United States


Environmental Protection
Agency          iewran
?ffice of water
(4601)
                                                      0
                                                      815-R-98_005

                                                  August 1998
- _ ___ ________^

-------

-------
                                                          EPA-815-R-98-005
                                                           August 1998
            EMPIRICALLY BASED MODELS FOR PREDICTING

            CHLORINATION AND OZONATION BY-PRODUCTS:

TRIHALOMETHANES, HALOACETIC ACIDS, CHLORAL HYDRATE, AND BROMATE
                              By:

                 Gary Amy, Mohamed  Siddiqui,
        Kenan Ozekin, Hai Wei Zhu,  and  Charlene Wang

              University of Colorado  at Boulder
                       Number CX 819579
                      Project Officers:

               James Westrick and Hiba Shukairy
          Office of Ground Water and Drinking Water
             U.S.  Environmental  Protection Agency
                      Cincinnati,  Ohio
                         August 1998

-------
                                  DISCLAIMER
      The information in this document has been funded wholly or in part by
the United States Environmental Protection Agency under Cooperative Agreement
 CX 819579 with the University of Colorado- It has been subjected to the
Agency's peer and administrative review, and it has been approved for
publication as an EPA document. Mention of trade names or commercial products
does not constitute1 endorsement or recommendation for use.

-------
                                    FORWARD

      This study was initiated to help the U.S. EPA, as a regulatory agency,
and U.S. water utilities, inpacted by EPA drinking water regulations,
promulgate and meet new standards^orr-dlsinfection by-products  (DBPs). At the
time of this report/ new/revised regulations were being proposed for both
chlorination and ozonation by-products. Most of the analytical and modeling
work was performed over the 1994-1995 time_,frame.
                                      ill

-------
                                    ABSTRACT

       This report documents a series-of statistically-based empirical models
 for use in predicting disinfection by-products (DBFs)  formed during water
 treatment disinfection using chlorine  or ozone.  The.models were created and
 calibrated from a data base derived from bench-scale assessment of a diverse
 range of waters,  including both surface water and groundwater sources.  Each
 model was formulated through multiple  step-wise  regression analysis,  and as
 such takes the form of a multiple regression equation.  After formulation and
 calibration,  model simulations were performed to compare predicted versus
 measured values,  employing the same data base used in model calibration.
 Finally,  each model was validated by using data  derived from the literature.
 The relevant  chlorination DBFs include haloacetic acids (HAAs),
 trihalomethanes (THMs),  and chloral hydrate 
-------
                                   CONTENTS

For award	iii
Abstract	 iv
Figures	 vi
Tables 	viii
List of Acronyms	ix
      1. introduction	  1
            Background	•. . .  1
            Research Obj ectives	  2
            Data Base and Models	  2
            Intended Model Users	  3
      2 . Experimental Methods and Procedures	  4
            Analytical Methods	  4
            Analytical Quality Control	 12
            Bench-Scale Testing Methods	 14
            Statistical Methods	 20
      3. Source Waters and Data Base Summary	 24
            Raw/Untreated Waters	 25
            Coagulated Waters	 28
      4. Haloacetic Acid Models 	 33
            Parameters Affecting Haloacetic Acid Formation	 33
            General Modeling Approach	 37
            Total Haloacetic Acids; Raw/Untreated Waters	 38
            HAA Species; Raw/Untreated Waters	 44
      Coagulated-Water Models	48
            HAA Speciation Models; Br'/DOC as Master Variable	60
      5 . Trihalomethane Models	 68
            Individual Parameter Effects on TTHM and THM Species	 69
            Total Trihalomathanes; Raw/Untreated Waters	 69
            THM Species; Raw/Untreated Waters	 73
            Coagulated Waters	 75
            Effects of Bromide on THM Formation	 86
            Simulation and Validation of TTHM Predictive Models	 89
      6. Chloral Hydrate Models	 97
            Raw/Untreated Waters	 97.
            Coagulated Waters	 99
            Surrogate Correlations between CH and TTHM or CHC13	104
      7 . Chlorine Decay Models	109
            Chlorine Residual Decay Models	109
            DBP Formation versus Chlorine Exposure  (C-T)	114
      8 . Bromate and Ozone Decay Models	118
            Parameters Affecting Bromate Formation	118
            Comparison of Reactor Types	125
            Modeling Efforts	125
            Comparison of True-Batch with Semi-Batch Models	143
            Evaluation of Control Options: Model Simulations	143
            Organo-Br Formation	146
      9. Model Applications	148
            Chlorination By-Product and Chlorine Decay Models	148
            Ozonation By-Product and Ozone Decay Models	149
Re f erences	150
                                      V

-------
                                     FIGURES

 2 .1 Typical GC Chromatogram Showing HAA. Species Peaks	  5
 2.2 Typical GC Chromatogram Showing THM Species and CH Peaks 	  6
 2.3 Typical 1C Chromatogram for Bromide Ion	  8
 2.4 Typical 1C Chromatogram for Bromate Ion	  9
 2.5 Calibration Curve for Bromide Ion.	.".".".".",".".*.".".".".".' 10
 2.6 Calibration Curve for Bromate Ion	......... 11
 3.1 Location of Utilities/Source Waters	*  	 26
 3.2 Results of Alum Coagulation Screening Experiments	 30
 3 .3 Results of Iron Coagulation Screening Experiments .	 31
 4.1 Individual Parameter Effects on THAA Formation: Effects of Chlorine]"""
       pH,  Temperature,  Bromide,  DOC, and Reaction Time	   34
 4.2 Individual Parameter Effects on HAA Species Formation: Effects"of	
 Chlorine,  pH,  Temperature, Bromide,  and Reaction Time	35
 4.3 Predicted versus Measured Values for Raw/Untreated Water THAAs;
       Weight-Based (ug/L)  Model	 41
 4.4 Predicted versus Measured Values for Raw/Untreated Water THAAs;	
 Identification of Individual Sources	          42
 4.5 Predicted versus Measured Values for Raw/Untreated Water THAAs;
       Molar-Based (umoles/L)  Model	 43
 4.6 Overall External Validation  using JMM Data with Raw-Water Modei........ 45
 4.7 External Evaluation of Kinetics  using JMM Data with  Raw-Water Model	46
 4.8 Predicted versus Measured Values for Raw-Water TCAA,  DCAA,  and BCAA	 47
 4.9 Summation  of  Predicted Individual HAA Species vs Predicted
       Raw-Water THAAs;  Weight-Based  Models (ug/L)	t	 49
 4.10 Predicted versus Measured Values of  THAA for Coagulated/Treated	
       Waters Using Combined Alum plus Iron Treated-Water Models	 54
 4.11 Predicted versus Measured Values of  TCAA,  DCAA,  and BCAA for
 Coagulated/Treated Waters	 55
 4.12 Predicted versus Measured Values of  THAA for Coagulated/Treated	
       Waters Using Raw/Untreated Water-Models Combined with * Concept;
       24-Hour  Predictions	"	 57
 4.13 Predicted versus Measured Values of  THAA for Coagulated/Treated	
       Waters Using Raw/Untreated Water-Models Combined with cf> Concept ;
       96-Hour  Predictions	 58
 4.14 Comparison of Predictions from  Treated-Water Models  versus
       Raw/Untreated Water  Models Combined with   Concepts	 59
 4.15  External  Validation for Coagulated Water Model using JMM Data.."."."."."."."." 61
 4.16  Simulated Effects  of  Coagulation on  THAA Formation	  62
 4.17  Fractional-Concentration  Speciation  Models;  24-Hour  Predictions!......  63
 4.18  Fractional-Concentration  Speciation  Models;  96-Hour  Predictions	  64
 5.1  Individual  Parameter Effects on  TTHM  Formation in HMR and VRW Sources..  70
 5.2  Individual  Parameter Effects  on  THM Species Formation in  VRW  Source	71
 5.3  Predicted versus Measured  Values  for Raw/Untreated Water  TTHMs;
      Weight-Based (ug/L)  Model	 74
 5.4  Summation of Predicted THM Species vs  Predicted Raw-Water TTHMs;
      Weight-Based Models  (ug/L)	 76
 5.5 Predicted versus Measured  Values of JTTHM  for Coagulated/Treated"waters"
      Using Combined Alum  plus Iron Treated-Water  Models	 81
 5.6 Predicted versus Measured  Values of TTHM  for Coagulated/Treated Waters*
      Using Combined Alum  plus Iron Treated-Water  Models;
      Individual Sources	         32
 5.7 Predicted versus Measured Values of TTHM  for Coagulated/Treated"waters"
      Using Raw/Untreated Water-Models with  Concept	 84
 5.8 Comparison of Predictions  from Treated-Water Models versus
      Raw/Untreated Water Models with $ Concept	 85
5.9  Fractional-Concentration Speciation Models; 24-Hour Predictions!....... 90
5.10 Fractional-Concentration  Speciation Models; 96-Hour  Predictions	 91
5.11 Overall External Validation using JMM Data with Raw-Water TTHM Model.. 92
5.12 External Validation of Kinetics using JMM Data with Raw-Water
      TTHM Model	 94


                                      vi

-------
 5.13  Simulated Effects  of  Coagulation on TTHM Formation	  95
 5.14  External  Validation of  Alum Treated-Water Model  and Combined Alum
       plus  Iron Treated-Water  Model  using JMM Data	  96
 6.1  Individual Parameters  Effects on CH Formation:  Effects  of Chlorine,
       pH, Temperature,  Bromide,  DOC,  and Reaction Time	  98
 6 .2  Predicted  versus  Measured  Values for Raw/Untreated Water CH	101
 6.3  Predicted  versus  Measured  Values of CH for Coagulated/Treated Waters
       Using Combined  Alum  plus Iron  Treated-Water Models	103
 6.4  Predicted  versus  Measured  Values of CH for Coagulated/Treated Waters
       Using Raw/Untreated  Water-Models Combined with   Concept;
       24-Hour  Predictions	106
 6.5  Simulated  Effects of Coagulation on CH Formation	107
 6.6  Correlations Between CH  and Chloroform (top)  or TTHMs (bottom)	108
 7.1  Predicted  versus  Observed  Chlorine Decay	113
 7.2  THHA as a  Function  of  Chlorine Exposure (C-T)  for Raw/Untreated (top)
       or Treated (bottom)  Waters	115
 7.3  TTHM as a  Function  of  Chlorine Exposure (C-T)  for Raw/Untreated (top)
       or Treated (bottom)  Waters	116
 7.4  CH as a Function  of Chlorine Exposure (C-T)  for Raw/Untreated (top)
       or Treated (bottom)  Waters	117
 8.1  Individual Parameter Effects on  Bromate Formation;  Effects of pH,
       Ozone Dose,  Bromide  Concentration,
       pH Depression/Ammonia  Addition,  and DOC	120
 8.2  Effect  of  Dissolved Ozone  on Bromate Formation	121
 8.3  Effect  of  Reactor Type on  Bromate Formation 	126
 8.4  Predicted  versus  Measured  Dissolved Ozone Using Semi-Batch,
       True  Batch-EPA, and  True Batch-EPA+EBMUD Models	130
 8.5  Predicted  versus  Measured  Bromate Using Semi-Batch,
       True  Batch-EPA, and  True Batch-EPA+EBMUD Models	132
 8.6  Predicted  versus  Measured  Bromate Using Semi-Batch,  True Batch-EPA,
       and True Batch-EPA+EBMUD Models;  without Ammonia	133
 8.7  Predicted  versus  Measured  CT (Exposure Time)  Using
       True  Batch-EPA  and True  Batch-EPA+EBMUD Models	136
 8.8  Predicted  versus  Measured  Bromate Using True Batch-EPA  and
       True  Batch-EPA+EBMUD CT  (Exposure Time)  Models	138
 8.9  Bromate Formation as a Function  of Ozone Exposure (CT) ,-
       Variable Bromide	139
 8.10  Bromate Formation  as  a  Function of Ozone Exposure (CT);
       Constant Bromide	140
 8.11  External  Validation of  Models with Literature Data
       (Data from Table  8.5)	142
.8.12  Predicted Bromate  Using True Batch-EPA Model  versus Semi-Batch Model..144
 8.13  Bromate Control  Options;  Simulations of True  Batch-EPA Model	145
 8.14  Individual  Parameter  Effects on Bromoform Formation (Semi-Batch);
 Effects  of  pH,  Ozone  Dose, Bromide Concentration	147
                                     vi 1

-------
                                    TABLES

2 .1 Minimum Reporting Levels	 13
2.2 Coefficients of Variation (C.V. )  for HAA Species 	 15
2.3 Coefficients of Variation (C.V.)  for THM Species and CH	 16
2.4 Coefficients of Variation (C.V.)  for Br03" and DO3	 17
2.5 Performance Evaluation of EPA Samples	 18
3 .1 Source Water Characteristics:  Raw/Untreated	 27
3.2 Source Water Characteristics:  After Coagulation  	 32
4.1 Predictive Raw-Water Models for Haloacetic Acids (HAA):
      Total HAAs (THAA) HAA Species	 39
4.2 Predictive Coagulated-Water Models for THAA and HAA Species:
      Alum Models	 50
4.3 Predictive Coagulated-Water Models for THAA and HAA Species:
      Iron Models	 51
4.4 Predictive Coagulated-Water Models for THAA and HAA Species;
      Combined Alum plus Iron Models	 52
4.5 Summary of Reactivity Coefficient, , Values for THAAs	 56
4.6 Summary of Fractional-Concentration HAA Speciation Models;
    24-Hour Predictions	65
4.7 Summary of Fractional-Concentration HAA Speciation Models;
    96-Hour Predictions	66
5.1 Predictive Raw-Water Models for Trihalomethanes  (THM):
      Total THMs (TTHM) and THM Species	 72
5.2 Predictive Coagulated-Water Models for TTHM and THM Species:
      Alum Models	 77
5.3 Predictive Coagulated-Water Models for TTHM and THM Species:
      Iron Models	 78
5.4 Predictive Coagulated-Water Models for TTHM and THM Species;
      Combined Alum plus Iron Models	 80
5.5 Summary of Reactivity Coefficient, 
-------
BCAA
Br~
BrO3"
CH
CHBr3
CHBrCl2
CHBr2Cl
CHC13
Cla
DBAA
DBFs
DCAA
D/DBP
DOC
F
HAAs
HAA
MBAA
MCAA
MCLS
03
r
R
SEE
SSE
TCAA
THAA
THMS
TOC
TTHM
UVA

a
                   LIST OF ACRONYMS

Bromochloroacetic Acid
Bromide  (Ion)
Bromate
Chloral Hydrate
Bromoform
Bromodichloromethane
Dibromochloroacetic Acid
Chloroform
Chlorine
Dibromoacetic Acid
Disinfection By-Products
Dichloroacetic Acid
Disinfectants/ Disinfection By-Products  (Rule)
Dissolved Organic Carbon
F-Statistic
Haloacetic Acids
Sum Of TCAA + DCAA + MCAA + DBAA -f MBAA  + BCAA
Sum of TCAA + DCAA + MCAA + DBAA + MBAA
Monobromoacetic Acid
Monochloroacetic Acid
Maximum Contaminant Levels
Ozone
Simple Correlation Coefficient
Multiple Correlation Coefficient
Standards Error of Estimate
Sum Squares Error
Trichloroacetic Acid
Total HAAs (corresponding to HAAg in this report)
Tr i ha1omethanes
Total Organic Carbon
Total THMs
UV Absorbance (@ 254 nm)
Significance
Reactivity Coefficient
                                      IX

-------

-------
                                   SECTION 1

                                  INTRODUCTION

      A number  of past efforts have  focused on development  of predictive
 trihalomethane  (THM) models  for  systems  using chlorine. Most of  the THM models
 developed have  been based on raw/untreated waters as  opposed to  waters treated
 for precursor removal. Less  work has been done in developing predictive models
 for other chlorination by-products such  as haloacetic acids (HAAs) and chloral
 hydrate  (CH), as well as ozonation by-products such as bromate  (BrO3') and
 bromoform.

      The major objective of this research was to develop empirical models  to
 define:  (i)  the kinetics of  disinfection by-product  (DBF) formation during
 chlprination, with additional emphasis on DBFs beyond THMs,  (ii)  the kinetics
 of chlorination DBF formation in waters  subjected to  treatment for DBF
 precursor removal, and  (iii)  the kinetics of brominated DBF formation during
 ozonation, with an emphasis  on bromate.

      As part of the Disinfectants/Disinfection By-Products (D/DBP) Rule
 Cluster, EPA regulations have been proposed with new  restrictive maximum
 contaminant  levels  (MCLs) for total THMs and sum of five HAA species  (HAA5)  of
 80 and 60 ug/L  (0.08 and 0.06 mg/L), respectively; these may be  further
 lowered to 40 and 30 ug/L (0.04  and 0.03 mg/L) as part of a second stage  of
 the regulations. It is possible  that CH  may also be regulated at a later  time.
 Moreover, all utilities with a total organic carbon  (TOC) of greater than 2
 mg/L  (at the point of first  disinfectant application)  must  evaluate
 implementation  of enhanced coagulation for precursor  removal. Thus,'there is a
 strong need  for models capable of predicting THMs and HAAs, with a particular
 need  for models relevant to  coagulated waters. EPA has also proposed a MCL of
 10 ug/L for  bromate. Given the increasing use of ozone, there is  a need for
 predictive models to assess  potential control options such  as pH  depression.

 BACKGROUND

      Empirically-based THM  prediction models have been developed (Amy,  et
 al.,  1987a; Amy et al.,  1987b) which presently form the basis for EPA and AWWA
 sponsored efforts to develop overall DBF formation models.  Whereas these
 original models described the chlorination of raw/untreated water, other
 efforts (Chadik and Amy,  1987; Moomaw et al.,  1993)  have attempted to address
 treatment effects on THM formation kinetics. Only modest progress has been
made  in modeling THM speciation  (Chowdhury,  et al.,  1991).  Recent work
                                   1

-------
 (Siddiqui and Amy, 1993; Siddiqui et al.,  1994) has  focused on developing a
quantitative understanding of ozonation by-products  such as BrO3".

      Major deficiencies of existing models are:  (i) there has been little
work on chlorination DBFs other than THMs,  (ii) only limited work has been
done to describe how treatment  (e.g., coagulation, adsorption, ozonation)
affects THM  (and other chlorination DBF)  formation,  and  (iii) little progress
has been made in modeling the kinetics of  ozonation  by-products.

      In summary, existing empirical models can provide accurate predictions
of the formation of THMs in chlorinated waters as a  function of reaction time.
However, these models have been largely based on raw/untreated waters; when
they have been applied to treated waters,  it is often assumed that the
character of the precursor  remaining after treatment is the same as that
found in the raw water. Our knowledge of the formation kinetics of DBFs other
than THMs, as well as ozonation DBFs such  as BrO3", is sparse. Even less is
understood about THM and J3BP speciation in general and, more specifically,
treatment effects on speciation.

RESEARCH OBJECTIVES

      The major objectives of the proposed research  were to:  (i) develop
predictive models for haloacetic acid formation kinetics (and speciation),
trihalomethane formation kinetics (and speciation),  and chloral hydrate
formation kinetics during free chlorination; and  (ii) develop a predictive
model for bromate formation kinetics during ozonation. A secondary objective
is to ascertain how DBF precursor removal, with an emphasis on coagulation,
affects the formation kinetics of haloacetic acids,  trihalomethanes,  and
chloral hydrate, and the speciation of haloacetic acids and trihalomethanes.

DATA BASE AND MODELS

      A range of natural waters was selected to reflect a diversity of
sources, including surface waters and groundwaters.  These waters were studied
within bench-scale assessments of chlorination and ozonation; chlorination
studies were augmented by coagulation studies to appraise DBF precursor
removal. A large and robust data base was  developed  for statistical analysis
and modeling.

      In this report, we present statistically-based predictive models for
predicting bromate formation kinetics during the ozonation of
bromide-containing waters. Using the selected source waters, a bench-scale
                                    2

-------
parametric  study was  performed.  The  resultant  data were  statistically analyzed
by multiple regression, yielding eolations  for use in  predicting  BrO3" as a
function of ozone  dose, bromide,  dissolved  organic carbon  (DOC),  pH,
temperature, and reaction  time;  other  variables  investigated  included ozone
residual, ammonia, and alkalinity. Ozone  residuals and corresponding  ozone
demands were also  determined, with these  data  also analyzed  to develop models
of ozone decay.

      We also present empirical  models for  predicting  the  formation of
haloacetic  acids  (HAAs), trihalomethanes  (THMs),  and chloral hydrate  (CH)  in
chlorinated drinking  water, based on water  quality parameters  (DOC, pH,  Br",
temperature) and treatment parameters  (C12 dose, reaction time).   (At the time
of this research,  the measurement of only six  HAA species  (HAA6J  out of a
possible nine  (HAA9)  was analytically possible; hence,  models presented for
total HAAs  (THAA)  actually correspond to  HAAg)  . These models were supplemented
with models to predict HAA and THM speciation,  with a  particular  emphasis  on
the influence of Br" and DOC.  Finally,  the models were adapted to the arena of
(alum or iron) coagulation, and  its  effects on chlorination  DBF formation  and
speciation.

INTENDED MODEL USERS

      The models developed through this work are  intended  to be used by
utilities and regulators. The models have relevance in assessing  how both
water quality and  treatment conditions affect  the kinetics and extent  of DBF
formation.  Users can  assess the  influence of changes in water  quality
conditions  (e.g.,  DOC and/or bromide (Br~)) , and evaluate the effectiveness of
changes in  treatment  conditions  in reducing DBF formation. For chlorination
DBFs,  one can assess  chlorine dose and contact  time as treatment variables,
along with  coagulation for precursor removal.  For ozonatibn DBFs
(i.e., Br03~), one  can assess how the ozone  dose and contact  time  associated
with a contactor translate into bromate formation, and the effectiveness of
potential control options such as pH depression and ammonia addition.

-------
                                   SECTION 2

                      EXPERIMENTAL METHODS AND PROCEDURES

   This chapter highlights the key analytical methods and experimental
procedures used in generating the chlorination by-product and ozonation by-
product data bases. Emphasis is placed on measurement of DBFs, non-routine
parameters affecting their formation, and protocols used in the bench-scale
assessments.

ANALYTICAL METHODS

Haloacetic Acids  (HAAs)
   We measured HAA6/  consisting of trichloroacetic acid (TCAA),  dichloroacetic
acid (DCAA),  monochloroacetic acid  (MCAAJ,  dibromoacetic acid (DBAA),
monobromoacetic acid  (MBAA), and bromochloroacetic acid (BCAA). The extraction
of haloacetic acids was based on EPA method 552 which involves an acidic,
salted medium extracted with ether  (MTBE).  The method requires that the
samples be acidified to a pH of less than 0.5 prior to extraction with ether
so that the HAAs are not in their dissociated form. Prior to extraction and GC
analysis, the compounds were derivatized (esterified) with diazomethane to
produce methyl ester derivatives which are more amenable to GC analysis .  A
Hewlett Packard 5890 gas chromatograph {GO with an electron capture detector
(ECD) was used with a DB-5 megabore column. A typical HAA chromatogram is
shown in Figure 2.1.

Trihalomethanes (THMs)
   The extraction of chloroform  (CHC13) ,  dichlorobromomethane (CHCl2Br) ,
dibromochloromethane (CHClBr2) ,  and bromoform (CHBr3) was accomplished  by
liquid-liquid extraction with MTBE using a modification of EPA method  551.
Sodium sulfate was used to decrease the solubility of ether in water and to
increase the partitioning of THMs into the solvent phase. The sample bottles
were filled and sealed in such a way as to ensure that there was no head
space. Method 551 also permits simultaneous extraction and measurement of
chloral hydrate (CH). A Hewlett Packard 5890 GC with an ECD was used with a
DB-1 megabore column. A typical THM and CH chromatogram is shown in Figure
2.2.

-------
      vvaa
WOL
              vvoa
                 vvoa
                                                                             c
                                                                             o
                                                                             0.

                                                                             
o
o

o
£
a

o
a
>
                                                                             CM


                                                                             B)

-------
                             4.467
                   CHBrCI2
           CHCIg
                                             CHBr2CI
                                                   7.738 CHBr3
T	T
0
T	r
                           4             6

                             Retention Time (Min)
T	1	|
 8             10
Figure 2.2  Typical GC Chromatogram Showing THM Species and Chloral
           Hydrate Peaks

-------
 Bromide (Br") and Bromate  fBrQ,"i

    Br" and BrO3"  ion measurements were  accomplished by ion  chromatography (1C)
 using a Dionex  4500i  series system coupled with Al-400 software and an lonPac
 AS9-SC column (EPA method 300).  For bromide ion,  a 2.0 mM  carbonate/0.75 mM
 bicarbonate  eluent was used in conjunction with a flow rate of 2 mL/min and a
 sample size  (injection volume)  of  100  uL.  For bromate ion,  a 40 mM borate/20
 mM hydroxide eluent was used;  the  minimum  detection limit  at the time of this
 research was 2  ug/L.  For samples with  high chloride ion content,  a silver
 cartridge was used to remove chloride  ions prior to 1C analysis for BrO3~.
 Conductivity detector response was almost  perfectly linear (r2  s  0.99)  for
 standards ranging from 25  to 500 ug/L  BrVL and from  5  to 50 ug/L BrO3VL.
 Typical  chromatograms for  bromide  and  bromate are shown in Figures 2.3 and
 2.4,  respectively. Calibration curves  for  bromide and bromate  are shown in
 Figures  2.5  and 2 . 6, .respectively.

 Total  Organic Carbon  (TOC)  and Dissolved Organic  Carbon (DOC) ,

    TOC was measured using  the  combustion infrared method as described in
 Method 5310  B (Standard Methods) with  a  Shimadzu  TOC-5000  analyzer fitted with
 an autosampler.  Samples were filtered  through a pre-washed 0.45 1m nylon
 membrane filter  in order to operationally  define  DOC.  Samples were sparged to
 remove  inorganic carbon after  acidification to pH < 2.

 UV Absorbance

    UV measurements at  254 nm were made using  a Shimadzu UV/VIS  160
 spectrophotometer and  a 1-cm quartz cell.  Analysis  was  conducted  at  ambient  pH
 (typically 6  to  8) . To minimize interferences  caused  by particulate  matter,
 samples  were  filtered  as described above with  respect  to TOC/DOC  analyses.

 Free Ammonia

   Ammonia measurements were performed with an ammonia  ion-selective electrode
using Method  4500-NH3  F (Standard Methods,  17th edition, 1989).
   pH was measured with a pH meter using Method 4500-H* as described in
Standard methods, 17th Edition (1989) .

-------
153304
131884
£ 106904
c
>- 81834
.">
§ 56804
-a
c
£ 31904
8804
-13088
Eluent: 2.0 mM Na2CO3/0.75 mM NaHCO3

1.85







(CI-)

6.6 (S042-)
3.5 (NO,") A
jfc 11
fl A
3.1(Br-)/ /I
jk 1 \ 1 \
/\l\ 11

i .1 i i . i ii i
 1.00   2.00   3.00   4.00   5.00   6.00   7.00   8.00   9.00
                     Elution Time (minutes)

Figure 2.3. Typical 1C Ion Chromatogram for Bromide Ion (Br~).
10.00

-------
 on
I
 c
 o
U
    3.0
    2.0
    1.0
   -1.0
   -2.0
          Eluent: 40 mM H3BO3/20 mM NaOH
                                            2.61  (CD
           2.2 (BrO3")


Unknown Peak  A
       0      0.5     1.00    1.5    2.00   2.50   3.00    3.50    4.00   5.00

                                 Elution Time (minutes)
         Figure 2.4. Typical 1C Ion Chromatogram for Bromate Ion (6103").

-------
  T.2
         Column: AS9-SC
         Eluant: 2.0 mM Na^O 70.75 mM NaHC03
  0.8
S0.6
  0.4-
  0.2
                                                       r -0.9995
     0       50
100      150     200     250      300     350
           Br
                Rgure 2.5. Calibration Curve for Bromide Ion.
                                  10

-------
        Column: AS9-SC
        Eluant: 40 mM H3B03/20 mM NaOH
1.5
0.5
                                                   r  -0.9998
20        40
60
80
                                                        100       120
                Figure 2.6. Calibration Curve For Bromate Ion.
                                 11

-------
 Alkalinity

   Alkalinity analyses were  conducted  by titration using method  2320 as
 described in  Standard Methods,  17th  edition  (1989).

 Turbidity

   Turbidity  was measured with  a turbidity meter using method 2130 as
 described in  Standard Methods,  17th  edition  (1989).

 Residual Ozone

   Ozone residuals were measured using Method 4500-O3 A, the indigo
 trisulfonate  method,  as described in Standards Methods, 17th Edition (1989) .

 Free and Total Chlorine

   Both free  and total chlorine residuals were measured by using Method 4500-
 Cl, the DPD method, as described in  Standard Methods, 17th edition (1989).
 Chlorine demand was calculated by the  difference between the applied dose and
 the measured  residual.

 ANALYTICAL QUALITY CONTROL

   Quality assessment is the process of using external and internal quality
 control measures to determine the quality of the data produced in this
 research. It  includes such items as performance evaluation samples, laboratory
 intercomparison samples and internal quality control samples. Internal quality
 control includes recovery of known additions, analysis of externally supplied
 standards, analysis of reagent blanks, calibration with standards, and
 analysis of duplicates.

   Minimum detection limits or minimum reporting limits (MRLs)  were determined
 by spiking a  series of concentrations  in high purity water and in source
waters examined. The least concentration which was detected above the noise
 level was assigned as the minimum reporting level of that contaminant.  This is
 an average of  several injections. A strict preventive maintenance program was
 in force to reduce instrument malfunctions,  maintain calibration, and to
reduce downtime. The minimum reporting levels/detection limits for each of
 these parameters are summarized in Table 2.1.

                                       12

-------
Table 2.1 Minimum Reporting Levels  (MRLs).
Analyte
MRL
Units
                        Method
CHC13
CHCl2Br
CHClBr2
CHBr3
CH
MCAA
DCAA
TCAA
MBAA
DBAA
BCAA
C12
Br~
Br03"
03
Alkalinity
Ammonia
0.6
0.5
0.5
0.3
0.5
1.2
1.0
1.5
0.5
0.7
1.5
0.1
5.0
2.0
0.05
10
0.1
ug/L
ug/L
ug/L
ug/L
ug/L
ug/L
ug/L
ug/L
ug/L
ug/L
ug/L
mg/L
ug/L
ug/L
mg/L
mg/L
mg/L
EPA551
EPA551
EPA551
EPA551
EPA551
EPA552
EPA552
EPA552
EPA552
E PAS 52
EPA553
4500-C1*
EPA300
EPA300
4500-03*
2320
4500-NH3*
*Standard Methods
                         13

-------
   As a minimum, four different dilutions of the standards were measured when
an analysis was initiated. Reportable analytical results are those within the
range of the standard dilutions used. A minimum of 3-4 replicate analyses of
an independently prepared sample having a concentration of between 5 and 50
times the method detection limit was used.

  When most samples were found to have measurable levels of constituents being
measured, selected analysis of duplicate samples was employed for determining
precision; 10% or more of the samples were analyzed in duplicate. Precision,
reported in terms of coefficients of variation (C.V.), is summarized in Table
2.2 for HAA species. Table 2.3 for THM species and chloral hydrate, and Table
2.4 for bromate (and ozone residual; simultaneous analysis of split samples).

   Data reduction,  validation, and reporting are the final features of a good
QA program. The concentrations obtained were always adjusted for such factors
as extraction efficiency, sample size, and background value. As part of the QC
program, performance evaluation standards provided by the EPA were analyzed;
these results are reported in Table 2.5.

BENCH-SCALE TESTING METHODS

Chlorination

   The protocol for each sample aliquot involved dosing with free chlorine at
a chlorine/DOC ratio (by weight) of 0.5-3.0 mg/mg and incubating for 2 to 168
hrs. A 1 mM phosphate buffer was used to maintain pH at either 6.5, 7.5, or
8.5. A water bath was used to maintain temperature at either 15,  20, or 25 °C.
In the presence of ammonia, the chlorine dose was increased 7.6 times the
concentration of ammonia-nitrogen to provide for breakpoint conditions.

   The chlorine dosing solution was prepared from reagent-grade sodium
hypochlorite and the stock chlorine solution was standardized by the DPD
titrimetric method  (Standard Methods, 1989).  The chlorine concentration of the
dosing solution was between 1.5-2.5 mg/ml, which is about 50 times the
concentration needed in the samples, in order to minimize dilution errors in
the reaction bottles. The chlorination experiments were conducted in glass
serum bottles with teflon septa. The bottles were pre-soaked in 40% sulfuric
acid for 24 hours and washed with phosphate-free detergent and rinsed with
deionized water followed by NOM-free Milli-Q water. The serum vials were
maintained headspace-free throughout the incubation period. After incubation,
an aliquot of the sample was withdrawn for determination of chlorine residual
(by DPD} and estimation of corresponding chlorine demand.
                                      14

-------
Table 2.2 Coefficients of Variation (C.V.) for HAA Species
Compound injection #

CAA 1
2
BAA 1
2
DCAA 1
2
TCAA 1
2
BCAA 1
2
DBAA 1
2
Compound Extraction #

CAA ~~l
2
BAA 1
2
DCAA 1
2
TCAA 1
2
BCAA 1
2
DBAA 1
2
Area
Count
29429
28093
326972
353867
803671
819567
2139600
2250493
1304154
1324020
867278
789652
Area
Count
8242
9150
67661
75671
243040
246181
1161904
1252530
516646
533885
311082
398810
s % C.V.


668 2,3

13447 4.0

7984 1 .0

55447 2.5

9933 0.8

38813 4.7
S % CV


454 5.2

4005 5.6

1571 0.6

45313 3.8

8620 1.6

43864 	 12.4
                           15

-------
Table 2.3  Coefficients of Variation (C.V.)  for THMs  and Chloral Hydrate
Injection:
Compound
CHCIs

CHBrCl2

CHBr2CI

CHBr3

CH


Injection #
1
2
1
2
1
2
1
2
1
2

Area
Count
41842
41637
171257
1 68328
1 36205
1 37836
17684
17507
22335
21871

Concentration s
foo/U
6.697
6.665 0.022627
6.911
6.793 0.083438
6.7130
6.7930 0.056569
2.198
2.176 0.015556
0.9269
0.9080 0.013364

% C.V.

0.339

1.218

0.838

0.711

1.46
Extraction:
Compound
CHCIa

CHBrCl2

CHBr2CI

CHBr3

CH

Extraction
#
1
2
1
2
1
2
1
2
1
2
Area
Count
236217
248703
1307833
1304055
191628
191758
8962
8874
203136
206052
Concentration s
(W/L)
41.14
43.90 1.9516
69.10
68.86 0.1697
8.667
8.673 0.004243
0.911
0.902 0.006364
5.678
5.759 0.057276
% C.V.

4.590

0.246

0.0490

0.702

1.00
                                 16

-------
Table 2.4 Coefficients of Variation  (C.V.)
          For Bromate and Dissolved Ozone.
Analyte
Bromate
Bromate
Ozone*
Ozone*
Concentration
25 ug/L
10 ug/L
0.50 mg/L
0.25 mg/L
# of Measurements
2
3
2
' 3
C.V. {%)
0.91
1.11
0.15
0.25
*Based on replicate samples taken from true-batch reactor
                               17

-------
     Table 2.5. Performance Evaluation of EPA Samples (9/2/93)
Analyte
Trihalomethanes
CHC13
CHBrCU
CHBr2Cl
CHBr3
Chloral Hydrate
Haloacetic Acids
MCAA
MBAA
DCAA
DBAA
BCAA
Sample #
I
1
1
1
1
2
1
1
1
1
1
1
Reported Value
(ug/L)
107.1
29.1
32.9
22.4
22.7
5.8
80.3
7.4
8.6
17.7
16.6
30.4
EPA True Value
(ug/L)
83.8
22.4
26.4
17.9
17.1
4.6
72.0
11.8
12.9
15.4
14.1
17.8
Except THMs and BCAA, all results were found to be acceptable. All of
 the THM species were off by the same factor indicating a dilution error
                               18

-------
Coagulation

   The experimental plan for the coagulation experiments was designed to
provide data on DOC removal using various alum  (A12 (SO4)3-18H2O) and ferric
chloride (FeCl3-6H2O)  doses at ambient pH to remove 25-50% of the DOC. Reagent
grade alum and ferric chloride were applied at different doses to each water
at 20 °C. A conventional jar test apparatus was used with the following
conditions: rapid mixing at 100 rpm for one minute, flocculation at 30 rpm for
30 minutes, and one hour of settling. Settled aliquots were  filtered through
prewashed 0.45 um filters for further analytical characterization.

Ozonation

   Two modes of bench-scale ozone application were employed, semi-batch and
true-batch.  The general system consisted of an OREC O3V5-O  (Ozone Research &
Equipment Corporation)  0.25 Ib/day generator and a 0.5 liter capacity glass
reactor (washing bottle) with a glass frit. Ozone was generated from pure
oxygen. Samples were buffered with a 1 mM phosphate buffer.

   Semi-batch experiments involved continuous application of ozone and carrier
gas admitted to a batch of water within the reactor. Contact time was
controlled by the mass  flow rate into the reactor; thus, applied dose  (mg/L)
was a function of mass  application rate  (mg/L-min) and application time  (min).
Applied and utilized ozone were determined by the classical  iodometric method
(Standard Methods, 1989). Typical transfer efficiencies were 30 to 60 %, with
an average of 50 %. Dissolved ozone residuals  (DO3)  present after cessation of
ozone application were  measured by the indigo method  (APHA,  1989)

   True-batch experiments were conducted by first generating a concentrated
ozone stock solution  (30-40 mg/L) by exhaustively ozonating  Milli-Q water at 2
to 3 °C.  Aliquots of the stock solution were then applied to a sample of raw
water to achieve a final initial ozone concentration  (typically in the range
of 3 - 5 mg/L, confirmed by an initial measurement). In this procedure, the
applied dose is equal to the transferred dose.  Dilution of the raw water
constituents by the ozone stock aliquot must be considered;  dilution was
generally kept to below 10 %. In contrast to the semi-batch  ozonation where
ozone is applied over a period of time  (5 to 15 minutes) and reactions can
occur during ozone application, true batch experiments involve introduction of
100% of the aqueous ozone to the system at time zero. The DO3  measured after
a designated reaction time corresponds to one point along the overall ozone
decay curve.
                                      19

-------
STATISTICAL METHODS

   For each of the DBF groupings; chlorination by-products including
haloacetic acids, chloral hydrate, and trihalomethanes; and ozonation by-
products represented by bromate; we employed a similar approach to statistical
analysis. In our previous work, we used a comparable approach in both data
base design and statistical analysis. It is noteworthy that our work on the
development of THM prediction models  (Amy, et al., 1987a; Amy et al., 1987b)
presently forms the basis for EPA and AWWA sponsored efforts.to develop
overall DBF formation models. The following sections discuss how the data base
was generated, and statistical approaches used in model development.

Data Base Design

   While models and their formulations can sometimes have a theoretical basis
in terms of a defined functionality {e.g., first order reaction kinetics),
empirical models are often necessary if the underlying mechanisms are too
complicated. When this approach is taken, caution must be exercised in terms
of boundary conditions since empirical models are not intended to be severely
extrapolated but rather used over limited ranges of the variables (generally
corresponding to ranges within' the data base). The requisite data base must be
designed to reflect important variables, ranges of variables, and interactive
effects among variables. Appropriate replication is needed to address
experimental error so that true parameter effects can be discerned.

   In developing a data base, one may choose a factorial design or an
orthogonal design. A factorial design involves defining a comprehensive set of
experiments within the context of a full matrix. For example, one could
perform a 3 x 4 factorial design to study rats exposed to three different
poisons and four different treatments. In this case, the complete matrix,
without replication, would involve 12 experiments; initially, both factors
would be considered of equal interest, and the possibility that the factors
interact is acknowledged. In an orthogonal design, only one parameter is
varied at a time while other parameters are maintained at some designated
"baseline" condition. The above "rat experiment" would require only 6
experiments using this approach.

   The problem with a full factorial design is the large number of experiments
required. For example, in the ozonation by-product, chlorination by-product,
and precursor-removal tasks being proposed herein, a full factorial design
would require over ten- thousand cases (n). A compromise can be reached by
                                      20

-------
 employing a partial (fractional)  factorial design.  An attribute of this
 approach is the ability to discern interactive effects.

    One can argue that  an orthogonal design corresponds to one version of a
 fractional (partial)  factorial design.  While an orthogonal design elucidates
 individual parameter  effects,  it  does not identify interactive effects. One
 can argue,  however, that an additional  set of randomly selected experiments
 can help fulfill this  need.  On the other hand,  a strictly random approach to
 identifying a fractional number of experiments from within a full factorial
 matrix leads to a "haphazard"  data base in which both individual-parameter and
 interactive effects are difficult to discern.

    We  have elected to  emphasize an orthogonal design with selected additional
 experiments performed  to elucidate interactive effects;  thus,  in actuality,
 the data base reflects some  movement toward a partial factorial design.

 General Modeling Approach

    Step-wise multiple  linear and  multiple nonlinear regression were used to
 develop models,  with a given DBF  (e.g.,  bromate)  designated as the dependent
 variable (Y).  Various  independent variables (X)  included both water quality
 (e.g.,  pH)  and treatment variables (e.g.,  ozone dose).

    We  attempted to 'develop models assuming various  mathematical formulations,
 including:

    Linear Models:  Y =  b0 + bjXj  +  ...

    Logarithmic  Models:  log Y =  log b0 + bt log X±  +  ...
       (equivalent  to power function:  Y  =  lO^fXi)1*1 ...)

    Nonlinear Models: Y =  b0 + biX^ +  ...

   A PC-based software package, SPSS  (Statistical Package  for  the  Social
Sciences), was  used for  statistical analysis. Statistical  fit  was  defined
through  examination of various  statistical  parameters, including SSE  (sum
squares  error),  SEE (standard error of estimate), the F  statistic,
significance  (a), and  R2 (multiple coefficient of determination) for linear
regression. A rigorous discussion  of  statistical parameters  (SSE,  etc.)
associated with each model appears in three project-related references
 (Ozekin, 1994; Wang, 1994; and  Zhu, 1995),  representing  three
                                      21

-------
 dissertations/theses  that  evolved from,  and were  funded  by,  this  cooperative
 EPA  agreement.

   Using  SPSS,  a  correlation  matrix was  first  developed  to elucidate  simple
 linear  correlations between the  dependent  variable and each  of  the  independent
 variables;  and  between  the independent variables  themselves.  Such a matrix
 provides  preliminary  insight  into the relative importance of the  independent
 variables as well as  co-linearity between  them. Correlation  matrices  were also
 used to examine relationships between transformed variables  (e.g.,  log Y vs
 log  X).

   "Step-wise"  multiple regression places  independent variables into  the
 equation  in order of  their partial correlation coefficients  with  the  dependent
 variable. Thus, the most important predictive  parameters are identified in the
 process;  this is  important because it is advantageous to keep the number of
 predictor-parameters  (each requiring an analytical measurement) to  a  minimum.
 Moreover, if two  independent  variables are  themselves correlated,  the most
 important parameter (based on its partial  correlation coefficient)  will be
 placed  into the equation first and the partial  correlation coefficient' for the
 remaining parameter will be adjusted downward  accordingly. Thus,  using the
 stepwise  approach, it is unlikely that two correlated independent parameters
 will be included  in the final equation. A  "tolerance" can be  specified (e.g.,
 AR2)  to  dictate when stepwise  inclusion ceases.

   The  above SPSS  efforts  correspond to model  calibration. Subsequent model
 testing involved  developing scatterplots of predicted versus  measured values,
 as well as  performing a sensitivity analysis to ascertain how predicted
 dependent-variable values  responded to a range  in specified values  for each
 independent parameter. Model  validation involved obtaining source waters other
 than those  used in developing the  equations to  create additional data based on
 a random  set of experimental conditions; model  simulation was ascertained by
 comparing experimentally-derived kinetic curves (concentration versus time)
versus predicted curves. We also evaluated the ability of the models to
predict pertinent DBF data found  in the literature.   A key concern  in model
validation  testing is the  "robustness" of a model; i.e.,  its ability to make
accurate predictions under extreme experimental conditions.  In testing the
 "robustness" of a model, boundary conditions were also tested in terms of
independent data taken from the literature.
                                      22

-------
Haloacetic Acid, Trihalomethane, and Chloral Hydrate Predictive Models

   Step-wise multiple linear and multiple nonlinear regression were used to
develop models with either total HAA (THAA), chloral hydrate  (CH), or total
THMs (TTHM) designated as the dependent variable  (Y). Independent variables
(Xi)  included DOC (and/or UV absorbance),  bromide concentration,  pH,
temperature, chlorine dose  (and/or utilized chlorine), and reaction time.
Moreover, we developed submodels to predict individual HAA species and THM
species  (Chowdhury, et al., 1991). As above, we examined linear models,
logarithmic models, and nonlinear models.

HAA, THM, and CH Predictive Models Accounting for Precursor Removal

   These "submodels" were similar in format to the raw/untreated water models
discussed above for HAA and CH. Additional modeling features highlighted
precursor reactivity, with delineation of a reactivity coefficient, 0, such
that an adjustment of 4>(DOC) can account for a different reactivity of the
precursor  (DOC) pool of material. A priori, it was expected that  would
likely range from 0 to 1.0 for coagulant-treated waters. Another important
modeling consideration was the effect of an increased ratio (after
coagulation) of Br'/DOC on HAA and THM  speciation.  (Besides the BrVDOC ratio,
The Br"/Cl2 ratio also affects speciation).

Br oma t e_ Pr edi c tive^Models

   As above, step-wise multiple linear and multiple nonlinear regression were
used to develop models, with bromate formation designated as the dependent
variable (Y). Independent variables tX) included DOC  {and/or UV absorbance),
bromide concentration,  pH, temperature, ozone dose, and reaction time.
Dissolved ozone residual  (DO3),  ammonia,  alkalinity,  and peroxide were also
considered as additional Xi parameters.
                                      23

-------
                                     SECTION  3

                       SOURCE WATERS AND DATA BASE SUMMARY

        The data base created through this  research was derived from a broad
  range of natural  waters acquired from throughout the United States.  Single
  samples  of twelve waters were  obtained specifically as part of this  EPA study.
  An  additional  four waters were obtained as  part  of  a related study sponsored
  by  the East Bay Municipal Utilities District  (EBMUD).  One  additional  water was
  obtained as part  of an  American Water Works Association Research  Foundation
  (AWWARF)  study. All water samples obtained  {EPA  and EBMUD)  corresponded to
  raw/untreated  aliquots  of source waters used by  operating  water treatment
  plants under the  jurisdiction  of cooperating utilities. The sequence/ordering
  under which samples were  obtained did not take into account any considerations
  of  seasonally; thus, while they capture source-related differences,  they  do
  not reflect  seasonal variations.  The  EPA sources included 10  surface waters
  and 2 groundwaters.  The EBMUD sources were all surface waters while the AWWARF
  source was a groundwater. Of these twelve sources, eight (all  surface waters)
 were evaluated within the coagulation part of this study.  A total of thirteen
 waters, eight EPA waters  (6 surface waters and 2 groundwaters), all four EBMUD
 waters, and the one AWWARF wate'r, were evaluated as  part of the ozonation part
 of this study.  The EBMUD sources and the AWWARF source were only studied
 within the context of the ozonation component  of the study. The specific
 waters evaluated and their abbreviation identifiers  (used  hereafter)  are
 summarized below:

 EPA  Sources:

 • Silver  Lake (CO);  SLW
 • State Project Water (CA); SPW
 • Brazos River  (TX);  BRW
 • Susquehanna River  (PA);  SRW
 * Burton Groundwater  (MI); BGW
 • Manhattan Groundwater  (KS) ,- MGW
 • Harwood Mill Reservoir  (VA) ;  HMR
 • Palm Beach Reservoir (FL);  PBW
 • Ilwaco Reservoir (WA);
 • Verde River (AZ); VRW
• Salt River (AZ);  STW
• Sioux River (SD); SXW
                                      24

-------
 EBMUD Sources:
 •  Pardee  Reservoir {CA);  EIS
 •  San Leandro  Reservior  (CA);  ESL
 •  Bixler  Reservoir/Mokelumne Aqueduct  (CA);  EWC
 *  San Pablo  Reservoir  (CA) ; EES
 AWWARF Source:

 • Teays Aquifer  (IL); TYA

       The geographical  distribution of the source waters is portrayed in
 Figure 3.1.  The  EBMUD sources are all shown clustered in northern California.
 The EPA source shown in southern California actually reflects Sacramento River
 Delta Water  transported via the California State Project to the Metropolitan
 Water District.

 RAW/UNTREATED WATERS

       Important  characteristics of the raw/untreated source waters are shown
 in  Table  3.1.  It can be seen that the selected sources cover a wide range of
 important water  quality characteristics.  Here,  both DOC and UVArserve as
 indices of (organic)  DBF precursors present.  The ratio of UVA/DOC,  the
 specific  absorbance,  indicates the character  of the precursor DBP material
 present;  a higher ratio reflects a greater humic-substances content.  Bromide,
 of  course, represents the inorganic precursor whose collective action with the
 organic precursor is manifested in DBP formation.  Based on a recent estimate
 of  a national  average of almost 100 ug/L  (Amy et al.,  1994),  the  ambient Br'
 levels range  from below to above average;  it  is important to note that ambient
 conditions represent the baseline condition 'for bromide within the  orthogonal
 experimental matrix.

      Background levels  of ammonia are  important insofar as  they  exert a free
 chlorine  demand,   and may impact  bromate formation.  The  ambient pH conditions
 shown do  nothing more than provide an  indication of  the  actual pH levels
 expected  under water treatment  conditions;  ambient pH was  not part  of  the
 orthogonal matrix. Turbidity here  is primarily  important  in  terms of its
 effect on coagulation and  its ability to remove  DBP precursors. Alkalinity
 affects both the  coagulation process as well as  ozone chemistry; When water
 samples were adjusted to reflect the pH conditions specified in the orthogonal
matrix, these adjustments were accompanied by changes in alkalinity.

                                      25

-------
                                                •H
                                                aJ
                                                •H
                                                c
                                                o
                                                -H
                                                JJ
                                                (0
                                                O
                                                O
                                                CT)





                                                 i-l





                                                •H
26

-------
Table 3.1 Source  Water Characteristics:  Raw/Untreated
Raw/Untreated
Water

Sources
SLW
SPW
BRW
SRW
BGW
MGW
HMR
PBW
ISW
VRW
STW
SXW
Date

Selected for
4/92
6/92
10/92
11/92
1/93
1/93
2/93
4/93
5/93
7/93
6/93
7/28
DOC
(mg/L)
EPA Study
7.15
4.19
3.54
4.18
1.20
2.44
5.73
10.6
2.86
3.67
4.36
10.55
UVA
(cm'1)

0.258
0.169
0.198
0.129
0.010
0.161
0.223
0.280
0.102
0.102
0.099
0.318
PH


7.0
8.2
8.2
7.5
7.7
7.3
7.1
7.8
6.5
8.4
8.2
8.1
NH3-N
(mg/L)

0.25
0.12
0.21
0.14
0.12
0.69
0.21
0.10
0.065
0.09
0.145
0.25
Alk
(mg/L)

15
83
153
43
186
205
50
99
14
156
140
248
Bromide
(ug/L)

7
312
250
50
143
206
40
97
83
71
54
68
Turbidity
(f^TU)

1.1
0.3
20.0
0.9
0.2
2.2
2.5
0.3
0.65
5
20.0
0.94
EBMUD Sources
EIS
ESL
EWC
EES
AWWARF
TYA
3/93
3/93
3/93
3/93
Source
2/93
6.3
5.1
2.1
4.9

3.0
0.24
0.14
0.04
0,14

0.12
6.4
7.8
8.6
7.6 .

8.6
0.11
0.05
0.00
0.13

0.79
142
185
44
104

330
90
28
7
24

90
6
3.1
1.3
12

1.0
                           27

-------
    The orthogonal matrices employed  for  the  raw/untreated water,  coagulated
water, and ozonation work are  summarized  below (*  =  baseline condition):

Chlorination  Conditions; Raw/Untreated Waters:

• DOC = ambient
• C12/DOC  =  0.5,  1.0*,  1.5,  2.0,  and 3.0 mg/itig
• pH = 6.5,  7.5*, 8.5
• Temperature = 15,  20*, 25 °C
• Br"  = ambient*,  amb.  + 100 ug/L,  amb. +  200  ug/L, amb.  + 300 ug/L
* Time = 2,  12, 24,  48, 96,  168 hrs

Chlorination  Conditions; Coagulated Waters:

• DOC = ambient
• C12/DOC  =  1  and 3 mg/mg
• pH = 7.5
• Temperature = 20 °C
• Br"  = ambient
• Time =2,  12, 24,  48, 96,  168 hrs

Ozonation Conditions:

• DOC = ambient
• O3/DOC = 0.5,  1.0*,  2.0 mg/mg (transferred)
• pH = 6.5,  7.5*, 8.5
• Temperature = 15,  20*, 25 °C
• Br"  = ambient*,  amb.  + 100 ug/L*,  amb. + 200 ug/L
   (* = amb. or amb. +  100 ug/L for  Br" > 80 or <. 80 ug/L, respectively)
• Time = 1,  5, 10, 20, 30,  60 min
• NH3/O3 = 0*, 0.35,  0.50 mg/mg (addition)

COAGULATED WATERS

      Screening experiments were performed to evaluate the removal of  DBF
precursors by alum or  iron coagulation; experiments  were done at ambient pH,
with pH allowed to drift upon  coagulant addition.  The major objective  was  to
determine a coagulant  dose that would provide a targeted DOC reduction within
the range of  25 to 50  %. Upon  defining an appropriate dose,  a larger batch of
coagulated water was produced  for assessment  of Chlorination by-products.
                                      28

-------
      This targeted range was selected because it approximately encompasses
the range of DOC removal expected at plants either optimized for turbidity
removal as opposed to plants which practice enhanced coagulation. Draft Stage
1 of the D/DBP Rule specifies that plants with a TOC of greater than 2 mg/L
must evaluate enhanced coagulation  (EC). Required removals are specified as a
function of initial TOC and alkalinity (a 3 x 3 matrix),  with TOC removals
ranging from 15 to 50 % (Step 1). If these removals are not attained, enhanced
coagulation is defined by a "point of diminishing returns" corresponding to a
ATOC/alum of 0.3 mg/L/10 mg/L (Step 2).

      Figures 3.2 and 3.3 summarize the results of coagulation screening
experiments. The graphs show DOC versus dose, DOC removal (%) versus dose,
ADOC versus dose, and ADOC/Adose versus dose. For comparative purposes, it can
be seen that DOC removals observed in four of the eight waters reflected the
attainment of enhanced coagulation by alum, based on the slope criterion and
the assumption that DOC ~ TOC.

      In several cases, the targeted DOC reductions (% removal or point of
diminishing returns) were not achieved over the range of doses evaluated, 0 to
100 mg/L. In these cases,  a pragmatic selection of targeted dose was made. The
final characteristics of coagulated waters are shown in Table 3,2 Selected
coagulant doses ranged from 25 to 100 mg/L; iron provided slightly better DOC
removal than alum. It is noteworthy that bromide was virtually conservative
through the coagulation process.
                                      29

-------
                              x a. ^ w v •*•

                              HWi
                               «oov/ooav
 (0(QCOTCL>a)CO

  HHIH
mmtoxiL^www

H   HHt!
                                                  (0

                                                  £
                                                  e

                                                  E
                                                  h.
                                                  0
                                                  a
                                                  x
                                                  UJ

                                                  a

                                                  JE
                                                  "c
                                                  9
                                                  «

                                                  U
                                                  (0

                                                  c
                                                  o
                                                 O)
                                                 a
                                                 o
                                                 U

                                                 E
                                                 O
                                                 3
                                                 O)

                                                 ii
N  O  
-------
a cc cr 2 CD tc
01 CD to z o. >
(0
H   HIM
                               5555
                       03.

                       HHIMI
                      ui
                                            5 8
                                            « O
                                           o

                                            ffl 0)
                                            in tu

                                            ^ "
                                           O D O
                                           n
                                            >  •=

                                               a
                                               X
                                               ui
                         0   coco>

                HIHH
                                                 O)

                                                 s
                                                 *c
                                                 O
                                                 O

                                                 O
                                                 CO
                                                 D)
                                                 to
                                                 O
                                                 O

                                                 c
                                                 O
                                                 CO

                                                 "5
                                                 (A
                                                 0)
                                                 tc
                                                n

                                                o
                                                k.
                                                3
                        31

-------
               Table 3.2. Coagulated Waters - Water Quality
After Coagulation
Water

SPW

BRW

SRW

HMR

PBW

ISW

VRW

STW

SXW

Coagulant

Alum
Iron
Alum
Iron
Alum
Iron
Alum
Iron
Alum
Iron
Alum
Iron
Alum
Iron
Alum
Iron
Alum
Iron
Dose*

70
60
100
100
65
65
45
45
100
100
25
25
75
75
50
50
100
100
pH

7.4
7.8
7.2
7.2
6.8
6.2
7.1
7.0
7.8
7.7
5.5
4.8
7.4
7.3
7.5
7.4
7.4
7.5
DOC
(mg/L)
2.61
2.58
2.72
2.63
2.56
2.55
4.29
4.23
4.60
4.20
1.0
1.03
2.56
2.35
2.91
2.76
7.77
7.63
NH3-N
Xrng/L)
0.05
0,07
0.01
0.01
0.03
0.01
0.08
0.10
0.05
0.02
0.015
0.055
0.06
0.03
0.11
0.07
0.19
0.21
UVA
(cm-1)
0.115
0.104
0.039
0.055
0.030
0.021
0.078
0.108
0.075
0.073
0.016
0.039
0.051
0.043
0.080
0.078
0.215
0.196
Bromide
(H9/L)
306
308
245
245
45
44
36
37
n/a
n/a
n/a
n/a
n/a
n/a
n/a
n/a
n/a
n/a
* mg/L as AI2(SO4)3-18H2O or FeCI3-6H2O.
n/a: not available.
                                             32

-------
                                   SECTION 4

                             HALOACETIC ACID MODELS

       There is a need to model both the formation of total haloactic acids
 (THAA)  as well as that of haloacetic acid (HAA)  species.  (As previously
 mentioned,  THAA corresponds to HAA6  in  this report).  In addition, the effects
 of treatment (e.g.,  coagulation)  on subsequent HAA formation after
 chlorination need to be elucidated through a model. The primary purpose of
 this chapter is to develop and present a modeling scenario for assessing the
 formation and control of HAAs when free chlorine is used  as a disinfectant.

 PARAMETERS AFFECTING HALOACETIC ACID FORMATION

       Various water quality (pH,  temperature,  DOC,  Br~) and treatment
 conditions (C12) affect the total yield,  formation  kinetics, and speciation  of
 HAAs (Figures 4.1 and 4.2,  24-hour reaction  time).  Results of the effects of
 various parameters on the formation of total  haloacetic acids (HAAg) and
 individual HAA species are discussed under the following  separate headings.
 Experiments were conducted to evaluate the effect of pH,  temperature, bromide,
 DOC,  chlorine dose,  and reaction  time on the  formation of THAA and  HAA
 species.  These were later augmented by experiments  to assess  the effects  of
 coagulant type and dose on HAA:formation.  Lower  levels of HAAs were observed
 in the  BGW source,  a result attributable to  the  relatively low DOC  and  UV
 absorbance  which represent HAA precursors. The highest levels  of HAAs were
 generally found in the SXW source,  a result attributable  to  the relatively
 high DOC,  and the  highest UV  absorbance observed.

 Effect  of pH

      Figure 4.1 shows  pH effects  on THAA formation for two source  waters  (HMR
 and MGW). For one  source,  a trend  of decreasing  THAA with increasing pH is
 seen; for another  source, pH-dependent  variations appear  to be minor. These
 differences  can be attributed to the effects of  pH  on individual HAA species
 (Figure 4.2). TCAA was  found  to strongly decrease on increasing the pH  from
 6.5 to  8.5 while DCAA and CAA {not shown in figure)  were  relatively
 insensitive  to pH. Brominated species,  BCAA, DBAA and BAA, all  slightly
 increased with pH. Reckhow and Singer  (1984) observed a decrease in TCAA and
DCAA with increasing pH.  There is little  information in the literature  on pH
effects on brominated HAA species. Since TCAA and DCAA are the  dominant
species, a model which captured their behavior will  also  likely do well in
simulating THAA  (i.e., HAA6) .
                                      33

-------
    la
350


300





200


iso :


wo -
       so
                                       -MOW
                       1.5   2

                       C^/DOC
     220


     200


     1M


  3  1*0

  a.


  1
  =E  120


     100


     so
                  18     30     33

                      Temp(°C)




i
3
is



320 1

200 -
ISO '
160 •
140 '
130 '
100 '
to •


	 _• 	 • — o— HMR
. 	 5




— .— MGW
	 -~ T^1 *

                                                               300    300    400

                                                                   Bf (jig/L)
                                                                                       500    COO
                                                                    SO       100       ISO

                                                                        Reaction Time 
-------
   35
30 *
25 •
20 •
15 '
 to •
   60
   50-
   40-
(0  30 -
   20-
    10
70


60 -


so -




30 -


20 -


10 -


 0
           0£
                                                  80


                                                  70 '

                                                  60 -


                                               I  so,
                                               (I

                                               I  40 •
                                               m

                                               S  30 '


                                                  20 •


                                                  10 '


                                                   0
                       1.5     2     ZJS     3
                        OJ.-DOC
                                                  50 '
                                                  40'
                                                  30-
                                                  20 •
                                                   10
                                                    200
      14      16     t8     20     22
                     Temperature PC)
                                   24
                                          26
                                                  60
                                                  50'
                                                  40-
                                                0}
                                                5 so i
                                                  20-
                                                  10
               50         100        150

                      Reaction Time (hr)
                                          200
                                                            300
                                                                    400
                                                                            500
                                                                                   •OCAA

                                                                                   •TCAA

                                                                                   -BCAA

                                                                                   •DBAA
                                                                                    -OCAA

                                                                                    -TCAA

                                                                                    •BCAA

                                                                                    •OBAA
                                                                                    600
                                                                                           700
                                                                                      DCAA
                                                                                      TCAA
                                                                                      BCAA
                                                                                      DBAA
                                                    0.06     0.08
                                                                0.1     O.t2     O.t4

                                                                  Bf/DOC (mg/mg)
                                                                                      0.16    0.18
         Fig. 4.2. Individual Parameter Effects on  HAA Species  Formation in BRW
                  Sources: Effects of  Chlorine, pH, Temperature,  Bromide  and
                  Reaction Time (All other Parameters held at Baseline Conditions)
                                                35

-------
Effect of Bromide Concentration

      The concentration of bromide ion (Br~)  in raw water is a significant
factor in the formation of chlorination by-products such as HAAs and THMs. Br"
influences both the total HAA  (and THM) yield as well as the species
distribution of chlorine and bromine-containing species when chlorine oxidizes
the bromide to hypobromous acid  (HOBr), which behaves in a manner analogous to
hypochlorous acid (HOCl). Generally, HOBr is a more effective substitution
agent than HOCl, while HOCl is a better oxidant (reduced to Cl~) .  HOCl is
typically present in great abundance relative to HOBr. Bromine substitution is
favored over chlorine, even when chlorine is present in large excess compared
with the Br~ concentration.  Any bromide present will immediately be oxidized
by HOCl/OCl" to  HOBr/OBr". Therefore, if the HOBr/OBr" is involved in an
oxidation/reduction reaction and Br* is reduced to Br", it would be rapidly
reoxidized to Br* with an excess of HOCl/OCl".

      Increased amounts of BCAA and DBAA were generally observed at higher
levels of bromide; TCAA and DCAA decreased with an increase in bromide
concentration (Figure 4.2). The net effect was generally a slight increase in
THAA with increasing Br",  as shown in Figure 4.1.

      The ratio of Br'/DOC for raw and treated waters is the parameter most
influential in controlling HAA speciation basis (the ratio of an individual
species to THAA). As the ratio increases, a shift to the more bromo-
substituted species occurs  (Figure 4.2).  The ratio of Br"/Cl2 also  influences
speciation.

Effect of Chlorine Dose

      The specific chlorination conditions affect both THAA and HAA species
formation. In our work, we have elected to represent chlorination conditions
through the use of the C12/DOC ratio,  whereby chlorine dose is normalized to
precursor  (DOC)  concentration.

      Figure 4.1 shows the results of changing the chlorine to DOC ratio on
THAA formation for two source waters. Figure 4.2 shows effects on individual
HAA species. The results are supported by the concept that an increase in the
chlorine dose results in a decrease in the total bromide to chlorine ratio
(Br'/Cl2),  and as the chlorine  dose  increases,  speciation shifts to the
chloro-substituted. species, THAA-C1.

                                      36

-------
 Effect of Temperature

       Figure 4.1 shows the impact of temperature on THAA formation.  The
 formation of TCAA,  DCAA,  and BCAA increased with temperature (Figure 4.2);  for
 DBAA,  however,  temperature had little effect.

 Effect of DOC

       A clear correlation was found between THAA and DOC (Figure 4.1).
 Speciation effects  were manifested through the Br"/DOC ratio. UVA provided
 poorer precursor-related  prediction capabilities than DOC;  because of the
 colinearity between DOC and UVA,  UVA was excluded from the  HAA models.

 Effect of Reaction  Time

       The kinetic response of THAAs is a composite effect of the effects  of
 reaction time on individual HAA species.  Figure 4.1 shows the results of  THAAs
 formation for two source  waters,  at varying chlorination reaction times.  These
 THAA kinetic curves show  the composite effects of individual HAA species  which
 form at different rates (Figure 4.2).  Generally,  DBAA increases to a plateau
 after  about 24  hours and  then remains relatively unchanged  thereafter.  TCAA,
 DCAA,  and BCAA all  increase continuously with  increasing reaction time  up to
 about  168 hours.  The formation of the bromo-substituted  DBFs (THAA-Br)  is
 generally  faster than the chloro-substituted  (THAA-Cl).  Thus,  for the  first
 24 hours,  the molar ratio of the  THAA-Br to the THAA-Cl  increases with
 increasing reaction time.  After 24  hours,  there is a decrease in this ratio,
 indicating a shift  to the chloro-substituted species at  longer  reaction times.

 GENERAL MODELING  APPROACH

       When the  various  parameters are  considered,  either  molar  or weight  based
 THAAs  could theoretically serve as  the dependent  variable whereas the other
 variables,  in either their arithmetic  or  transformed state,  represent
 candidate  independent variables.  The general strategy adopted in formulating
 each model  was  to include  single  terms to  describe the roles  of  precursor,
 chlorine,  temperature,  pH,  bromide, and reaction  time in  the  formation  of
 THAAs.  In  keeping with  the philosophy  of developing  chemically rational
models, both molar based as well  as weight  based  THAAs were used as  the
 dependent variable THAA. However, as will be shown,  little difference was
observed between statistical correlations based on molar  versus weight basis
THAAS.
                                       37

-------
 TOTAL  HALOACETIC ACIDS;  RAW/UNTREATED WATERS

       This  section focuses  on models  which can predict  the formation of
 haloacetic  acids in untreated source  waters subjected to chlorination.  Their
 relevance is  severalfold:  (i)  they can be  used to  assess pre-chlorination;
 (ii) they can be used to describe the behavior of  treated waters where  little
 precursor removal has taken place (e.g., direct filtration);  (iii)  they can be
 used to predict  treated-water response if  treated-water precursor levels are
 input, even though treatment  affects  both  the  amount  and type  of precursor;
 and  (iv) they provide a  framework for further  modeling  efforts where changes
 in the amount and type of precursor can be incorporated.

       As part of the model  building (formulation)  process,  the individual
 effects of  independent parameters on  total HAAs were  evaluated singularly.
 Selected results were previously  discussed and shown  in Figure 4.1.  Such
 evaluations were,used to help define  individual parameter effects,  linear
 versus nonlinear effects, and positive versus  inverse effects. Positive
 effects were  exerted by  C12 dose,  temperature,  Br"  concentration, DOC (or
 UVA),  and reaction time. The  effects  of pH were generally mixed,  presenting
 additional  modeling challenges.

       Engerholm  and Amy  (1983) found  that  formation of  chloroform from  humic
 acid under  different conditions of  pH,  temperature, precursor  concentration,
 and chlorine-to-DOC ratio could be  accurately  modeled by transforming both
 dependent and independent variables into logarithmic  forms. This approach was
 later  modified to account for  bromide effects  (Amy et al.,  1987). These
 previously  successful  modeling approaches  were applied  to  the  entire data base
 (738 cases).  Ordinary step-wise multiple-regression modeling efforts
 highlighted the  logarithmic (power-function) formulation  shown below:

            Y =  10bo(X1)bl(X2)b2                                      (4.1)
            where Y =  dependent variable,
            Xi=  independent variable(s), and bi= regression coefficient(s)

       The power-function models,  expressed on  both a weight and  molar basis,
 derived from  the  above approach are shown  in Table 4.1.   (The THAA models
 correspond  to HAAg; HAA5  predictions can be based on summation  of predictions
 for each of the  five relevant  individual species). Perusal  of  the model
 exponents indicates  the positive  influence of  chlorine  dose, temperature,
reaction time and DOC  (DOC was selected over UVA because UVA did not provide
better correlations),  inverse  influence of pH,   and the mixed influence  of
bromide.  These trends are generally consistent with the results  shown in
Figure 4.1.  While pH affects individual species differently, its effects  on
                                       38

-------
  Table 4.1  PREDICTIVE RAW-WATER MODELS FOR HALO ACETIC ACIDS (HAA): TOTAL
                          HAAS (THAA) AND HAA SPECIES
Weight-Based ((ig/L) Models:

     [CAA] = 0.45 f0-009 [Temp]0 m pH"*279 [CU0*7 [DOCf m [Erf029
     R2 = 0.14  F = 18  a < 0.0001 N = 738
     [BAA] = 6.21 x 10-5 10090 [Temp]0-707 pH*804 [Clzf754 [DOCf1584 [Br]1'100
     R2»0.43  F = 83  a£0.0001 N = 738
     [DCAA] = 0.30 10-218 [Tempf485 pH0-200 [Cy0-379 [DOC]1'396 [BrT° '149
     R2 = 0.83  F = S89  a 5 0.0001   N = 738
     [TCAA] = 92.68 1°-180 [Temp]0-299 pH'1-627 [Cya331 [DOC]1-152 [BC]-*229
     R2 = 0.87  F = 821  a < 0.0001   N=738
     [BCAAJ s 5.51  x lO-3!^220 Uemp]0379 pH0-581 [CI/-522 pOCf463 [Br]0-667
     R2 = 0.76  F = 360  a £0.0001   N = 738
     [DBAA] = 3.59 x 10*5 10-085 [Temp]0-360 pH"*001 [Clg]0-673 POCT1-888 [Br]2-052
     R2=0.77  F = 370  a £0.0001   N = 738
     UHAA] = 9.98 10-178 Uemp]0^7 pH"0*5 [Cy0443 [DOC]0-935
     R2 = 0.87  F = 831  a£0.0001   N = 738
Molar-Based ((imoles/L) Model:
     UHAA] = 4.58 1° -156 [Tempf 4S1 pH*"3 [Cy0-865 [DOCf639
     R2 = 0.80  F = 490  a £0.0001  N = 738
Symbols are defined below:
[CAA]; [BAA]; [DCAAJ; fTCAA]; [BCAAJ; [DBAA]: Individual HAA Species (ng/L);
[THAA]: Total Haloacetic Acids Qig/L or ^moles/L);
t:   Reaction Time (hr); 2 < t £ 168
[Temp]: Temperature (*C); 15 £ [Temp] £ 25
pH:  6.5£pH£8.5
[Cy: Applied Chlorine Dose (mg/L), 2.11 < [Cy < 26.4
[DOC]:  Dissolved Organic Carbon (mg/L; 1.2 < [DOC] < 10.7
[Br]: Concentration of Br (HO/L); 7£[Br]£560
[Cy:[DOC]:  0.5 £ [Cy:[DOC] < 3 (mg/mg)

"Source Waters: SLW, SPW, BRW, SRW, BRW, MGW, BGW. PBW.ISW, STO, and SXW
                                         39

-------
THAAs was generally inverse for most source waters (MGW being an exception).
Other researchers (Miller and Uden, 1983) have shown that pH, chlorine dose,
and reaction time have similar effects on TCAA and DCAA formation similar to
those observed herein. With a few exceptions, the model exponents generally
reflect the expected effects of individual parameters. Expressed in this form,
the log-log model cannot yield a negative prediction. Moreover, if t  (reaction
time), DOC, C12  dose,  pH,  Temp,  or Br~ are zero, the multiple-parameter power
function predicts a THAA of zero. For the parameters t, DOC, C12,  these  zero
predictions strictly conform to theoretical expectations. For pH and
temperature,  it is also reasonable to expect that, as these conditions
approach zero, THAA formation should likewise approach zero. On the other
hand, the presence of bromide is not absolutely necessary for TCAA, DCAA, and
CAA formation; other modelers have used a (Br + 1) term, which reduces to
unity, to compensate for this limitation. On the other hand, vitually all
natural waters contain some level of bromide (Amy et al., 1994). If necessary,
the user can simply input a very low value (5 ug/L) near the detection limit
to represent a zero level for bromide.

Internal Data Simulation

      Each of the models has been subjected to a model testing procedure by
plotting predicted versus measured values, employing the same data base used
in model calibration. These data simulations (internal validations) are
summarized in Figures 4.3, 4.4 and 4.5. Figures 4.3 and 4.5 describe weight
and molar based THAA models, respectively. Figure 4.4 is an elaboration of
Figure 4.3 whereby individual-source data sets are shown. A perfect model
simulation would be represented by a plot with an intercept of zero, a slope
of 1.0, and a r2 of  1.0;  a positive intercept suggests overprediction at lower
concentrations while a slope of less than 1.0 generally suggests
underprediction. With the entire data base (738 cases), regressions of
predicted versus measured THAAs were conducted for the log-log models on both
a weight and molar basis, yielding r2 values  of 0.90  and 0.91,  respectively;
intercepts of 2.57 and 0.056, respectively; and slopes of 0.97 and 0.98,
respectively. These results are portrayed in Figure 4.3 and 4.5. Although
correlations between measured and predicted values are very good, the
intercepts and slopes of the above regression equations indicate that all of
the models showed a tendency to overpredict at low THAA levels and
underpredict at high THAA levels. Thus, the models tend to overpredict for
conditions least conducive to THAA formation and underpredict for conditions
most conducive to THAA formation.  Figure 4.4 shows the identification of
individual-source data points used in the testing of the weight basis model,
allowing observation of sources which conform to or diverge from predictions.
                                      40

-------
O>
    800  1
    700  ~
    600  ~
    500  ~
   .
I   400
0)


T3   300
k.
CL
    200 ~
    100 1
           [PREDICTED]  =  2.57  -t- 0.97[MEASURED]

                      R2«0.90   (n=738)
                                                  O
                                    O
                          o
                   o
                   o
                      o
                     o
                                              o
                                                                    o
          ~i  ( i  i |  i  I I  i [—i—i—i—i—|—i—i—i—i—|—i—i—i—i—|—i—i—i—i—|—i—i—i—i—I—j—|—r

         0      100     200     300     400     500     600     700     800

                                  Measured THAA
              Figure 4.3.  Predicted versus Measured Values for Raw/Untreated

                         Weight-Based  (|ig/L) Model
                                     41

-------
                1   1
               o
                                         1
                                                      XD
       D
                D
                                                                       o
                                                                       o
                                                                       CO
                                                                       o
                                                                       o
                                                                                    0

                                                                                    TO




                                                                                    T3
                                                                                    0

                                                                                    CO

                                                                                    0
                 n
                        »*
            O
Ul
cc


CO


UJ

o
UJ
  o
  UJ
                                      o
                                                                       o
                                                                       o
                                                                       CD
                                                                     o
                                                                     o
                                                                     in
                                                                       o
                                                                       o
                                                                       O
                                                                       o
                                                                       CO
                                                                       o
                                                                       o
                                                                       CM
                                                                       O
                                                                       o
                                                                            ^
                                                                            O)
                                                                     s
                                                                     a
                                                                    =    s
                                                                     i_    O
                                                                    £    |

                                                                     8   »
                                                                    |    o

                                                                    >   -o
                                                                     1   1
                                                                     (0
                                                                     0   t
                                                                     0   O

                                                                    E   c
                                                                     o>   O
                                                                     3   -s
                                                                     £   S

                                                                     ?   5
                                                                     •^   •-

                                                                    TS   «
                                                                     *   2
                                                                     o   —
                                                                                      0

                                                                                     £
                                                                                    0

                                                                                    3
                                                                                    O)
                                                                                          0)
o
o
       o
       o
o
o
o
o
o
o
o
o
o
o
o
o
                       (-y6r1) VVH1
                                           42

-------
[PREDICTED]=0.056  +  Q.98[MEASURED]
            R2=0.91
                    2-3           4
                     Measured THAA (jimoles/L)
Figure 4.5.  Predicted versus  Measured Values for Raw/Untreated Water
            THAAs:  Molar-Based (umoles/L)  Model.
                             43

-------
 External Data Validation

       The next step in model validation was external validation with data not
 used in the calibration of the model.  Selected data were taken from the
 literature (James M. Montgomery Engineers,  1991 to test and validate the
 original model (Figure 4.6). The symbols shown in Figure 4,6 represent actual
 data from a range of utility source waters; the line shown represents the
 correlation between our predictions (Table  4.1; THAA weight-basis equation)
 and their measurements.  This attempt at model validation indicated that the
 model overpredicted at lower levels and underpredicted at higher levels.  Thus,
 the model appears to be most applicable to  chlorination of waters with a
 propensity to form THAAs at levels in the general vicinity of the USEPA
 proposed primary drinking water standard (60 ug/L),  although there is a trend
 toward modest overpredictions.

       A similar validation of the model's ability to capture reaction kinetics
 is  shown in Figure 4.7.  Divergence between  measured and predicted values  was
 more apparent at higher  HAA levels.

 HAA SPECIES;  RAW/UNTREATED WATERS

       As part of  the total HAA measurements,  the  concentrations  of six
 individual species were  measured:  trichloroacetic acid (TCAA); dichloroacetic
 acid {DCAA),  monochloroacetic acid (CAA); bromochloroacetic  acid (BCAA);
 dibromoacetic acid (DBAA);  and monobromoacetic  acid  (BAA).   In low bromide
 waters,  the only  significant species were TCAA  and DCAA;  in  waters with
 moderate bromide,  BCAA was also significant;  DBAA was  only an important
 constituent in experiments involving spiked levels of  bromide. In almost  all
 experiments,  CAA  and BAA were trace constituents  present  near detection
 limits.  Individual  parameter effects were highlighted  in  Figure  4.2 as  part of
 the model  building  process.  Table  4.1 also  shows  the individual  HAA species
 models.  For the various  models,  higher values of  R2 were obtained for DCAA,
 TCAA,  BCAA, and DBAA than  CAA and  BAA because CAA and  BAA were only minor
 (present at very  low concentration) constituents  in all cases.

       Predicted versus measured values for TCAA,  DCAA  and BCAA are plotted in
 Figure 4.8. Several  outliers  for both TCAA and DCAA in the figures were
 identified as being  from SLW, one  of the first source waters evaluated which
 involved a higher C12/DOC baseline (3:1)  than the other waters.  Near the
beginning of the research, we elected to lower the baseline  condition from 3:1
to 1:1 mg/mg.
                                      44

-------
_
to
•D
(D
TJ
     600
     500
$    400
re
cc
CC
>.    300
200
     100
           [PREDICTED] =  44.9 +  0.75[MEASURED]
                       R2=0.97 (n=60)
    Seven  Source Water:
     DOC=  3.0 to 11.0 mg/L
     CI2= 3.0 to  25.3 mg/L
    pH= 7.2 to 8.3
    Temp = 20°c
     Br'=5 to 430
                                                 Time * 0.2 to 98.7 hr
                                                      ~
                                                      L-J
                                                                   98.
                                                                    r
                100
                    200
500
600
700
                                                                           800
                             300      400
                                JMM Data
Figure 4.6. Overall External Validation Using JMM Data  with Raw Water Model
           (Final Report: Disinfection By-Products Database and Model  Project,
           1991)
                                      45

-------

1 1 ' 1 1 1 1 I 1 I I 1 1 1 1 1 1 t 1 1 1 f I
1 1 1 ' ••
0>
— "5
5
Q>
o
3
O
CO
tt
0)
kip
.E
•^B
H
_

>
i
— i
•
^
•

—
-
; >

—
V

\
\
—
\
\
III]
1 1 1 1 1 1 1 1 1 f F 1 1 1 1 1 1 P 1 1 \ 1 \

O O O O O
o o o o o
DO r*- CD to ^~
0>
1 ll t 1 1 1 1 1 1 1 1 III 1
^j «si
o» _j o) -
E •& E _
CO CO
co" « £ in
o • o 3 « -
** £ ^ o 5 -
m *Z -.
CO CM « 0 «? -
ii i^ °- to 10
O n E ii 11
O I * '*- — w
a a. H ca o


. • -



A
.


_
L .:





I .:
	
1
> m\*~

\ 7~1 ifct. tfc. 1 1 I.J PsL^k
TJ
o
o E
*«
a
o w
CM QC
•r-
£
•;
^
CD

CO (0
0 O
E
^ -3
£, O)
0 ®^ ^
co £ 5
P
c *
1 =
O 9t
as c
S £ =
"5
c
tfft
o 5

** "5
>
IU
c
tw
O Q>
w "5
UJ
K
^
o
o -t
at
1L
O O O o
000
CO CM -i-



1
a
II
«
o
JQ
E
>*
CO
..
"m
•D
O

«
IJ
a>
^

<
$
<
o
*•
».
o
Q.
€9
oc

"JJ
c
E
?
^






46

-------
    s
    1
    Q.
        500
        400 -
        300 -
200 -
        100 -
             [PREDICTED]=0.49 * 0.96 [MEASURED]
                     R2»0.91
                                                   TCAA
        300
        250 '
                    100       200       300
                        Measured TCAA Oig/L)
                                       400
500
    8
              [PREDICTED]=3.« + 1.08 [MEASURED]
                     Ra*0.86
                   50     100    150     200     250    300
                         Measured DCAACig/U
         100
         80 -
         60 ^
         40*
         20-
              {PREDICTED] = 4.59 + 0.74 [MEASURED]
                      ** xO.85
                                                  BCAA
            0     20     40    60    80    100    120    140
                          MEASURED BCAA (fig/L)

Figure 4.8.  Predicted versus Measured Values for Raw-Water
             TCAA, DCAA and BCAA
                                       47

-------
        Theoretically, predictions derived  from the  individual HAA-species
 models should be consistent with total THAA model  predictions; in other words,
 the summation of predicted values of individual HAA species should be  equal  to
 the measured and/or predicted values of total  THAAs. Figure 4.9 shows  the
 relationship between the summation of predicted individual species (from the
 respective individual species models) versus directly predicted THAA values
 {from  the overall model). As can be seen, a good relationship was observed.
 This suggests two alternative approaches for predicting THAA: either directly
 (with  the THAA model) or indirectly (with summation of predictions from
 individual species models).

 COAGULATED-WATER MODELS

       Precursor (DOC) removal influences both the kinetics and yield of HAAs
 formed. Some precursor removal processes such as coagulation,  adsorption,  and
 membrane separation remove precursor molecules intact;  others such as
 ozonation transform (partially oxidize)  precursor molecules.  The precursor
 remaining after treatment may be less (or possibly more)  reactive in forming
 DBFs.

       While removing precursors,  coagulation has little effect  on bromide  ion,
 Br"  (Amy et al., 1991); thus, coagulated waters have a greater amount of
 bromide ion relative to organic  precursor.  Ultimately,  HAA speciation is
 affected by both precursor and bromide  levels;  as  the ratio of  BrVDOC  (or
 Br"/Cl2) increases,  the formation of brominated species  is favored.

       These  "submodels" are generally similar in format to the  raw/untreated
 water  models discussed  above.  However, since  pH and temperature were
 maintained constant  at  their baseline conditions,  these parameters do not
 appear in the coagulated-water models, a model  limitation.  The THAA and HAAs
 models  for alum  and  iron  coagulated waters are  shown in Tables 4.2 and  4.3,
 respectively. There  are 144 cases  (n = 144) for each coagulant-specific data
 base. As can be  seen, both sets of models have  similar functionalities
 associated with  each parameter; moreover, model  simulations by each provided
 comparable results. Thus, a decision was made to combine the data bases
 together for development  of a combined alum plus iron set  of models. The
 corresponding total HAA models, on both a weight and molar basis, and HAA
 species  models for all of the treated waters are listed in Table 4.4. As
 discussed before, more accurate models were found for DCAA, TCAA, BCAA, and
DBAA than for CAA and BAA. Little difference was observed between weight and
molar basis models for THAAs.
                                      48

-------
     700
_   600


s
(0
•§    500

-------
TABLE 4.2 PREDICTIVE COAGULATED-WATER MODELS FOR THAA AND HAA SPECIES:
                                 ALUM MODELS'           __

Alum Coagulated Water Models:
     [HAA] = 7.05 t°'159 [DOCf581 [Brf080
     R* = 0.90  F = 313  a^O.0001  N =
     [CAA] = 12.82 r0066 [DOC]*377 [BrT0303
     1^ = 0.30  F = 15  a^O.0001  N = 144
     [BAA] = 3.97 x 10* t0132 [DOC]0409 [Brf834 [CljJ0095
     R2 = 0.33  F = 17  a<0.0001  N = 144
     [DCAA] = 10.96 10*30 [DOCf704 [fir?0-514 [CIJ° 751
     R2 = 0.84  F = 178  a£0.0001  N = 144
     [TCAA] = 6.22 1°'164 [DOCf900 [BrT0567 [CIJ°
     R2 = 0.93  F = 460  a^O.0001  N = 144
     [BCAA] = 0.13 10-193 [DOC]1286 [Brf675 [Cy0-251
     R2 = 0.86  F = 204  a<0.0001 N = 144
     [DBAA] = 4.84 x 10'5 10-077 [DOC]*424 [Brf222 [Claf379
     R2 = 0.85  F = 188  a<0.0001 N = 144
Symbols are defined below:
[HAA]:    Total Concentration of Haloacetic Acids ftig/L); Sum of Six Species
[DCAA]; [TCAA]; [BCAA]: Individual Haloacetic Acid Species
t:        Reaction Time (hr); 2 < t £ 168
[Cy:     Applied Chlorine Dose (mg/L), 1.11 < [Cy < 14.19
[DOC]:    Dissolved Organic Carbon (mg/L); 1 < [DOC] < 4.6
[Br]:     Concentration of Br (^g/L); 36 < [Br] < 308
[Cy:[DOC]:  1 < [Cy:[DOC] £ 3 (mg/mg)

"Source Waters: SPW, BRW, SRW, HMR, PBW, ISW, STW and SXW
                                         50

-------
Ol
CO
0>
'o

ts
T3
£
0-
«ta-
o

E
3
CO
     700
     600
500
     400
     300
200
100
           [SUM]=-7.39  + 1.04 [PREDICTED]


                     R2=0.99
                100    200    300    400    500

                             Predicted THAA Qig/L)
                                                600    700    800
  Figure 4.9. Summation of Predicted Individual HAA Species

            vs. Predicted Raw-Water THAAs; Weight-Based Models
                                     49

-------
TABLE 4.2 PREDICTIVE COAGULATED-WATER MODELS FOR THAA AND HAA SPECIES:
                                 ALUM MODELS*

Alum Coagulated Water Models:
     [HAA] = 7.05 f -159 [DOCf581 [Brf080
     R2 = 0.90   F = 313
     [CAA] = 12.82 f0-066 [DOCJ*877 [BrT0303 [Cy0671
     R2 = 0.30   F = 15 a < 0.0001  N = 144
     [BAA] = 3.97 x 10* t0132 [DOCf409 [Brf ** [Clj""5
     Rz = 0.33   F = 17  cx^O.0001  N = 144
     [DCAA] = 10.96 10530 [DOCf704
     R2 = 0.84  F = 178  a£0.0001
     [TCAA] = 6.22 1° 164 [DOCf900 [BiT*287 [Cy°-
                      a<0.0001  N = 144
     [BCAA] = 0.13 10193 [DOC]*286 [Brf ^ [Cy° -251
     R2 = 0.86  F = 204  a<0.0001  N = 144
     [DBAA] = 4.84 x 10'5 1"77 [DOC]*484 [Brf222 [CIJ0379
     R2 = 0.85  F = 188  a<0.0001  N = 144
Symbols are defined below:
[HAA]:    Total Concentration of Haloacetic Acids (^g/L); Sum of Six Species
[DCAA]; [TCAA]; [BCAA]: Individual Haloacetic Acid Species
t:        Reaction Time (hr); 2 
-------
   TABLE 4.3 PREDICTIVE COAGULATED-WATER MODELS FOR THAA AND HAA SPECIES: IRON
                                         MODELS*
  Iron Coagulated Water Models:
       [THAA] = 3.84t0-163 [DOC]0682 [Brf170 [CIJ0-551
       R2 = 0.94  F = 525  a £ 0.0001  N = 144
       [CAAJ = 11.94 r0021 [DOC]-0666 [BiT0415 [Cy1 °61
       R2 = 0.44  F = 27   a<0.0001  N = 144
       [BAA] = 3.33 x 10* f0-020 [DOC]"11 [Br]0925 [Cya313
       R2 = 0.39  F = 22   a<0.0001  N = 144
       [DCAA] = 6.31t0-213 [DOC]0846 [BrT0416 [Cy0742
       R2 = 0.90  F = 329   a < 0.0001  N = 144
       [TCAA] = 3.971°163 [DOC]1 -083 [Brl^15 {CIJ0880
       R2 = 0.94  F = 523   a £0.0001  N = 144
       [BCAA] = 0.07510209 [DOC]° *" [Br'f ^ [Cy°199
       R2 = 0.90  F = 301   a < 0.0001  N = 144
      [DBAA] = 3.92 x 10* t0-068 [DOC]*318 [Br]2*56 [cy°-
      R2 = 0.92   F=189  a £0.0001 N * 144
(0.397
 Symbols are defined below:
 [HAA]:    Total Concentration of Haloacetic Acids (ng/L); Sum of Six Species
 [DCAA]; [TCAA]; [BCAA]: Individual Haloacetic Acid Species
 t:         Reaction Time (hr); 2 
-------
 TABLE 4.4 PREDICTIVE COAGULATED-WATER MODELS FOR THAA AND HAA SPE-
                  CIES: COMBINED ALUM PLUS IRON MODELS'
Weight-Based (nS/L) Models:
     [THAA] = 5.22 1° 153 [DOC]° •
     R2 = 0.92   F = 771   a £0.001  N = 288
     [CAA] = 12.30 r0-043 [DOC]"*522 [BrT™ [CU""
     R2 = 0.36   F = 40 a < 0.0001  N = 288
     [BAA] = 3.68 x 10* t0077 [DOC]0** Pfl"77 [CIJ""
     R2 = 0.35   F = 37  a<0.0001  N = 288
     [DCAA] = 8.38 t0-222 [DOC!"77 [Erf466 PU0 ?44
     R2 = 0.88  F = 461   a< 0.0001  N = 288
     [TCAA] = 4.98 f •« [DOCf""
     F^sO.gS  F = 460  a<0.0001  N = 288
     [BCAA] = 0.098 10^11 [DOCf368 [Br]0'713
     R2 = 0.87  F = 485  a < 0.0001  N = 288
     [DBAA] = 4.41 x 10'5 10'072 [DOC]^374 [Br]"37 [CIJ™
     R2 = 0.84  F = 382  a < 0.0001  N = 288
 Molar-Based (^moles/L) Models:
     UHAA] = 3.03 1° -153
     R2 = 0.92  F = 771  a < 0.001  N = 288
 Symbols are defined below:
 [THAA]: Total Haloacetic Acids fag/L); (^g/L or nmoles/L)
 [CAA]; [BAA]; [DCAA]; [TCAA]; [BCAA]; [DBAA]: Individual HAA Species
 t:   Reaction Time (hr); 2 < t £ 168
 [CIJ: Applied Chlorine Dose (mg/L),   1-11 < [Cfc] < 24
 [DOC]:  Dissolved Organic Carbon (mg/L);  1.0 < [DOC] < 4.6
 [Br]: Concentration of Br (jig/L);  36 < [Br] < 308
 [CIJ:[DOC]:  1 < [Cy:[DOC] < 3  (mg/mg)
 'Source Waters: SPW, BRW, SRW, HMR, PBW, ISW, STW and SXW
                                   52

-------
 Data Simulation/Internal Validation

       A data simulation for all of the treated waters is shown in Figure 4.10;
 comparisons of predicted versus measured THAAs were made with the entire data
 base (n = 288). Although correlations between measured and predicted values
 were good, the models showed a tendency to overpredict at lower THAA levels
 and underpredict at higher THAA levels. However, the simulations generally
 show least error at levels near the proposed standard (60 ug/L). Figure 4.10
 also elucidates data subsets derived from alum versus iron coagulation. The
 underprediction at higher THAA levels appeared to be associated with the SXW
 source which is the only source with a high DOC  remaining after coagulation.

       Predicted versus measured values for TCAA, DCAA and BCAA were also
 plotted (Figure 4.11). Compared to Figure 4.8,  better correlations were
 obtained for the coagulated waters than the raw waters.  However, these treated
 water models showed a tendency to: overpredict at lower THAA levels and to
 underpredict at higher THAA levels. Thus,  the models tend to overpredict for
 conditions least conducive to THAA formation and underpredict for conditions
 most conducive to THAA formation.

 Predicted Models Based on Reactivity Coefficient,  d>

      We also took a different (alternative)  approach to modeling HAA
 formation in treated waters,  based on changes (reductions)" in precursor
 reactivity (HAA/DOC)  after coagulation.  This  approach involves use of a
 reactivity coefficient,  ,  which is used to adjust the DOC term in the raw-
 water models,  based on the premise:

              = (HAA/DOC) trt/(HAA/DOC)Cftw      where c[>=0-1.0         (4.2)

      The  DOC term in the  raw water models  is adjusted to  reflect  a  different
 (lower)  reactivity such  that  DOC =  $(DOCtrC).  Based on an anticipated reduction
 in reactivity,  the parameter  $ would  be  expected to vary from 0  to 1.0  for
 coagulant-treated  waters.  Table 4.5 summarizes <|> values  for THAA  formation  at
 different  reaction times.  Model simulations based  on  this  approach are  shown
 in Figures 4.12  and 4.13  for  reaction times of 24  and  96 hours,, respectively.
As can be  seen,  there  is good  prediction by ^-adjusted raw water models for
 the treated waters. Figure  4.14 shows  one example  for  this type of simulation
with the BRW  source water.  Thus, as an alternative to the  treated-water
models,  we can directly predict treated water THAA formation  by coupling the  $
parameter with raw water models. A limitation to this approach is the need to
                                       53

-------
     700
     600
     500
     400
2    300
     200
     100
[PREDICTED]  = 28.44 + 0.66  [MEASURED] (OVERALL)



           R2 =  0.95 (n=288)



[PREDICTED] = 2.62 + 0.98  [MEASURED] (IF THAA < 100




           R2 =  0.92 (n=179)
          0
     	1	1	1	1	1	1	1	1	1	1	[	1	1	1	1	j	1	1	1	1	1	!	1	1	'	1	1	1  I I"





      100     200     300     400      500      600      700
                                   Measured THAAs (fig/L)




Figure 4.10.  Predicted versus Measured Values of THAA for Coagulated/Treated



            Waters Using Combined Alum plus Iron Treated-Water Models
                                       54

-------
  300




  250 H
[PREDICTEDJ-5.22 + 0.85[MEASUREO]
   200
       . [PREDICTED1=5.44 + 0.76[MEASUR EDI
   150 -
o
Q  100 H
1
a.
    50 -
    40
    35 -
 m  20 -




 *  1S"

 °-  10-


     5 -


     0
         [PREDICTED]=2.07 + 0.81 [MEASURED]
                 R!»0.93                o   o
                                    O  Oq-
                                      00
               O O

             O      00^°'        *
               9  o>  oa
                ^ «o>*  o

             °^o *V«o   00


                -°^ff o°
                  CD
        .<£.
      0     50    100    150    200    250   300    350

                       Measured TCAA (|ig/L)
                10
             50     100     150     200    250     300

                       Measured DCAA Oig/L)
                                             50
                               20       30       40

                             Measured BCAA (fig/L.)


Figure  4.11.   Predicted  versus  Measured  Values  of TCAA,  DCAA


               and BCAA for  Coagulated/Treated  Waters
                                  55

-------
TABLE 4.5 SUMMARY OF REACTIVITY COEFFICIENT, 


-------
    700
    600
    500
w
<   400
I
=6   300
    200
    100
            [PRED1CTED]=0,22  +  1.13[MEASURED]
                          R2=0.98
                          O
                                                           O
0  i  l  (  ' '  I  '
   0        100
500
600
700
                          200      300      400
                                Measured THAA
 Figure 4.12. Predicted versus Measured Values  of THAA for Coagulated/Treated
             Waters Using Raw/Untreated  Water-Models Combined with 0
             Concept; 24-Hour Prediction
                                     57

-------
58

-------
TJ
    180
    160
    140
    120
    100
     80
     60
     40
                                                   Raw Water Model
- 0 Model
 Treated Model
 Raw Data
 Coag. Data
    Source:  BRW
    DOC: 3.54 mg/L
    0: 0.77
    Baseline Conditions
20 "1 — ' — ' — ' — ' — ! — ' — ' — ' — ' — I — ' — ' — ' — ' — I — ' — ' — ' — '
   0         50
250
                  300
                            100       150       200
                                  Reaction Time (hr)
Figure 4.14.  Comparison of  Predictions from   Treated-Water Models versus
             Raw/Untreated Water Models  Combined  with 0 Concepts
                                   59

-------
 experimentally determine <|j for a given set of conditions.  Nevertheless,  this
 approach provides a framework for modeling precursor reactivity and associated
 reductions imparted by coagulation.

 External Model Val.idation

       An external validation of the  alum-plus-iron coagulated water model;
 employing literature data,  is shown  in Figure 4.15 {data shown are from
 coagulated waters; some boundary condition violations for  Br") . The model
 predictions appear to be reasonably  accurate for a range of  source waters
 subjected to coagulation.

 Model  Simulations for Coagulation of HAA Precursors

       The U.S.  EPA will require utilities to consider enhanced coagulation  of
 precursors as a DBP control  strategy.  A range of TOC removals,  ranging  from 15
 to  50  %  will be required,  depending  on initial TOC and alkalinity  conditions.
 Using  the alum-plus-iron coagulated-water models,  a simulation of  reducing  TOC
 from 4 to 2 mg/L is shown in Figure  4.16.  The models presented herein can be
 used to  assess  such control  strategies.

 HAA SPECIATION  MODELS;  BrVDOC AS MASTER VARIABLE

       In theory,  one would expect chlorinated HAA species  to  decrease as a
 function of increasing bromide;  brominated HAA species to  increase with
 increasing bromide;  and mixed chlorinated/brominated species  to first increase
 then decrease with bromide.  Thus,  for  the HAA species, we  examined alternative
 functionalities  to represent bromide effects  on  HAA species.  In particular,  we
 focused  on use  of a polynomial terra  to capture bromide effects  by  simply
 relating fractional concentrations of  individual species (HAA Species/THAA)  to
 the Br~/DOC ratio. Through prior examination of scatterplots,  we found that
 the ratio  of  BrVDOC, representing the ratio of inorganic to  organic
precursor, most  accurately captures  bromide effects  on HAA speciation. Figures
 4.17 and 4.18 show the  fraction  of the  six HAA species versus BrVDOC for all
 the raw  and  treated waters;  shown are  actual  data  (points)  along with model
simulations  (curves) .  As BrVDOC increases, TCAA and DCAA  decrease
exponentially, DBAA increases  exponentially after  a  lag, and  the intermediate
species  BCAA  increases then  decreases  in  a polynomial pattern.  Resultant
 fractional concentration models  are  shown in  Tables  4.6 and 4.7. Except for
TCAA,  polynomial  functions provided  the best  data  fit; however, in using these
functionalities,  one must not  violate  the  boundary conditions  (ranges of data)
used in  their formulation. In  other words, the overall polynomial  pattern
                                       60

-------
                                                                          CM

                                                                          ->»
                                                                                   Q
                                                                                   »*
                                                                                   a
                                                                                   O
                                                                                  -D


                                                                                  O>


                                                                                  «
                                                 O



                                                 «

                                                 «



                                                 •D
                                                 0)
                                                                                  O)
                                                                                  a
                                                                                  o
                                                                                  u
                                                                         tD
                                                                         >
                                                 X
                                                1U

                                                in


                                                *t

                                                 G)
O

CO
O

CO
O

CO
                            (1/Bri)
                                   61

-------
                                  o
                                  IT)
                                  CM
                                  O
                                  O

                                  CM
IB



O
1L
                                   o
                                   in    -c-
                                   §
                                   o
                                   u>
                                                  O
                                                  O)
                                                  CD
                                                  O
                                                 o
 IB
 **
 O



 UJ
 E

 CO

 «>
 ^

 *


 d>
62

-------
                     m
3;S     3    5     E
8    g     8    §     |
                            D
                                                                            CO

                                                                            6
                                                                            CO

                                                                            o
                                                                            u>
                                                                            CM
                                                                            CM

                                                                            O
                                                                            to

                                                                            6
                                                        o>

                                                       I
                                                       £   a
                                                             CO
                                                             c
                                                             £   ~
                                                             *»   CO
                                                             _o   c
                                                             •D   £
                                                             o   *-
                                                                                         O   CO
                                                             *   ^
                                                             o   c
                                                             *o   co
                                                             O   CO
                                                             «E   ®

                                                             g   51
                                                             2   tc
                                                       O
                                                       O
                                                       D
                                                       ~t
                                                       m
                                                                                         o   o>
                                                                                         £   |
                03

                C
                O
                                                                                             O
                     a
                                                                                         a   a
                                                                                         S   Q

                                                                                         £   +-
                                                                                         O   c
                                                                                         u   £
                                                                                         H   f|)
                                                                                         O   4)

                                                                                         0   a
                                                                                         •=   °
                                                                                         S   a
                                                                                         o   o>
                                                                            in
                                                                            o
                                                             o   c
                                                             £   p
                                                             u.   a
                                                                                        _O)
                                                                                        u.
CM
                     CO
                     6
    


-------
            <   <   o   o   o   m
            o   m   D   H   CD   o
                    D   6
                                      n
                              40 .   1
                               i  ,   /1 r
                               •V  /--.
                               •\/     /
                                « D V*
CM
                                                      m
                                                      co
                                                     cvi

                                                     d
                                        7 *

                                        — o
                                        "O 5
                                        £ £
                                        a. s

                                        >. I
                                                              (0 V
                                                              O>  O
                                        2 c
                                        V 0
                                        O «

                                        = £

                                        c &
                                        £ oc


                                    ?  5 «
                                    •I  o •
                                    O      w
                                    O   c  ..
                                5   o   5  S

                                d   m   £  S
                                        «  e
                                        o  «
                                        c  «
                                        O  £

                                -       21
                                        «£

                                        -  5
                                        "S  £
                                        S  o
                                10       "-  t
                                o
                                                              _D>

                                                              1L
              00

              d
(D

d
ex;

d
                S9i09ds WH (0 UOI
                             64

-------
     TABLE 4.6 SUMMARY OF FRACTIONAL-CONCENTRATION HAA SPECiATION MODELS'
                                    24-hour

                        SECOND ORDER POLYNOMIAL FITS
      SPECIES FRACTIONAL CONCENTRATION = a + bX 4CX2 WHERE X = Br/DOC (mg/mg)
Species
CAA
DCAA
BCAA
BAA
DBAA
a
0.045
0.32
0.052
0.0006
-0.041
b
0.75
-2.16
3.28
0.2
2.66
c
-2.68
4.56
-11.2
-0.25
-5.24
r2
0.29
0.72
0.84
0.73
0.80
                              EXPONENTIAL FITS
	SPECIES FRACTIONAL CONCENTRATION = a e" WHERE X = Br/DOC (mg/mg)
	Species	a	     b                  r2
       TCAA               0.52                -4.30	p.60

All Waters, Number of Cases (n) = 71; Reaction Time = 24 hr
                                      65

-------
    TABLE 4.7 SUMMARY OF FRACTIONAL-CONCENTRATION HAA SPECIATON MODELS"

                                    96-hour


                        SECOND ORDER POLYNOMIAL FITS
     SPECIES FRACTIONAL CONCENTRATION = a + bX +CX2 WHERE X = Br/DOC fng/mg)
Species
CAA
DCAA
BCAA
BAA
DBAA
a
0.028
0.364
0.050
0.003
-0.041
b
0.477
-2.604
3.738
0.093
2.800
c _
-1.581
6.150
-12.220
0.109
-4.604
r2
0.25
0.86
0.85
0.76
0.88
                              EXPONENTIAL FITS
       SPECIES FRACTIONAL CONCENTRATION =  a e** WHERE X = Br/DOC (mg/mg)

      Species	a	b-                  r2

      TCAA                Q.518               -4.496                0.74



All Waters, Number of Cases (n) = 71; Reaction Time * 96 hr
                                      66

-------
suggests an increase then decrease as a function of the independent variable,
a dependent variable which is known to only increase or decrease can be
modeled by an appropriate region/range of the polynomial function.

      Because coagulation does not remove bromide, Br~/DOC ratios are higher
in the treated waters. Consequently, data shown at higher Br'/DOC ratios in
Figure 4.17 and 4.18 corresponds to treated water data. The fractional
concentration models portrayed in Tables 4.6 and 4.7 were calibrated using
both raw/untreated and treated water data; thus, they pertain to, and are
applicable to both. No external validation of these models was performed.
                                      67

-------
                                   SECTION 5
                             TRIHALOMETHANE MODELS

      Trihalome thanes (THMs) are important by-products formed during the
chlorination of drinking water. The maximum contaminant level (MCL) for THMs
in drinking water was first regulated at 100 ug/L in 1979. A new MCL of 80
ug/L is being proposed for THMs. The mechanisms and kinetics of THM formation
have been long and extensively studied. THM precursors in natural waters have
been identified as predominantly humic substances (Rook, 1974; Oliver, 1979).
Humic substances typically comprise 50 % of the dissolved organic carbon  (DOC)
in natural water. Humic substances include two fractions; fulvic acids  (80-
90%) and humic acids (10-20%) . The THM formation potential for humic
substances has been investigated by Lynn (1982) who showed that fulvic acids,
particularly the molecular weight fraction between 5,000 - 10,000 daltons,
contributed most of the THM formation upon chlorination. Trends of greater THM
formation potential with high phenolic content and larger molecular size of
humic substances were observed by Oliver, et al . (1983); they also showed
that THM formation correlated with color. Miller and Uden in 1983 studied the
effects of reaction time, pH, and chlorine- to-carbon ratio (C12/DOC)  on
chloroform and other chlorination products.

      A predictive model for chloroform formation from humic acid was
developed by Engerholm and Amy  (1983). Amy et al. (1987b) developed a non-
linear regression model to predict THM formation in raw source waters; the
five independent variables involved in model development included pH,
temperature, bromide ion concentration (Br~) ,  applied chlorine concentration
and nonvolatile total organic carbon  (NVTOC) . Only one reaction time of 96
hours was considered. The general non-linear model based on a relationship
between 96-hour trihalome thanes formation potential  (THMFP) and individual
variables, was formulated front the following general equation:
THMFP = b0 + (Br-)bl + b2 Log  (C12)  + b3  (pH)  +  10[b4 (Ten*)]  + b5(NVTOC)     (5.1)

      Amy, et al.  (1987a) formulated_THM predictive models  for  raw waters,
which have formed  the basis  for EPA and AWWA  sponsored efforts  to develop
overall DBP formation models. Amy's models were based on nine geographically
distributed natural waters,  and included seven independent  variables:  pH,  TOC,
temperature, Br~ concentration,  UV254/ chlorine dose,  and reaction time. Models
based on a log-log transformation of variables were found to be more  accurate

                                       68

-------
  than  non-linear models.  Amy's  models  described  the  chlorination of
  raw/untreated water, whereas the  models  of  Chadik and Amy (1987)  attempted to
  address  treatment  effects  on THM  formation  kinetics.

       This  chapter presents statistically-based  empirical models  for
  predicting  the kinetics  of total  trihalomethanes (TTHM)  formation,  as well as
  individual  trihalomethane  species in  both raw/untreated  waters and  chemically-
  treated  (coagulated) waters.   Resultant  power-function TTHM and THM species
  predictive  models  for raw/untreated sources were based on data derived  from
  eleven source  waters; BGW, BRW, HMR,  ISW, MGW, SLW, SPW,  SRW, SXW,  VRW  and
  PBW. Eight  of  these source waters, excluding BGW, MGW, and SLW, were used  in
  developing  models  for coagulated  waters.

  INDIVIDUAL  PARAMETER EFFECTS ON TTHM AND THM SPECIES

       The model building process requires an understanding of the effects  of
 individual parameters (independent variables) on TTHM  formation as well THM
 species.  The effects of pH, temperature,  chlorine dose (C12/DOC) ,bromide
 concentration, reaction time,  and DOC are shown in Figure 5.1. Except for DOC
 effects.  Figure 5.1 contrasts two representative sources, VRW and HMR;  DOC
 effects are shown for 10 sources encompassing a broad range.  The positive
 effects of each of these parameters is apparent,  although some parameters have
 a linear  effect while others exert a nonlinear effect. DOC provided better
 correlations than UV absorbance as a precursor parameter; due to their
 colinearity, only DOC was included in the models.

       Figure 5.2 shows the effects of individual  parameters on each of  the
 four THM  species. These  effects are straightforward with the  exception  of
 bromide;  chloroform is  inversely related  (an exponential  decrease),
 bromodichloromethane  and dibromochloromethane increase then level off  (and
 would  be  expected to  further decrease  polynomially at  higher  Br-  levels),  and
 bromoform increases in an S-shape  manner  following a  logistic  function
 behavior.

 TOTAL  TRIHALOMETHANES; RAW/UNTREATED WATERS

      The weight-based and  molar-based models for predicting total THM  (TTHM)
 formation in raw/untreated  waters  are  shown  in Table 5.1.  These models  are
based on 11  source waters and a total  of  786 cases  (n  = 786). They include  six
independent  variables; dissolved organic  carbon (DOC),  chlorine dose (C12) ,
ambient/spiked  bromide levels (Bar), temperature  (Temp), pH, and reaction time
 (t) .  Assessment of each source water involved a total of  82 measurements,
based on an orthogonal matrix of one DOC  (ambient), five  chlorine doses
                                       69

-------
     240
     ZOO
     160
     120
      80
     240
  „  200
     160
     120
      SO
        0.0
                                                 240
                                                 200 -
        14.0   16.0    18.0   20.0   22.0   24.0

                    Temperature (°C)
                                           26.0
                                                 120 -
                                                    6.5
                                                 300
                                                 250
                                                 200
                                              S
                                              E  150
                                                 100
0.5    1.0    1.5    2.0    2.5    3.0

           CI:DOC
                                                  50
                                                    0.0
                                                             50.0
                                                                                        8.5
                                                                     100.0      150.0

                                                                   Tbiw (hour)
                                                                                        200.0
     350



     300



     250



     200



     150



     100



      5 0
                                                 500
                                    400  -
                                    300 -
                                    200
                                    100  -
               100
                      200

                       Bf
                             300     400
                                           500
                                        1.20 2.44 2.78 3.54 3.67 4.18 4.19 5.7310.5510.GO

                                                      DOC
Figure  5.1  Individal Parameter  Effects on TTHM Formation  in  HMR & VRW Sources:
             Effect of  Chlorine, pH,  Temperature, Bromide,  and Reaction Time (all
             other parameters  held  at  baseline  conditions).
                                               70

-------
  180
  150
  120  '
   3D
                                         200
   140
   120
3- 100
    80
    60
    40
    20
    0
    0.0
          0.5
               1.0
                    1.5   2.0

                     ChDOC
                               2.5
                                    ao
                                         3.5
   120
   100
I  BO
   60
   40
   20
     14
           16
                 18     20    22

                  Temperature fC)
                                   24
                                         26
   140



   120



3-  ioo
OS
a.
7  80
*


I  60



1  40



   20
                                                    6.0
                                                                7.0
                       7.5

                       PH
                                                                            8.0
                                                                                   8.5
          100   150  200   250  300   350  400  450
   140


   120


   100


    80


    60


    40


    20
    0.000   0.020  0.040   0.060

                       BrTCI
                                                                            0.080   0.100   0.120
Figure  5.2  Individual Parameter Effects on THM Species  Formation  in  VRW:
              Effects of Chlorine, pH, Temperature, Bromide, and  Reaction Time
              (all other parameters  held  at  baseline conditions).
                                            71

-------
         Table 5.1 Predictive Raw-Water Models for Trihalomethanes (THM):
                       Total THMs  (TTHM) and THM Species*

 Weight-Based fcg/L) Models:
 [TTHM] =10-1-385 [DOC]1-098 [Cl^0-152 [Br]0.068 TempO.609 prf'JBM #.263
        R2 = 0,90      N = 786         F = 1198        a< 0.0001
 [CHCI3] =10-1-205 [DOCJ1.617 [Cy0-094 [Brl-0-175 TempO-607 pHl-403 tO.306
        R2 = 0.87      N = 786         F = 847         a< 0.0001
          =10-2-874 [DOC]0.901 [CyO-01^ [Br]0.733 TempO-498 pH1-511 tO.199
        R2 = 0.90      N = 786         F = 1164        a<0.0001
 [CHBr2CI] =10-5-649 [DOCJ-0.226 [Cl^0-108 [Br]1-81 TempO.512 PH2-212 tO.146
        R2 = 0.89      N = 786         F = 1087        a< 0.0001
 [CHBr3] =10-7-83 [DOC]-0-983 [Cl^0-804 [Br]1-765 TempO-754 pn2.139 tO.566
       R2 = 0.61       N = 786         F= 199         a<0.0001
 Molar-Based ^moles/L)  Models:
 [THMs] =10-1-873 [rX)C]1-222 [Cla]0-104 [Br]0.016 TempO.604 pHl-538 tO.270
       R2 = 0.91      N = 786         F = 1198        a<0.0001
 UTHM] = Total Trihalomethanes (ng/L) or (^Mote/L).
 [CHCiaJ, [CHBrCa, [CHBr2CO, [CHBra] = Individuai Concentrations of THM Species (jig/L).
 [DOC] = Dissolved Organic Carbon (mg/L) or (mMote/L)
       1.2<[DOC(mg/L)]<10.6,  0.1 < [DOC (mMote/L)] <; 0.883
 [Pal = Applied Chlorine (mgA) or (mMote/L)
       1.51 < [C^ (mgA.)} £ 33.55,  0.0213 £ [CI2 (mMole/L)] <, 0.472
 [Br] = Concentration of Bromide Otg/L) or (nMole/L)
       7 £ [Br Qig/L)] < 600,  0.0876 < fBr friMote/L) £ 7.509
Temp = Incubation Temperature(°C)
       15 < Temp £25
pH: 6.5
-------
  (C12/DOC = 0.5,  1,  1.5,  2 and 3 mg/mg),  four bromide levels (ambient,  ambient
 + 100 , ambient + 200, ambient + 300 ug/L), three pH levels (6.5, 7.5, and
 8.5), three temperatures  (15, 20, 25 °C)  and six reaction times (2,  12,  24,
 48, 96 and 168 hours). Based on the exponents associated with the models, each
 of the independent variables exerts a positive influence on total THM
 formation. Both linear and non-linear models were developed preliminarily,
 with power function models (log/log transforms) deemed as providing the best
 fit of data. The TTHM models all exhibited a good coefficient of
 determination, R2  =  0.90.

       Figure 5.3 shows data simulation in which predicted  (modeled)  results
 are compared against measured (experimental) values; this internal validation
 demonstrates the good simulation capabilities of the TTHM model. It is
 noteworthy that the molar-based model for TTHM does not provide significantly
 better predictive capabilities than the weight-based model.

 THM SPECIES; RAW/UNTREATED WATERS

       The development of THM speciation models can provide an  indirect means
 of estimating total  THM formation (summation of individual species),  can help
 describe the relative importance of  each THM component behavior under various
 conditions,  and can  elucidate the influence of bromide ion on  THM species
 distribution.  Individual THM species models were first developed by  Chowdhury,
 et al.  (1991).  Their models allow a quantitative assessment of bromide effects
 on overall THM formation as well as THM speciation;  however, they are  only
 valid for raw/untreated waters.  Predictive models for THM species formation
 for both raw/untreated and treated water have been developed in this study.
 These models will  be discussed in this  (raw/untreated)  and the next  (treated)
 section.

      Table  5.1  also shows weight-based models  for predicting  individual  THM
 species  formation. The R2 values for the three models which predict chloroform
 (CHC13) ,  bromodichloromethane  (CHBr2Cl),  and dibromochloromethane (CHBr2Cl)
 formation  range  from 0.87  to  0.90, while  for  bromoform  (CHBr3) , the R2  was
 only  0.61.   Bromide  plays  a very important role  in THM  species  formation  and
 distribution. Bromide  has  a negative effect on chloroform  formation and a
positive influence on  the  brominated species, at  least  over the range of  the
 indicated  boundary conditions. Formation of brominated versus chlorinated THM
species is affected  by the  competition between bromine and chlorine. The
boundary conditions  for the TTHM and THM species models are listed in Table
5.1. Besides use of  the TTHM predictive model, there is an alternative
approach for predicting TTHM through summation of  individual (predicted) THM

                                      73

-------
  1000
   800
oi
            [Predicted] = 0.21788 + 0.98474[Measured]

            R2 = 0.96,   n = 786
                          Measured TTHM  (ug/L)
                                                              1000
    Figure 5.3  Predicted versus Measured Values for  Raw/Untreated
               Water  TTHM;  Weight-Based (\igtL)  Model
                               74

-------
species  from the THM species models. Figure  5.4 shows  the relationship between
the  summation of predicted THM  individual species  (from individual  species
models)  versus TTHM from  the overall model;  it is  apparent  that both
approaches have merit, although the TTHM model is  superior  in predictive
capability based on its R2 (0.90) compared to the R2 {0.61  - 0.90) values  for
each of  the four species  models. As shown in Figure 5.4, the summation
approach tends to overpredict.

COAGULATED WATERS

      One of the objectives of  this research was to define  coagulation effects
on the formation and kinetics of TTHM and THM species.   Coagulation was
emphasized in this effort because of the recent regulatory  emphasis placed  on
precursor removal. Proposed regulations will require that utilities evaluate
enhanced coagulation if their TOC level before post-disinfection is * 2 mg/L.
Jar  tests were employed to evaluate precursor removal  by both alum  and iron
(ferric  chloride) coagulation;  the precursor removals  observed were described
in Chapter 3. Coagulant doses were selected  through screening experiments as
those providing a target  DOC reduction of between  25%  to 50%. Chlorination  of
coagulated waters was carried out at a pH of 7.5,  a temperature  of 20 °C, and
ambient  bromide level. C12/DOC ratios of 1 and 3  mg/mg were employed over
reaction times of 2 to 168 hours. In these models,  temperature and  pH were  not
varied.

      A  major effect of chemical coagulation is that it  increases the ratio of
Br~/DOC,  'While removal of  the organic precursor (DOC) is achieved,  the
inorganic precursor (Br~)  passes conservatively through the coagulation
process. The net effect of this increase in  Br'/DOC ratio is a  shift toward
the  formation of brominated THMs.

      Table 5.2 shows predictive coagulated-water  models for TTHM and THM
species  formation based on alum as a coagulant for eight water sources (n =
143). Corresponding boundary conditions are  also indicated  in Table 5.2. In
the predictive model for  TTHM, the exponents show  the positive effect of DOC,
chlorine dose, bromide level and reaction time.  Bromide shows a negative
effect on CHC13  formation. In the THM species models, chlorine  dose  appears  to
negatively affect CHBrCl2  and CHBr3; this is  likely just  a statistical anomaly
arising  from the simultaneous interaction of both  chlorine  and bromine (from
chlorine oxidation of bromide) in forming brominated THM species.

      Table 5.3 shows coagulated-water models for  TTHM and THM species
formation with iron as a coagulant.  A comparison of model exponents shown in
                                      75

-------
   1200
             Y = 20.364+ 0.71739X
             R2 = 0.99
j" 100°
"&
0>
£   800
>   600
TJ
_O

a-
Q.-
"5
E
CO
    400
    200
                200      400      600      800     1000
                      Predicted TTHM by TTHM Model
                                                              1200
      Figure 5.4   Summation of Predicted THM Species vs Predicted
                  Raw-Water  TTHM; Weight-Based Models (\ig/L)
                              76

-------
       Table 5.2   Predictive  Coagulated-Water  Models  for TTHM and  THM Species:
                                      Alum  Models*

 Alum Coagulated Water Models:

        [TTHM] ^ 100-651 [DOC]°.7S2 [CI2]°-246 [Br]0.185 tO.258
                       N = 143     F = 224     aSO.0001
        [CHCI3] =101-331[DOCJ1-11 [Cl2]°-324[Br]-0.532 tO.341
                       N = 143      F = 323     a<0.0001
                 =10-0-203[DOC]0.504 [CI2]°-126[Br"]0.474 tO.187
             = 0.84     N = 143      F=181     asO.0001
        [CHBr2CI] =10-5-398[DOCp-943 [Cl2]-0-228EBr]2.678 tO.175
          R2 = 0.31      N = 143      F = 15      a£0.0001
        [CHBrs] =10-9-6[DOCp.203
          R2 = 0.73     N = 143      F = 92      a £ 0.0001
 [TTHM] = Total Trihalomethanes

 [CHCI3], [CHBrCI2], [CHBr2CI], [CHBr3] = Individual Concentrations of four THM Species

 [DOC] = Dissolved Organic Carbon in Coagulated Water (mg/L)
       1.00£[DOC(mg/L)]<;7.77

 [Cl2] = Applied Chlorine (mg/L)
       1.11 £ [CI2 (mg/L)] £ 24.75

 [Br] = Concentration of Bromide (|ig/L)
       36£[Br(ug/L)]<308

 pH = 7.5

Temp = 20 °C

t = Incubation Reaction Time (hour)
*Source Waters: BRW, HMR, ISW, SPW, SRW, SXW, VRW, PBW
                                            77

-------
      Table 5.3    Predictive Coagulated- Water Models for TTHM and  THM Species:
                                       Iron  Models*

Iron  Coagulated Water Models:

       [TTHM] = 100-387 [DOCJ9-839 [Clj*]0-2?7 {Br]0.259 Jp.270
          R2 = 0.88      N = 143      F=248       a £ 0.0001
       [CHCI3]=101-092 [DOC]1-179 [Cl^0-378 [Br]-0-454 tO.326
          R2 = 0.92      N = 143      F = 372       a £ 0.0001

       [CHBrCl2]=10-0-416 [DOCJO-599 [Cl?]0-125 [BrjO-533 tO.205
          R2 = 0.84      N = 143      F=177       a £ 0.0001
       [CHBr2CI]=10-5.127[DOC]0.194 [Cl^0-433 IBr]2.427 tO.294
                       N = 143      F=18        a 5 0.0001
       [CHBr3]=1 0-9-427 [DOCJ-0-329 [OaT0'035 [Br]4.335 tO.307
          R2 = 0.72     N = 143      F = 91         a5
[TTHM] = Total Trihalomethanes (jig/L)

[CHCI3], [CHBrClzl, [CHB^CI], [CHBra] = Individual Concentrations of four THM Species Oig/L)

[DOC] = Dissolved Organic Carbon in Coagulated Water (mg/L)
       1 .00 £ [DOC (mg/L)] £ 7.77
[Cfc] = Applied Chlorine (mg/L)
       1.1 1 5 [CI2 (mg/L)] < 24.75

[Br] s Concentration of Bromide (ug/L)
       36 £ [Br (ug/L)] < 308
pH = 7.5

Temp = 20 °C

t = Incubation Reaction Time (hour)
*Source Waters: BRW, HMR, ISW,  SPW,  SRW, SXW, VRW, PBW
                                             78

-------
Tables 5.3 and 5.2 indicate that the net effects of the two different
coagulants on TTHM and THM species formation are similar.  A comparison of
alum and iron shows that they both have similar capabilities of DOC reduction
and THM precursor removal.

      Table 5.4 presents coagulated-water models which are based on the
combined data base derived from both alum and iron coagulation  {n = 286). Data
simulation is shown in Figure 5.5 where predicted TTHM is based on the
combined alum plus iron treated-water model. Over the higher concentration
range, TTHM is under predicted; overpredictions are seen over lower
concentration ranges. In contrast to Figure 5.5, which shows data simulation
with the entire eight-water data base. Figure 5.6 portrays data simulation  for
data with each source water separately identified. It can be seen that the
model simulates TTHM data well for seven of eight waters, with the exception
of SXW,

      Coagulation not only removes bulk DOC but also may preferentially remove
more reactive THM precursors. Precursor reactivity can be described through
use of a reactivity coefficient, :

            tj> = (THM/DOC)trE/(THM/DOC)raw         where  = 0 to  1.0
            (5.2)

      Consideration of precursor reactivity can provide an alternative
modeling approach for coagulated waters. In this approach, THM  levels in
treated waters can be estimated by using raw water models in which a  {)DOCtrt
term is substituted for the general DOCraw term; the DOCtrt reflects the reduced
precursor level while the (J> term adjusts for the reduced reactivity of the
precursor remaining compared to the raw water precursor reactivity. The
precursor reactivities of the eight sources evaluated at 2, 24, and 96 hour
reaction times are summarized in Table 5.5.  Except for SRW and ISW, the $
values ranged from 0.55 to 0.86 for alum-treated and from 0.54 to 0.87 for
iron-treated waters.  Figure 5.7 shows predicted versus measured values for
alum and iron coagulated/treated waters using the raw/untreated water-model
coupled with the   (24-hour) concept.  A comparison of predictions from the
treated-water model versus the raw/untreated-water model coupled with   (24-
hour) and without 0 is shown in Figure 5.8, shown as an extension of Figure
5.7.  The predictions provided by the treated-water model are most accurate.
The predictions by the raw/untreated-water model without $ correspond to
overpredictions; predictions by the raw/untreated-water model coupled with  4>
represent underpredictions. Thus, the treated water models have most merit  in
predictive capabilities, while the 0 based models represent an alternative
                                      79

-------
       Table 5.4     Predictive Coagulated-Water Models for TTHM and THM Species:
                             Combined Alum plus Iron Models*

  Weight-Based  (|j.g/L)  Models:

         nTHM]=100.518poCJt>-801 [Cl^-261 (BrjO-223 tO.264
                        N = 287      F=458        a < 0.0001
         [CHCI3]=101.211 fDOCJ1.149 Pa]0'345 [BrJ-0-492 fO.333
           R2=r0.91      N = 287      F = 680        a < 0.0001
        [CHBrCl2j=1 0-0-311 [DOCJ0.556 [Clg]0-121 [BrjO-505 tO.196
           R2 = 0.83     N = 287     F = 351        a 5 0.0001

        [CHBr2CI]=1 0-5^48 [DOCp-55 [CI2]0-105 [Br]2.549 10^34
                                   F = 34        a £ 0.0001
        [CHBr3]=1 0-9-5 [DOC]-0-075 [Cy-0-34 [Br]4.409 tO.313
           R2 = 0.72      N = 287     F = 185        a £ 0.0001

 Molar-Based  (nmoles/L) Models:

        [TTHM}=100.188 [DOCJO-B87 [Cla]0-276 [BrjO-084 tO.276
           R2 = 0.89      N = 287     F = 561     a£0.0001
 [TTHM] = Total Trihalomethanes Oig/L) or

 [CHCIa], [CHBrCy, [CHBraQ], [CHBr3] = Individual THM Species (ng/L)

 DOC] = Dissolved Organic Carbon in Coagulated Water (mg/L) or (mMote/L)
        1 .00 < POC (rngfl.)] £ 7.77,  0.0833 <, [DOC (mMote/L)] 5 0.648
 [Cy = Applied Chlorine (mg/L) or (mMde/L)
        1 .1 1 S [Cfe (mg/L)] < 24.75,   0.0156 < [CI2 (mMole/L)] 5 0.349

 [Br] = Concentration of Bromide (ng/L) or (jiMote/L)
       36 S [Br (pg/L)l £ 308,   0.451 S [BT (pMoleA.)] 5 3.899
 pH =* 7.5

 Temp = 20 °C

 t = Incubation Reaction Time (hour)
•Source Waters: BRW, HMR, ISW, SPW, SRW, SXW, VRW, PBW
                                            80

-------
   400
   350
   300
_J
0) 250
           [Predicted] = 22.338 + 0.742[Measured]

           R2 = 0.93, n =
   200
•o
0)
   150
   100
                               O
                                C
        0     50    100    150    200    250    300    350    400


                          Measured TTHM  (|J.g/L)



    Figure  5.5    Predicted versus Measured Values of TTHM for
                 Coagulated/Treated Waters Using Combined Alum
                 plus Iron Treated-Water  Models
                              81

-------
400 .
350 ~
300 ~
-
I 25°"
£
X
Predicted T1
-* M
Ol O
O O
,,!,,,,!
-
100~

50 ,
O
o
0 °
Ofc
0 °
B o o
m a °
_^S i i II
1^^*"
*P^
m
f

0 VRW
D IWR
0 SXW
A SRW
X BRW
+ SPW
S HMR
ffl PBW
-p 	 1 	 1 i i r i 	 r ' •"•





i
     0      100     200     300     400     500     600     700


                       Measured  TTHM (\ig/L)
Figure 5.6    Predicted versus Measured Values of TTHM for
             Coagulated/Treated Waters Using Combined Alum
             plus  Iron Treated-Water Models; Individual Sources
                              82

-------
Table 5.5  Summary of Reactivity Coefficient, 0, Values for TTHM
Water

SPW Raw
Alum
Iron
BRW Raw
Alum
Iron
SRW Raw
Alum
Iron
HMR Raw
Alum
Iron
PBW Raw
Alum
Iron
ISW Raw
Alum
Iron
VRW Raw
Alum
Iron
SXW Raw
Alum
Iron
DOC
(mg/L)
4.19
2.56
2.58
3.54
2.72
2.63
4.18
2.56
2.55
5.73
4.29
4.23
10.6
4.6
4.2
2.78
1.00
1.03
3.67
2.56
2.35
10.55
7.77
7.63
Br
(ng/L)
312
306
308
250
245
245
50
48
46
40
37
38
97
97
97
84
83
83
71
68
68
68
67
67
0 (2 hr)

1.00
0.89
0.93
1.00
0.83
0.92
1.00
0.82
0.54
1.00
0.50
0.50
1.00
0.57
0.59
1.00
1.01
0.85
1.00
0.67
0.62
1.00
0.88
0.84
0 (24 hr)

1.00
0.77
0.67
1.00
0.81
0.80
1.00
0.28
0.38
1.00
0.56
0.54
1.00
0.67
0.62
1.00
1.03
0.90
1.00
0.69
0.63
1.00
0.86
0.87
0 (96 hr)

1.00
1.00
1.03
1.00
0.57
0.61
1.00
0.72
0.67
1.00
0.76
0.68
1.00
0.62
0.62
1.00
1.00
1.01
1.00
0.74
0.67
1.00
0-.85
0.90
                                 83

-------
    500
O
             [Predtelecfl = 7.967 + 0.726[Measured]

               = 0.94, n =
                     i—P——i—i—i—i——i—i—i—i——i—i—i—i——i—i—i—i——i—i—i—
               100     200     300     400     500     600     700
                             Measured TTHM


Figure 5.7   Predicted versus Measured  Values of TTHM for Alum and
            Iron Coagulated/Treated Waters  Using Raw/Untreated
            Water-Modelswith 0 (24-hour) Concept.
                               84

-------
   120"
   100"
_J
"o*
3  80'
X
I-
•o
0>
?   60-
    40'
       	Alum Treated Model
       	Raw Model with 0 (24 hr)
       	Raw Model without 0
       	Perfect Prediction
20 ~!  '   ' — ' — I — ' — ' — ' — I — ' — ' — ' — i — ' — ' — '
  20         40          60         80         100
                      Measured  TTHM ()ig/L)
                                                                120
  Figure  5.8   Comparison of Predictions from Treated-Water Models
              versus Raw/Untreated Water  Models with 0.
                              85

-------
framework  based on precursor activity.

EFFECTS,OF BROMIDE ON THM SPECIES FORMATION

      Bromide ion  (Br~)  is often found in drinking water sources through
various pathways including geochemical weathering, connate seawater, and
seawater intrusion. The concentrations of Br" in natural surface and ground
waters, except seawater, exhibit a wide range from less than  0.01 mg/L  to more
than 1 mg/L, with an average of about 0.1 mg/L. During chlorination, Br" is
oxidized by chlorine to form hypobromous acid  (HOBr) and/or hypobromite ion
{OBr->  (Gordon,  1987):

                    HOC1 + Br'   -  HOBr +  Cl-           (5.3)

      Hypobromous acid can react with natural organic matter  (NOM)  to produce
brominated DBFs such as bromoform. Collectively, the relative amounts of HOCl
and HOBr determine the THM species distribution.

      Gould et al. (1983) studied the effects of Br~ on total trihalomethane
and individual THM species formation kinetics. They formulated the  order of
THM species formation kinetics: CHC13 < CHBrCl2 < CHBr2Cl < CHBr3  (Gould,
1983) . In other words, the THM species having higher Br" concentration form
faster than those having less Br~  concentration.

      Of the four THM species, three are brominated.  The amount of bromide
ion present influences the overall THM formation as well as speciation.  Of
course, besides this inorganic THM precursor, DOC represents  the organic THM
precursor. Through examination of scatter-plot trends, we found that the ratio
of Br'/DOC,  the  ratio of inorganic to organic precursor,  most accurately
captures bromide effects on THM speciation.  We developed fractional
concentration models for each species as a function of the ratio of Br'/DOC.
In these models, fractional THM species concentrations, ranging from 0  to 1.0,
were defined by the ratio of THM species/TTHM; thus, the individual species
fractional concentrations should sum to 1.0  for TTHM = 2(THM  Species).
Fractional concentration models as function  of Br'/DOC at reaction times of 24
hours and 96 hours are shown in Tables 5.6 and 5.7, respectively. These models
are based on both raw and treated waters; thus, they can be used to predict
THM species for both raw and treated waters.  As Br'/DOC increases,  CHC13
decreases exponentially; the two intermediate species, CHBrCl2,  and CHBr2Cl,
increase then decrease in a polynomial pattern. In theory, CHBr3 should
increase in an inverse  (S-shaped)  manner as  a function of Br'/DOC.  However,  we
found a better statistical fit by using a polynomial function with  the
                                      86

-------
           Table 5.6 Summary of  Fractional-Concentration THM  Speciation
                                  Models;  24-Hours
                                    Exponential Fit
                          Species = (a)(expbx) where x = Br/DOC
Species
Chloroform
S
^^^^S^^^^^^^^^^^^^^^— s^^e*^-^^— *^^»
^^—••^^^^^^^^"^••^^^^^^^^fc^^^™^*™ ^^^« p^fl^UJ
Species
Bromodichloromethane
Dibromochloromethane
Bromoform
S^^^^SSS^^^^SpS^^^^SS^^^^^S^^^S^^^^^j^S
a b
0.670 - 9.667
2nd order
pecies = a + bx +
— •^^^^^HSS^^^^^H.^^E
a
0.218
- 0.0285
- 0.0315
Polynomial Fits
ex2 where x = Br
=SS^=S^^=5^=!
b
1.817
4.679
1.782
-/DOC
=^^=sss^=
C
r2
0.93

- 6.465
- 10.236
-2.619
=sas:^==:^=
r2
0.54
0.95
0.93
(n = 71 total; = 43 for raw waters; = 28 for treated waters)

-------
         Table  5.7  Summary  of Fractional-Concentration  THM Speciation
                                Models;  96-Hours
                                  Exponential Fit
                        Species = (a)(expbx) .where x = Br/DOC
Species
Chloroform
a
0.727
b
• 9.178
r2
0.95
                              2nd order Polynomial Fits
                       Species = a + bx + ex2 where x = Br/DOC
Species
Bromodichloromethane
Dibromochloromethane
Bromoform
a
0.170
- 0.0318
- 0.0266
b
2.375
4.524
1.516
c
-7.460
- 10.359
-1.496
r2
0.69
0.95
0.93
(n = 71 total; = 43 for raw waters; = 28 for treated waters)
                                         88

-------
  stipulation  of  boundary  conditions  to  prevent  decreasing predictions  at very
  high  BrVDOC levels. Data simulations are presented in Figures 5.9 and 5.10,
  showing measured and predicted THM  species  (fractional concentrations) versus
  Br'/DOC (mg/mg)  at 24-hour and 96-hour reaction times, respectively.

  SIMULATION AND  VALIDATION OF TTHM PREDICTIVE MODEL

       Internal  data simulation/validation has  been discussed in the previous
  sections. External validation can be conducted by employing literature data
  not used in  the calibration of the model. Based on available literature data,
  an external validation of the TTHM predictive  raw-water model was performed by
  using data from a data base created by James M. Montgomery Engineers  (1991).
 All measured TTHM values from eight utilities  in their data base are employed
  in the validation. These eight utilities have  the following raw water
  characteristics: pH values from 6.8 to 8.5 with an average of 7.59; TOC
 concentrations from 3.0 to 11 mg/L with an average of 5.51 mg/L; bromide
  levels from less than 10 to 430 ug/L with an average of 119 ug/L.  The THM
 formation experiments done by JMM were conducted by providing a chlorine dose
 from 3.0 to 25.3 mg/L (C12/TOC of  1.0 to  2.3 mg/mg) with an average of 10.93
 mg/L;  reaction temperature was kept constant at 20 °C; reaction  time was
 varied from 0.1  to 98.7 hours.  Figure 5.11 shows predicted TTHM versus
 measured TTHM provided using the JMM data base, here highlighting  molar
 predictions.  Eighty cases (n = 80)  are  involved.  Regression of predicted
 versus measured  values shows a^generally good fit of data,  with an intercept
 of 0.0218,  a  slope of  1.174  and a R value of 0.917.  The value  of the slope (>
 1)  indicates  that  the  model  slightly overpredicts.

       An attempt to compare  two  TTHM predictive models, the one  developed  in
 this study  and the  original  model  developed  by  Amy and Chadik  {1987 {hereafter
 the  "Amy model") was accomplished  using  these  same TTHM data  from the JMM
 data base. A  linear regression of predicted  TTHM  by the "Amy model"  against
 measured TTHM shows an intercept of  -0.181,  a slope of 1.31, and a  R value of
 0.958. The  "Amy  model", based on its regression slope,  appeared to
 overpredict even more than the model developed  in this study. The accuracy of
 the predictions  depends on the conditions underlying model development such as
 source water  selection. The experimental matrix design used in creating the
 data base is  also important. Fewer reaction times selected over shorter time
periods may cause inaccurate predictions over shorter time frames.  In this
study,  only three reaction times were selected  within 24 hours while in the
development of "Amy model" seven reaction times were selected within 24 hours.
This could be one of the reasons why the R value for the "Amy model" is
greater than the one developed in this study since 70% of JMM's data falls

                                      89

-------
1II11


S sT -w > "s
O CQ  m CD  t>

o o  o o  1

o a>  o o>
73 15  S ^>  "5
TJ T3  "S "O  C
£ £  £ £  §
Q. 0.  0- Q.  CO
 5 i I  5 I


 i^M
 E ow o  E *
 —J™ fc W  k- ^
 O CD CD  CO £
 X X X  X 3
 O O O  O CO
      • -4  B
 T—1—1—I—I—|—I


oo       to
        T—i—i—i—i—i—i—r



             CO
                                                      0)
                                                      c
                                                      o
                                                     TJ

                                                     0
                                                      3

                                                      O

                                                      X

                                                      4-
                                                      OJ
                                                      0>
                                                      •o
                                                      o
                                                      c
                                                      o
                                                      o
                                                      o
                                                      a
                                                      (O
                                                      a>
                                                      o
                                                      c
                                                      o
                                                      O
                                                      i

                                                      a
                                                      c
                                                      a

                                                      o
                                                      a
                                                      O)

                                                      in


                                                      a>
                                                      w
                                                      3

                                                      D)
                UOIIOBJJ
                           90

-------
o CD m co
i z x x
o o o o
O m co m
x i z i
o o o o eo
                 '' *£
                                 tJ
                                 o>
                                 O
                                 X
                                 1
                                 (O
                                 O)


                                 jo
                                 
-------
    a

    «   2
    •o
    o
                [Predicted] = 0.0218 + 1.174[Measured]


                R = 0.92
         0.00   0.50    1.00   1.50   2,00   2.50    3.00   3.50   4.00



                           Measured TTHM  (nMole/L)



Figure 5.11   Overall External Validation Using JMM Data with Raw Water

             TTHM Model (Data from JMM Utilities 3, 5, 6,  7, 12,  21, 26 and

             33 at TOC=3.0-11 mg/L,  pH=6.8-8.5, CI2=3.0-25.3  mg/L,


             Br'=10-430 M,g/L,  Temperature=20°C and Time=0.1-98.7 Hours)
                                  92

-------
within a reaction time  range  of  less  than 24  hours.

       Figure  5.12 shows an  external validation of  reaction  kinetics with the
TTHM raw water model  using  JMM utility 3  data in which  the  raw water
parameters  and experimental condition were as  follows:  TOC =3.53 mg/L,  C12
dose = 7.7  mg/L,  Br' = 10 ug/L, pH = 6.7 and  temperature = 20 °C,  All
predicted values  fall within  " 20 percent of  the measured values.

       Figure  5.13 shows a simulation  of the effects of  coagulation on TTHM
formation by  the  combined alum plus iron  treated-water model.  This simulation
shows  that  when DOC is  reduced from 4  to  2 mg/L, TTHM is reduced  by slightly
more than 50%  when all  other  independent  variables are constant.  In this  way,
the  coagulated-water models can be used to assess  the effects  of  coagulation
on THM formation.

       An external  validation  of the TTHM  treated water model was  conducted by
using  selected data from the  JMM data  base (JMM 1992). The  results of this
validation  analysis are  summarized in  Figure  5.14, which shows measured versus
predicted TTHM values derived from the alum treated-water model and the
combined alum  plus iron  treated-water  model.  The six measured values shown are
from six different utilities  reflecting the following water qualities: pH
values  from 6.8 to 7.5;  TOC (after coagulation) from 2.9 to 3.5 mg/L;  bromide
levels  from 30 to  360 ug/L. The TOC removal ranged from  17% to 50%. The
experiments were conduced at  a reaction temperature of 20 ^C,  a reaction time
of 16 hours, and chlorine dose from 2.7 to  8.4 mg/L.  Figure 5.14 shows
excellent predictive capabilities by both  the alum treated-water model and the
combined alum plus iron  treated-water model.
                                      93

-------
     200
     150
£
X

H


"B

c
o

o
  0>
  o
  c
  o
  o
     100
                                          O   Actual data


                                         	Predicted value

                    20
                              40         60


                               Time (hour)
80
100
Figure 5.12  External Validation of Kinetics Using JMM  Data with TTHM


            Raw Water Model (Data from JMM Utility 3 at TOC=3.53 mg/L,

            CL Dose=7.7 mg/L,  Br'=10 ^g/L, pH=6.7  and T=20  °C)
              2
                               94

-------
     200
£
X
l-
c
Q>
o
c
o
O
                    DOC = 4 (mg/L)



                    DOC = 2 (mg/L)
                      50
    100

Time (hour)
150
200
    Fig. 5.13  Simulated Effects of Coagulation on THM Formation
              (Simulated by Raw and Combined Alum plus Iron Treated

              Water Model at Bf = 100 |ag/L, pH = 7.5, CI^DOC = 1,

               Temperture = 20 °C)
                              95

-------

96

-------
                                   SECTION 6

                             CHLORAL  HYDRATE  MODELS

    Chloral hydrate {CH)  is a chlorination by-product of increasing regulatory
 interest.  Its molecular  formula is C2HC13(OH)2,  and it  is  an unstable compound
 which may decompose into chloroform  or undergo oxidation to trichloroacetic
 acid (Reckhow and Singer,  1985).  The actual  mechanism for chloral hydrate
 formation is unclear;  it has been suggested  that,  during chlorination,  chloral
 (trichloroacetaldehyde)  is first  formed which then hydrolyzes into chloral
 hydrate.  Upon high-pH  hydrolysis,  chloroform is the decomposition product of
 chloral hydrate {Miller  and Uden,  1983).

    This chapter presents a model  of  the kinetics of chloral hydrate (CH)
 formation.  The effects of influential independent variables on CH formation
 are discussed and modeled.  Predictive raw/untreated water and treated water
 {after coagulation)  models have been developed.

 RAW/UNTREATED WATERS

    Measurements of CH  were made for  each of  eleven source waters,  including
 two groundwaters,  BGW  and MGW.  Since the DOC values of the two groundwater
 sources were low,  1.20 and 2.44 mg/L,  detectable levels of CH were not
 observed upon chlorination in these  sources.  Statistical  analysis showed  that
 the experimental results derived  from these  two sources strongly influenced CH
 predictive model capabilities.  Thus,  we elected to exclude these data from the
 data base  for the CH predictive model.   Thus,  the  final predictive raw-water
 models for CH are based  on nine source  waters:  BRW,  HMR,  ISW,  SLW,  SPW, SRW,
 SXW,  VRW and PBW (n  =  622).

 Effects of  Independent Variables  on  CH  Formation

    The effects  of  six  independent  variables  {model parameters);  DOC,  chlorine
 dose (C12) , bromide level  (fir'), pH,  temperature (Temp),  and reaction time
 (t) ,-  on CH  formation have  been  investigated.  Figure 6.1 shows  that  four of
 these variables   exert positive effects;  bromide shows  a  negative  effect, and
pH  shows mixed  effects.  For  some waters,  pH exhibited a clearly, positive
effect while, for  other  waters, CH first  increased then decreased with
increasing pH from 6.5 to  8.5. A control  experiment demonstrated that
hydrolysis of CH becomes significant at pH levels  of higher  than 9.0, with
effects particularly pronounced at pH levels  of  higher  than  10.  Since the CH
models stipulate a boundary  condition of  pH =  8.5,  CH hydrolysis does not
                                      97

-------
  3D
  35
  25-
^ 20-|

O
  15-
         -VBW

         -HMR
                  1.5    2

                  O.: DOC
18    20    22    24     26

  Tompture (°C)
          100     200    300     400    500
                                            _

                                            S   :
                                            5 is:
                                                               7.5
                                                               PH

                                                 1.2' 2.44' 2.79' 3^4' 3.57
                                                             DOC(mgfl.)
 Figure  6.1   Individual Parameter Effects on  CH Formation:  Effects of
              Chlorine, pH, Temperature, Bromide,  DOC, and Reaction
              Time (other  parameters  held at baseline condition).
                                     98

-------
affect  the models developed herein. The effect of Br" is particularly
noteworthy; it is hypothesized  that a brominated analog  of  chloral hydrate  is
formed  in the presence of Br~ at the expense of chloral hydrate itself.

Predictive Raw/Untreated Water  Model for Chloral Hydrate

   Through statistical analysis by stepwise multiple  regression, we developed
models  in the following format:

         Logarithmic Models:   logY = log b0  +  bx log Xi + , ..     (6.1)
                                      or
               Power-function Models:  Y = 10*° X^1. . .      (6.2)

                                 where Y = CH
                         Xi = independent variable(s)
                         bi = regression coefficients

   The  logarithmic or power-function model format provided  the best simulation
of CH formation. Table 6.1 summarizes the predictive  raw-water model for CH.
The model exponents indicate the positive influences  of  DOC, chlorine  dose,
temperature, pH, and reaction time on CH. A polynomial function for the pH
term was evaluated; however, this approach did not improve  model simulation
capabilities. The bromide exponent shows a negative influence on CH formation.
Boundary conditions for all independent variables are shown in Table 6.1.
Modeling testing, in the form of scatterplots of predicted  versus measured
values, is summarized in Figure 6.2. The model shows  underprediction over
higher  concentration ranges.

COAGULATED WATERS

   Predictive treated-water models for CH formation were generated based on
eight source waters; BRW, HMR,  ISW, SPW, SRW,SXW, VRW and PBW. Alum, iron and
alum plus iron (combined) models are presented in Table  6.2.  In these
treated water models, there are only four independent variables: DOC, chlorine
dose (C12/DOCS=  1 and 3),  bromide  level  (amb.  +  spikes),  and reaction time.  In
chlorination experiments with coagulated waters, temperature was maintained
constant at 20 °C,  and pH at  7.5.  Generally,  alum and iron  coagulation  showed
similar capabilities of CH precursor removal. Figure 6.3 shows data
simulations of predicted versus measured values of CH for coagulated/treated
waters using the combined alum plus iron treated-water model; the model
generally tended to underpredict CH values.
                                      99

-------
            Table 6.1  Predictive Raw-Water Models for Chloral  Hydrate (CH)<
        [CHJ =10-1-971 [DOCf1-
-------
    120
    100
     80
 D)
 X
 o

 •o
 0)
 •o
 0>
60
Figure 6.2
            [Predicted] = 3.5803 + 0.7137[Measured]

            R2 = 0.89,   n = 622
                           O
                  50            100           150


                       Measured  CH (|ig/L)



        Predicted versus Measured Values  for Raw/Untreated
        Water CH.
                                                                  200
                              101

-------
         Table 6.2  Predictive  Coagulated-Water Models for  Chloral  Hydrate (CH);
                    Alum, Iron, and Combined  Alum plus Iron Models*
 Alum Coagulated Water  Model:
        [CH] = 100-816 [DOCJ0.806 [Cfcp-^fBrj-O.eOI tO.402
          R2 = 0.87     N = 143      F = 225

 Iron Coagulated  Water  Model:


        [CH] = 100.694 [DOC]0.76 [Cl2]a423[Br]-0-573 tO.404

          R2 = 0.88     N = 143      F = 247
 Combined Alum plus Iron Model:

        [CH] = 100.755 [DOC]0.785 [Cl^0-375 [Brj-0.586 tO.403
                      N = 286     F = 474     aSO.0001
 [CH] = Concentration of Qhloral Hydrate (jig/L)

 [DOC] = Dissolved Organic Carbon in Coagulated Water (mg/L)
       1.00 < [DOC (mg/L)] < 7.77

 [Cl2] = Applied Chlorine (mg/L)
       1.11 5 [C\2(mg/L)}<, 24.75

 [Br] = Concentration of Bromide (jig/L)
       37 £ [Br (ug/L)] < 308

 pH = 7.5

 Temp = 20 °C

 t = Incubation Reaction Time (hour)
       2
-------
80
       [Predicted] = 1.8166 + 0.8888[Measured]

       R2 = 0.91, n = 288
               20         40         60

                       Measured CH (|ig/L)
                                  80
100
  Figure 6.3
Predicted versus Measured Values of CH for
Coagulated/Treated  Waters Using  Combined
Alum plus  Iron Treated-Water Models.
                         103

-------
    An alternative approach,  based on the CH formation reactivity of the DOC,
 was also assessed,  based on a reactivity coefficient, :

                       $  =  (CH/DOC)trt/{CH/DOC)raw      (6.3)

    A summary of the reactivity coefficient,  $,  values for CH is shown in Table
 6.3.   Values of 4> at reaction times of 2,  24,  and 96 hours are calculated and
 listed.  For the eight waters,  one water (SXW)  had  values of greater than
 1.0;  one water (BRW)  had cj> values of less  than 0.3,  and the other six waters
 had $ values reflecting  an average of 0.67.  Figure 6.4 shows predicted versus
 measured values of  CH for  coagulated/treated waters using the raw/untreated
 water-model combined with  the $ (24-hour)  concept.  Using this approach,  the
 term (^(DOC^,.)  is  substituted  into the raw  model  for  the  DOC term.   The
 scatterplot (Figure 6.4) demonstrates that this  modeling approach results in
 some  degree of underprediction for CH.

    The effects of coagulation on  DOC reduction and CH precursor  removal can
 be  simulated by using the  predictive treated-water model for CH formation.
 Figure 6.5  shows  that after DOC is reduced from  4  to 2 mg/L by coagulation
 (TOC  = 2  mg/L represents the  proposed regulatory action  level),  CH values are
 reduced by  more than  50%,  based on simulation with the combined alum plus iron
 treated water model  (Br~ =  100 ig/L,  pH = 7.5,  C12/DOC  =  1 mg/mg, and
 temperature = 20  °C) .

 SURROGATE CORRELATIONS BETWEEN  CH AND TTHM OR CHLOROFORM

   Correlations between CH and  TTHM  or  chloroform are  shown in Figure 6.6,
 representing  plots of measured  CH versus measured TTHM or chloroform for the
 entire data base  and  ranges of parameters  such as pH (6.5 - 8.5) .   These plots
 suggest that  CH exhibits a strong linear correlation with chloroform or  TTHM.
This implies  that THM predictive  models can be used  to approximate CH
 formation through these correlations.  (Poorer correlations were  observed
between total HAAs and either CH  or  TTHM).
                                      104

-------

Table 6.3  Summary of Reactivity Coefficient,  0,  Values for CH
Water

SPW


BRW


SRW


HMR


PBW


ISW


VRW


SXW




Raw
Alum
Iron
Raw
Alum
Iron
Raw
Alum
Iron
Raw
Alum
Iron
Raw
Alum
Iron
Raw
Alum
Iron
Raw
Alum
Iron
Raw
Alum
Iron
DOC
(mg/L)
4.19
2.56
2.58
3.54
2.72
2.63
4.18
2.56
2.55
5.73
4.29
4.23
10.6
4.6
4.2
2.78
1.00
1.03
3.67
2.56
2.35
10.55
7.77
7.63
Br
(ng/L)
312
306
308
250
245
245
50
48
46
40
37
38
97
97
97
84
83
83
71
68
68
68
67
67
0 (2hr)

1.00
1.02
0.73
1.00
0.41
0.64
1.00
1.11
0.93
1.00
0.74
0.71
1.00
0.58
0.55
1.00
0.64
0.58
1.00
0.66
0.62
1.00
1.01
0.87
0 (24 hr)

1.00
0.50
0.40
1.00
0.22
0.25
1.00
0.98
0.96
1.00
0.61
0.59
1.00
0.73
0.63
1.00
0.54
0.53
1.00
0.84
0.75
1.00
1.12
1.12
0 (96 hr)

1.00
0.71
0.41
1.00
0.28
0.19
1.00
1.24
1.00
1.00
1.02
0.85
1.00
0.75
0.69
1.00
0.73
0.75
1.00
0.93
0.86
1.00
0.93
1.01
                         105

-------
     35
     30






     25

j"
"en


Q    20
X
o
-D    15
0>
0>
k.
D.
     10
             [Predicted] = 1.748 + 0.933[Measured]

             R = 0.98
   Figure 6.4
                         10      15      20      25

                            Measured CH (jig/L)
                                                          30
35
                Predicted versus Measured Values  of CH for
                Coagulated/Treated Waters  Using  Raw/Untreated
                Water-Models  Combined with 0 Concept;  24-Hour
                Predictions.
                              106

-------
      25
O)
z
O
c
O
+*
a
+M
c
0)
O
c
O
U
      20-
15-
10-
               	DOC = 4 (mg/L)


               	DOC = 2(mg/L)
                                          Treated.Water Model
                          1	1	r
                                       ~\	1	1	r
                      50
                              100


                          Time  (hour)
150
200
Fig. 6.5 Simulated Effects of Coagulation on  CH Formation (Simulated
         by Raw and Combined Alum plus Iron Treated Water Model at

         Br' = 100 fig/L, pH =  7.5, CI2/DOC =  1, Temperture = 20 °C)
                             107

-------

       200
       150-
              [CH] = 0.906 + 0.129[CHCy

              R2 = 0.94, n = 786
                  200      400      600      800     1000     1200

                              Chloroform
       200
              [CH] = -4.220 + 0.117[TTHM]

              R2=0.92, n = 786
                   200     400      600      800     1000     1200
Figure 6.6  Correlations between CH and Chloroform (top)
             or TTHM (bottom).
                                 108

-------
                                   SECTION 7

                             CHLORINE DECAY MODELS

      Chlorine is the most widely used chemical for disinfection of drinking
waters in the U.S. Evaluation of chlorine disappearance during chlorination
not only can help drinking water treatment plants optimize disinfection
conditions, but also can help one understand the kinetics and amounts of
disinfection by-products  formed during chlorination.

      This chapter presents and discusses the experimental results of chlorine
consumption during DBF formation, and the development of chlorine decay
models.  Also, this chapter describes DBF formation from the perspective of
the concept of chlorine exposure time, represented by C-T, corresponding to
integration of the chlorine residual versus time decay curve. This concept
represents a different perspective on chlorine demand.

CHLORINE RESIDUAL DECAY MODELS

      The chlorine decay model proposed by Quails and Johnson  (1983)
represented a semi-theoretical approach to predict free chlorine
concentrations as a function of time in natural waters. However, this model
was based on a total reaction time of only five minutes. They have indicated
that reaction of chlorine with fulvic acid occurs in two stages: a first stage
(within one minute) representing a fast reaction  following by a second stage
(from 1 minute to 5 minutes) representing a slow  reaction. The rate of
chlorine disappearance was derived as the sum of  two first-order reactions
{fast and slow reactions)  (Quails and Johnson, 1983):
                 d[C!2J/dt  =  kr[Cl2][Fl]  +  k2[C!2][F23
(7-1)
where d[C!2]/dt is the rate of chlorine disappearance, [Cl2]  is  the molar free
chlorine residual concentration, ki and k2  are the rate constants  for  fast and
slow reaction, respectively, and  [Fl] and  [F2] are  the molar concentrations of
reactive sites on fulvic acids  for  fast and slow  reactions, respectively.
      In this study, the bench-scale experiments were performed to evaluate
chlorine decay, based on five different C12/DOC ratios (Cl2/DOC  =  3,  2,  1.5,
I, 0.5 mg/mg) and six reaction times  (2, 12,  24, 48, 96, 168 hours)  at  a
temperature of 20 °C,  a pH of 7.5,  ambient bromide levels,  and ambient DOC
values. A total of 11 source waters were evaluated in these experiments.
                                      109

-------
        Through an examination of all chlorine decay curves,  the  resultant
  kinetics generally suggested pseudo first order disappearance due  to  the
  presence  of DOC. However, chlorine decay was significantly more rapid within
  the first twelve hours, with much slower disappearance noted thereafter> The
  "intersection" of rapid and slow decay occurs somewhere between reaction times
  of 2 and 12 hours. Two additional chlorine decay experiments were designed to
  more precisely determine the location of this -intersection point-. In these
  additional experiments, the reaction times chosen were 0.5, 1,  2, 5, 9,  12,
  24,  96 and 216 hours; other experimental conditions were maintained the same
  as conditions in the previous  experiments.  Curve fitting indicated that the
  intersection point between rapid and slow decay occurred at about five hours
  of reaction time.  Within the first  five hours of reaction time,  chlorine
  concentrations  dropped from 30% to  80% depending on C12/DOC  ratio;  30% to 50%
  when C12/DOC = 3, 2 or  1.5 mg/mg, and  50% to 80% when C12/DOC =  1 or 0.5
  mg/mg. Based on  this  analysis,  the  first order chlorine  decay model for  -fast
  decay? was  defined as:
                       =  C0 exp <-
0 <. t
                                                   5 hours
                    (7-2)
 where GI is the predicted chlorine residual (mg/L as cia)  at  time t (hours),  c
 is the initial concentration of chlorine when reaction time  is zero  (i.e.,
 chlorine dose), and k, is the first order reaction rate constant with the'
 units of hour-*. The reaction rate  constant k, was found to have a very strong
 dependency on DOC, ammonia concentration (mg/L as N), chlorine dose, and
 C12/DOC  ratio.  An  extensive tabulation of k, values appears in Table  7.1 for
 the baseline experiment along with other experiments within the orthogonal
 matrix for each of the source waters. This tabulation clearly shows the
 positive effects of DOC and temperature on chlorine short-term chlorine decay-
 the effects of pH  and bromide are mixed.  The tabulation of k, values  shown in
 Table 7.1 are both source- and experiment-specific,  in an attempt to
 generalize,  the following empirical relationship for predicting k± was
 derived:
                      0.442 + 0.889 In(DOC) + 0.345 ln(7 . 6* (NH3-N) )
                       1.082 ln(CJ +  0.192  ln(C!2/DOC)      (7-3)
                                   R2 = 0.62

      After five hours of reaction time, chlorine decay exhibits  slower
kinetics. The concentration of chlorine residual after t = 5 hours  shows  first
order decay with a lower reaction rate:
                                      110

-------
CO
o
0)
a

eu
O
<••»
O

I

n
«

o
O
c
o

o

oc


lO


•*



CO



CM




a>
e
3
O
CO
10
CO
X
Q.
®
iq
I
a
®

O
ID
CM
II
a
f
®

O
in
I'-
ll
I
®

*
0)
.£
CD
CO
CD
£
0)
I
s

j£
2

Si

SJ

5

2

^£

S!

T-
J£

CO
O
LJJ
in
CO
O
*
CM
CD
LU
CD
O
CO
CM
CO
O
UJ
«t
CO
CO
9
LU
CO
•*•
CM
CO
O
ill
CM
CM
CO
CO
s
LU
CT>
r^
O
CO
CO
O
LU
O
•
CD
Oi
*~
8
LU
O)
ai
O)
CM
9
LJJ
GO
CO
CM
CO
CO
O
LLJ
Ol
CD
CO
CO
s
LLJ
CO
CO
O
LLJ
S
CD
CO
9
LU
<•*
CM
CO
CO
O
LLJ
TT
CD
1^
CO
9
LU
CO
CD
CD
^5-
CO
O
LU
CM
CO
in
C\J
CM
o
LU
•
o>
CO
CM
CD
LLJ
CO
CO
•
CO
r^

o
LU
r^
o
T—
^"
9
LU
CM
in
CO

o
LJJ
CO
¥
CO
9
LLJ
CD
r^
•^
CSJ
TT
O)
in
CO
?
LLJ
O
CO
in
1~
3
LJJ
T-
LO
CO
?
LJJ
^t
CO

s
LU
O)
O
CO
O)
3
LLJ
R
-*

s
LJJ
CO
o
CD
CM
O
LLJ
£
•*•

9
LU
1^-
CO
CD
CO
CO
CM
O
r
LU
CO
CO
CO
*-
CM
O
<
LU
CO
t
f1^
CO
9
LU
^f
O
CM
?
LLJ
O)
1^
CO
o>
9
LU
ai
O)

o
LU
CD
5
*~
S
LLJ
LO
m
CO

9
LU
CO
8
O)
8
LU
O)
m
01

o
UJ
N-
m
q
LT
£
I
CM
o
LLJ
CO
CD
"fr
T-
?
LLJ
CO
CO
cn
a>
?
LLJ
CO
CO
CO
*•
9
LU
r^
r^
a>
S
LU
CM
CM
^~

O
LJJ
O)
h-
CM
*~
S
LLJ
m
in

T—
O
1
LU
N.
CO
o
•••
?
LLJ
in
m
U3

CM
O
LLJ
fc
CM
a>
CC
CO
CM
CD
LU
r^
TJ-
in
••"
?
LU
CM
"fr
CM
CD
CM
O
LLJ
Ol
r^.
TT
^~
?
LU
CD
O
O)
CD
s
LU
S
CD

S
UJ
CM
in
O)
CO
CM
o
LU
O)
cn
CM

s
LU
^~
CO
o
h*
s
LLJ
N-
CO
CD

?
LLJ
CD
N:
i
a.
                                                                           LT
                                                                           m

                                                                           •o
                                                                           c
                                                                           CO


                                                                           a.
                                                                           CO
                                                                           CO
                                                                            c
                                                                            o
                                                                           CO
                                                                            II

                                                                           o
                                                                           o
o


c
u
Id


CO

II

CO
2



2
CD



CD m
£ OL

E-o
03 c

II *°

0<=
O GC

Q W

in LT
                                                                            CM
                                                                              cn
                                                                            3 CD
                                                                            £ 03
                                                                            c -*
                                                                            o  c
                                                                            S  o
                                                                            O  II
                                                                            0 O

                                                                            18
         S

         jg


         1
                                              111

-------
 O
O
 03
 O
 O
o
 «


 o
£
u

•5
 0
+*

 ID
O
U

£
(D
cn

0


o



O)





CD


rv



CO
(D
2
3
O
CO

o
c
o
©

u>
O
II
O
u
n
*i
o
O

^,
2
o
in
li
k_
CD
<
@
«
a."
o
in
CM
ii
CO
<3
O

1
O
IO
11
m
<
@
Waters

r-.
CO
CM
o
U]
to
o
•*
CM
8
LU
CO
cn
o
^-
9
111
1^-
r-.
in
CM
CD
s
LLJ
OS
en
CD
3
LU
h-
r*-
o>
CM
O
LU
CO
in
in
in

o
LJJ
O)
CO
'-

<
Z

^
Z
CO
O
LJJ
h-
CO
in
CO
9
LJJ
O>
O
CO
•*
CO
o
LIJ

9
LJJ
r-
o
o
u>
CO
o
LU
o
r*.
CO
**"
9
LU
§
•*
CO
o
LU

CO
"*"
CM
O
LU
CO
O>
•*
CC
CO
<
Z
<
z
CM
O
LU
CO
CO
*•

o
LU
CO
Ol
o
*~
3
UJ
h-
CO
CO
^
9
LU
•^t
CO
r-
CO
CM
o
LLJ
CO
^~
CM
"•"
CM
O
LLJ
S
CD
CM
O
LLJ
O
CO
*~
S
LU
CM
CO
CD
in
QC
>
<
Z
<
z
CO
o
LU
m
CO
in
CD

9
LU
in
^~
CD
O>
CO
O
LU
O
CO
o
tn
?
UJ
CO
CO
CO
CO
CO
o
Lil
o>
CD
CO
in
3
Lil
•*•
in
i^
m
CO
o
LLJ
S
CD
•*
CM
O
LJJ
CO
O
CD
O
<
Z
<
z
CO
O
LJJ
r*.
CO
h-
CO

S
LLJ
•*•
•t—
CM
•«•
CO
o
LU
O)
CO
CO
S
UJ
CO
*•
at
CO
S
LLJ
g
CM
CO
CM
O
LLJ
CO
•*
in
CO
CO
o
LU
at
•«•
CM
CO
3
LU
CO
?
CO
O
m
<
z
<
z
3
LU
CM
^~
r*.
CO

o
LU
CO
in
CM
CO
o
UJ
r*-
00
in
en
o
LLJ
o
CO
o
CM
9
LLJ
CD
O9
O

O
LJJ
5
CO
*- '
9
LLJ
OJ
CO
CM

O
LLJ
CO
CO
in
t~
CO
<
z
<
z
CM
O
LJJ
m
•*
CM
••-

9
LU
•^I
t-
f-
-^
3
LU
T*
CO
•^>
•^
g
LJJ
CO
5
oi
3
LU
O>
CD
Tf

CM
O
LU
N.
CD
in
en
9
LLJ
at
CO
•*

S
LU
in
o
CO
CO
CO
<
Z
<
z
CO
o
LJJ
cn
T—
CD
h-

O
LU
CO
CM
CM
3
UJ
in
cn
h-
••-
o
LJJ
CO
in
01
9
LU
CD
CD
CO

O
LLJ
S
CM
•^
9
LU
CO
CO
CO

o
LJJ
O
CM
CM
CC
X
<
z
<
z
CO
o
LLJ
CO
in
-«*•
CM

o
LU
CO
O)
—
CM
O
LU
1
-*-
O
LLJ
r^
CO
CO
3
LU
O>
O>
•*

O
LU
cn
CO
^~
CM
O
LLJ
CM
CO
CO

*—
O
LLJ
in
CO
CC
CO
<
Z
<
z
CM
O
LLJ
1^
CD
*•
••-

O
LU
CO
CO
CM
•^
S
LJJ
m
CO
in
••-
CM
O
LJJ
cn
CM
at
r^
CM
o
UJ
r*-
CM
in

9
LU
CM
in
r-
°
LU
S
CD

CM
O
LJJ
S
r^
i^
m
D.
                                                                              CO

                                                                              c
                                                                              CO
                                                                              Q_
                                                                              CO
CO
 c
 o
CO
 II
o
O
O
fc
o

"c
 (D
O
P


g

CO

                                                                             O C
                                                                             :-E o
         o>
                                                                                •w-    &
                                                                                 n
                                                                             0 0
                                                                             18
         I

-------
•o
0)
o
o
                                                                                                  cc
                                                                                                  CD

                                                                                                  T3
                                                                                                  C
                                                                                                  03



                                                                                                  Q_
                                                                                                  CO
 
^
CM
O
UJ
^
CO
CM
^
CO
o
UJ
m
CM
O
00
0
UJ
R
CM
^

>
_|
CO








^
z


CO
o
r
UJ
CO
^
^
CM
O
Ul
CD
CM
CO
^
CM
O
Ul
o
CO
CM
^
CM
O
LJJ
in
8
o>

5
Q_
CO



$
Z



^
Z


CO
o
1
UJ

q

CM
O
LJJ

CO
O
CO
CO
0
UJ

CO
CO
m*
8
LJJ
CM
CO
^

^
DC
CD
CO
o
t
UJ
00

CM"
Si
LJJ
p
CM
CM"
CO
o
1
UJ
en
CM
|v
CO
9
LJJ
fv.
CO
OS
CM"
CO
o
UJ
CM
0)
CO
CO
s
LJJ
1
CO

5
IT
>
CO
o
1
UJ
CO
CD
o
CM*
8
i
LJJ
O
m
in
-*
CO
o

LJJ
m

CM*
9
LJJ
CO
CM
O
tn*
CO
o
UJ
CO
in
CO
CM
3
i
LJJ
in
o>
T-
cd

5
o
s
CO
o
1
LJJ
CM
in

•""
8
i
LJJ
CO
CM
CM
CM
CO
O
i
LJJ
O>
^
CM*
9
LJJ
•^
""4"
^~
CM*
CO
O
UJ
m

tn
CM
3
i
LJJ
R
CO
CM

^
O
m
CO
o

LIJ
m
CO
in
co"
s
UJ
O)
CO
CO
CO
o

UJ
i
o
^
g
1
UJ
CM
o>
CO
in
CM
0
i
UJ
^f
CM

W
3
i
LJJ
CO
CM
l>-

>
CO

CO
0

UJ
CM
|v
CD
"*
?
LJJ
O

CO
CO
0
1
LJJ

^
CO
s
UJ

^
CO
^
CM

x
CO
CO
o

LJJ
CM

CD
"*
3
i
LJJ
^
00
^"
CO
o
1
LJJ
CM

O>
9
LJJ
^~
fi.
O
CO
CM
0
LJJ
00
o
o
CM
CM
O
LJJ
I
CD

LT
HE
X
CO
o
1
LJJ
N.
o>
CO
8
LIJ
O
CD
in
CO
o

UJ
o
CD
^-
CO
9
LJJ
^>
CO
CO
CO*
CM
0
LIJ
^
CD
CM

CM
O
LLJ
O
00

^
or
CO
CO
O

UJ
•r—
O
CO*
8
LJJ
CD
O>
CO
CO
o

UJ
(O
CM

in"
s
1
UJ
CD
T-
^~
^**
CO
o
LLJ
fx.
CO
CM
f^
C\l
LLJ
CO
in

5
CD
Q_
                                                                                                             "6

-------
                C2 = C0.s exp{-k2t)      5  <.  t ^  168 hours     (7-4)

  where C2 is the predicted chlorine residual (mg/L)  as C12> after  t  =  5  hours,
  C0_5  is  the chlorine  concentration when  reaction time is 5 hours,  and k2 is the
  first order  chlorine decay rate constant for t > 5 hours.

        C0_s can be obtained using the boundary condition of d  = C2 at t  = 5
  hours;  by adding Equation (7-2) and (7-4) at t - 5 hours, we obtain:
                         C0_5 = C0 expEStkz-kJ]      (7-5)

       Thus, Equation  (7-4)  can  be expressed without C0_5  term as:

        C2 = C0  exp[5{k2 - k^]  exp(-k2t)       5  <;  t  <. 168  hours     (7-6)

       Table 7.1 also shows  an extensive tabulation of  k2 values; long-term
 chlorine decay is positively correlated with DOC,  temperature,  and pH.
 Generalizing beyond the source- and experiment-specific  values  tabulated in
 Table 7.1,  k2  can be estimated from the following empirical expression:

             In(k2) = -4.817  +  1.187 In(DOC)  + 0.102 ln(7 . 6* (NH3-N) )
                  - 0.821 ln(C0)  -  0.271 ln(C!2/DOC)     (7-7)

                                    R2 = 0.72

       Figure 7.1  shows predicted and observed chlorine decay at five different
 C12/DOC ratios  using data derived  from  ISW and  the  predictive models shown in
 equations (7-2) and  (7-4).

       Statistical analysis  indicated no correlation between bromide and
 chlorine  decay.

       A comparison between  the Quails and  Johnson model in equation (7-1) and
 the models  developed in this study,  equations (7-2) and (7-4), was conducted
 for predictions of chlorine  residual. At C12/DOC  ratios of  0.5,  1.0, 2.0,  and
 3.0, the  fits of  Quails and  Johnson model  were  not  as good for prediction of
 free chlorine residual,  particularly  for long reaction times,  because the
Quails & Johnson model  was established  on  the conditions  of low C12/DOC ratio
and a  5-minute short reaction period. In full scale treatment,  a higher
C12/DOC ratio and  longer reaction time are involved. Therefore,  the
application of the models developed in  this  study are considered to be more
appropriate than the Quails  and Johnson model.
                                      112

-------
   10
O)
E
o
             D
             O
- Predicted CI2/DOC=3   o    Data CI2/DOC=1.5


  Data CI2/DOC=3     	Predicted CI2/DOC=1


 - Predicted CI2/DOC=2   A    Data CI2/DOC=1

  Data CI2/DOC=2     	Predicted CI2/DOC=0.5


 - Predicted CI2/DOC=1.5  A    Data CI2/DOC=0.5
             D
                    D
                                  D
        L~O
                                                      n
             Tr-
                                  o
       *  A
                    50
                   100

               Time (hour)
150
200
     Figure 7.1    Predicted and Observed Chlorine Decay at Five
                  Chlorine/DOC  Ratios, ISW.
                                113

-------
 DBF FORMATION VERSUS CHLORINE EXPOSURE  (C-T).

       The formation of chlorination by-products can potentially be evaluated
 through use of the chlorine exposure  (C-T) concept which is the integration of
 the chlorine residual versus reaction time curve, with units of mg/L*min. We
 attempted to discern possible correlations between chlorination DBF formation
 and C-T, a function of both chlorine dose and decay.

       Figure 7.2 portrays total HAA as a function of C-T value for
 raw/untreated or treated waters at a temperature of 20 °C,  a pH of  7.5,
 ambient DOC,  ambient bromide,  a C12/DOC  =  1 mg/mg, and reaction times  of  2,
 12,  and 24 hours.  It is apparent that the pattern of chlorine
 decay/consumption represented by C-T does not provide an accurate indication
 of HAA formation.  Differences  may be partly attributable to variations in
 ambient Br" and its role in forming brominated HAAs. It  is  interesting,
 however,  that there  is  some  clustering of results for coagulated  waters,  with
 all  sources  except SXW  behaving  comparably.

       Figures 7.3  and 7.4  show total  THMs  and chloral  hydrate,  respectively,
 as a  function of chlorine  exposure for raw/untreated and treated waters under
 the same  conditions as  shown for HAAs. Raw-water  responses  are  highly  variable
 for both  THMs and  CH while treated waters  showed  some  more  comparable  trends
 except SXW.

       All three figures  (Figures 7.2,  7.3, and 7.4) show that each source
water  exhibits a somewhat  unique chlorine  demand. Also,  the ambient Br'
manifests itself differently in  each source water in forming brominated DBFs
 (see Chapters 4 and 5). It appears that, after coagulation, the NOM remaining
in each of the sources exhibits more similar chlorine demands.  Establishment
of simple correlations between chlorination DBPs and the C-T parameters did
not prove viable.
                                     114

-------
              1000    2000     3000    4000
                          CT V«ioes
-------
                                                         BOW
                                                         BGW
                                                         HMR
                                                         ISW
                                                         MGW
                                                         sxw
                                                         SLW
                                                     m   VRW
                                                         paw
                   1000
2000
                                 3000    4000

                                 C-T (mg/L-mm)
                     5000
6000
                                    7000
        250
                   1000
                           2000     3000     4000
                                C-T (mg/L-min)
                           5000
                                   6000
Figure  7.3  TTHM as a Function of Chlorine Exposure (C-T) for
            Raw/Untreated (top) or Treated  (bottom) Waters
                               116

-------
                                               BRW
                                               BGW
                                               HMR
                                               ISW
                                               MGW
                                               sxw
                                               SLW
                                               VRW
                                          —e— PBW
35
        30 -

        25 -

        20 -

        15 '

        10 :

         5 -

         0
Figure  7.4
         1000   2000
                               3000    4000
                                C-T (mg/L-min)
5000
6000
7000
                                               -BRW
                                               -HMR
                                               -ISW
                                               -SXW
                                               -VRW
                                               •PBW
                                               •SPW
                                               -SRW
          1000
                                                     6000
                        C-T(mg/L-min)
     CH as Function of Chlorine Exposure (C-T) for
     Raw/Untreated (top) or Treated  (bottom) Waters.
                       117

-------
                                    SECTION 8

                         BROMATE AND  OZONE DECAY MODELS

    During the oxidation of natural waters containing bromide ion (Br~) with
 ozone  (O3),  bromate is formed at concentrations ranging from 0-150 ug/L under
 normal water treatment conditions. Bromide itself  occurs  ubiquitously, with an
 average concentration in the U.S. of almost  100  ug/L {Amy et al.,  1994).
 Current studies show that bromate is a carcinogen. The Environmental
 Protection Agency  (EPA) is currently considering a Maximum Contaminant Level
 (MCL) of 10 ug/L in U.S. drinking waters.  Considering that the  average bromide
 ion concentration in U.S. waters is  100 ug/L, it is  expected that  detectable
 bromate will form in a majority of waters which are  subjected to ozonation.
 Therefore, an understanding of bromate formation during ozonation  and the
 quantitative effects of water quality parameters (pH,  alkalinity,  DOC, etc.)
 is crucial for evaluating various bromate control strategies.

    Haag and Hoigne (1983)  suggest that bromide can be  oxidized, by ozone,
 according to the following reaction:

                    03 + Br- -  02 + OBr-  k= 160 irV1   (8.1)

    The above reaction  is pH  dependent because  hypobromite  ion  (OBr-)  is in
 equilibrium with hypobromous  acid (HOBr)  according to the  following reaction:

                    HOBr  - BT + OBr- pKa= 8.7  at  25°C    (8.2)

    Hypobromite ion  can  further react  with  ozone to  form bromate;

                  203  +  OBr- - 202 + BrO3-   k=  160 M^s'1    (8.3)
   The  formation of bromate during ozonation can be influenced by water quality
parameters   (pH,  DOC,  Br~,  temperature)  and  various  operational/treatment
conditions  {O3 dose, dissolved O3 residual,  contact time).

PARAMETERS AFFECTING BROMATE  FORMATION

   An assessment of water quality characteristics  (e.g.  Br~, pH, DOC) can help
determine if a bromate problem is likely to  occur.  An  understanding of
treatment options can help  reduce bromate formation. The effects of each
parameter on bromate formation and ozone demand are discussed below.
                                      118

-------
 Effect  of  pH

    The  effects  of  pH on  bromate  formation and ozone decay have been analyzed
 for each water  source, and it was found that bromate concentration increases
 upon an increase of  pH from 6.5  to 8.5,  a trend largely attributable to the
 high OBr'/HOBr ratio at higher pH  levels. pH has been controlled in  our
 experiments  with a 10"3 M phosphate buffer; pH was monitored during
 experiments  to  assure that pH was constant within ± 0.1 units.

    On the  other hand, ozone decays faster in high pH waters than  low pH
 waters.  This is consistent with  what  other researchers  have found (Weis,  1935;
 Hoigne  and Bader,  1977;  Staehelin et  al.,  1984;  Tomiyasu et al.,  1985;  Grasso,
 1987; Gordon, 1987).  This  behavior is attributed to direct reaction of
 hydroxide  ions  with  ozone:

                   03 + OH' - H02 + 02-   k=  70 M"1 s"1    (8.4)

    Figure  8. la  shows  the effects of pH on bromate formation for SPW;       ^
 increasing pH from 6.5 to  8.5 almost  doubled the bromate concentration.

 Effect  of  Ozone Dose

    For  low DOC  waters (DOC £ 210 mg/L) ,  an O3/DOC = 2 mg/mg was selected as
 the baseline condition;  however  for the higher  DOC waters "(DOC  >  2  mg/L),  an
 O3/DOC = 1 mg/mg was selected in order to provide realistic ozone doses.

    Bromate concentration was observed to  increase  with  an increase  in 03/DOC
 ratio (Figure 8.1b). This  is due to the direct  reaction of ozone with OBr' to
 form BrO3". Bromate does  not form after the ozone residual becomes zero
 (Figure  8.2). This is an important finding because this suggests  that bromate
primarily  forms  in treatment plants whereas most of the organic DBFs (e.g.,
bromoform) form within distribution systems after  several hours of  reaction
 time. This is also potentially advantageous because bromate  can potentially  be
removed  before  treated water leaves the treatment  plant.

    In the  case of high 03/DOC ratios with low DOC waters, bromate formation
steadily increases with  time whereas  ozone  decays  very  slowly until  all
bromide  is converted into bromate.

Effect of Bromide Concentration

   Bromate formation and ozone decay  in natural waters  have been studied at
                                      119

-------

iq in 10
GO h* to
i 44
o n *

UJ
X
o.
"S"
_i
^ » Ob
s *? ^
i **• o ^
co m Q o"





1











1 «

\ \
k 1
\ 1 \

• CD

LO
10
: w
rg

LO
vf>
• ^r
* O
i- CO
LO
- CM

L O


O O O O Q
CD CM CO ^





•E-
I

g
p






       o to


         4
   o  & 1.
                                m
                                CM
 ooooooooo
 CD
 O


 O

O
2



m

 of
 CO
 o
o

 0)


 8
o


 Q.

"5
 co ,
                                                  -S


1 *
£ £
UJ
1 H
i
1
£
^
lit
§ « n
S J! O "«
£ (500
....,....,,.


^J
u
-
> i



> i
•



1 '


g i
n
— h
1 <



•
1 -
J, -
»n

\\
• • •• i " '


>
3
i
1







\
fm-r)
O
- CO
L o
I
L O
- CD
L 0 £
: w E
-§1
0 K
m
_ o
CM
_ O


— O
o    in   o   in   o
ro    CM   01   i-   *-
                                               2
                                               m
                                               m e-
                                               sS
                                               
   120

-------
              85
           o>
           o
           k
           ffl
              80
              75-
              70
              65-
              60
                    SPW
                    pH=7.5

                    Br= 260 jig/L
                                                            •£}
                                              2.2
                                               1.8
                                                                      1.4
                                                 to
                                                                      0.6
                                                                      0.2
• i  • •  • •  i • '  • •  I '  • • '  i '  ' •  • I  • •  ' ' i  • •  • •  i -0.2

10    20     30     40     50    60     70
                                      Time  (min)
Figure 8.2.  Effect of Dissolved Ozone  on Bromate  Formation
                                        121

-------
 ambient fir", and at spiked levels of ambient + 0.1 mg/L  (100 ug/L) and ambient
 +  0.2 rag/L  (200 ug/L) . In the case of high ambient Br~ levels  (Br~ > 80 ug/L),
 ambient Br" was chosen as the baseline condition, whereas for the low ambient
 Br~ levels,  ambient +0.1 mg/L was the baseline condition used in order to
 produce measurable quantities of bromate. A national survey  indicates  that Br"
 levels in raw US drinking water supplies ranges  from <0.005  mg/L  to almost 0.5
 mg/L with a national average of slightly less than 0.1 mg/L  (Amy  et al., '
 1994); therefore, the baseline conditions were maintained close to these
 levels.

    Bromate concentration increased with increasing bromide ion concentration
 (Figure S.lc),  due to direct reaction of ozone with bromide  to produce OBr"
 which further reacts with ozone to produce bromate.

    Siddiqui and Amy (1993)  previously showed that Br
-------
 effects of temperature on bromate formation were studied at 15, 20, and  25°C,
 with 20°C  chosen as  the baseline condition.  The results show that bromate
 formation increases with increasing temperature  (Figure 8.Id). This is
 generally consistent with what previous researchers have found. Siddiqui and
 Amy (1993) attribute this to a combination of the temperature-dependent
 increase in rate constants and a decrease in the pKa of HOBr-OBr" with an
 increase in temperature.

    Temperature also has a direct effect on ozone decay; an increase in
 temperature brings about a decrease in the dissolved ozone concentration
 (Roth and Sullivan,  1981; Hewes, C. et al. ,1971; Sotelo, J.L. et al., 1989).
 This occurs due to a drop in the liquid phase driving force and to a higher
 ozone decomposition rate. Our results agree with the literature, and a higher
 ozone decomposition was observed with high temperature values.

 Effect of  Ammonia

    The effect of ammonia on the formation of bromate and ozone decay was
 studied. NH3/O3 ratios of 0.2, 0.35, and 0.5 mg/mg were used in the
 experiments;  the NH3/O3 = 0.35 mg/mg ratio corresponds to the stoichiometric
 conversion of Br"  to OBr' and stoichiometric conversion of OBr" to
 monobromamine  (i.e.,  1 mol of O3 produces 1 mol  of  OBr" which reacts with 1
 mol of NH3 to produce monobromamine, and thus the stoichiometric  ratio of
 NH3/03 can  be calculated as; NH^/Oj = (Imol)/(Imol) =  (17g)/(48g) - 0.35
 mg/mg).  It has been  observed that addition of ammonia decreases bromate
 formation  (Figure B.le).  This is likely due  to  direct reaction of OBr" with
 ammonia to form monobromamine or reaction  of HOBr with ammonia and a
 corresponding conversion of OBr' to HOBr. Adding  excess ammonia  (NH3/03 =0.5
 vs  0.35 mg/mg)  did not  decrease the bromate  formation further.  This may
 possibly be attributable  to the reaction competition between OBr' and  O3
 versus  OBr' and NH3 :

     OBr' +  203 - Br03- + 202     k= 100 M'1 s'1  (Haag  and Hoigne,1983)     (8.6)

    OBr- +  NH3 - NH2Br +  OH'    k= 20 M'1 s'1 (Hoigne and Bader,  1985)     (8.7)
Effect of DOC

   DOC exerts a very clear negative  influence on bromate  formation (Figure
8. If) . This effect is largely in response to its positive influence  on  ozone
decay; i.e., DOC-related ozone demand. However, this effect  is very  source

                                      123

-------
specific with DOC  from more  humic  sources generally exerting a greater ozone
demand.

Effect of Hydrogen Peroxide

   Peroxide effects on the formation of  bromate and ozone decay were studied.
Our results have shown that  addition of  peroxide increases bromate formation
in some waters and decreases it  in other waters.  Thus,  more work should be
performed to understand  the  mechanisms of the influence of peroxide on bromate
formation. Hydrogen peroxide may potentially react with OBr"  to  produce
bromate:

                       H2O2 + OBr"  - BrO2- + H2O     (8.8)
                       BrO2-  + OBr" - Br~ + BrO3-   (8.9)

   Ozone reaction with hydroperoxoide ion {HOa'J  is very fast (k = 5.5xl06 M^s"1};
thus, the presence of hydrogen  peroxide in water  significantly affects ozone
decay. The ozone decomposition rate increases with increasing pH values. Peroxide
effects have been  studied only at baseline pH values, and results  show that ozone
decomposition is faster  in the peroxide  containing waters.

   Peroxide effects are also  manifested by its role in promoting hydroxyl radical
(OH-)  formation in the presence  of ozone (Hoigne and Bader,  1985),  and by its
independent role in reducing hypobromite to bromide (Haag and Hoigne,  1983):

                       H202 + 2O3   -  2OH-  + 3O2    (8.10)
                     H2O2 + OBr~  - Br- +  O2 + H2O    (8.11)

   The stochiometric ratio for peroxide  production of OH- radicals  is H2O2/O3 =
0.35 mg/mg.  Above  this ratio, there  is an excess of peroxide in the system.

Effect of Reaction Time

   Reaction time is another  parameter which affects bromate formation and
ozone decay. Results show that bromate formation is directly related to the
dissolved ozone concentration in water,  and bromate does not form after
dissolved ozone concentration goes to zero.  In a majority of the water sources
studied,  ozone concentration decreased to zero in less  than an hour (Figure
8.2), with an ozone half-life ranging from approximately 10 seconds to 30
minutes.  Thus, bromate formation does not occur beyond  a time frame of about
one hour. Ozone reaction with bromide is fast,  and most of the bromate
formation occurs in the  first five minutes  after ozone  application.
                                      124

-------
COMPARISON OF REACTOR TYPES

   In work performed herein, two different reactor configurations have been
employed: (i) semi-batch and (ii) true-batch. In the semi-batch ozonation
system, ozone is applied continuously (mg/L-min) to a batch of water (L) for a
predetermined application time (min) to achieve a targeted applied dose
(mg/L); corresponding measurement of transfer efficiency (typically = 35% to
65%) allows determination of transferred/utilized ozone dose. In the true-
batch approach, a stock solution of ozone {= 35-40 mg/1) is prepared by
exhaustive ozonation of Milli-Q water at 2-3°C  and then added to a batch of
the source water of interest, and a teflon disk cover is used to prevent any
transfer of ozone from the liquid phase to the gas phase.

   Since dissolved ozone is directly responsible for bromate formation, the
dissolved ozone profile at any time in the reactor is very important. In the
semi-batch approach, dissolved ozone concentration increases with time during
application  (with a possible lag while most ozone demand is being met)  and
thereafter decreases with time after ozone application has ceased. Bromate
also forms in response to the dissolved ozone present in water at any time
during the application of ozone. In contrast, in the true batch approach, at
time zero (the time of stock introduction), an initial system exists within
which the reactants are at their maximum and the products do not exist; after
time zero, ozone decays and bromate forms over time. Figure 8.3 shows bromate
formation and ozone decay for semi batch and true batch modes of applications.
Both reactor configurations have some advantages and disadvantages. The semi-
batch mode is physically more similar to pilot-scale or full-scale continuous-
flow ozone contactors than true-batch in terms of the continuous introduction
of gas. In these systems  (semi-batch or continuous-flow),  it is necessary to
measure transfer efficiency in order to determine utilized ozone dose.  On the
other hand,  transfer efficency is not a concern in the true-batch system,
because applied and transferred ozone are identical. In the semi-batch
application, one should always be concerned about head space in the reactor
and any possible transfer of ozone from the liquid to gas phase. The biggest
concern with the true batch approach can be associated with the dilution
effect caused by adding a small amount of ozone stock solution to an aliquot
of source water; however, dilution effects can be accounted for by redefining
the original water characteristics.

MODELING EFFORTS

   The experimental methods used in this study along with statistical
approaches are explained in Chapter 2. Three type of models that are discussed

                                      125

-------
           ^  40
           01
           n
           O
           k.
           m
                               10
                                      Time (min)
Figure 8.3.   Effect of Reactor Type on Bromate Formation
                                       126

-------
herein were developed  (i) semi-batch,  (ii) true-batch-EPA,  (iii) true-batch-
EPA+EBMUD models. Semi-batch models were developed during the beginning stages
of the project with data derived from  semi-batch ozonation  of the first five
source waters selected as part of the  EPA data base  (n=113). True-batch models
were developed during later stages with data derived  from true-batch
ozonation, consisting of two data bases; the EPA data base  alone  (n= 116) or
the combined EPA+EBMUD data base  (n= 176). The two true-batch models are
similar in format except that the former was calibrated using data  from a
smaller yet diverse data base derived  from six source waters; the latter was
calibrated with a larger data base which was influenced by  the  four California
waters comprising the EBMUD sources.

   Step-wise multiple regression was used to develop  the models according to  a
power function format; relevant statistical parameters include  the  number of
cases (n) and the multiple coefficient of determination  (R2) .  Each model was
tested through an internal validation, representing a data  simulation with
data used in model calibration. A perfect data simulation between predicted
and measured values would be presented by a regression line with a  slope of
1.0, an intercept of zero, and an r2 of 1.0;  this analysis also allows an
assessment of model over- or under-prediction.

      An important stipulation of the  following models is the indicated
boundary conditions, defined by the ranges over which each  independent
parameter was varied. These ranges were defined by the orthogonol matrix
employed: O3/DOC = 0.5,  1.0* 2.0 mg/mg; Br~ = amb. , amb.  + 0.1*, amb. + 0.2
mg/L; pH = 6.5, 7.5*, 8.5; NH3/O3  =  0*, 0.20, 0.35, 0.50 mg/mg;  temperature =
15, 20*, 25 C  (* = baseline condition; for high ambient Br"  > 80 ug/L,  amb.
was used as the baseline; for low DOC  waters ^ 2.0 mg/L, O3/DOC = 2  mg/mg was
used as the baseline) . The reaction times evaluated ranged  from 1 to 60
minutes. For the semi-batch approach,  ozone is applied at "absolute" time
"zero" until some targeted dose is achieved; thereafter  (post-03) ,  the DO3
residual is allowed to decay. The semi-batch models define  reaction time  (t)
as beginning at the end of the application period, a  "redefined" time zero.

Ozone Demand
   Bromide directly reacts with ozone to form bromate, therefore an
understanding of ozone decay and its relationship to bromate formation is
crucial. Three different ozone decay models  (semi-batch, true-batch-EPA, true-
batch-EPA+EBMUD) are shown in Table 8.1. The model exponents shown in Table
8.1 demonstrate the ozone demand of DOC and the effect of pH on ozone
decomposition. The coefficient associated with alkalinity demonstrates its
stabilizing influence on O3  decomposition.  While temperature is not included
                                      127

-------
                 TABLE 8.1-  PREDICTIVE MODELS FOR  OZONE  DECAY

 a- Semi Batch Model
 (Source Waters: BRW, SPW, BGW, MGW, HMR)
        [DOs] = 182 (DOC)'2-66 (pH)-2.66 (Qg)1.52 {Br)0-176 (t+i)-0.53
        R2 = 0.61, F= 34, N = 113, a< 0.0001
 b- True Batch Model; EPA Data Only
 (Source Waters: BGW, PBW, ISW, SPW, TYA, HMR)
        [D03] = 357 (DOC)-1-79 (pH)^ (O^'-H (Br)0-08 (l)-°-59 (Alk)0-22^
        R2 = 0.66, F= 65, N = 116, a< 0.0001
 c-  True Batch Model; EPA+EBMUD Data
 (Source Waters: BGW, PBW, ISW, SPW, TYA, HMR, EIS,  ESL, EBC, EES)
       [DOa] = 2831 (DOC)-1-942(pH)-4-79(O3)1-81 (Br)0-163^)-0-678 (Alk)0-236
       R2 = 0.68, F= 83, N = 176, a< 0.0001

 where
 Br" = Bromide (ng/L); 70 < Pr"] < 440
 DOs = Dissolved Ozone (mg/L); 0.05 ^ [DOs] < 4.6
 t = Time(min); 1 ^t<120
 pH; 6.5 ^pH^ 8.5
 DOC= Dissolved Organic Carbon (mg/L); 1.1 £ DOC < 8.4
 Os = Transferred/Utilized Ozone (mg/l); 1.1 < Os £ 10
Alk = ABtalinfty (mg/l); 13 < Alk < 316
                                       128

-------
  as a parameter in the models, it exhibits a significant influence on O3
  decomposition  (slower at lower temperature), in the semi-batch model,  the
  reaction time  (t) is based on post-ozonation (after O3 application has ceased)
  conditions. However, some ozone demand is met during ozonation; hence, the use
  of the  (t + 1) term in the model permits calculation of the ozone residual
  after ozone application at t = 0. Data simulations, in the form of predicted
  versus measured comparisons,  are shown in Figure 8.4.  All of the models show
  some overpredictions at lower values with underpredictions  observed at higher
  values; predictions are most accurate under conditions leading to medium
  levels. The ozone residual models have potential relevance in making C-T
  predictions,  a point discussed later in more detail.

  Bromate Formation Models

     Table 8.2  shows  bromate  formation models  for  semi-batch,  true-batch-EPA
  and true-batch-EPA+EBMUD data bases.  Data simulations  for  all  models  are  shown
  in  Figure 8.5.  Again, the reaction  time,  t,  is based on post-ozonation
  conditions  in  the semi-batch  bromate  model.  However, bromate can  form during
  ozonation;  thus,  use of  the  (t +  1,  term  in  the model permits  time-zero (the
  instant  when ozone application has  ceased) predictions. The regression  lines
  suggest  some propensity  towards overpredictions at  lower values, and
  underprediction towards higher values  for all the models, with best
  predictions provided under mid-level conditions. In all of the models, while
  reaction time is shown to exert a positive effect, bromate does not form after
  the ozone residual goes to zero.   Therefore, the bromate prediction models may
 potentially be coupled with the corresponding ozone residual models shown in
 Table 8.1. Once again,  temperature effects are not encompassed by the models-
 greater bromate is observed at higher temperature.

    Bromate formation models without ammonia as a parameter are also shown for
 true-batch-EPA and true-batch-EPA+EBMUD data sets in Table  8.2.  Those  models
 provide flexibility to water utilities that do not measure  ammonia as  a  water
 quality parameter,  or that do  not  contemplate ammonia addition  as  a control
 option.  Data simulations  using these models are shown in Figure 8  6  Again
 overprediction  at  lower values, and  underprediction  at higher values are
 observed.
CT Models
   In the Surface Water Treatment Rule  (SWTR), disinfection is addressed by
EPA through the use of CT values (defined as  the product of residual ozone
concentration in mg/L and effective contact time in minutes). Therefore
having a CT-based model would be useful in helping to understand the
                                      129

-------
                   ~   2-

                   |

                   -~ 1.5J
                            {Predicted^ 0.148 + O^Measured]
a)- Semi-Batch Model

            "*   .
                                     Measured  DO, (mg/L)
                        0 I i"i r-T •! i i i i f r i  i i | i i i i [ 'I i i i J i ' '  ' I
                          I     0.5    1     1-5    2    2.5     3
                                             D0a (mg/L)
                                                                2.5
                            [Predfcted]= 0.197 + 0.6(Measured]
                                      3.5
                                     M«a*urad DO, (mg/L)
Figure  8.4.   Predicted versus Measured Dissolved Ozone Using Semi Batch,
              True  Batch-EPA,  and True Batch  EPA+EBMUD Models
                                           130

-------
                 TABLE 8.2-  PREDICTIVE  MODELS  FOR BROMATE FORMATION
   a- Semi Batch  Model
   (Source Waters: BRW,  SPW,  BGW, MGW,  HMR)
         [BrOs] = 5.5x10-6 (DOC)'1-61 (pH)«-54 (03)1.2 (Br)1-06 (t+i)0.35

         R2 = 0.62, F= 38, N = 113, ct< 0.0001
  b- True Batch Model; EPA Data Only
  (Source Waters: BGW,  PBW,  ISW, SPW,  TYA, HMR)
         1- without  Ammonia
         [BrOa] = 2.74x10-6 (DOC)-1-32 (pH)5.4 (Q3)1-36 (Br}0.86 (t)0.31

         R2 = 0.73, F= 61, N = 116, 
-------
                      200
                   O
                   m
                      150-
                      100-
                           [Pre
-------
            250
                   [Predicted]= 11.65 + 0.66[MeasuredJ

                   R2= 0.82
                  a)- EPA Model Without Ammonia
                         I '  ' '  '  I '  ' '  ' I  '
                               100     150     200     250
300     350
                                    Measured BrO3"(ng/L)
            200
         ~.  150-
          O)
          GO
          •o
             100-
                   [Predicted}= 10.3 + 0.59[Measured]
                  b)- EPA+EBMUD Model Without Ammonia
             50-
                                                               300     350
                                   Measured  BrO * (|ag/L)
Figure  8.6.   Predicted versus  Measured Bromate Using  True Batch-EPA,
              and True Batch EPA+EBMUD  Models without Ammonia
                                          133

-------
  relationship between DBFs  and disinfection levels  under certain water quality
  and operating conditions.  The SWTR interpretation  of  CT involves the DO3
  residual  (C)  leaving a  continuous-flow contactor and  the t10 residence time
  (T) .  In a  batch system,  this  parameter can be  approximated by using  the
  exposure time concept developed by Von Gunten  and  Hoigne (1993)  which
  corresponds  to integration of the  ozone decay  curve  (C  vs.  t).

    Table 8.3  shows CT models  calibrated with true-batch-EPA and  true-batch-
  EPA+EBMUD  data bases. Data simulations with these  models  are  shown in Figure
  8.7;  in general, there  is very good agreement. Model exponents show  that
  increasing pH  and DOC will lower CT values, reflecting both the  lesser
  stability of ozone at higher pH values  and  the ozone demand of DOC.  CT values
  also can be translated  to bromate levels by using CT as a variable in  bromate
 prediction models. Table 8.4 shows bromate models with CT as a variable in the
 model for true-batch-EPA and true-batch-EPA+EBMUD data sets. Data simulations
 using these models are shown in Figure  8.8.  Regression lines suggest
 overpredictions at lower values and underpredictions at higher values.

    Figure 8,9 shows  (ultimate) bromate  formation as a function of (calculated)
 CT values for different water sources. The calculated values represent
 integration of DO3 vs t  curves from t  = 0  to several selected  points  along  the
 decay curve.  Besides differences in Br~ levels, the fact that,  for the same CT
 values,  each water shows different bromate formation suggests that type of DOC
 in these waters also is  important.  This CT concept  captures the ozone-demand
 characteristics of the specific DOC present. Figure 8.10 shows (ultimate)
 bromate  formation as a function of CT  for  different water sources with equal
 (adjusted)  bromide levels.  In  comparison to Figure  8.9 (variable bromide),
 there is less divergence in the bromate formation response of the different
 waters.  Nevertheless, while the divergence is  less, it is still significant,
 indicating  the influence of (type  and  amount)  of DOC.  Even for those  plots
 corresponding to almost  equal  DOC  levels,  there are still differences,
 indicating  the influence of type of DOC.

 External Validation of Models

   External validation of models is  important to check validity of the models.
 External validation involves use of  data external to the  data  base to test  the
predictive capabilities  of models. We  elected to use data  from several  studies
 summarized in Table 8.5; these include pilot studies (Siddiqui et  al.,  1993;
Krasner et al.,  1993) and continuous-flow bench-scale studies  (James  M.
Montgomery Engineers, 1992), conducted over a range of water quality  and
treatment conditions. Temperature conditions were not considered  in these
simulations; reaction time was set equal to the hydraulic residence time
                                     134

-------
                          TABLE 8.3- PREDICTIVE  MODELS FOR CT
 a- True Batch Model; EPA Data Only
 (Source Waters: BGW,  PBW, ISW, SPW, TYA, HMR)
        [CT] = 9.2 (DOC)-°-809 (pH)'1-053 (03)1-304 (t)0.691
        R2 = 0.96, F= 602,  N = 116, ct< 0.0001


 b- True Batch Model; EPA+EBMUD Data
 (Source Waters: BGW,  PBW,  ISW, SPW, TYA, HMR, EIS, ESL, EWC, EES)
       [CT] = 14.5 (DOC)-°-753 (pH)-L266 (O3)1-265 (t)0.677
       R2 = 0.96, F= 954, N = 176, a< 0.0001
 where

 CT= (Concentration)(Time) (mg-min/L); 0.7 £ CT £ 155
t = Time (min); 1 < t < 120
pH; 6.5
-------
             200
             160-
                    [Predicted^ 2.4 + 0.85[Measured]
                     .2
                                           100
                                  Measured CT (mg-mln/L)
200
            200
                    [Predicted}. -1.0 + 1.1IMeasured]
                   b)- EPA + EBMUD Model
                              50           100          150
                                  Measured CT (mg-mln/L)
200
Figure  8.7.  Predicted versus  Measured CT, True  Batch-EPA,
             and  True Batch EPA+EBMUD  Models
                                         136

-------
          TABLE 8.4- PREDICTIVE MODELS FOR BROMATE FORMATION  WITH  CT
a- True Batch Model; EPA Data Only
(Source Waters: BGW,  PBW, ISW, SPW, TYA, HMR)
       [BrO3] = 1.0X10-6 (DOCJ-0-368 (pH)6-185 (Br)0-793 (CT)°-553

       R2 = 0.7, F= 66, N = 116, a< 0.0001


b- True Batch Model; EPA+EBMUD  Data
(Source Waters: BGW,  PBW, ISW, SPW, TYA, HMR, EIS, ESL, EWC, EES)
       [BrOa]^ 3.0x10-7(DOC)-°-2H (pH)6-818(Br)°-684 (CT)°-603

       R2 = 0.7, F= 100, N = 176, a< 0.0001
where
BrOa" = Bromate (ngrt_);  2 < [BrOa"]  £ 314
Br" = Bromide (ng/L); 70 £ [Br~] < 440
t = Time(min); 1 £t^120
pH;  6.5^pH<8.5
DOC= Dissolved Organic Carbon (mg/L); 1.1 £ DOC < 8.4
CT= (Concentratbn)(Time) (mg-min/L); 0.7 £ CT < 155
                                        137

-------
            200
            160  ~
                    [Predicted}-: 8 + 0-69{Measured]


                    R2= 0.82
         0" 120
         i_
         m


         •o
         0


         1  80

         9
             40  -
        1 00


Measured  BrO
                                                         1 50
                                   200
            150
                    [Predicted^ 7.1 + 0.67[Measured]


                    B2=0.81
            120 -
                   bV EPA+ EBMUD Model wrthCT
                                                                       150
                                    Measured  BrO." (ng/L)
                                                 3
Figure  8.8.  Predicted  versus Measured Bromate  Using True  Batch-EPA,

             and True  Batch EPA+EBMUD  Models  with  CT
                                         138

-------
             80
             70-
             60-
             50-
           D>
           3 40-
          o
           k_
          m
             30-
             20-
             10-
                                                X
                                         6^142
                               Br=l03
TYA
HMR
PBW
SPW
1SW
                                  6      8     10    12    14    16
                                  CT (mg-min/L)
Figure 8.9.  Bromate Formation as a Function of Ozone Exposure (C-T);
            Variable Bromide
                                     139

-------
              110
               90
            ~  70T
            O
            CO  50"
               30~
               10'
     -TYA(DOC=£6mg/L)
     - EES (DOC= 4.1 mgfl.)
— $- - ISW(DOC=2.6mgfl_)
- -X- - EIS(DOC=3.1 mgl)
- - +- - SPW{DOC=3-2irsrt_)
                                                  Br=270>ig/L
                                                  pH*7.5
                                      X--'
                                                       -H
      T-I—i i r j i  r i r i \ i  i r-i i )  i i i  i i  j i  i i i  r
                                  6       9       12
                                     CT (mg-min/L)
                                     15      18
Figure 8.10. Bromate Formation as a Function of Ozone Exposure (C-T);
             Constant Bromide
                                       140

-------
TABLE 8.5- EXTERNAL VALIDATION OF MODELS WITH LITERATURE DATA
Water
Source
Utility #4


Utility #5

Utility #23
Utility #24


Utility #27
Utility #30
Sacramento
River Delta







Colorado
River




Ref
1


1

1
1


1
1
2







3




PH
8.0
7.0
7.0
8.4
7.4
7.3
7.2
7.4
8.2
8.1
7.3
7.3
8.3
6.8
8.8
8.8
6.5
6.5
8.8
8.0
8.0
8.0
8.0

Br
(H9/L)
360
360
360
90
90
400
260
260
260
30
30
30
160
60
500
500
500
500
500
290
290
290
290

03
(mg/L)
7.81
2.35
4.45
2.39
4.63
1.98
0.42
0.66
0.41
9.8
5.1
9.7
6.1
6.6
4.0
4.0
4.0
4.0
2
1
2
3
4

DOC
(mg/L)
4.2
4.2
4.2
5.2
5.2
3.3
0.6
0.6
0.6
4.5
4.5
4.5
6.6
3.7
5.5
5.5
5.5
5.5
5.5
3.3
3.3
3.3
3.3

Measured
BrO3(mg/L)
148
8
32
5
8
10
15
29
14
10
6
14
10
7
100
121
31
37
32
16
28
117
122

Predicted BK>3(ng/L)
EPA EPA+EBMUD
164
16
37
8
12
23
17
37
33
25
6
14
39
13
71
88
14
17
28
14
36
63
93

148
11
30
7
10.5
16
11
27
21.5
29
6
16
33
14
53
65
11
13
18
9
25
48
76

1- James M. Montgomery Engineers, " Effects of Cagulation and Ozonation on the Formation of
Disinfection By-Products", Prepared for AWWA, January 1992.
2- Siddiqui, M., Amy, G., Ozekin, K., Westerhoff. P., Miller, K., " The Role of Tracer Studies in Relating
Laboratory and Pilot-Scale Ozonation Data", 11 Ozone World Congress Proceedings, San Francisco,
1993.
3- Krasner, S.W., Gramrth, J.T., Coffey, B.M., Yates, R.S., " Impact of Water Quality and Operational
Parameters on the Formation and Control of Bromate During Ozonation", International Water Supply
Association Proceedings : Bromate and Water Treatment, Paris, November 22-24, 1993.
                            141

-------
                180'
               150'
               120'
            O

            ffi


            TJ
            2
           a
           TJ

           O
                60'
                30'
                              [Preo5cted] = 8.6 + 0.75{Measured] R*s 0.83




                              [Predicted] = 6.1 + 0.6[Measured]  R2^ 0.77
-•— True Batch-EPA Model without Ammonia



 B~ - True Batch-EPA+ EBMUD Model without Ammonia I
                                                                 D
                          20      40      60      80     100    120



                              Measured Literature BrO'ftig/L)
                                                 140
Figure  8.11. External Validation of Models with Literature Data

              (Data from Table 8.5)
                                           142

-------
 (HRT); ozone doses are transferred doses; and the data reflect a mix of pre-
and intermediate-O3 tests.  Figure 8.11 shows predicted values versus pilot
 (and  continuous-flow bench) scale values for true-batch-EPA  {without ammonia)
and true-batch-EPA+EBMUD (without ammonia) models. These model predictions
show  some underprediction at higher values and some overprediction at low
values, suggesting reasonably good predictions at intermediate values.

COMPARISON OF TRUE BATCH WITH SEMI-BATCH MODELS

   It was previously stated that bromate formation is reactor specific, a
premise that can be demonstrated by comparing the semi-batch and true-batch
models under the same water quality and operational conditions. Figure 8.12
shows predictions of bromate for different source waters using semi-batch and
true-batch-EPA models. These simulations show that the true batch model (and
its mode of ozone application) results in greater predictions of bromate
formation than the semi-batch model (and its mode of ozone application). The
difference can range from as low as 30 % to as high as 100 %. One reason for
this  difference is that there is a higher driving force  (higher DO3)  under
true-batch mode of ozone application. Another possible reason is the presence
of headspace in the semi-batch system, permitting transfer from the liquid to
the gas phase after ozone application has ceased. The hydrodynamics of the
reactor and associated mixing conditions are another possible reason for these
differences. Still another reason is how reaction time is defined in modeling
semi-batch data. The previous predictions applied to pilot-plant data are
further revealing in that the true-batch models were reasonably capable of
predicting bromate formation under continuous-flow pilot-plant conditions.

   Thus, even at the laboratory scale, bromate formation is reactor specific.
While the semi^batch mode of application more closely simulates a pilot- or
full-scale reactor in terms of continuous ozonation, the DO3  profile is much
different; a continuous-flow system provides a steady state profile across the
reactor. The true-batch mode of application results in a time-dependent DO3
profile which is similar to the space-dependent profile observed in ozone
contactors with a point of admission near the contactor entrance. We have
found that true-batch models more accurately simulate pilot- and full-scale
contactors. Such simulations were discussed in the previous section.
Admittedly, part of the poor simulations of the semi-batch reactor may be
attributed to its non-steady behavior and our assumptions on how to define the
reaction time (defined as post-application herein}.

EVALUATION OF CONTROL OPTIONS: MODEL SIMULATIONS
   An important attribute of the bromate prediction models is that they allow
                                      143

-------
10        20       30        40



  Semi  Batch BrO ' Predictions
                 9
                                                                 50
Figure 8.12. Predicted  Bromate Using True  Batch-EPA Model versus

            Semi  Batch  Model
                                     144

-------
           o>
          O
          *.
          m
              35 .
              30"
              25 ~
              20 .
15 .
               5 .
DOC=3mg/L
0 = 3mg/L
Time=10min

E3
0
m
Q
NH /O = 0.0 mg/mg
33 "
NH /O = 0.2 mg/mg
3 3
NH /O = 0.35 mg/mg
33 "
NH /O = 0.5 mg/mg
3 3
                        6.5
Figure 8.13. Bromate Control Options (Simulations of True Batch-EPA Model)
                                      145

-------
assessment of bromate control options, including pH depression and ammonia
addition. Figure 8.13 shows such simulations based on a defined set of
conditions. pH reduction from 8.5 to 7.5 to 6.5 is very influential; the major
constraint to such an approach would be the amount of acid required,
particularly for high alkalinity waters. Ammonia addition clearly has a lesser
effect. These simulations are consistent with control option assessments -
performed by others, and the designation of pH depression as the best
available technology (BAT).

ORGANO-Br FORMATION

   In the presence of DOC and Br~,  ozonation of  natural  waters can lead to
organo-Br by-products such as bromoform, bromoacetic acids, bromoacetones, and
bromoacetonitriles.  Bromoform, one of the brominated organic by-products, can
form through reaction of HOBr, acting as a substitution agent, with DOC:

                         HOBr + DOC - CHBr3    (8.12)

   It is evident from above reaction that bromoform formation is influenced by
certain water quality and operational/treatment conditions such as pH and the
presence of DOC. Figure 8.14 shows the effects of various parameters on
bromoform formation. As can be seen from Figure 8.14,  bromoform forms only in
the case of a relatively high bromide concentration and application of
relatively high ozone doses. Considering that the MCL for total THMs is 100
ug/L, with a proposal to lower the MCL to 80 ug/L, bromoform formation during
ozonation will not be a significant problem for water utilities.
                                      146

-------
                                  10  1*15    20    25   30    35
                                      •*Time (hr)
                                  10   15    20    25   30    35
                                       Time (hr)
Figure  8.14. Individual Parameter Effects on Bromoform Formation  (Semi Batch);
             Effects of pH, Ozone Dose, and Bromide Ion  Concentration
                                        147

-------
                                    SECTION 9

                                MODEL APPLICATIONS

       The models  developed herein  can  be used  to assess  the  ability  of water
 utilities to meet existing and future  DBF regulations. These models  have most
 relevance to inorganic ozonation by-products formed under pre-ozonation
 conditions, and to chlorination by-products formed either before or  after
 significant DBF precursor  removal  has  been accomplished.

 CHLORINATION BY-PRODUCT AND CHLORINE DECAY MODELS

       The chlorination by-product models can be used to assess both  in-plant
 and distribution system formation of THMs,  HAAs (HAA«) ,  and CH.  (As mentioned
 previously,  THAA models correspond to HAAe,-  HAAs regulated under Stage 1 of
 the D/DBP Rule can be estimated by summation of predictions the five relevant
 individual HAA species}.  Water quality conditions such as DOC,  pH,
 temperature,  and bromide are needed as inputs to the models;  such data then
 allow assessment of chlorination DBF formation as a function of reaction time.
 Within this  context,  time can reflect the hydraulic residence (HRT) of a
 chlorine contact basin or the average HRT of a distribution system.

       The models can be used to assess pre-chlorination scenarios  involving
 raw/untreated waters. Post-chlorination scenarios  can be assessed using either
 rsw/untreated water models, if little precursor removal has been achieved.
 Otherwise, either treated (coagulated)  water models  or reactivity-coefficient
 adjusted raw-water models can be employed (the  latter recommended if
 temperature and pH variations are significant).

       Chloramination  scenarios involving the sequence  of  free chlorine
 followed by ammonia addition  can be approximated by considering  the lag time
 between  addition of the respective  chemicals; such an  approximation would
 simply show DBPs  formed in  the presence of free  chlorine  before  ammonia
 addition.

       The effects  of  precursor removal  by chemical coagulation can  be
 assessed through use  of the treated water models. One can either predict DBPs
 formed under a given  degree of precursor removal, or can define  the degree of
precursor removal required  to  meet  given DBF regulations. The impact  of
bromide  ion on meeting regulations can  also be assessed.  If one makes the
assumption that precursor reactivity (i.e.,  DBF/DOC) changes as a function of
treatment type,  one can also assess other precursor removal processes such GAC
or membranes  through use of the raw/untreated water models  (i.e., (J)^,

                                     148

-------
  *MEMBR«JES) •  Otherwise,  if  one  assumes  comparable  effects  on reductions in
  precursor reactivity, the coagulated water models can be used to approximate
  the performance of these other precursor removal processes. Although it can be
  envisioned that there is a 020HE, it is recommended that one should not one use
  the models to approximate  post-chlorination by-products following an
  ozonation step, given the complexity of ozone effects on subsequent chlorine
  reactivity of NOM.

        Through use of the chlorine decay models, one can assess CT requirements
  and CT conditions provided under various water quality conditions.  Moreover,
  the chlorine decay models can be used to assess dosing requirements to ensure
  maintenance of distribution system residuals.

        Another potential  application  is assessing impacts on DBF formation of
  the lead and copper rule through the pH parameter.

  BROMATE AND OZONE DECAY  MODELS

        The  bromate formation  models presented herein can be  used  to  approximate
  bromate formation under  varying  water  quality  (DOC, PH,  Br") and treatment
  conditions  (O3 dose). The models can also be used to assess potential control
  strategies  (pH  depression, NH3 addition) . The major constraint to such use of
  these models  is that bromate  formation  has been  found to be  largely reactor
  specific. It  is the dissolved ozone  time/space profile within a reactor which
  is most influential in determining the  degree of bromate formation;  these
 profiles are  established through the mode of ozone application and  contactor
 hydrodynamics  (mixing). Given the two forms of models developed herein, true-
 batch and semi-batch, it is the former  that comes closer to simulating
 continuous flow contactors, either at the pilot or full-scale.

       The effects of DOC on bromate formation can be assessed to determine
 ozone point of application; either pre-O3  before any DOC removal  has been
 realized, or intermediate-Q, after chemical coagulation has achieved some DOC
 removal. In applying the models to post-coagulated waters,  one must  assume (as
 an  approximation)  that the character of the DOC remaining after coagulation
 resembles that of  the raw water.  The  effects of pH,  either as a water quality
 condition or treatment option,  can easily be assessed.  Those models  which
 include an  ammonia term were  developed to permit assessment  of NH3 addition.
 Ozone application  strategies  such as  tapered ozonation  can be assessed through
 stepwise application of the model.

The ozone decay models have relevance in both bromate minimization as well  as
CT aspects of  ozone use. The models can  be used  to approximate contactor DO3
residuals expected after a given HRT  (or t10).

                                      149

-------
                                  REFERENCES

Amy, G.; Chadik, P. and Chowdhury, Z., -Developing Models for Predicting
Trihalomethane Formation Potential and Kinetics-, Journal AWWA, 79:7:89
(1987a).

Amy, G.; Minear, R., and Cooper, W.,  -Testing and Validation of Multiple
Linear Regression Model For Trihalomethane Formation Potential", Water
Research, 21:649 (1987b).

Amy, G., et al., "The Effect of Ozonation and Activated Carbon Adsorption on
Trihalomethane  Speciation", Water Research, 25:2:191  (1991).

Amy, G., et al., "Threshold Levels for Bromate Formation in Drinking Water",
IWSA Proceedings: Bromate  and Water Treatment, Paris  (1993).

Amy, G., et al., "Bromide  Occurence:  Nationwide  Bromide Survey", AWWARF Report
 (1994) .

AI>HA,  Standard  Methods  for the  Examination of Water and Wastewater (1989).

Box, G., et al., Statistics For Experimenters, Wiley  Interscience  (1978).

Chadik,  P., and Amy,  G.,  "Coagulation and Adsorption  of Aquatic Organic Matter
and Humic  Substances: An Analysis of  Surrogate Parameters  for Predicting
Effects on Trihalomethane  Formation Potential",  Environmental Technology
Letters,  87:8:261  (1987).

Chowdhury, Z.;  Amy,  G.  &  Siddigui,  M.,  "Modeling  Effects  of Bromide Ion
Concentration on The Formation of Brominated Trihalomethanes",  Proceedings.
AWWA  Conference (1991).

 Engerholm, B.,  and Amy,  G.,  "A Predicative Model for Chloroform Formation from
Humic Acids", Journal AWWA 75:8:418  (1983).

 Gordon, G.,  "The very Slow Decomposition of Aqueous Ozone in Highly Basic
 Solutions",  Proceedings,  8th Ozone World Congress, IOA,  Zurich, Switzerland
 (1987).

 Gordon, G. Cooper, W.,  Rice,  R.,  and Pacey,  G.,  "Disinfectant Residual
 Measurement  Methods", AWWARF,  Research Report (1987).
                                      150

-------
 Gould, j., Fitchhorn, L., and Urheim, E.,  "Formation of Brominated
 Trihalomethanes: Extent and Kinetics", Water Chlorination Environmental  Impact
 and Health Effects/Vol. 4, pp. 297-310, Ann Arbor, MI: Ann Arbor Science
 Publishers, Inc. (1983).

 Grasso, D., "Ozonation Dynamics in Water Treatment: Autocatalytic
 Decomposition, Mass Transfer and Impact on Particle Stability", Ph.D.
 Dissertation, The University of Michigan, Ann Arbor, Mich. (1987).

 Haag,  W., and Hoigne, J., "Ozonation of Bromide Containing Waters: Kinetics of
 Formation of Hypobromous Acid and Bromate", Envir. Sci. Technol., 17:261
 (1983) .

 Hewes, C., et al.,  "Kinetics of Ozone Decomposition and Reaction with Organics
 in Water", A.I.Ch.E., 17:141 (1971).

 Hoigne, J.,  and Bader,  H., "Ozonation of Water: Selectivity and Rate of
 Oxidation of Solutes",  Proceedings,  3rd IDA Congress,  Paris,  France (1977).

 Hoigne,  J.,  and Bader,  H., "Rate Constants of Reactions of Ozone with Organic
 and Inorganic Compounds in Water:  III:  Inorganic Compounds and Radicals",
 Water  Res.,  19:993  (1985).

 Krasner,  S.,  et al.,  "The Occurrence Of  Disinfection By-Productsln  U.S.
 Drinking Water",  Journal AWWA,  81:  8:41  (1989).

 Krasner,  S.,  et al.,  "Impact  of  Water Quality and Operational Parameters  on
 the Formation  and Control  of  Bromate During Ozonation",  IWSA  Proceedings:
 Bromate  and Water Treatment,  Paris  (1993).

 Lynn,  S.W. "An Analytical  Survey of  Chloroform  Formed from the Chlorination of
 Humic  Substances", Ph.D. Dissertation,  University of Massachusetts  (1982).

 Miller, J., and Uden, P.,  "Characterization of Nonvolatile Aqueous
 Chlorination Products of Humic Substances", Environ. Sci. Technol. 17:150
 (1983).

Moomaw, C., Amy( G.,  Krasner, S., and Najm, I.,  "Predictive Models for
Coagulation Efficiency in DBP Precursor Removal",  Proceedings, AWWA
Conference, pp. 221-233  (1993).
                                      151

-------
Montgomery  Engineers,  "Disinfection By-Products  Database  and Model Project",
AWWA Project  Final  Report,  James  M.  Montgomery,  Consulting Engineers,Inc.
 (1991).

Montgomery  Engineers,  "Effect  of  Coagulation and Ozonation on the Formation of
Disinfection  By-Products*,  AWWA Project  Final Report, James M.  Montgomery,
Consulting  Engineers,  Inc.  (1992).

Oliver, B., and Lawrence, J.,  "Haloforms in Drinking Water:  A Study of
Precursors  and Precursor Removal",  Journal  AWWA,.  71:161-163 (1979).

Oliver, B., et al.,  "Influence of Aquatic Humic  Substance Properties on
Trihalomethane Potential",  Water Chlorination Environmental Impact and Health
Effects Vol.  4:1, R.L. Jolley., Ann Arbor,  MI: Ann Arbor  Science  Publishers,
Inc., pp. 231-242  (1983).

Ozekin, K., "Modeling  Bromate  Formation  during Ozonation  and Assessing its
Control*, Ph.D. Dissertation,  University of Colorado, Boulder (1994).

Quails, R., and Johnson, D., "Kinetics of the Short-Term  Consumption of
Chlorine by Fulvic Acid", Environ.  Sci.  & Technol., 17:692  (1983).

Reckhow, D.,  and Singer, P., "Mechanisms  of Organic Halide  Formation During
Fulvic Acid Chlorination and Implications with Respect to Preozonation*, Water
Chlorination: Environmental Impact  and Health Effects, Vol.  5  , Lewis  Publ.,
Chelsea, Mich. (1985) .

Reckhow, D.,  and Singer, P., "The Removal of Organic Halide  Precursors by
Preozonation  and Alum Coagulation",  Journal  AWWA,  76:4:151  (1984).

Rook, J., "Formation of Haloforms During  Chlorination of  Natural  Waters",
Water Treat.  Exam. 23:234-243  (1974).

Roth, J., and Sullivan, D., "Solubility of  Ozone  in Water",  Indus^ Engrg.
Chem.Fund.,  20:137  (1981).

Siddiqui, M.,  and Amy,  G.,  "Factors  Affecting DBP  Formation  During  Ozone-
Bromide Reactions", Journal AWWA,  85:1:63 (1993).

Siddiqui, M.,  et al.,  "The Role of Tracer Studies  in Relating Laboratory and
Pilot Scale Ozonation Data", llch Ozone World Congress,  Volume 1:  S-2-45
                                      152

-------
 (1993).

 Sotelo, J.,  et  al.,  "Henry's Law Constant for the Ozone-Water System", Water
 Research,  23:1239  (1989).

 Staehelin, J.,  et  al.,  "Ozone Decomposition in Water Studied by Pulse
 Radialysis to OH and HO4 as Chain Intermediates",  Jour.  Phys.  Chem.,  88:5999
 (1984).

 Tomiyasu,  H. ( et al.,  "Kinetics  and Mechanisms of Ozone Decomposition in Basic
 Aqueous Solution", Inorg.  Chem.,  24:2962 (1985).

von Gunten,  U., and  Hoigne,  J.,  "Bromate Formation During Ozonation of Bromide
 Containing Waters",  11th Ozone World Congress,  Volume 1:  S-9-42 (1993).

Wang, H.,  "Empirically  Based Kinetic Models for Predicting the Formation of
 Chlorination By-Products:  Haloacetic Acids",  M.S. Thesis,  University of
Colorado,  Boulder  (1994).

Weis, J.,  "Investigation on  the  Radical  HO2 in  Solution",  Trans. Faraday  Soc.,
 31:668  (1935).

Zhu, H., "Modeling the  Effects of Coagulation on  Chlorination By-Product
Formation",  Ph.D. Dissertation,  University  of Colorado,  Boulder (1995).
                                      153
                                              M US. GOVERNMENT PRINTING OWICE: 1999 -TM-UUMM5

-------


-------

-------

-------

-------
CO
33

i
(jo


               o -t m c

               5" n  3 —
               £?. 3- =• m
               2 2. o Q.
               3 O  2 rn
opS


cn 8 3
l\3 W g.
                         CD =5

                         - >
                            (Q
                            CD
                            3
                            O
                    CO.—   m Q. TJ
                      Sa^<   »a »

                      B °   ?S S
                    s; Q S.   o Q f
                    SO     3 -X ffi

                    g.5 *   I ? ffi.

                    8 SJ   ^ S 3
                    3 « 5    5.8
                    


-------