1 SALTWATER INTRUSION IMPACTS ON BROMIDE CONCENTRATION AND DISINFECTION BYPRODUCT FORMATION: MODEL EVALUATION AND LABORATORY SCALE ANALYSIS By EVAN CHARLES GED A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF ENGINEERING UNIVERSITY OF FLORIDA 2013
2 2013 Evan Charles Ged
3 To my fa ther
4 ACKNOWLEDGEMENTS I would like to thank my parents and family for their constant support and encouragement throughout my graduate career. I would also like thank Emily Morton for her friendship and support over the last two years. In addition to my friends and family I would like to thank my advisor Dr. T reavor Boyer, without his guidance none of this would have been possible. I also give thanks to Dr. Paul Chadik, Dr. Louis Motz, Dr. Jonathan Martin, Dr. Kathryn Frank, Jack Kurki Fox, Pedro Palomino, and the Boyer research group for all of their valuable input. I would like to acknowledge those who helped with sample collection including the Cedar Key Water and Sewer District, the United States Geological Survey (USGS), Broward County Natural Resources Planning & Management Division, and Advanced Environme ntal Laboratories for performing THM and HAA analysis. This work was supported by the Research Opportunity Seed Fund fice of Research and Florida Sea Grant award PD 12 22.
5 TABLE OF CONTENTS Page ACKNOWLEDGEMENTS ................................ ................................ ............................... 4 LIST OF TABLES ................................ ................................ ................................ ............ 7 LIST OF FIGURES ................................ ................................ ................................ .......... 8 LIST OF ABBREVIATIONS ................................ ................................ ............................. 9 ABSTRACT ................................ ................................ ................................ ................... 11 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ .... 14 2 CRITICAL EVALUATION OF DISINFECTION BYPRODUCT MODELS: VALIDATION AND STATISTICAL ANALYSIS ................................ ........................ 17 2.1 Overview of DBPs and DBP Modeling ................................ .............................. 17 2.2 Background on DBP modeling ................................ ................................ .......... 18 2.3 Experimental Approach ................................ ................................ ..................... 22 2.3.1 DBP Models ................................ ................................ ........................... 23 2.3.2 DBP Formation Data Sets ................................ ................................ ..... 24 2.3.3 Statistical Analysis ................................ ................................ ................. 24 2.4 Results ................................ ................................ ................................ .............. 26 2.4.1 Overview of DBP Mode ls ................................ ................................ ....... 26 2.4.2 Trihalomethanes ................................ ................................ .................... 27 22.214.171.124 THM4 ................................ ................................ ........................ 27 126.96.36.199 Individua l THM species ................................ ............................. 29 2.4.3 Haloacetic Acids ................................ ................................ .................... 30 188.8.131.52 HAA5, HAA6, and HAA9 ................................ ........................... 30 184.108.40.206 Individual HAA species ................................ ............................. 31 2.5 Discussion ................................ ................................ ................................ ........ 32 2.6 Summary of the Critical Evaluation of DBP Models ................................ .......... 36 3 EFFECT OF SALTWATER INTRUSION ON BROMINE SPECIATION OF TRIHALOMETHANES AND HALOACETIC ACIDS ................................ ................ 62 3.1 Overview of Saltwater Intrusion Impa cts ................................ ........................... 62 3.2 Experimental Section ................................ ................................ ........................ 69 3.2.1 Sampling Location ................................ ................................ ................. 69 3. 2.2 Chlorine Demand and Chlorination Under Uniform Formation Conditions ................................ ................................ .............................. 70 3.3 Analytical Methods ................................ ................................ ............................ 71 3.3.1 DOC Measurements ................................ ................................ .............. 71
6 3.3.2 Ultraviolet Absorbance at 254 nm (UVA 254 ) ................................ ........... 71 3.3.3 Inorganic Anions ................................ ................................ .................... 71 3.3.4 pH ................................ ................................ ................................ .......... 72 3.3.5 Chlorine Residual ................................ ................................ .................. 72 3.3.6 THM Analysis ................................ ................................ ........................ 72 3.3.7 HAA Analysis ................................ ................................ ......................... 73 3.4 Results and Discussion ................................ ................................ ..................... 74 3.4.1 Effects of Saltwater Intrusion on Bromide Concentration ....................... 74 3.4.2 Effects of Saltwater Intrusion on THM and HAA Formation ................... 76 3.4.3 Effects of Saltwater Intrusion on Bromine Incorporation Fac tor ............. 79 3.4.4 Effects of Saltwater Intrusion on Health Risks of DBPs ......................... 81 3.5 Implications of the Impacts of Saltwater Intrusion on THM and HAA Formation ................................ ................................ ................................ ......... 83 4 CONCLUSIONS AND RECOMMENDATIONS ................................ ....................... 92 LIST OF REFERENCES ................................ ................................ ............................... 95 BIOGRAPHICAL SKETCH ................................ ................................ .......................... 104
7 LIST OF TABLES Table Page 2 1 Published DBP models for THM4, HAAs, and individual THM and HAA species. ................................ ................................ ................................ .............. 38 2 2 Summary of water quality reported for validation data sets. ............................... 47 2 3 Descriptive statistics for THM4 models from Boyer and Singer (2005) .............. 48 2 4 Descriptive statistics for THM4 models from Amy et al. (1993 ) ......................... 49 2 5 Descriptive statistics for individua l THM species models. ................................ ... 50 2 6 Descriptive statistics for HAA models. ................................ ................................ 51 2 7 Descriptive statistics for individual HAA species mod els. ................................ ... 52 3 1 Water quality parameters for various mixing ratios in the chlorination experiments. ................................ ................................ ................................ ....... 87
8 LIST OF FIGURES Figure Page 2 1 Measured versus predicted THM4 concentrations from Boyer and Singer (2005) ................................ ................................ ................................ ................ 53 2 2 Measured versus predicted THM4 concentrations from Amy et al. (1993) ........ 54 2 3 Measured versus predicted THM4 concentrations for best fitting models from Amy et al. (1993) .. ................................ ................................ .............................. 55 2 4 Measured ve rsus predicted concentrations of chloroform (CHCl 3 ). .................... 56 2 5 Measured versus predicted concentrations of bromodichloromethane (CHCl 2 Br). ................................ ................................ ................................ ........... 57 2 6 Measured versus predicted concentrations of dibromochloromethane (CHClBr 2 ). ................................ ................................ ................................ ........... 58 2 7 Measured versus predicted concentrations of bromoform (CHBr 3 ). ................... 59 2 8 Measured versus predicted concentrations of HAAs. ................................ ......... 60 2 9 Measured versus predicted concentrations of individual HAA species for both raw and treated wat ers. ................................ ................................ ...................... 61 3 1 DBP yields as a function of seawater concentration.. ................................ ......... 88 3 2 DBP speciation as a function of seawater concentration. ................................ ... 89 3 3 Bromine incorporation factor (BIF) for THM4, X 2 AA, and X 3 AA as a function of seawater added by volume. ................................ ................................ ............ 90 3 4 Risk (exces s cancer cases/10 6 people) as a function of seawater concentration.. ................................ ................................ ................................ .... 91
9 L IST O F ABBREVIATIONS BCAA Bromochloroacetic acid BDCAA Bromodichloroacetic acid BDCM Bromodichloromethane BIF Bromine incorporation factor DBCA A Dibromochloroacetic acid DBCM Dibromochloromethane DBP Disinfection byproducts DBPFP Disinfection byproduct formation potential DBAA Dibromoacetic acid DCAA Dichloroacetic acid DOC Dissolved organic carbon DOM Dissolved organic matter HAA Haloace tic acids MBAA Monobromoacetic acid MCAA Monochloroacetic acid MCL Maximum contaminant level MPSD Marquardt's percent standard deviation NOM Natural organic matter NVTOC Non volatile total organic carbon RO Reverse osmosis SE Standard error SUVA 254 Specific ultraviolet absorbance at 254 nm
10 TBAA Tribromoacetic acid TCAA Trichloroacetic acid THM Trihalomethanes TDS Total dissolved solids TOC Total organic carbon UFC Uniform formation conditions UV 254 Ultraviolet absorbance at 254 nm X 2 AA Dihalo acetic acid X 3 AA Trihaloacetic acid
11 Abstract of Thesis Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Master of Engineering SALTWATER INTRUSION IMPACTS ON BR OMIDE CONCENTRATION AND DISINFECTION BYPRODUCT FORMATION: MODEL EVALUATION AND LABORATORY SCALE ANALYSIS By Evan Charles Ged August 2013 Chair: Treavor H. Boyer Major: Environmental Engineering Sciences The overall objective of this work was to evalua te the effect of saltwater intrusion on bromide concentrations in fresh groundwater and how those concentrations impact the formation and speciation of disinfection byproducts (DBPs). This study involved two separate but related components (i) evaluating p ublished predictive models for trihalomethanes (THMs) and haloacetic acids (HAAs) using two comprehensive data bases and (ii) conducting laboratory experiments to demonstrate increases in THM and HAA formation as a function of saltwater intrusion as well a s the shift toward bromine containing species. The first part of this work focused on evaluating 85 peer reviewed DBP models for THMs and HAAs from 23 different publications. Of the 85 models studied, 38 were for THM4, 15 were for either HAA5, HAA6, or HAA 9, 22 for individual THM species, and nine for individual HAA species. The models were evaluated using two data sets where the model output was DBP concentrations (g L 1 ) and then compared this output to measured data to assess model accuracy. The three s tatistical metrics used for determining accuracy were coefficient of determination (R 2 ), standard error (SE), and
12 Marquardt's Percent Standard Deviation (MPSD). For both data bases used in this analysis, the best performing models for THM4 were capable of predicting DBP formation with an SE of approximately 45 g L 1 an R 2 of > 0.95, and an MPSD of about 30%. For HAAs the best models yielded an SE of 15 g L 1 an R 2 > 0.93, and an MPSD of also 30%. Individual species occur at much lower concentrations res ulting in lower SE; however, MPSD and R 2 were generally not as good as THM4 or HAA models. To investigate the effects of bromide, models were plotted separately as models including explanatory variable for bromide and models that did not. For the models ex cluding a term for bromide it was found that they tended to under predict DBP formation in waters high in bromide although some of these models were still accurate over a range of water sources. The models that included bromide as an explanatory variable w ere capable of more accurate predictions over a range of measured values, as indicated by high R 2 values. The second part of this study aimed to simulate saltwater intrusion in the laboratory by mixing fresh groundwater with Gulf of Mexico seawater at vari ous ratios ranging from 0% to 2.0% by volume saltwater content. The protocol consisted of performing chlorine demand tests to determine the amount of chlorine demand exerted by the sample matrix and then forming DBPs under uniform formation conditions (UFC ). At the no saltwater intrusion scenario THM4 and HAA5 were 43.4 g L 1 and 24.3 g L 1 respectively. When increasing the saltwater content to 2% (approximately 1000 g L 1 bromide) the THM4 and HAA5 increased to 206.5 g L 1 and 26.0 g L 1 a 376% and 7% increase respectively. The THM4 change was substantial, exceeding the maximum contaminant level (MCL) of 80 g L 1 Increases in HAA5 were small since the five
13 regulated HAAs do not include the most brominated species. When considering HAA9 there was an observed increase from 39.2 g L 1 to 75.2 g L 1 (MCL = 60 g L 1 ), a 92% increase. The bromine incorporation factor (BIF) suggested the molar yields of each DBP class consisted of >90% organic bromine at 974 g L 1 bromide but only 7 to 31 % at the bas eline conditions of 38 g L 1 bromide. Considering the well documented adverse health effects of brominated THMs and HAAs, the implications for saltwater impacted groundwaters could pose an increased risk to human consumers. In addition to increased cancer risk, utilities experiencing sudden or progressive increases in bromide can be in danger of exceeding MCLs for THMs and generate higher concentrations of HAA9. Overall this work demonstrated that many coastal areas may be threatened by saltwater intrusion and in the future it will be necessary to move to alternative water sources and/or implement advanced treatment processes such as ion exchange or reverse osmosis to remove DBP precursors.
14 CHAPTER 1 INTRODUCTION The phenomenon of saltwater intrusion has been o bserved in coastal communities worldwide, replacing fresh groundwater with unpalatable water high in dissolved solids. Researchers have investigated the mechanisms, both natural and anthropogenic, that have resulted in saltwater intrusion and subsequent sa linization of coastal aquifers. These mechanisms can include over pumping to meet drinking water demand, land development which alters aquifer recharge, and climate induced factors such as severe drought, land subsidence, and sea level rise (Oude Essink et al., 2010) In addition to high quality freshwater becoming scarc er, global population projections are stressing water resources. These stresses are considerably worse in coastal regions where approximately 40% of all people live in the near coastal zone that is within 100 km of the coastline and less than 100 m above s ea level (Small & Nicholls, 2003). In the future it will be necessary to utilize alternative water sources such as brackish water, desalinated seawater, and reclaimed water to augment water supplies. During saltwater intrusion, bromide ion (Br ) infiltrate s drinking water aquifers and alters the chemistry of the groundwater matrix (Liu et al., 2011) When amb ient levels of bromide are elevated there is the potential to alter the formation and speciation of disinfection byproducts (DBPs) during water treatment resulting in adverse health effects on human consumers. The first DBPs were discovered in treated drin king water in the early 1970's by Rook and colleagues (1974); soon after, studies published by the National Cancer Institute (1976) demonstrated these byproducts were carcinogenic to laboratory animals. As a result of these studies the first DBPs, trihalom ethanes (THMs), were regulated in 1979 and spurred new interest in studying the byproducts of chlorine
15 disinfection (U.S. EPA, 1979). Today there are 11 DBPs regulated in the U.S. including THM4 (the summation of the four regulated species), five haloaceti c acid species (HAA5), bromate, and chlorite (U.S. EPA, 2006), the respective byproducts of ozonation and chlorine dioxide disinfection. This study focuses on the two classes of halogenated organic byproducts, THM4 and HAA5 which are regulated at a maximum contaminant level (MCL) of 80 g L 1 and 60 g L 1 respectively (U.S. EPA, 2006). Since the recognition of these compounds as a threat to public health researchers have sought to develop mathematical models to predict DBP formation as a means to better un derstand disinfection practices, reaction mechanisms, effectiveness of treatment process to remove DBP precursors, regulatory compliance, and risks to consumers. There are currently over 100 models that have been developed for predicting THMs, HAAs, indivi dual THM and HAA species, bromate, and unregulated species of DBPs including haloacetonitriles (HANs) and N nitrosodimethylamine (NDMA). These models can be useful tools for making the connection between saltwater intrusion, elevated levels of bromide, and the subsequent formation of brominated DBPs (Br DBPs). Several recent articles have focused on demonstrating the toxicity of Br DBPs, where many publications suggest these compounds are more toxic than their chlorinated analogues (Richardson et al., 2002; 2007). When present at sub mg L 1 concentrations, bromide can cause dramatic increases in the overall magnitude of THM4 and HAA9 and shift the speciation towards the bromine containing species. The mass yields of these byproducts can create issues with re gulatory compliance and pose a greater risk to human health. Hua et al. (2006) demonstrated that spiking natural waters with bromide can increase the levels of THM4, HAA5, and HAA9 above
16 regulatory thresholds. Similar effects have been observed by Cowman a nd Singer (1996) and Wu and Chadik (1998). The concentration difference between HAA9 and HAA5 also increases as bromide concentrations increase due to the shift of the relative speciation toward the four unregulated HAAs (Hua et al., 2006). It is hypothesi zed that as saltwater intrusion becomes more prominent, Br DBPs will be more of a concern to utilities, engineers, and water resources planners. Thus, it is important to understand the changes in THM4, HAA5, and HAA9 that will occur as a function of saltwa ter intrusion. The title of Chapter 2 is "Critical Evaluation of Disinfection Byproduct Models: Validation and Statistical Analysis." Chapter 2 discusses the existing DBP models found in the literature and assesses their predictive capabilities for THMs an d HAAs as a function of various precursor concentrations and chlorination conditions. The title of Chapter 3 is "Effects of Saltwater Intrusion on Bromine Speciation of Trihalomethanes and Haloacetic Acids." Chapter 3 discusses bench scale chlorination exp eriments that were designed to mimic various saltwater intrusion scenarios and quantify the magnitude of THM4, HAA5, and HAA9 formed as well as the individual speciation. Major findings from Chapters 2 and 3 are presented in Chapter 4, along with suggestio ns for future work.
17 CHAPTER 2 CRITICAL EVALUATION OF DISINFECTION BYPRODUCT MODELS: VALIDATION AND STATISTICAL ANALYSIS 2.1 Overview of DBPs and DBP Modeling One of the major advancements in public health was the implementation of chlorine disinfection in drinking wa ter treatment. The ability of chlorine to inactivate pathogenic microorganisms with high efficiency and at low cost resulted in global the formation of harmful disinfection byproducts (DBPs) has been well documented in drinking water systems employing chlorine. Excellent reviews on the history and timeline of DBP discovery, regulation, and occurrence are available in the literature (Richardson et al., 2002; 2007). Key events in the his tory of DBPs include the discovery by Rook (1974) of chloroform as a byproduct of chlorinating waters containing natural organic matter (NOM) and evidence that chloroform was carcinogenic to laboratory animals (National Cancer Institute, 1976). The outcome of these studies raised awareness about harmful contaminants in drinking water and resulted in the recognition of DBPs by public health professionals and researchers. Following the discovery of DBPs and the first regulation of trihalomethanes (THMs) in 19 79 by the U.S. Environmental Protection Agency (U.S. EPA 1979 ), researchers began investigating the formation mechanisms and health risks associated with DBPs leading to numerous publications, e.g., Zoeteman, 1985; Boorman et al., 1999; Plewa et al., 2002 ; Hua et al., 2006; Krasner et al., 2006; Richardson et al., 2007; Baytak et al., 2008; Parinet et al., 2012; Pan and Zhang, 2013 A specific field of study that has emerged from DBP research is the development of mathematical models for predicting DBP f ormation under various water quality and treatment scenarios. DBP models can be useful tools for water utility managers,
18 operators, engineers, epidemiologist, and water resources planners by helping them comply with DBP regulations, such as the Stage 2 Dis infectants and Disinfection Byproduct Rule (Stage 2 D/DBP Rule) (U.S. EPA, 2006), and reduc ing health risks to the general population. The most studied DBPs are THMs and haloacetic acids (HAAs), two classes of compounds that are currently regulated by the U.S. EPA as being subject to guidelines in Canada, Australia, the European Union, and the World Health Organization. Bromate and chlorite are the only other DBPs regulated in the U.S. which form as a result of ozonation and chlorine dioxide disinfection, r espectively. Although more than 100 DBP formation models have been published in the last 30 years, many of these models are not applicable to a range of drinking water sources and can often underestimate DBP formation by excluding key explanatory variables This work aims to critically evaluate existing DBP models, identify the best models using external data and statistical metrics, and assess their accuracy as predictive tools in drinking water treatment. 2.2 Background on DBP modeling A recent review articl e by Chowdhury et al. (2009) discusses existing DBP models from 48 peer reviewed journals: 42 models focus on THMs, 8 on HAAs, 5 on bromate, and 1 on chlorite. For a comprehensive review of existing DBP models and their advantages and limitations the reader is directed to Chowdhury et al. (2009) and an earlier a rticle by Sadiq and Rodriguez (2004). Both articles provide a chronological summary of existing models demonstrating that several of the models are capable of generating linear correlations (R 2 ) of > 0.90 based on results from their respective studies. In addition to a handful of review articles many publications are research articles that discuss the development of new models for total THMs (TTHMs or THM4)
19 (Morrow & Minear, 1987; Golfinopoulos & Arhonditsis, 2002; Hong et al., 2007; Chowdhury, 2009) HAAs (Le kkas & Nikolaou, 2004; Sohn et al., 2004) brominated THM and HAA species (Siddiqui et al., 1998; Fabbricino & Korshin, 2009; Chowdhury et al., 2010) and unregulated and emerging classes of DBPs (Chen & Westerhoff, 2010) However, many models may not be applicable to a wide range of source waters because of the experimental design or methodology involved in the model formulation. For example, some researchers have correlated explanatory variables (e.g., DBP precursors) to DBP formation potential (DBPFP) using low bromide waters creating inaccurate model predictions when using the models with waters that contain higher levels of b romide (Rathbun, 1996; Nikolaou et al., 2004) A study by Chen and Westerhoff (2010) generated accurate models for predicting total THMs ( TTHM or THM4), HAA9 (the summation of all 9 HAA species), as well as unregulated nitrogenous DBPs (N DBPs) which include haloacetonitriles (HANs) and N nitrosodimethylamine (NDMA). These multivariate power models resulted in moderately strong positive corr elations with R 2 values ranging from 0.62 to 0.88 across all DBP species and used a large number of samples in the experimental design, ranging from 134 to 210. Although the Chen and Westerhoff (2010) models demonstrated reasonable accuracy during model va lidation, they only incorporated three independent variables: dissolved organic carbon (DOC), ultraviolet absorbance at 254 nm (UV 254 ), and bromide ion (Br ). For N DBPs a fourth variable, dissolved organic nitrogen (DON), was used to improve model perform ance by accounting for nitrogenous precursors. The methodology used by Chen and Westerhoff (2010) simplified the model development process by creating a uniform set of DBP formation conditions including
20 specified pH, temperature, chlorine dose, and reactio n time (all variables that influence DBP formation), but hinder the application of the models to other disinfection scenarios. For many DBP formation potential studies, models such as Chen and Westerhoff (2010) can be used and may give accurate results und er controlled disinfection conditions; however, the impacts of an authentic drinking water treatment and distribution system cannot be anticipated. For chlorination experiments in the Chen and Westerhoff (2010) study the uniform formation conditions (UFC) specified a 24 h reaction time, but in reality DBPs can form for much longer periods in a distribution system, resulting in an underestimation of DBP formation if the model does not include reaction time as an explanatory variable. Another variable importa nt to full scale systems is temperature which affects reaction rates. For municipalities in the southern U.S., and other parts of the world, it is realistic to expect temperatures to exceed 30 C in the summer months and time periods approaching 7 d in lar ge distribution systems. External factors such as these decrease the predictive capability of DBP models when the models do not include chlorine dose, reaction time, or temperature. Following the publication of empirically based DBP models several resear chers have employed other techniques for deriving more accurate predictions of DBP formation. Researchers have used first principles approaches to devise chemical mechanism based models for DBP formation with the theoretical basis that DBPs are formed unde r kinetically controlled conditions. Nokes et al. (1999) published one of the first kinetic models, deriving a series of expressions that calculates the relative concentration of the four THM species based on the ratio of bromid e ion to chlorine dose. Although these models demonstrated good agreement with experimental data,
21 the models of Nokes et al. (1999) lack common water chemistry inputs th at are typically measured (i.e., DOC and UV 254 ) or chlorination conditions (i.e., tempe rature, pH, chlorine dose, reaction time) and rely only on theoretical expressions of reaction rates and the presence of oxidizing agents (HOCl and HOBr). Li and Zhao (2006) developed a second order reaction equation based on the first order relationship with chlorine d ecay and THM formation and the first order reaction with humic acid precursor and THM formation. Since kinetic models involve variables that are difficult to measure, the predictive capabilities are not robust and can be difficult to apply to an external data set. Although kinetic models are most useful for determining formation mechanisms, empirical rate constants can be difficult to determine considering the heterogeneous nature of NOM. Perhaps one of the most interesting techniques for predicting DBP fo rmation is artificial neural networks explored by Kulkarni and Chellam (2010) Artificial neural networks are a modeling techniq ue capable of handling noisy, distorted, multivariate data by mimicking biological neurons, similar to the human brain. Using a feed forward, back propagation algorithm the model iteratively solves for DBP concentrations while minimizing the sum of squared errors (SSE) to converge to the optimal solution. When comparing model results with experimental data the artificial neural network models achieved linear correlations of R 2 = 0.96 and 0.90 for THM4 formation in raw and treated waters, respectively, and achieved R 2 = 0.92 and 0.71 for HAA6 in raw and treated waters. The Kulkarni and Chellam (2010) study demonstrated that artificial neural networks are useful tools for predicting DBP formation in waters having a wide range of water chemistry but lack the m echanistic understanding of kinetic models and
22 regression models based on fundamental chemistry. Although these models are comprehensive and have the potential to save cost in experimental and pilot test determination of DBP formation, they are not widely used and require proprietary software for model development. In this respect multivariable regression models are easier to use and implement in a wide range of settings when compared to artificial neural network models and kinetic models. To gain valuable insights on the accuracy and applicability of existing DBP models it is necessary to compare all peer reviewed models using a common data set. The key limitation to the DBP modeling literature is the lack of validation with external data sets and a cross comparison of predictability based on statistical metrics. The research objective of this work was to compare DBP models published over the last 30 years using two external data sets on DBP formation and identify the best fitting models using three statis tical metrics 2.3 Experimental Approach In an effort to identify the most accurate DBP models a procedure was developed to compare all existing models using two common data sets and evaluate the accuracy of the models through three statistical metrics. This procedure involved compiling all applicable models and indentifying relevant data sets that include raw water quality measurements as well as concentration and speciation of THMs and HAAs. The raw water data were used as inputs to the models and the measur ed THM/HAA data were used to compare the accuracy of the model output. A description of the model selection process is discussed in section 2. 3 .1 and a summary of the water chemistry for each data set is discussed in section 2. 3 .2.
23 2.3.1 DBP Models A thorough li terature review was conducted to investigate the currently published DBP models and which ones are applicable to this study. Several studies have focused on developing new mechanistic models based on kinetics and chemical reactions between DBP precursors a nd disinfectants; however, some models were not applicable to this study because they use inputs that are not commonly measured in drinking water. For the purposes of this study models were excluded if they involved parameters for fulvic acid, chlorophyll a, fluorescence, geographic region, seasons, kinetic rate constants, flow rate and/or tank volume, or any of several dummy variables. The application of these models is not practical because the number of assumptions necessary for obtaining a solution incr eases model error. It should also be noted that models which predict total DBP concentrations in mol L 1 were excluded because it is not possible to convert to mass based units without knowing the DBP speciation. A complete list of the models considered i n this study is tabulated in Table 2 1. A total of 85 models were selected for this study, although several others were considered in the screening process. Additionally, all models are based on chlorine disinfection because this is the most widely publish ed form of DBP modeling and most widely used disinfectant. Other types of DBP models include chloramination models that predict NDMA formation (Chen & Westerhoff, 2010) as well as dichloroacetonitrile (DCAN) and certain THM and HAA species (Yang et al., 2008) and ozonation models for predicting the formation of bromate (Siddiqui et al., 1998; Sohn et al., 2004) and bromoform (Siddiqui et al., 1994).
24 2.3.2 DBP Formation Data Sets Two sets of data were used to validate the e xisting DBP models. The first data set is from a 2005 study by Boyer and Singer comprising four surface waters that represent a range of DOC and bromide concentrations. DBP formation potential experiments were conducted on all four raw waters as well as for the same waters under various treatment scenarios. The chlorination procedure involved uniform formation conditions (UFC) whic h yielded 1 mg L 1 chlorine residual after 24 h of incubation in the dark at 20 C. The treatment included alum coagulation, magnetic ion exchange (MIEX) treatment, and a combination of alum and MIEX treatment. The primary function of alum is to reduce DOC and turbidity whereas MIEX is capable of reducing levels of DOC and bromide by ion exchange. The second set of data was from a 1993 survey of bromide in drinking water published by the American Water Works Association Research Foundation (Amy et al., 1993 ), which included 100 utility participants from a variety of raw water sources including surface waters, groundwaters, and drinking water reservoirs, resulting in 145 discrete samples. Additionally, the survey spanned four seasons over an 18 month period g enerating four subsets of data that can be used in validating the DBP models. A list of the water chemistry parameters recorded for the two data sets is included in Table 2 2. The data used for DBP model validation for Boyer and Singer (2005) is included a s Supplementary data; the data used for DBP model validation for Amy et al. (1993) can be found in the original report. 2.3.3 Statistical Analysis Model developers refer to the fit of their models based on various statistical metrics. For research papers focusi ng on the development of a new model the researchers may quantify the fit of their model using a linear coefficient of determination
25 (R 2 ) (Al omari et al., 2004; Sohn et al., 2004) standard error (SE) (Chen & Westerhoff, 2010) F test (Obolensky & Singer, 2008) students t test (Rathbun, 1996) analytical variance (Chen & Westerhoff, 2010) and various methods to determine goodness of fit such as Kolmogorov Smirnov (K S) test (Nikolaou et al., 2004) Pearson correlation (Semerjian, Dennis, & Ayoub, 2009) and Chi square test. Determination of the best performing model is difficult because many previous publications report the goodness of fit based on the data set used for model development or use an externa l data set for model validation that may be from a similar water source. Also some authors quantify model performance using metrics that report absolute error, such as sum of squared errors (SSE), instead of relative error which can be compared across dif ferent publications. To gauge the ability of models for predicting DBP formation under a wide range of water chemistry scenarios it is necessary to validate the models using a common data set that contains a range of DBP precursor concentrations and disin fection conditions and using statistical analyses that yield a relative error. For this study model fit and predictive capability were assessed using standard error (SE), on (R 2 modified percent standard deviation that accounts for the number of explanatory variables in a model, and the R 2 value quantifies how well a model predicts measured DB P concentrations. The three statistical metrics provide a concise analysis of model performance by quantifying errors in prediction, imposing a penalty on models that use more input variables, and demonstrating how well predicted versus measured values are linearly correlated. The equations used in the statistical analysis are shown below
26 where n is the number data points, p is the number of variables, x is the measured value, and y is the predicted value. Eq. 2 1 from Chen and Westerhoff (2010), and Eq 2 2 from Marquardt (1963). (2 1) (2 2) (2 3) 2.4 Results 2.4.1 Overview of DBP Models The general results of this study demonstrate the level of accuracy across all models. Models were classif ied as including a term for Br or excluding a term for Br to evaluate the effect of neglecting Br precursor on THM and HAA formation and speciation. It is well known that during chlorine disinfection the presence of Br forms HOBr, a strong halogenating agent, which subsequently results in brominated DBPs (Br DBPs) that are believed to pose a greater health risk than chlorinated analogues (Chang et al., 2001; Duong et al., 2003; Chowdhury et al., 2010) Although models excluding Br may still be accurate for waters containing low concentrations of Br it is
27 posited that models including a term for Br will exhibit a higher degree of accuracy across a wide range of Br concentrations when compared to the models that exclude this term. 2.4.2 Trihalomethanes 220.127.116.11 THM4 The four individual THM species and THM4 were measured in the Boyer and Singer (2005) data set while the Amy et al. (1993) data set only recorded THM4. Results for THMs are illustrated with figures of measured versus predicted concentrations with y = x line showing 1:1 correspondence between predicted and measured values. Figures 2 1 A 2 1 D are for the Boyer and Si nger (2005) data set, showing model results for raw waters and treated waters separately. Figures 2 1 A and 2 1 B are for models that include a term for Br (15 models) whereas Figures 2 1 C and 2 1 D are from the same data set but show results for models excl uding a Br term (18 models). Accompanying Figures 2 1 A 2 1 D are results of the statistical analysis (Table 2 3) on the measured versus predicted THM4 concentrations. Using the Boyer and Singer (2005) data set, the average SE for THM4 models that include Br is 93.4 g L 1 (15 models) and for models not including Br the average SE is 117 g L 1 (16 models). It should also be noted that models 26 and 27 were excluded from the calculation of average SE for models which exclude Br because the error predict ed in these models exceeded 10 8 g L 1 Using the same data set to compare R 2 results, models including Br had an average R 2 value of 0.74 whereas models excluding Br had an average R 2 value of 0.67. Data in all four parts of Figure 2 1 show that the maj ority of DBP models under predict THM4 formation.
28 Based on the lowest SE, the best models for predicting THM4 in Figure 2 1 are models 5, 24, and 38 for models including Br and models 18 and 19 which do not include Br All of these models have SE < 60 g L 1 interpretation of these results suggests that these models are capable of predicting THM4 within 60 g L 1 and exhibit a percent standard deviation of less than 50% across a range of water chemistry inputs. A common feature of all five models, with the exception of model 38, is that they all incorporate key precursors as inputs (TOC/DOC and/or UV 254 ) as well as disinfection conditions (Cl 2 dose, pH, reaction time, and temperature). Model 38, generated by Chen and Westerhoff (2010), includes terms for DOC, UV 254 and Br The formation conditions involved pH, temperature, reaction time, and chlorine dose similar to those used in Boyer and Singer (2005) which likely contributes to the accurate model predictions. Although there is a sp read in the predictive capability of the models in Figure 2 1, the average of the three statistical metrics across all models favors models including Br shown by lower SE, MPSD, and higher R 2 (Tables 2 3 A and 2 3 B ). Also included in the THM4 modeling effo rt was the Amy et al. (1993) data set which used 145 discrete raw water samples to cover over 100 different source waters across the U.S. To illustrate the spread in model prediction Figure 2 2 A and 2 2 B show the measured versus predicted concentrations of THM4 for models including Br term and models not including Br term, respectively. In contrast to Figure 2 1, the data in Figure 2 2 are scattered above and below the y = x line indicating both over prediction and under prediction. Individual plots were also generated for the most accurate models using the same data (Figures 2 3 A 2 3 D ). Three of the four models in Figure 2 3 are
29 models that include Br (models 15, 17, and 34) and one model that did not include Br (model 30). Table 2 4 tabulates the res ults of the statistical analysis for Figure 2 2. Models that include a term for Br performed better in SE and MPSD but slightly worse for R 2 The lower average R 2 is because two of the models (models 31 and 35) have poor individual R 2 values (< 0.03), whe n excluding these models the R 2 increases to 0.81 and surpasses the average R 2 of models excluding Br (R 2 = 0.77). The statistical metrics are lower in Table 2 3 than Table 2 4 because of the smaller sample size. 18.104.22.168 Individual THM s pecies Speciation of THM4 is believed to be dependent on concentration of Br at the point of disinfection. For individual THM models it is necessary to incorporate a term for Br ; all of the species models drawn from the literature contained Br as a variable. The results for this section are presented in two figures for each THM species, separately presenting species concentrations for raw and treated waters from Boyer and Singer (2005). Figure 2 4 A and 2 4 B present chloroform (CHCl 3 ) predictions for raw and treated waters, respec tively. Figures 2 5 A and 2 5 B 2 6 A and 2 6 B and 2 7 A and 2 7 B present results for bromodichloromethane (CHCl 2 Br), dibromochloromethane (CHClBr 2 ), and bromoform (CHBr 3 ) in the same format. The predictive capability of individual THM species models is gene rally less accurate than the THM4 models as shown by the statistical analysis of THM species models in Table 2 5. Looking at the statistical metrics it appears that species models perform comparably to THM4 ; however, individual species occur at lower conce ntrations and should result in lower SE than THM4. The best performing CHCl 3 model is model 45 which has an SE of 14.4 g L 1 a MPSD of 47%, and an R 2 of 0.95. The good performance of this model is largely due to no time input term but the model was devel oped using the same time as the UFC used
30 in the measured values. Other CHCl 3 models suffer in comparison because the time variable is spread over a large interval and the modeling effort was to minimize error over the entire interval. The MPSD and R 2 are s imilar to THM4 models but the SE is lower, as expected, because of the lower concentration. For the other THM species the best models are 48 for CHCl 2 Br (SE = 8.7 g L 1 MPSD = 37%, and R 2 = 0.82) and model 61 for CHBr 3 (SE = 4.1 g L 1 MPSD = 58%, and R 2 = 0.97). Of the four models investigated for CHClBr 2 none yielded accurate results with no models having a SE < 35 g L 1 or an MPSD < 116%. 2.4.3 Haloacetic Acids 22.214.171.124 HAA5, HAA6, and HAA9 The nine HAA species were measured in Boyer and Singer (2005) and further categorized into HAA5 (chloroacetic acid, dichloroacetic acid, trichloroacetic acid, bromoacetic acid, and dibromoacetic acid), HAA6 (HAA5 + bromochloroacetic acid), and HAA9 (HAA6 + bromodichloroacetic acid, dibromochloroacetic acid, and tribromoacetic ac id) (Wu & Chadik, 1998) The U.S. EPA currently regulates HAA5, but the formation and occurrence of the four unregulated HAA species is well documented in published literature (Nikolaou et al., 2004; Sohn et al., 2004) and is critical to consider for the full impact of Br on HAA speciation. For the 15 HAA models reviewed in this article one is for HAA5, ten are for HAA6, and four are for HAA9. Of the 15 models, six do not include a term for Br ; all models are plotted together in Figure 2 8. Similar to the presentation of results for THM4, figures are separate for raw and treated waters. Additionally, models 65 and 66 are not plotted because they predict HAAs > 10 6 g L 1 and model 67 is not plotted since all predictions are essentially 0 g L 1 These models were developed in Greece by Lekkas and Nikolaou (2004) and incorporated a range of
31 pH values, reaction times, bromide concentrations, and chlorine doses in the model formulation; the high error associated with these models is probably due to the logarithmic expression that can yield errors spanning orders of magnitude. Measured versus predicted HAA concen trations are presented in Figures 2 8 A and 2 8 B and the statistical analysis for the models is tabulated in Table 2 6. HAA5 is regulated at an MCL of 60 g L 1 specified in the Stage 1 and Stage 2 D/DBP rule. Models 68/73, 69/74, and 70/75 (raw/treated) m ay be appropriate for predicting HAA formation at a level significant for regulatory purposes. These models all come from a publication by Sohn et al. (2004) and are DOC, UV 254 and DOC + UV 254 models for both raw and treated waters. The only other publica tion that generated a model (model 76) capable of accurately predicting HAAs is Chen and Westerhoff (2010), where the SE associated with this model is 35.8 g L 1 The overall best fitting models for HAA predictions are models 69 and 74 which generated a S E of 13.5 g L 1 and an R 2 of 0.94. 126.96.36.199 Individual HAA s pecies Only two known publications have presented a formulation of individual HAA species models (Montgomery Watson, 1993; Obolensky & Singer, 2008) The Obolensky and Singer (2008) models were not applicable to this study because they include terms for turbidity, alkalinity, and complex disinfection variables including a categorical variable for chlorination point (raw or settled water), chlorine res idual, and chlorine consumed. These variables were not recorded in the Boyer and Singer (2005) data set, and in general are not usually available. The only application of species models for HAAs is using this data set for the Montgomery Watson models. In a ddition to being the only publication with usable HAA species models there are only models for
32 the five regulated HAA species. Figure 2 9 shows the results for the five HAA species models and Table 2 7 tabulates the results from the statistical analysis. The best performing model for HAA species is model 81 (dibromoacetic acid), which had SE of 2.1 g L 1 and R 2 of 0.95. The SE for models 77 (chloroacetic acid) and 80 1 but demonstrate no correlation with measured data h aving R 2 concentrations being in the range of 0 to 9.6 g L 1 Most of the measurements were also below the minimum reporting level of 2 g L 1 which makes it difficult to correlate measured versus predicted concentrations. Models 78 (dichloroacetic acid) and 79 (trichloroacetic acid) have R 2 values of 0.95 but they do not fall along the 1:1 line, underestimating species concentrations as indicated by the SE. 2.5 Discussion Research in the field of DBP formation modeling has far reaching impacts for the management of drinking water sources and human health risk. The U.S. and other countries worldwide have recognized DBPs as a threat to public health, creating a need for new tools that can better h elp manage the occurrence of these harmful byproducts of drinking water disinfection. Considering the extensive research efforts that have been placed on developing models for predicting DBPs in drinking water, it is critical to evaluate the predictive cap ability of the existing models and compare models from numerous publications. Using SE, MPSD (%), and R 2 as statistical metrics for predictive accuracy, the best performing models for THM4 were 5, 15, 17, 18, 19, 24, 30, and 38. Of these models, 18, 19, a nd 30 did not include a term for bromide ion (Br ). The best performing THM4 models are capable of predicting THM4 with an SE less than the regulatory
33 threshold of 80 g L 1 Comparing error to the regulatory limit is useful for knowing the extent to which a model can be used for meeting compliance requirements and how accurately a model can predict below the MCL. For individual THM species the best performing models include 45 (chloroform), 48 (bromodichloromethane), and 61 (bromoform) based on the same st atistical metrics; no model for dibromochloromethane yielded an SE < 35 g L 1 As for HAA models, the three best performing models come from the publication by Sohn et al. (2004) which included models 68/73, 69/74, and 70/75, all for HAA6. The errors asso ciated with these HAA6 models are less than the regulatory MCL of 60 g L 1 for HAA5. The overall best performing model for an individual species of HAA is model 81 for dibromoacetic acid; however, it should be noted that only five of the nine HAA species had models available for to this study. All of the THM and HAA models listed above are in the form of multivariate power law models to DBP formation potential in the mode l development either using laboratory experiments or field data. Power law models proved to generate a better fit to measured data as opposed to linear or logarithmic expressions; however, in some cases logarithmic models can be converted to power law mode ls. A common characteristic of these models is the incorporation of key precursors as inputs (TOC/DOC, UV 254 and/or Br ) as well as operational conditions for disinfection (pH, reaction time, temperature, and chlorine dose). Generally models with more inp results with the exception of models from Chen and Westerhoff (2010), where the models from that study used chlorination under UFC which was similar to the chlorination conditions used in Boyer and Singer (2005) the dat a base used for the
34 model comparisons The statistical analysis in this work shows that it is also necessary to include Br in the model to have accurate results across a variety of waters. For example, models 18 and 19 do not include a term for Br but ar e accurate with respect to SE. However, these models tend to over predict THM4 in low bromide waters and under predict THM4 in high bromide waters. It is recommended that model developers include all key precursors in order to generate robust models that c an handle a wide range of water quality. For future work in DBP modeling researchers should focus on developing power law models that include TOC and/or DOC, UV 254 and Br ; it is also important to formulate the models using a range of disinfection scenari os to ensure the models are robust to predict DBPs over a range of chlorination conditions. Other important factors include using a range of concentrations for organic precursors ( as measured by TOC DOC and UV 254 ) and inorganic precursors (Br ). To prevent limitations on model application it is necessary that the model, or perhaps set of models, can capture DBP formation potential in sources containing variable levels of precursors. Some of the most accurate models investigated in this study are specified f or either raw or treated waters. Model developers should consider this when formulating new models since treatment processes, e.g., coagulation, selectively remove portions of NOM that are more reactive, aromatic, or likely to form halogenated byproducts. In addition to including all necessary explanatory variables it is also important that researchers focus on filling gaps in the modeling literature. Based on a thorough literature review it is evident that THM4 models are abundant but there is a need for m ore DBP species models, HAA5 (or HAA6 and HAA9) models, bromate models, or models for major
35 unregulated DBPs such as NDMA, HANs, halonitromethanes, and iodinated species of THMs and HAAs. Finally, model developers and DBP researchers should place an emphas is on reporting all experimental methods and water quality data in their publications or supplementary electronic data. For example, see PANGAEA (http://www.pangaea.de/) as option for open access archiving of primary data. Tabulating and making data availa ble will allow researchers to use previous data sets for model development, calibration, or validation in future studies; larger data sets used for model formulation will improve overall model performance. Improved models can be useful tools for DBP manag ement at the municipality scale or can be used to make preliminary assessments for new water management practices. An example scenario where DBP models can be useful for predicting changes in THM or HAA formation is in drinking water sources that experienc e sudden or progressive increases in bromide concentration due to saltwater intrusion. Coastal rivers can exhibit higher levels of salinity during periods of drought or low flow resulting in elevated levels of bromide far upstream (Walker and Boyer, 2011). For utilities on the coast that use groundwater for their potable water supply, saltwater intrusion could elevate levels of bromide that may cause shifts in speciation or increases in THM4 or HAA5 and result in compliance issues. A similar phenomenon may occur in areas where high total dissolved solids (TDS) fracking flowback waters are discharged to surface waters or infiltrate into shallow groundwater (Warner et al., 2012) The flowback waters have salinity levels higher than that of seawater and even relatively small volumes coul d create problems with brominated species of DBPs if they enter into a drinking water source. The best fitting models from this study suggest that THM4 can
36 double when bromide concentration increases by one order of magnitude higher and all other variables remain constant. A model capable of accurately predicting DBPs over a range of bromide concentrations can help manage the formation of regulated DBPs and minimize human health risk to brominated DBPs in particular. It is becoming common practice that muni cipalities search for alternative water supplies as supplies of high quality fresh water diminish (Boyer et al., 2012); accordingly, DBP models can be useful for predicting the effects of changing water supplies or treatment processes. The potential uses o f these models not only include operational purposes for utility managers, but can be coupled with epidemiological studies to assess human health risk and exposure, or used by regulatory agencies to balance DBP risk with economic cost of upgrading treatmen t and removal of precursors. For instance, Black et al. (1996) used various raw water and treated water quality to model THM formation, which in turn was used to model cancer risk. As another example, Wang et al. (2007) used THM and HAA data collected from full scale water plants to evaluate the lifetime cancer risk from DBP exposure; additional scenarios for DBP exposure could be investigated by combining risk models with DBP models. 2.6 Summary of the Critical Evaluation of DBP Models Over the past 30 years researchers have placed the most emphasis on developing THM4 models and have neglected HAAs which are of equal importance from a regulatory perspective and even less emphasis has been placed on individual THM and HAA species modeling. The best performing models from this study were capable of predicting THM4 with accuracy of approx. 0.5MCL and for HAA5 with accuracy of approx.
37 0.25MCL while yielding a MPSD at approx. 30% for both THM4 and HAA5. With this level of accuracy, the models have the potential t o be calibrated to individual water sources and used as a tool for managing water sources, considering changes in treatment processes, regulatory compliance, or health risk assessment. Several models were also capable of generating R 2 values > 0.90 but may have high errors in their prediction accuracy. Nevertheless, models with high R 2 can be used to predict behavioral trends across a data set but perhaps with a fixed error. Multivariate power law models generated the best predictions for THM4, HAAs, and i ndividual species. Additionally, models that included at least five of the seven explanatory variables (DOC, UV 254 Br pH, chlorine dose, reaction time, and temperature) resulted in lower SE and MPSD. The best models were also developed with a broad rang e of precursor concentrations and chlorination conditions. A next step in evaluating the model would be to assess the weight of the explanatory variables contributing to DBP formation. Models including a term for Br generally performed better than model s that did not. Although models 18, 19, and 30 (which did not include Br ) were able to predict THM4 with errors less than the MCL in the Boyer and Singer (2005) data set, the error increased by nearly an order of magnitude in the Amy et al. (1993) data se t because of over estimating THM4 in low bromide waters. Additionally, models that did not include Br tended to underestimate DBP formation in high bromide waters (models 6 14).
38 Table 2 1. Published DBP models for THM4, HAAs, and individual THM and HAA sp ecies. Models unable to be evaluated in this study as a result of insufficient data are indicated by *. Models Units Source THM Models 1* THMs = 3.91 + (Br ) 0.15 + 0.23log(Cl 2 ) + 0. 24(pH) + 10 0.009T + 0.26(NVTOC) mol/L Morrow & Minear (1983) (conditions @ t = 96 h) 2 THMs = 0.00082(pH 2.8)TOC(Cl 2 ) 0.25 (t) 0.36 g/L Urano et al. (1983) 3* THMs = 0.0031(UV 254 .TOC) 0.440 (Cl 2 ) 0.409 (t) 0.265 (T) 1.06 (pH 2.6) 0.715 (Br +1) 0.0358 mol/L Amy et al. (1987) 4* TTHM = 0.00325(t) 0.252 (Cl 2 ) 0.517 (TOC.UV 254 ) 0.378 (T) 0.985 (pH 2.6) 0.64 (Br +1) 0.0536 mol/L Amy et al. (1987) 5 THMs = 7.21(TOC) 0.004 (UV254) 0.534 (Cl 2 7.6NH 3 ) 0.224 (t) 0.255 (Br +1) 2.01 (T) 0.480 (pH 2.6) 0.719 g/L Malcolm Pirnie Inc. (1993) 6 THMs = 12.72(TOC) 0.291 (t) 0.271 (Cl 2 ) 0.072 g /L Chang et al. (1996) 7 THMs = 108.8(TOC) 0.2466 (t) 0.2956 (UV 254 ) 0.9919 (Cl 2 ) 0.126 g/L Chang et al. (19 96) 8 THMs = 131.75(t) 0.2931 (UV 254 ) 1.075 (Cl 2 ) 0.1064 g/L Chang et al. (1996) 9 THMs = 14.6(pH 3.8) 1.01 (Cl 2 ) 0.206 (UV 254 ) 0.849 (t) 0.306 g/L Rathbun (1996) 10 THMs = 0.00412(DOC) 1.10 (Cl 2 ) 0.152 (Br ) 0.068 (T) 0.61 (pH) 1.60 (t) 0.260 g/L Amy et al. (1998) 11 THMs = 0.044(DOC) 1.030 (t) 0.262 (pH) 1.149 (Cl 2 ) 0.277 (T) 0.968 g/L Rodriguez et al. (2000) 12 THMs = 1.392(DOC) 1.092 (pH) 0.531 (T) 0.255 g/L 13 THMs = 16.9 + 16.0(TOC) + 3.319(Cl 2 ) 1.135(T) + 1.139(t) g/L Srodes, et al. (2003)
39 Table 2 1. continued Models Units Source 14 THMs = 21.2 + 2.447(Cl 2 ) + 0.499(t) g/L S rodes et al. (2003) 15 TTHM = 10 1.385 (DOC) 1.098 (Cl 2 ) 0.152 (Br ) 0.068 (T) 0.609 (pH) 1.601 (t) 0.263 g/L Sohn et al. (2004) (DOC based model for raw waters) 16 TTHM = 0.42(UV 254 ) 0.482 (Cl 2 ) 0.339 (Br ) 0.023 (T) 0.617 (pH) 1.609 (t) 0.261 g/L Sohn et al. (2004) (UV 254 based model for raw waters) 17 TTHM = 0.283(DOC.UV 254 ) 0.421 (Cl 2 ) 0.145 (Br ) 0.041 (T) 0.614 (pH) 1.606 (t) 0.261 g/L Sohn et al. (2004) (UV 2 54 and DOC based model for raw waters) 18 TTHM = 0.00253(DOC) 1.22 (Cl 2 ) 0.442 (T) 1.34 (pH) 1.75 (t) 0.34 g/L Sohn et al. (2004) (short term DOC based model from 1987 EPA database) 19 TTHM = 0.012(DOC.UV 254 ) 0.47 (Cl 2 ) 0.48 (T) 1.10 (pH) 2.38 (t) 0. 35 g/L Sohn et al. (2004) (short term DOC and UV 254 model based on 1987 EPA database) 20 TTHM = 3.296(DOC) 0.801 (Cl 2 ) 0.261 (Br ) 0.223 (t) 0.264 g/L Sohn et al. (2004) (DOC based model for coagulated waters) 21 TTHM = 75.7(UV 254 ) 0. 593 (Cl 2 ) 0.332 (Br ) 0.060 (t) 0.264 g/L Sohn et al. (2004) (UV 254 based model for coagulated waters) 22 TTHM = 23.9(DOC.UV 254 ) 0.403 (Cl 2 ) 0.225 (Br ) 0.141 (t) 0.264 g/L Sohn et al. (2004) (DOC and UV 254 based model for coagulated waters)
40 Table 2 1. continued Models Units Source 23 TTHM = (TTHM @ pH = 7.5, T = 20 C )1.156 (pH 7.5) 1.0263 (T 20) g/L Sohn et al. (2004) (temperature and pH correction for models 20 22) 24 THMs = 4.527(t) 0.127 (Cl 2 ) 0.595 ( TOC) 0.596 (Br ) 0.103 (pH) 0.66 g/L Al omari et al. (2004) 25 log(THMs) = 0.33(pH) 0.02(pH) 2 + 0.12(t) 0.04(t) 2 g/L Nikolaou et al. (2004) 26 log(THMs) = 0.44(pH) + 7.53log(pH) 1.10(Cl 2 ) + 0.20(Cl 2 ) 2 g/L 27 log(THMs) = 1.546 + 0.631(pH) 2 + 0.569log(t) + 0.385log(Cl 2 ) g/L Lekkas & Nikolaou (2004) 28 THMs = 0.0001(Cl 2 ) 3.14 (pH) 1.56 (TOC) 0.69 (t) 0.175 g/L Kolla (2004) 29 THMs = 0.0707(TOC + 3.2) 1.314 (pH 4.0) 1.496 (Cl 2 2.5) 0.197 (T + 10) 0.724 g/L Uyak et al. (2005) 30 log(THMs) = 1.078 + 0.398log(TOC) + 0.158log(T) + 0.702log(Cl 2 ) g/L Toroz & Uyak (2005) (models 29 and 30 were developed from samples collected at various locations in a drinking water distribution system with varying reaction time, t) 31 TTHMs = 10 1.375 (t) 0.258 (Cl 2 /DOC) 0.194 (pH) 1.695 (T) 0.507 (Br ) 0.218 g/L Hong et al. (2007) 32 THM = 10 0.038 (Cl 2 ) 0.654 (pH) 1.322 (t) 0.174 (SUVA) 0.712 g/L Uyak et al. (2007) 33 THMs 2 = 471.11 + 0.48t 2 + 1856.07(Br ) 2 + 404.38(Cl 2 ) 2 g/L Semerjian et al. (2008) (models 33 36) 34 THM = 12.66 + 0.48(t) + 0.3506(Br ) + 10.26(Cl 2 ) g/L 35 ln(THM) = 6.11 + 0.211ln(t) + 1.64ln(Br ) + 0.34ln(Cl 2 ) 0.80ln(T) g/L
41 Table 2 1. continued Models Units Source 36 ln(THM) = 1.43 + 0.02(t) 0.04(T) + 2.78(Br ) + 0.48(Cl 2 ) g/L (models 33 36 are for laboratory simulated distribut ion system (SDS) experiments where Br is in mg L 1 ) 37* log(THM4) = 1.371 + 0.015(T) 0.0005(Alk) + 0.188log(TOC) + 0.326log(UV 254 ) + 0.291log(Cl 2 ) + 0.119log(t) + 0.087(pH) + 0.167log(Cl 2,res ) g/L Obolensky & Singer (2008) 38 THM4 = 1147(DOC) 0.00 (UV 254 ) 0.83 (Br +1) 0.27 g/L Chen & Westerhoff (2010) (conditions @ t = 24 h, T = 25 C, pH = 8.2 ) CHCl 3 Models 39 CHCl 3 = 0.078(TOC.UV 254 ) 0.616 (Cl 2 ) 0.391 (t) 0.265 (T) 1 .15 (pH 2.6) 0.8 (Br +1) 2.23 g/L Malcomn Pirnie Inc. (1992) 40 CHCl 3 = 0.997(TOC) 0.58 (UV 254 ) 0.58 (Cl 2 ) 0.814 (t) 0.278 (Br +1) 4.27 (T) 0.569 (pH 2.6) 0.759 g/L Malcomn Pirnie Inc. (1993) 41 CHCl 3 = 0.064(TOC) 0.329 (UV 254 ) 0.854 (Br + 0.0 1) 0.404 (pH) 1.161 (Cl 2 ) 0.561 (t) 0.269 (T) 1.018 g/L Montgomery Watson (1993) 42 CHCl 3 = 0.442(pH) 2 (Cl 2 ) 0.229 (DOC) 0.912 (Br ) 0.116 g/L Rathbun (1996) (conditions @ T = 25 C and t = 7 d) 43 CHCl 3 = 10 0.748 (t) 0.21 (Cl 2 /DOC) 0.221 (p H) 1.374 (T) 0.532 (Br ) 0.184 g/L Hong et al. (2007)
42 Table 2 1. continued Models Units Source 44* log(CHCl 3 ) = 1.935 + 0.017(T) 0.0012(Alk) + 0.1993log(TOC) + 0.4450log(UV 254 ) + 0.3824log(Cl 2 ) + 0.0921log(t) + 0.1133(pH) g/L Obolensky & Singer (2008) 45 CHCl 3 = 1805(DOC) 0.11 (UV 254 ) 1.22 (Br +1) 2.19 g/L Chen & Westerhoff (2010) (conditions @ t = 24 h) BDCM Models 46 BDCM = 0.863(TOC.UV 254 ) 0.177 (Cl 2 ) 0.309 (t) 0.271 (T) 0.72 (pH 2.6) 0.925 (Br +1) 0.722 g/L Malcolm Pirnie Inc. (1992) 47 BDCM = 4.05(TOC) 0.567 (UV 254 ) 0.567 (Cl 2 7.6NH 3 ) 0.351 (t) 0.366 (Br ) 0.291 (T) 0.568 (pH 2.6) 0.568 g/L Malcolm Pirnie Inc. (1993) 48 BDCM = 0.0098(Br ) 0.181 (pH) 2.55 (Cl 2 ) 0.497 (t) 0.256 (T) 0.519 (for Cl 2 /Br <75) BDCM = 1.325(TOC) 0.725 (Br ) 0.794 (Cl 2 ) 0.632 (t) 0.204 (T) 1.441 (for Cl 2 /Br >75) g/L Montgomery Watson (1993) 49 BDCM = 17.5(pH) 1.01 (Cl 2 ) 0.0367 (DOC) 0.228 (Br ) 0.513 g/L Rathbun (1996) (conditions @ T = 25 C and t = 7 d) 50 BDCM = 10 3.201 (t) 0.297 (pH) 2.878 (T) 0.414 (Br ) 0.371 g/L Hong et al. (2007) 51* log(BDCM) = 0.984 + 0.0483log(turb) + 0.4723log(Br ) + 0.0088(T) + 0.4013log(TOC) + 0.2493log(Cl 2 ) g/L Obolensky & Sin ger (2008) 52 BDCM = 3.6 2.43(DOC) + 3.47(Cl 2 ) pH + 0.58(T) + 0.167(t) + 0.144(Br ) + 0.0012(t 19.93)(Br 88.18) g/L Chowdhury (2009)
43 Table 2 1. continued Models Units Source 53 BDCM = 137(DOC) 0.16 (UV 254 ) 0.94 (Br +1) 3.66 g/L Chen & Westerhoff (2010) DBCM Models 54 DBCM = 2.57(UV 254 /TOC) 0.184 (Cl 2 ) 0.046 (t) 0.252 (T) 0.57 (pH 2.6) 1.35 (Br +1) 2.08 g/L Malcom Pirnie Inc. (1992) 55 DBCM = 22.9(TOC) 0.253 (UV 254 ) 0.253 (Cl 2 7.6NH 3 ) 0.352 (t) 0.292 (Br ) 1.04 (T) 0.491 (pH 2.6) 0.325 g/L Malcom Pirnie Inc. (1993) 56 DBCM = 14.998(TOC) 1.665 (Br ) 1.241 (Cl 2 ) 0.729 (t) 0.261 (T) 0.989 (for Cl 2 /Br < 50) DBCM = 0.028(UV 254 ) 1.175 (TOC) 1.078 (Br ) 1.573 (pH) 1.956 (Cl 2 ) 1.072 (t) 0.2 (T) 0.596 (for Cl 2 /Br > 50) g/L Montgomery Watson (1993) 57 DBCM = 26.6(pH) 1.80 (Cl 2 ) 0.0928 (DOC) 0.758 (Br ) 1.2 g/L Rathbun (1996) 58* log(DBCM) = 1.247 + 0.0501log(turb) + 0.9070log(Br ) + 0.0073(T) + 0.0009(Alk) + 0.1624log(TOC) 0.2569log(UV 254 ) + 0.1174log(t) g/L Obolensky & Singer (2008) CHBr 3 Models 59 CHBr 3 = 1.28(TOC) 0.167 (UV 254 ) 0.167 (Cl 2 7.6NH 3 ) 2.22 (t) 0.294 (Br ) 1.48 (T) 0.553 (pH 2.6) 0.198 g/L Malcom Pirnie Inc. (1993) 60 CHBr 3 = 6.533(TOC) 2.031 (Br ) 1.388 (pH) 1.60 3 (Cl 2 ) 1.057 (t) 0.136 g/L Montgomery Watson (1993)
44 Table 2 1. continued Models Units Source 61 CHBr 3 = 0.29(pH) 3.51 (Cl 2 ) 0.347 (DOC) 0.330 (Br ) 1.84 g/L Rathbun (1996) HAA Models 62 HAA6 = 2.72(TOC) 0.653 (Cl 2 ) 0.458 (t) 0.295 g/L Serodes et al. (2003) 63 HAA6 = 1.33(TOC) 2.612 (Cl 2 ) 0.102 (T) 0.255 (t) 0.102 g/L 64 HAA6 = 8.202 + 4.869(TOC) + 1.053(Cl 2 ) + 0.364(t) g/L 65 log(HAA9) = 0.33(pH) 0.02(pH) 2 + 0.48(t) + 0.09(Cl 2 ) g/L Nikolaou et al. (2004) 66 log(HAA9) = 0.98log(pH) + 1.10log(t) 0.01(t)(Cl 2 ) + 1.59(Cl 2 ) g/L 67 log(HAA9) = 0.00189 1.7(pH) 2 + 1.5log(pH) 0.9(Br ) + 0.875(pH)(Br ) + 0.710log(t) 0.28(pH)(t) + 0.215log(Cl 2 ) g/L Lekkas and Nikolaou (2004) 68 HAA6 = 9.98(DOC) 0.935 (Cl 2 ) 0.443 (Br ) 0.031 (T) 0.387 (pH) 0.655 (t) 0.178 g/L Sohn et al. (2004) (DOC based model for raw waters) 69 HAA6 = 171.4(UV 254 ) 0.584 (Cl 2 ) 0.398 (Br ) 0.091 )(T) 0.396 (pH) 0.645 (t) 0.178 g/L Sohn et al. (2004) (UV 254 based model for raw waters) 70 HAA6 = 101.2(UV 254 .DOC) 0.452 (Cl 2 ) 0.194 (Br ) 0.0698 (T) 0.346 (pH) 0.623 (t) 0.180 g/L Sohn et al. (2004) (DOC and UV 254 based model for raw waters) 71 HAA6 = 5.22( DOC) 0.585 (Cl 2 ) 0.565 (Br ) 0.031 (t) 0.153 g/L Sohn et al. (2004) (DOC based model for coagulated waters)
45 Table 2 1. continued Models Units Source 72 HAA6 = 63.7(UV 254 ) 0.419 (Cl 2 ) 0.640 (Br ) 0.066 (t) 0.161 g/L Sohn et al. (2004) (UV 254 based model for coagulated waters) 73 HAA6 = 30.7(DOC.UV 254 ) 0.302 (Cl 2 ) 0.541 (Br ) 0.012 (t) 0.161 g/L Sohn et al. (2004) (DOC and UV 254 based model for coagulated waters) 74 HAA6 = (HAA6 @ pH = 7.5, T = 20 C)0.932 (pH 7.5) 1.021 (T 20) g/L Sohn et al. (2004) (temperature and pH correction for models 18 20) 75 HAA5 = 10 0.874 (Cl 2 ) 0.351 (pH) 1.248 (t) 0.172 (SUVA) 0.469 g/L Uyak et al. (2007) 76 HAA9 = 1151(DOC) 0.17 (UV 254 ) 0.89 (Br + 1) 0.60 g/L Chen & Westerhoff (2010) Individual HAA Species Models 77 MCAA = 1.634(TOC) 0.753 (pH) 1.124 (Cl 2 ) 0.509 (Br + 0.01) 0.085 (t) 0.300 g/L Montgomery Watson (1993) 78 DCAA = 0.605(TOC) 0.291 ( T) 0.665 (Cl 2 ) 0.480 (Br + 0.01) 0.568 (UV 254 ) 0.726 (t) 0.239 g/L Montgomery Watson (1993) 79 TCAA = 87.182(TOC) 0.355 (pH) 1.732 (Cl 2 ) 0.881 (Br + 0.01) 0.679 (UV 254 ) 0.901 (t) 0.264 g/L Montgomery Watson (1993) 80 MBAA = 0.176(TOC) 1. 664 (pH) 0.927 (T) 0.450 (Br ) 0.795 (UV 254 ) 0.624 (t) 0.145 g/L Montgomery Watson (1993)
46 Table 2 1. continued Models Units Source 81 DBAA = 84.945(TOC) 0.602 (T) 0.657 (Cl 2 ) 0.200 (Br ) 1.073 (UV 254 ) 0.651 (t) 0.120 g/L Montgom ery Watson (1993) 82* log(DCAA) = 1.284 0.2779log(Br ) + 0.0069(T) 0.0009(Alk) + 0.2864log(TOC) + 0.3189log(UV 254 ) + 0.3898log(Cl 2 ) 0.1010(Cl 2,pnt ) g/L Obolensky & Singer (2008) 83* log(BCAA) = 1.456 + 0.0472log(turb) + 0.4083l og(Br ) + 0.0054(T) + 0.0007(Alk) + 0.2820log(TOC) + 0.1710log(Cl 2 ) + 0.1890log(Cl 2,res ) 0.1010(Cl 2,pnt ) g/L Obolensky & Singer (2008) 84* log(TCAA) = 0.653 0.0681log(turb) 0.2814log(Br ) 0.0015(Alk) + 0.1645log(TOC) + 0.4376log(UV 254 ) + 0.5563log(Cl 2 ) + 0.0673log(t) 0.0775(pH) + 0.1732log(Cl 2,res ) g/L Obolensky & Singer (2008) 85* log(BDCAA) = 0.532 + 0.3224log(Br ) + 0.3320log(TOC) + 0.1966log(Cl 2 ) + 0.0526log(t) 0.1247(pH) + 0.1987log(Cl 2,res ) g/L Obolensky & Sing er (2008) THMs = trihalomethanes (g/L or mol/L); TTHMs = total trihalomethanes; THM4 = summation of CHCl 3 (chloroform), CHBr 3 (bromoform), DBCM (dibromochloromethane), and BDCM (bromodichloromethane); HAAs = haloacetic acids (HAA9, HAA6, HAA5 equ al summation of individual HAA species); MCAA = monochloroacetic acid; DCAA = dichloroacetic acid; TCAA = trichloroacetic acid; MBAA = monobromoacetic acid; DBAA = dibromoacetic acid; BCAA = bromochloroacetic acid; BDCAA = bromodichloroacetic acid; DBCAA = dibromochloroacetic acid; TBAA = tribromoacetic acid; Br = bromide ion concentration (mg/L); Cl 2 = applied chlorine dose (mg/L as Cl 2 ); pH = pH in standard units; T = temperature (C); NVTOC = non volatile total organic carbon (mg/L); TOC = total organi c carbon (mg/L), DOC = dissolved organic carbon (mg/L); IC = inorganic carbon (mg/L); t = reaction time (hours); t m = reaction time (minutes); Alk = alkalinity (mg/L as CaCO 3 ); UV 254 = ultraviolet absorbance at 254 nm wavelength; NH 3 = ammonia (mg/L); turb = turbidity (NTUs); Cl 2,res = chlorine residual (mg/L as Cl 2 ); Cl 2,pnt = categorical variable either 1 or 0 for raw or settled water.
47 Table 2 2. Summary of water quality reported for validation data sets. Parameter Units Amy et al. (1993) Boyer and Singer (2005) Br (g L 1 ) 2.5 446 43 540 Cl (mg L 1 ) 0.2 158 DOC (mg L 1 ) 0.112 17.94 0.8 5.1 TOC (mg L 1 ) 1.0 5.5 UV 254 (cm 1 ) 0.003 0.533 0.02 0.193 SUVA 254 (L mg 1 m 1 ) 1.4 3.8 pH 7 8.1 8.3 temperature (C) 20 20 reaction time (hours) 96 24 Cl 2 dose (mg L 1 ) 0.336 53.8 3 9 alkalinity (mg L 1 as CaCO 3 ) 0 338.7 hardness (mg L 1 as CaCO 3 ) 10 610 NH 3 N (mg L 1 ) 0 1.82 THM4 (g L 1 ) 0.93 1006 39.5 294 % CHCl3 (%) 0 98.29 Individual THM species (g L 1 ) <1 262 HAA9 (g L 1 ) 32.6 224 Individual HAA species (g L 1 ) <2 87.2
48 Table 2 3. Descriptive s tatistics for THM4 models from Boyer and Singer (2005) A ) M odels including term for Br B ) M odels not including term for Br Models 15/20, 16/21, 17/22 are grouped as raw/treated water models for a DOC, UV 254 and DOC+UV 254 based models, respectively. The values for SE, MPSD, and R 2 are the average of 16 data points (12 treated waters and 4 r aw waters) A ) Model SE (g L 1 ) MPSD (%) R 2 5 47.7 36% 0.93 10 159 120% 0.97 15/20 67.4 58% 0.98 16/21 85.6 56% 0.71 17/22 77.2 61% 0.92 24 57.1 50% 0.95 31 122 75% 0.24 33 84.6 45% 0.59 34 125 73% 0.56 35 151 107% 0.50 36 88.3 74% 0.79 38 54 .8 31% 0.69 Average 93.4 65% 0.74 B ) Model SE (g L 1 ) MPSD (%) R 2 2 169 115% 0.92 6 137 80% 0.96 7 136 95% 0.62 8 146 96% 0.58 9 148 100% 0.61 11 74.3 58% 0.94 12 140 93% 0.96 13 90.1 52% 0.93 14 130 69% 0.56 18 51.3 50% 0.90 19 53.5 35% 0.73 25 169 107% 0.05 28 166 114% 0.38 29 108 62% 0.73 30 86.9 49% 0.79 32 74.6 89% 0.18 Average 117 79% 0.68
49 Table 2 4. Descriptive statistics for THM4 models from Amy et al. (1993) A ) M odels including term for Br B ) models not including a term for Br The values for SE, MPSD, and R 2 are the average of 145 data points. A ) Models a SE (g L 1 ) MPSD (%) R 2 5 76.8 174% 0.76 10 186 94% 0.79 15 75.3 93% 0.79 16 68.2 169% 0.84 17 68.8 105% 0.85 24 328 293% 0.79 31 171 354% 0.02 33 83.8 672% 0.82 34 74.7 455% 0.82 35 200 97% 0.03 38 69.4 181% 0.78 Average 127 244% 0.66 a model 36 excluded since SE and MPSD exceed 10 6 B ) Models a SE (g L 1 ) MPSD (%) R 2 2 205 101% 0.80 6 166 326% 0.67 7 107 68% 0.82 8 143 77% 0.82 9 167 82% 0.83 11 107 11 2% 0.79 12 166 85% 0.82 13 84.5 1141% 0.82 14 134 707% 0.82 18 432 129% 0.73 19 322 131% 0.82 28 932 158% 0.44 30 64.5 158% 0.82 32 121 496% 0.72 Average 272 262% 0.77 a models 25, 26, 27, and 29 were excluded since SE and MPSD were either unable to be calculated or exceeded 10 8
50 Table 2 5. Descriptive statistics for individual THM species models. The values for SE, MPSD, and R 2 are the average of 16 data points (12 treated waters and 4 raw waters) from Boyer and Singer (2005). CHCl 3 Models SE ( g L 1 ) MPSD (%) R 2 39 67.4 104% 0.94 40 23.7 54% 0.96 41 75.1 111% 0.36 42 161 1009% 0.16 43 72.2 72% 0.03 45 14.4 47% 0.95 CHCl 2 Br Models 46 82.9 389% 0.50 47 13.0 60% 0.70 48 8.7 37% 0.82 49 55.8 167% 0.79 50 20.4 58% 0.70 52 21.2 60% 0.31 53 16.2 60% 0.77 CHClBr 2 Models 54 902 7295% 0.92 55 55.3 116% 0.95 56 35.8 200% 0.47 57 40.2 156% 0.88 CHBr 3 Models 59 28.5 131% 0.95 60 21.0 636% 0.31 61 4.1 58% 0.97
51 Table 2 6. Descriptive statistics for HAA m odels. Models 68/73, 69/74, and 70/75 are grouped as raw/treated water models for DOC, UV 254 and DOC+UV 254 based models, respectively. The values for SE, MPSD, and R 2 are the average of 16 data points (12 treated waters and 4 raw waters) from Boyer and Si nger (2005). Models SE (g L 1 ) MPSD (%) R 2 62 44.7 58% 0.68 63 103 166% 0.59 64 52.7 72% 0.59 65 2.2E+13 3.4E+11 0.79 66 9.2E+13 4.5E+11 0.59 67 93.9 115% 68/73 23.3 53% 0.68 69/74 13.5 30% 0.94 70/75 16.4 34% 0.86 75 65.6 109% 0.70 76 35.8 9 6% 0.93 Average 49.9 81% 0.73
52 Table 2 7. Descriptive statistics for individual HAA species models. The values for SE, MPSD, and R 2 are the average of 16 data points (12 treated waters and 4 raw waters) from Boyer and Singer (2005 ). Models SE (g L 1 ) MPSD (%) R 2 77 2.3 56% 0.06 78 9.9 53% 0.95 79 6.1 67% 0.95 80 2.1 96% 0.06 81 2.1 50% 0.95
53 Figure 2 1. Measured versus predicted THM4 concentrations from Boyer and Singer (2005) A ) R aw waters with models including Br B ) T reated waters with models including Br C ) R aw waters with models not including Br D) T reated waters with models not including Br n is number of individual waters per model. D ) C ) B ) A )
54 Figure 2 2. Measured versus predicted THM4 concentrations from Amy et al. (1993) A ) M odels including Br B ) M odels not including Br n is number of individual waters per model. B ) A )
55 Figure 2 3. Measured versus predicted THM4 concentrations for best fitting models fr om Amy et al. (1993) A) M odel 15 B ) M odel 17 C ) M odel 30 D ) M odel 34. n is number of individual waters per model. C ) D ) B ) A )
56 Figure 2 4. Measured versus predicted concentrations of chloroform (CHCl 3 ) A ) R aw waters and B ) T reated waters. Measured data from Boyer and Singer (2005). n is number of individual waters per model. B ) A )
57 Figure 2 5. Measured versus p redicted concentrations of bromodichloromethane (CHCl 2 Br) A ) R aw waters B ) T reated waters. Measured data from Boyer and Singer (2005). n is number of individual waters per model. A ) B )
58 Figure 2 6. Measured versus predicted concentrations of dibromochloromethane (CHClBr 2 ) A ) R aw waters B ) T reated waters. Measured data from Boyer and Singer (2005). n is number of individual waters per model. A ) B )
59 Figure 2 7. Measured versus predicted concentrations of bromoform (CHBr 3 ) A ) R aw waters B ) T reated waters. Model 59 does not appear in the figure because predicted values are <1 g L 1 Measu red data from Boyer and Singer (2005). n is number of individual waters per model. A ) B )
60 Figure 2 8. Measured versus predicted concentrations of HAAs A) R aw waters B ) T re ated waters. The only HAA5 model is 75; HAA6 models include 62, 63, 64, 68, 69 and 70; and HAA9 models include 65, 66, 67 and 76. Models 65 and 66 are not shown in figures because the predictions exceed 10 6 g L 1 and model 67 is excluded because all predi ctions are ~0 g L 1 Measured data from Boyer and Singer (2005). n is number of individual waters per model. A ) B )
61 Figure 2 9. Measured versus predicted concentrations of individual HAA species for both raw and treated waters. Model 77 is chloroacetic acid, model 78 is dichloroacetic acid, model 79 is trichloroacetic acid, model 80 is bro moacetic acid, and model 81 is dibromoacetic acid. Measured data from Boyer and Singer (2005). n is number of individual waters per model.
62 CHAPTER 3 EFFECT OF SALTWATER INTRUSION ON BROMINE SPECIATION OF TRIHALOMETHANES AND HALOACETIC ACIDS 3.1 Overview of Saltwater I ntrusion Impacts The chlorination of drinking water was one of the major advancements in public health in the early 20 th century, thereby providing a mechanism to inactivate pathogenic microorganisms and create a residual disinfectant throughout public wa ter systems. A byproducts (DBPs) when natural organic matter is present (Rook et al., 1974). Recent research has focused on studying the formation of the DBPs in waters enri ched in precursors such as dissolved organic carbon (DOC) and halide ions like bromide and iodide that originate from seawater, connate groundwater, or anthropogenic waste streams (Richardson et al., 1999; Krasner et al., 2006) Additionally, the use of high quality freshwater for drinking water sources is becoming scarcer as global populations continue to grow and refractory pollutants infiltrate water bodies. With approximately 40% of the worlds' popu lation living within 100 km of the coast it is becoming increasingly necessary to utilize coastal groundwaters and brackish surface waters for drinking water supply (Small & Nicholls, 2003) Furthermore, the combined effects of growing populations, increased pumping and urban development are stressing water supplies and resulting in saltwater intrusion, elevating salinity levels in freshwater aquifers. It is well documented that saltwater intrusion can compromise drinking water sources by increasing levels of total dissolved solids (TDS) and imparting a salty taste in finished waters (Terrazas, 1995; Murgulet & Tick, 2007) Although the majority of previous work studying the thr eats of saltwater intrusion has focused on the landward movement of the freshwater/saltwater interface (Darnault & Godinez, 2008; Park & Aral,
63 2008; Motz & Sedighi, 2013) few studies to date have investigated the impacts of saltwater intrusion on DBP formation and speciation (Krasner et al., 1993; Kampioti & Stephanou, 2002) Previously, researchers have quantified formation of bromate in ozonated drin king waters high in bromide (Krasner et al., 1993) or the formation of various classes of DBPs in waters from different geographic regions, some of which may be near the coast (Kampioti & Stephanou, 2002). The work presented in this manuscript is novel sin ce no other study has isolated a fresh groundwater near the coast and simulated saltwater intrusion in a laboratory setting to quantify the effects of saltwater intrusion on bromide concentrations and the formation of regulated (THM4 and HAA5) and unregu lated (HAA9) classes of DBPs. Saltwater intrusion can occur in both coastal groundwaters and tidally influenced surface waters in low lying areas. The infiltration of seawater into fresh groundwater can take place due to over pumping of the aquifer, alter ing recharge through changes in land use, or can happen over longer time periods as a result of climate induced sea level rise and land subsidence (Essink et al., 2010) In addition to the salinization of fresh groundwater, this same phenomenon has been observed in surface waters near the coast which can be affected by drought, periods of lo w flow, or tidal fluctuations which can all be exacerbated by climate change over longer time scales (Bonte & Zwolsman, 2010) A major concern of saltwater intrusion is the replacement of freshwater with lower quality waters high in salt content. The majority of literature on saltwater intrusion has focused on chlorid e or sodium ions, the predominant constituents of seawater, but generally neglects other minor components of seawater like bromide or iodide (Rasmussen et al., 2013). In the past the major problem with
64 saltwater intrusion has been taste associated with hig h levels of chloride; halide ions like bromide and iodide occur at much lower concentrations and do not contribute to taste which is why they have not been previously monitored. Now, there is evidence to support that these trace halides can lead to the for mation of strong halogenating agents resulting in brominated or iodinated DBPs when present in drinking water supplies (Hua et al., 2006) Along with chloride ions, bromide is co transported during saltwater intrusion, elevating background levels in freshwater aquifers creating the potential t o form THMs and HAAs (Alcal & Custodio, 2008) Although it has been documented in recent studies (Hua et al., 2006; Krasner et al., 2006) the effects of iodi de on DBP formation and the subsequent health risks are beyond the scope of this work. The presence of bromide is problematic from a regulatory perspective where brominated species of DBPs are currently regulated by the U.S. Environmental Protection Agen cy (EPA), the European Union (EU), and other developed countries worldwide; in addition to these regulatory entities, the World Health Organization (WHO) provides guidelines for allowable DBP concentrations in treated drinking water. U.S. EPA regulated b romine containing DBPs (Br DBPs) resulting from chlorine disinfection include three species of trihalomethanes (THMs) and two species of haloacetic acids (HAAs) (there are also four unregulated bromine containing HAA species); the only other regulated Br D BP is bromate from ozonation. In addition to the handful of regulated Br DBPs there are hundreds of unregulated species that have been identified in recent research articles including brominated carboxylic acids and 2,3,5 tribromopyrrole (Richardson et al., 2003) new polar aromatic and unsaturated aliphatic Br DBPs like 2,4,6 trib romophenol, 3,5 dibromo 4 hydroxybenzoic acid, 2,6 dibromo
65 1,4 hydroquinone, and 3,3 dibromopropenoic acid (Zhai & Zhang, 2011) brominated halobenzoquinones (Zhao et al., 2012) and new putative aromatic halogenated DBPs including bromoaleic acids, hydroxybenzaldehydes, hydroxybenzoic acids, bromo salicylic acids, and trihalo phenols (Ding et al., 2013; Pan & Zhang, 2013) Although these new Br DBPs have been identified in chlorin ated waters, their health effects are unknown and may potentially be more toxic than regulated counterparts. Bromide can also be problematic from a water treatment perspective where it is difficult to remove if saltwater intrusion occurs and reverse osmos is (RO) is not in place. The formation of Br DBPs occurs when bromide ion is present at the point of disinfection; strong oxidants, such as the acid/base pair hypochlorous acid/hypochlorite (HOCl/OCl ), can oxidize bromide ion to aqueous bromine (hypobromo us acid/hypobromite or HOBr/OBr ) a reactive species that leads to the formation of brominated byproducts. A set of reactions are shown to demonstrate the chemical mechanisms that lead to Br DBPs from chlorine disinfection; there are also several intermedi ate reactions that can occur, but the most important reactions are highlighted below. (Kumar & Margerum, 1987) (Kumar & Margerum, 1987) (Westerhoff et al., 2004; Hua et a l., 2006)
66 (Westerhoff et al., 2004; Hua et al., 2006) Equations 3 1 and 3 2 are the oxidation of bromide to aqueous bromine, the respective reactions rate coefficients indicate that the formation of aqueous bromine is cataly zed a t lower pH values. Equations 3 3 and 3 4 are generic chemical reactions demonstrating the rate of formation of halogenated byproducts in the presence of HOCl or HOBr. Based on the reaction rate coefficients, brominated byproducts are formed at a rate nearly two orders of magnitude greater than chlorinated analogues. Additionally, the reaction rate coefficients are from Westerhoff et al. (2004) where brominating reactions were described as having two stage reactions kinetics. The initial reactions are r apid and happen too quickly to quantify K values, but were estimated at 500 5000 M 1 s 1 ; equation 3 4 shows the reaction rate for the slower, second stage brominating reactions. Comparing the reaction kinetics for HOBr/NOM and HOCl/NOM indicates a great er efficiency for bromine substitution into byproduct molecules. It is necessary to remove bromide ion precursors prior to chlorine disinfection to prevent the formation of HOBr/OBr and subsequent Br DBPs. Techniques for removing bromide fall into three c ategories: membrane separation, electrochemical, and adsorptive techniques (Watson et al., 2012). Membrane processes can include reverse osmosis, nanofiltration, and electrodialysis membrane techniques; electrochemical methods include electrolysis, capacit ive deionization and membrane capacitive deionization; and adsorption techniques include layered double hydroxides, impregnated activated carbon, ion exchange, carbon aerogels, and aluminum coagulation (Boyer & Singer, 2005; Watson et al., 2012)
67 According to a national survey of bromide in drinking water sources, the average concentration of bromide ion (Br ) is approximately 60 g L 1 across lakes, rivers, and gr oundwaters used for potable purposes (Amy et al., 1993). However, it is realistic to observe concentrations in the range of 500 g L 1 in coastal areas with some studies reporting Br concentrations in excess of 2 mg L 1 (Nikolaou et al., 2004; Agus et al., 2009; Walker & Boyer, 2011) Researchers have demonstrated that increasing bromide concentrations can have a significant impact on the formation and speciation of regulated and unregulated B r DBPs. Studies by Wu and Chadik (1998) and Hua et al. (2006) have demonstrated that a ten fold increase in Br concentrations are capable of doubling concentrations of THM4 or HAA9 and shifting speciation to the brominated analogues. In addition to cr eating compliance issues for utilities, these Br DBPs are believed to pose a greater cancer risk than chlorinated species as demonstrated in bioassays of Chinese hamster ovaries (Plewa et al., 2002) and chromosomal abberations in Chinese hamster lung cells ( Echigo et al., 2004) Considering the implications of Br DBPs on human health and regulatory compliance it is critical and timely to consider the potential for elevated levels of Br DBPs in source waters affec ted by saltwater intrusion. The key limitation of previous published work on saltwater intrusion is the lack of information regarding formation and speciation of Br THMs and Br HAAs in drinking water sources that will experience sudden or progressive incr eases in Br There are a handful of research articles that have examined the effects of chlorinating natural and synthetic seawater samples (Shi et al., 2012) saline sewage effluents (Ding et al., 2013) and seawater swimming pools (Parinet et al., 2012) having Br concentrations
68 ranging from approx. 4 mg L 1 to 90 mg L 1 more than two orders of magnitude higher than that of a typical drinking water. There are also articles focusi ng on quantifying increases in the magnitude of THMs and HAAs as well as the shift in speciation as a function of Br ; however, these articles typically involve natural waters low in Br that are spiked with a KBr or NaBr solution (Cowman & Singer, 1996 ; Liang & Singer, 2003 ; Hua et al., 2006 ; Hua & Reckhow, 2012 ; Pan & Zhang, 2013) Th e work in this study uses natural seawater as the source of Br spike to a potable groundwater supply to simulate various scenarios of saltwater intrusion. The use of real seawater is important to mimic the natural process of saltwater intrusion in which n ot only Br is transported into the aquifer but also chloride ion (Cl ). A study by Sivey et al. (2013) demonstrated excess Cl in the presence of free available chlorine catalyzed the formation of bromine chloride (BrCl) a brominating agent that is up to 10 6 times more reactive than hypobromous acid (HOBr), the agent assumed to be responsible for forming Br DBPs. Spiking fresh groundwater with seawater provides the Cl that may faci litate the formation of Br DBPs and ultimately alter the overall speciation and magnitude of DBPs formed. The overall goal of this work was to provide new insights on the effects of saltwater intrusion on the formation and speciation of DBPs. The specific objectives of the research were to (1) quantify the formation and speciation of THM4, HAA5 and HAA9 for different ratios of fresh groundwater and Gulf of Mexico seawater; (2) elucidate the inclusion of bromide in DBPs by using the DBP data to calculate br omine incorporation factors (BIF); and (3) estimate the cancer risk of DBPs as a result of saltwater intrusion.
69 3.2 Experimental Section 3.2.1 Sampling Location Groundwater samples were collected from well 4 at the Cedar Key Water Treatment Plant (CKWTP) in Cedar K ey, FL on April 8th, 2013. On the same day that samples were collected from the production well a seawater sample was collected approximately one mile offshore at about 2/3 the depth of the water column. All samples were bottled, placed in a cooler on ice, returned to the University of Florida (approximately 1 h drive), filtered immediately using 0.45 m membrane filters (Millipore), and stored at 4 C in the dark until used for analysis and experiments. Simulation of saltwater intrusion was achieved by mi xing the Cedar Key groundwater at various ratios with the Gulf of Mexico seawater in volumetric flasks. It should be noted that prior to the seawater spike all groundwater samples were diluted with three parts deionized (DI) water to one part groundwater to achieve DOC levels comparable to finished drinking water (approximately 1.4 mg L 1 ). The high Br concentration in seawater allowed for small spiking volumes which helped ensure minimal changes in water chemistry such as pH, DOC, and ultraviolet absorba nce at 254 nm (UVA 254 ). The ratios of groundwater to seawater were adjusted to result in six samples that contained 0%, 0.1%, 0.2%, 0.4%, 1%, and 2% seawater by volume yielding theoretical Br concentrations of 25, 50, 100, 200, 500, and 1000 g L 1 based on conservative mixing of the groundwater and seawater This range was chosen based on a thorough review of saltwater intrusion literature and is believed to be a representative concentration spectrum for coastal drinking waters impacted by seawater while also remaining under the secondary MCL for chloride at 250 mg L 1
70 3.2.2 Chlorine Demand and Chlorination Under Uniform Formation Conditions Chlorine demand test and chlorination under uniform formation conditions (UFC) were conducted following the procedure as described by Boyer and Singer (2005) and Summers et al., (1996) The UFC were conditions that yielded a 1.0 0.4 mg L 1 free chlorine resi dual at pH 8 after 24 h of incubation in the dark at 25 C. A chlorine dosing solution was prepared from a 4 6% NaOCl stock solution (Fisher Scientific) by diluting to approximately 1000 mg L 1 The stock NaOCl solution was standardized prior to all chlori nation experiments by titrating with 0.1 N sodium thiosulfate according to Standard Method 4500 Cl B. Iodometric Method ( Eaton 2005 ). All glassware was made chlorine demand free before chlorination experiments by acid washing and soaking in a 50 mg L 1 N aOCl solution over night. Determination of chlorine demand was performed by choosing six different chlorine doses (in the range of 1 3 mg Cl 2 per mg DOC) and dosing samples in 300 mL glass stoppered, headspace free BOD bottles. The requisite chlorine dose was chosen based on the dose that yielded a 1.0 mg L 1 free chlorine residual in the 0% seawater sample (the 3+1 Cedar Key groundwater). The 1% and 2% seawater samples (503 and 974 g L 1 Br respectively) exerted additional chlorine demand resulting in a free chlorine residual of approximately 0.4 mg L 1 as Cl 2 Although the free chlorine residual for these two samples fell outside the range of the UFC (1.0 0.4 mg L 1 Cl 2 ) it was important to use the same dose across all samples as established in the background matrix. In reality utilities would apply a greater chlorine dose to high bromide waters in order to achieve a 1 mg L 1 Cl 2 residual and produce even higher concentrations of DBPs All samples chlorinated for THM and HAA analysis were done in 300 mL BOD bottles transferred to 40 mL vials with screw caps and Teflon faced silicone sept a, and
71 filled headspace free The 40 mL vials were preserved with ammonium chloride for HAAs and sodium thiosulfate for THMs in order to quench the residual free chlo rine after 24 h. All samples were collected in triplicate, stored in a cooler on ice until analysis, and all samples were extracted within two days of chlorination. 3.3 Analytical Methods 3.3.1 DOC M easurements DOC was measured using a Shimadzu Total Organic Carbon Analyzer (TOC V CPH ) equipped with an ASI V auto sampler and analyzed according to Standard Method 5310 B high temperature combustion method ( Eaton et al., 2005 ). Prior to the analysis samples were filtered with a 0.45 m membrane filter and acidified to a pH < 2 using 2 N HCl. Accuracy was determined by measuring calibration standard checks and independent control standards (Ricca chemical company) and precision was evaluated by running sample duplicates All percent recoveries were in the acceptable margi n of 10% and all duplicates had a relative percent difference (RPD) of less than 10% 3.3.2 Ultraviolet A bsorbance at 254 nm (UV A 254 ) UV A 254 was measured using a 1 cm quartz cuvette on a Hitachi U 2900 spectrophotometer following Standard Method 5910 ( Eaton 2005 ). Similar to DOC measurements, all samples were filtered using 0.45 m membrane filters. 3.3.3 Inorganic A nions Measurements for chloride and bromide were performed on a Dionex ICS 3000 ion chromatograph equipped with an AS40 auto sampler, an IonPac ASS22 c olumn, IonPac AG22 guard, and ASRS 300 4mm suppressor, the system was r u n with a 4.5 mM Na 2 CO 3 /1.4 mM NaHCO 3 eluent (Dionex concentrated AS22 eluent) ; all samples were analyzed according to Walker and Boyer (2011). Instrument accuracy was
72 determined by mea suring seven anion standard (Dionex) and precision was evaluated by measuring sample duplicates. All duplicates had a RPD less than 10% and accuracy checks for the seven anion standard were within 20% for each analyte. 3.3.4 pH All samples were buffered with a p H 8.0 borate solution and adjusted to a final pH of 8.0 0.05 with H 2 SO 4 or NaOH prior to all chlorination experiments. Measurements were performed using a Accumet Basic AB15 pH meter and probe and the instrument was calibrated with a pH 4.0, 7.0, and 10. 0 buffer solution (Fisher Scientific). 3.3.5 Chlorine R esidual After 24 h of incubation the free chlorine residual of each sample was measured using a HACH DR 850 pocket colorimeter and DPD free chlorine foil pillow reagent for a 10 mL sample size (HACH ). 3.3.6 THM A n alysis Samples generated by chlorination under UFC were analyzed for the four bromine and chlorine containing THM species (THM4). The analysis was performed in accordance with EPA method 524.2 Volatile Organic Compounds by Gas Chromatography/Mass Spectro metry (GC/MS) at Advanced Environmental Laboratories (AEL) (Jacksonville, FL) (USEPA, 1995) The method involved bubbling ultrapure helium through the aqueous sample, the purged sample components are trapped in a tube containing suitable sorbent materials. Once completely purged the tube is heated and back flushed with helium to desorb trapped components onto a capillary gas chromatography (GC) column interfaced to a mass spectrometer. The column is temperature programmed to facilitate the separation of ana lytes. Compounds eluting
73 from the GC column are identified by comparing their measured mass spectra and retention times to reference spectra and retention times in a data base. The extracts were analyzed on a Shimadzu GC 17A gas chromatograph with a Shimad zu GCMS QP5000 mass spectrometer. The capillary column was a Restek RTX VRX 60 m by 0.32 mm fused capillary column with a phase of 1.8 m. All samples analyzed were subject to a rigorous QA/QC protocol established by AEL. Laboratory control samples (LCS) a nd matrix spikes/matrix spike duplicates (MS/MSD) were run along with the samples. All sample duplicates has a relative percent difference (RPD) of less than 20% and LCS and MS/MSD recoveries fell within the acceptable margin of 20%. The method detection limit (MDL), as determined by AEL, for the four THM analytes were as follows: chloroform = 0.66 g L 1 bromodichloromethane = 0.60 g L 1 dibromochloromethane = 0.56 g L 1 and bromoform = 0.89 g L 1 3.3.7 HAA A nalysis Samples generated by chlorination und er UFC were analyzed for the nine bromine and chlorine containing HAA species (HAA9). The analysis was performed in accordance with EPA method 552.2 Determination of Haloacetic Acids in Drinking Water by Liquid liquid Extraction, Derivatization and Gas C hromatography with Electron Capture Detection at Advanced Environmental Laboratories (AEL) (Jacksonville, FL) (USEPA, 1990) The method involved adjusting the pH of the 40 mL sample to a pH < 0.5 and extracting with 4 mL of methyl tert butyl ether (MTBE). The HAAs that have been partitioned into the organic phase are converted to their methyl esters by adding acidic methanol and heating. The acidic extract is neutralized by a back extraction with a saturated solution of sodium bicarbonate and target analy tes are identified and
74 measured by a capillary column gas chromatography and electron capture detector (GC/ECD). The derivatized sample extracts plus an internal standard of 1,2,3 trichloropropane, and the surrogate compound 2,3 dibromopropionic acid were analyzed on a Pekin Elmer GC dual column with dual ECD and data system for measuring peak areas. The guard column from the injector was connected with "y" splitter into two columns where column 1 was a 30 m by 0.32 mm Restek fused silica capillary column with 0.5 m film thickness and column 2 was a 30 m by 0.32 mm Phenomenex Zebron Multi Res fused silica capillary column also with 0.5 m film thickness. Along with surrogates and internal standards, LCS and MS/MSD were run with the samples with recoveries of 20% and the RPD between duplicates was < 20%. The MDLs for the nine HAA species were as follows: monochloroacetic acid = 0.887 g L 1 monobromoacetic acid = 0.516 g L 1 dichloroacetic acid = 0.894 g L 1 trichloroacetic acid = 0.670 g L 1 dibr omoacetic acid = 0.731 g L 1 bromodichloroacetic acid = 0.5 g L 1 chlorodibromoacetic acid = 0.5 g L 1 tribromoacetic acid = 1.0 g L 1 and bromochloroacetic acid = 1.0 g L 1 3.4 Results and Discussion 3.4.1 Effects of Saltwater Intrusion on Bromide Concent ration Several studies worldwide have reported the sudden or progressive increase in the total dissolved solids (TDS) of coastal aquifers and rivers due to saltwater intrusion Barlow & Reichard, 2009) The intrusion of Br is often overlooked or of minor concern ; however some researchers have made a link to Br using chloride to bromide ratios (Cl /Br ) and comparing them to the well known ratio in seawater of 655 4 on a molar basis or 287 on a mass basis (Davis et al., 1998; Millero
75 et al., 2008) The authors of this wo rk are not aware of another study that places primary focus on increases in Br as a result of saltwater intrusion and the impacts it has on potable groundwater. The CKWTP is a small public water system located approximately two miles from the Gulf of Mex ico and provides drinking water to approximately 1000 customers. On May 28th, 2012 the conductivity in the production wells increased by three times and over the next month the Cl concentrations increased from approximately 25 mg L 1 to 500 mg L 1 The s udden increase in Cl was attributed to a complex combination of pumping and drought which lead to severe saltwater intrusion, making this an ideal location for this study Since the May 2012 saltwater intrusion event at Cedar Key weekly samples have been collected from the two production wells at the facility and measured for chloride and bromide concentrations. Background levels of Br at this study site were generally < 25 g L 1 but reached as high as 580 g L 1 in July of 2012. For the experiments cond ucted in this study, t he background concentration of Br in the Cedar Key groundwater was 84 g L 1 and the Br concentration of the seawater was 49.7 mg L 1 Table 3 1 lists the water quality parameters measured in the chlorination study for the undilute d Cedar Key groundwater, the undiluted seawater, and the different mixing ratios where the values reflect the water quality after dilution with DI water. Bromide and chloride values measured in the seawater sample were approximately 80% of typical seawater concentrations, but yielded a Cl / Br ratio of 283 which is nearly identical to the 287 mass ratio reported in the literature (Davis et al., 1998 ; Millero et al., 2008 ) These results suggest an exchange of freshwater with
76 seawater at the sampling location, approximately one mile from the coastline. The concentrations of bromide and chloride are still high enough to require only small volumes for dilution and the c onsistent ratio with model seawater makes them ideal for sample spiking. The observed Cl /Br ratio in the undiluted Cedar Key groundwater was 345, indicating a chloride enrichment of about 20% in the vicinity of the production wells. For the various groun dwater/seawater mixes the ratios fell within the range of 205 292 with four of the six saltwater intrusion scenarios being within 10% of the seawater Cl /Br ratio. The ratios reported in this study are comparable to ratios previously reported for potabl e groundwaters. Davis et al. (1998) tabulated the Cl /Br ratios for 251 potable groundwaters where it was observed that groundwaters with low chloride (< 200 mg L 1 ) generally had ratios less than 200, but as chloride concentration increased the ratio app roached that of seawater. The resulting bromide concentrations, which will affect DBP formation, were generally within 10% of the theoretical values calculated based on conservative mixing. For the 0% seawater sample the bromide concentration was higher th an expected, yielding 38 g L 1 as opposed to 25 g L 1 (Table 3 1). The only other water quality parameters that affect DBP formation are DOC, UVA 254 and SUVA 254 ; the measured concentrations for DOC and UVA 254 remained nearly constant across all samples and SUVA 254 remained in the narrow range of 2.74 3.12. The consistency of these precursor concentrations makes the formation of DBPs solely a function of bromide concentration. 3.4.2 Effects of Saltwater Intrusion on THM and HAA Formation The severity of saltw ater intrusion is represented by increases in Br starting at 38 g L 1 (no saltwater intrusion) and reaching the highest concentration of 974 g L 1 Figure 3 1 A and 3 1 B show the THM4, HAA5, and HAA9 yields in mass and molar
77 units. Mass units are useful to compare results to drinking water maximum contaminant levels (MCLs) where THM4 = 80 g L 1 and HAA5 = 60 g L 1 as specified in the Stage 2 Disinfectants and Disinfection Byproduct Rule (Stage 2 D/DBP Rule) (U.S. EPA, 2006). The molar units give insigh t to the increase in halogenation that occurs when bromide is present in high concentrations. The increase in mass concentration for each DBP class is not only attributed to substitution of one chlorine atom for one heavier bromine atom (molecular weights of 35.45 g mol 1 and 79.90 g mol 1 respectively), but HOBr/OBr forms at higher bromide concentrations and is a stronger halogenating agent than HOCl/OCl resulting in a greater conversion of precursors to byproducts. The results show a greater molar form ation of DBPs as seawater content increases, but the molar concentration of organic precursors remains constant. Using THM4 as an example, the molar yields increase from 0.33 mol L 1 to 0.85 mol L 1 over the range of 0% seawater to 2% seawater; however, the concentrations of DOC and UVA 254 remain constant. For the no saltwater intrusion scenario (Br = 38 g L 1 ) the resulting THM4 concentration was 43.4 g L 1 and HAA5 was 24.3 g L 1 approximately half the MCLs. When increasing Br to 974 g L 1 the m ass yields for THM4 increased to 206.5 g L 1 and HAA5 increased to 26.0 g L 1 a 376% and 7% increase, respectively. The overall increase in the magnitude of THM4 is substantial, exceeding the regulatory threshold by 126.5 g L 1 ; however, the small incr ease in HAA5 is a result of speciation shifting to the bromine containing species which are not included in the summation of HAA5. When referring to the molar yields of HAA5, there was a 27% decrease when Br increases from 38 g L 1 to 974 g L 1 but ove r the same Br range HAA9 increased by
78 32% (a 92% increase on a mass basis) These results agree with previous studies by Hua et al. (2006) where they observed that the molar yield of HAA5 decreased while the molar yield of HAA9 increased in two different water sources with the addition of a bromide spi ke. An important distinction between the Hua et al. (2006) results and the results presented in this study are the increases in THM4 and HAA9 as a function of bromide spike. Hua et al. (2006) reports an increase in THM4 of 58% and 74% for two natural water s that had 30 mol L 1 of bromide added when using a NaBr spike. For the seawater spiked samples in this study the molar concentrations of THM4 increased by 155% for the equivalent of a 11.7 mol L 1 bromide added. Increases in HAA9 were not as substantial where the Hua et al. (2006) study reported 18% and 35% increases in the two natural waters with 30 mol L 1 bromide added, and the present study found a 32% increase in HAA9 for 11.7 mol L 1 bromide added. Comparing the percent increases of THM4 and HAA 9 to the Hua et al. (2006) study indicates a positive result for the hypothesis that the chloride in the seawater spike catalyzes the formation of byproducts. The bromide added in the present study was less than half of that in Hua et al. (2006) study but the percent increases in this study were greater than or equal to the percent increases in Hua et al. (2006). The results should be interpreted with caution, however, since a direct comparison cannot be made where the chlorination procedure in Hua et al. ( 2006) (chlorine dose = 5.0 6.2 mg Cl 2 L 1 reaction time = 48 h, pH = 7, temperature = 20 C, and Cl 2 residual = 0.5 mg L 1 ) and organic precursor concentrations (DOC = 5.1 8.5 mg L 1 ) are different from the present study. Figures 3 2 A and 3 2 B illu strate THM4 species concentrations and HAA9 species concentration as a function of saltwater intrusion respectively. As
79 demonstrated in these figures, there is not only an increase in the overall concentration of THM4 and HAA9, but the speciation is shift ed toward the brominated species as Br increased. The five regulated HAAs include monochloroacetic acid (ClAA), dichloroacetic acid (Cl 2 AA), trichloroacetic acid (Cl 3 AA), monobromoacetic acid (BrAA), and dibromoacetic acid (Br 2 AA) whose concentrations are only 34% of HAA9 when Br is at 974 g L 1 The increase in HAA9 is more dramatic compared to HAA5, increasing from 39.2 g L 1 to 79.2 g L 1 a 92% increase. The four additional HAA species included in HAA9 are tribromoacetic acid (Br 3 AA), bromodichloro acetic acid (BrCl 2 AA), chlorodibromoacetic acid (ClBr 2 AA), and bromochloroacetic acid (BrClAA), species all containing bromine. At the most severe saltwater intrusion scenario ( 974 g L 1 Br ) the bromine containing species are the most abundant, bromoform consisting of 85% of THM4 and tribromoacetic acid consisting of 52% of HAA9. Studies by Wu and Chadik (1998) showed similar results for HAA9 speciation where Br 3 AA, Br 2 AA and ClBr 2 AA increased and concen trations of ClAA, Cl 2 AA, and Cl 3 AA decreased as the ambient bromide levels were elevated. In addition to overall increases in the magnitude of HAA9, which may go unnoticed from a regulatory perspective, several studies have demonstrated that the unregulate d bromine containing species pose greater risks to human health than the regulated species (Echigo et al., 2004 ; Richardson et al., 2007) The implications of these results are increased exposure of HAA9 to consumers where Br is high in sou rce waters and no detection of HAA MCL exceedances by regulatory entities because HAA5 remains low. 3.4.3 Effect s of Saltwater Intrusion on Bromine Incorporation Factor The bromine incorporation factor (BIF) is the ratio of the number of moles of organic bromine to the number of moles of total organic halogen (neglecting iodinated
80 species) for a specified class of DBP compounds. Equations used for calculating the BIF are as follows (Boyer & Singer, 2005) : (3 5) (3 6) (3 7) Figure 3 3 shows the BIF for THM4, dihalogenated HAAs (X 2 AAs), and trihalogenated HAAs (X 3 AAs) as a function of saltwater intrusion (Br = 38 g L 1 to Br = 974 g L 1 ) where the value s can range from zero to one. The BIF has also been referred to as the bromine substitution factor in previous studies, mathematically e xpressed the same as equations 3 5 3 7 (Hua & Reckhow, 2012) The BIF gives insights to the extent of bromine substitution for each cla ss of DBPs ranging from 0.08 to 0.92 for THM4, 0.07 to 0.93 for X 2 AA, and 0.31 to 0.92 for X 3 AA. These results agree with bromine substitutions reported by Obolensky and Singer (2005) and Boyer and Singer (2005) where the BIF was comparable for THM4 and X 2 AA. However, t he X 3 AA substitution reported by Obolensky and Singer (2005) was approximately 10% lower than THM4 where X 3 AA substitution in this study and the study by Boyer and Singer (2005) demonstrated higher BIF when compared to THM4 and X 2 AA. Another study investi gating BIF Hua and Reckhow (2012) report ed BIF for THM4, X 2 AA, X 3 AA, and dihaloacetonitriles (DHAN) as a function of bromide spike, chlorine dose, reaction time, temperature, and pH. The results from Hua and Reckhow (2012) agree with results from the study where the behav ior of THM4 and X 2 AA are
81 similar in terms of BIF, and X 3 AA exhibits a different behavior across various water chemistry scenarios. The Hua and Reckhow (2012) study showed that THM4 and X 2 AA experienced greater chlorine incorporation at basic pH (pH = 10) a nd resulted in lower BIF; alternatively, X 3 AA formation was catalyzed at acidic pH (pH = 5) which suggests that formation mechanisms for THM4 and X 2 AA are different than that of X 3 AA. It can also be expected that B I F may decrease at longer reaction times s ince the brominating reactions are typically faster than chlorine reactions and the incorporation of bromine containing species may account for a higher percentage of total halogenated byproducts at shorter reaction times. The strong agreement between the present study and the Boyer and Singer study (2005) is due to the similarities in experimental design (i.e. chlorination under UFC where t = 24 h), accounting for similar BIF results across the three classes of DBP compounds that were investigated. The Obo lensky and Singer study (2005) calculated BIF using the Information Collection Rule (ICR) database that yielded a wide range of BIF results across the 6565 samples from 297 drinking water utilities, where samples were collected in the distribution system. Differences in BIF for X 3 AA from the present study and the Obolensky and Singer study (2005) arise from the use of field data where the reaction times were likely greater than 24 h and the chlorine dose was not uniform. 3.4.4 Effect s of Saltwater Intrusion on He alth Risks of DBPs Health risks for various classes of DBPs have been well documented in the literature demonstrating cytotoxicity and genotoxicity in Chinese hamster ovary cells (Plewa et al., 2002) and chromosomal aberrations in Chinese hamster lung cells (Echigo et al., 2004) To estimate human health impacts some researchers have developed formulas for calculating a probabilistic risk based on DBP exposure and
82 dose response data published in toxicological or epidemiological studies. Using DBP concentrations in water (CW) and standard values for body weight (BW), ingestion rate (IR), exposure frequency (EF), exposure factor (AF, assumed to be 1), average expo sure time (AT), and exposure duration (ED) a standard exposure dose (EXP) can be calculated by the following equation (Hamidin et al., 2008 ) : (3 8) The EXP has also been referred to as the chronic daily intake (CDI) in other studies (Lee et al., 2006) Using the EXP and risk slope factors (SF, (mg kg 1 d 1 ) 1 ) developed by the USEPA in the Integrate Risk Information System (IRIS), the risk attributed to individual compounds can be calculated by multiplying EXP by SF to y ield a dimensionless risk factor (USEPA, 2005) The risk attributed to the ingestion of drinking waters containing the four THM species can be estimated by summing the individual risk associated with each THM compound. Excess cancer cases per one million p eople can be calculated using the THM4 concentrations from this study and the standard val ues for EXP listed in equation 3 8 Figure s 3 4 A and 3 4B shows the increase in hypothetical cancer cases as a function of saltwater intrusion for THM4 and individua l THM species, respectively The values presented in Figure s 3 4 A and 3 4B are a result of ingestion, however theoretical risk may actually be higher due to exposure from inhalation and dermal contact with THMs (i.e. showering or swimming) (Black et al., 1996; Lee et al., 2006) The relative toxicity of b r omine containing THMs (Br THMs) can also be evaluated by simply
83 comparing SF where the three bromine containing species have a SF more than one order of magnitude greater tha n CHCl 3 Also looking at theTHM4 data from this study, the no saltwater intrusion scenario (Br = 38 g L 1 ) yields total Br THMs of 12 g L 1 and the most sever saltwater intrusion scenario (Br = 974 g L 1 ) yields total Br THMs of 206 g L 1 Based on t he SF for the Br THMs and the shift in speciation towards the brominated species, the effects of saltwater intrusion create a two fold impact on the health risk associated with THM ingestion. Illustrated in Figure 3 4 B the excess cancer cases attributed t o CHCl 3 never exceed six cases per one million people, but cancer cases attributed to CHBr 3 reach 400 cases per one million people at the most severe scenario of saltwater intrusion. Since there are no MCLs for individual THM species it is more important t o consider the risks attributed to the ingestion of the entireTHM4 (Figure 3 4A ). It should also be noted that these estimates for cancer risk are not considering exposure to HAAs (IRIS does not currently have slope factors for all HAA species), or other u nregulated DBPs that form as a result of high bromide concentrations. 3.5 Implications of the Impacts of Saltwater Intrusion on THM and HAA Formation Utilities in coastal regions are experiencing progressive increases in the TDS of their freshwater aquifers ei ther as a result of increased pumping and decreased recharge (Murgulet & Tick, 2007) Additionally, it has been suggested that the effects of saltwater intrusion will be exacerbated in the future by climate induce sea level rise and land subsidence (Darnault & Godinez, 2008 ; Oude Essink et al., 2010) Researchers have developed groundwater models to predict the increase in TDS or chloride as a function of increased head at the coastal boundary or changes in land use that affect recharge. Modeling techniques, or fi eld data from observation wells, can be used to
84 estimate concentrations of bromide in the vicinity of the drinking water production wells. Additionally, historic monitoring data for chloride along with Cl /Br ratios can be useful tools for estimating brom ide concentrations when no bromide data is available. These preliminary estimates could then be used to assess the potential to exceed regulatory MCLs or increase risks to human health. At 2% seawater (v/v), the most severe saltwater intrusion scenario in this study, the chloride concentrations were approximately equal to the secondary MCL of 250 mg L 1 but THM4 was a factor of 2.5 greater than the primary MCL. It can be expected that utilities will experience compliance issues with DBPs well before the ae sthetic concerns with chloride or TDS pose issues. The results of this work demonstrate that regulated DBPs, like THM4, and unregulated DBPs, like HAA9, can be problematic for drinking water utilities that use chlorine disinfection and experience increases in bromide caused by saltwater intrusion. Estimates from this study suggest a coastal aquifer intruded with ~0.5% seawater by volume can cause THM4 to increase by a factor of two and is estimated to cause nearly 200 excess cancer cases attributed to the i ngestion of THMs. To avoid compliance issues it may be desirable for utilities to switch to alternative disinfectants to manage levels of THM4. Research articles have already been published demonstrating reductions in DBPs using alternative disinfectants s uch as chloramines, chlorine dioxide, and ozonation (Hua & Reckhow, 2007) Although these methods have demonstrated reductions in THM4 and HAA5, they may pose a new health risk associated with DBPs that are not regulated or even monitored. Facilities using chloramines may form iodinated THMs at levels harmful to human health if iodide is
85 present in the source water, which it would be from saltwater intrusion (Bichsel & von Gunten, 2000) Other studies have demonstrated the potential for chloramines to form nitrogenous DBPs, like nitrosamines, when organic nitrogen precursors are present (Chang et al. 2011) Bromide precursor is also a problem for utilities that switch to ozonation as a mode of disinfection where it is well known that ozone forms bromate, one of the regulated oxyhalides, a problematic inorganic byproduct that cannot be degraded by bio logical filters (von Gunten, 2003) Facilities considering switching from chlorine disinfection to an alternative disinfectant need to consider the precursors in their source waters and the removal efficiency of the precursors prior to disinfection. These me asures are to ensure they do not form unregulated byproducts that are potentially more harmful than regulated DBPs that arise from chlorine disinfection. Water resources managers, utilities, and engineers will have to account for bromide in source waters a nd adapt to changing water quality in years to come In addition to improving disinfection practices it may be necessary to implement advanced treatment processes such as ion exchange or reverse osmosis for bromide precursor removal or increase the efficie ncy of NOM removal as measured by DOC or UVA 254 A recent review article by Watson et al. (2012) discusses the difficulties of removing halides like bromide and iodide, as well as NOM, to minimize DBP formation. It is concluded in the Watson et al. (2012) article that there is a general need for further development of halide removal techniques in commercial water treatment as more utilities are moving toward alternative water sources. Emerging Br DBPs have been well documented and the proliferation of new t echnologies and water treatment practices will be necessary to manage the impacts on human health. Future research in this field will
86 need to focus on the effects of bromide on the formation of emerging, unregulated species and also iodide which can form b yproducts that are potentially more toxic than Br DBPs (Richardson et al., 2007) Finally, c oastal utilities must be aware of changes in water quality induced by saltwater intrusion and anticipate the formation of new brominated and iodinated DBPs to protect p ublic health.
87 Table 3 1. Water quality parameters for various mixing ratios in the chlorination experiments. Sample DOC (mg L 1 ) UV 254 (cm 1 ) SUVA 254 (L mg 1 m 1 ) pH Br (g L 1 ) Cl (mg L 1 ) Cl /Br Undiluted CK Groundwater 5.73 0.164 2.86 8.08 84 29 34 5 Undiluted Seawater a 0.194 8.02 51,600 14,600 283 0% Seawater 1.39 0.04 2.88 7.98 38 7.8 205 0.1% Seawater 1.33 0.039 2.93 7.96 59 14 237 0.2% Seawater 1.35 0.037 2.74 8.01 106 28 264 0.4% Seawater 1.35 0.041 3.04 8.02 197 57 289 1% Sea water 1.41 0.043 3.05 8.04 503 146 290 2% Seawater 1.41 0.044 3.12 7.98 974 284 292 a DOC not measured in undiluted seawater samples.
88 Figure 3 1. DBP yields as a function of seawater concentration. A ) M ass based yields of THM4, HAA5, and HAA9 B ) M olar yields of THM4, HAA5, and HAA9. A ) B )
89 Figure 3 2. DBP speciation as a function of seawa ter concentration. A ) speciation of THM4 compounds where chloroform = CHC l 3 bromodichloromethane = CHBrCl 2 dibromochloromethane = CHBr 2 Cl, and bromoform = CHBr 3 B ) speciation of HAA9 compounds where monochloroacetic acid = ClAA, dichloroacetic acid = Cl 2 AA, trichloroacetic acid = Cl 3 AA, monobromoacetic acid = BrAA, dibromoacetic acid = Br 2 AA, tribromoacetic acid = Br 3 AA, bromodichloroacetic acid = BrCl 2 AA, chlorodibromoacetic acid = ClBr 2 AA, and bromochloroacetic acid = BrClAA. A ) B )
90 Figure 3 3. Bromine incorporation factor (BIF) for THM4, X 2 AA, and X 3 AA as a function of seawater added by volume.
91 Figure 3 4. Risk (excess cancer cases/10 6 people) as a function of seawater concentration A ) THM4 B) I ndividual THM species. Slope factors: CHCl 3 = 6.1 10 3 CHBrCl 2 = 6.2 10 2 CHBr 2 Cl = 8.4 10 2 and CHBr 3 = 7.9 10 2 (USEPA, 2005). A ) B )
92 CHAPTER 4 CONCLUSIONS AND RECOMMENDATIONS The thr eat of saltwater intrusion presents challenges for engineers and water resources planners, creating the need for innovative treatment technologies and alternative water sources. Coastal areas experiencing saltwater intrusion have been monitoring chloride c oncentrations and conductivity for years; however, as coastal resources become increasingly stressed it will be necessary to monitor bromide concentrations to assess the consequent health effects of DBP formation and devise strategies to control DBP format ion. An improved understanding of the formation chemistry of brominated DBPs in salt impacted waters is necessary to quantify human risk and facilitate compliance with the Stage 2 D/DBP Rule. Monitoring bromide, as well as other precursors like DOC and UV 2 54 is also necessary to determine if treatment facility upgrades are needed. Treatment alternatives such as membrane processes, electrochemical techniques, and adsorptive processes have proven to effectively remove bromide as well as DOC in drinking wat ers (Boyer & Singer, 2005; Watson et al., 2012) The diff iculties of predicting DBP formation were discussed in Chapter 2 where the best models evaluated could predict THM4 concentrations at 0.5 MCL ( 40 g L 1 ) and HAA5 at 0.25 MCL ( 15 g L 1 ). The best mod els were power law models that included at least five terms including DOC, UV 254 bromide, pH, reaction time, temperature, and chlorine dose. Modeling inaccuracy may arise due to the exclusion of key DBP formation explanatory variables, such as bromide, or due to the heterogeneous nature of NOM. It is well known that waters containing more aromatic, high SUVA 254 organic matter are those that are most reactive with chlorine and form
93 DBPs. The majority of publications on DBP modeling used an empirical approa ch to formulate their models following either linear or non linear regression techniques; much fewer models involved reaction kinetics. Model developers should focus on modeling over a broad range of precursor concentrations and chlorination conditions. Ad ditionally, it is recommended that future modeling work focus on new emerging DBPs like iodinated THMs and HAAs, NDMA, HANs, and ozonation byproducts like bromate. The applications for DBP models include operational purposes for water treatment facilities, water quality management, a measure of precursor removal efficiency, and connecting these models with epidemiological studies for assessing human risk and exposure. For a utility that wants to use a model to predict and manage DBP formation it is recommen ded that they choose one model formulation that generates low error and contains all relevant inputs. The utility should calibrate that model to their particular water source by adjusting coefficients and exponents in a way that minimizes error. This can b e done by conducting laboratory experiments or collecting field data and using a statistical software package that can determine coefficients based on minimization of error. This study demonstrated that bromide concentrations at approximately 200 g L 1 an d greater can be problematic for DBP compliance, elevating THMs well above the MCL based on the DOC concentrations and NOM chemistry of the experimental waters used. When reaching bromide concentrations of 1 mg L 1 organic bromine represents over 90% of to tal organic halogens on a molar basis as indicated by the BIF. The resulting species are more toxic and heavier compounds that make it difficult to stay under MCLs and increases health risk to consumers. The analysis performed in this
94 study estimated up to 500 excess cancer cases for every one million people when bromide concentrations reach 1 mg L 1 in source waters. In addition to high levels of regulated DBPs, the presence of bromide can result in the formation of new emerging brominated byproducts as di scussed in Chapter 3. These aromatic and polar brominated species often act as intermediates that can degrade to form THMs and HAAs, increasing the ultimate concentrations of THM4 and HAA9 (Pan & Zhang, 2013) Overall re sults from this study can lead to a fundamental understanding of how saltwater intrusion leads to increases in bromide and subsequent THM and HAA formation which can be used by engineers and utilities to manage DBPs and human health.
95 L IST OF R EFERENCES Agus, E., Voutchkov, N., Sedlak, D. L. 2009. Disinfection by products and their potential impact on the quality of water produced by desalination systems: A literature review. Desalination 237 (1 3), 214 237. Amy G.L., Chadik P.A., Chowdhury Z .K. 1987. Developing models for predicting trihalomethane formation potential kinetics. J Am Water Works Ass. 79(7): 89 97. Amy G., Siddiqui M., Zhai W., DeBroux J., Odem W. 1993. Survey of Bromide in Drinking Water and Impacts on DBP Formation. AWWA Res earch Foundation. Amy G.L., Siddiqui M., Ozekin K., Zhu H.W., Wang C. 1998. Empirical based models for predicting chlorination and ozonation byproducts: haloacetic acids, chloral hydrate, and bromate. U.S. EPA report CX 819579. Alcal, F. J., Custodio, E. 2008. Using the Cl/Br ratio as a tracer to identify the origin of salinity in aquifers in Spain and Portugal. J Hydrol. 359 (1 2), 189 207. Al omari, A., Fayyad, M., Abdel, A. 2004. Modeling trihalomethane f ormation for Jabal Amman water supply in Jordan. Environ Monit Assess. 9, 245 252. Barlow, P. M., Reichard, E. G. 2009. Saltwater intrusion in coastal regions of North America. Hydrogeol J. 18 (1), 247 260. Baytak, D., Sofuoglu, A., Inal, F., Sofuoglu, S.C ., 2008. Seasonal variation in drinking water concentrations of disinfection by products in IZMIR and associated human health risks. Sci Total Environ. 407, 286 296. Bichsel, Y., von Gunten, U. 2000. Formation of Iodo Trihalomethanes during Disinfection an d Oxidation of Iodide Containing Waters. Environ Sci Technol. 34 (13), 2784 2791. Black, B.D., Harrington, G.W., Singer, P.C., 1996. Reducing cancer risks by improving organic carbon removal. J Am Water Works Ass. 88, 40 52. Bonte, M., Zwolsman, J. J. G. 2010. Climate change induced salinisation of artificial lakes in the Netherlands and consequences for drinking water production. Water Res. 44 (15), 4411 24. Boorman, G.A., Dellarco, V., Dunnick, J.K., Chapin, R.E., Hunter, S., Hauchman, F., Gardner, H. 1 999. Drinking Water Disinfection Byproducts: Review and Approach to Toxicity Evaluation. Environ Health Persp. 107, 207 217. Boyer, T.H., Singer, P.C. 2005. Bench scale testing of a magnetic ion exchange resin for removal of disinfection by product precur sors. Water Res. 39(7), 1265 1276.
96 Boyer, T.H., Overdevest, C., Christiansen, L. Ishii, S.K.L., 2012. Expert stakeholder attitudes and support for alternative water sources in a groundwater depleted region. Sci Total Environ. 437, 245 254. Chang, E.E., Lin, Y.P., Chiang, P.C. 2001. Effects of bromide ion on the formation of THMs and HAAs. Chemosphere. 43, 1029 1034. Chang, H., Chen, C., Wang, G. 2011. Identification of potential nitrogenous organic precursors for C N DBPs and characterization of their DBPs formation. Water Res. 45 (12), 3753 64. Chen, B., Westerhoff, P. 2010. Predicting disinfection by product formation potential in water. Water Res. 44(13), 3755 3762. Chowdhury, S., Champagne, P., McLellan, P.J. 2010. Investigating effects of bromid e ions on trihalomethanes and developing model for predicting bromodichloromethane in drinking water. Water Res. 44(7), 2349 2359. Chowdhury, S., Champagne, P., McLellan, P.J. 2009. Models for predicting disinfection byproduct (DBP) formation in drinking waters: a chronological review. Sci Total Environ. 407(14), 4189 4206. Chowdhury, S.H. 2009. Modeling Trihalomethane Formation in Drinking Water with Application to Risk Based Decision Making. PhD Thesis. Queen's University, Kingston, Ontario, Canada. Cow man, G. A., Singer, P. C. 1996. Effect of Bromide Ion on Haloacetic Acid Speciation Resulting from Chlorination and Chloramination of Aquatic Humic Substances. Environ Sci Technol. 30 (1), 16 24. Darnault, C. J. G., Godinez, I. G. 2008. Coastal Aquifers and Saltwater Intrusion. Nato Sci Peace Secur 18, 185 201. Davis, S. N., Whittemore, D. O., Fabryka Martin, J. 1998. Uses of Chloride/Bromide Ratios in Studies of Potable Water. Ground water 36 (2), 338 350. Ding, G., Zhang, X., Yang, M., Pan, Y. 2013. Forma tion of new brominated disinfection byproducts during chlorination of saline sewage effluents. Water Res. 47 (8), 2710 8. Duong, H.A., Berg, M., Hoang, M.H., Pham, H.V., Gallard, H., Giger, W., Gunten, U.V. 2003. Trihalomethane formation by chlorination o f ammonium and bromide containing groundwater in water supplies of Hanoi, Vietnam. Water Res. 37(13), 3242 3252
97 fluoride, bromide and chloride contents in public drinking w ater supplies in Sicily (Italy). Environ Monit Assess. 145 (1 3), 303 13. Eaton, A.D., Clesceri, L.S., Rice, E.W., Greenberg, A.E. (Eds.). 2005. Standard Methods for the Examination of Water & Wastewater, 21st ed. APHA, AWWA, WEF, Washington, DC. Echigo, S., Itoh, S., Natsui, T., Araki, T., Ando, R. 2004. Contribution of brominated organic disinfection by products to the mutagenicity of drinking water. Water Sci Technol 50 (5), 321 328. Fabbricino, M., Korshin, G.V. 2009. Modeling disinfection by product s formation in bromide containing waters. J Hazard Mater. 168(2 3), 782 786. Golfinopoulos, S.K., Arhonditsis, G.B. 2002. Multiple regression models: a methodology for evaluating trihalomethane concentrations in drinking water from raw water characterist ics. Chemosphere. 47(9), 1007 18. Hamidin, N., Yu, Q. J., Connell, D. W. 2008. Human health risk assessment of chlorinated disinfection by products in drinking water using a probabilistic approach. Water Res. 42 (13), 3263 74. Hong, H.C., Liang, Y., Han, B .P., Mazumder, A., Wong, M.H. 2007. Modeling of trihalomethane (THM) formation via chlorination of the water from Dongjiang River 3), 48 54. Hua, G., Reckhow, D. A. 2007. Comparison of disinfection byproduct formation from chlorine and alternative disinfectants. Water Res. 41 (8), 1667 78. Hua, G., Reckhow, D. A. 2012. Evaluation of bromine substitution factors of DBPs during chlorination and chloramination. Water Res. 46 (13), 4208 16. Hua, G., Reckhow, D. A., Kim, J. 2006. Effect of bromide and iodide ions on the formation and speciation of disinfection byproducts during chlorination. Environ Sci Technol. 40 (9), 3050 6. Kampioti, A. A, Stephanou, E. G. 2002. The impact of bromide on t he formation of neutral and acidic disinfection by products (DBPs) in Mediterranean chlorinated drinking water. Water Res 36 (10), 2596 606. Kolla, R.B. 2004. Formation and modeling of disinfect ion by products in Newfoundland communities. Masters Thesis; Memorial University of Newfoundland, NL, Canada.
98 Krasner, S. W., Glaze, W. H., Weinberg, H. S., Daniel, P. A., Najm, I. N. 1993. Formation and Control of Bromate during Ozonation of Waters Containing Bromide. J Am Water Works Ass. 85(1), 73 81. Krasner, S .W., Weinberg, H.S., Richardson, S.D., Pastor, S.J., Chinn, R., Sclimenti, M.J., Onstad, G.D. 2006. Occurrence of a New Generation of Disinfection Byproducts. Environ Sci Technol. 40(23), 7175 7185. Kulkarni, P., Chellam, S. 2010. Disinfection by product formation following chlorination of drinking water: artificial neural network models and changes in speciation with treatment. Sci Total Environ. 408(19), 4202 4210. Kumar, K., Margerum, D. W. 1987. Kinetics and Mechanism of General Acid Assisted Oxidati on of Bromide by Hypochlorite and Hypochlorous Acid. Inorg Chem 26 2706 2711. Lee, H.K., Yeh, Y.Y., Chen, W.M. 2006. Cancer risk analysis and assessment of trihalomethanes in drinking water. Stoch Env Res Risk A. 21 (1), 1 13. Lekkas, T.D., Nikolaou, A. D. 2004. Development of Predictive Models for the Formation of Trihalomethanes and Haloacetic Acids during Chlorination of Bromide Rich Water. Water Qual Res J Can. 39(2), 149 159. Li, X., Zhao, H. Bin. 2006. Development of a model for predicting trihalom ethanes propagation in water distribution systems. Chemosphere. 62(6), 1028 1032. Liang, L., Singer, P. C. 2003. Factors influencing the formation and relative distribution of haloacetic acids and trihalomethanes in drinking water. Environ Sci Technol. 37 ( 13), 2920 8. Liu, S., Zhu, Z., Fan, C., Qiu, Y., Zhao, J. 2011. Seasonal variation effects on the formation of trihalomethane during chlorination of water from Yangtze River and associated cancer risk assessm ent. J Environ Sci. 23 (9), 1503 1511. Malcolm Pirnie Inc. 1992. Water Treatment Plant Simulation Program, Version 1.21, User's manual.Washington, D.C.: U.S. Environmental Protection Agency (EPA). Malcolm Pirnie Inc. 1993. Bay Delta water quality modeling report No. 15 041. Los Angeles: Metropolitan Water District of Southern California. Marquardt, D.W. 1963. An Algorithm For Least Squares Estimation of Nonlinear Parameters. J Soc Ind App Math. (2), 431 441. Millero, F. J., Feistel, R., Wright, D. G., & McDougall, T. J. 2008. The composition of Standard Seawater and the definition of the Reference Composition Salinity Scale. Deep Sea Res. Oceanogr., A 55 (1), 50 72.
99 Morrow, C.M., Minear, R.A. 1987. Use of regression models to link raw water characteristi cs to trihalomethane concentrations in drinking water. Water Res. 21(1), 41 48. Motz, L. H., Sedighi, A. 2013. Saltwater Intrusion and Recirculation of Seawater at a Coastal Boundary. J Hydrol Eng. 18 (January), 10 18. Murgulet, D., Tick, G. 2007. The exten t of saltwater intrusion in southern Baldwin County, Alabama. Environ Geol. 55 (6), 1235 1245. National Cancer Institute. 1976. Report on the carcinogenesis bioassay of chloroform (CAS no. 67 66 3), TR 000 NTIS Rpt No. PB264018, Bethesda, MD. Nikolaou, A .D., Golfinopoulos, S.K., Arhonditsis, G.B., Kolovoyiannis, V., Lekkas, T.D. 2004. Modeling the formation of chlorination by products in river waters with different quality. Chemosphere. 55(3), 409 420. Nokes, C.J., Fenton, E., Randall, C.J. 1999. Modell ing the Formation of Brominated Trihalomethanes in Chlorinated Drinking Waters. Water Res. 33(17), 3557 3568. Obolensky, A., Singer, P. C. 2005. Halogen substitution patterns among disinfection byproducts in the information collection rule database. Enviro n Sci Technol. 39 (8), 2719 30. Oude Essink, G. H. P., van Baaren, E. S., de Louw, P. G. B. 2010. Effects of climate change on coastal groundwater systems: A modeling study in the Netherlands. Water Resour Res. 46 (10), 1 16. Pan, Y., Zhang, X. 2013. Four groups of new aromatic halogenated disinfection byproducts: effect of bromide concentration on their formation and speciation in chlorinated drinking water. Environ Sci Technol. 47 (3), 1265 73. Parinet, J., Tabaries, S., Coulomb, B., Vassalo, L., Boudenne J.L. 2012. Exposure levels to brominated compounds in seawater swimming pools treated with chlorine. Water Res. 46 (3), 828 36. Park, C., Aral, M. M. 2008. Saltwater Intrusion Hydrodynamics in a Tidal Aquifer. J Hydrol Eng. 13(9) 863 873. Plewa, M.J., K argalioglu, Y., Vankerk, D., Minear, R.A., Wagner, E.D., 2002. Mammalian cell cytotoxicity and genotoxicity analysis of drinking water disinfection by products. Environ Mol Mutagen. 40, 134 142.
100 Rasmussen, P., Sonnenborg, T. O., Goncear, G., Hinsby, K. 201 3. Assessing Impacts of Climate Change, Sea Level Rise, and Drainage Canals on Saltwater Intrusion to Coastal Aquifers. Hydrol Earth Syst Sc. 17(1), 421 443. Rathbun, R.E. 1996. Regression equations for disinfection by products for the Mississippi, Ohio an d Missouri rivers. Sci Total Environ. 191(3), 235 244. Richardson, S. D., Plewa, M. J., Wagner, E. D., Schoeny, R., Demarini, D. M. 2007. Occurrence, genotoxicity, and carcinogenicity of regulated and emerging disinfection by products in drinking water: a review and roadmap for research. Mutat Res. 636 (1 3), 178 242. Richardson, S. D., Thruston, A. D., Caughran, T. V., Chen, P. H., Collette, T. W., Floyd, T. L., Schenck, K. M., et al. 1999. Identification of New Drinking Water Disinfection Byproducts Forme d in the Presence of Bromide. Environ Sci Technol. 33 (19), 3378 3383. Richardson, S. D., Thruston, A. D., Rav Acha, C., Groisman, L., Popilevsky, I., Juraev, O., Glezer, V. 2003. Tribromopyrrole, brominated acids, and other disinfection byproducts produce d by disinfection of drinking water rich in bromide. Environ Sci Technol. 37 (17), 3782 93. Rodriguez M.J., Serodes J. Morin M. 2000. Estimation of water utility compliance with trihalomethane regulations using modeling approach. J Water Supply: Res Techn ol; 49(2), 57 73. Rook, J.J., 1974. Formation of haloforms during chlorination of natural waters. Water Treatment and Examination. 23, 234 243. Sadiq, R., Rodriguez, M.J., 2004. Disinfection by products (dbps) in drinking water and predictive models for their occurrence: A review. Sci Total Environ. 321, 21 46. Semerjian, L., Dennis, J., Ayoub, G. 2009. Modeling the formation of trihalomethanes in drinking waters of Lebanon. Environ Monit Assess. 149(1 4), 429 436. Srodes, J. B., Rodriguez, M. J., Li, H., Bouchard, C. 2003. Occurrence of THMs and HAAs in experimental chlorinated waters of the Quebec City area (Canada). Chemosphere. 51(4), 253 263. Shi, H., Qiang, Z., Adams, C. 2012. Formation of haloacetic acids, halonitromethanes, bromate and iodate du ring chlorination and ozonation of seawater and saltwater of marine aquaria systems. Chemosphere 90 (10), 2485 2492. Siddiqui M, Amy G, Ozekin K,Westerhoff P. 1994. Empirically and theoretically based models for predicting brominated ozonated by products. Ozone: Sci Eng. 16, 157 78.
101 Siddiqui, M., Amy, G., Ozekin, K., Westerhoff, P. 1998. Modeling dissolved ozone and bromate ion formation in ozone contactors. Water Air Soil Poll. 108, 1 32. Sivey, J. D., Arey, J. S., Tentscher, P. R., Roberts, A. L. 2013. R eactivity of BrCl, Br 2 BrOCl, Br 2 O, and HOBr toward dimethenamid in solutions of bromide + aqueous free chlorine. Environ Sci Technol. 47 (3), 1330 8. Small, C., Nicholls, R. J. 2003. A Global Analysis of Human Settlement in Coastal Zones. J Coastal Res. 19 (3), 584 599. Sohn, J., Amy, G., Cho, J., Lee, Y., Yoon, Y. 2004. Disinfectant decay and disinfection by products formation model development: chlorination and ozonation by products. Water Res. 38(10), 2461 2478. Summers, S. R., Hooper, S. M., Shukairy, H. M., Solarik, G., Owen, D. 1996. Assessing DBP 88 (6), 80 93. Taylor, P., Chang, E.E., Chao, S.H., Chiang, P.C., Fwu, J. 2013. Chemistry Effects of chlorination on THMs formation in raw water. Toxicol Environ Chem. 56(1 4), 221 225. Terrazas, American City and County. 110 (2), 46 59. Toroz, I., Uyak, V. 2005. Seasonal variations of trihalomethanes (THMs) in water distribution networks of Istanbul City. Desalination. 176(1 3), 127 141 Urano, K., Wada, H., Takemasa, T. 1983. Empirical Rate Equation for Trihalomethane Formation with Chlorination of Humic Substances in Water. Water Res. 17(12), 1797 1802. United States Environmental Protection Agency (USEPA). 1995. Method 524.2 Revisio n 4.1, Volatile Organic Compounds by Gas Chromatography/Mass Spectrometry (GC/MS). Cincinnati, OH. United States Environmental Protection Agency (USEPA). 1990. Method 524.2 Revision 1.0, Determination of Haloacetic Acids in Drinking Water by Liquid liquid Extraction, Derivatization and Gas Chromatography with Electron Capture Detection. Cincinnati, OH. U.S.EPA. Integrated Risk Information System (IRIS). 2005. Online.Washington, DC: National Center for Environmental Assessment.
102 U.S. Environmental Protection Agency. 2006. National Primary Drinking Water Regulations: Stage 2 Disinfectants and Disinfection Byproducts Rule. Fed. Regist. 2006, 71, 387 493. Uyak, V., Toroz, I., Meri, S. 2005. Monitoring and modeling of trihalomethanes (THMs) for a water treatmen t plant in Istanbul. Desalination. 176(1 3), 91 101. Uyak, Vedat, Ozdemir, K., Toroz, I. 2007. Multiple linear regression modeling of disinfection by products formation in Istanbul drinking water reservoirs. Sci Total Environ. 378(3), 269 80. von Gunten U. 2003. Ozonation of drinking water: part II. Disinfection and by product formation in presence of bromide, iodide or chlorine. Water Res. 37 (7), 1469 87. Walker, K.M., Boyer, T.H., 2011. Long term performance of bicarbonate form anion exchange: Remova l of dissolved organic matter and bromide from the St. Johns River, FL, USA. Water Res. 45(9), 2875 2886. Warner, N.R., Jackson, R.B., Darrah, T.H., Osborn, S.G., Down, A., Zhao, K., White, A. 2012. Geochemical evidence for possible natural migration of M arcellus Formation brine to shallow aquifers in Pennsylvania. P Nat Acad Sci USA. 109(30), 11961 11966. Watson, K., Farr, M. J., Knight, N. 2012. Strategies for the removal of halides from drinking water sources, and their applicability in disinfection by product minimisation: a critical review. J Environ Manage. 110 276 98. Westerhoff, P., Chao, P., Mash, H. 2004. Reactivity of natural organic matter with aqueous chlorine and bromine. Water Res. 38(6), 1502 1513. Wu, W., Chadik, P. 1998. Effect of Bro mide Ion on Haloacetic Acid Formation During Chlorination of Biscayne Aquifer Water. J Environ Eng. 124, 932 938. Yang, X., Shang, C., Lee, W., Westerhoff, P., Fan, C. 2008. Correlations between organic matter properties and DBP formation during chloram ination. Water Res. 42(8 9), 2329 2339. Zhai, H., Zhang, X. 2011. Formation and decomposition of new and unknown polar brominated disinfection byproducts during chlorination. Environ Sci Technol. 45 (6), 2194 201. Zhao, Y., Anichina, J., Lu, X., Bull, R. J ., Krasner, S. W., Hrudey, S. E., Li, X.F. 2012. Occurrence and formation of chloro and bromo benzoquinones during drinking water disinfection. Water Res. 46 (14), 4351 60.
103 Zoeteman, B.C.J. 1985. Drinking water and health hazards in environmental perspec tive. Sci Total Environ. 41, 487 503.
104 BIOGRAPHICAL SKETCH Evan Ged graduated m agna c um l aude from North Carolina State University in Raleigh, NC in May 2011, with a BS in e nvironmental e ngineering. In August 2011, he egree in e nvironmental e ngineering at the he plans to practice environmental engineering for CH2M Hill in Dallas, TX.