Page 1
Accepted Manuscript
Title: Investigating effects of bromide ions on trihalomethanes and developing modelfor predicting bromodichloromethane in drinking water
Authors: Shakhawat Chowdhury, Pascale Champagne, P. James McLellan
PII: S0043-1354(09)00854-9
DOI: 10.1016/j.watres.2009.12.042
Reference: WR 7824
To appear in: Water Research
Received Date: 14 September 2009
Revised Date: 12 November 2009
Accepted Date: 23 December 2009
Please cite this article as: Chowdhury, S., Champagne, P., McLellan, P.J. Investigating effects ofbromide ions on trihalomethanes and developing model for predicting bromodichloromethane in drinkingwater, Water Research (2010), doi: 10.1016/j.watres.2009.12.042
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service toour customers we are providing this early version of the manuscript. The manuscript will undergocopyediting, typesetting, and review of the resulting proof before it is published in its final form. Pleasenote that during the production process errors may be discovered which could affect the content, and alllegal disclaimers that apply to the journal pertain.
Page 2
MANUSCRIP
T
ACCEPTED
ACCEPTED MANUSCRIPT
1
Investigating effects of bromide ions on trihalomethanes and developing 1
model for predicting bromodichloromethane in drinking water 2
Shakhawat Chowdhury1*, Pascale Champagne1 and P. James McLellan2 3
4
* Corresponding Author 5
1*Shakhawat Chowdhury 6
Department of Civil Engineering 7
Queen’s University, Kingston, ON, K7L 3N6 8
Email: [email protected] 9
Phone: 1-613-533-2144; Fax: 1-613-533-2128 10
11
1Pascale Champagne 12
Associate Professor, Department of Civil Engineering 13
Queen’s University, Kingston, ON, K7L 3N6 14
Email: [email protected] 15
16
2 P. James McLellan 17
Professor and Head, Department of Chemical Engineering 18
Queen’s University, Kingston, ON, K7L 3N6 19
Email: [email protected] 20
21
22
23
Page 3
MANUSCRIP
T
ACCEPTED
ACCEPTED MANUSCRIPT
2
ABSTRACT 24
Chlorination for drinking water can form brominated trihalomethanes (THMs) in the presence of 25
bromide ions. Recent studies have reported that bromodichloromethane (BDCM) has a stronger 26
association with stillbirths and neural tube defects than other THMs species. In this paper, the 27
results of an experimental investigation into the factors forming THMs in the presence of 28
bromide ions are presented. The experiments were conducted using synthetic water samples with 29
different characteristics (e.g., pH, temperature, dissolve organic content). Different combinations 30
of these characteristics were considered in the experimental program. The results showed that 31
increased bromide ion concentrations led to increases in the formation of total THMs, with 32
higher BDCM and dibromochloromethane (DBCM), and lower chloroform formation. By 33
increasing the pH from 6 to 8.5, increased chloroform and decreased BDCM and DBCM 34
formation were observed. Higher bromide ions to chlorine ratios increased BDCM and DBCM 35
and decreased chloroform formation, while higher temperatures increased BDCM, DBCM and 36
chloroform formation. In most cases, bromoform (CHBr3) concentrations were found to be 37
below the detection limit. Significant factors influencing BDCM formation were identified using 38
a statistical analysis. A model for BDCM formation was estimated from 44 experiments and 39
statistical adequacy was assessed using appropriate diagnostics, including residual plots and an 40
R2 of 0.97. The model was validated using external data from 17 water supply systems in 41
Newfoundland, Canada. The predictive performance of the model was found to be excellent, and 42
the resulting model could be used to predict BDCM formation in drinking water and to perform 43
risk-cost balance analyses for best management practices. 44
45
Page 4
MANUSCRIP
T
ACCEPTED
ACCEPTED MANUSCRIPT
3
KEYWORDS: Brominated trihalomethanes, bromide ions effects, bromodichloromethane 46
formation model, design of experiments (DOE), statistical model 47
48
INTRODUCTION 49
Chlorine is widely used in the municipal water supply systems in Canada and the USA due to its 50
excellent disinfection performance and low cost (Clark et al., 1994, 1998; Reiff, 1995; USEPA, 51
2006; Chowdhury and Husain, 2006; Chowdhury et al., 2007; Health Canada, 2008). However, 52
natural organic matter (NOM) in the water can react with chlorine during disinfection, forming 53
disinfection by-products (DBPs), such as trihalomethanes (THMs), haloacetic acids (HAAs), 54
haloacetonitriles (HANs), haloketones (HKs), as well as other known and unknown compounds 55
(Richardson, 2005). The possible effects of THMs on animal and human health have been 56
extensively studied from toxicological and epidemiological perspectives since their discovery in 57
1974 (King and Marrett, 1996; Wigle, 1998; Mills et al., 1998; Waller et al., 1998; King et al., 58
2000; Richardson et al., 2002; Dodds et al., 2004; Villanueva et al., 2004). More than 90% (87-59
98%) of the THMs in drinking water supplies across the Canadian provinces typically consist of 60
CHCl3 and BDCM, while BDCM alone contributes 2.1 to 14% of the THMs. The occurrences of 61
BDCM often exceed the Canadian regulatory limit of 16 ppb (Health Canada, 2008). It has been 62
reported that BDCM in drinking water has a much stronger association with stillbirths and low 63
birth weights than the other THMs species (King et al., 2000). The BDCM targets human 64
placental trophoblasts that produce a hormone which is required during pregnancy. A decrease in 65
bioactive levels of this hormone can lead to adverse effects during pregnancy (Health Canada, 66
2007). Dodds and King (2001) reported an increased risk of neural tube defects from BDCM at 67
exposure concentrations of 20 ppb or higher. Toxicological studies have characterized BDCM as 68
Page 5
MANUSCRIP
T
ACCEPTED
ACCEPTED MANUSCRIPT
4
a probable human carcinogen with a slope factor (upper bound lifetime probability of an 69
individual developing cancer) of 0.062 (mg/kg/day)-1 and a reference dose (maximum safe dose 70
to human) of 0.02 mg/kg/day (IRIS, 2008). 71
72
The reaction pathways and factors influencing THMs formation are well established in the 73
existing literature (Stevens et al., 1976; Minear and Morrow, 1983; Clark et al., 2001; Rodrigues 74
et al., 2007; Chowdhury et al, 2008, Chowdhury et al, 2009). To address the effects of bromide 75
ion (Br-) concentrations in water, a number of previous studies incorporated bromide ion in 76
THMs formation model development (Minear and Morrow, 1983; Amy et al., 1987; 77
Golfinopoulos et al., 1998; Westerhoff et al., 2000; Elshorbagy et al., 2000; Rodriguez et al., 78
2003; Lekkas and Nikolaou, 2004). Bromide ions form brominated THMs following complex 79
reaction pathways during the chlorination of drinking water (Liang and Singer, 2003). Uyak and 80
Toroz (2007) reported that Br- produces hypobromous acid (HOBr) in chlorinated water, which 81
is approximately 20 times more reactive with NOM than hypochlorous acid (HOCl). Increases in 82
bromide ion concentrations gradually shift chlorinated THMs to mixed bromochloro THMs. As 83
such, the bromide ions to chlorine ratio in water may be an important factor, which may describe 84
the relative distributions of brominated and chlorinated THMs (Uyak and Toroz, 2007; Hellur-85
Grossman et al., 2001; Nokes et al., 1999). Further complexity arises due to the partial 86
conversion of Br- into brominated THMs (18 to 28%), which also depends on water pH, 87
temperature and relative distributions of hydrophobic and hydrophilic fractions of NOM in water 88
(Sohn et al., 2006; Liang and Singer, 2003). As such, assumptions of complete conversion of Br- 89
into brominated THMs may not reflect the actual observed conversion products. 90
91
Page 6
MANUSCRIP
T
ACCEPTED
ACCEPTED MANUSCRIPT
5
The aquatic NOM, which is the main precursor of THMs formation, contains both hydrophobic 92
and hydrophilic fractions. The hydrophobic fractions generally represent the higher molecular 93
weight NOM with activated aromatic rings, phenolic hydroxyl groups and conjugated double 94
bonds, while the hydrophilic fractions are typically lower molecular weight NOM with aliphatic 95
ketones and alcohols (Liang and Singer, 2003). The hydrophobic fractions of NOM exhibit 96
higher UV254 and SUVA (specific ultraviolet absorbance, which is defined 97
as 100/DOCUV254× ), values and may be more reactive with HOCl than HOBr, while the 98
hydrophilic fractions have lower UV254 and SUVA values and may be more reactive with HOBr 99
than HOCl (Liang and Singer, 2003). The formation pathways for the brominated THMs can be 100
hypothesized following Figure 1. Either of the reaction pathways can be significant for the 101
brominated THMs formation depending on the types of NOM, concentrations of bromide ions 102
and chlorine doses. Source water without bromide ions form chlorinated THMs (CHCl3), 103
possibly due to reactions between HOCl and the hydrophobic fractions of NOM; thus, a 104
significant portion of the hydrophilic fractions of NOM may be left unreacted in this water. 105
Conversely, water with bromide ions may form brominated THMs through reactions between the 106
hydrophilic fractions of NOM and HOBr, as well as a result of a shift from chlorinated THMs to 107
brominated THMs. However, these brominated THMs may not be adequately characterized by 108
the corresponding low SUVA or UV254 measurements. For example, Hellur-Grossman et al. 109
(2001) noted that the Sea of Galilee (Lake Kinneret) in Israel has TOC concentrations ranging 110
between 4 to 6 mg/L and uniquely high bromide ion concentrations (up to 1.9 mg/L). This water 111
is composed of mostly hydrophilic NOM. The hydrophilic NOM favored brominated THMs 112
formation resulting in more than 96% of total THMs in Lake Kinneret water (summer: 113
brominated THMs = 595.5 ppb and total THMs = 606 ppb; winter: brominated THMs = 487 ppb 114
Page 7
MANUSCRIP
T
ACCEPTED
ACCEPTED MANUSCRIPT
6
and total THMs = 507 ppb). They reported that the measured UV254 and SUVA values could not 115
effectively be employed to characterize THMs formation for this source water, possibly due to 116
the fact that the hydrophilic fractions of NOM could not be adequately represented by the 117
corresponding low SUVA or UV254 values (Hellur-Grossman et al., 2001). 118
119
To date, investigations into the effects of bromide ions on specific THMs are limited in the 120
existing literature. Previous studies mostly focused on the effects of bromide ion concentrations 121
on total THMs formation (Minear and Morrow, 1983; Amy et al., 1987; Clark et al., 2001; Gang 122
et al., 2002; Serodes et al., 2003; MWH, 2005; Navalon et al., 2008). In the present paper, effects 123
of bromide ion and other factors (DOC, chlorine dose, pH, temperature, reaction time and ratio 124
of bromide ion to chlorine) on THMs species were investigated under controlled laboratory 125
conditions using synthetic water. The synthetic water samples with different characteristics were 126
prepared using a factorial design approach. The formation of THMs with and without bromide 127
ion was compared statistically and significant factors for BDCM formation were identified. 128
Using the significant factors, a model for predicting BDCM formation was developed. Model 129
adequacy was tested statistically using graphical and numerical diagnostics and the prediction 130
performance of the model was tested systematically using the full ranges of the experimental 131
data and by applying data from 17 water distribution systems from Newfoundland and Labrador. 132
133
EXPERIMENTAL METHODS 134
The experiments conducted consisted of synthetic water samples of differing characteristics. 135
Commercial humic acid (Suwannee River Humic Acid) was used to generate water samples with 136
different DOC concentrations, using a concentrated stock solution prepared by dissolving humic 137
Page 8
MANUSCRIP
T
ACCEPTED
ACCEPTED MANUSCRIPT
7
acid (HA) into 0.1 N NaOH solutions. The stock solution was used at varying concentrations in 138
deionized water and the corresponding DOC concentrations were determined using Standard 139
Method 5310B (APHA, 1995) in the Analytical Services Unit (ASU) at Queen’s University, 140
Kingston, Ontario. Different levels of DOC were adjusted using HA and confirmed by laboratory 141
analyses. The UV254 (cm-1) of each DOC concentration was measured using Biochrom Ultrospec 142
1000 UV/Visible Spectrophotometer and the corresponding SUVA was calculated. A chlorine 143
stock solution of 5mg Cl2/mL was prepared using a 5% aqueous sodium hypochlorite solution 144
following Standard Method 5710B (APHA, 1995). A bromide ion stock solution of 100 µg/mL 145
was prepared by diluting Sigma-Aldrich bromide standard (Catalog no. 17355) with deionized 146
water. The stock solutions were stored in a refrigerator at 2 ± 0.1°C and discarded after 7 days. 147
148
Two different DOC concentrations (3 and 5.25 mg/L, with corresponding SUVA of 6.34 and 149
7.12 L/mg-m), were investigated with multiple levels of Br- concentrations (0, 40, 80, 120 ppb), 150
chlorine doses (4.05, 5.1 mg/L), pH (6, 8.5), temperatures (8, 16, 25 °C) and reaction times (3, 8, 151
28, 48 hours). While the experimental combinations do not conform to a formal factorial 152
experimental design (e.g., Montgomery, 1993), the combinations of experimental runs in the 153
experimental design do support the estimation of higher-order models having interaction 154
(product) terms, as will be shown later in the paper. 155
156
For each level of DOC, a chlorine dose-chlorine residual curve was established to determine the 157
breakpoint chlorine concentration. Measurements of total and free residual chlorines were 158
performed following Standard Methods 4500-Cl F (APHA, 1995). A free residual chlorine 159
concentration of 2.0 mg/L was added to the breakpoint chlorine concentrations throughout the 160
Page 9
MANUSCRIP
T
ACCEPTED
ACCEPTED MANUSCRIPT
8
experiments to provide adequate free chlorine residuals. The pH of the samples was adjusted by 161
adding 2 N HCl or 2 N NaOH and measured using a 3000 VWR scientific pH meter. The pH 162
meter was calibrated daily using standard buffer solutions (pH = 4, 7, 10) covering the 163
experimental range. Temperature was controlled by performing the reactions in a water bath and 164
the reaction time was controlled using timer. 165
166
For each experiment, 1 L of water sample was prepared by mixing the required amount of stock 167
HA solution to deionized water. The desired amount of chlorine stock solution was added and 168
mixing was performed mechanically using a magnetic stirrer. Immediately after mixing, 100 mL 169
samples were placed in amber glass bottles and sealed, headspace free, with polyseal lined caps. 170
The reactions were suspended by adding 10 mg sodium thiosulphate to prevent further formation 171
of THMs after the designated reaction time. The samples were preserved in a refrigerator at 2 ± 172
0.1°C for analysis. The analyses for THMs were performed within two weeks. The ASU-0 173
method, which is a selective ion monitoring (SIM) approach, was applied for the determination 174
of chloroform, bromodichloromethane, dibromochloromethane and bromoform in a liquid matrix 175
(CAEAL, 2007). For this method, a 35 mL sample was placed in a tube containing 8 g of sodium 176
chloride. Two milliliters of tert-methyl butyl ether (MTBE) was added as the extraction solvent 177
to the sample and fluorobenzene was used as an internal standard (IS) to verify the efficiency of 178
extraction. The extracted sample aliquots were analyzed by gas chromatography with mass 179
spectrometry (GC/MS) using a VOCOL column. An autosampler was used to ensure precise 180
injections into the GC/MS. Prior to use, GC/MS was calibrated for CHCl3, BDCM, DBCM and 181
CHBr3 using calibration standards. The calibration standards, quality control samples and MTBE 182
Page 10
MANUSCRIP
T
ACCEPTED
ACCEPTED MANUSCRIPT
9
were used in each group of samples to confirm the analytical results. The detection limit of the 183
approach was 2 ppb at 95% probability of detection. 184
185
RESULTS 186
Effects characterization 187
A systematic investigation to determine the effects of bromide ions and other factors on CHCl3, 188
BDCM and DBCM formation in drinking waters was performed. The formation of CHBr3 was 189
found to be below the detection limit in most cases throughout the experiments and as such was 190
not included in the investigation. The effects of different factors are presented below. 191
192
Effects of bromide ions 193
The formation of THMs was investigated using different levels of bromide ion concentrations (0, 194
40, 80, 120 ppb). These reactions were performed at constant pH (6), temperature (8°C), DOC 195
(3.0 mg/L) and chlorine dose (4.05 mg/L). Increases in bromide ion concentrations resulted in 196
increased THMs formation during each reaction period (3, 8, 28, 48 hours). The graph in Figure 197
2 shows the THMs concentrations after a 48 hour reaction period for bromide ion concentrations 198
of 0, 40, 80 and 120 ppb. Repeated measurements of THMs were made for a run at a given 199
bromide ion concentration. The mean concentrations of THMs obtained from these repeated 200
measurements (as shown in Figure 2) were found to be 75.7, 84.4, 91.2 and 97.1 ppb for bromide 201
ion concentrations of 0, 40, 80 and 120 ppb respectively. The Kruskal-Wallis (K-W) test showed 202
that the THMs concentrations for 0, 40, 80 and 120 ppb bromide ion were statistically different 203
with a p-value of 0.0001 (Minitab, 2004). The increase in THMs may be attributed to the 204
reactions between the NOM and HOBr as well as shifts in chlorinated THMs to brominated 205
Page 11
MANUSCRIP
T
ACCEPTED
ACCEPTED MANUSCRIPT
10
THMs as hypothesized in Figure 1. The effects of bromide ion concentrations on specific THMs 206
showed variable results for CHCl3, BDCM and DBCM formation. Increase in bromide ion 207
concentration from 40 to 120 ppb decreased CHCl3 formation from 68 to 60.1 ppb (-7.9 ppb) 208
during the 48 hour reaction period (Figure 3a). During the same reaction period, BDCM 209
formation was increased from 13.5 to 29.2 ppb (+15.7 ppb) when bromide ion was increased 210
from 40 to 120 ppb (Figure 3b). Increase in the formation of DBCM was observed by 4.8 ppb 211
(2.7 to 7.5 ppb) during this reaction period (Figure 3c). However, the decrease in chlorinated 212
THMs (CHCl3) and the increase in brominated THMs (BDCM, DBCM, CHBr3) were not 213
balanced (weight and molecular basis). For example, 0.07 µmol/L CHCl3 was reduced during 214
this 48 hour reaction period, while BDCM and DBCM were increased by 0.1µmol/L and 215
0.02µmol/L respectively. This variation may be due to the differences between the reactions of 216
HOBr and HOCl with NOM as well as conversion of bromide ions into brominated THMs (Sohn 217
et al. 2006). 218
219
The current study shows decrease of CHCl3 and increase of BDCM with increase in bromide 220
ions. However, Uyak and Toroz (2007) reported decrease of CHCl3 and BDCM with increase of 221
bromide ions. Their study showed initial increase of DBCM, which was decreased with further 222
increase in bromide ions, while CHBr3 was increased in most cases. THMs was increased in all 223
cases (Uyak and Toroz 2007), possibly due to increase of higher order brominated species (e.g., 224
DBCM and CHBr3). Increase in THMs with increase in bromide ions is consistent to the current 225
study (Figure 2). The difference between the current study and Uyak and Toroz (2007) may be 226
due to higher contents of bromide ions in their water (0.05-4.0 mg/L) than the current study 227
(0.04-0.12 mg/L), pH, temperature and site specific property of the NOM in their source water 228
Page 12
MANUSCRIP
T
ACCEPTED
ACCEPTED MANUSCRIPT
11
(Buyukcekmece lake water). The current study was controlled at pH of 6 and temperature of 229
8°C, while Uyak and Toroz (2007) performed their reactions at higher pH and temperature (pH = 230
7; temperature = 25°C). Higher pH and higher temperature might have affected the conversion of 231
bromide ions in their study (Sohn et al., 2006). In addition, higher contents of bromide ions 232
might have shifted THM speciation from chlorinated and lower order brominated species to 233
higher order brominated species in their study as discussed in Cowman and Singer (1996). 234
235
Effects of ratio between bromide ions and chlorine dose 236
The effects of bromide ion concentration to chlorine concentration ratio on CHCl3, BDCM and 237
DBCM are shown in Figure 4. Increasing the bromide ion to chlorine ratio decreased CHCl3 238
formation during each reaction period (Figure 4a). Increasing the bromide ion to chlorine ratio 239
from 9.9 to 29.6 decreased CHCl3 formation from 68 to 60.1 ppb over the 48 hour reaction 240
period. Similar trends were observed for the other reaction periods (3, 8 and 28 hours). The 241
formation of BDCM was increased for each of these reaction periods (Figure 4b). For example, 242
an increase of bromide ion to chlorine ratio from 9.88 to 29.6 increased BDCM from 13.5 to 29.2 243
ppb over the 48 hour reaction period. For similar conditions, DBCM formation increased from 244
2.7 to 7.5 ppb (Figure 4c). Increasing the bromide ion to chlorine ratio gradually shifted the 245
chlorinated THMs into mixed brominated THMs (Uyak and Toroz, 2007; Nokes et al., 1999). 246
However, the conversion of bromide ions to brominated THMs also depends on pH and 247
temperature (Sohn et al., 2006; Liang and Singer, 2003). Consequently, shifting of chlorinated 248
THMs to brominated THMs may also be a function of pH and temperature. 249
250
251
Page 13
MANUSCRIP
T
ACCEPTED
ACCEPTED MANUSCRIPT
12
Effects of pH 252
The effects of pH on CHCl3, BDCM and DBCM formation were investigated using various 253
reaction times and bromide ion concentrations. Figure 5 shows the effects of pH at different 254
bromide ion concentrations. At an 80 ppb bromide ion concentration, the formation of CHCl3 255
increased from 63 to 87.6 ppb during the 48 hour reaction period when pH was increased from 6 256
to 8.5 (Figure 5a). When pH was increased from 6 to 8.5, BDCM formation decreased from 22.8 257
to 17.5 ppb over 48 hour reaction period (Figure 5b) and DBCM formation decreased from 5.4 to 258
3.1 ppb (Figure 5c). Similar patterns were also observed for 40 ppb bromide ion concentrations 259
(Figure 5). In addition, at 25°C and 120 ppb bromide ion concentration, CHCl3 formation was 260
increased from 104.6 to 118.9 ppb when pH was increased from 6 to 8.5. Under the same 261
condition, BDCM formation was decreased from 34.4 to 32.3 ppb and DBCM was found 262
unchanged to 6 ppb (not shown). 263
264
Effects of temperature 265
The effects of temperature on CHCl3, BDCM and DBCM formation were observed to be 266
consistent. For samples with 120 ppb bromide ion concentration and 48 hour reaction period, 267
CHCl3 concentration increased from 89.6 ppb to 104.6 ppb when temperature was increased 268
from 16°C to 25°C. The formation of BDCM increased from 27.4 ppb to 34.4 ppb and DBCM 269
increased from 3.9 ppb to 6.1 ppb under the same conditions. Reaction rates appeared to increase 270
with increasing temperature, leading to the increase in CHCl3, BDCM and DBCM formation. 271
272
273
274
Page 14
MANUSCRIP
T
ACCEPTED
ACCEPTED MANUSCRIPT
13
Overall correlations 275
The correlation analysis in Table 1 shows the overall correlation structure of different factors 276
with BDCM formation. The BDCM concentrations had higher correlations with chlorine dose 277
(0.72), bromide ions (0.78) and temperature (0.59), while reaction time and DOC had relatively 278
smaller correlations (Table 1). The pH had negative effect (correlation coefficient = -0.2) on 279
BDCM formation. The similar effects of pH on BDCM have been shown in the preceding 280
sections of this paper. However, the negative effect of pH on BDCM formation has not been well 281
investigated in the existing literature. The correlations among the other factors can also be found 282
in Table 1. Further details of the correlation can be found in Montgomery and Runger (2007) 283
284
Model development 285
A full first-order plus interaction terms model for BDCM formation including six factors (DOC, 286
chlorine dose, pH, temperature, reaction time and bromide ion concentration) was developed 287
using an ordinary least squares regression applied to the data from the experiments. The 288
calculations were conducted using the JMPTM statistical package (SAS Inc., 2007). The 289
significant factors for the BDCM formation were determined through graphical and quantitative 290
diagnostics. The results for the regression analysis using JMPTM are summarized in Table 2, 291
which presents the estimated parameter values, their standard errors, and summarizes the 292
hypothesis test results for the significance of each parameter. The statistical significance is 293
represented in terms of Prob>|t| in Table 2. If the value of Prob>|t| for a factor is less than 0.05, 294
the factor is considered to be statistically significant (Montgomery and Runger, 2007). From 295
Table 2, it can be seen that DOC, chlorine dose, pH, temperature, reaction time, bromide ion 296
concentration and the interaction between reaction time and bromide ion concentration are 297
statistically significant. The statistical significance of the parameters can also be assessed 298
Page 15
MANUSCRIP
T
ACCEPTED
ACCEPTED MANUSCRIPT
14
graphically using a Half-Normal probability plot shown in Figure 6. From Figure 6, the largest 299
effect is seen to be associated with chlorine dose, while DOC, bromide ion concentration, 300
temperature, reaction time and pH have shown moderate effects on BDCM formation. The 301
interaction between reaction time and bromide ion concentration has a comparatively less, but 302
still statistically significant effect on BDCM formation (Figure 6, Table 2). 303
304
The contour plots in Figure 7 show the effects of simultaneous variability of chlorine dose, DOC, 305
temperature, reaction time, pH and bromide ion on BDCM formation. For example, by adjusting 306
chlorine dose and bromide ion simultaneously, the same concentration of BDCM is obtained 307
(Figure 7a). By varying the other factors simultaneously, such as chlorine dose and reaction time 308
(Figure 7b), reaction time and bromide ion (Figure 7d) and chlorine dose and DOC (Figure 7e), 309
desired results can be found. Curvature in the contour indicates the presence of two-factor 310
interactions. Contour plots in Figure 7d indicate that reaction time and bromide ion have an 311
interaction effect (presence of curvature), which is consistent with the results presented in Table 312
2. Figure 8 is an interaction plot for all factors that can be used to visually assess interaction 313
effects between factors (SAS Inc., 2007). An interaction effect exists when the impact of one 314
factor on the response depends on the level of another factor. Such an effect shows up in an 315
interaction plot when the lines of perturbation in one factor with the other held constant are not 316
parallel. From Figure 8, it can be seen that reaction time and bromide ion concentration have 317
interaction effects. At higher bromide ion concentration, increasing reaction time increased 318
BDCM formation (Figure 8). 319
320
Page 16
MANUSCRIP
T
ACCEPTED
ACCEPTED MANUSCRIPT
15
The parameter estimates for the final model are summarized in Table 3. Further details on the 321
generalized multiple regression technique and parameter estimation is provided in supplementary 322
materials. The details of parameter estimates can also be accessed through the 323
statistical/modeling books (e.g., Montgomery and Runger, 2007). The predictive model for 324
BDCM formation can be represented as: 325
326
)18.88)(93.19(7
6543221
−−
+++++++=−
−
Brtβ
BrβtβTβpHβClβDOCββBDCM ο (1) 327
328
where BDCM is the concentration of bromodichloromethane in water (µg/L), DOC is the 329
dissolved organic carbon (mg/L), Cl2 is the administered chlorine dose (mg/L), T is the reaction 330
temperature (°C), t is the reaction time (hour), Br - is the concentration of bromide ion in water 331
(µg/L), and β are the model parameters (regression coefficients). 332
333
The analysis of variance of the model and model lack of fit (LOF) test are shown in Table 4. 334
Tables 3 and 4 indicate that the model is statistically significant (Prob>F < 0.0001) and the lack 335
of fit is statistically insignificant (P = 0.398). The coefficient of determination (R2) for the model 336
was found to be 0.97, which represents strong linear trend among the model predictions and 337
experimental data. The lack of fit test decomposes the residual sum of squares into two parts – 338
one associated strictly with noise in the data that is seen in the replicate runs (known as the “pure 339
error”), and the other containing the effects of noise along with possible model mis-specification 340
error (known as the “mean square lack of fit”). The pure error mean square provides a good 341
estimate of the noise variance. If the mean square lack of fit is larger than the pure error variance 342
in a statistical sense, this indicates a lack of fit. In Table 4, the lack of fit test shows that the mean 343
Page 17
MANUSCRIP
T
ACCEPTED
ACCEPTED MANUSCRIPT
16
square lack of fit is not significantly different from the pure error variance, providing additional 344
evidence that the model provides an adequate fit to the data. Additional statistical diagnostics, 345
including the coefficient of determination and MSR/MSE ratio are available elsewhere 346
(Montgomery, 1993; Montgomery and Runger, 2007). 347
348
The model predictions were compared with the experimental data and observed to have a very 349
good fit to the data, with modest variability but strong agreement in trend ((R2 = 0.97). The 350
residuals indicate no discernable trends, again suggesting that the model is providing an adequate 351
fit to the data. The residuals were confirmed to be normally distributed, and additional plots of 352
residuals versus factors indicated no discernable trends. The experimental data and model 353
predictions were plotted in increasing order of THMs concentrations in order to verify the model 354
performance throughout the range of THMs concentrations. The predictions of the model were 355
observed to be consistent with the experimental data throughout the full range of THMs 356
concentrations, which indicates negligible systematic under/over predictions. On the basis of the 357
graphical and quantitative diagnostics, the final model was confirmed to provide an adequate fit 358
to the data. 359
360
Model validation 361
The model was validated by using the data from 17 water supply systems across Newfoundland 362
and Labrador, Canada. The Department of Environment (DOE) collected these data throughout 363
2004 to 2007 (DOE, 2008). Newfoundland is located in the Eastern part of Canada and 364
surrounded by the Atlantic Ocean. Most part of Newfoundland is situated in the coastal region. 365
Approximately 75 percentage of population in Newfoundland are served with more than 400 366
Page 18
MANUSCRIP
T
ACCEPTED
ACCEPTED MANUSCRIPT
17
water supply systems throughout the province (DOE, 2008). More than 95 percent of the water 367
supply systems in the province use chlorine as the primary disinfectant. Many of these water 368
supply systems have reported bromide ions in their source waters in the ranges between 7.6 and 369
550 ppb (DOE, 2008). The variability of bromide ions can be attributed to the inorganic 370
compounds, bio-geochemical processes, leaching due to the ocean water intrusion and 371
environmental conditions. The wide ranges of bromide ions as well as other water quality and 372
operational parameters produce wide ranges of BDCM concentrations in drinking waters. The 373
Department of Environment (DOE) in Newfoundland has also reported variable concentrations 374
of BDCM (0-176 ppb) across the province (DOE, 2008). Application of the developed model to 375
these water supply systems for validation will assist in understanding the model’s performance 376
under such a wide range of bromide ions. The summary of the values for the model parameters 377
of the selected 17 water supply systems are shown in Table 5. The model predictions and the 378
measured values of BDCM concentrations are shown in Figure 9. The correlation coefficient (r) 379
for the predicted and measured values of BDCM was found to be 0.93. The data were found to 380
be modestly consistent with the line of equal concentration in the graph (Figure 9). The data 381
orders and respective measured and predicted concentrations of BDCM are plotted in Figure 10. 382
From Figure 10, it can be seen that the model predictions follow the similar trend to the 383
measured data for the whole range of concentrations. To check the possible systematic under or 384
over predictions, the model predictions were compared by plotting in the increasing order of the 385
modeled data (not shown). The model predictions were found to be fairly consistent to the 386
measured data, which indicated no systematic under or over predictions with this model. 387
It is to be noted here that the R2 of the model was 0.97 (based on the experimental data), while 388
the correlation coefficient (r) between the measured and predicted BDCM in the validation study 389
Page 19
MANUSCRIP
T
ACCEPTED
ACCEPTED MANUSCRIPT
18
was 0.93, which is approximately 11% reduction in the efficiency in characterizing data 390
variability. The reduction in the model performance may be due to the differences in the 391
characteristics of NOM, uncertainty as well as other factors. 392
393
SUMMARY AND CONCLUSIONS 394
Chlorinated waters produce known and unknown disinfection byproducts, some of which are 395
often characterized as human health concerns. Recent epidemiological studies on specific THMs 396
show stronger association of BDCM with still births, low birth weights and neural tube defects, 397
which prompted the Health Canada to regulate BDCM at 16 ppb in drinking water. This study 398
has presented the results of a lab-scale investigation using synthetic waters to determine the 399
effects of chlorine dose, humic substances (DOC), pH, temperature, reaction time and bromide 400
ion concentrations on THMs species formation, from which a model to predict BDCM formation 401
in the chlorinated supply waters was developed. 402
403
The formation of THMs in the presence of bromide ions is complex. The hydrophilic NOM have 404
lower molecular weights, which makes their removal through coagulations less effective. As 405
such, hydrophilic NOM tends to remain in finished waters; thus, there exists increased possibility 406
of higher BDCM formation in drinking water. In this study, the formation of THMs was 407
increased with increases in bromide ions, possibly due to reactions of HOBr with the NOM and 408
shifting of chlorinated THMs into the brominated THMs. CHCl3 was observed to decrease and 409
BDCM and DBCM were observed to increase with increasing bromide ion concentrations. As 410
the bromide ions to chlorine ratio increased, chlorinated THMs gradually shifted to brominated 411
THMs. The effects of pH on the formation of CHCl3, BDCM and DBCM were observed to be 412
Page 20
MANUSCRIP
T
ACCEPTED
ACCEPTED MANUSCRIPT
19
variable. CHCl3 formation increased with pH while BDCM and DBCM formation decreased. 413
CHCl3 formation can be reduced by lowering pH, which represents THMs for waters having 414
insignificant bromide ion concentration. However, lowering the pH for waters with significant 415
bromide ion concentrations needs to be carefully monitored as this action could lead to an 416
increase in BDCM formation. Thus, the possibility of regulatory non-compliance of BDCM 417
concentration increases. The kinetics and mechanisms for reduction of brominated THMs at 418
increased pH are not well documented in the literature and should be the focus of future research. 419
420
The control of THMs and BDCM in drinking water is important from a regulatory perspective. 421
Despite the availability of a number of models for predicting total THMs and BDCM formation, 422
THMs formation models for source waters, which do not have significant bromide ion 423
concentrations, may not be efficiently applicable to waters having significant bromide ion 424
concentrations. In this study, following the characterization of the effects of different factors, a 425
model for predicting BDCM formation was developed. The model adequacy was tested 426
statistically using numerical and graphical diagnostics including residual plots and the 427
determination of a corresponding coefficient of determination, for which R2 value of 0.97 was 428
obtained. The model was found to be statistically significant and the lack of fit was insignificant. 429
The residuals plot demonstrated no visible trend. The predicted and experimental BDCM graph 430
showed no systematic deviation, indicating that the model served well throughout the entire 431
range of experiment conditions. This model was validated using external data from 17 water 432
supply systems in Newfoundland, Canada. The correlation coefficient for the measured and 433
predicted values was found to be 0.93. The model performance was found to be excellent under a 434
wide range of bromide ions and BDCM concentrations and no systematic under or over 435
Page 21
MANUSCRIP
T
ACCEPTED
ACCEPTED MANUSCRIPT
20
prediction was noticed throughout the whole range of BDCM concentrations. This model could 436
be used in predicting BDCM formation, performing complex risk-cost tradeoff studies and 437
maintaining regulatory compliance. 438
439
ACKNOWLEDGEMENT 440
Financial support from Natural Sciences and Engineering Research Council of Canada (NSERC) 441
in the form of Canada Graduate Scholarship (CGS) and Queen’s University in the form of a 442
Queen’s Graduate Award (QGA) is gratefully acknowledged. 443
444
REFERENCES 445
APHA (American Public Health Association), AWWA (American Water Works Association), 446
and WEF (Water Environment Federation)., 1995. Standard methods for the examination of 447
water and wastewater, 19th edition, Washington DC, USA 448
Amy, G.L., Chadik, P.A., Chowdhury, Z.K., 1987. Developing models for predicting 449
trihalomethane formation potential kinetics. J. Am. Water Works Assoc.79(7), 89-96. 450
CAEAL (Canadian Association for Environmental Analytical Laboratories)., 2007. Available online at 451
http://69.20.236.164/welcome2.lasso?s=city&-session=caeal:820F70B7169bf1FC36jlSl2402ED 452
Chowdhury, S., Husain T., 2006. Evaluation of drinking water treatment technology: An 453
entropy-based fuzzy application. Journal of Environmental Engineering 132(10), 1264-1271. 454
Chowdhury, S., Champagne, P., Husain, T., 2007. Fuzzy approach for selection of drinking 455
water disinfectants. Journal of Water Supply: Research and Technology (Aqua) 56(2), 75-93 456
Chowdhury, S., Champagne, P., McLellan, J., 2008. Factors influencing formation of 457
trihalomethanes in drinking water: Results from a multivariate statistical investigation of the 458
Page 22
MANUSCRIP
T
ACCEPTED
ACCEPTED MANUSCRIPT
21
Ontario Drinking Water Surveillance Program database. Water Qual. Res. J. of Canada 459
42(2/3), 189-199. 460
Chowdhury, S. Champagne, P., McLellan, J., 2009. Models for predicting disinfection 461
byproducts (DBPs) formation in drinking waters: A chronological review. Sci. of the Total 462
Environment 407(14): 4189-4206 463
Cowman, G.A., Singer, P.C., 1996. Effect of bromide ion on haloacetic acid speciation resulting 464
from chlorination and chloramination of aquatic humic substances. Environ. Sci. Technol. 465
30(1), 16–24. 466
Clark, R.M., Adams, J.Q., Lykins, B.W. Jr., 1994. DBP control in drinking water: cost and 467
performance. Journal of Environmental Engineering 120(4), 759-781. 468
Clark, R.M., Adams, J.Q., Sethi, V., Sivaganesan, M., 1998. Control of microbial contaminants 469
and disinfect ion by-products for drinking water in the US: cost and performance. J Water 470
SRT-Aqua 47(6), 255-265. 471
Clark, R.M., Thurnau, R.C., Sivaganesan, M., Ringhand, P., 2001. Predicting the formation of 472
chlorinated and brominated by-products. Journal of Environmental Engineering 127(6), 493-473
501. 474
Dodds, L., King, W.D., 2001. Relation between trihalomethane compounds and birth defects 475
Occup Environ Med 58, 443-446 476
Dodds, L., King, W.D., Allen, A.C., Armson, B.A., Fell, D.B., Nimrod, C., 2004. 477
Trihalomethanes in Public Water Supplies and Risk of Stillbirth. Epidemiology 15, 179-186. 478
DOE (Department of Environment). 2008. Personal communication with Department of 479
Environment: Municipal Water Supply, St. John’s, Newfoundland, Canada. This 480
Page 23
MANUSCRIP
T
ACCEPTED
ACCEPTED MANUSCRIPT
22
communication was made during October 2008. Partial information is also available at 481
(http://www.env.gov.nl.ca/env/SourceToTap/THMReport/THMReport.asp) 482
Elshorbagy, W.E., Abu-Qdais, H., Elsheamy, M.K., 2000. Simulation of THMs species in water 483
distribution systems. Water Research 34(13), 3431-3439 484
Gang, D.D., Segar, R.L. Jr., Clevenger, T.E., Banerji, S.K., 2002. Using chlorine demand to 485
predict THM and HAA9 formation. J. Am. Water Works Assoc. 94(10), 76-86 486
Golfinopoulos, S.K., Xilourgidis, N.K., Kostopoulou, M.N., Lekkas, T.D., 1998. Use of a 487
multiple regression for predicting trihalomethane formation. Water research 32(9), 2821-488
2829. 489
Health Canada, 2008 Guidelines for Canadian Drinking Water Quality. Prepared by the Federal-490
Provincial-Territorial Committee on Health and the Environment: March, 2008 491
Hellur-Grossman, L., Manka, J., Lamoni-Relis, B., Rebhun, M., 2001. THM, haloacetic acids 492
and other organic DBPs formation in disinfect ion of bromide rich Sea of Galilee (Lake 493
Kinneret) water. Water Science and Technology: Water Supply 1(2), 259-266 494
IRIS (Integrated Risk Information System)., 2008. Online database, Available online at: 495
http://www.epa.gov/iris/subst/index.html; US Environmental Protection Agency, Washington 496
D.C. 497
King, W.D., Marrett, L.D., 1996. Case-control study of bladder cancer and chlorination by 498
products in treated water (Ontario, Canada). Cancer Causes and Controls 7, 596-604 499
King, W.D., Dodds, L., Allen, A.C., 2000. Relation between stillbirths and specific chlorinated 500
byproducts in public water supplies. Environmental Health Perspectives 108(9), 883-886 501
Page 24
MANUSCRIP
T
ACCEPTED
ACCEPTED MANUSCRIPT
23
Lekkas, T.D., Nikolaou, A.D. 2004. Development of predictive models for the formation of 502
trihalomethanes and haloacetic acids during chlorination of bromide ion rich water. Water 503
Qual. Res. J. Canada 39(2), 149-159. 504
Liang, L., Singer, P.C., 2003. Factors influencing the formation and relative distribution of 505
haloacetic acids and trihalomethanes in drinking waters. Environmental Science and Technol. 506
37, 2920-2928 507
Mills, C.J., Bull, R.J., Cantor, K.P., Reif, J., Hrudey, S.E., Huston, P., An Expert Working 508
Group. 1998. Health Risks of Drinking Water Chlorination By-products. Chronic Dis. in 509
Canada 19(3), 91-102. 510
Minear, R., Morrow, C. 1983. Raw water bromide: levels and relationship to distribution of 511
trihalomethanes in finished drinking water. Research Report No. 91, Water Resources 512
Research Center, University of Tennessee. 513
Minitab Inc. Minitab Statistical Software., 2004. Available online at: http://www.minitab.com, 514
2004 515
MOE (Ministry of Environment, Ontario, Canada)., 2006. Drinking Water Surveillance Program 516
(DWSP) monitoring data for 2000-2004 on 179 municipal water supply systems (MWSS) in 517
Ontario. (Personal communication with Dave Fellowes, MOE, Ontario, Canada and website 518
http://www.ene.gov.on.ca/envision/water/dwsp/0002/index.htm) 519
Montgomery, D.C., 1993. Design and analysis of experiments, 3rd edition, John Wiley and Sons 520
Inc., NY, USA, 1993 521
Montgomery, D.C., Runger, G.C., 2007. Applied statistics and probability for engineers. 4th 522
edition. John Wiley and Sons, New York, USA. 523
Page 25
MANUSCRIP
T
ACCEPTED
ACCEPTED MANUSCRIPT
24
MWH (Montgomery Watson Harza)., 2005. Water Treatment: Principles and Design, John 524
Wiley & Sons, NJ, USA 525
Navalon, S, Alvaro, M., Garcia, H. 2008. Carbohydrates as trihalomethanes precursors. Influence 526
of pH and the presence of Cl- and Br- on trihalomethane formation potential. Water Research 527
42(14): 3990-4000. 528
Nokes, C.J., Fenton, E., Randall, C.J., 1999. Modeling the formation of brominated 529
trihalomethanes in chlorinated drinking waters. Water Research 33(17), 3557-3568 530
Reiff, F.M., 1995. Balancing the chemical and microbial risks in the disinfection of drinking 531
water supplies in developing countries. In: Assessing and Managing Health Risks from 532
Drinking Water Contamination: Approaches and Applications. IAHS Pub. No. 233. 533
Richardson, S.D., 2005. New disinfection by-product issues: emerging DBPs and alternative 534
routes of exposure. Global Nest Journal 7(1), 43-60. 535
Richardson, S.D., 2002. Simmons JE and Rice G. Disinfection byproducts: the next generation, 536
Environ. Sci. Technol. 36 (9), 198A-205A. 537
Rodrigues, M.S.M., Silva, J.C.G., Antunes, M.C.G., 2007. Factorial analysis of the 538
trihalomethane formation in water disinfection using chlorine. Analytica Chemica Acta 595, 539
266-274 540
Rodriguez, M., Milot, J., Serodes, J.B., 2003. Predicting trihalomethane formation in chlorinated 541
waters using multivariate regression and neural network. Journal of Water Supply: Research 542
and Technology (Aqua) 52(3), 199-215 543
SAS Inc., 2007. Statistical discovery software. SAS Institute Inc. (http://www.jmp.com/), NC, 544
USA 545
Page 26
MANUSCRIP
T
ACCEPTED
ACCEPTED MANUSCRIPT
25
Serodes, J.B., Rodriguez, MJ.., Li, H., Bouchard, C., 2003. Occurrence of THMs and HAAs in 546
experimental chlorinated waters of the Quebec City area (Canada). Chemosphere 51, 253–547
263. 548
Sohn, J., Amy, G., Cho, J., Yoon, Y., 2006. Bromide ion incorporation into brominated 549
disinfection by-products. Water, Air, and Soil Pollution 174, 265–277 550
Stevens, A.A., Slocum, C.J., Seeger, D.R., Robeck, C.B., 1976.. Measurement of THM and 551
precursor concentration changes. J. Am. Water Works Assoc. 68, 546-554 552
USEPA (US Environmental Protection Agency)., 2006. National Primary Drinking Water 553
Regulations: Stage 2 Disinfectants and Disinfection By products Rule: Final Rule. Federal 554
Register 71(2), January 4, 2006. 555
Uyak, V., Toroz, I., 2007. Investigation of bromide ion effects on disinfection by-products 556
formation and speciation in an Istanbul water supply. Journal of Hazardous Materials 149, 557
445–451 558
Villanueva, C.M., Cantor, K.P., Cordier, S., Jaakkola, J.J.K., King, W.D., Lynch, C.F., Porru, S., 559
Kogevinas, M., 2004. Disinfection Byproducts and Bladder Cancer: a Pooled Analysis. 560
Epidemiology 15(3), 357-367. 561
Waller, K., Swan, S.H., DeLorenze, G., Hopkins, B. 1998. Trihalomethanes in drinking water 562
and spontaneous abortion. Epidemiology 9(2), 134-140. 563
Westerhoff, P., Debroux, J., Amy, G.L., Gatel, D., Mary, V., Cavard, J., 2000. Applying DBP 564
models to full-scale plants. J. Am. Water Works Assoc. 92(3), 89-102 565
Wigle, D.T., 1998. Safe Drinking Water: A Public Health Challenge. Chronic Dis. in Canada 566
19(3), 103-107. 567
568
Page 27
MANUSCRIP
T
ACCEPTED
ACCEPTED MANUSCRIPT
Table 1. Correlation analysis for the factors of BDCM formation
DOC Cl2 dose Temperature Time pH Bromide ion BDCM DOC 1.00 0.13 0.88 -0.07 0.01 0.52 0.35 Cl2 dose 1.00 0.33 0.13 -0.06 0.38 0.72 Temperature 1.00 0.04 0.08 0.62 0.59 Time 1.00 0.002 0.08 0.44 pH 1.00 -0.21 -0.20 Bromide ion 1.00 0.78 BDCM 1.00
Page 28
MANUSCRIP
T
ACCEPTED
ACCEPTED MANUSCRIPT
Table 2. Screening effects of the factors for BDCM formation
Factor Estimate Std Error t Ratio Prob>|t| Comments DOC -2.45 0.56 -4.36 0.0001 Significant Cl2 dose 3.81 0.34 11.33 <.0001 Significant pH -0.78 0.27 -2.88 0.0074 Significant Temperature 0.56 0.12 4.61 <.0001 Significant Time 0.15 0.02 10.18 <.0001 Significant Bromide ion 0.14 0.02 8.20 <.0001 Significant (DOC-4.18)*(pH-7.08) 0.60 0.36 1.66 0.1079 (DOC-4.18)*(Time-19.93) -0.06 0.03 -1.81 0.0801 (DOC-4.18)*(Bromide ion-88.18) 0.00 0.02 -0.11 0.9130 (Cl2 dose-4.19)*(Time-19.93) 0.02 0.02 0.82 0.4194 (pH-7.08)*(Time-19.93) 0.00 0.01 -0.15 0.8808 (pH-7.08)*(Bromide ion-88.18) 0.00 0.01 -0.29 0.7761 (Temperature-14.02)*(Time-19.93) 0.00 0.01 0.39 0.7017 (Time-19.93)*(Bromide ion-88.18) 0.00 0.00 2.89 0.0073 Significant
Page 29
MANUSCRIP
T
ACCEPTED
ACCEPTED MANUSCRIPT
Table 3. Parameter Estimates for BDCM formation model
Parameter Estimate Std Error t Ratio Prob>|t|
Intercept βo -3.6 2.99 -1.20 0.2372
DOC β1 -2.43 0.54 -4.53 <.0001
Cl2 dose β2 3.47 0.3 11.66 <.0001
pH β3 -1 0.26 -3.79 0.0006
Temperature β4 0.58 0.1 5.62 <.0001
Time β5 0.167 0.016 10.49 <.0001
Br- β6 0.144 0.013 11.36 <.0001
(Time-19.93)*(Br- -88.18) β7 0.0012 0.00051 2.36 0.0239
Page 30
MANUSCRIP
T
ACCEPTED
ACCEPTED MANUSCRIPT
Table 4. Quantitative Diagnostics for Final Second-order Model for BDCM formation
Source DF Sum of Squares Mean Square F Ratio R2 Model 7 3458.9307 494.133 161.0003 Error 36 110.4892 3.069
C. Total 43 3569.4199 Prob > F <.0001
Lack Of Fit 33 104.51982 3.16727 1.5918 Pure Error 3 5.96935 1.98978 Total Error 36 110.48917
Prob > F 0.3981
0.97
Page 31
MANUSCRIP
T
ACCEPTED
ACCEPTED MANUSCRIPT
Table 5. Values of the factors for the model validation study
Parameter Minimum Most Likely Maximum
DOC (mg/L) 1.3 3.7 8
Cl2 dose (mg/L) 2.4 5.1 8.9
pH 5.4 7.1 8.5
Temperature (°C) 5.9 15.1 25.7
Time (Hour) 10.7 16.1 22.4
Br- (µg/L) 7.6 100.2 550
Page 32
MANUSCRIP
T
ACCEPTED
ACCEPTED MANUSCRIPT
Figure 1. Possible formation pathways for brominated THMs in drinking water
NOM
Hydrophobic NOM
Hydrophilic NOM
+ HOCl
+ HOBr
Chlorinated THMs
→
Brominated THMs
→
+ HOBr
Brominated THMs
Chlorinated THMs
Page 33
MANUSCRIP
T
ACCEPTED
ACCEPTED MANUSCRIPT
Br- (ppb)
TH
Ms
(ppb
)
12080400
105
100
95
90
85
80
75
70
97.145
91.2267
84.3583
75.7167
Figure 2. Effects of bromide ion on THMs formation (Temperature =8°C, pH = 6, DOC = 3.0
mg/L; [SUVA = 6.34 L/mg-m], chlorine dose = 4.05 mg/L, reaction time = 48 hours)
Page 34
MANUSCRIP
T
ACCEPTED
ACCEPTED MANUSCRIPT
0
20
40
60
80
0 20 40 60Time (hour) .
CHCl3
(ppb) .
40 ppb
80 ppb
120 ppb
(a)
0
10
20
30
40
0 20 40 60Time (hour) .
BDCM
(ppb) .
40 ppb80 ppb120 ppb
(b)
0
2
4
6
8
10
0 20 40 60Time (hour) .
DBCM
(ppb) . 40 ppb 80 ppb
120 ppb
(c)
Figure 3. Effects of Br- on CHCl3, BDCM and DBCM (Cl2 dose = 4.05 mg/L; pH = 6, DOC = 3.0 mg/L; [SUVA = 6.34 L/mg-m], Temperature=8°C)
Page 35
MANUSCRIP
T
ACCEPTED
ACCEPTED MANUSCRIPT
0
40
80
120
5 12 19 26 33Br- (ppb)/Cl2 (ppm) .
CHCl3 (ppb) .
3-Hour 8 Hour28 Hour 48 Hour
(a)
0
9
18
27
36
45
5 12 19 26 33Br- (ppb)/Cl2 (ppm) .
BDCM (ppb) .
3-Hour 8 Hour28 Hour 48 Hour
(b)
0
2
4
6
8
10
12
5 12 19 26 33Br-(ppb)/Cl2 (ppm) .
DBCM (ppb) .
3-Hour 8 Hour28 Hour 48 Hour
(c)
Figure 4. Effects of Br-/chlorine on CHCl3, BDCM and DBCM formation (pH = 6, DOC =3.0 mg/L; [SUVA = 6.34 L/mg-m], Temperature=8°C)
Page 36
MANUSCRIP
T
ACCEPTED
ACCEPTED MANUSCRIPT
0
20
40
60
80
100
0 20 40 60Time (hour) .
CH
Cl3
(ppb) .
80 ppb;pH=8.5 80 ppb; pH=640 ppb;pH=6 40ppb;pH=8.5
(a)
0
10
20
30
40
0 20 40 60Time (hour) .
BD
CM
(ppb) .
80 ppb;pH=8.5 80 ppb; pH=640 ppb; pH = 6 40 ppb; pH=8.5
(b)
0
2
4
6
8
10
0 20 40 60Time (hour) .
DB
CM
(ppb) .
80 ppb;pH=8.5 80 ppb; pH=640ppb;pH=6 40ppb;pH=8.5
(c)
Figure 5. Effects of pH on CHCl3, BDCM and DBCM formation (DOC = 3.0 mg/L; [SUVA = 6.34 L/mg-m], chlorine dose = 4.05 mg/L; Temperature=8°C)
Page 37
MANUSCRIP
T
ACCEPTED
ACCEPTED MANUSCRIPT
Figure 6. Half- normal plot for the factors effects on BDCM formation
0
5
10
15
20
25
DOC
Cl2 dose
pH
Temp. Time
Br-
(Time-19.93)*(Br--88.18)
0.0 0.5 1.0 1.5 2.0 2.5 3.0
Normal Quantile
No
rma
lize
d E
stim
ate
s (O
rtho
g t
)
Page 38
MANUSCRIP
T
ACCEPTED
ACCEPTED MANUSCRIPT
40
50
60
70
80
90
100
110
120
Bro
mid
e io
n
BDCM
2.350549
5.509496
8.668444
11.82739
14.98634 18.14529 21.30423 24.46318
1.5 2 2.5 3 3.5 4 4.5 5Cl2 dose
(a). Chlorine dose and bromide ion
10
20
30
40
Tim
e
BDCM
5.509496
8.668444
11.82739
14.98634
18.14529
21.3042324.46318
1.5 2 2.5 3 3.5 4 4.5 5Cl2 dose
(b). Chlorine dose and reaction time
10
15
20
25
Tem
pera
ture
BDCM
5.509496
8.668444
11.82739
14.9863418.14529 21.30423 24.46318 27.62213
1.5 2 2.5 3 3.5 4 4.5 5Cl2 dose
(c). Chlorine dose and temp.
40
50
60
70
80
90
100
110
120
Bro
mid
e io
n
BDCM
11.82739
14.98634
18.14529
21.30423 24.46318 27.62213
10 20 30 40Time
(d). Reaction time and bromide ion
3
3.5
4
4.5
5
DO
C
BDCM
8.668444
5.509496
11.82739 14.9863418.14529
21.30423
1.5 2 2.5 3 3.5 4 4.5 5Cl2 dose
(e). Chlorine dose and DOC
40
50
60
70
80
90
100
110
120
Bro
mid
e io
n
BDCM
14.98634
18.14529
21.30423
24.46318
11.82739
3 3.5 4 4.5 5DOC
(f). DOC and bromide ion
Figure 7. Effects of simultaneous variability of factors on BDCM formation
Page 39
MANUSCRIP
T
ACCEPTED
ACCEPTED MANUSCRIPT
010203040
BD
CM
010203040
BD
CM
010203040
BD
CM
010203040
BD
CM
010203040
BD
CM
010203040
BD
CM
DOC
1.1
5.1
68.5
8
25
348
40
120
33.
5 44.
5 55.
5
35.25
Cl2 dose
68.5
8
25
348
40
120
1.5
2.5
3.5
4.5
5.5
35.25
1.1
5.1
pH
8
25
348
40
120
6 7 8 9
35.25
1.1
5.1
68.5
Temperature
348
40
120
10
15
20
25
35.25
1.1
5.1
68.5
8
25
Time
40
120
10 20 30 40 50
35.25
1.1
5.1
68.5
8
25
348
Bromide ion
40 60 80 100
120
DO
CC
l2 dosepH
Tem
peratureT
ime
Brom
ide ion
Figure 8. Interaction of reaction time and Br- concentration on BDCM formation
Page 40
MANUSCRIP
T
ACCEPTED
ACCEPTED MANUSCRIPT
0
40
80
120
0 40 80 120Measured (ppb) .
Pre
dic
ted
(p
pb
)
.
Measured vs predicted data
Line of equal concentration
Figure 9. Model predictions and measured concentrations of BDCM in 17 water supply systems
Page 41
MANUSCRIP
T
ACCEPTED
ACCEPTED MANUSCRIPT
0
40
80
120
1 9 17
25
33
41
49
57
65
73
81
89
97
10
5
11
3
12
1
Data .
BD
CM
(p
pb
) .
Measured Predicted
Figure 10. Model predictions and measured BDCM concentrations (Data order)
