Mitigation of Disinfection By-Product Formation through

Mitigation of Disinfection By-Product Formation through Development

of a Multiple Regression Equation and a Bayesian Network

by

Brett Harper

A Thesis

presented to

The University of Guelph

In partial fulfillment of requirements

for the degree of

Master of Science

in

Engineering

Guelph, Ontario, Canada

© Brett Harper, May, 2012

ABSTRACT

MITIGATION OF DISINFECTION BY-PRODUCT FORMATION

THROUGH DEVELOPMENT OF A MULTIPLE REGRESSION EQUATION

AND A BAYESIAN NETWORK

Brett Harper Advisor:

University of Guelph, 2012 Professor Ed McBean

Disinfection By-Product (DBP) formation due to chlorination in drinking water is a

common issue encountered by WTP operators. Efforts to minimize DBP formation

are complicated by the presence of zebra mussels, which may inhabit the raw water

intake of WTPs.

Multivariate models for Total Trihalomethane (TTHM) and Haloacetic Acid (HAA)

subspecies are employed to show that in some instances pre-chlorination can be

reduced to lower DBP formation, while post-chlorination can be increased.

A second regression model for TTHM which includes bromide, a variable which was

previously unused, is described and demonstrates that DBP levels can be reduced by

lowering pre-chlorination levels.

Finally, a Bayesian network is developed using the Webweavr-IV Toolkit, utilizing

causal relationships between raw water quality parameters in the form of conditional

probabilities.

The results show that the average cancer risk can be decreased by while still

maintaining zebra mussel control and simultaneously decreasing the incremental

cancer risk.

iii

Acknowledgements

The author wishes to thank Ed McBean, Professor of Engineering at the University of

Guelph and Canada Research Chair in Water Supply Security, for his guidance,

support and ruthless editing throughout this study. The author is also grateful to Erica

Hamilton of the Ontario Ministry of the Environment (MOE), who provided useful

data for the project. The author is also thankful for the comprehensive database

supplied by the Drinking Water Surveillance Program (DWSP), which supplied the

majority of the data for the project.

The author would also like to thank Yang Xiang, who supplied the Bayesian Network

Software, Webweavr IV, and finally Zoe Zhu, my co-advisor and Adjunct Professor

at the University of Guelph, for her kind guidance, support, and technical help.

iv

Table of Contents

Chapter 1 .............................................................................................................................. 1

Introduction .......................................................................................................................... 1

1.1 Occurrence of Disinfection By-Products (DBPs) in Drinking Water ........................... 1

1.2 Health Effects of Human Exposure to DBPs ............................................................. 3

1.3 Zebra Mussels ........................................................................................................ 4

1.4 Multiple Regression Models for DBPs ..................................................................... 5

1.5 The Bromide Ion ..................................................................................................... 6

1.6 Bayesian Network Models for TTHMs ..................................................................... 6

1.7 Purpose of this Study .............................................................................................. 7

Chapter 2 .............................................................................................................................. 9

Paper #1: Attaining Zebra Mussel Control and Mitigating Disinfection By-Product Formation

in Drinking Water Treatment ................................................................................................. 9

Chapter 3 ............................................................................................................................ 38

Paper #2: Mitigation of Disinfection By-Product Formation by the Development of a

Regression Equation with the Bromide Ion .......................................................................... 38

Chapter 4 ............................................................................................................................ 61

Paper #3: Modelling and Mitigation of Disinfection By-Product Formation through the

Development of a Bayesian Network ................................................................................... 61

Chapter 5 ............................................................................................................................ 83

Conclusions ........................................................................................................................ 83

5.1 Summary ............................................................................................................. 83

5.2 Recommendations ............................................................................................... 85

References .......................................................................................................................... 87

List of Tables

Table 1: The health effects related to the major DBPs (modified from US EPA 1999a) .......... 3

Table 2: Current WHO Guidelines for DBPs in Drinking Water ............................................ 4


in Drinking Water Treatment

Table 1: The health effects related to the major DBPs (modified from US EPA 1999).......... 12

Table 2: Average Concentrations of TTHMs in DWSP – Monitored Surface Water Systems in

Ontario ................................................................................................................................ 15

v

Table 3: The significance of multivariate regression ............................................................ 17

Table 4: Datasets from Kenora water treatment plant before treatment strategies applied ..... 21

Table 5: Datasets from Kenora water treatment plant after treatment strategies applied ........ 22

Table 6: Slope factor for each subspecies and average slope factor for TTHM ..................... 23

Table 7: Cancer risk level of surface water drinking sources of TTHMs in Ontario WTPs.... 26

Table 8: Average Carcinogenic Risks from HAA and TTHM subspecies ............................. 30

Table 9: Average Carcinogenic Risks from TTHM and Sum of Trihalomethane Subspecies 31

Table 10: Reduction of TTHM Cancer Risk, Relative to Current Scenario ........................... 32


Regression Equation with the Bromide Ion

Table 1: The health effects related to the major DBPs (modified from USEPA 1998) .......... 40

Table 2: Slope factor for each TTHM subspecies ................................................................. 42

Table 3: Non Compliance for TTHM at various specified levels in all WTPs which report

TTHM ................................................................................................................................ 46

Table 4: The significance of multivariate regression for Equation 2 ..................................... 49


Development of a Bayesian Network

Table 1: The health effects related to the major DBPs (modified from USEPA 1999) .......... 63

Table 2: Raw Water Quality Parameter Assigned Ranges .................................................... 69

Table 3: Bayesian Network Results ..................................................................................... 73

Table 4: Comparison of Current, Average Conditions with the Four Scenarios .................... 74

List of Figures


in Drinking Water Treatment

Figure 1: Measured TTHM vs. Predicted TTHM for Calibration Dataset ............................. 18

Figure 2: Measured TTHM vs. Predicted TTHM for 2004 ................................................... 18

Figure 3: Seasonality of TTHM formation ........................................................................... 20


Regression Equation with the Bromide Ion

Figure 1: Sum of Trihalomethane Subspecies in Ontario WTPs ........................................... 56

vi

Figure 2: Subspecies formation based on bromide ion concentration (From Health Canada,

1995) .................................................................................................................................. 57

Figure 3: Percent speciation of TTHM subspecies given bromide concentrations in raw

drinking water ..................................................................................................................... 57

Figure 4: Cancer Risk for TTHM Concentration of 20µg/L Given Bromide Concentration in

Raw Water .......................................................................................................................... 58

Figure 5: TTHM Concentration when Bromide Concentration in Raw Water is Known ....... 58

Figure 6: Measured TTHM vs. Predicted TTHM from Equation 2 ....................................... 59

Figure 7: Measured TTHM vs. Predicted TTHM for Equations 1 and 2 ............................... 59

Figure 8: TTHM Cancer Risk with Altered Pre-Chlorination Dosage ................................... 60


Development of a Bayesian Network

Figure 1: Bayesian Network to Determine TTHM levels and Resulting Cancer Risk ............ 82

1

Chapter 1

Introduction

1.1 Occurrence of Disinfection By-Products (DBPs) in Drinking

Water

Utilization of chlorine at the beginning of the 20th century was likely the most

important beneficial action to safeguard human health. Prior to chlorination,

waterborne diseases were widely prevalent (e.g. the cholera outbreak in Chicago in

the 1800‟s) and were a major cause of death. Chlorine and its compounds continue as

the most commonly used disinfectant for water treatment facilities. The popularity of

chlorine is not only due to its low cost, but also to its high oxidizing potential, which

provides a chlorine residual throughout the distribution system and protects against

microbial recontamination (Sadiq and Rodriguez 2003). However in the 1970s, it was

discovered that chlorine used during water treatment reacts with organic matter, such

as humic and fulvic acids, and produces disinfection by-products (DBPs).

DBPs, unlike pathogens such as E.coli, do not display immediate health effects on

those that are exposed to them. DBPs, some of which are carcinogens, cause illness in

people over a lifetime, which makes it harder to quantify the risk they present to the

public, and more difficult to stress to governing bodies the need to regulate these

contaminants.

Haloacetic acids and trihalomethanes are the two major groups of chlorinated

disinfection by-products found in drinking water and generally, are at the highest

levels. Together, these two groups can be used as indicators for the presence of all

2

chlorinated disinfection by-products in drinking water supplies, and their control is

expected to reduce the levels of all chlorinated disinfection by-products and the

corresponding risks to health. This thesis will largely focus on THMs, but will address

HAAs as well.

Total Trihalomethanes (TTHMs) are by-products created when the chlorine used in

the water treatment plant (WTP) disinfection process reacts with naturally occurring

organics (e.g. organics formed by the decay of algae and vegetation) in raw water. In

addition to chlorination, TTHMs are known to also be influenced by temperature,

bromide, and the concentration of the dissolved organic carbon in the water (Health

Canada, 2006). The most common forms of trihalomethanes created are chloroform,

bromodichloromethane (BDCM), chlorodibromomethane (CDBM) and Bromoform

(Harper et al., 2012). Some trihalomethanes are volatile, though inhalation of

trihalomethanes due to volatilization is negligible when compared to consumption

rates through drinking water (WHO, 2011). While not all trihalomethanes can be

absorbed, chloroform can be absorbed through the skin, and contributes a significant

amount of the THM exposure that swimmers experience (New Hampshire DES,

2006).

Haloacetic Acids are by-products that are also created when chlorine used in water

treatment plant (WTP) disinfection processes react with naturally occurring organics.

In treated surface waters, Haloacetic acid concentrations are typically higher than

those found in treated ground water; which is due to the higher levels of organic

matter in surface waters. In general, HAA concentrations will also be higher in

warmer months because of higher concentrations of organic matter and, more

3

importantly, because DBP formation rates increase at higher temperatures. Due to the

fact that HAAs are not volatile, and cannot be absorbed by the skin, exposure through

routes other than consumption is considered negligible.

Currently, attempts to reduce exposure to THMs and HAAs are generally focused on

reducing their formation. Disinfection By-Product concentrations in drinking water

can be reduced at WTPs by removing the organic matter from the water before

chlorine is added, by optimizing the disinfection process, by using alternative

disinfection methods or by using a different water source. However, it is very

important that the method used to mitigate DBP formation does not interfere with the

effectiveness of the disinfection. This thesis will address the mitigation of DBP

formation by optimizing the disinfection process.

1.2 Health Effects of Human Exposure to DBPs

Through considerable efforts within epidemiological and toxicological studies, it has

been reported that there are potential adverse effects from exposure to some of the

DBPs. Table 1 summarizes the harmful effects of some of the important DBPs.

Table 1: The health effects related to the major DBPs (modified from US EPA 1999a)

Class of DBPs Compound Potential Health Effects

Total

Trihalomethanes

Chloroform Cancer, liver, kidney and

reproductive effects

Dibromochloromethane Nervous system, liver, kidney and


Bromodichloromethane Cancer, liver, kidney and


Bromoform Cancer, nervous system, liver and

kidney effects

Haloacetic acids Dichloracetic acid Cancer, reproductive, developmental

effects

Trichloracetic acid Liver, kidney, spleen, developmental

effects

4

These health risks have prompted several countries to establish maximum acceptable

concentration levels of DBPs, in particular Trihalomethanes (THMs) and Haloacetic

acids (HAAs) in treated drinking waters. In 2011, the World Health Organization

published drinking water quality guidelines for several DBPs including THMs and

HAAs, which establish a foundation for the countries around the world to promulgate

their regulations on these chemicals. Table 2 summarizes standards/guidelines related

to DBPs that are currently endorsed by the WHO. It is noted that these guidelines are

for all nations and that it is recommended that DBP levels in drinking water be kept as

low as practicable. Developed nations typically impose more stringent guidelines on

DBPs.

Table 2: Current WHO Guidelines for DBPs in Drinking Water

Class of DBPs Compound WHO Guideline (µg/L)

Total Trihalomethanes Chloroform 300

Dibromochloromethane 100

Bromodichloromethane 60

Bromoform 100

Haloacetic acids Dichloracetic acid 50

Trichloracetic acid 200

The current standards for TTHMs and HAAs set by Health Canada are 100µg/L and

80µg/L, respectively (Health Canada 2006). By comparison, the US EPA has set their

standards at 80µg/L and 60µg/L for TTHMs and HAAs respectively (US EPA, 1999).

1.3 Zebra Mussels

Zebra mussels (Dreissena polymorpha) are a class of mollusc similar to oysters,

clams, and scallops, which originated in the Black and Caspian seas and have been

inadvertently transported into the Great Lakes water by cargo ships. The mussels

grow to 1 inch in length and produce 35,000 eggs per season per female (Dermott et

al. 1993). The proliferation of zebra mussels is causing serious problems by clogging

5

raw-water intakes and discharge lines, increasing pipe corrosion, and producing

massive bio-fouling. As a result, zebra mussels are a recent problem for surface water

treatment facilities in parts of the US and Canada, particularly along the western

shores of Lake Erie. In response, most of these water treatment facilities chlorinate

their source water to control zebra mussels in the intake. As a consideration, options

which would involve timed dosages to control zebra mussels, and not pre-chlorinating

when not needed, would have favourable benefits in terms of reducing the formation

of disinfection by-products. This thesis evaluates several strategies to accomplish

lesser formation rates of the disinfection by-products while still controlling zebra

mussels.

1.4 Multiple Regression Models for DBPs

Of interest is to use multivariate regression analyses to develop predictive models for

DBP formation by relating DBP concentrations to various combinations of

explanatory variables, which include: chlorine, dissolved organic matter, pH,

temperature and bromide. When a validated model is derived, it can be used to

identify the significance of diverse operational and water quality parameters which

control or reduce DBP formation. For example, treatment plant operators can lower

DBP formation by reducing unnecessary chlorine dosages, while maintaining enough

residues to fight microbes.

McBean et al. (2008) and Harper et al. (2012) have developed regression equations to

model formation of TTHM concentrations in drinking water. However, these papers

did not include bromide in the regression equation because bromide data over a four

year period (2005-2008) have only recently been released by the DWSP.

6

1.5 The Bromide Ion

Health Canada has listed the bromide ion as one of the contributing factors in the

formation of TTHM (Health Canada, 1995). Until recently, however, only very

limited bromide data have been available; of interest is the structure of regression

models in response to the addition of the bromide ion.

Brominated subspecies have been shown to have significantly higher health effects

compared to non-brominated subspecies. A number of researchers (Morrow, 1987;

Chang, 2001; Kampioti, 2002; Uyak, 2007; Wang, 2007; and Sun, 2009) have

reported that speciation shifted to the bromine-substituted THMs as a function of

bromide concentration when all other parameters were held constant. Under

conditions of high natural organic matter (NOM) and low bromide concentrations,

chlorine-substituted by-products predominated, especially during longer reaction

times. In the presence of chlorine and organic material, as much as 50% of the

bromide ion may become incorporated into the brominated trihalomethane subspecies

(Chang et al., 2001); this efficiency of bromide incorporation implies that 100 µg/L of

bromide may result in up to 50 µg/L of THM-bound bromine (THM-Br). A reduction

in bromide concentration will have a significant impact on the concentration and

speciation of formed TTHM.

1.6 Bayesian Network Models for TTHMs

Chapter 4 employs a Bayesian network to predict TTHM concentration level by

utilizing causal relationships between raw water quality parameters and TTHMs, and

causal relationships between the water quality parameters.

7

Bayesian networks are a type of intelligent system that represents domain knowledge

with a graphical structure that uses nodes to represent variables and arcs between the

nodes to represent dependencies between variables. A Bayesian network quantifies

this knowledge structure with probabilistic expressions of the interaction among

variables. Since probabilities are logical ways to quantify the unknown, Bayesian

networks are ideal intelligent systems. Bayesian networks also possess the unique

ability to incorporate expert estimates, as well as observed evidence, which allows

them to unite different kinds of uncertainty in a single theoretical environment (Olson

et al, 1990; van der Gaag, 1996).

The WebWeavr IV Toolkit will be employed to model a Bayesian network that

utilizes the concentration levels of raw water parameters to predict concentration

levels of TTHM.

1.7 Purpose of this Study

The following primary study objectives are addressed in this thesis:

1) A regression equation is developed to model the formation of two known

DBPs: Total Trihalomethanes (TTHMs) and Haloacetic Acids (HAAs), in

treated Ontario drinking waters.

2) Zebra mussels are identified as an obstacle to optimizing the disinfection

process. The weaknesses of Zebra mussels are examined and exploited in

Chapter 2.

8

3) The seasonality of TTHM formation is examined to determine the most

effective times to combat zebra mussels with pre-chlorination.

4) Bromide is identified as a contributor to TTHM levels, and bromide data are

used along with other raw water quality parameters in the development of a

regression equation model to characterize the formation of TTHMs.

5) The contribution of bromide to the subspecies of total trihalomethanes

(TTHM) is examined. Bromide compounds have been demonstrated to

contribute to a large percentage of TTHM as bromide concentrations increase.

This is problematic since cancer slope factors for bromide compounds are ten-

fold higher than chloroform, meaning the potential cancer rates when bromide

is present, are greatly increased.

6) Current cancer risks from Trihalomethanes in treated Ontario drinking waters

are calculated and compared to current allowable Trihalomethane levels as

well as the „de minimus‟ risk.

7) A Bayesian network is developed using the Webweavr-IV Toolkit, utilizing

causal relationships between raw water quality parameters in the form of

conditional probabilities to predict TTHM levels and cancer risk.

9

Chapter 2

Paper #1: Attaining Zebra Mussel Control and Mitigating

Disinfection By-Product Formation in Drinking Water

Treatment Brett Harper, E. McBean and Zoe J. Y. Zhu

10

Abstract

A complicating factor to reduce disinfection by-product (DBP) formation arising from

chlorination for drinking water treatment systems using surface water as a supply

source, may be the presence of large populations of zebra mussels. Many treatment

facilities control zebra mussels via chlorination of the water to elevated levels at the

intake. The chlorine acts as an anti-foulant, but this approach increases the DBP

formation.

Methods for reducing DBPs are explored, including adjusting the location for chlorine

additions in the treatment sequence. Multivariate models for Total Trihalomethane

(TTHM) and Haloacetic Acid (HAA) subspecies are employed to show that pre-

chlorination can be reduced, and post-chlorination increases. Regression models (R2

of 0.75) predict that DBP levels can be lowered by post-chlorination rather than

chlorinating raw water for portions of the year except during the combatable life stage

for zebra mussel control. The current TTHM incremental cancer risks in Ontario are

demonstrated as higher than „de minimus‟ risk, and range from 1 in 50,000 to

100,000. The results show that the current incremental excess lifetime cancer risk can

be decreased by approximately 24 percent while still maintaining zebra mussel

control and simultaneously decreasing the incremental cancer risk.

Keywords: disinfection by-products (DBPs), Trihalomethanes (TTHM), Haloacetic

acids (HAA), Organic matter, Chlorine, Multivariate regression

11

1.0 Introduction

1.1 An Overview of Disinfection By-Products in Drinking Water and the

Zebra Mussel Problem

Utilization of chlorine at the beginning of the 20th century was likely the most

important beneficial action ever undertaken to safeguard human health. Prior to

chlorination, waterborne diseases were widely prevalent (e.g. the cholera outbreak in

Chicago in the 1800‟s) and were a major cause of death. Chlorine and its compounds

are the most commonly used disinfectants by water treatment facilities. The

popularity of chlorine is due to the combination of low cost, high oxidizing potential,

and the chlorine residual which exists throughout the distribution system and protects

against microbial recontamination (Sadiq and Rodriguez 2003). However in the

1970s, it was discovered that chlorine used during water treatment reacts with organic

matter, such as humic and fulvic acids, and produces carcinogenic by-products. Given

this, there is merit in reducing the formation of these by-products, as feasible, which

may include post-chlorination of treated water as opposed to pre-chlorination of raw

water, since organic matter levels are much lower in treated water. Nevertheless, there

may exist additional issues which counter the potential to post-chlorinate rather than

pre-chlorinate. One of these issues is zebra mussels (Dreissena polymorpha), a class

of mollusc similar to oysters, clams, and scallops, which originated in the Black and

Caspian seas and have been inadvertently transported into the Great Lakes water by

cargo ships. The mussels grow to 1 inch in length and produce 35,000 eggs per season

per female (Dermott et al. 1993). The proliferation of zebra mussels is causing serious

problems by clogging raw-water intakes and discharge lines, increasing pipe

corrosion, and producing massive bio-fouling. As a result, zebra mussels are a recent

12

problem for surface water treatment facilities in parts of the US and Canada,

particularly along the western shores of Lake Erie. In response, most of these water

treatment facilities chlorinate their source water to control zebra mussels in the intake.

As a consideration, options which would involve timed dosages to control zebra

mussels, and not pre-chlorinating when not needed, would have favourable benefits in

terms of reducing the formation of disinfection by-products. This paper evaluates

several strategies to accomplish lesser formation rates of the disinfection by-products

while still controlling zebra mussels.

2.0 Literature Review and Background

2.1 Known Health Effects of DBPs

Table 1 below outlines the known health effects associated with different classes of

DBPs.

Table 1: The health effects related to the major DBPs (modified from US EPA 1999)


Total Trihalomethanes Chloroform Cancer, liver, kidney and reproductive effects

Dibromochloromethane Nervous system, liver, kidney and reproductive effects

Bromodichloromethane Cancer, liver, kidney and reproductive effects

Bromoform Cancer, nervous system, liver and kidney effects

Haloacetic acids Dichloracetic acid Cancer, reproductive, developmental effects

Trichloracetic acid Liver, kidney, spleen, developmental effects

Inorganic compounds Bromate Cancer

Chlorite Developmental and reproductive effects

13

The current standards for Total Trihalomethanes (TTHMs) and Haloacetic acids

(HAAs) set by Health Canada are 100µg/L and 80µg/L, respectively (Health Canada

2006). By comparison, the US Environmental Protection Agency (USEPA) has set

their standards at 80µg/L and 60µg/L for TTHMs and HAAs respectively (USEPA,

1999).

2.2 Research on chlorination and zebra mussels control

Verween et al. (2009) have shown that chlorination is an effective tool for eliminating

zebra mussels. One strategy to control zebra mussels is “Continuous Treatment”,

where chlorine is applied consistently to the water at concentrations around 2 mg/L at

the intake. This is very effective at killing all biological life in the water. It does,

however, produce the greatest levels of DBPs. It is likely that if Ontario WTPs

employ zebra mussel control, that this is the method that is most commonly used; and

is used year-round.

Alternatively, zebra mussels can be controlled by killing the adult mussels by either

intermittent treatment, or periodic control chlorination (Klerks et al. 1993).

“Intermittent Treatment” includes killing the larvae before they settle and change into

their more-resilient juvenile forms. Sprecher et al. (2000) demonstrate that

chlorinating for 30 minutes every 12 hours is effective in controlling veliger (zebra

mussel larvae) populations and preventing new zebra mussels from settling; however,

this approach is not effective at killing established adult zebra mussels. By

chlorinating heavily for 2-3 weeks in late spring/early summer, the established adults

can be killed. At this point, intermittent treatment can be used to kill the young zebra

mussels looking to “settle”.

14

It has been demonstrated that chlorination rates can be altered by incorporating

knowledge of the changing seasonal tolerances of zebra mussels (Costa et al., 2008).

A promising alternative to continuous treatment is “Periodic Treatment”, which

involves chlorinating “on” and “off” during the year. This treatment usually consists

of three treatment applications spread across the months of April to October. Each

treatment gives a dose of chlorine between 0.5 and 2 mg/L and lasts for 2-4 weeks.

Costa et al. showed that in one experiment, a dose of 0.3 mg/L was applied in 3 sets

of 2-3 week periods, resulting in a 95% mortality rate for zebra mussels.

3.0 Study Area and Approach

3.1 Multivariate Regression

Predictive multivariate regression models for DBPs were developed based on field-

data obtained from across Ontario, from 2000 to 2003. The data are from 25 water

treatment plants (WTPs), and include both raw water and treated water measurements

from these facilities. The models are validated using data from 2004 in the same area

to assess their predictive ability of DBP formation. With a validated model, the

significance of diverse operational and water quality parameters which control or

reduce DBP formation can be identified. For example, treatment plant operators can

lower DBP formation by reducing chlorine dosages, while maintaining sufficient

residuals for disinfection.

3.2 Ontario Drinking Water Surveillance Program (DWSP)

15

The DWSP is a scientifically-based water monitoring program which examines

drinking water quality with a focus on non-regulated drinking water quality

parameters, and possible new contaminants.

3.3 Current TTHM Lifetime Carcinogenic Risk

Over a four year period (2004-2007), 130 WTPs that participated in the DWSP were

analyzed for total THMs as well as four total THM subspecies: Bromoform,

Chlorodibromomethane (CDBM), Bromodichloromethane (BDCM), and Chloroform.

Data from 2004-2007 were taken in lieu of data from 2000-2003 because chlorine

dosage data (which is not readily available after 2004) was not required for this

particular analysis. Only data from surface water sources were employed herein

because zebra mussels are only present in surface source waters. The censored data

present in the dataset which was at, and below, the detection limits, (~15%) was

removed from the dataset, entirely. It is noted that censored data was not removed

from bromoform as it was highly censored (~97%). Shown in Table 2 are the average

concentrations calculated for the THM subspecies and total THMs from the datasets.

Table 2: Average Concentrations of TTHMs in DWSP – Monitored Surface Water Systems in

Ontario

Measured Concentration (µg/L)

Contaminant Surface Water Sources

Bromoform 0.7

Chloroform 25.8

BDCM 4.0

CDBM 2.6

16

TTHMs 33.1

As per Table 2, TTHM levels for surface water treatment plants fall well under the

guideline of 80µg/L, on average. However, about 20% of TTHMs are over 50 µg/L

and about 5% are over 80µg/L. Table 2 shows that chloroform represents the majority

(78%) of the total THMs for surface drinking water sources.

3.4 Derivation of the predictive equation and validation thereof

Using the multivariate regression method, datasets for 2000–2003 were used to derive

the predictive equation. In order to improve the quality of the dataset, data that were

outliers in the dataset were removed; the residual (є) between measured values versus

predicted values was restricted such that TTHMs were set as the predictive variable

and many potential dependant variables were analyzed using multivariate regression.

Through trial-and-error, a number of statistically significant dependent variables were

identified. These variables proved to be: pre-chlorination dose (PreCl2), post-

chlorination dose (PostCl2), treated temperature (Treated Temp), dissolved organic

carbon in raw water (DOCR) and dissolved organic carbon in treated water (DOCT).

A statistically significant non-linear model relating these variables was obtained and

is shown in Equation 1 below.

TTHM = 100.825

*(PreCl2)0.238

*(PostCl2)-0.099

*(TreatedTemp)0.225

*(DOCR)0.362

*(DOCT)0.585

(1)

17

Where n=206 and R2 = 0.60

The negative power value for PostCl2 suggests that post chlorination will lower the

DBPs; however, it is noted that this negative power is derived from a dataset with a

total chlorination such that pre- and post- total attain acceptable levels (approximately

3 mg/L).

Table 3 lists the „p‟ values that indicate the statistical significance of the

corresponding variables. Any variable that exhibited a p value of 0.05 or less was

deemed statistically significant, which indicates that all the noted constituents,

dissolved organic carbon in raw (DOCR) and treated water (DOCT), pre-chlorine

dose (PreCl2), post-chlorine dose (PostCl2), and temperature (Temp) are statistically

significant in this application.

Table 3: The significance of multivariate regression

Figure 1 shows how the predicted TTHM values for the calibration dataset compare

favourably with the measured TTHM values.

Coefficients Standard

Error

t Stat P-value Lower95% Upper

95%

Intercept 0.824 0.0512 16.1 5.31E-38 0.724 0.925

PreCl2 0.238 0.0283 8.41 7.68E-15 0.182 0.294

PostCl2 9.89E-02 0.0319 -3.10 2.21E-03 -0.162 -3.60E-02

Temperature 0.225 0.0235 9.57 4.29E-18 0.179 0.271

DOC (Raw) 0.362 0.0496 7.29 7.16E-12 0.264 0.460

DOC(Treated) 0.585 0.0755 7.75 4.67E-13 0.436 0.733

18

Figure 1: Measured TTHM vs. Predicted TTHM for Calibration Dataset

The dataset from 2004 was used to validate the predictive equation. Figure 2 shows

how the predicted TTHM values for 2004 compare favourably with the measured

TTHM values.

Figure 2: Measured TTHM vs. Predicted TTHM for 2004

y = 1.0542x R² = 0.6078

0.00

20.00

40.00

60.00

80.00

100.00

120.00

140.00

0.00 20.00 40.00 60.00 80.00 100.00

Mea

sure

d T

THM

µg/L

Predicted TTHM µg/L

Measured TTHM vs. Predicted TTHM for Calibration Dataset

y = 1.0189x R² = 0.7507

0

10

20

30

40

50

60

70

80

90

100

0 20 40 60 80 100

Mea

sure

d T

THM

µg/L


Measured TTHM vs. Predicted TTHM for 2004

19

From Figure 2, the predicted TTHM values for 2004 using Equation 1 were graphed

against the measured TTHM values, and the R-Squared value was calculated from the

line of best fit, 75%, validating the regression model.

3.5 The seasonal impact on zebra mussel control

To effectively control zebra mussels using chlorination, the seasonal variation of

zebra mussel susceptibility to chlorination patterns must be considered. According to

experiments by Bryant et al. (1985), Heinonen (2001), Rajagopal et al. (2002a), and

Costa et al. (2008), the high susceptibility of zebra mussels to chlorine is observed in

the U.S. to some extent in June and peaks in July and August, after reproduction

because of their low body weight, high filtration activity and the high water

temperatures. However, given Ontario‟s very long, cold winters and short summers,

the peaks of susceptibility move to between August and September as evident in

Figure 3 which shows the yearly profile of TTHMs and subspecies, where it is evident

that TTHMs peak between August and September. Hence, when zebra mussels are

most susceptible to chlorine, and thus August and September represent the optimal

period to pre-chlorinate to reduce zebra mussel levels.

As demonstrated by the regression equation, as well as historical records, TTHM

concentrations fluctuate as temperature changes. In other portions of the year,

however, zebra mussels are not as susceptible to chlorine and pre-chlorination may be

lowered as a means of reducing TTHM formation.

20

Figure 3: Seasonality of TTHM formation

As seen in Figure 3, there is a strong correlation between increasing temperature and

increasing chloroform and TTHM concentrations; the correlation is not as strong

between the other trihalomethane subspecies and temperature. Temperature has an

effect on TTHM formation, because it increases the reaction kinetics. As well, it is

noted that the raw DOC levels remain relatively constant indicating that raw DOC is

not responsible for peaks in TTHMs.

0

4

8

12

16

20

0

5

10

15

20

25

30

35

40

Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

Tem

pe

ratu

re (°

C)

Co

nce

ntr

atio

n (µ

g/L)

Month

Seasonality of TTHM Formation for All Sites 2004-2007

Bromodichloromethane Chlorodibromomethane

TTHM Chloroform

Raw DOC Temperature

21

Given the above evidence, consideration of both zebra mussel control and treatment

strategies is feasible.

Consider as an example, Kenora‟s water treatment facility which currently operates

with one or two pre-chlorine applications in the fall (August and September), when

the temperature is high (above 20oC). However, this causes DBP formation due to the

high level of resulting mussel fouling increasing the organic matter in the raw water.

Table 4 shows some selected water quality parameters before the control and

treatment strategies were applied to Kenora‟s water treatment plant. As seen in Table

4, both HAAs and TTHM levels are over the MCL, 80µg/L, and 100µg/L respectively

for individual months at Kenora. Consequently, placing an emphasis on pre-

chlorination during the spawning season, and shifting the emphasis to post-

chlorination when not in the spawning season represents an opportunity to reduce

DBP formation. Table 5 shows the same selected water quality parameters after the

control and treatment strategies have been undertaken.

Table 4: Datasets from Kenora water treatment plant before treatment strategies applied

Date Flow Pre

Cl2(mg/L)

Post Cl2

(mg/L)

TEMP.

(oC)

HAAs

(µg/L)

TTHM

(µg/L)

Feb-2000 11.5 1.75 1.16 2.9 87.8 57.50

May-2000 11 4.46 1.25 10.3 - 99

Aug-2000 11.2 3.26 2.18 22.5 78.9 126

Nov-2000 8.8 3.61 1.55 5 112 79.5

Feb-2001 11 2.34 1.53 1.9 79.9 57.5

Jun-2001 9.9 4.63 1.8 13.1 - 112

Sept-2001 11.8 4.7 2.21 21.2 94.5 138

Dec-2001 9.2 3.4 2.57 2 198 84.5

22

May-2002 10 2.85 2.64 5.8 - 96

Jun-2002 10.5 3.54 2.62 10.7 - 104

Sept-2002 11.2 5.29 2.97 20.8 163 135

Table 5: Datasets from Kenora water treatment plant after treatment strategies applied

Date Flow Pre

Cl2(mg/L)

Post Cl2

(mg/L)

TEMP.

(oC)

HAAs

(µg/L)

TTHM

(µg/L)

Feb-2006 1.35 2.25 0.9 - 43.5

Jul-2006 2 2.9 24.2 48.1 78

Oct-2006 1.8 2.6 6.3 34 49

Feb-2007 1.3 2.2 2.7 32.2 38.5

May-2007 1.7 2.2 15 - 47

Jul-2007 2.1 2.6 23.2 25.1 75

Oct-2007 2 2.4 10 44.9 59

May-2008 1.5 2.1 5.4 23.1 44.5

Aug-2008 1.6 2.6 21.6 41.7 56

Nov-2008 1.6 2.2 10.3 40.9 49.5

The Brockville water treatment plant and the Toronto R.C. Harris water treatment

plant in Ontario rely on an alternative process to Kenora‟s, for eliminating zebra

mussels. As soon as the temperature exceeds 10oC, pre-chlorination is initiated at a

concentration of 0.5mg/L to 0.8mg/L. This usually starts in the spring and ends in the

fall when the water temperature drops below 10oC. Based on practical experience, this

has been an effective method of eliminating zebra mussels (Richards, 2010).

4.0 Integrated Analysis

23

4.1 Integrated Risk Information System (IRIS)

The USEPA has created the Integrated Risk Information System (IRIS), which is

designed to assess the risks of different chemicals to human health. For carcinogenic

compounds they have established a cancer slope value, which has units of

(mg/kg/day)-1

. They have identified cancer slope values for four trihalomethane

subspecies. The TTHMs identified are BDCM, CDBM, Chloroform and Bromoform,

as listed in Table 5. The cancer slope factors for each subspecies were normalized

using the percentage of total subspecies that each individual subspecies forms on

average, and then these values were summed to calculate their weighted average.

Table 6: Slope factor for each subspecies and average slope factor for TTHM

(1)Cancer Slope Factors taken from IRIS Database

In this research, statistically significant multivariate regression models for the

carcinogenic HAAs and total THMs subspecies were also developed, and are shown

in Equations 2 through 5, below. Other HAA and THM subspecies were not added

due to current lack of measured data.

Compound

Average

Concentration

(µg/L)

% Slope Factor

(1)(mg/kg/day)-1

Weighted Slope

Factor

(mg/kg/day)-1

Average Slope

Factor for

TTHM

(mg/kg/day)-1

Bromodichloromethane 4.0 12 0.062 0.0074

Chlorodibromomethane 2.6 8 0.084 0.0067

Chloroform 25.8 78 0.006 0.0047

Bromoform 0.7 2 0.079 0.0016

Sum 33.1 0.0204

24

HAA

Dichloroacetic acid (DCAA) (HAA):

DCAA = 102.34

* (PreCl2)0.09

*(Raw pH)-2.18

*(DOCR)0.79

*(DOCT)0.46

(2)

N = 329 R2 = 0.61

THM Subspecies

Bromodichloromethane (BDCM):

BDCM = 100.36

*(Treated Temp)0.21

*(DOCR)-0.59

*(DOCT)0.78

(3)

N = 323, R2 = 0.14

Chlorodibromomethane (CDBM):

CDBM = 100.35

*(Treated Temp)0.15

*(DOCR)-0.91

(4)

N = 237 R2 = 0.38

Chloroform:

Chloroform = 101.88

* (PreCl2)0.20

*(PostCl2)0.10

*(Raw pH)-1.88

*

(Treated Temp)0.22

*(DOCR)0.89

*(DOCT)0.70

(5)

N= 326 R2 = 0.76

Where,

PreCl2 = Pre-Chlorine Dose,

PostCl2= Post-Chlorine Dose,

25

Raw pH = pH of Raw Water,

DOCR = Dissolved Organic Carbon in Raw Water,

DOCT = Dissolved Organic Carbon in Treated Water and,

Treated Temp = Temperature of Treated Water.

Once the concentrations of the subspecies were determined by averaging each

subspecies‟ mean monthly concentrations, the cancer risk was calculated using the

cancer slope factor.

Exposure to the compounds, given by the Lifetime Average Daily Dose (LADD) is

defined as:

LADD = (Cw x IR x EFx ED)/(BW x AT)

Where,

Cw = Contaminant Concentration,

IR = Intake Rate,

EF = Exposure Frequency,

ED = Exposure Duration,

BW = Body Weight and,

AT = Averaging Time.

26

Assuming a body weight of 70kg, an intake rate of 2L of water a day (Gleick 1996)

for 365 days per year, a lifespan of 70 years (ED), and an average lifetime of 25,550

days; the consumption of water per mass per day (L/(kg*day)) is:

LADD = Cw x consumption of water per mass per day.

The consumption of water per mass per day is:

0.029L/(kg*day).

The corresponding incremental excess lifetime cancer risk, IECLR, for the

trihalomethane subspecies concentrations listed in Table 2 as shown in Table 7 (Note

there is no cancer slope factor available for TTHMs, and hence the IECLR of TTHMs

is calculated from the sum of the major trihalomethanes).

Table 7: Cancer risk level of surface water drinking sources of TTHMs in Ontario WTPs

Current Situation for Chlorination

Contaminant Excess Cancer Risk (10-6

)

Bromoform 1.6

Chloroform 4.5

BDCM 7.2

CDBM 6.3

Sum of Subspecies 19.6

Table 7 demonstrates that all subspecies and TTHMs currently exist in concentrations

that result in cancer risk levels greater than de minimus risk (one in a million). This is

27

problematic because although the predicted TTHM levels are under current

guidelines, they still yield cancer risks higher than one in a million. To estimate an

approximate TTHM level that would yield a cancer risk less than de minimus risk, a

cancer slope of 0.029 (see Table 2) for TTHMs and for a water consumption of 2L per

day, a TTHM concentration of 1.7µg/L is determined. This finding indicates that

TTHM concentrations would have to be reduced to this level to attain a cancer risk at

de minimus risk.

4.2 Comparison of varying pre- and post-chlorination levels versus the status

quo

Consider six scenarios with varying levels of pre- and post-chlorination in an effort to

mimic alternative approaches to chlorination while also incorporating mussel control.

The results are compared to control values as shown in Table 9. The control value for

predicted TTHMs was found by calculating the average value for each variable in

Equation 1, for each month of the year. The variables being examined were: pre-

chlorination, post-chlorination, raw pH, treated water temperature, raw DOC, and

treated DOC. Using the regression equations, namely Equations 1 to 5, values of

Dichloroacetic acid (DCAA), Bromodichloromethane, Chlorodibromomethane,

Chloroform and TTHM were calculated. The cancer risk was then calculated for each

species for each month of the year. A mean value for each species was also found for

the whole year. The six scenarios are as follows:

28

Scenario 1: pre-chlorination was regular (the same as the control) but post-

chlorination was set to be 0.001mg/L (0.001mg/L had to be used to preserve the

integrity of the model).

Scenario 2: During the months of May, July and September, pre-chlorination was set

to 0.5mg/L and the rest of the year it was set to 0.001mg/L, and post chlorination was

set to 3mg/L throughout the year. This test follows “Intermittent Treatment”

(Sprecher, 2000), where there are three treatments a year of pre-chlorination.

Scenario 3: This was very similar to scenario 2 but adopted a more aggressive

approach to pre-chlorination. Scenario 3 assumed pre-chlorination at a concentration

of 0.75mg/L during the months of April, June, August and October, and during the

rest of the year the pre-chlorination was set to 0.001mg/L. Post-chlorination was set to

3mg/L.

Scenario 4: This test was similar to scenarios 2 and 3, but instead utilized a process

known as “End-of-season treatment” (after Sprecher, 2000). Pre-chlorination was

assumed at a concentration of 0.75mg/L during the months of May and July, and at a

concentration of 3.0mg/L in September. The purpose of the higher dose in September

was to flush out any remaining adult zebra mussels prior to the next season beginning.

Again, the post-chlorination was 3mg/L.

29

Scenario 5: This scenario is quite different from the preceding scenarios. In this

scenario, “Continuous Treatment” was used during the months where the water

temperature exceeded 12oC. During the months of May to November the dose of pre-

chlorine was set to 3mg/L, and set to 0.001mg/L for the remainder of the year. During

the months of May to November the dose of post-chlorine was set to 0.3mg/L, and set

to 3.3mg/L for the remainder of the year. This scenario was designed to be the most

secure test in terms of microbial disinfection and zebra mussel control.

Scenario 6: This test was set up to replicate operation of the City of Brockville‟s

water treatment plant similar to scenario 5 but with lower concentrations of chlorine.

During the months of May to November the pre-chlorination was set to 0.5mg/L, and

set to 0.001mg/L for the remainder of the year. Throughout the year the post-

chlorination was set to 1.25mg/L. This approach has proven to be effective in

Brockville.

Below are three tables which summarize the results of the six chlorination

approaches. Table 8 summarizes the DCAA and THM subspecies concentrations

calculated from Equations 2 through 5, and the corresponding IECLRs for each of the

scenarios.

30

Table 8: Average Carcinogenic Risks from HAA and TTHM subspecies

Dichloroacetic Acid (DCAA)

Bromodichloromethane

(BDCM)

Chlorodibromomethane

(CDBM) Chloroform

Concentration

(µg/L) IECLR

Concentration

(µg/L) IECLR

Concentration

(µg/L) IECLR

Concentration

(µg/L) IECLR

Original 12 5.22E-06 4 7.19E-06 1 2.44E-06 23 4.00E-06

Scenario 1 12 5.22E-06 4 7.19E-06 1 2.44E-06 10 1.74E-06

Scenario 2 8 3.48E-06 4 7.19E-06 1 2.44E-06 8 1.39E-06

Scenario 3 8 3.48E-06 4 7.19E-06 1 2.44E-06 10 1.74E-06

Scenario 4 8 3.48E-06 4 7.19E-06 1 2.44E-06 9 1.57E-06

Scenario 5 10 4.35E-06 4 7.19E-06 1 2.44E-06 12 2.09E-06

Scenario 6 9 3.92E-06 4 7.19E-06 1 2.44E-06 10 1.74E-06

As seen in Table 8, scenarios 2 and 4 are most effective at reducing the subspecies

concentrations as well as the associated cancer risk. It is noteworthy that BDCM and

CDBM are not affected by chlorination dosages. Table 9 displays the known,

measured concentrations of TTHMs for each scenario, and compares these values to

TTHM concentrations calculated from Equation 1, as well as the sum of subspecies

calculated from summing the THM subspecies concentrations from Table 8. The

corresponding IECLR has also been calculated for each concentration value.

31

Table 9: Average Carcinogenic Risks from TTHM and Sum of Trihalomethane Subspecies

Measured

TTHM (µg/L) IECLR

Eq. 1 Predicted TTHM

(µg/L) IECLR

Sum of THM Subspecies

Excluding Bromoform

(Table 7) (µg/L)

IECLR

Original 32 1.89E-05 34 2.01E-05 28 1.36E-05

Scenario 1 32 1.89E-05 75 4.44E-05 15 1.14E-05

Scenario 2 13 7.69E-06 10 5.92E-06 13 1.10E-05

Scenario 3 15 8.87E-06 13 7.69E-05 15 1.14E-05

Scenario 4 15 8.87E-06 12 7.10E-06 14 1.12E-05

Scenario 5 23 1.36E-05 26 1.54E-05 17 1.17E-05

Scenario 6 19 1.12E-05 34 2.01E-05 15 1.14E-05

As seen in Table 9, Scenarios 2 and 4 once again yield the lowest TTHM levels as

well as the lowest corresponding cancer risks. Table 10 displays the percentage

reduction in TTHM concentrations for the TTHM concentrations predicted using

Equation 1, and for the sum of TTHM subspecies calculated from Table 8. The

relative cancer risk for each scenario is also displayed, and has been calculated by

dividing IECLR of each scenario by the IECLR from the original control values.

32

Table 10: Reduction of TTHM Cancer Risk, Relative to Current Scenario

Scenario

Relative Cancer Risk for

Predicted TTHM

Relative Cancer Risk for TTHM

(Sum of Subspecies)

Reduction(2)

% for

Predicted TTHM

Reduction % for TTHM

(Sum of Subspecies)

Original 1 1 0% 0%

Scenario 1 2.21 0.84 -121% 46%

Scenario 2 0.29 0.81 71% 54%

Scenario 3 0.38 0.84 62% 46%

Scenario 4 0.35 0.82 65% 50%

Scenario 5 0.76 0.86 24% 39%

Scenario 6 1 0.84 0% 46%

(2) Relative to Current Water Treatment Scenario

According to the results summarized in Table 10, scenario 2 would cause the greatest

improvement in TTHM formation (71% using Equation 1, and 54% from sum of

subspecies, see line 4 of Table 10). There is concern that although Test 2 is the most

effective in reducing TTHM levels, it may be too lenient in terms of chlorination and

could possibly allow zebra mussel infestation. Stricter, more thorough scenarios such

as 5 and 6 might be more appropriate for zebra mussel removal, while still resulting in

lower levels of TTHMs than current methods.

5.0 Conclusions

1) Regression models for the formation of TTHMs and its subspecies as determined

from 2000 to 2003 data were validated using data from 2004, and shown to be

33

robust. TTHM models can also be used to estimate how TTHM levels could be

reduced, while still continue to provide adequate zebra mussel removal.

2) The regression equation that was developed to predict TTHM concentrations

results in the pre-chlorination variable having a higher exponent than the post-

chlorination variable, indicating that TTHM concentrations are more heavily

influenced by pre-chlorination dosages which indicate reducing the pre-

chlorination dose when feasible, and making up for this reduction by increasing

the post-chlorination dose has merit. This can be accomplished in winter months

as zebra mussels are considerably less active during this time.

3) Cancer risks from TTHMs and its subspecies with current disinfection practices

exceed one in a million. The incremental cancer risk ranged from 1 in 50,000 to

100,000. However, lowering the incremental cancer risk from to 1 in 1,000,000,

would require the average concentration of TTHMs to be 1.7µg/L to reach 1 in

1,000,000.

4) There is a strong correlation between increasing temperature and increasing

chloroform and TTHM concentrations. There is not as strong a correlation

between the other TTHM subspecies and temperature.

5) A strategy for effectively controlling zebra mussels is to use pre-chlorination at

3mg/L when the temperature exceeds 12oC, and a year-round post-chlorination.

This scenario results in a reduction of TTHM formation of 24% and reduces

34

cancer risk by 24%, while maintaining effective disinfection and zebra mussel

removal.

6.0 Acknowledgements

The authors would like to thank the DWSP of Ontario for their contributions in the

data collection phase. This research was funded by the Canada Research Chair

Program.

7.0 References

Berthouex, P. Mac. Statistics for environmental engineers. Boca Raton: Lewis, 1994.

Print.

Costa, A., Aldridge, D. C., Moggridge, G, D. “Seasonal variation of zebra mussel

susceptibility to molluscicidalagnets”, Journal of Applied Ecology, 2008, 45,

1712-1721.

Dermott, R., and Munawar, M., "Invasion of Lake Erie Offshore Sediments by

Dreissena, and Its Ecological Implications." Department of Fisheries and

Oceans, Canada 50(1993): pp 2298-2304.

"Drinking Water Surveillance Program Data Report 2003 and 2004." 21 October

2008. DWSP. 6 Aug 2009

<http://www.ene.gov.on.ca/envision/water/dwsp/0304/>.

Fisher, S. W., Dabrowska, H., Waller, D. L., Babcock-Jackson, L., and Zhang, X.

(1994). “Sensitivity of Zebra Mussel (DreissenaPolymorpha) Life Stages to

35

Candidate Molluscides”. Journal of Shellfish Research, Vol. 13, No. 2, 373-

377.


"Sensitivity of Zebra Mussel (DreissenaPolymorpha) Life Stages to Candidate

Molluscicides." Department of Entomology. 13(1994): pp 373-377.

Gleick, P.H., Basic Water Requirements for Human Activities: Meeting Basic Needs.

International, 1996. 21: p. 83-92.

"Guidelines for Canadian Drinking Water Quality - Summary Table." 30 May 2008.

Health Canada. 6 Aug 2009 <http://www.hc-sc.gc.ca/ewh-semt/pubs/water-

eau/sum_guide-res_recom/index-eng.php>.

"Guidelines for Canadian Drinking Water Quality: Guideline Technical Document:

Trihalomethanes." 01 May 2006. Health Canada. 6 Aug 2009 <http://www.hc-

sc.gc.ca/ewh-semt/pubs/water-eau/trihalomethanes/index-eng.php>.

Hamilton, D. J., Ankney, C. D., and Bailey, R. C. 1994. “Predation of Zebra Mussels

by Diving Ducks: An Exclosure Study”. The Ecological Society of America.

Pp 521-531.

Kilgour, B.W., and Mackie., G. L., Colonization of Different Construction Materials

by the Zebra Mussel. Lewis Publishers, 1993.

McBean, E., Zhu, Z., and Zeng, W., "Systems Analysis Models for Disinfection By-

Product Formation in Chlorinated Drinking Water in Ontario." Civil

Engineering and Environmental Systems 25(2008): pp 127-138.

Metcalf, Eddy. Wastewater Engineering. New York: McGraw-Hill College, 2002.

Print.

36

"National Primary Drinking Water Regulations:" 16 December 1998. US EPA. 6 Aug

2009 <http://www.epa.gov/OGWDW/mdbp/dbpfr.html>.

"National Survey of Chlorinated Disinfection By-Products in Canadian Drinking

Water." 12 September 2008. Health Canada. 7 Aug 2009 <http://www.hc-

sc.gc.ca/ewh-semt/pubs/water-eau/byproducts-sousproduits/index-eng.php>.

On, Conference. Water chlorination environmental impact and health effects:

proceedings of the Conference on the Environmental Impact of Water

Chlorination, Oak Ridge National Laboratory, Oak Ridge, Tennessee, October

22-24, 1975. Ann Arbor, Mich: Ann Arbor Science, 1978. Print.

Rajogopal, S., Van der Velde, G., Van der Gaag, M., Jenner, H. A., “How Effective is

Intermittent Chlorination to Control Adult Mussel Fouling in Cooling Water

Systems?”. Water Res.(2003) 37, 329-338.

Ram, J. L., Fong, P. P and Garton, D. W., "Physiological Aspects of Zebra Mussel

Preproduction: Maturation, Spawning, and Fertilization." American Zoologist

36(1996): pp 326-338.

Richards, D. Chief Operator, Brockville Water Treatment. Phone: (613) 342-7819

Sprecher, S., and Getsinger, K. D., "Zebra Mussel Chemical Control Guide." US

Army Corps of Engineers - Environmental Laboratory (2000): 1-116.

Verween ., , Vincx , M., Degraer, S., “Comparative toxicity of chlorine and peracetic

acid in the biofouling control of Mytilopsisleucophaeata and

Dreissenapolymorpha embryos (Mollusca, Bivalvia)”, International

Biodeterioration& Biodegradation, 63 (2009) 523–528.

37

Wang, G-S, Deng, and Lin. "Risk Assessment from Trihalomethanes in Drinking

Water." Science of the Total Environment 387(2007): 86-95.

38

Chapter 3

Paper #2: Mitigation of Disinfection By-Product Formation by

the Development of a Regression Equation with the Bromide Ion Brett Harper, Zoe J. Y. Zhu and E. McBean

Abstract

The contribution of bromide to the subspecies of total trihalomethanes (TTHM)

demonstrates that bromide compounds approach fifty percent of TTHM as bromide

reaches 10 µg/L. This is problematic since cancer slope factors for bromide

compounds are ten-fold higher than chloroform, meaning the potential cancer rates

when bromide is present, are greatly increased.

A multivariate regression model for TTHM (R2 = 0.91) which includes bromide, is

described and demonstrates that water temperature, dissolved organic carbon, pH,

pre-chlorination and raw water bromide levels are statistically significant variables for

prediction of TTHM levels.

Keywords: Disinfection By-Products (DBPs), Trihalomethanes (THMs), Bromide,

Organic matter, Chlorine, Multivariate regression

39

1.0 Introduction

1.1 An Overview of Disinfection By-Products in Drinking Water

Total Trihalomethanes (TTHM) are by-products created when the chlorine used in the

disinfection process reacts with naturally occurring organics (e.g. organics formed by

the decay of algae and vegetation) in raw water. The most common forms of

trihalomethanes created are chloroform, bromodichloromethane (BDCM),

chlorodibromomethane (CDBM) and bromoform and hence the bromide ion is

obviously a contributing factor to the formation of TTHM. While chlorine is added

for water disinfection, the chlorine also causes the formation of TTHM and these by-

products are carcinogens. To decrease the health risk, the Ontario government (and

many others) is considering lowering its guideline from a maximum acceptable

concentration of 100 µg/L to a maximum acceptable concentration of 80 µg/L. Of

interest is the degree to which the bromide ion affects the speciation of the TTHM.

1.2 Related Literature on TTHM formation

McBean et al. (2008) and Harper et al. (2012) have developed regression equations to

model TTHM concentrations in drinking water. However, these papers did not

include bromide in the regression equation because bromide data over a four year

period (2005-2008) have only recently been released by the Ontario-based Drinking

Water Surveillance Program (DWSP). The DWSP is a scientifically-based water

monitoring program which examines drinking water quality with a focus on non-

regulated drinking water quality parameters, and possible new contaminants.

40



Table 1 outlines the known health effects associated with different classes of DBPs.

Table 1: The health effects related to the major DBPs (modified from USEPA 1998)


Total Trihalomethanes Chloroform Cancer, liver, kidney and reproductive effects



Bromodichloromethane Cancer, liver, kidney and reproductive

effects

Bromoform Cancer, nervous system, liver and kidney

effects

The current standard for TTHM set by Health Canada is 100 µg/L (Health Canada,


their standards at 80 µg/L for TTHM (USEPA, 1998).

2.2 Trihalomethane Subspecies vs. TTHM

The DWSP reports TTHM as the sum of trihalomethane subspecies (bromoform,

bromodichloromethane, chlorodibromomethane, and chloroform). Figure 1

demonstrates the average composition of the TTHM subspecies concentrations from

2004-2007. These results were obtained by averaging over all WTP data records for

which bromide data were available, as well as all THM subspecies (119 records

include bromide in raw water, and 123 records include bromide concentrations in

treated water), and calculating the average concentration of each of the THMs.

41

2.3 Bromide Ion


formation of TTHM (Health Canada, 1995). Until recently, however, only very

limited bromide data have been available; of interest is the structure of regression

models in response to the addition of the bromide ion.

The TTHM regression model reported in Harper et al. without bromide data, for water

as released to the distribution system.

TTHM =100.825

* (PreCl2)0.238

* (PostCl2)-0.099

* (Temp)0.225

* (DOCR)0.362

*

(DOCT)0.585

(1)

Where n=206 R2 = 0.75

The negative power value for PostCl2 suggests that post chlorination will lower the

DBPs; however, it is noted that this negative power was derived from a dataset with a

total chlorination such that pre- and post- total attain specified levels (approximately 3

mg/L).

Figure 2 illustrates how alternative disinfection by-product subspecies vary in percent

speciation given different bromide concentrations. The data used by Health Canada

was taken from 52 drinking water treatment facilities across Canada, on two

occasions in 1993. Looking specifically at TTHM, when the Bromide ion is < 0.01

42

mg/L, the chloroform concentration is nearly 100% of the TTHM formed (see Figure

2). Chloroform has the lowest cancer slope (0.0061(mg/kg/day)-1

) of the TTHM

subspecies (USEPA, 1998). When the bromide ion reaches 0.5 mg/L the subspecies

form in the following percentages: chloroform 20%, bromodichloromethane 20%,

chlorodibromomethane 25%, bromoform 35%. Considering how the three bromo-

subspecies have cancer slopes that are at least 10-fold greater than chloroform, the

implication is a significant increase in the overall cancer risk when bromide

concentrations are high. Table 2 indicates the cancer slope factor of each THM

subspecies (from USEPA, 1998).

Table 2: Slope factor for each TTHM subspecies

Data from all Ontario WTPs (which recorded bromide data at any point from 2004 –

2007), given bromide concentrations in raw drinking water, are used to illustrate how

TTHM speciation favours chloroform as bromide levels decrease, and are shown in

Figure 3. Concentration data were used for Ontario WTPs (21 surface, 4 ground, 2

mixed) for the years of 2004 – 2007 for bromide, as well as all TTHM subspecies.

The TTHM subspecies concentrations were summed, and the percent speciation of

each TTHM subspecies was calculated by dividing each individual subspecies by the

Compound

Slope Factor

(mg/kg/day)-1

Chloroform 0.006

Chlorodibromomethane 0.084

Bromoform 0.079

Bromodichloromethane 0.062

43

sum. The percent speciation was then graphed against the corresponding bromide

concentration. As seen in Figure 3, when bromide levels are at <5µg/L, chloroform

concentrations exceed 70% of TTHM formed, whereas when bromide levels double to

>10µg/L, the percentage of chloroform drops to approximately 50%. Figure 3

demonstrates that as bromide levels increase, the brominated THM subspecies

(CDBM, BDCM and bromoform) all increase in percent speciation. These findings

differ from those of Health Canada because in the Health Canada study the TTHM

subspecies concentrations were taken from the distribution system, as opposed to the

treated water.

Figure 4 demonstrates how cancer risk is affected when raw water bromide

concentrations increase. For TTHM concentration of 20µg/L, for purposes of

estimation in Figure 4, the percent speciation of TTHM is multiplied by this

concentration to obtain an estimated concentration for each THM subspecies. From

these concentrations, the exposure to each of these compounds, given by the Lifetime

Average Daily Dose (LADD), can be calculated as:


Where,


IR = Intake Rate,


44




Assuming a body weight of 70kg, an intake rate of 2L of water a day (Gleick, 1996)





0.029L/(kg*day).

The corresponding incremental excess lifetime cancer risk, IECLR, for the

trihalomethane subspecies concentrations is calculated by multiplying the LADD by

the corresponding CSF of each trihalomethane subspecies given in Table 2. The

resulting cancer risks for all trihalomethane subspecies, as well as TTHM, are shown

in Figure 4. As Figure 4 demonstrates, cancer risk roughly doubles when raw water

bromide concentrations are greater than 10µg/L, and increases by 35% when treated

water bromide concentrations are greater than 10µg/L.

When bromine is low, chloroform predominates TTHM, but when bromine is high,

there is both higher bromine-related TTHM and higher TTHM, as demonstrated in

45

Figure 5. This indicates that bromine acts as an additive effect. In addition, the

brominated subspecies have been shown to have significantly higher health effects

compared to non-brominated subspecies. A number of researchers (Morrow, 1987;

Chang, 2001; Kampioti, 2002; Uyak, 2007; Wang, 2007; and Sun, 2009) have

reported that speciation shifted to the bromine-substituted THMs as a function of

bromide concentration when all other parameters were held constant. Under

conditions of high natural organic matter (NOM) and low bromide concentrations,

chlorine-substituted by-products predominated, especially during longer reaction

times. In the presence of chlorine and organic material, as much as 50% of the

bromide ion may become incorporated into the brominated trihalomethane subspecies

(Chang et al., 2001); this efficiency of bromide incorporation implies that 100 µg/L of

bromide may result in up to 50 µg/L of THM-bound bromine (THM-Br). A reduction

in bromide concentration will have a significant impact on the concentration and

speciation of formed TTHM.


3.1 Water Treatment Plant MCL Compliance

Over a four year period (2005-2008), 130 WTPs that participated in the DWSP were

analyzed to identify frequency for various levels of TTHM. Shown in Table 3 are the

number of records and WTPs that exceeded various levels on any one instance for

monitoring. Four different specified levels were used for this comparison: 100µg/L,

80µg/L, 60µg/L and 1.2µg/L. Regarding 1.2µg/L, Harper et al. (2012) refer to the

concentration level that TTHM would have to attain for the cancer risk to be at „de

minimus‟ (one in a million) cancer risk.

46

Table 3: Non Compliance for TTHM at various specified levels in all WTPs which report TTHM

TTHM 2005 % Pass 2006 % Pass 2007 % Pass 2008 % Pass

Number of Data Records 369

338

339

321

Number of Records Failing Specified Level of 100 µg/L 3 99.19% 2 99.41% 1 99.71% 1 99.69%



Number of Records Failing Specified Level of 1.2 µg/L 352 4.61% 318 5.92% 316 6.78% 293 8.72%

Number of Plants Reporting TTHM 122

119

116

107

Number of Plants Failing Specified Level of 100ug/L 2 98.36% 2 98.32% 1 99.14% 1 99.07%



Number of Plants Failing Specified Level of 1.2ug/L 121 0.82% 117 1.68% 115 0.86% 102 4.67%

As apparent in Table 3, TTHM for water treatment plants fall under the specified

level of 100µg/L in nearly every instance. The same holds true for a specified level of

80µg/L. However, if the guideline were to be lowered to 60µg/L about 10% of

records would violate the specified level, and roughly 15% of all WTPs fall into non-

compliance. This number is worsened when a specified level of 1.2µg/L is used. This

demonstrates that lowering the specified level for TTHM from 100µg/L to 80µg/L

would not create large issues, as well as demonstrating that further reducing the MCL

to 60µg/L would not appear likely to cause huge increases exceedance frequency.

47

Notably, there exists a WTP with very high average TTHM (104µg/L, corresponding

cancer risk of 6.15 x 10-5

). WTPs with very high average TTHM will likely see the

greatest TTHM mitigation, and thus the greatest reduction in cancer risk.

3.2 Multiple Regression Analysis

A multiple regression model was developed to estimate the various disinfection by-

product concentrations in chlorinated drinking waters, using data from six WTPs (five

from surface water sources, one from mixed surface water and groundwater sources)

in Ontario during 2005 to 2008. Only for six treatment plants were all chlorination

dosage data and bromide data available and hence used in the regression analysis.

3.3 Derivation of the predictive equation and validation thereof

Using the multivariate regression method, 2005 – 2008 data were used to derive a

predictive equation for TTHM released to the distribution system. The backward

elimination multivariate regression method (for log-transformed variables) was

employed as follows:

1. Run Microsoft Excel‟s regression analysis, testing all potential variables in the

equation.

2. Check to see if all variable p-values are below 0.05, and if not, remove the

variable with the highest p-value.

48

3. Re-run the regression analysis with all remaining variables.

4. Continue this procedure, removing the variable with the highest p-value and

re-running the regression model, until all remaining variables have a p-value

less than 0.05, at which point, all remaining variables are statistically

significant.

Through trial and error, a number of statistically significant independent variables

were identified. These variables were: dissolved organic carbon in raw water in mg/L

(DOCR), pre-chlorine dosage in mg/L (PreCl2), bromide in raw water in µg/L

(RawBr), temperature of raw water in °C (Raw Temp), and pH of treated water being

released to the distribution system (Treated pH). A statistically significant non-linear

model relating these variables was obtained as per Equation 2.

TTHM = 101.62

* (DOCR)0.663

x (PreCl2)0.653

x (RawBr)0.282

x (Raw Temp)0.173

x

(Treated pH)-1.40

Where n = 26 (2)

Table 4 lists the „p values‟ that indicate the statistical significance of the

corresponding variables. This indicates that all the constituents, intercept, dissolved

organic carbon in raw water (DOCR), pre-chlorine dosage (PreCl2), bromide in raw

water (RawBr), temperature of raw water (Raw Temp), and pH of treated water prior

to being released to the distribution system (Treated pH) are statistically significant in

this application. It should be noted that when the acceptable significance was

49

increased from 0.05 to 0.1, that no new variables entered the equation. Post-

chlorination was tested for statistical significance in place of pre-chlorination and was

determined not to be statistically significant (p=0.88). This implies that if pre-

chlorination were lowered to decrease TTHM, and post-chlorination would thus have

to be increased, the increase in post-chlorination should not affect TTHM levels.

Table 4: The significance of multivariate regression for Equation 2

With a validated model, the significance of alternate operational and water quality

parameters which control or reduce DBP formation can be identified. For example,

treatment plant operators can lower DBP formation by reducing pre-chlorination

dosages, while maintaining sufficient residual for adequate disinfection.

From Figure 6, the predicted TTHM values using Equation 2 were graphed against the

measured TTHM values, and the R-Squared value was calculated from the line of best

fit, 91%, validating the regression model. Figure 7 demonstrates how the predictive

capabilities of Equation 1 and Equation 2 compare to each other. As apparent from

Figure 7, Equation 2 demonstrates improved accuracy over Equation 1. To

demonstrate how altering water quality parameters may affect cancer risk, Figure 8

shows how altering pre-chlorination levels can reduce cancer risk. A reduction of

10% can lower cancer risk by roughly 6.5%, a reduction of 20% lowers cancer risk by

Coefficients Standard

Error

t Stat P-value Lower95% Upper

95%

Intercept 1.62 0.635221 2.549665 0.01909 0.294553 2.944649

DOC Raw 0.663 0.078765 8.411051 5.33E-08 0.498198 0.826802

Pre

Chlorine

0.653 0.224138 2.911509 0.008629 0.185036 1.120125

Raw Bromide

0.282 0.104202 2.708164 0.013534 0.064835 0.499558

Raw Temp 0.173 0.041359 4.181989 0.00046 0.08669 0.259239

Treated pH -1.40 0.579789 -2.41619 0.025368 -2.6103 -0.19146

50

13.5%, and a reduction up to 30% can lower cancer risk by 20.8%. Through a

monthly bromide level analysis, it was determined that bromide levels do not

significantly vary within the year. This implies that it would not be feasible to alter

pre-chlorination levels during periods of low bromide, as these periods do not occur.

While bromide does not vary within the year, it is important to note that several

individual WTPs have noticeably higher bromide concentrations than others, and thus

have higher TTHM levels.

4.0 Conclusions

1) When bromide is at low concentrations, chloroform concentrations are nearly

100% of the TTHM formed, but when bromide is at higher concentrations

(approximately 0.5 mg/L), all trihalomethane subspecies are typically at

approximately equal levels, until bromide approaches 10 mg/L, the bromide

subspecies approach 100% of the TTHM formed.

2) A regression model (R2 = 0.91) was determined to characterize the formation of

TTHM with statistically significant variables of water temperature, dissolved

organic carbon, pH, pre-chlorination and raw water bromide concentrations; and

hence can be used to predict TTHM formation.

3) The number of water treatment plants that fall into non-compliance if the MCL for

TTHM was to be lowered from 100 µg/L to 80 µg/L is minimal, implying that an

MCL reduction would be relatively straightforward to attain. A reduction from

51

100µg/L to 60µg/L would result in approximately 15% of all WTPs falling into

non-compliance and would be more difficult to accomplish. To reduce the MCL

to a level where the cancer risk would be de minimus, or 1 in 1,000,000 would

result in nearly all WTPs being out of compliance, and would not be feasible.




Program and the Ontario Research Foundation.

6.0 References

“A National Survey of Chlorinated Disinfection By-Products in Canadian Drinking

Water”. 1995. Health Canada. 12 Dec 2011 <http://www.hc-sc.gc.ca/ewh-

semt/pubs/water-eau/byproducts-sousproduits/results-resultats-eng.php>.


Print.



1712-1721.

Chang, E., Lin, Y., and Chiang, P., “Effects of bromide on the formation of THMs

and HAAs.” Chemosphere 43(2001): pp 1029-1034.

52

Dermott, R., and Munawar, M.,"Invasion of Lake Erie Offshore Sediments by


Oceans, Canada 50(1993): pp 2298-2304.


2008. DWSP. 6 Aug 2009





377.












53

Harper, B., McBean, E., and Zhu, Z. (2012). Attaining Zebra Mussel Control and

Mitigating Disinfection By-Product Formation in Drinking Water Treatment.

Unpublished master‟s thesis, University of Guelph, Guelph, Ontario.


by Diving Ducks: An Exclosure Study”. The Ecological Society of America.Pp

521-531.

Jolley, R.L., Conference. Water chlorination environmental impact and health effects,

Vol. 1: Proceedings of the Conference on the Environmental Impact of Water



Kampioti, A., and Stephanou, E., “The impact of bromide on the formation of neutral

and acidic disinfection by-products (DBPs) in Mediterranean chlorinated

drinking water.” Water Research 36(2002): pp 2596-2606.

Kilgour, B.W., and Mackie., G. L.,Colonization of Different Construction Materials






Print.

Morrow, C., and Minear, R., “Use of Regression Models to Link Raw Water

Characteristics to Trihalomethane Concentrations in Drinking Water.” Water

Research 21(1987): pp 41-48.

54

"National Primary Drinking Water Regulations:" 16 December 1998. USEPA. 6 Aug







Systems?”.Water Res.(2003) 37, 329-338.

Ram, J. L., Fong, P. P and Garton, D. W.,"Physiological Aspects of Zebra Mussel


36(1996): pp 326-338.



Sun, Y., Wu, Q., Hu, H., and Tian, J., “Effect of bromide on the formation of

disinfection by-products during wastewater chlorination.” Water Research

43(2009): pp 2391-2398.

Tutt, A. (2009). TTHM Formation Modeling Using Regression Analysis in Ontario

Water. Unpublished master‟s thesis, University of Guelph, Guelph, Ontario.

Uyak, V., and Toroz, I., “Investigation of bromide ion effects on disinfection by-

products formation and speciation in an Istanbul water supply.” Journal of

Hazardous Materials 149(2007): pp 445-451.

55

Verween, Vincx , M., Degraer, S., “Comparative toxicity of chlorine and peracetic




Wang, G-S., Deng, and Lin, T., "Risk Assessment from Trihalomethanes in Drinking


56

Figure 1: Sum of Trihalomethane Subspecies in Ontario WTPs

0

5

10

15

20

25

30

Co

nce

ntr

atio

n (µ

g/L

)

Species

Sum of Trihalomethane Subspecies from 2004-2007 in Ontario WTPs

Bromoform

CDBM

Chloroform

BDCM

57

Figure 2: Subspecies formation based on bromide ion concentration (From Health Canada, 1995)

Figure 3: Percent speciation of TTHM subspecies given bromide concentrations in raw drinking water

0

0.2

0.4

0.6

0.8

1

<5 5-10 >10

Per

cen

t Sp

ecia

tio

n

Raw Water Bromide Concentration (ug/L)

Percent Speciation of TTHM Subspecies for all Ontario WTPs for 2004 - 2007

CF

CDBM

BF

BDCM

CF BF CDBM BDCM

58

Figure 4: Cancer Risk for TTHM Concentration of 20µg/L Given Bromide Concentration in Raw Water

Figure 5: TTHM Concentration when Bromide Concentration in Raw Water is Known

0.00E+00

5.00E-06

1.00E-05

1.50E-05

2.00E-05

2.50E-05

<10 >10

Can

cer

Ris

k

Raw Water Bromide Concentration (ug/L)

Cancer Risk for TTHM Concentration of 20µg/L Given

Bromide Concentration in Raw Water

CF

CDBM

BF

BDCM

TTHM

R² = 0.2321

0

20

40

60

80

100

120

0 10 20 30 40 50 60

TTH

M C

on

cen

trat

ion

(µ

g/L)

Raw Bromide Concentration (µg/L)

TTHM Concentration when Bromide Concentration in Raw Water is Known

59

Figure 6: Measured TTHM vs. Predicted TTHM from Equation 2

Figure 4: Measured TTHM vs. Predicted TTHM for Equations 1 and 2

R² = 0.9136

0

10

20

30

40

50

60

70

80

90

100

0 20 40 60 80 100

Mea

sure

d T

THM

µg/L


Measured TTHM vs. Predicted TTHM

0

10

20

30

40

50

60

70

80

90

100

0 20 40 60 80 100

Mea

sure

d T

THM

µg/L


Measured TTHM vs. Predicted TTHM

Equation 1

Equation 2

60

Figure 5: TTHM Cancer Risk with Altered Pre-Chlorination Dosage

1.00E-05

1.20E-05

1.40E-05

1.60E-05

1.80E-05

2.00E-05 C

ance

r R

isk

TTHM Cancer Risk with Altered Pre-Chlorination Dosage

TTHM Existing Chlorine

TTHM 10% Pre-Chlorine Reduction



61

Chapter 4

Paper #3: Modelling and Mitigation of Disinfection By-Product

Formation through the Development of a Bayesian Network Brett Harper, Zoe J. Y. Zhu and E. McBean

Abstract

Issues of Disinfection by-product (DBP) formation in response to chlorination in

drinking water treatment systems is a common issue encountered by WTP operators.

This is complicated by the presence of zebra mussels, which may inhabit the raw

water intake of WTPs. While chlorination at the intake to control zebra mussel

populations is effective, the formation of DBPs is exacerbated. A Bayesian network is

developed using the Webweavr-IV Toolkit, utilizing causal relationships between raw

water quality parameters in the form of conditional probabilities.

Four alternative chlorination scenarios are analyzed, one of which demonstrates the

probability of high TTHM concentrations (>80µg/L) can be reduced from 25.2% to

24.5%; and the probability of high cancer risk from TTHM (>10-5

) was reduced from

96.6% to 96.2%.

Keywords: Disinfection by-products (DBPs), Total Trihalomethanes, Bayesian

Network, Haloacetic acids (HAA), Organic matter, Chlorine.

62

1.0 Introduction

1.1 An Overview of Disinfection By-Products in Drinking Water

Total Trihalomethanes (TTHMs) are by-products created when the chlorine used in

the disinfection process reacts with naturally occurring organics (e.g. organics formed

by the decay of algae and vegetation) in raw water. In addition to chlorination,

TTHMs are known to also be influenced by temperature, bromide, and the

concentration of the dissolved organic carbon in the water (Health Canada, 2006). The

most common forms of trihalomethanes created are chloroform,

bromodichloromethane (BDCM), chlorodibromomethane (CDBM) and bromoform

indicating that the bromide ion is a contributing factor to the formation of TTHM

(Harper et al., 2012).

Chlorine and its compounds are the most commonly used disinfectants by water

treatment facilities. The popularity of chlorine is due to the combination of low cost,

high oxidizing potential, and the chlorine residual which exists throughout the

distribution system and protects against microbial recontamination (Sadiq and

Rodriguez, 2003). However in the 1970s, it was discovered that chlorine used during

water treatment reacts with organic matter, such as humic and fulvic acids, and

produces carcinogenic by-products. Given this, there is merit in reducing the

formation of these by-products, as feasible, by altering chlorination dosages.

However, additional issues which may counter the potential to alter chlorination

dosages is zebra mussels (Dreissena polymorpha), a class of mollusc similar to

oysters, clams, and scallops, which originated in the Black and Caspian seas and have

been inadvertently transported into the Great Lakes water by cargo ships. The mussels

63

grow to 1 inch in length and produce 35,000 eggs per season per female (Dermott et

al., 1993). The proliferation of zebra mussels is causing serious problems by clogging

raw-water intakes and discharge lines, increasing pipe corrosion, and producing

massive bio-fouling. As a result, zebra mussels are a recent problem for surface water

treatment facilities in parts of the US and Canada, particularly along the western

shores of Lake Erie. This paper evaluates several chlorination strategies to accomplish

lesser formation rates of the disinfection by-products while still controlling zebra

mussels.



Table 1 outlines the known health effects associated with different classes of DBPs.

Table 1: The health effects related to the major DBPs (modified from USEPA 1999)


Total Trihalomethanes Chloroform Cancer, liver, kidney and reproductive

effects



Bromodichloromethane Cancer, liver, kidney and reproductive

effects

Bromoform Cancer, nervous system, liver and kidney

effects

64

The current standard for TTHM set by Health Canada is 100 µg/L (Health Canada


their standards at 80 µg/L for TTHM (USEPA, 1999).

1.1 Research on chlorination and zebra mussels control

Verween et al. (2009) have shown that chlorination is an effective tool for eliminating

zebra mussels. One strategy to control zebra mussels is “Continuous Treatment”,

where chlorine is applied consistently to the water at concentrations around 2 mg/L at

the intake. This is very effective at killing all biological life in the water. It does,

however, produce high levels of DBPs (Harper et al., 2012).

Alternatively, zebra mussels can be controlled by killing the adult mussels by either

intermittent treatment, or periodic control chlorination (Klerks et al., 1993).

“Intermittent Treatment” includes killing the larvae before they settle and change into

their more-resilient juvenile forms. Sprecher et al. (2000) demonstrate that

chlorinating for 30 minutes every 12 hours is effective in controlling veliger (zebra

mussel larvae) populations and preventing new zebra mussels from settling; however,

this approach is not effective at killing established adult zebra mussels. By

chlorinating heavily for two to three weeks in late spring/early summer (in addition to

the preceding), the established adults can be killed. At this point, intermittent

treatment can be used to kill the young zebra mussels looking to “settle”.

It has been demonstrated that chlorination rates can be altered by incorporating

knowledge of the changing seasonal tolerances of zebra mussels (Costa et al., 2008).

A promising alternative to continuous treatment is “Periodic Treatment”, which

involves chlorinating “on” and “off” throughout the year. This treatment usually

65

consists of three treatment applications spread across the months of April to October.

Each treatment gives a dose of chlorine between 0.5 and 2 mg/L and lasts for 2-4

weeks. Costa et al. showed that in one experiment, a dose of 0.3 mg/L was applied in

3 sets of 2-3 week periods, resulting in a 95% mortality rate for zebra mussels.

2.3 Bromide Ion


formation of TTHMs (Health Canada, 1995). In the presence of chlorine and organic

material, as much as 50% of the bromide ion may become incorporated into

brominated trihalomethane subspecies (Chang et al., 2001; Harper et al., 2012). This

implies that a reduction in bromide concentration will have a significant impact on the

concentration TTHM. Until recently, only very limited bromide data have been

available. However, bromide data over a four year period (2005-2008) have recently

been released by the Ontario-based Drinking Water Surveillance Program (DWSP).

This study utilizes this recent bromide data in the development of a Bayesian network.

2.4 Bayesian Networks

This study employs the use of a Bayesian network to predict TTHM concentration

level by utilizing causal relationships between raw water quality parameters and

TTHMs, and causal relationships between the water quality parameters.

Bayesian networks are a type of intelligent system that represents domain knowledge

with a graphical structure that uses nodes to represent variables and arcs between the

66

nodes to represent dependencies between variables. A Bayesian network quantifies

this knowledge structure with probabilistic expressions of the interaction among

variables. Since probabilities are logical ways to quantify the unknown, Bayesian

networks are ideal intelligent systems. Bayesian networks also possess the unique

ability to incorporate expert estimates, as well as observed evidence, which allows

them to unite different kinds of uncertainty in a single theoretical environment (Olson

et al, 1990; van der Gaag, 1996).

Bayesian networks rely on Bayes‟ theorem to send information between nodes.

Bayes‟ theorem is displayed below.

The belief P(H|E) in a hypothesis given some evidence depends on the likelihood of

observing evidence, E, given both the hypothesis and its negation, P(E|H) and

P(E| H), and on the prior probability of the hypothesis P(H) (Pearl, 1982). The

denominator expresses a normalizing constant that indicates the prior probability of

the evidence, P(E).

Bayesian networks do not require that relationships between every variable be

specified. Rather, each variable in a Bayesian network requires only a local joint

probability distribution that reflects the possible configurations of its immediate

parents, if any, in the graph. A node with no parents is termed as „marginally

67

independent‟, meaning that the probabilities used to describe the node may be derived

from observed data or an estimate, but it is not dependant on any other nodes in the

network itself. Alternatively, conditional probabilities are used to describe the various

states of a node with parents, where each state of such a node is described in terms of

the various combinations of its parents‟ states. Ergo, the state of each child node is

conditional on the knowledge of the parents‟ states. This results in a network that

requires limited quantitative parameters to describe the influence of variables on one

another; conditional probabilities are described locally, but, through propagation of

probabilities from one node to the next, have global influence.

The WebWeavr IV Toolkit will be employed to model a Bayesian network that

utilizes the concentration levels of raw water parameters to predict concentration

levels of TTHM.


3.1 WEBWEAVR-IV

The Bayesian network was created using the WEBWEAVR-IV Toolkit (Xiang,

2006); a Java-based research toolkit developed for decision support systems based on

graphical models. WEBWEAVR-IV supports the construction of Bayesian networks,

inference in standard and dynamic Bayesian networks and decomposable Markov

networks, construction and verification of multiply-sectioned Bayesian networks

(MSBNs), and inference in multi-agent MSBNs.

68

3.2 Bayesian Network Development

Shown in Figure 1 is the Bayesian network developed for TTHMs, and cancer risk

prediction. The network displays how the various water quality parameters, namely

cancer risk and zebra mussels rely on the other parameters in the network. For

example, Zebra mussels are dependent on raw water temperature, and TTHMs are

dependent on bromine and DOC in raw water, and pre- and post- chlorine dosages.

3.3 Conditional Probabilities

The conditional probabilities for the Bayesian network were developed using data

from six WTPs (five from surface water sources, one from mixed surface water and

groundwater sources) in Ontario during 2005 to 2008. Only for six treatment plants

were all chlorination dosage data and bromide data available and hence used in the

Bayesian Network.

The probability values were developed from the raw water data as follows:

A series of ranges was assigned to each raw water quality parameter, and cancer risk,

and is shown in Table 2.

69

Table 2: Raw Water Quality Parameter Assigned Ranges

Parameter Range

Temperature Low: <12°C Med:12°C – 20°C High: >20°C

Zebra Mussels Low, Med, High

Raw DOC Low: <4.1µg/L Med: 4.1µg/L – 6.3µg/L High:

>6.3µg/L

Bromine Low: <12.9µg/L Med:12.9µg/L – 21.5µg/L

High: >21.5µg/L

Pre-

Chlorination

Dose

Low: <1.6mg/L Med:1.6mg/L – 1.76mg/L

High: >1.76mg/L

Post-

Chlorination

Dose

Low: <1.28mg/L Med: 1.32mg/L – 2.2mg/L

High: >2.2mg/L

Total

Trihalomethanes

Low: <50µg/L Med: 50µg/L – 80µg/L High:

>80µg/L

Cancer Risk Low: <10-6

Med: 10-6

– 10-5

High: >10-5

Each individual raw water datum was converted from a magnitude to either low,

medium, or high, depending on where each datum fell in the range.

Probability levels for three of the four parent nodes (Bromine, Pre-chlorination and

Post-chlorination) in the Bayesian network was determined to be equal split among

low, medium and high (0.333 probability for each level). The other parent node,

temperature, was assigned range values that did not reflect an equal frequency

between temperature levels, but rather correlated with zebra mussel spawning rates, as

zebra mussel populations tend to increase when the water is above 12°C.

70

Given the range values for the four parent nodes, as well as the ranges of each child

node in the Bayesian network, the conditional probabilities for each child node were

determined by calculating the probability the dependent water quality parameter,

given the value of the parent water quality parameter.

3.4 The Seasonal Impact on Zebra Mussel Control

To effectively control zebra mussels using chlorination, the seasonal variation of

zebra mussel susceptibility to chlorination patterns must be considered. According to

experiments by Bryant et al. (1985), Heinonen (2001), Rajagopal et al. (2002a), and

Costa et al. (2008), the high susceptibility of zebra mussels to chlorine is observed in

the U.S. to some extent in June and peaks in July and August, after reproduction. This

occurs because of mussels‟ low body weight, high filtration activity and the high

water temperatures. However, given Ontario‟s very long, cold winters and short

summers, the peaks of susceptibility move to between August and September (Harper

et al., 2012). Hence, it is in August and September when zebra mussels are most

susceptible to chlorine and is the optimal period to pre-chlorinate to reduce zebra

mussel levels. Given this, consideration of both zebra mussel control and treatment

strategies is feasible.

3.5 Cancer Risk

From TTHM concentrations, the exposure to each of these compounds, given by the

Lifetime Average Daily Dose (LADD), can be calculated.

The LADD is defined as:

71


Where,


IR = Intake Rate,





Assuming a body weight of 70kg, an intake rate of 2L of water a day (Gleick 1996)





0.029L/(kg*day).

The corresponding incremental excess lifetime cancer risk, IECLR, for TTHM

concentrations is calculated by multiplying the LADD by its corresponding CSF.

Harper et al. (2012) calculated the CSF for TTHM to be 0.0204 (mg/kg/day)-1

. The

resulting cancer risk probabilities for TTHM are shown in Table 3.

72

4.0 Integrated Analysis

4.1 Comparison of Altered Pre- and Post- Chlorination Levels for TTHM

Reduction

Consider four scenarios with varying levels of pre- and post-chlorination in an effort

to mimic alternative approaches to chlorination while also incorporating mussel

control. The variables being examined were: pre-chlorination, post-chlorination,

treated water temperature, raw DOC, and bromine. Using the Bayesian network,

values of TTHM and cancer risk were calculated. The four scenarios are as follows:

Scenario 1: pre-chlorination was regular (the same as current conditions) but post-

chlorination was set to the low level (<1.28mg/L).

Scenario 2: During the months of May, July and September, pre-chlorination was set

to 0.5mg/L and the rest of the year it was set to 0.001mg/L, and post-chlorination was

set to 3mg/L. This test follows “Intermittent Treatment” (Sprecher, 2000), where there

are three treatments a year using pre-chlorination.

Scenario 3: This test was similar to scenario 2, but instead utilized a process known as

“End-of-season treatment” (after Sprecher, 2000). Pre-chlorination was assumed at a

concentration of 0.75mg/L during the months of May and July, and at a concentration

of 3.0mg/L in September. The purpose of the higher dose in September was to flush

73

out any remaining adult zebra mussels prior to the next season beginning. Again, the

post-chlorination was 3mg/L.

Scenario 4: This scenario is quite different from the preceding scenarios. In this

scenario, “Continuous Treatment” was used during the months where the water

temperature exceeded 12oC. During the months of May to November the dose of pre-

chlorine was set to 3mg/L, and at 0.001mg/L for the remainder of the year. During the

months of May to November the dose of post-chlorine was set to 0.3mg/L, and

3.3mg/L for the remainder of the year. This scenario was designed to be the most

secure test in terms of microbial disinfection and zebra mussel control.

4.2 Bayesian Network Results

Shown in Table 3 are the probability results generated by the Bayesian Network for

current, average water quality conditions of reported by WTPs across Ontario.

Table 3: Bayesian Network Results

Parameter Range Low Medium High

Temperature Low: <12°C Med:12°C – 20°C High: >20°C 0.517 0.207 0.276

Zebra Mussels Low, Med, High 0.359 0.334 0.307

Raw DOC Low: <4.1µg/L Med: 4.1µg/L – 6.3µg/L High:

>6.3µg/L 0.313 0.334 0.353

Bromine Low: <12.9µg/L Med:12.9µg/L – 21.5µg/L

High: >21.5µg/L 0.333 0.333 0.333

Pre-

Chlorination

Dose

Low: <1.6mg/L Med:1.6mg/L – 1.76mg/L

High: >1.76mg/L 0.333 0.333 0.333

74

Post-

Chlorination

Dose

Low: <1.28mg/L Med: 1.32mg/L – 2.2mg/L

High: >2.2mg/L 0.333 0.333 0.333

Total

Trihalomethanes

Low: <50µg/L Med: 50µg/L – 80µg/L High:

>80µg/L 0.459 0.289 0.252

Cancer Risk Low: <10-6

Med: 10-6

– 10-5

High: >10-5

0 0.034 0.966

The results in Table 3 display the predicted probability of each outcome occurring.

For example, temperature has a probability of 0.517, or a 51.7% chance of being

under 12°C throughout any given year. Of interest is that given the input parameters

of average WTP raw water concentrations from the data from 2005 – 2008, cancer

risk has a 0% chance of being less than „de minimus‟ (one in a million), a 3.4%

chance of being between one in a million and 1 in 100,000, and a 96.6% chance of

being over 1 in 100,000. Shown in Table 4 is a comparison of how the current,

average conditions relate to those in Scenarios 1 through 4.

Table 4: Comparison of Current, Average Conditions with the Four Scenarios

Scenario Magnitude

Parameter

Pre-

Chlorination

Dose

Post-

Chlorination

Dose

Total

Trihalomethan

es

Cancer Risk

Original

Data

Low 0.333 0.333 0.459 0.000

Medium 0.333 0.333 0.289 0.034

High 0.333 0.333 0.252 0.966

Scenario 1

Low 0.333 1.000 0.510 0.000

Medium 0.333 0.000 0.245 0.038

High 0.333 0.000 0.245 0.962

Scenario 2 Low 0.750 0.000 0.354 0.000

75

Medium 0.250 0.000 0.338 0.027

High 0.000 1.000 0.308 0.973

Scenario 3

Low 0.750 0.000 0.348 0.000

Medium 0.167 0.000 0.340 0.026

High 0.083 1.000 0.312 0.974

Scenario 4

Low 0.500 0.500 0.435 0.000

Medium 0.000 0.000 0.296 0.033

High 0.500 0.500 0.270 0.967

Table 4 displays the probabilities of pre-chlorination dose, post-chlorination dose,

TTHM, and cancer risk being low, medium, or high (Refer to Table 2 for the range of

magnitudes of each level). For example, the probability of TTHM being low in

Scenario 1 is 0.510, or 51.0%. For each Scenario, the probabilities of a given

parameter being low, medium or high, all sum to 1.

It was found that cancer risk is at its lowest in Scenario 1, as the probability of a high

cancer risk (>10-5

) is 96.2%, which is lower than the probability of high cancer risk of

96.6% that is associated with current WTP conditions. All other Scenarios

demonstrate probabilities of high cancer risk greater than those in Scenario 1. The

small change in cancer risk can be attributed to the lack of data available for the

Bayesian network. Lack of data requires that the Bayesian network have very coarse

grid spacing (only three levels: low, medium and high), which results in the output of

the Bayesian network being equally coarse. This means that either very large, or very

minute changes in cancer risk will be observed.

76

It was also found that current WTP conditions and all Scenarios demonstrated a

probability of 0, or a 0% chance for cancer risk to be low (<10-6

). Given this, it is

evident that under current WTP conditions, as well as in all possible Scenarios, that

cancer risk is above „de minimus‟ (one in a million).

The probability of a high level of TTHM (80µg/L) being encountered is lowest in

Scenario 2, at 24.5%. Original conditions have a 25.2% chance of high trihalomethane

concentrations, followed by Scenario 4 at 27.0%, Scenario 2 at 30.8%, and Scenario 3

at 31.2%.

5.0 Conclusions

1. A Bayesian network was developed to predict TTHM concentration levels.

This was done using the Webweavr-IV Toolkit, and by utilizing the causal

relationships between raw water quality parameters and TTHM, as well as the

causal relationships between the water quality parameters, themselves.

2. Under current, average WTP conditions, cancer risk was found to be above „de

minimus‟ (one in a million).

3. Scenario 1 was found to be the most effective chlorination alternative to

current practices because the probability of high TTHM concentrations was

reduced from 25.2% to 24.5%; and the probability of high cancer risk from

TTHM was reduced from 96.6% to 96.2%. The small change in cancer risk

can be attributed to the lack of data available for the Bayesian network, which

results in the output of the Bayesian network being coarse.

77




Program and the Ontario Research Foundation.

7.0 References





Print.



1712-1721.



Dermott, R., and Munawar, M.,"Invasion of Lake Erie Offshore Sediments by


Oceans, Canada 50(1993): pp 2298-2304.

78


2008. DWSP. 6 Aug 2009





377.













by Diving Ducks: An Exclosure Study”. The Ecological Society of America.Pp

521-531.

79

Harper, B., McBean, E., and Zhu, Z. (2012). Attaining Zebra Mussel Control and

Mitigating Disinfection By-Product Formation in Drinking Water Treatment.

Unpublished Manuscript, University of Guelph, Guelph, Ontario.

Harper, B., McBean, E., and Zhu, Z. (2012). Mitigation of Disinfection By-Product

Formation by the Development of a Regression Equation with the Bromide

Ion. Unpublished Manuscript, University of Guelph, Guelph, Ontario.








Kilgour, B.W., and Mackie., G. L.,Colonization of Different Construction Materials






Print.



Research 21(1987): pp 41-48.

80

"National Primary Drinking Water Regulations:" 16 December 1998. USEPA. 6 Aug





Olson, R.L., Willer, J.L. and Wager, T.L. 1990. A Framework for Modeling

Uncertain Reasoning in Ecosystem Management: Bayesian Belief Netowrks.

AI Applications 4(4):11-23.



Systems?”.Water Res.(2003) 37, 329-338.

Ram, J. L., Fong, P. P and Garton, D. W.,"Physiological Aspects of Zebra Mussel


36(1996): pp 326-338.





43(2009): pp 2391-2398.



81




Van der Gaag, L., 1996. Bayesian Belief Networks: Odds and Ends. The Computer

Journal 39(2):97-113.

Verween, Vincx , M., Degraer, S., “Comparative toxicity of chlorine and peracetic




Wang, G-S., Deng, and Lin, T., "Risk Assessment from Trihalomethanes in Drinking


Xiang, Y. (2002). Probabilistic Reasoning in Multiagent Systems: A graphical models

approach. Cambridge, UK: Cambridge University Press.

Xiang, Y. (2006). Webweaver-IV Research Toolkit. [Software]. Available from

http://www.socs.uoguelph.ca/~yxiang/

82

Figure 1: Bayesian Network to Determine TTHM levels and Resulting Cancer Risk

83

Chapter 5

Conclusions

5.1 Summary

In Ontario, cancer risks from TTHMs and its subspecies with current disinfection

practices currently exceed one in a million and the incremental cancer risk ranges

from 1 in 50,000 to 100,000. In an effort to reduce cancer risk from TTHMs, two

multiple regression models which predict the formation of TTHMs were developed

using water quality data from Ontario WTPs. The first TTHM model was developed

using data from 2000 to 2003; and water temperature, dissolved organic carbon in raw

water, dissolved organic carbon in treated water, pre-chlorination and post-

chlorination were all found to be statistically significant parameters. The other TTHM

model was derived from data from 2005 to 2008; and water temperature, dissolved

organic carbon, pH, pre-chlorination and raw water bromide concentrations were

found to be statistically significant.

The first model (R2

= 0.75) was used to estimate how TTHM levels could be reduced,

while still continuing to provide adequate zebra mussel removal. The model that was

developed results in the pre-chlorination variable having a higher exponent than the

post-chlorination variable, indicating that TTHM concentrations are more heavily

influenced by pre-chlorination dosages. This indicates that reducing the pre-

chlorination dose when feasible, and making up for this reduction by increasing the

post-chlorination dose has merit. This can be accomplished in winter months as zebra

mussels are considerably less active during this time. The model was used to develop

a strategy for effectively controlling zebra mussels, which is to pre-chlorinate at

84

3mg/L when the temperature exceeds 12oC, and post-chlorinate year-round. This

strategy results in a reduction of TTHM formation of 24% and reduces cancer risk by

24%, while maintaining effective disinfection and zebra mussel removal. During the

development of this model, it was found that there is a strong correlation between

increasing temperature and increasing chloroform and TTHM concentrations and that

there is not as strong of a correlation between the other TTHM subspecies and

temperature.

The second model (R2 = 0.91) was used to estimate how TTHM levels could be

reduced, but does not take zebra mussels into account due to post-chlorination

concentrations not being statistically significant in the model. During the development

of this model, further investigation was done on the newly-documented bromide ion.

It was found that when bromide is at low concentrations, chloroform concentrations

are nearly 100% of the TTHM formed, but when bromide is at higher concentrations

(approximately 0.5 mg/L), all trihalomethane subspecies are typically at

approximately equal levels, until bromide approaches 10 mg/L, when the bromide

subspecies approach 100% of the TTHM formed.

A Bayesian network was developed to predict TTHM concentration levels. This was

done using the Webweavr-IV Toolkit, and by utilizing the causal relationships

between raw water quality parameters and TTHM, as well as the causal relationships

between the water quality parameters, themselves. Holding pre-chlorination the same

as current conditions and setting post-chlorination to the low level (<1.28mg/L), was

found to be the most effective chlorination alternative to current practices. This

85

reduced the probability of high TTHM concentrations from 25.2% to 24.5%; and

reduced the probability of high cancer risk from TTHM from 96.6% to 96.2%.

5.2 Recommendations

It may be advisable for the DWSP to encourage participating WTPs to record water

quality results on a monthly basis, rather than 2-4 times per year. This would make it

easier to identify the seasonal changes of DBP formation, as well as enhance the

predictive capabilities of the multiple regression equations and the Bayesian network.

Consequently, zebra mussel mitigation could be enhanced, as DBP levels could be

more accurately calculated, allowing pre-chlorination levels to be optimized for both

zebra mussel control and minimal DBP formation.

Further documentation of the bromide ion would be beneficial to future studies. It has

been shown in this study that bromide plays a statistically significant role in the

formation of TTHMs in Ontario. Currently, bromide is monitored by few WTPs in

Ontario and thus the available data is still rather sparse. Further documentation of the

bromide ion could improve the accuracy of the regression equations and Bayesian

network, which would provide more information for WTP operators to mitigate

TTHM formation.

Greater amounts of water quality data (WTP data, expert opinions, or wet lab

experiments) are required to appropriately train the Bayesian network. Currently, the

amount of data available from the DWSP is limited when compared to the large

86

requirements of the Bayesian network. Obtaining this data would be beneficial to

creating a more useful Bayesian network.

87

References





Print.





1712-1721.

Dermott, R., and Munawar, M., "Invasion of Lake Erie Offshore Sediments by


Oceans, Canada 50(1993): pp 2298-2304.


2008. DWSP. 6 Aug 2009





377.

88













by Diving Ducks: An Exclosure Study”. The Ecological Society of America.

Pp 521-531.








Kilgour, B.W., and Mackie., G. L., Colonization of Different Construction Materials


89





Print.



Research 21(1987): pp 41-48.

"National Primary Drinking Water Regulations:" 16 December 1998. US EPA. 6 Aug





Olson, R.L., Willer, J.L. and Wager, T.L. 1990. A Framework for Modeling

Uncertain Reasoning in Ecosystem Management: Bayesian Belief Netowrks.

AI Applications 4(4):11-23.



Systems?”. Water Res.(2003) 37, 329-338.

Ram, J. L., Fong, P. P and Garton, D. W., "Physiological Aspects of Zebra Mussel


36(1996): pp 326-338.

Richards, D. Chief Operator, Brockville Water Treatment. Phone: (613) 342-7819

90





43(2009): pp 2391-2398.






Van der Gaag, L., 1996. Bayesian Belief Networks: Odds and Ends. The Computer

Journal 39(2):97-113.

Verween ., , Vincx , M., Degraer, S., “Comparative toxicity of chlorine and peracetic




Wang, G-S, Deng, and Lin. "Risk Assessment from Trihalomethanes in Drinking


Documents

Mitigation of Disinfection By-Product Formation through