currently unregulated mixed bromochloro HAA species. Experimental Work. All NF experiments were conducted using procedures specified by U.S. EPA for bench-scale treatment studies under the ICR (2). Cross-flow filtration experiments were conducted for four quarters over a 1-year period using a stainless steel cell fitted with feed and permeate spacers employed in commercial spiral-wound elements. A portion of the concentrate water was recycled to the membrane influent in order to maintain a manufacturer specified cross-flow velocity (∼0.1 m/s). During each quarter, constant pressure experiments were conducted at four different recoveries (30%, 50%, 70%, and 90%) using two membranes. Manufacturer specifications of various membrane characteristics are given in Table 1. The experiment at 70% recovery was run for a minimum of 78 h during which duplicate water quality samples were obtained after steadystate operation was achieved. Experimental duration at other recoveries was adjusted in order to obtain at least 4 L of permeate necessary to perform all water quality analyses including simulated distribution system (SDS) tests. One feedwater quality sample was collected during operation with each membrane. Each participating utility sent ∼350 L of water (collected prior to adding any oxidants or chemical disinfectants) during each quarter. The only NF pretreatment for groundwaters (A-D) was depth filtration using a 5 µm cartridge. Surface waters E, F, and G were coagulated (using alum) and flocculated, settled (∼3 h), and filtered using a 5 µm cartridge prior to nanofiltration. Important NF feedwater quality information following pretreatment is summarized in Table 2. SDS conditions for groundwaters A-D were not dependent on season (Table 2). However, surface waters exhibited seasonal changes in distribution system temperature and/ or residence time (Table 2). Each of the four membranes was not employed with each of the seven source waters. Therefore, the number of water quality measurements made for each membrane is different in this study. More detailed information on materials and methods employed are available in previous publications (6, 17).
Results and Discussion Specific Ultraviolet Absorbance. Specific UV absorbance (SUVA, defined as UV254 expressed in m-1 divided by the
DOC concentration expressed in mg/L) has been found to be a good surrogate for a water’s humic content (18). In addition, SUVA can provide an indication of the reactivity of NOM to form DBPs. Because these NF membranes rejected UV254 to a greater extent than DOC (6), permeate waters should have a lower SUVA than the feedwater. Figure 1a confirms this expectation for the membranes under study. These data suggest that DBP control by NF is achieved by reductions in the reactivity of NOM with chlorine in addition to the removal of precursors measured as DOC. These observations are further evaluated in Figure 1b, where specific yields of TOX, THMs, and HAA9 in feed and permeate waters are depicted in the form of box plots. (The observed THM and HAA concentrations in µg/L were converted to molar concentrations, and the yields were normalized to the carbon concentration on a molar basis; i.e., µΜ of DBPs produced per mM of DOC. TOX mass concentration in µg/L was normalized by 35.5, the atomic weight of chlorine, to obtain molar concentration.) The median reduction in TOX, THM, and HAA9 specific yields were calculated to be 18, 11, and 11% using matched-pair data. However, t-tests showed that there was no difference in the average DBP specific yields in the feed and permeate waters at the 95% significance level. Thus, reductions in SUVA by NF did not appear to significantly decrease DBP specific yield. Bromide Utilization. One method of quantifying changes in bromine substitution is the concept of bromide utilization that was originally proposed for THMs (19). This concept can be extended to include HAAs 3
bromide utilization )
∑ i × [CHCl
3-iBri]/[Br
-
]+
i)1
{[BrAA] + 2[Br2AA] + 3[Br3AA] + [BrClAA] +
[BrCl2AA] + 2[Br2ClAA]}/[Br-] (1)
where the concentrations are on a molar basis. Changes in bromide utilization between feed and permeate waters are depicted in Figure 2a. Membranes II, III, and IV achieved much greater DOC removals (>90%) than membrane I (∼70%), whereas none of the membranes removed bromide ion appreciably (6, 17). As observed in Figure 2a, NF membranes that were highly retentive to NOM VOL. 35, NO. 19, 2001 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
3989
FIGURE 1. Box plots of reduction in SUVA by various NF membranes (a) and TOX, THM, and HAA yields in feed and NF permeate waters (b). TOX mass concentrations were divided by the atomic weight of chlorine (35.5 g/mol) to obtain molar concentrations. Open circles denote the maximum and minimum values, closed circles denote the 1% and 99% percentile values, and the cross denotes the average value. The box encompasses the 25% and 75% percentiles, and the whiskers are determined by the 5% and 95% percentiles. The horizontal line inside the box is the median value. The number of observations (n) for each box are shown underneath the respective x-axis label. Median percent reductions in SUVA and DBP yields based on matched paired data (measurements made in feed and permeate waters) are shown in parentheses above appropriate boxes. The maximum and minimum SUVA values for membranes III and IV are coincident with the 1% and 99% percentiles thereby obscuring the respective symbols. allowed a smaller fraction of Br- (∼20-30% versus ∼40% on a median basis) to be incorporated into THMs and HAAs. Bromide utilization decreased by 49-63% (on a median basis) for membranes II, III, and IV permeates, whereas only a 4% change was observed for membrane I permeate. Permeates that are NOM limited possess very few sites for bromine substitution. Because hypobromous acid (HOBr) is a more powerful halogenating agent than hypochlorous acid (HOCl), the brominated DBPs are formed first with bromine consuming the available sites on NOM. In precursor limited permeates, bromide utilization is reduced because excess Br- cannot react once available reactive sites on NOM are occupied. Similar observations have been made for THMs where bromide utilization decreased from 44 to 30% following granular activated carbon treatment (19). For NF membranes evaluated in this study, permeate DOC concentrations generally increased with feedwater recovery (6, 17). Thus, increasing recovery reduced the extent of precursor limitation by making more NOM reactive sites available in permeate waters, ultimately resulting in higher bromide utilization (Figure 2b). This trend of increasing bromide utilization with increasing recovery was observed consistently with membranes II and III on all feedwaters employed in this study. However, because membrane I permeates were not all precursor limited (median DOC ) 0.13 mM), bromide utilization did not always increase with increasing recovery. 3990
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 35, NO. 19, 2001
THM Concentrations and Speciation. Br- and DOC concentrations in NF feedwaters from various participating utilities varied in the range 0.45-8.01 µM and 0.20-1.20 mM, respectively. These concentrations were further changed by nanofiltration. Because of the wide range of NOM concentrations and feedwater quality (e.g., raw-water ammonia levels), the chlorine demand was also different. All these factors can be expected to confound DBP formation and speciation. THM formation and speciation were evaluated (as shown in Figure 3) for permeate samples from varying locations and membrane types with 0.1 < Cl2/DOC < 0.4 mM/mM and 8 < pH < 9 using Br-/DOC as the independent variable. These samples span a wide range of Br-/DOC ratios, in part, because of two clusters of waters at either end. Source G (represented by 1 on the right side of Figure 3a corresponding to Br-/DOC > ∼80 µM/mM) had the highest bromide concentration in this study (5.2 µM on average), and all the permeates for this water had DOC concentrations e ∼0.08 mM. For source G, THM bromide utilization was low ( 0.9, 21% e CV e 44%). Lower correlation coefficients (R ∼ 0.5) and higher coefficient of variations (>75%) were observed for fits between chlorine consumption and DBP formation in permeate waters (Figure 2 (Supporting Information)). However, four of the source waters for our study were from the Biscayne Aquifer. This high-DOC groundwater is also high in inorganic chlorine demand caused by both NH3 and H2S. Thus, chlorine consumption may not be a good surrogate for DBP production in waters with low precursor concentrations and high inorganic chlorine demand. Alternatively, relationships between chlorine consumption and DBP production are depicted in Figure 3 (Supporting
FIGURE 7. Site specific correlations for NF permeates using feedwater from source B (24 h holding time, 24 °C temperature, 8.4 e pH e 9.2). Note the improved correlation coefficients and coefficients of variation for TTHM, di- and trihalogenated HAAs, halogen equivalent TTHM+HAA9, and TOX compared to Figure 6. VOL. 35, NO. 19, 2001 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
3997
FIGURE 8. Correlations between TTHM, HAA9, and halogenated TTHM+HAA9 with TOX concentrations for NF feed [(a), (c), and (e)] and permeate [(b), (d), and (f)] waters. Information) for permeate waters from one of the surface waters (source E) that contained no measurable NH3-N and was also low in bromide ion concentrations (0.45-0.63 µM). The site-specific correlation coefficients and coefficients of variation (for source E alone) for various DBPs with chlorine consumption as the independent variable improved substantially (0.84 e R e 0.89 and 15 e CV e 19%). Thus, chlorine consumption is a good predictor of DBP formation in nanofilter permeates in the absence of inorganic chlorine demand causing substances. Figure 8 shows excellent quantitative relationships between TOX and TTHM, HAA9, and halogen equivalent THM+HAA9 concentrations for both feed and permeate 3998
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 35, NO. 19, 2001
waters (R g 0.93) even though 16 e CV e 32%. It should be emphasized that correlations using TOX as the independent variable have been obtained using a wide range of SDS conditions (see Table 2). Thus, observed scatter can be attributed to both varying SDS conditions and water quality. Both correlation coefficients and coefficients of variations improved substantially (R g 0.99, 11 e CV e 14%) when only results from source water B are considered (see Figure 4 (Supporting Information)), underscoring the need for sitespecific correlations prior to making quantitative predictions. TOX can be interpreted as a cumulative amount of halogenated organics in drinking water. Hence, even though TOX is not currently regulated, good water treatment practices
suggest that its formation should be minimized. These findings suggest that reducing THM formation may also result in controlling HAA9 and the cumulative overall halogenated DBPs formed upon chlorination of nanofiltered waters. As observed in Figure 8e,f, the slope and 95% confidence intervals of the regression line for the sum of the halogen equivalent THMs and HAAs in NF feedwater was 0.43 ( 0.015 and for the permeate waters was 0.43 ( 0.010. Thus, only 43% of the TOX in both feed and permeate waters was quantified as THMs and HAAs in this study. These agree very well with values (35 and 38%) measured in North Carolina distribution systems following conventional treatment (11, 27). Further, because these two slopes do not differ at the 95% confidence level, NF does not appear to change the proportion of TOX comprised by THMs and HAAs upon chlorination. It should be emphasized that these good correlations between various DBPs and UV254, DOC, and TOX have been derived even though widely different source water characteristics, membranes characteristics, and NF pretreatment techniques were employed. Stronger correlations were obtained when a site-specific evaluation was conducted and when chlorination conditions such as temperature and reaction time were also maintained at constant values.
Acknowledgments This research was possible made possible by the partial support of the Texas Higher Education Coordinating Board (Project Number 003652-0412-1999). Gerry Filteau, Stuart McClellan, Terry Smith, Tom Stocker, and Mark Wilf provided membrane samples for testing. Jennifer Abrajano, Daniel Bush, Joseph Jacangelo, Eric Landsberg, Jason Radgowski, and David Wilkes assisted during some of the experiments. The constructive comments of Zaid Chowdhury and three anonymous reviewers on an earlier version of the manuscript are appreciated.
Supporting Information Available Various correlations observed in NF permeate waters as well as feedwaters: Figures 1-3, those with UV254 and chlorine consumed as independent variables and various DBPs as dependent variables, and Figure 4, site-specific correlations between TOX and THMs and HAAs. This material is available free of charge via the Internet at http://pubs.acs.org.
Literature Cited (1) USEPA. National Primary Drinking Water Regulations: Disinfectants and Disinfection By-Products, Final Rule; Washington, DC, 1998; pp 69390-69476. (2) USEPA. ICR Manual for Bench- and Pilot-Scale Treatment Studies; Office of Ground Water and Drinking Water: Cincinnati, OH, 1996; pp 1-1-3-108. (3) Pourmoghaddas, H.; et al. J.sAm. Water Works Assoc. 1993, 85(1), 82-87. (4) Cowman, G. A.; Singer, P. C. Environ. Sci. Technol. 1996, 30(1), 16-24.
(5) Wu, W. W.; Chadlik, P. A. J. Environ. Eng. 1998, 124(10), 932938. (6) Chellam, S. Environ. Sci. Technol. 2000, 34(9), 1813-1820. (7) Tan, L.; Amy, G. L. J.sAm. Water Works Assoc. 1991, 83(5), 7479. (8) Blau, T. J.; et al. J.sAm. Water Works Assoc. 1992, 84(12), 104116. (9) Allgeier, S. C.; Summers, R. S. J.sAm. Water Works Assoc. 1995, 87(3), 87-99. (10) Chellam, S.; et al. J.sAm. Water Works Assoc. 1997, 89(10), 7789. (11) Singer, P. C.; et al. J.sAm. Water Works Assoc. 1995, 87(10), 83-92. (12) Singer, P. C.; et al. J.sAm. Water Works Assoc. 1981, 72(8), 392401. (13) Singer, P. C.; Chang, S. D. J.sAm. Water Works Assoc. 1989, 81(8), 61-65. (14) Krasner, S. W.; et al. J.sAm. Water Works Assoc. 1989, 81(8), 41-53. (15) Edzwald, J. K.; et al. J.sAm. Water Works Assoc. 1985, 77(4), 122-132. (16) Najm, I. N.; et al. J.sAm. Water Works Assoc. 1994, 86(6), 98106. (17) Chellam, S.; Taylor, J. S. Water Res. 2001, 35(10), 2460-2474. (18) Edzwald, J. K.; Van Benschoten, J. E. Aluminum Coagulation of Natural Organic Matter. in Fourth International Gothenburg Symposium on Chemical Treatment; Madrid, Spain, 1990. (19) Symons, J. M.; et al. J.sAm. Water Works Assoc. 1993, 85(1), 51-62 (20) Minear, R. A.; Bird, J. C. Trihalomethanes: Impact of Bromide Ion Concentration on Yield, Species Distribution, Rate of Formation, and Influence of Other Variables, in Water Chlorination: Environmental Impact and Health Effects; Jolley, R. L., Brungs, W. A., Cumming, R. B., Eds.; Ann Arbor Science: Ann Arbor, MI, 1980; pp 151-160. (21) Krasner, S. W.; et al. J.sAm. Water Works Assoc. 1994, 86(6), 34-47. (22) Oliver, B. G. Effect of Temperature, pH, and Bromide Concentration on the Trihalomethane Reaction of Chlorine with Aquatic Humic Material, in Water Chlorination: Environmental Impact and Health Effects; Jolley, R. L., Brungs, W. A., Cumming, R. B., Eds.; Ann Arbor Science: Ann Arbor, MI, 1980; pp 141-149. (23) Laine, J.-M.; et al. J.sAm. Water Works Assoc. 1993, 85(6), 8799. (24) Hwang, C. J.; et al. DBP Formation and Reactivities of NOM Fractions of a Low-Humic Water, in Natural Organic Matter and Disinfection By-Products: Characterization and Control in Drinking Water; Barrett, S. E., Krasner, S. W., Amy, G. L., Eds.; American Chemical Society: Washington, DC., 2000; pp 173187. (25) Reckhow, D. A.; Singer, P. C. Mechanisms of Organic Halide Formation During Fulvic Acid Chlorination and Implications with Respect to Preozonation, in Water Chlorination: Chemistry, Environmental Impact and Health Effects; Jolley, R. L., Ed.; Lewis Publishers Inc.: Chelsea, MI, 1985. (26) Stevens, A. A.; et al. J.sAm. Water Works Assoc. 1989, 81(8), 54. (27) Reckhow, D. A.; Singer, P. C. J.sAm. Water Works Assoc. 1990, 82(4), 173-180.
Received for review March 23, 2001. Revised manuscript received July 6, 2001. Accepted July 16, 2001. ES010775N
VOL. 35, NO. 19, 2001 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
3999
