Development of Predictive Models for the Formation of ...

Water Qual. Res. J. Canada, 2004

•

Volume 39, No. 2, 149–159

Copyright © 2004, CAWQ

Development of Predictive Models for the Formation of Trihalomethanes and Haloacetic Acids during Chlorination of Bromide-Rich Water Themistokles D. Lekkas and Anastasia D. Nikolaou* Water and Air Quality Laboratory, Department of Environmental Studies, University of the Aegean, Karadoni 17, 81100 Mytilene, Lesvos, Greece

The formation of disinfection by-products (DBPs) during chlorination of drinking water is an issue which has drawn significant scientific attention over the last few years, due to the adverse effects of these compounds on human health. The complicated mechanisms of the formation of DBPs are still under investigation, and modelling attempts are being made. One of the major factors affecting the yields of DBPs is the concentration of bromide ion. In bromide-rich waters, the brominated species— bromoform, dibromoacetic acid—which are considered more toxic than their chlorinated analogues, may be the major species formed and therefore the consumption of these waters may be more harmful to human health than the consumption of water containing chlorinated species (chloroform, dichloroacetic acid). During this study, the effect of bromide ion on DBP formation was investigated by bench-scale chlorination experiments and statistically evaluated for individual species of DBPs. The importance of interactions between pairs of chlorination factors, pH, bromide concentration, chlorine dose and time for DBP formation was highlighted by statistical analysis results. Moreover, multiple regression models were developed for the concentrations of total trihalomethanes and total haloacetic acids, the two major DBP groups formed. The models developed provide satisfactory estimations of, but also indicate different formation mechanisms for, these two groups of DBPs. Key words: trihalomethanes, haloacetic acids, bromide, multiple regression models

Introduction

triles (HANs), chloropicrin (CP), as well as MX, iodinated THMs and iodinated HAAs have also been reported to exist in chlorinated drinking waters, but at significantly lower concentrations (Krasner et al. 1989; Oliver 1983; Kronberg et al. 1991; Nikolaou et al. 1999; Cancho et al. 2000; Weinberg et al. 2002). Scientific research is still being continued to determine the complicated mechanisms of the formation of DBPs, including reactions of disinfectants with natural organic matter (NOM) in water (Morris 1986; Reckhow and Singer 1986; Canale et al. 1997). The nature and concentration of NOM, the chlorine dose, the chlorine contact time, pH, temperature and the concentration of bromide have been reported to be the main parameters affecting DBP formation (Singer 1994). Based on these parameters, several predictive models have been developed to describe the formation of DBPs (Harrington et al. 1992; Greiner et al. 1992; Roberson et al. 1995; Rathbun 1996; Golfinopoulos et al. 1998; Clark 1998; Westerhoff et al. 2000; Sung et al. 2000; Lin and Hoang 2000; Elshorbagy et al. 2000). Most of these models predict the concentrations of total or individual species of THMs and some of them predict the concentrations of the major species of HAAs formed in chlorinated waters. To the authors’ knowledge, predictive models for the sum concentration of all nine HAAs in chlorinated waters have not yet been published.

Disinfection of drinking water with chlorine has been reported to result in the formation of a large number of organic by-products (disinfection by-products—DBPs) (Krasner et al. 1989; Oliver 1983; Nikolaou et al. 1999; Richardson et al. 1999; Richardson 2002; Roberts et al. 2002; Weinberg et al. 2002). Most of these compounds are harmful to human health, being possible or probable human carcinogens (Premazzi et al. 1997; U.S. EPA 1998), correlated to increased miscarriage risk during the first terms of pregnancy (U.S. EPA 1998; Swan et al. 1998; Waller et al. 1998, 2001), and having toxic effects on algae, crustaceans, bacteria and mice (Becher et al. 1992; Hashimoto et al. 1998; Glezer et al. 1999). Results from toxicological and epidemiological studies have resulted in regulations of their concentrations in drinking water by the U.S. EPA, the WHO and the EU (U.S. EPA 1998; WHO 1995; EEC 1998). DBPs most frequently detected in water include trihalomethanes (THMs) (Rook 1974; Alawi et al. 1994; Golfinopoulos and Nikolaou 2001) and haloacetic acids (HAAs) (Christman et al. 1983; Miller and Uden 1983; Reckhow and Singer 1990; Nikolaou and Lekkas 2001). Haloketones (HKs), chloral hydrate (CH), haloacetoni* Corresponding author; [email protected] 149

150

Lekkas and Nikolaou

The occurrence of brominated species of DBPs in drinking water is gaining increasing scientific concern recently, because of the higher toxicity of brominated compounds compared to their chlorinated counterparts (Buijs et al. 1984). Positive impact of bromide ions on genotoxicity of halogenated DBPs was observed (Nobukawa and Sanakida 2001). Therefore, emphasis should be given to the determination of the influence of bromide ion on the formation of DBPs. It has been documented by many investigators that in the presence of bromide ion (Br-), more brominated and mixed chlorobromo derivatives are formed (Krasner et al. 1989; Peters et al. 1991; Heller-Grossman et al. 1993; Pourmoghaddas and Stevens 1995). In bromide-rich waters, the brominated species—bromoform, dibromoacetic acid—may be the major species formed (Singer 1994; Cowman and Singer 1996), and therefore the consumption of these waters may be more harmful to human health than the consumption of water containing chlorinated species (chloroform, dichloroacetic acid). Natural water usually contains bromide ion originating from natural sources (geological sources, intrusion of seawater into groundwater) and from human activities (use of methyl bromide for agricultural applications, dibromoethylene as additive in petrol, use of salts to avoid the formation of ice in roads). Moreover, bromide ion occurs as an impurity in chlorine used for disinfection of water (Cooper et al. 1985; Siddiqui et al. 1995). Bromide ion in water is oxidized by chlorine to hypobromous acid or hypobromous ion, both of which react with natural organic matter (NOM) forming DBPs. The following reactions occur, resulting in the formation of species H2OBr+, BrCl, Cl2, Cl2O, Br2, HOBr, and HOCl (Haag and Hoigne 1983; Voudrias and Reinhard 1988): Kƒ HOBr + H+ + Cl- ←–––→ BrCl + H2O, Kf = 3.16 x 104 M-2, 20oC

(1)

Kƒ HOBr + H+ + Br- ←–––→ Br2 + H2O, Kf = 2.53 x 108 M-2, 20oC

(2)

Due to the high electrophilic affinity of BrCl and H2OBr+, the initial bromination rate is higher than the initial chlorination rate, even in relatively low bromide ion concentrations. Subsequently, the speciation of DBPs shifts to more brominated species (Cooper et al. 1985; Cowman and Singer 1996). The main objective of the present study was to determine and statistically evaluate the influence of bromide ion concentration on the formation of different species of DBPs during varying chlorination conditions, and furthermore to develop predictive models for the formation of these compounds during chlorination of bromide-rich river water. Most predictive models reported in the literature regard waters with low or not detectable bromide ion content. This is also the case for the multiple regres-

sion models proposed by Harrington et al. (1992), for which Greiner et al. (1992) reported that they were sensitive to bromide ion concentration changes. During the present bench-scale chlorination study, not only the influence of chlorine dose and contact time on DBP formation was investigated, but also particular emphasis was given to the parameters bromide ion concentration and pH, in order to obtain experimental data for a wider range of these parameters than that used in most previous model development studies.

Materials and Methods Glassware The glassware used during analysis was washed with detergent, rinsed with tap water, ultrapure water (Millipore: Milli-Ro 5 plus and Milli Q plus 185), acetone (Mallinckrodt Chemical Works, St. Louis, Mo.) and dried in an oven at 150°C for 2 hours.

Reagents-Standard Solutions Methanol (purge and trap grade) was purchased from Sigma-Aldrich, methyl-tert-butyl ether (MTBE) suprasolv grade, potassium dichromate, potassium iodide, potassium bromide, sodium sulfite, ammonium chloride, sodium sulfate, copper (II) sulfate pentahydrate and sulfuric acid concentrated ISO for analysis from Merck and boric acid (analytical grade) from Ferak. Ultrapure water was from a Milli-Q water purification system (Millipore: Milli-Ro 5 plus and Milli Q plus 185). Stock solutions were prepared in MTBE (Merck, for organic trace analysis) by addition of the appropriate amounts of monochloroacetonitrile (MCAN), dichloroacetonitrile (DCAN), bromochloroacetonitrile (BCAN), monobromoacetonitrile (MBAN), dibromoacetonitrile (DBAN), 1,1-dichloropropanone (1,1-DCP), 1,3-dichloropropanone (1,3-DCP), 1,1,1trichloropropanone (1,1,1-TCP), chloral hydrate (CH) and chloropicrin (CP) (Chemservice, purity >99%). For THMs, certified commercial mix solutions (Chemservice, purity >99%) containing known concentration of the four compounds chloroform (CHCl3), dichlorobromomethane (CHCl2Br), dibromochloromethane (CHClBr2) and bromoform (CHBr3) were used. Certified commercial stock solutions of HAAs and their methyl esters in MTBE were purchased from Supelco (purity >99%), containing the nine compounds monochloroacetic acid (MCA), monobromoacetic acid (MBA), dichloroacetic acid (DCA), bromochloroacetic acid (BCA), trichloroacetic acid (TCA), dibromoacetic acid (DBA), bromodichloroacetic acid (BDCA), dibromochloroacetic acid (DBCA) and tribromoacetic acid (TBA). From the stock solutions, standard solutions of DBPs 100 mg/L in MTBE were prepared, known volumes of which were injected into ultrapure water, giving standard solutions for system calibration.

THMs and HAAs in Bromide-Rich Water

Sample Preparation Water samples were collected in March 2000 from Tsiknias River in Mytilene Island, Greece. This river is the largest of the island, however flow is not constant throughout the year, and does not exist during summer, leading to increased organic matter content, and therefore enhanced potential for the formation of DBPs during chlorination. During this study, the organic matter content was measured as UV absorbance at 272 nm (UV-272). UV-272 has been reported to be an equivalent or better indicator than total organic carbon (TOC) of the organic matter content of the water, especially when the quenching agent sulfite exists in the water sample (Eaton 1995; Korshin et al. 1997). Samples were stored in 1-L amber glass bottles and, kept at 4oC, they were transported to the Water and Air Quality Laboratory of the University of Aegean. pH measurements were performed with a Crison MicropH2001 pH meter, and sample filtration with Whatman GF/A glass microfibre filters 4.7 cm. Samples were analyzed for chloride, bromide and nitrate ions by a modification of EPA Method 300.0 (O’Dell et al. 1984). UV absorbance measurements were performed at 272 nm by use of a Cary 1E UV-visible spectrophotometer. The samples were spiked with bromide ion (KBr) at concentration levels 1, 3, 15 and 30 mg/L, chlorinated according to the procedure described in Standard Methods for the Examination of Water and Wastewater (Iodometric Method I 4500B) (APHA 1992) with 3, 7.5, 15 and 30 mg/L Cl and divided into 40-mL amber glass bottles with polypropylene screw caps and TFE-faced septa (Pierce 13075). The reaction times were 2, 16, 24 and 48 h (room temperature, approximately 20oC). The same experiment was simultaneously conducted for three pH values: (a) not adjusted (pH 7.64), (b) pH 4 (adjusted with HCl solution) and (c) pH 10 (adjusted with NaOH solution). After the required reaction times had been completed, determination of residual chlorine was performed with the DPD colorimetric method (APHA 1992) and the quenching agent for depletion of residual chlorine was added. The quenching agent for the samples analyzed for THMs and other volatile DBPs was sodium sulfite, and for the samples analyzed for HAAs, ammonium chloride (100 mg/L of sample in both cases).

Sample Analysis For THMs and other volatile DBPs, a modification of EPA Method 551.1, which includes liquid-liquid extraction (LLE) with MTBE was performed (EPA 1998; Nikolaou et al. 2002a). For HAAs, acidic methanol esterification was used, as described elsewhere (Nikolaou et al. 2002b). Method recoveries and detection limits have been presented elsewhere (Nikolaou et al. 2002a,b). Recover-

151

ies of DBPs were satisfactory, generally ranging from 87.6 to 112.8% for THMs, from 60.4 to 124.3% for the other volatile DBPs and from 78.1 to 123.7% for HAAs. Exceptions are CH, CP and MCA, where low recoveries were observed (ranges 86.7–144.5%, 58.7–124.2% and 24.9–126.3%, respectively). Calibration curves of high linearity were obtained in all cases (R2 ≥ 0.99).

Statistical Analysis Multifactor analysis of variance (MANOVA) was applied for four factors: bromide ion concentration, chlorine dose, reaction time and pH, in order to determine the statistically significant parameters for the formation of each compound, and also the statistically significant combinations of parameters that could further influence this formation, and therefore should be considered during the development of predictive models. The software used for MANOVA application was Statgraphics 4.0. The development of predictive models for the formation of DBPs was performed by use of multiple regression (Draper and Smith 1981), with the Statistica 1999 program. Before the development of models, normal distribution of values of dependent variables was tested and log-transformation was applied wherever necessary (Golfinopoulos et al. 1998). For the elimination of statistically insignificant variables, the F-criterion was used, based on the ratio of the sum of squares due to regression (SSreg) to the sum of squares of residuals divided into the corresponding degrees of freedom (se2). For each regression coefficient, F = SSreg/se2. High F values denote that the particular coefficient is statistically significant for the model, thus the variable must be included in the model (Golfinopoulos et al. 1998). After the development of each model, the conditions that should be met by the residuals were tested: whether they follow the normal distribution and have zero mean value with constant dispersion (Draper and Smith 1981), in order to ensure the reliability of the model. These conditions are valid for both models developed during this study.

Results and Discussion DBPs Formation and Speciation The raw water characteristics of the samples are shown in Table 1. The organic matter content of the water was significant (UV-272 0.058 cm-1), denoting high potential of this water for DBP formation upon chlorination. This was confirmed by the rapidly decreasing values of residual chlorine in chlorinated samples, given in Table 2 for the bromide ion concentration of 1 mg/L. For the other concentration levels of bromide, residual chlorine concentrations were similar.

152

Lekkas and Nikolaou

TABLE 1. Raw water characteristics (pH, T, UV-272, Cl-, Br-, NO3-) of the Tsiknias River pH

T (°C)

UV-272 (cm-1)

Cl- (mg/L)

Br- (mg/L)

NO3- (mg/L)

7.64

9

0.058

64.5

NDa

5

a

ND; Not detectable concentration.

None of the compounds studied occurred in the raw water samples. However, in chlorinated samples many DBPs were detected: CHCl 3, CHCl 2Br, CHBr 3, CH, 1,1-DCP, 1,1,1-TCP, BCAN, MCAN, CP, MBAN, BCA, TCA, DBA, BDCA, DBCA and TBA. Brominated species predominated, especially at high bromide ion concentration, which is in accordance with results from several studies reported (Heller-Grossman et al. 1993; Cowman and Singer 1996; Kampioti and Stephanou 2002). The concentrations of the main species of DBPs detected in chlorinated water as a function of bromide ion concentration (chlorine dose 15 mg/L, reaction time 16 h, pH 7.64) are presented in Fig. 1, and the speciation of DBPs for bromide ion concentrations 1 and 15 mg/L (Cl dose 30 mg/L, reaction time 24 h) in Fig. 2. Similar changes in concentrations of DBPs as a function of bromide ion concentration were also observed at the two other pH values examined. Increase in bromide ion concentration resulted in decrease of concentrations of CHCl 3 , CHCl 2 Br and CHClBr 2 , but in increase of CHBr3 concentration. However, increase of bromide ion concentration above 15 mg/L did not cause further increase of CHBr3 concentration. The concentrations of the HAAs MCA, BCA, TCA and BDCA decreased with increasing bromide ion concentration, while the opposite is true for DBA and TBA. Bromide ion concentration 30 mg/L did not result in further increase of TBA concentration in relation to that at 15 mg/L.

tion, since at high bromide ion concentration, formation of brominated species is enhanced. CHCl3 concentration increased with increasing chlorine dose and was higher for higher pH (10). For CHCl2Br formation, all factors studied were statistically significant. Interactions existed between bromide-chlorine dose, bromide-reaction time, chlorine dose-reaction time and bromide-pH. CHCl2Br and CHClBr2 concentrations were significantly higher at low bromide concentration, high chlorine dose, high pH and long reaction time. Formation of CHBr3 was influenced by all studied parameters, with interactions between bromide-chlorine dose, bromide-contact time, bromide-pH, chlorine dosepH and contact time-pH. Higher CHBr3 concentrations were detected at high bromide ion concentration, high chlorine dose, high pH and long contact time. For CH formation, the statistically significant parameters were bromide and pH with interactions between them. Significantly higher concentration of CH was formed at lower bromide ion concentration. 1,1,1-TCP formation was affected by bromide ion concentration, chlorine dose and pH, with interactions between bromide-chlorine dose, bromide-pH, and chlorine dose-pH. The concentration of 1,1,1-TCP was significantly higher at low bromide ion concentration (non-

Statistical Analysis

TABLE 2. Residual chlorine concentrations (mg/L) in chlorinated samples from Tsiknias River (temperature 20°C, bromide concentration 1 mg/L)

The results of the MANOVA analysis for the four factors bromide ion concentration, chlorine dose, reaction time and pH are shown (only for the DBPs showing statistically significant differences) in Table 3 and the results of the interactions between the pairs of factors in Table 4. These interactions, which had not been statistically evaluated during previous studies, proved to be important for the formation of many DBPs, especially for pairs of factors including bromide ion concentration (bromide-chlorine dose, bromide-pH and bromide-time). THMs and other volatile DBPs. CHCl3 formation was significantly affected by bromide concentration, chlorine dose and pH, with interactions between bromide-chlorine dose, bromide-pH and chlorine dose-pH. CHCl3 concentration was significantly higher at low (1 and 3 mg/L) than at high (15 and 30 mg/L) bromide concentra-

Time (h) A. pH 4 Chlorine dose 3 mg/L Chlorine dose 7.5 mg/L Chlorine dose 15 mg/L Chlorine dose 30 mg/L B. pH 7.64 (not adjusted) Chlorine dose 3 mg/L Chlorine dose 7.5 mg/L Chlorine dose 15 mg/L Chlorine dose 30 mg/L Γ. pH 10 Chlorine dose 3 mg/L Chlorine dose 7.5 mg/L Chlorine dose 15 mg/L Chlorine dose 30 mg/L a

2

16

24

48

1.2 5 10 12.5

0.3 5 10 12.5

0.1 5 8 12.5

0.1 5 8 12.5

1.4 5 10 12.5

0.5 5 10 12.5

0.2 5 10 12.5

NDa 5 10 12.5

1.4 4 10 12.5

0.4 4 10 12.5

0.2 3.6 10 12.5

0.1 3.2 10 12.5

ND; Not detectable concentration.


153

Fig. 1. Concentrations of DBPs in chlorinated water from Tsiknias River as a function of bromide ion concentration (chlorine dose 15 mg/L, contact time 16 h, pH 7.64).

brominated species), higher chlorine dose and low pH. At high pH, 1,1,1-TCP is subject to hydrolysis, forming chloroform (Nikolaou et al. 2001). For MCAN formation, the significant parameters were bromide ion concentration, contact time and pH, with correlations between bromide-reaction time, bromide-pH and time-pH. The concentration of MCAN increased with increasing contact time and pH and was higher at low bromide concentration. CP formation was affected by bromide ion concentration, contact time and pH, with interactions between bromide-pH and time-pH. Higher concentrations of CP were formed at higher bromide concentrations, despite that CP is not a brominated species. This fact implies that bromide ion has a positive effect on CP formation via another mechanism. CP formation was significantly higher at longer contact time and low pH. MBAN formation is affected by bromide ion, chlorine dose and pH. Interactions exist between bromidechlorine dose, bromide-pH and chlorine dose-pH. Optimum value of bromide concentration for MBAN formation was 3 mg/L. Higher concentrations of MBAN

were observed for high chlorine dose (30 and 15 mg/L) and low pH (4). Haloacetic acids. For MCA, all factors studied were statistically significant with interactions between bromidechlorine dose, bromide-contact time, chlorine dosecontact time and bromide-pH. MCA concentration decreased with increasing bromide ion concentration and increased with increasing chlorine dose. It also increased over time and reached a maximum at pH 7.64. MBA formation was affected by bromide ion, reaction time and pH, with interactions between bromidepH. Higher concentrations of MBA were observed at higher bromide concentration and for longer contact time, while for pH 10, the MBA concentration decreased, probably due to decomposition. DCA formation was significantly affected by bromide ion, chlorine dose and pH with interactions between bromide-chlorine dose, bromide-pH and chlorine dosepH. Significantly higher concentration was observed at the lowest bromide ion concentration, the highest chlorine dose and the lowest pH value (4).

b

a

Fig. 2. Speciation of DBPs for bromide ion concentrations (a) 1, and (b) 15 mg/L (Cl dose 30 mg/L, reaction time 24 h).

154

Lekkas and Nikolaou

TABLE 3. MANOVA results for DBPs concentrations as function of bromide ion, chlorine dose, contact time and pH Factors DBPs CHCl3 CHCl2Br CHClBr2 CHBr3 CH 1,1-DCP 1,1,1-TCP MCAN CP MBAN MCA MBA DCA BCA TCA DBA BDCA DBCA TBA

Bromide ion Chlorine dose *a * * * *

* * * *

* * * * * * * * * * * * *

*

Time * * *

* * * *

* *

* * * * *

* * * *

pH

Degrees of freedom

* * * * * * * * * * * * * * * * * * *

135 135 135 135 135 135 135 135 135 135 135 135 135 135 135 135 135 116 135

a

*; Statistically significant difference (significance level F0.05 = 2.10, F0.01 = 2.80 for the HAAs model. The R2 values were 0.87 and 0.51, the standard errors of estimate 0.098 and 0.222 and the Durbin-Watson estimate 1.74 and 1.83 for the THMs and for the HAAs model, respectively. The predicted versus observed values for the two models, as derived from the dataset of measurements used for the development of the model, are presented in Fig. 3. It can be observed that both models provide good fit. According to the statistical analysis, the residuals (errors) were shown to have zero mean value and constant dispersion. Moreover, the Durbin-Watson values close to 2, further confirm the reliability of the models. The THM model seems to be very satisfactory, in agreement with the results of the previous study (Golfinopoulos et al. 1998), confirming that the multiple regression technique is useful for modelling THM concentrations in chlorinated waters. The HAA model, giving relatively good results, has a moderate R2 value, implying that another modelling approach may be needed to more accurately predict HAA concentrations. Different

mechanisms of formation of HAAs than THMs could be the reason for this result. Another possible explanation is the use of the sum of all nine species of HAAs, including BDCA, DBCA which have been reported to be unstable, decomposing to the corresponding THMs at ambient temperature (Heller-Grossman et al. 1993; Zhang and Minear 2002). It is interesting to note that bromide ion concentration was not a statistically significant factor for inclusion in the multiple regression model predicting total THM concentration. Although bromide ion concentration significantly affected the speciation of THMs, it did not influence their total concentration. On the contrary, for the total HAA model, bromide concentration was a statistically significant variable and was included in the model. The influence of the pairs of factors was tested during model development, since it was shown to have a statistically significant effect on DBPs formation. For the total THM model, none of the products of pairs of factors was statistically significant, although shown by MANOVA analysis to have statistically significant influence on the concentration of each individual species (Table 4). However, for the total HAAs model, the products pH x Br and pH x time were statistically significant variables, which improved its performance. This fact indicates the different mechanism for the formation of HAAs than that for the THMs, which needs to be taken into account during optimization of drinking water chlorination conditions to min-

156

Lekkas and Nikolaou

a

b

Fig. 3. Predicted versus observed values (a) for the THMs model, (b) for the HAAs model.

imize the DBPs formed. Moreover, it appeared that the inclusion of products of factors into the multiple regression models can, in some cases, support the accuracy of model estimations. Similar observations have been reported in two other modelling studies, one for the product (temperature) x (season) (Golfinopoulos et al. 1998), and another for the product UV x TOC (Lin and Hoang 2000).

Validation of the models

and 8, the chlorine dosages 2 and 4 mg/L and the reaction times ranged from 0 to 24 h. The validation results are presented in Fig. 4a and 4b for the THM and the HAA model, respectively. The comparison between predicted and observed values showed very satisfactory results for the THM model and relatively good results for the HAA model. Specifically, for the THM model, 24 of the 24 observed values lay within ±20% of the predicted values, with 17 of the 24 observed values lying between ±15% of the predicted values. For the HAA model, 13 of the 24 observed values lay within ±20% of the predicted values, with 10 of the 24 observed values lying between ±15% of the predicted values.

A. Validation against another dataset from chlorinated river water. The predictive models developed for THMs and HAAs were validated against an individual dataset of DBP measurements in water samples taken during another period (October 2000) from the same river and chlorinated under conditions different than those used in the present study. The organic matter content of the river (0.048 cm-1) was slightly lower than in March (0.058 cm-1) and bromide ion was not detectable. The pH values tested were 7

B. Validation against a dataset from a water treatment plant. Validation of the developed models was also performed against a dataset from an actual water treatment plant (in Athens, Greece). The characteristics of the raw water entering the plant are presented in Table 5. The

a

b

Fig. 4. Model validation against different dataset of (a) THMs, and (b) HAAs measurements in chlorinated water from the Tsiknias River.


157

TABLE 5. Quality characteristics of the water from the water treatment plant used for model validation Parameter Color (Pt-Co) Turbidity (NTU) Temperature (oC) pH

Range

Parameter

Range

3–4 8.95–48.3 13–18 7.94–8.31

Total organic carbon (mg/L) UV-272 absorbance (cm-1) Bromide ion (mg/L)

0.91–1.39 0.018–0.046 0.017–0.032

level of organic matter content of the water, expressed as UV-272, was similar to that of Tsiknias River. Table 6 presents the validation of the models against a 12-month dataset of DBP measurements in the effluent of the water treatment plant, for the year 2002. The validation results were satisfactory for both models. For the THM model, 11 of the 12 observed values lay within ±20% of the predicted values, with 10 of the 12 observed values lying between ±15% of the predicted values. For the HAA model, 10 of the 12 observed values lay within ±20% of the predicted values, with 9 of the 12 observed values lying between ±15% of the predicted values. The percent deviations between predicted and observed values were calculated by the formula 100x[(observed value)-(predicted value)]/ (predicted value), and ranged from 0.1 to 26.7% for the THM model and from 0.2 to 60.6% for the HAA model. Possible explanation for the high deviations observed in a limited number of cases is the influence of season and temperature, which was not investigated during the present study. Future experimental work taking into account the temperature and season could possibly improve the performance of the proposed models for the prediction of DBP formation in chlorinated waters.

Conclusions The influence of bromide ion concentration on the formation of DBPs was investigated in chlorinated river water with significant organic matter content. The results showed that increase in bromide ion concentration resulted in decrease of concentrations of CHCl3, CHCl2Br, CHClBr2, MCA, BCA, TCA and BDCA, and in increase of CHBr3, DBA and TBA concentrations. MANOVA was applied for four factors: bromide ion concentration, chlorine dose, reaction time and pH, in order to determine the statistically significant parameters and combinations of parameters for the formation of individual species of DBPs. Statistically significant influence was observed in the majority of cases for all factors as well as for many combinations of factors, especially bromide ion-pH, chlorine dose-pH and bromide ion-chlorine dose. Predictive models were developed for the concentrations of total THMs and total HAAs, which were the

major groups of compounds formed. Statistical evaluation of the models showed they provide satisfactory estimations. The lower R2 value of the HAA model compared to that of the THM model indicates that a modelling approach other than multiple regression might be needed to describe the formation of HAAs. Different formation mechanisms of HAAs and THMs are implied also from the statistically significant parameters included in each model, especially bromide ion, which affected total HAAs and individual THM formation, but not total THM concentration. The influence of combinations of factors, expressed as products of pairs of factors, was taken into account during model development. Pairs of factors were not statistically significant for total THM formation, but only for the formation of individual species of THMs. On the contrary, the pairs of factors pH x Br and pH x time were included in the predictive model for total HAA concentration and significantly improved its performance. Validation of the proposed models against DBP measurements both in chlorinated river water and drinking water treatment plant effluent gave satisfactory results for the prediction of DBPs in chlorinated waters.

Acknowledgements We would like to thank Prof. Maria Kostopoulou and Prof. Spyros Golfinopoulos for their scientific and technical assistance.

TABLE 6. Percent deviations between predicted and observed values of THMs and HAAs in the samples from the water treatment plant effluent on a monthly basis for the year 2002 Month January February March April May June July August September October November December

THMs (% deviation)

HAAs (% deviation)

2,9 -11,6 16,1 2,8 0,1 -10,2 -6,2 -4,8 1,1 0,1 -10,9 26,7

-0,2 60,6 31,7 2,1 6,7 -2,9 -0,7 5,3 7,5 1,7 -5,0 17,5

158

Lekkas and Nikolaou

References Alawi MA, Khalill F, Sahili I. 1994. Determination of trihalomethanes produced through the chlorination of water as a function of its humic acid content. Arch. Environ. Contam. Toxicol. 26:381–386. APHA. 1992. Standard methods for the examination of water and wastewater. 18th edition. Becher G, Ovrum NM, Christman RF. 1992. Novel chlorination by-products of aquatic humic substances. Sci. Tot. Environ. 117/118:509–520. Buijs W, Van der Gen S, Mohn GR, Breimer DD. 1984. The direct mutagenic activity of α-ω-dihalogenoalkanes in Salmonella typhimurium: strong correlation between chemical properties and mutagenic activity. Mutat. Res. 141:11–14. Canale RP, Chapra SC, Amy GL, Edwards MA. 1997. Trihalomethane precursor model for Lake Youngs, Washington. J. Water Resour. Plan. Manage. 123:259–265. Cancho B, Ventura F, Galceran M, Diaz A, Ricart S. 2000. Determination, synthesis and survey of iodinated trihalomethanes in water treatment processes. Water Res. 34:3380–3390. Christman RF, Norwood DL, Millington DS, Johnson JD, Stevens AA. 1983. Identity and yields of major halogenated products of aquatic fulvic acid chlorination. Environ. Sci. Tech. 17:625–628. Clark RM. 1998. Chlorine demand and TTHM formation kinetics: a second-order model. J. Environ. Eng. Div. 124:16–24. Cooper WJ, Zika RG, Steinhauer MS. 1985. Bromide oxidant interactions and THM formation: a literature review. J. Am. Water Works Assoc. 77:116. Cowman GA, Singer PC. 1996. Effect of bromide ion on haloacetic acid speciation resulting from chlorination and chloramination of humic substances. Environ. Sci. Tech. 30:16–24. Draper NR, Smith H. 1981. Applied regression analysis. John Wiley and Sons, New York. p. 169–171. Eaton A. 1995. Measuring UV-absorbing organics: a standard method. J. Am. Water Works Assoc. 87:86–90. EEC. 1998. EEC Council Directive 98/83/EC of 3 November 1998 on the quality of water intended for human consumption. Official Journal of the European Communities, L 330/32, 5.12.98. Elshorbagy WE, Abu-Qdais H, Elsheamy MK. 2000. Simulation of THM species in water distribution systems. Water Res. 34:3431–3439. EPA. 1998. EPA Method 551.1. Determination of chlorinated disinfection by-products, chlorinated solvents, and halogenated pesticides/herbicides in drinking water by liquid-liquid extraction and gas chromatography with electron capture detection. U.S. EPA, Office of Water, Technical Support Center. Glezer V, Harris B, Tal N, Iosefzon B, Lev O. 1999. Hydrolysis of haloacetonitriles: linear free energy relationship, kinetics and products. Water Res. 33:1938–1948.

Golfinopoulos SK, Nikolaou AD. 2001. Disinfection byproducts and volatile organic compounds in the water supply system of Athens, Greece. Env. Sci. Health A36:483–499. Golfinopoulos SK, Xylourgidis NK, Kostopoulou MN, Lekkas TD. 1998. Use of a multiple regression model for predicting trihalomethane formation. Water Res. 32:2821–2829. Greiner AD, Obolensky A, Singer PC. 1992. Technical note: comparing predicted and observed concentrations of DBPs. J. Am. Water Works Assoc. 84:99–102. Haag WR, Hoigne J. 1983. Ozonation of bromide-containing waters: kinetics of formation of hypobromous acid and bromate. Environ. Sci. Tech. 17:261. Harrington GW, Chowdhurry ZK, Owen DM. 1992. Developing a computer model to simulate DBP formation during water treatment. J. Am. Water Works Assoc. 84:78. Hashimoto S, Azuma T, Otsuki A. 1998. Distribution, sources and stability of haloacetic acids in Tokyo Bay Japan. Environ. Toxicol. Chem. 17:798–805. Heller-Grossman L, Manka J, Limoni-Relis B, Rebhun M. 1993. Formation and distribution of haloacetic acids, THM and TOX in chlorination of bromide-rich lake water. Water Res. 27:1323–1331. Kampioti A, Stephanou E. 2002. The impact of bromide on the formation of neutral and acidic disinfection byproducts (DBPs) in Mediterranean chlorinated drinking water. Water Res. 36:2596–2606. Korshin G, Li C, Benjamin M. 1997. The decrease of UV absorbance as an indicator of TOX formation. Water Res. 31:946–949. Krasner S, McGuire M, Jacangelo J, Patania N, Reagan K, Aieta E. 1989. The occurrence of disinfection by-products in U.S. drinking water. J. Am. Water Works Assoc. 81:41–53. Kronberg L, Christman RF, Singh R, Ball LM. 1991. Identification of oxidized and reduced forms of the strong bacterial mutagen (Z)-2-Chloro-3-dichloromethyl-4oxobutenoic acid (MX) in extracts of chlorine-treated water. Environ. Sci. Tech. 25:99–104. Lin T-F, Hoang S-W. 2000. Inhalation exposure to THMs from drinking water in south Taiwan. Sci. Tot. Environ. 246:41–49. Miller JW, Uden PC. 1983. Characterization of non-volatile aqueous chlorination products of humic substances. Environ. Sci. Tech. 17:150–157. Morris JC. 1986. Reaction dynamics in water chlorination, p. 701–711. In Water chlorination: chemistry, environmental impact and health effects, Vol. 5. Lewis Publishers, U.S.A. Nikolaou AD, Golfinopoulos SK, Kostopoulou MN, Lekkas TD. 2002b. Determination of haloacetic acids in water by acidic methanol esterification-GC-ECD method. Water Res. 36:1089–1094. Nikolaou A, Kostopoulou M, Lekkas T. 1999. Organic byproducts of drinking water chlorination: a review. Global NEST The Int. J. 1:143–156.


Nikolaou AD, Lekkas TD. 2001. The role of natural organic materials during formation of chlorination byproducts: a review. Acta Hydrochimica et Hydrobiologica 29:63–77. Nikolaou AD, Lekkas TD, Golfinopoulos SK, Kostopoulou MN. 2002a. Application of different analytical methods for determination of volatile chlorination by-products in drinking water. Talanta 56:717–726. Nikolaou AD, Lekkas TD, Kostopoulou MN, Golfinopoulos SK. 2001. Investigation of the behavior of haloketones in water samples. Chemosphere 44:907–912. Nobukawa T, Sanakida S. 2001. Effect of bromide ions on genotoxicity of halogenated by-products from chlorination of humic acid in water. Water Res. 35:4293–4298. O’Dell J, Pfaff J, Gales M, McKee G. 1984. Test method: the determination of inorganic anions in water by ion chromatography, Method 300.0. United States Environmental Protection Agency, Environmental Monitoring and Support Laboratory, Cincinnati OH 45268, EPA-600/4-84-017, March 1984. Oliver B. 1983. Dihaloacetonitriles in drinking water: algae and fulvic acid as precursors. Environ. Sci. Tech. 17: 80–83. Peters RJB, Erkelens C, De Leer EWB, DeGalan L. 1991. The analysis of halogenated acids in Dutch drinking water. Water Res. 25:473–477. Pourmoghaddas H, Stevens AA. 1995. Relationship between trihalomethanes and haloacetic acids with total organic halogen during chlorination. Water Res. 29:2059–2062. Premazzi G, Cardoso C, Conio O, Palumbo F, Ziglio G, Meucci L, Borgioli A. 1997. Standards and strategies in the European Union to control trihalomethanes in drinking water. Environment Institute, European Commission Joint Research Centre and Techware, Italy. Rathbun RE. 1996. Regression equations for disinfection by-products for the Mississippi, Ohio and Missouri rivers. Sci. Tot. Environ. 191:235–244. Reckhow DA, Singer PC. 1986. Mechanisms of organic halide formation during fulvic acid chlorination and implications with respect to preozonation, p. 1229–1257. In Water chlorination: chemistry, environmental impact and health effects, Vol. 5. Lewis Publishers, U.S.A. Reckhow DA, Singer PC. 1990. Chlorination by-products in drinking waters: from formation potentials to finished water concentrations. J. Am. Water Works Assoc. 82:173–180. Richardson SD. 2002. The role of GC-MS and LC-MS in the discovery of drinking water disinfection by-products. J. Environ. Monit. 4:1–9. Richardson SD, Thruston Jr AD, Caughran TV, Chen PH, Collette TW, Floyd TL, Schenck KM, Lykins Jr BW. 1999. Identification of new ozone disinfection byproducts in drinking water. Environ. Sci. Tech. 33: 3368–3377.

159

Roberson JA, Cromwell JE, Krasner SW, McGuire MJ, Owen DM, Regli S, Summers RS. 1995. The D-DBP rule: where did the numbers come from? J. Am. Water Works Assoc. 87:46–57. Roberts MG, Singer PC, Obolensky A. 2002. Comparing total HAA and total THM concentrations using ICR data. J. Am. Water Works Assoc. 94:103–114. Rook JJ. 1974. Formation of haloforms during chlorination of natural waters. Water Treatment Examination 23: 234–242. Siddiqui MS, Amy GL, Rice RC. 1995. Bromate ion formation: a critical review. J. Am. Water Works Assoc. 87:58. Singer P. 1994. Control of disinfection by-products in drinking water. J. Environ. Eng. Div. 120:727–744. Sung W, Reilley-Matthews B, O’Day D, Horrigan K. 2000. Modeling DBP formation. J. Am. Water Works Assoc. 92:53–63. Swan SH, Waller K, Hopkins B, Windham G, Fenster L, Schaefer C, Neutra RR. 1998. A prospective study of spontaneous abortion: relation to amount and source of drinking water consumed in early pregnancy. Epidemiology 9:126–133. U.S. EPA. 1998. National primary drinking water regulations: disinfectants and disinfection by-products notice of data availability. Office of Ground Water and Drinking Water, http://www.epa.gov/OGWDW/mdbp/dis.html. Voudrias EA, Reinhard M. 1988. A kinetic model for the halogenation of p-xylene in aqueous HOCl solutions containing Cl- and Br-. Environ Sci. Tech. 22:1056–1062. Waller K, Swan SH, DeLorenze G, Hopkins B. 1998. Trihalomethanes in drinking water and spontaneous abortion. Epidemiology 9:134–140. Waller K, Swan SH, Windham GC, Fenster L. 2001. Influence of exposure assessment methods on risk estimates in an epidemiologic study of total trihalomethane exposure and spontaneous abortion. J. Expo. Anal. Environ. Epidemiol. 11:522–31. Weinberg HS, Krasner SW, Richardson SD, Thruston Jr AD. 2002. The occurrence of disinfection-by-products (DBPs) of health concern in drinking water: results of a nationwide DBP occurrence study, EPA/600/R02/068, U.S. Environmental Protection Agency, National Exposure Research Laboratory, Athens, Ga. Westerhoff P, Debroux J, Amy G, Gatel D, Mary V, Cavard J. 2000. Applying DBP models to full-scale plants. J. Am. Water Works Assoc. 92:89. WHO. 1995. Desinfection de l’eau. Local authorities, health and environment briefing pamphlet series, No. 3. Zhang X, Minear RA. 2002. Decomposition of trihaloacetic acids and formation of the corresponding trihalomethanes in drinking water. Water Res. 36:3665–3673.

Received: May 15, 2003; accepted: February 16, 2004.