Chemosphere 242 (2020) 125227

Contents lists available at ScienceDirect

Chemosphere
journal homepage: www.elsevier.com/locate/chemosphere

Formation and interdependence of disinfection byproducts during

chlorination of natural organic matter in a conventional drinking
water treatment plant
Xiaoxiao Zhang a, b, Zhonglin Chen a, Jimin Shen a, Shengxin Zhao a, Jing Kang a, *,
Wei Chu b, **, Yaoyu Zhou c, Binyuan Wang a
a
State Key Laboratory of Urban Water Resource and Environment, School of Environment, Harbin Institute of Technology, Harbin, 150090, PR China
b
Department of Civil and Environmental Engineering, The Hong Kong Polytechnic University, Hung Hom, Kowloon, Hong Kong
c
College of Resources and Environment, Hunan Agricultural University, Changsha, 410128, China

h i g h l i g h t s g r a p h i c a l a b s t r a c t

A hyperbola model was applied to

describe the TCM, DCAA and TCAA
formation.

TCM exhibited the highest initial
formation rate and maximum theo-
retical yield.

The TCM yield can serve as a good
indicator for the yields of measured
DBPs.

Specific unregulated DBP yields
increased after a filtration process.

Soluble microbial by-product-like
compounds increased the specific
DCAN yield.

a r t i c l e i n f o a b s t r a c t

Article history: It is crucial to explore the source, formation process and interdependence of disinfection byproducts
Received 10 July 2019 (DBPs) to reduce their risk on public health. In this investigation, a source water was chlorinated to
Received in revised form evaluate the initial formation rates and the maximum yields of trichloromethane (TCM), dichloroacetic
23 October 2019
acid (DCAA), and trichloroacetic acid (TCAA) based on a hyperbola model. The results showed that TCM
Accepted 24 October 2019
Available online xxx
achieved the highest initial formation rate and maximum theoretical concentration compared with
DCAA and TCAA. The TCM yield can be used to forecast the yields of DCAA and TCAA throughout the
Handling Editor: Xiangru Zhang whole reaction process, and the yields of chloral hydrate (CH), dichloroacetonitrile (DCAN) and 1,1,1-
trichloropropanone (1,1,1-TCP) within the initial reaction stage. Besides, the raw water, settled water
Keywords: and filtered water collected from a drinking water treatment plant were divided into five fractions,
Disinfection byproducts respectively, by ultrafiltration membranes to evaluate their DBP formation after chlorination. Compared
Hyperbola model with the medium molecular weight species, high and low molecular weight organic matters exhibited
Natural organic matter relatively high specific regulated and unregulated DBP yields (expressed as mg/mg C), respectively. Humic
Molecular weight distribution
acid-like compositions predominantly contributed to regulated DBP yields, while soluble microbial by-
Toxicity
product-like compounds preferentially generated DCAN. The correlation study revealed that the TCM
could also serve as an indicator for the measured DBPs from chlorination of sample fractions with

* Corresponding author.
** Corresponding author.
E-mail addresses: jingkanghit@163.com (J. Kang), wei.chu@polyu.edu.hk
(W. Chu).

https://doi.org/10.1016/j.chemosphere.2019.125227
0045-6535/© 2019 Elsevier Ltd. All rights reserved.
2 X. Zhang et al. / Chemosphere 242 (2020) 125227
different molecular weight. Finally, it was found that the theoretical cytotoxicity was enhanced during
chlorination of filtered water compared with chlorination of settled water.
1. Introduction
To ensure the safety of the public water supply is of great
importance (Y. D. Xu et al., 2019). Chlorine disinfection is commonly
used to kill pathogens in drinking water. However, the interreaction
of disinfectants with ubiquitous nature organic matter (NOM) in
water leads to the formation of disinfection byproducts (DBPs) (Liu
et al., 2019; Yang et al., 2019b), which are genotoxic, cytotoxic and
developmentally toxic (Han and Zhang, 2018; Li and Mitch, 2018).
Trihalomethanes (THMs) and haloacetic acids (HAAs) are most
common among DBPs, and specialized standards have been made
to evaluate their concentrations. Haloacetaldehydes (HALs) are the
next prevalent class of DBPs (Krasner et al., 2006). The World
Health Organization has listed a guideline concentration for the
chloral hydrate (CH), which is the representative of HALs. Besides,
unregulated DBPs, e.g. nitrogenous DBPs, were found to show
significantly higher toxicity than carbonaceous THMs, HAAs and
HALs (Plewa et al., 2008). The unregulated DBPs, including hal-
oketones (HKs), haloacetonitriles (HANs) and halonitroalkanes,
were also frequently detected in drinking water distribution sys-
tems (Huang et al., 2017).
Various water treatment technologies were exploited to control
the DBP formation by removing corresponding precursors, such as
the coagulation, sedimentation, sand filtration, ultrafiltration (D. Y.
Xu et al., 2019), granular activated carbon (GAC) adsorption (Jiang
et al., 2017), etc. The humic-like macromolecules were preferen-
tially removed by the conventional coagulation, sedimentation and
sand filtration processes. Organics with medium molecular weight
(MW) were predominantly reduced by the ultrafiltration process. In
contrast, the GAC adsorption was usually used in the advanced
water treatment process to remove DBP precursors with the low
MW.
To better control DBP formation, it is necessary to decipher their
precursors through specific characterization technologies (Mao
et al., 2016). The ultraviolet (UV) absorption spectrometry was
usually applied for the characterization of DBP precursors. For
example, the differential UV absorbances at 272 nm were found to
correlate well with many DBP yields (Beauchamp et al., 2018).
Compared with the UV absorption spectrometry, the fluorescence
excitationemission matrices (EEMs) provided more comprehen-
sive information of DBP precursors (Vera et al., 2017). The data
processing techniques of EEMs, such as, the fluorescence indexes
(Roccaro et al., 2011), fluorescence regional integration (FRI) tech-
nique (F. Trueman et al., 2016), parallel factor analysis (Watson
et al., 2018) and fluorescence relationship-Neural networks
(Peleato et al., 2018) were widely utilized to decode DBP precursors
(Cai et al., 2017, 2016; Ma et al., 2018). Among these data processing
approaches, the FRI technique divides the fluorescence EEM spectra
of NOM into five EEM regions: two aromatic protein regions, humic
acid-like region, fulvic acid-like region and soluble microbial by-
product-like region (Chen et al., 2003). Upon this division, the
characteristics of NOM can be comprehensively demonstrated. To
data, information regarding to the relationship between fluores-
cence data via FRI analysis and DBP yields is still limited.
Furthermore, it should be noted that most studies about the
evaluation of the control effect on DBP formation in practice
focused on the regulated THMs and HAAs (Han et al., 2015).
© 2019 Elsevier Ltd. All rights reserved.
Compared with HAAs, THMs are easier to be detected (Yang et al.,
2019a). No pretreatment is needed before analyzing THMs by the
head space gas chromatography according to the standard method
recommended by the Ministry of Health of the People’s Republic of
China. In comparison, the derivatization reaction should be done
prior to the analysis of HAAs. A previous study revealed a certain
degree of correlation between THMs and HAAs (Roccaro et al.,
2014). It suggested that the easily detected THMs might be
considered as a substitute for the hardly detected HAAs. Besides,
relationships were also found between THMs and the unregulated
DBPs (Roccaro et al., 2014; Smith et al., 2015). Recently, correlations
of THMs with other detected DBPs and even with the toxicity of
disinfected waters aroused great concern (Li and Mitch, 2018).
This study aimed at 1) exploring the initial formation rates and
maximum theoretical concentrations of regulated DBPs, 2)
revealing the relationships of trichloromethane (TCM) with
dichloroacetic acid (DCAA), trichloroacetic acid (TCAA), CH,
dichloroacetonitrile (DCAN), or 1,1,1-trichloropropanone (1,1,1-
TCP), 3) evaluating the variations of DBP precursors along the wa-
ter treatment processes in a drinking water treatment plant.
2. Materials and methods
2.1. Collection and pretreatment of sample
Three water samples: S1: raw water (RW), S2: settled water
(SW) and S3: filtered water (FW) were collected from a drinking
water treatment plant in northeast China. In the drinking water
treatment plant, the polyaluminium chloride (PAC) was selected as
the coagulant, and a sand filtration process was conducted. The
process flow chart of the drinking water treatment plant and
sampling spots were shown in Fig. S1.
All samples were collected on 2017 March 15 and transferred to
the laboratory within 2 h. After arriving at the laboratory, glass fiber
membranes (GF/F, 0.7 mm, Whatman, Clifton, NJ) were used to
remove particulate matters in water samples. Each water sample
was filtered through a 0.45 mm syringe filter for the analysis of
dissolved organic carbon (DOC). The DOC concentrations of RW, SW
and FW were 5.6, 2.5 and 2.1 mg/L, respectively. The Br concen-
trations of these effluents were all lower than the detection limit
(0.1 mg/L).
2.2. Chemicals
The standard chemicals of DBPs including TCM, chloroacetic
acid (MCAA), DCAA, TCAA, CH, chloroacetonitrile (MCAN), DCAN,
trichloroacetonitrile (TCAN), 1,1,1-TCP and 1,3-dichloropropanone
(1,3-DCP) were purchased from Sigma-Aldrich. Water used during
the experiments was prepared by the Milli-Q ultrapure water sys-
tem. Chlorine stock solution was obtained by diluting sodium hy-
pochlorite (Sigma-Aldrich, 10e15% available free chlorine).
2.3. Fractionation of sample
The ultrafiltration process was conducted by using a stirred ul-
trafiltration cell (Millipore, 8400) and a series of ultrafiltration
membranes with MW cutoffs at 1, 3, 5 and 10 kDa (kDa) (Millipore,
 X. Zhang et al. / Chemosphere 242 (2020) 125227
MA, USA). In each ultrafiltration step, 400 mL sample was added in
the ultrafiltration cell and 300 mL permeate was collected for the
further study. Pressure was kept at 0.25 MPa by the nitrogen gas
(99.99%). After ultrafiltration, five sample fractions with different
MW were received as follows: F1: MW < 1 kDa, F2: MW < 3 kDa,
F3: MW < 5 kDa, F4: MW < 10 kDa and F5: particle size <0.45 mm.
2.4. Disinfection
The solution pH of each fraction was adjusted to 7.0 ± 0.2 with
2 mM phosphate buffered solution. Then the chlorination reaction
was conducted with the ratio of chlorine to DOC set at 5:1 (based on
the weight) in a headspace-free reactor at 20  C in the dark. Under
the high chlorine dose and long reaction time, the final concen-
trations of DBPs reliably reflected the DBP formation potential. The
ascorbic acid was used as the terminator of the chlorination reac-
tion (Kristiana et al., 2014). All experiments in this study were
conducted in duplicates, and average values were taken for
analysis.
2.5. Instruments and analytical methods
DOC was measured by Jena TOC-Vcsh instrument (Germany).
The pH value was determined by pH 7110 m (WTW, Germany). The
free chlorine concentration was measured through the DPD/FAS
titration method using the UV spectrophotometer at 515 nm. The
UV absorbance was measured by Purkinje General T6 spectropho-
tometer (Beijing, China). Inorganic anions were measured by the
ion chromatography (IC) (Thermo Dionex ICS-2100). DBP yields
were detected by the gas chromatography instrument (Agilent GC
7890A, Santa Clara, USA) equipped with a DB-35 MS capillary col-
umn (30 m  0.25 mm  0.25 mm, Agilent, USA) and an electron
capture detector (ECD). The GC temperature program was set up at
40  C for 5 min, which was increased by 5  C/min to 100  C, then
50  C/min to 200  C and hold for 2 min with a constant nitrogen
carrier gas flow at 1.0 mL/min under a split mode (split ratio 25:1).
To obtain the fluorescence EEM spectra, parameters of the instru-
ment (Jasco, FP-6500, Japan) were set as follows: the excitation
wavelength increased linearly from 220 to 450 nm with 5-nm in-
tervals; for each excitation wavelength, the emission wavelength
was detected from 220 to 500 nm with 1-nm intervals. To reduce
Raleigh and Raman scatterings, the ultrapure water blank was
subtracted. ‘
Fig. 1. Changes of DBP yields with reaction time. (a), time-dependent DBP yields; (b), curve fitting of TCM, DCAA or TCAA yield over time.
3
2.6. Calculation of theoretical cytotoxicity
LC50 was the concentration of each DBP leading to a 50%
decrease in the growth of Chinese Hamster Ovary (CHO) cells after
72 h of exposure, which could be utilized to indicate the toxic po-
tency (Plewa et al., 2008). The LC50 values of measured DBPs were
listed in Table S1. The theoretical cytotoxicity was calculated using
the ratio of the individual DBP molar concentration and its LC50
value (Gao et al., 2019).
3. Results and discussions
3.1. Effect of reaction time on DBP formation
3.1.1. DBP yields based on reaction time profiles
The RW was chlorinated to investigate the changes of DBP yields
with time. As depicted in Fig. 1(a), the TCM yield kept increasing for
the first 120 h, and then leveled off. The generation of DCAA and
TCAA with time also exhibited similar trends as TCM. These results
are consistent with a recent report (Padhi et al., 2019). Notably, the
yields of HAAs from chlorination of RW had the following order:
TCAA > DCAA, which was different from a recent report where di-
HAAs > tri-HAAs (Padhi et al., 2019). Different chlorine dose and
reaction time (3 d vs. 30 min to 24 h) could account for this
difference.
With respect to HANs and HKs, only DCAN and 1,1,1-TCP were
detected, which reached their maximum concentrations at 12 h
and then decayed. Particularly, MCAN, TCAN and 1,3-DCP were not
detected, which may be attributed to the addition of chlorine at a
high dose. The resultant MCAN and 1,3-DCP (if any) could be easily
transformed into di- or trichloro-products and even TCM (Yang
et al., 2007), while TCAN could be rapidly decomposed after its
formation under a high dose of the disinfectant (Yu and Reckhow,
2015). Furthermore, excessive chlorine was found to react with
DCAN through nucleophilic attack on the nitrile carbon, which
produced DCAA or TCAA as the major end product (Yu and
Reckhow, 2015). The decrease of DCAN and 1,1,1-TCP could also
result in the increase of TCM as demonstrated in previous studies
(Yang et al., 2007; Yu and Reckhow, 2015).
To predict the formation of THMs in chlorinated water, a first
order model was developed as a hyperbolic growth function
(Gallard and von Gunten, 2002). This model described the forma-
tion dynamic of THMs, which also gave the theoretical maximum
concentration of THMs. The formation rates of THMs and HAAs in
this study followed a decreasing trend with the reaction time,
4 X. Zhang et al. / Chemosphere 242 (2020) 125227

which complied with the hyperbola model well. Therefore, the
hyperbola model was applied to describe the TCM, DCAA and TCAA
formation as shown in Eq. (1).
t were well correlated with the TCM yield within the whole reaction
CðtÞ ¼
ða þ b  tÞ
where C(t) was the concentration of TCM, DCAA or TCAA at time t
(mg/L); a and b were two characteristic constants. The Eq. (1) can
also be linearized by plotting t/C(t) versus t, with the intercept and
slope as a and b, respectively. As t was set as zero, by taking the
derivation of Eq. (1), the initial formation rate could be calculated
as:

(Jeong et al., 2015). In China, 10 mg/L was set as the maximum
dCt¼0
¼1 a
dt
where the physical meaning of 1/a (mg/(L,h)) was the initial for-
mation rate of DBPs. On the other hand, when t was approaching
infinity, the theoretical maximum concentrations of DBPs can be
obtained, as shown in Eq. (3).
Ct/∞ ¼ 1=b
The modelled curves were shown in Fig. 1(b). Based on the
observation of a and b, the initial formation rates of TCM, TCAA, and
DCAA were 23.1, 18.5, and 9.6 mg/(L,h), respectively; while the
theoretical maxima of TCM, TCAA and DCAA yields were 304.9,
239.2 and 104.8 mg/L, respectively.
3.1.2. Relationships between DBP yields based on reaction time
profiles
To verify whether TCM can be an indicator of other measured
DBPs, the time-dependent linear relationships of DCAA, TCAA or CH
vs. TCM yield were shown in Fig. 2(a). The DCAA and TCAA yields
(1)
process with R2 > 0.85. The slope of the correlation equation be-
tween TCAA and TCM was higher than the slope of that between
DCAA and TCM, suggesting that the generation of TCM in certain
amount was accompanied by more TCAA yield than DCAA. In
addition, the 1,1,1-TCP or DCAN yield displayed good correlations
with the TCM yield within the first 12 h; however, likely due to the
involvement of additional side-reactions, the correlation abruptly
disengaged after 12 h, as demonstrated in Fig. 2(b).
CH is more toxic and carcinogenic than TCM, DCAA and TCAA
contaminant level (MCL) for CH. The relationships of the CH yield
(2)
with the regulated DCAA, TCAA or TCM yield were depicted in the
Fig. 2(c), (d) and (a), respectively. The CH yield was linearly
dependent on the DCAA, TCAA or TCM yield within 48 h
chlorination.
According to the stage 2 of the D/DBP rule in USA (Agency,
2006), 80 and 60 mg/L were set as the MCL for THMs and HAAs,
respectively. In China, total THM level was regulated depending on
the ratio (Karanfil et al., 2008):
(3)
CCHCl3 CCHCl2 Br CCHClBr2 CCHBr3
þ þ þ <1 (4)
60 60 100 100
Based on above equation, MCL for TCM was 60 mg/L. When the
yield of TCM reached to 60 mg/L, CH concentration could be calcu-
lated as 3.9 mg/L according to the regression equation in Fig. 2(a). It

Fig. 2. Relationships between DBP yields. (a), relationships between DCAA, TCAA or CH yield and TCM yield within 168 h; (b), relationships between DCAN or 1,1,1-TCP yield and
TCM yield within 12 h; (c), relationship between CH yield and DCAA yield within 48 h; (d), relationship between CH yield and TCAA yield within 48 h. The n is the free degree of the
regression equation.
 X. Zhang et al. / Chemosphere 242 (2020) 125227
suggested that, in this chlorination environment, meeting the TCM
standard could guarantee CH yield to meet the standard of China.
Similarly, meeting HAA standards would indicate that CH could also
meet the regulation based on the regression equations in Fig. 2(c)
and (d). Locally, the hydraulic retention time (HRT) after post-
chlorination in practice was generally longer than 30 min but
surely less than 48 h, which was within the time window of the
proposed regression equations, so the models may be practical for
application.
3.2. Effect of NOM MW on DBP formation
3.2.1. Reactivity of NOM toward chlorine in DBP formation
According to the previous discussions, most DBPs reached
relative constant concentrations after 72 h chlorination, and hence
72 h was selected as the chlorination time for the further study.
After 72 h chlorination, the specific DBP yields (expressed as mg/mg
C) for different MW fractions were obtained as shown in Fig. 3. For
RW, F5 showed the highest specific TCM, DCAA and TCAA yields.
After the sedimentation process, as shown in Fig. 3(b), F5 still kept a
high reactivity toward chlorine in DCAA and TCAA formation, while
F4 dominated TCM generation. After the filtration process, F4
demonstrated the highest reactivity with chlorine in TCM, DCAA
and TCAA formation (Fig. 3(c)). Generally, the higher the average
MW, the higher the NOM reactivity toward chlorine in TCM, DCAA
and TCAA formation. Compared with TCM, DCAA and TCAA, the
specific CH yield was relatively independent on the MW of NOM.
Furthermore, for RW and SW, NOM fractions with the lower MW
showed the relatively higher specific DCAN and 1,1,1-TCP yields. For
FW, however, the specific DCAN yield was independent of the MW
of NOM. N-compounds, such as amino acid-like substances, were
the key precursors of DCAN (Ma et al., 2018), which might be
originated from the backwash flow or the aging of the filter media,
Fig. 3. Effect of NOM MW on DBP formation in (a) RW, (b) SW, and (c) FW; (d) the theoretical cytotoxicity per DOC.SDBP corresponds to the specific DBP yields.
5
thus leading to the independence of specific DCAN yield from the
MW of organics in FW. Moreover, the theoretical cytotoxicity (in
terms of LC50) per DOC of organics with varying MW were shown in
Fig. 3(d). For RW and SW, NOM fractions with the lower MW
showed relatively higher theoretical cytotoxicity per DOC during
chlorination. For FW, however, the theoretical cytotoxicity per DOC
was independent of the MW of NOM, which was consistent with
the independence of the specific DCAN yield from the MW of NOM.
It could be ascribed to the domination of the theoretical cytotox-
icity by DCAN.
3.2.2. Relationships between DBP yields from chlorination of NOM
in varying MW
Relationships between DBP yields from chlorination of NOM in
varying MW were investigated as shown in Table 1. The relation-
ship of the TCM yield with the DCAA or TCAA yield was remarkable
(R2 > 0.96), which was consistent with a previous study where the
yields of THMs and HAAs correlated well with each other during
chlorination (Roccaro et al., 2014). For the unregulated DCAN or
1,1,1-TCP, significant correlations could also be respectively estab-
lished with the TCM yield, with R2 around 0.88. Additionally, an
extremely strong correlation was found between the TCM and CH
yield with R2 ¼ 0.99. It confirmed that the TCM could serve as an
indicator for the measured DBPs in chlorination of NOM with
different MW.
3.3. Variations of NOM along water treatment processes
3.3.1. Changes of average MW
The distribution of DOC according to the MW interval was
shown in Fig. 4(a). For RW, NOM segments with MW in 0e1, 1e3,
3e5, 5e10 and >10 kDa accounted for 22.5%, 16.7%, 8.0%, 39.0%, and
13.8% of total DOC, respectively. After coagulation and
6 X. Zhang et al. / Chemosphere 242 (2020) 125227

Table 1
R2 (n ¼ 15, p < 0.1) of linear regression equations between the DBP yields during chlorination of different NOM fractions.

Fig. 4. Changes of NOM along water treatment processes. (a), DOC distributions based on MW intervals; (b), changes of the EEM volume; (c), (d) and (e), fluorescence EEM spectra of
RW, SW, and FW; (f), (g) and (h), variations of organic components in RW, SW, and FW, respectively. I: aromatic protein I region, II: aromatic protein II region, III: fulvic acid-like
region, IV: soluble microbial by-product-like region, V: humic acid-like region.

sedimentation processes, 55.0% of DOC in RW was removed, in
which the NOM segment with the MW > 5 kDa (especially MW in
5e10 kDa) was significantly removed. The decrease of the NOM
MW after the coagulation and sedimentation processes may be
mainly attributed to the absorption and bridge effect induced by
the flocculants (Han et al., 2015). Gradually, the proportion of NOM
with MW < 5 kDa raised from 45.7% of RW to 50.6% of SW, and
continuously climbed up to 72.5% of FW.
3.3.2. Changes of organic compositions
The FRI analysis was developed to distinguish NOM into five
components (Chen et al., 2003). The percentage of normalized
excitationemission area volumes (Fi,n) represented the relative
content of each component (Chen et al., 2003). As depicted in
Fig. 4(b), FV,n took the highest percentage (56.0%) of the cumulative
EEM volume for RW. After the sedimentation process, fulvic and
humic acid-like compositions (FIII,n and FV,n) were significantly
reduced. Unexpectedly, the subsequent filtration process increased
the soluble microbial by-product-like compositions (FVI,n), which
might account for the increase of specific DCAN and 1,1,1-TCP yields
(Goslan et al., 2017; Zheng et al., 2018).
To visually depict the variations of NOM components, the fluo-
rescence EEM spectra of RW, SW and FW were shown in
Fig. 4(c)e(e). Notably, the removal of humic acid-like components
 X. Zhang et al. / Chemosphere 242 (2020) 125227
after the sedimentation process was mainly ascribed to the
reduction of compounds with fluorescence response under UVC
(220e275 nm) excitation. There were bad removal efficiencies for
compositions with fluorescence responses under UVB/UVA exci-
tation, which might be one of the reasons leading to the high
specific DCAN and 1,1,1-TCP yields during chlorination of SW.
Furthermore, the MW distributions of different organic com-
positions were shown in Fig. 4(f)e(h). Compared with RW, the
reduction of humic acid-like compositions (Fv, n) in SW was pre-
dominantly present in the >10 kDa interval. After the filtration
process, the unexpected increase of soluble microbial by-product-
like compounds (FIV,n) were also present in the >10 kDa interval,
which might account for the independence of the specific DCAN
yield from the MW of organics in FW.
3.4. Relationships between optical properties and DBP yields
Recently, optical parameters were widely applied to predict DBP
formation in both bench and full scales (Beauchamp et al., 2018;
Peleato et al., 2018). In this part, the relationships between DBP
yields and the readable fluorescence responses in region “i” (Ri ¼
PP
Iðlex lem Þ, I(lexlem) refers to the fluorescence response at lex
ex em
and lem) were investigated for 15 NOM fractions. As shown in
Table 2, for carbonaceous DBPs, including TCM, DCAA, TCAA, CH
and 1,1,1-TCP, R2 reached its maxima at RV, implying that humic
acid-like compositions might be their key precursors. For DCAN, its
yield correlated best with RIV, which supported the previous
assumption that the augment of soluble microbial by-product-like
compounds in >10 kDa segment accelerated the DCAN generation.
The slope of the regression equation was the unit increment of
Table 2
Relationships between UV254, Ri and the DBP yield during chlorination of 15 NOM fractions. The R2 (n ¼ 15, p < 0.1), slope and intercept are obtained based on the linear
regression equation between the value of each optical parameter and the DBP yield.
7
DBPs per Ri or UV254. The reciprocal of the slope represented the
sensitivity of the characteristic method, i.e. the slope had negative
relationship with the sensitivity to characterize the DBP yields.
Compared to UV254, the slope of Ri was decreased by 4 logs.
Therefore, the sensitivity was greatly improved by using Ri. As for
intercept of the regression equation, it represented the range of the
characteristic method for evaluating DBP yields and the least value
of DBP yields above which the DBP yields could be characterized by
the Ri or UV254. With respect to Ri, the intercepts of TCM, DCAA,
TCAA were all negative and the intercepts of CH were relatively
small. It indicated that TCM, DCAA, TCAA and CH could be char-
acterized in a wide range by Ri.
Combined with the small slope, the Ri showed a significant
advantage in evaluating the TCM, DCAA, TCAA and CH yields over
the use of UV254. For DCAN and 1,1,1-TCP, all the intercepts using Ri
and UV254 were high. It suggested the DCAN and 1,1,1-TCP could
only be detected and characterized in a certain degree. Therefore,
the Ri (especially RV) can serve as an effective and sensitive eval-
uation for TCM, DCAA, TCAA and CH yields.
3.5. Potential cytotoxicity assessment
The theoretical cytotoxicity of measured DBPs from chlorination
of RW, SW or FW were depicted in Fig. 5. The measured DBP yields
and the corresponding cytotoxicity from chlorination of different
NOM fractions were demonstrated in Table S2. The results showed
that the cytotoxicity was mainly caused by chlorination of low MW
DBP precursors (e.g. < 1 kDa). The theoretical cytotoxicity was
reduced by 50.3% during chlorination of SW compared with chlo-
rination of RW. The reduction was mainly attributed to the removal
of >5 kDa DBP precursors through the coagulation and
8 X. Zhang et al. / Chemosphere 242 (2020) 125227
Fig. 5. The theoretical cytotoxicity of DBPs.
sedimentation processes. However, an unwanted increase of the
theoretical cytotoxicity during chlorination of the FW was noticed,
which may be due to the augment of soluble microbial by-product-
like components through the filtration process.
4. Conclusions
The reaction time profile of DBP yields from chlorination of the
RW verified that the TCM exhibited the highest initial formation
rate of 23.1 mg/(L,h) and maximum theoretical concentration of
304.9 mg/L, compared with DCAA and TCAA. Depending on the
correlations of TCM with DCAA, TCAA, CH, DCAN or 1,1,1-TCP, it was
possible to select TCM as an indicator for drinking water quality.
Notably, a continuously increasing trend for CH yields occurred
when the chlorination time was prolonged beyond 120 h. There-
fore, the CH concentration in the long-distance drinking water
distribution system deserves further concern. The specific DBP
yields were found to be dependent on the NOM MW, which was
related to the organic compositions. The unexpected increase of
>1 kDa soluble microbial by-product-like components after the
filtration process greatly contributed to the theoretical cytotoxicity
of the measured DBPs during chlorination. Accordingly, the filter
aging and microbial growth in practical water treatment structures
deserve further attention.
Acknowledgements
The work was supported by the National Key R&D Program of
China (Grant No. 2017YFA0207203 and 2017YFA0207204), the State
Key Laboratory of Urban Water Resource and Environment, Harbin
Institute of Technology (Grant No.2017TS02), the National Natural
Science Foundation of China (Grant No. 51608148), the Hong Kong
Polytechnic University (Grant No. GSB0L), and the Hong Kong
Polytechnic University (Grant No. 1-ZVH6).
Appendix A. Supplementary data
Supplementary data to this article can be found online at
https://doi.org/10.1016/j.chemosphere.2019.125227.
References
Agency, U.S.E.P., 2006. National primary drinking water regulations: stage 2 disin-
fectants and disinfection byproducts rule: final rule. Fed. Regist. 71, 388e493.
Beauchamp, N., Laflamme, O., Simard, S., Dorea, C., Pelletier, G., Bouchard, C.,
Rodriguez, M., 2018. Relationships between DBP concentrations and differential
UV absorbance in full-scale conditions. Water Res. 131, 110e121. https://doi.org/
10.1016/j.watres.2017.12.031.
Cai, W., Liu, J., Zhang, X., Ng, W.J., Liu, Y., 2016. Generation of dissolved organic
matter and byproducts from activated sludge during contact with sodium hy-
pochlorite and its implications to on-line chemical cleaning in MBR. Water Res.
104, 44e52. https://doi.org/10.1016/j.watres.2016.07.065.
Cai, W., Liu, J., Zhu, X., Zhang, X., Liu, Y., 2017. Fate of dissolved organic matter and
byproducts generated from on-line chemical cleaning with sodium hypochlo-
rite in MBR. Chem. Eng. J. 323, 233e242. https://doi.org/10.1016/
j.cej.2017.04.099.
Chen, W., Westerhoff, P., Leenheer, J.A., Booksh, K., 2003. Fluorescence excitation-
emission matrix regional integration to quantify spectra for dissolved organic
matter. Environ. Sci. Technol. 37, 5701e5710. https://doi.org/10.1021/es034354c.
Gallard, H., von Gunten, U., 2002. Chlorination of natural organic matter: kinetics of
chlorination and of THM formation. Water Res. 36, 65e74. https://doi.org/
10.1016/S0043-1354(01)00187-7.
Gao, Z.-C., Lin, Y.-L., Xu, B., Xia, Y., Hu, C.-Y., Zhang, T.-Y., Cao, T.-C., Chu, W.-H.,
Gao, N.-Y., 2019. Effect of UV wavelength on humic acid degradation and
disinfection by-product formation during the UV/chlorine process. Water Res.
154, 199e209. https://doi.org/10.1016/j.watres.2019.02.004.
Goslan, E.H., Seigle, C., Purcell, D., Henderson, R., Parsons, S.A., Jefferson, B., Judd, S.J.,
2017. Carbonaceous and nitrogenous disinfection by-product formation from
algal organic matter. Chemosphere 170, 1e9. https://doi.org/10.1016/
j.chemosphere.2016.11.148.
Han, J., Zhang, X., 2018. Evaluating the comparative toxicity of DBP mixtures from
different disinfection scenarios: a new approach by combining freeze-drying or
rotoevaporation with a marine polychaete bioassay. Environ. Sci. Technol. 52,
10552e10561. https://doi.org/10.1021/acs.est.8b02054.
Han, Q., Yan, H., Zhang, F., Xue, N., Wang, Y., Chu, Y., Gao, B., 2015. Trihalomethanes
(THMs) precursor fractions removal by coagulation and adsorption for bio-
treated municipal wastewater: molecular weight, hydrophobicity/hydrophily
and fluorescence. J. Hazard Mater. 297, 119e126. https://doi.org/10.1016/
j.jhazmat.2015.04.070.
Huang, H., Zhu, H., Gan, W., Chen, X., Yang, X., 2017. Occurrence of nitrogenous and
carbonaceous disinfection byproducts in drinking water distributed in Shenz-
hen, China. Chemosphere 188, 257e264. https://doi.org/10.1016/
j.chemosphere.2017.08.172.
Jeong, C.H., Postigo, C., Richardson, S.D., Simmons, J.E., Kimura, S.Y., Marin ~ as, B.J.,
Barcelo, D., Liang, P., Wagner, E.D., Plewa, M.J., 2015. Occurrence and compar-
ative toxicity of haloacetaldehyde disinfection byproducts in drinking water.
Environ. Sci. Technol. 49, 13749e13759. https://doi.org/10.1021/es506358x.
Jiang, J., Zhang, X., Zhu, X., Li, Y., 2017. Removal of intermediate aromatic haloge-
nated DBPs by activated carbon adsorption: a new approach to controlling
halogenated DBPs in chlorinated drinking water. Environ. Sci. Technol. 51,
3435e3444. https://doi.org/10.1021/acs.est.6b06161.
Karanfil, T., Krasner, S.W., Westerhoff, P., Xie, Y., 2008. Recent advances in disin-
fection by-product formation, occurrence, control, Health effects, and regula-
tions. In: Disinfection By-Products in Drinking Water, ACS Symposium Series.
American Chemical Society, pp. 2e19. https://doi.org/10.1021/bk-2008-
0995.ch001.
Krasner, S.W., Weinberg, H.S., Richardson, S.D., Pastor, S.J., Chinn, R., Sclimenti, M.J.,
Onstad, G.D., Thruston, A.D., 2006. Occurrence of a new generation of disin-
fection byproducts. Environ. Sci. Technol. 40, 7175e7185. https://doi.org/
10.1021/es060353j.
Kristiana, I., Lethorn, A., Joll, C., Heitz, A., 2014. To add or not to add: the use of
quenching agents for the analysis of disinfection by-products in water samples.
Water Res. 59, 90e98. https://doi.org/10.1016/j.watres.2014.04.006.
Li, X.-F., Mitch, W.A., 2018. Drinking water disinfection byproducts (DBPs) and
human Health effects: multidisciplinary challenges and opportunities. Environ.
Sci. Technol. 52, 1681e1689. https://doi.org/10.1021/acs.est.7b05440.
Liu, J., Zhang, X., Li, Y., Li, W., Hang, C., Sharma, V.K., 2019. Phototransformation of
halophenolic disinfection byproducts in receiving seawater: kinetics, products,
and toxicity. Water Res. 150, 68e76. https://doi.org/10.1016/
j.watres.2018.11.059.
Ma, C., Xu, H., Zhang, L., Pei, H., Jin, Y., 2018. Use of fluorescence excitationeemission
matrices coupled with parallel factor analysis to monitor C- and N-DBPs for-
mation in drinking water recovered from cyanobacteria-laden sludge dew-
atering. Sci. Total Environ. 640e641, 609e618. https://doi.org/10.1016/
j.scitotenv.2018.05.369.
Mao, Y., Wang, X., Guo, X., Yang, H., Xie, Y.F., 2016. Characterization of hal-
oacetaldehyde and trihalomethane formation potentials during drinking water
treatment. Chemosphere 159, 378e384. https://doi.org/10.1016/
j.chemosphere.2016.05.088.
Padhi, R.K., Subramanian, S., Satpathy, K.K., 2019. Formation, distribution, and
speciation of DBPs (THMs, HAAs, ClO 
2 , and ClO3 ) during treatment of different
source water with chlorine and chlorine dioxide. Chemosphere 218, 540e550.
https://doi.org/10.1016/j.chemosphere.2018.11.100.
Peleato, N.M., Legge, R.L., Andrews, R.C., 2018. Neural networks for dimensionality
reduction of fluorescence spectra and prediction of drinking water disinfection
by-products. Water Res. 136, 84e94. https://doi.org/10.1016/
j.watres.2018.02.052.
Plewa, M.J., Wagner, E.D., Muellner, M.G., Hsu, K.M., Richardson, S.D., 2008.
Comparative mammalian cell toxicity of N-DBPs and C-DBPs. In: Disinfection
By-Products in Drinking Water: Occurrence, Formation, Health Effects, and
 X. Zhang et al. / Chemosphere 242 (2020) 125227
Control, vol. 995. ACS Symposium Series, pp. 36e50. https://doi.org/10.1021/bk-
2008-0995.ch003. American Chemical Society.
Roccaro, P., Vagliasindi, F.G.A., Korshin, G.V., 2011. Quantifying the formation of
nitrogen-containing disinfection by-products in chlorinated water using
absorbance and fluorescence indexes. Water Sci. Technol. 63, 40e44. https://
doi.org/10.2166/wst.2011.006.
Roccaro, P., Vagliasindi, F.G.A., Korshin, G.V., 2014. Relationships between tri-
halomethanes, haloacetic acids, and haloacetonitriles formed by the chlorina-
tion of raw, treated, and fractionated surface waters. J. Water Supply Res.
Technol. - Aqua 63, 21e30. https://doi.org/10.2166/aqua.2013.043.
Smith, R.B., Bennett, J.E., Rantakokko, P., Martinez, D., Nieuwenhuijsen, M.J.,
Toledano, M.B., 2015. The relationship between MX [3-Chloro-4-(dichlor-
omethyl)-5-hydroxy-2(5H)-furanone], routinely monitored trihalomethanes,
and other characteristics in drinking water in a long-term survey. Environ. Sci.
Technol. 49, 6485e6493. https://doi.org/10.1021/es5062006.
Trueman, B.F., MacIsaac, S.A., Stoddart, A.K., Gagnon, G.A., 2016. Prediction of
disinfection by-product formation in drinking water via fluorescence spec-
troscopy. Environ. Sci. Water Res. Technol. 2, 383e389. https://doi.org/10.1039/
C5EW00285K.
Vera, M., Cruz, S., Boleda, M.R., Mesa, J., Martín-Alonso, J., Casas, S., Gibert, O.,
Cortina, J.L., 2017. Fluorescence spectroscopy and parallel factor analysis as a
dissolved organic monitoring tool to assess treatment performance in drinking
water trains. Sci. Total Environ. 584 (585), 1212e1220. https://doi.org/10.1016/
j.scitotenv.2017.01.184.
Watson, K., Farre , M.J., Leusch, F.D.L., Knight, N., 2018. Using fluorescence-parallel
factor analysis for assessing disinfection by-product formation and natural
organic matter removal efficiency in secondary treated synthetic drinking
waters. Sci. Total Environ. 640e641, 31e40. https://doi.org/10.1016/
9
j.scitotenv.2018.05.280.
Xu, D., Bai, L., Tang, X., Niu, D., Luo, X., Zhu, X., Li, G., Liang, H., 2019. A comparison
study of sand filtration and ultrafiltration in drinking water treatment: removal
of organic foulants and disinfection by-product formation. Sci. Total Environ.
691, 322e331. https://doi.org/10.1016/j.scitotenv.2019.07.071.
Xu, Y., Ma, J., Liu, D., Xu, H., Cui, F., Wang, W., 2019. Origami system for efficient solar
driven distillation in emergency water supply. Chem. Eng. J. 356, 869e876.
https://doi.org/10.1016/j.cej.2018.09.070.
Yang, X., Shang, C., Westerhoff, P., 2007. Factors affecting formation of hal-
oacetonitriles, haloketones, chloropicrin and cyanogen halides during chlor-
amination. Water Res. 41, 1193e1200. https://doi.org/10.1016/
j.watres.2006.12.004.
Yang, M., Liberatore, H.K., Zhang, X., 2019a. Current methods for analyzing drinking
water disinfection byproducts. Drinking water contaminants and health effects
Beyond Buy and Dry: the Aftermath of Cowley County, Colorado Curr. Opin.
Environ. Sci. Health 7, 98e107. https://doi.org/10.1016/j.coesh.2018.12.006.
Yang, M., Zhang, X., Liang, Q., Yang, B., 2019b. Application of (LC/)MS/MS precursor
ion scan for evaluating the occurrence, formation and control of polar haloge-
nated DBPs in disinfected waters: a review. Water Res. 158, 322e337. https://
doi.org/10.1016/j.watres.2019.04.033.
Yu, Y., Reckhow, D.A., 2015. Kinetic analysis of haloacetonitrile stability in drinking
waters. Environ. Sci. Technol. 49, 11028e11036. https://doi.org/10.1021/
acs.est.5b02772.
Zheng, J., Lin, T., Chen, W., Tao, H., Tan, Y., Ma, B., 2018. Removal of precursors of
typical nitrogenous disinfection byproducts in ozonation integrated with bio-
logical activated carbon (O3/BAC). Chemosphere 209, 68e77. https://doi.org/
10.1016/j.chemosphere.2018.06.018.