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
------- |