Water Research 36 (2002) 330–342
Development of chlorine dioxide-related by-product models
for drinking water treatment
Caroline Korna, Robert C. Andrewsb,*, Michael D. Escobarc
aAcres and Associated Environmental Limited, Suite 525, 21 Four Seasons Place, Toronto, Ont., Canada M9B 6J8
bDepartment of Civil Engineering, University of Toronto, 35 St. George Street, Toronto, Ont., Canada M5S 1A4
cDepartment of Public Health Sciences, University of Toronto, Canada
Received 1 January 1999; received in revised form 1 April 2001; accepted 1 April 2001
Factorial experiments were conducted using source waters from seven drinking water treatment plants in Ontario,
Canada to develop statistically based model equations capable of predicting chlorine dioxide consumption and chlorite
and chlorate formation upon chlorine dioxide application. The equations address raw water quality and operational
parameters including pH, temperature, chlorine dioxide concentration, reaction time and water organic content (as
described by non-purgeable organic carbon?ultraviolet absorbance measured at 254nm, NPOC?UV254). Terms
describing two-factor interaction effects were also included, improving the accuracy of the predictive equations in fitting
measured response concentrations as evaluated through internal and external validations. Nearly 80% of the
predictions for chlorine dioxide consumption and chlorite formation were observed to be within 20% of the measured
levels. Over 90% of the predicted chlorate levels were within 70.1mg/L of the measured levels. Chlorine dioxide
concentration and NPOC?UV254were key parameters when developing the predictive models. r 2001 Elsevier
Science Ltd. All rights reserved.
Keywords: Chlorine dioxide; Chlorite; Chlorate; Disinfection; Modeling; Water treatment
The use of chlorine as a primary disinfectant has
historically been viewed as common practice. However,
during treatment, chlorine reacts with naturally occur-
ring organic material to produce numerous by-products,
some of which are suspected carcinogens. Alternative
disinfectants such as chlorine dioxide (ClO2) could be
considered as viable alternatives to chlorine use.
Chlorine dioxide has been reported to be effective in
the inactivation of pathogenic organisms including
Cryptosporidium parvum . Further, unlike chlorine,
chlorine dioxide is not known to react with humic
substances to form trihalomethanes (THMs) [2–6].
While employing chlorine dioxide as a disinfectant
presents many advantages, concern exists over the
formation of by-products including chlorite (ClO2
and chlorate (ClO3
anemia at low levels of exposure; while higher levels of
exposure can result in an increase in methemoglobin .
Tradeoffs between achieving a desired disinfection
level and forming disinfection by-products (DBPs)
encourage the development and use of forecasting
models . In view of proposed regulatory limits for
DBPs such as chlorite ion [9,10] as well as potential
limits to be set for chlorate [11,12], the development of
equations which are capable of predicting DBPs
associated with chlorine dioxide treatment will become
increasingly useful. Moreover, the fact that a maximum
residual disinfectant level (MRDL) of 0.8mg/L has been
?) which may lead to hemolytic
*Corresponding author. Tel.: +1-416-978-5399; fax: +1-
E-mail addresses: email@example.com (C. Korn),
firstname.lastname@example.org (R.C. Andrews), escobar@utstat.
toronto.edu (M.D. Escobar).
0043-1354/01/$-see front matter r 2001 Elsevier Science Ltd. All rights reserved.
PII: S 0043 -1354(01 )0 0194-4
proposed for residual chlorine dioxide , may require
plants to further optimize their treatment processes.
While previous research into the development of
predictive by-product formation models related to
chlorination has been conducted by others [13–19],
similar model equations are lacking in the literature for
chlorine dioxide-related by-product formation.
This study presents disinfection by-product formation
results (chlorite and chlorate) obtained from statistically
designed experiments using natural waters. Equations
were developed in terms of water quality and chlorine
dioxide disinfection-related parameters including tem-
perature, pH, chlorine dioxide dose, contact time and
organic content. Significant two-factor interaction ef-
fects were also examined and included in the equations.
Care was taken to develop equations capable of
predicting chlorine dioxide consumption, chlorite and
chlorate formation within boundary conditions of the
examined parameters, while maintaining reasonable
chemical justification and ‘‘statistical soundness’’. Mod-
el validations involved comparing response levels as
predicted from the equations with ‘‘internal’’ (levels
employed as part of the original model data base) as
well as with ‘‘external’’ (levels reported in the literature
and observed upon testing waters that were not part
of the database) response levels. Sensitivity analyses
consisted mainly of evaluating parameter contributions
to the ‘‘explanatory power’’ of the relevant model
2. Experimental design
Experiments were conducted at bench-scale using
250mL amber bottles serving as batch reactors. The
selected ‘‘factorial design’’ approach differs from the
conventional experimental design of varying ‘‘one-
factor-at-a-time’’ (OFAT). Factorially designed experi-
ments allow the simultaneous variation of individual
parameters within a given experimental run . Using a
factorial design allowed the identification of not only
main parameter effects (such as pH or temperature), but
also two parameter-interaction effects (such as pH-
chlorine dioxide concentration) on DBP formation, in
both a time- as well as resource-efficient manner.
To ‘‘screen’’ for significant main parameter and
interaction effects on chlorite and chlorate formation,
fractional factorially designed experiments were con-
ducted using ‘‘synthetic’’ waters. The parameters that
were examined included pH, temperature, ClO2:NPOC
ratio, chlorine concentration, chlorate concentration
and reaction time. During these preliminary experi-
ments, sufficient data was provided to show that
two-factor interactions may be important in chlorite
and chlorate formation when employing chlorine
dioxide. As such, two-factor interactions were consid-
ered when developing subsequent DBP predictive model
Subsequent to the screening experiments, individual
24(2 levels, 4 factors) full factorial experiments were
conducted using raw waters (7 surface waters and one
groundwater). All possible parameter combinations
were examined, thus allowing the determination of
main parameter and two-factor interaction effects on
DBP formation. The experiments consisted of ‘‘two-
level’’ factorials where the effect of each of the
investigated parameters was assessed by varying the
parameter between a pre-selected ‘‘low’’ and ‘‘high’’
level (Table 1). Predictive equations were developed
from experimental data pooled from all eight factorially
designed experiments. Hence, unless otherwise noted,
each equation was developed from a database of 128
different responses (i.e. 16 experimental conditions?8
Experimental factors and levels investigated in raw water experiments
Factor Parameter values
Lowest level examineda
Highest level examineda
Chlorine dioxide concentration (mg/L)
Reaction time (h)
aSome factors were adjusted to their specified low or high levels.
bWater obtained from the Toronto R.L. Clark Treatment Plant had the lowest NPOC and UV254of the seven surface waters
examined in this study.
C. Korn et al. / Water Research 36 (2002) 330–342331
the University of Toronto, Department of Civil En-
gineering for providing technical assistance.
 Liyanage LRJ, Gyurek LL, Finch GR, Belosevic M.
Modeling cryptosporidium parvum inactivation by chlor-
ine dioxide. In: Proceedings of the Forth Environmental
Engineering Speciality Conference, Edmonton, AB, May
29–June 1, 1996. p. 17–25.
 Lafrance P, Duchesne D, Arcouette N, Pr! e evost M. The use
of ClO2: case study of Laval. In: Chlorine dioxide:
drinking water issues. Proceedings of the Second Interna-
tional Symposium, American Water Works Association,
Houston, TX, 1993. p. 147–62.
 Lykins Jr BW, Griese MH. Using chlorine dioxide for
trihalomethane control. J Am. Water Works Assoc
 Muttamara S, Sales CI, Gazali Z. The formation of
trihalomethane from chemical disinfectants and humic
 Werdehoff KS, Singer PC. Chlorine dioxide effects on
THMPF, TOXFP, and the formation of inorganic by-
products. J Am Water Works Assoc 1987;79(9):107–13.
 Li J-W, Yu Z, Cai X, Gao M, Chao F. Trihalomethanes
formation in water treated with chlorine dioxide. Water
 Gallagher DL, Hoehn RC, Dietrich AM. Sources,
occurrence and control of chlorine dioxide by-product
residuals in drinking water. Denver, CO: AWWA Re-
search Foundation, 1994. 184pp.
 Harrington GW, Chowdhury ZK, Owen DM. Developing
a computer model to simulate DBP formation during
water treatment.J Am
 Federal Register (Fed. Reg.). National primary drinking
water regulations: disinfectants and disinfection bypro-
ducts; final rule. vol. 63, No. 241, pp. 69389–69476.
Environmental Protection Agency, USA, December 164,
 Pontius FW. New horizons in federal regulation. J Am
Water Works Assoc 1998;90(3):38–50.
 Pontius FW. Disinfectant-disinfection by-product rule
update. J Am Water Works Assoc 1991;83(12):24–115.
 Health Canada, Guidelines for Canadian drinking water
quality, 6th ed. Ottawa, ON: Canada Communications
 Amy GL, Chadik PA, Chowdhury ZK. Developing
potential and kinetics. J Am Water Works Assoc
 Morrow CM, Minear RA. Use of regression models to link
raw water characteristics to trihalomethane concentrations
in drinking water. Water Res 1987;21(1):41–8.
 Hutton PH, Chung FI. Simulating THM formation
potential in sacramento delta, part I. J Water Resour Plan
 Lyn TL, and Taylor JS. Modeling compliance of chlorine
residual and disinfection by-products. Proceedings of
American Water Works Association Water Quality
Technology Conference, Miami, FL, November 7–11,
1993. p. 513–23.
 Hofmann R. Minimising chlorination and chloramination
DBPs while ensuring microbiological control. M.A. Sci.
Thesis. University of Toronto, 1996.
 Amy GL, Minear RA, Cooper WJ. Testing and validation
of a multiple nonlinear regression model for predicting
 Adin A, Katzhendler J, Alkaslassy D, Rav-Acha Ch.
Trihalomethane formation in chlorinated drinking water: a
kinetic model. Water Res 1991;25(7):797–805.
 Montgomery DC. Design and analysis of experiments. 4th
ed. Toronto, ON: Wiley, 1997. 704pp.
 Stat-Ease Inc., Design Eases. Version 3.0.0. Minneapolis,
 Minitab Inc., The student edition of Minitabs
windows software. Release 9. Reading, MA: Addison-
 American Public Health Association, American Water
Works Association and Water Environment Federation
(APHA-AWWA-WEF). Standard methods for the exam-
ination of water and wastewater, 19th ed. Washington,
 Andrews RC, Hofmann R, Di Scipio R. Drinking water
treatment to minimize disinfection by-products and ensure
microbiological control. Report, Environmental Health
Centre, Ottawa, ON: Health Canada, 1997. 246 pp.
 Masschelein WJ, Fransolet G, Laforge P, Savoir R.
Determination of residual ozone or chlorine dioxide in
water with ACVKFan updated version. Ozone Sci Eng
 Hofmann R, Andrews RC, Ye Q. Comparison of
 United States Environmental Protection Agency (USE-
PA). Determination of inorganic anions by ion chromato-
graphy, Revision 2.1. In: Methods for the determination of
inorganic substances in environmental samples, EPA/600/
 Hautman DP, Bolyard M. Using ion chromatography to
analyze inorganic disinfection by-products. J Am Water
Works Assoc 1992;84(11):88–93.
 Engerholm BA, Amy GL. A predictive model for chloro-
form formation from humic acid. J Am Water Works
 Engerholm BA, Amy GL. An empirical model for
predicting chloroform formation from humic and fulvic
Cumming RB, Mattice JS, Jacobs VA, editors. Water
chlorination: environmental impact and health effects,
Vol. 4, Bk. 1. Chelsea, MI: Ann Arbor Science Publishers,
1983. p. 243–52.
 Peixoto JL. A property of well-formulated polynomial
regression models. The Am Stat 1990;44(1):26–30.
 Griepentrog GL, Ryan JM, Smith LD. Linear transforma-
tions of polynomial regression models. The Am Stat, Part
C. Korn et al. / Water Research 36 (2002) 330–342341
 McKenzie J, Schaefer RL, Farber E. The student edition
of MINITABsfor windows users manual. Reading, MA:
 Greiner AD,Obolensky
note: comparing predicted and observed concentra-
tions of DBPs. J Am Water Works Assoc 1992;
A, Singer PC. Technical
 Aieta EM, Roberts PV, Hernandez M. Determination of
chlorine dioxide, chlorine, chlorite, and chlorate in water. J
Am Water Works Assoc 1984;76(1):64–70.
 Myers RH. Classical and modern regression with applica-
tions, 2nd ed. In: The Duxbury advanced series in statistics
and decision sciences. Boston, MA: PWS-KENT 1990,
C. Korn et al. / Water Research 36 (2002) 330–342342