Performance of an integrated process combining ozonation with ceramic

membrane ultra-ltration for advanced treatment of drinking water

Xiaojiang Fan
a
, Yi Tao
a
, Lingyun Wang
a
, Xihui Zhang
a,
, Ying Lei
a
, Zhuo Wang
a
, Hiroshi Noguchi
b
a
Research Center for Environmental Engineering & Management, Graduate School at Shenzhen, Tsinghua University, Shenzhen 518055, China
b
Meidensha Corporation, Meiko Building, 5-5-5 Osaki, Shinagawa-ku, Tokyo 141-8616, Japan
H I G H L I G H T S
Integrated process includes coagulation,
ozonation, ceramic UF and GAC ltration.
Full-performance evaluations of the pilot-
scale integrated process of 120 m
3
/d
Turbidity, DOC, ammonia, geosmin,
2-MIB, EDCs and PPCPs are efciently
removed.
UF membrane fouling can be controlled
in situ by ozonation.
G R A P H I C A L A B S T R A C T
The integration of UF with ozonation (i.e. UF/ozonation) in this paper merged 5 units (including pre-oxidation,
sedimentation, ltration, main oxidation and UF) into one unit. Such a hybrid treatment train is much shorter
and suitable for the upgrade of conventional water treatment plants in dealing with micro-polluted water
sources for better supply of drinking water.
a b s t r a c t a r t i c l e i n f o
Article history:
Received 18 September 2013
Received in revised form 13 December 2013
Accepted 14 December 2013
Available online 9 January 2014
Keywords:
Ceramic membrane
Ozonation
Drinking water
Nano-reactor
An integrated process including coagulation, ozonation, ceramic membrane ultra-ltration and activated carbon
ltration was investigated for the treatment of drinking water from the micro-polluted surface water in
the southern China. A pilot-scale plant with the capacity of 120 m
3
/d was set up and operated. Submerged
at-sheet ceramic membranes were selected with the average pore diameter of 60 nm and the ltration area
of 50 m
2
. Quite a number of water quality parameters were evaluated including turbidity, particle counts,
coliformbacteria, Cryptosporidium and Giardia cysts, dissolved organic carbon (DOC), ammonia, geosmin,
2-methylisoborneol (2-MIB), 4 trihalomethanes, 6 haloacetic acids, 8 endocrine disrupting compounds
(EDCs) and 14 pharmaceutical and personal care products (PPCPs). The experimental results showed
that the removal efciencies of all the evaluated parameters ranged from 64% to 100%. The ozonation in mem-
brane tank is very helpful both inenhancing the performance of removing multiple contaminants and incontrol-
ling the membrane fouling withozone dosage of 25 mg/L at the permeate ux of 100 L/m
2
h. The innovation of
the integrated process was discussed based on the concept of nano-reactor to probe into the reaction activities
inside the membrane pores. However, the detailed mechanisms need to be explored in the future.
Crown Copyright 2013 Published by Elsevier B.V. All rights reserved.
1. Introduction
In some fast-developing and densely-populated areas in the south of
China, many drinking water sources have been polluted with a variety
of contaminants [1,2]. Natural organic matters (NOM) and ammonia
are the main ones that cause many water quality problems. It was also
Desalination 335 (2014) 4754
Corresponding author. Tel.: +86 755 2603 6707.
E-mail address: zhangxh@sz.tsinghua.edu.cn (X. Zhang).
0011-9164/$ see front matter. Crown Copyright 2013 Published by Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.desal.2013.12.014
Contents lists available at ScienceDirect
Desalination
j our nal homepage: www. el sevi er . com/ l ocat e/ desal
found that geosmin and 2-mehtylisoborneol (2-MIB) are typical com-
pounds that cause taste and odor problems [3]. With the help of ad-
vanced analytical instruments, many endocrine disrupting compounds
(EDCs) and pharmaceuticals and personal care products (PPCPs) have
also been detected at the level of ng/L in drinking water [4,5]. In a pre-
vious research, fteen PPCPs were identied with the concentration of
036 ng/L in the source water and twelve PPCPs in the product water
in a southern city of China [5]. The inuence of both EDCs and PPCPs
on water safety has attracted wide attention in China. Thus, the removal
of PPCPs and EDCs along with regular contaminants should be taken
into account in the treatment of drinking water.
The current technologies available for drinking water include some
conventional methods, for example, coagulation, sedimentation, sand
ltration and chlorination; and some advanced processes of ozonation
and granular activated carbon. In addition, per-oxidation with oxidants,
like ozone, chlorine or permanganate, and pretreatment with powdered
activated carbon (PAC) were frequently used to control algal growth,
color, taste or odor. In recent years, the whole treatment train will be
too long and too costly for the complete removal of all the pollutants
in the micro-polluted surface water. On the contrary, low pressure
membrane ultra-ltration (UF) might provide a promising way to
make the treatment train shorter by integrating multiple treatment
units, for example, only one membrane unit can replace sedimentation
and sand ltration operations. Membrane ltration is effective in re-
moving particles, microorganisms and some organic matters with rela-
tively smaller footprint and lower cost [6,7]. Therefore, the application
of membrane ltration to drinking water treatment has developed
quite fast in recent years [7]. However, membrane fouling is still one
of the major challenges in membrane application, for it leads to the in-
crease of operational cost with the decrease of specic permeate ux.
Furthermore, chemical oxidation is usually used to alleviate membrane
fouling [810]. But it may shorten the life of polymeric membranes, for
it is weak in oxidation resistance.
Ceramic membranes are resistant to chemical oxidationsothey have
longer life time than polymeric membrane [11]. In this paper, ceramic
membranes with ozonation process are set up in one tank to replace
regular sedimentation, sand ltration, pre-ozonation and major ozona-
tion. In this way, the whole treatment train would be much shorter. It is
reported that the hybrid process combining ozonation with ceramic
membranes is helpful to control membrane fouling [1215]. When the
ceramic membrane surface was coated with manganese, iron or titani-
um oxides, membrane fouling was alleviated more effectively [16,17].
Such a hybrid process has been demonstrated to be viable for dispersive
supply of drinking water under tropic conditions [18]. However, most of
the previous studies were focused on the alleviation of membrane
fouling and the removal of regular organic matters, and very less atten-
tion was paid to the removal of contaminants like ammonia, EDCs and
PPCPs. The evaluation of the whole treatment train for drinking water
treatment has not been found in any literatures.
The objective of this study is to develop a highly integrated process
including coagulation, ozonation, ceramic UF and GAC ltration to deal
with micro-polluted source water. A pilot scale plant with the capacity
of 120 m
3
/d was set up and was operated like a normal water treatment
plant in Dongguan City in the southern China. Full removal perfor-
mances were evaluated, such as turbidity particles, bacteria, organic
matters, DBPFPs, ammonia, geosmin, 2-MIB, EDCs, and PPCPs.
2. Materials and methods
2.1. Raw water
Raw water was collected from the Dongjiang River, an important
drinking water source in the southern China, and a city canal in
Guangdong province of SouthChina, whichis highly pollutedby domes-
tic sewage and industrial efuents. During rainy seasons, some water in
the city canal overows into the Dongjiang River. The raw water was
taken directly by two submerged pumps to simulate the seasonal
pollution; one was put in the Dongjiang River, while the other in the
city canal; then mixed in a tank for the experiments. Two xed sieves
with a mesh width of 2 mm were used before the submerged pumps
respectively to prevent small impurities from entering the pipes of the
treatment equipment. Table 1 shows the composition of the raw
water used in the test.
2.2. Membrane modules
Flat-sheet ceramic membranes with multi-channels (Meidensha,
Japan) were used in this study. They are made of Al
2
O
3
with a
mean pore size of 60 nm. A single membrane sheet is 1046 mm
(length) 260 mm (width) 6 mm (thickness). One membrane
module contains 50 membrane sheets with the total ltration area of
25 m
2
. Two membrane modules are selected for this study.
2.3. Experimental setup and system operation
Fig. 1 shows the schematic of the pilot scale treatment process in this
study. The process contains coagulation, ozonation, ceramic UF, GAC
ltration as well as disinfection. Coagulation is used as pretreatment
for the destabilization of colloidal particles, for it is necessary to make
the small particles form larger ocs in order to prevent them from
blocking the membrane pores. Then, ceramic UF is used to remove
ocs and microorganisms in the water. Ozonation is set up with ceramic
UF in one reactor so that ozonation can take place not only on the mem-
brane surface but also inside the membrane pores, to oxidize the organ-
ic contaminants into smaller molecules, and to detach the adsorbed
foulants. Meanwhile, granular activated carbon ltration can further
remove the residual organic matters, ammonia, EDCs, PPCPs, etc.
The water sample from the Dongjiang River and the city canal was
mixed in a storage tank of 7.5 m
3
by an agitator, then, pumped into
the coagulation tank with the prepared polymeric aluminum chloride
as coagulant; the dosage was 2.4 mg/L, and the hydraulic retention
time was 12 min. After the coagulated water was fed into the membrane
reactor for ultra-ltration with a constant suction-ux of 100 L/m
2
h,
ozone was aerated into the membrane reactor via a diffuse disc at the
bottom of the membrane module. The permeate entered the following
activated carbon lter (0.8 0.8 m in cross section, 2 m in depth) with
a ltration velocity of 10 m/h. The granular activated carbon was selected
with the diameter ranged from 0.6 mmto 2.0 mm; the mean equivalent
diameter was 1.19 mm, iodine sorption value was 950 mg/g. Finally, the
efuent of GAC ltration was disinfected with sodium hypochlorite. The
hydraulic retention time of coagulation, ozonation, membrane and GAC
ltration is 48 min, while 120 min for disinfection with an available
chlorine dosage of 1.5 mg/L.
During the experiments, the membranes were back ushed automat-
ically for 3 min at intervals of 240 min with a ush ux of 15 m
3
/h, in
order to remove the cake layer on the membrane surface. The sludge
water was discharged periodically after back ushing.
Pure oxygen gas was vaporized from a liquid cylinder, and then fed
into the ozone generator (OZONIA CFS-3 2G). The ozone concentration
was modiedby varying the voltage applied to the ozone generator. The
excess ozone was vented through an ozone destruction unit (OZONIA
ODT-003).
The electricity requirement for the pilot plant was 0.51 kWh/m
3
including raw water supplying, ultra-ltration, stirring, dosing, ozone
generation and destruction of the residual ozone. The cost on chemicals
was less than 0.16 RMB/m
3
including coagulants, disinfectants and pure
oxygen.
2.4. Analytical methods
The ow rate of raw water and permeate as well as the cross-
membrane pressure were monitored continuously with electronic
48 X. Fan et al. / Desalination 335 (2014) 4754
on-line owmeters or pressure gauges respectively. The gaseous ozone
concentration was measured by an Ozone Analyzer BMT 964 BT
(Germany). DOC was analyzed by TOC-VCPH (Shimadzu, Japan). Parti-
cle counts were detected by a GR-1000A desktop laser particle counter
(IBR, America).
Water samples were taken fromrawwater and the efuents of each
treatment units of coagulation, UF, GAC ltration and disinfection. All
the samples were stored at 4 C, they were pre-treated in 24 h before
further analyses.
Geosmin and 2-MIB were detected using the method of head-
space solid phase micro-extraction, gas chromatography and mass
spectrometry (HSPME-GC/MS) as reported in literatures [19]. Head-
space extractions were carried out with 25 mL water sample in a
40 mL vial, screw-capped, straight-sided via a pre-pierced PTFE-faced
silicone rubber septum. EDCs were measured by solid phase extraction,
gas chromatography and mass spectrometry (SPE-GC/MS) according to
Wang [4]. Eight typical EDCs in water samples, that is, bisphenol A
(BPA), nonyl phenol (NP), 4-tert-octylphenol (OP), estrone (E1), estra-
diol (E2), 17a-estradiol (17a-E2), estriol (E3) and 17a-ethinylestradiol
(EE2), were simultaneously determined. PPCPs were measured by
solid phase extraction, high performance liquid chromatography and
mass spectrometry/mass spectrometry (SPE-HPLC/MS/MS) as reported
[5]. Fourteen PPCPs were detected in raw water, including eight antibi-
otics (erythromycin, lincomycin, roxithromycin, sulfamethoxazole,
sulfamerazine, sulfapyridine, griseofulvin, trimethoprim), one anti-
inammatory drug (indomethacin), and the others (benzabrate,
diethyltoluamide (DEET), metoprolol, sulpride, 2-quinoxalinecarboxylic
acid (2-QCA)). Water samples with a volume of 2.5 L were ltrated
through microglass ber membranes (GF/B, Whatman, USA) before
solid phase extraction in the measurement of EDCs and PPCPs
respectively.
The water samples for quantication of DBP formation potential
(DBPFP) were chlorinated for 120 h at 25 2 C at pH 7.0 0.2; the
chlorine concentration was dosed to ensure the residual chlorine
concentration within the range of 35 mg/L. Four trihalomethanes
(THMs), including chloroform (CHCl
3
), bromodichloromethane
(CHBrCl
2
), dibromochloromethane (CHBr
2
Cl), and bromoform
(CHBr
3
), were extracted from the water samples with methyl tert-
butyl ether (MTBE), and then analyzed by gas chromatography (Agilent
GC 6790N) with a micro-electron capture detector (USEPA 551.1).
Six haloacetic acids (HAAs), including monochloroacetic acid
(MCAA), monobromoacetic acid (MBAA), dichloroacetic acid (DCAA),
bromochloroacetic acid (BCAA), trichloroacetic acid (TCAA) and
dibromoacetic acid (DBAA), were determined by liquid chromatogra-
phy and double sets of mass spectrometry (LC/MS/MS). The chromato-
graphic separation of the components was performed on a Shim-pak
XR-ODS column (3 mm 75 mm, 2 m) by applying methanol and
Milli-Q water (containing 0.1% formic acid) as the mobile phase with
Table 1
Characteristics of raw water and product water during test.
Parameters Units Raw water Product water Standards for drinking water in China
pH 6.67.1 6.47.0 6.58.5
Temperature C 14.318.1 14.117.7 Not dened
Conductivity S/cm 241269 239274 Not dened
Turbidity NTU 1879.9 b0.2 1
Ammonia mg/L 2.434.15 0.010.65 0.5
Nitrite mg/L 0.0910.324 00.043 1.00
Fe mg/L 0.40.8 b0.05 0.3
Mn mg/L 0.040.06 b0.01 0.1
Br

mg/L b0.02 b0.01 Not dened

BrO
3

mg/L b0.005 b0.005 0.01

Cl

mg/L 2030 2030 250

SO
4
2
mg/L 2024 2023 250
Total hardness mg/L 4552 4452 450
TOC mg/L 2.23.4 b1 Not dened
COD
Mn
mg/L 1.743.97 b0.5 3
UV
254
1/m 3.84.6 b1 Not dened
Coliform group MPN/100 mL N16,000 None
a
None
Standard plate-count bacteria CFU/mL N860 b30
a
100
a
After disinfection with NaClO.
Fig. 1. Schematic representation of the pilot-scale hybrid process for drinking water treatment.
49 X. Fan et al. / Desalination 335 (2014) 4754
gradient elution mode. Electrospray ionization was operated in the
negative multiple reaction monitoring (MRM) mode.
Multiple-tube fermentation technique was used in the Coliform
group detection. 1 mL water sample (or an appropriate dilution) was
added into the tube with 9 mL of lactose broth (SCRC, China); then,
the tubes were incubated at 36 1 C for 24 2 h. Quantication of
the Coliform group was carried out by most probable number (MPN).
Plate count was performed in triplicate by the spread plate technique
by using 1 mL of the water sample spreading over the surface of stan-
dard plat-count agar. Colony-forming units (CFUs) were counted after
culturing for 48 h at 36 1 C.
Cryptosporidiumand Giardia cysts were detected through concentra-
tion, immunomagnetic separation (IMS), and immunouorescence
assay (FA) microscopy (USEPA1623, 2005). 20 L raw water or 40 L
membrane permeate was collected by Envirochek lter capsule, and
then it was eluted and concentrated. Dynabeads Giardia-Combo and
Dynabeads Crypto-Combo were used in the IMS. Direct orescent
antibody (DFA) staining, with EasyStain (BTF-bio), was performed
according to the manufacturer's instructions. Cysts were enumerated
at 2001000 magnication with a biological microscope BX51W
(Olympus, Japan), with4,6-diamidino-2-phenylindole (DAPI) anduo-
rescein differential interference contrast (DIC) lters.
All the water samples were analyzed in duplicate. Quality control
was conducted during the analyses, and the parameters for quality con-
trol are listed in Table 2. UV
254
, DOC, ammonia, nitrite, nitrate, turbidity
and particle counts were tested twice a day, while Coliformgroup, stan-
dard plate-count bacteria, geosmin, 2-MIB, EDCs, PPCPs and DBPs were
detected every three days. Cryptosporidium and Giardia cysts were
detected twice during the experiments.
3. Results and discussion
3.1. Degradation and separation of water components
3.1.1. Removal of turbidity and particles
Fig. 2 shows the removal of turbidity and particles along the treat-
ment train: coagulation, ozonation/UF and activated-carbon ltration.
It was found that UF could remove turbidity almost completely. The ef-
uent turbidity was always below 0.1 NTU, although the turbidity of
raw water was uctuated. After activated carbon ltration, the efuent
turbidity slightly rose up but was less than 0.2 NTU. The counts of par-
ticles larger than 2 m in raw water and in the coagulation efuent
were almost 60,000 and 9000 CNT (count)/mL respectively, and it
decreased to below 10 CNT/mL in the membrane permeate. Under all
conditions, particles larger than 7 m were hardly detected, while the
particles between 2 and 5 mwere still found in the permeate samples.
The counts of particles larger than2 minthe GACefuent were around
40 CNT/mL; this may indicate that either ne carbon particles or bio-
fragments leaked from GAC ltration bed, but not many.
In this study, the Coliform group and standard plate-count bacteria
in rawwater were more than 16,000 MPN/100 mL and 860 CFU/mL re-
spectively, while the number of Cryptosporidium or Giardia cysts was
more than 0.3 or 0.5 per 10 L raw water. However, all of them were
not found in the membrane permeate (data not shown). It is indicated
that almost all the microorganisms, including pathogenic microorgan-
isms, are retained by the ceramic membranes in this study.
3.1.2. Removal of DOC and DBPs precursors
Fig. 3 is the result of DOC removal during the treatment processes
with or without ozonation. It can be seen that about 20% DOC was re-
moved by coagulation, about 5% by UF either withor without ozonation,
about 51% by GAC ltration if without ozonation, while about 64% if
with ozonation. Obviously, ozonation is very signicant in the improve-
ment of DOC removal in the hybrid processes.
The removal of THMFP and HAAFP with or without ozonation is
shown in Fig. 4. The removal ratio of THMFP and HAAFP were less than
20% by coagulation, 7% THMFP and 4% HAAFP by UF only, 11% THMFP
and 13% HAAFP by UF with ozonation. During the following GAC ltra-
tion in the integrated process, the removal efciencies reached up to
46% for THMFP and 45% for HAAFP respectively if without ozonation,
they increased up to 73% for THMFP and 75% for HAAFP respectively if
with ozonationin the treatment train. Apparently, the formation poten-
tials of either THMs or HAAs can be reduced effectively during the acti-
vated carbon ltration with the presence of ozonation in advance.
As a great number of studies have already shown, organic matters
are removed not only by the adsorption of granular activated carbon
Table 2
Quality control performances during multiple parameters analyses.
Compound Precursor
ions
Transition ions Recovery
(%) (n = 5)
LOD
(ng/L)
LOQ
(ng/L)
Quant Conrm
DBPs
CHCl
3
95 0.5 0.02
a
0.5
a
CHCl
2
Br 96 1 0.016
a
0.4
a
CHClBr
2
95 1 0.016
a
0.4
a
CHBr
3
93 1 0.025
a
0.6
a
MCAA 92.8 36.7 98 0.5 0.53
a
1.5
a
MBAA 136.8 78.7 101 2 0.016
a
0.053
a
DCAA 126.9 82.6 95 3 0.025
a
0.083
a
BCAA 172.6 80.8 97 3 0.04
a
0.12
a
TCAA 162.8 118.8 100 2 0.079
a
0.26
a
DBAA 218.7 80.8 102 4 0.16
a
0.51
a
Geosmin 182
b
112.0 115.0 79 12 2.8 8.4
2-MIB 168
b
95.0 135.0 86 8 2 6.5
EDCs
BPA 228.3
b
372.0 357.0 98 5 0.45 2
NP 220.35
b
207.0 91 6 0.5 3
OP 206.3
b
292.0 179.0 88 2 0.4 2
E1 270.4
b
257.0 342.0 75 6 0.45 4.5
E2 272.4
b
416.0 285.0 74 7 0.58 5.8
17-E2 272.4
b
326.0 285.0 86.3 4 0.6 6
E3 288.4
b
504.0 311.0 88 3 1.2 12
EE2 296.4
b
440.0 425.0 78 4 1.05 5
PPCPs
Benzabrate 362.1 121.1 139.1 103 10 0.2 0.5
DEET 192.1 119.0 73.0 66 5 0.4 1.2
Erythromycin 734.8 158.2 83.1 65 8 0.1 0.5
Griseofulvin 353.1 69.0 76 4 0.3 1
Indomethacin 359.1 126.2 77 4 0.2 0.5
Lincomycin 407.1 126.2 70 3 0.2 0.5
Metoprolol 268.2 77.1 74.0 84 7 0.2 1
Roxithromycin 837.5 158.2 83.1 85 2 0.3 0.9
Sulfamethoxazole 254.0 92.1 65.0 80 7 0.3 1
Sulfamerazine 265.1 92.2 108.0 85 4 0.2 1.1
Sulfapyridine 250.184 65 70 5 0.6 1.2
Sulpiride 342.146 112.2 214.1 80 4 0.3 1
Trimethoprim 291.102 230.2 123.2 80 6 0.2 0.6
2-QCA 176.2 91.0 117.1 77 2 0.4 1
a
Unit: g/L.
b
Molecular weight.
1
10
100
1000
10000
100000
0
5
10
15
20
25
30
Raw water After
coagulation
After
ozone/UF
After GAC
filtration
>
2

m

p
a
r
t
i
c
l
e
s

(
c
o
u
n
t
/
m
L
)
T
u
r
b
i
d
i
t
y

(
N
T
U
)
Turbidity >2m particles
Fig. 2. Removal of turbidity and particles during treatments.
50 X. Fan et al. / Desalination 335 (2014) 4754
but also by the degradation of biological reactions [20], which denoted
as biological activated carbon ltration (BAC) [10]. Organic matters
like humic acids may be converted into biodegradable alcohols and
ketones [20] during the ozonation process in the experiment. Conse-
quently, the removal ratios of DOC and DBP precursors are increased
in the presence of ozonation.
3.1.3. Removal of ammonia
During raining seasons, the ammonia concentration in source water
can reach up to over 3 mg/L as shown in Fig. 5. The amount of ammonia
was hardly reduced during coagulation and ceramic UF with or without
ozone added. However, ammonia was efciently removed by activated
carbon ltration. The removal ratio of ammonia was below40% if with-
out ozonation in advance. The concentration of nitrite which is harmful
to human health was as high as 0.75 mg/L in the product water. Com-
paratively, the removal ratio of ammonia increased to 98% if combined
with ozonation in advance, and no more nitrite was found present in
the product water. Dissolved oxygen (DO) is considered to be an elec-
tron acceptor in the bio-nitrication process; 4.57 mg O
2
/mg NH
3
N is
needed by nitriers in stoichiometric proportions. Since ozone was
generated by pure oxygen instead of air, the concentration of DO in
water increased to 1113 mg/L during ozone aeration. Ammonia was
transformed to nitrate in GAC ltration when DO present enough. The
actual DO amount utilized in removing NH
4
+
N was 4.52 mg O
2
/mg
NH
3
N which is less than the total DO demand in theory in this study.
Also, the produced amount of NO
3

N was slightly less than the sum

of NH
3
N and NO
2

N removed. There may be some other possible

pathways like assimilation or denitrication for ammonia removal [1].
3.1.4. Removal of taste and odor causing substances
The problem of taste and odor of drinking water was occasionally
complained by the consumers in the southern China. Geosmin and
2-MIB are two typical substances causing taste and odor problems;
both were investigated in the pilot-scale treatment process as shown
in Fig. 6.
The removal efciencies were 47% for geosmin and 50% for 2-MIB
respectively after coagulation, while they rose up to 96% and 88%
respectively after ozonation/UF combination. Relatively the activated
carbon ltration had little inuence on the further removal of geosmin
and 2-MIB.
Most studies inliterature reportedthat coagulationwas inefcient in
removing geosmin and MIB [21]. In this study the removal of geosmin
and 2-MIB may be caused by a fast stirring during the coagulation pro-
cess, in this way, some geosmin or 2-MIB may volatilize. Bruchet and
Laine [22] have found that UF alone would not reduce the substances
causing taste and odor problems in the treated water. Thus, ozonation
plays a major role in the oxidative removal of geosmin and 2-MIB. The
average concentrations of geosmin and 2-MIB in the product water
were 2.7 ng/L and 15.1 ng/L respectively in this study.
3.1.5. Removal of EDCs and PPCPs
Eight typical EDCs that were widely detected in local aquatic envi-
ronment [4,23] were tested in the experiment. Fig. 7 shows that NP
and BPA are the main EDC contaminants in the tested water. The
average concentrations of NP and BPA in the raw water were over
1000 ng/L and 70 ng/L respectively, while the concentrations of the
other six EDCs including OP, 17-E2, E1, E2, EE2 and E3 were less
than 15 ng/L in the raw water, particularly, the average concentrations
of E2 and 17-E2 under 5 ng/L. Coagulation was inefcient in the re-
moval of EDCs, for the total removal ratio was around 10%. The removal
ratio was 50% with UF alone, while it increased to 90% when combined
with ozonation. The total concentration of the eight EDCs in the product
water after GAC ltration was 45 ng/L if without ozonation, while
25 ng/L if with ozonation respectively. Obviously, ozonation improved
the removal of EDCs in UF and GAC ltration.
Fourteen PPCPs were detected in the rawwater including eight anti-
biotics and the others. The amount of sulfapyridine, indomethacin and
lincomycin in the raw water was accounted for over 70% of the total
PPCPs in proportion. The removal ratio of PPCPs by coagulation was
less than 20%. UF alone can reduce PPCPs by 50%, while UF/ozonation
can reduce PPCPs up to 86%. Five PPCPs with the total amount of
less than 3 ng/L were found left in the GAC efuent when ozonation
presented in the treatment train. Comparatively seven PPCPs with
total amount of 5 ng/L were found when without ozone present. The
removal of PPCPs was enhanced by ozonation in combination with UF
and GAC ltration.
Some conventional treatment processes like coagulation, sedimen-
tation and sand ltration were not efcient in the removal of EDCs
and PPCPs [24]. The removal efciencies of EDCs and PPCPs by coagula-
tion in this study agree well with those in previous studies [5].
Ceramic membrane UF showed the removal efciencies of EDCs and
PPCPs as high as over 50% in this study. Since the mean pore size of the
membrane is 60 nm, EDCs and PPCPs cannot be removed by size
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
Raw water After
coagulation
After UF After GAC
filtration
D
O
C

(
m
g
/
L
)
without ozone
with ozone
Fig. 3. Removal of organic matters during treatments with or without ozonation.
0
40
80
120
160
Raw water After
coagualtion
After UF After GAC
filtration
T
H
M
F
P

(

g
/
L
)
without ozone
with ozone
0
50
100
150
200
250
Raw water After
coagulation
After UF After GAC
filtration
H
A
A
F
P

(

g
/
L
)
without ozone
with ozone
Fig. 4. Removal of DBPFP during treatments with or without ozonation.
51 X. Fan et al. / Desalination 335 (2014) 4754
exclusion. EDCs and PPCPs may be removed by sorption on the cake
layer or the membrane inter-surface. The removal efciencies of NP, in-
domethacin and lincomycin were 55%, 99% and 98% respectively in UF
process. However, the removal efciencies of BPA and sulfapyridine
were 7% and 16%, respectively. The molecular properties of EDCs and
PPCPs, including hydrophilic/hydrophobic interactions of the contami-
nants at the membrane inter-surface, may have mechanistic inuences
on the removal behaviors during ceramic membrane ultra-ltration.
Ozone was efcient inthe removal of most EDCs and PPCPs. Also, the re-
moval of EDCs and PPCPs was improved during GAC ltration combined
with ozonation; that is, probably because the biodegradability of the ox-
idation products of EDCs and PPCPs increased [5,20].
3.2. Effect of ozonation on trans-membrane pressure (TMP)
The effect of ozonation on cross-membrane pressure was evaluated
as shown in Fig. 8. The pure water ux under 20 kPa suction pressure
for the membrane is roughly 120 L/m
2
h, which is higher than that in
Sartor's study [18]. In this study, a constant membrane ltration ux
of 100 L/m
2
h was maintained through the experiment via adjusting
the frequency of the suction pump automatically.
Fig. 8 shows the TMP trend throughout a test period of time. The
TMP increased nearly 10 kPa in 120 h during the ultra-ltration if
without ozonation. At an ozone dosage of 2 mg/L, the increasing rate
of TMP slowed down, and the TMP increased by 3.5 kPa in the test
period of time. When the ozone dosing raised to 5 mg/L for a short
period of 5 min every 55 min, the TMP slightly increased at the initial
time and then remained stable for a longer period of time. The total
increase of TMP was less than 2 kPa in 120 h.
It is shown in Fig. 8 that the ozone dosage of 2 mg/L was not enough
for complete control of the membrane fouling; thus the TMP slowly
increased. With the intermittent increase of ozone concentration to
5 mg/L, the TMP can remain stable for a longer period of time.
3.3. Discussion
For micro-polluted rawwater, the full treatment train conventional-
ly adopted in practice includes pre-oxidation, enhanced coagulation,
sedimentation, sand ltration, main oxidation, GAC ltration, mem-
brane ultra-ltration and disinfection. Such a treatment train is too
long to be afforded for the developing countries like China. The hybrid
treatment train developed in this paper, where the integration of UF
with ozonation (i.e. UF/ozonation) merged 5 units (including pre-
oxidation, sedimentation, ltration, main oxidation and UF) into one
unit, is equivalent to the whole conventional trains. In this way, the ox-
idation is not necessary to be carried out twice, and the particular mat-
ters can be removed by membrane ultra-ltrationone time for all, while
dissolved contaminants can be removed through adsorption and bio-
degradation in GAC ltration. Obviously, the treatment train developed
in this study is muchshorter. It is reasonably believed that such a hybrid
treatment train is suitable for the upgrade of conventional water treat-
ment plants in dealing with micro-polluted water sources for better
supply of drinking water.
The key for the new treatment train is the at-sheet ceramic mem-
brane, which is oxidation durable and can be put into the conventional
sedimentation tank in practice. According to Fig. 2, putting ceramic
membrane before GAC ltration in the treatment train is able to greatly
decrease the particles owing into GAC ltration unit, thus preventing
any microbiological risks in advance. Conventionally, the membrane is
putting behind ltration to prevent microbes leaking from ltration
bedinto water distributionsystem. It is believed that putting membrane
before GAC ltration bed is effective in preventing microbes fromenter-
ing the ltrationbed to further growor leak out. Furthermore, the ozone
can kill pathogens before GAC ltration.
The integration of ozonation with ceramic membrane UF may
strengthen the role of ozonation. According to Figs. 3 and 4, oxidation
in the treatment train can substantially change the molecular character-
istics of organic matters at mg/L level, and further lead to their removal
effectively in the following GAC ltration in terms of DOC and DBPFPs.
The removal efciency of DOC in GAC ltration is almost doubled than
usual. In Figs. 6 and 7, oxidation in the treatment train is able to directly
degrade some micro-pollutants like geosmin, 2-MIB, EDCs and PPCPs at
ng/L level, resulting in almost complete removal of these substances in
the following GAC ltration. In addition, aeration of the ozone that
was generated from pure oxygen can signicantly enhance the concen-
tration of DO in water, which benets the biological removal of ammo-
nia in GAC ltration.
The combination of ozonation with membranes may bring a promis-
ing way for in situ control of membrane fouling, which is a critical issue
in efcient operation of membrane ultra-ltration in engineering prac-
tice. As ozone present in the membrane tank, part of ozone would enter
the membrane pores with water ow; thus, ozonation would happen
inside the membrane pores, which may help to prevent organic
0
1
2
3
4
5
Raw water After
flocculation
After UF After GAC
filtration
C
o
n
c
e
n
t
r
a
t
i
o
n

(
m
g
/
L
)
Ammonia Nitrite Nitrate
0
1
2
3
4
5
Raw water After
flocculation
After
Ozon./UF
After GAC
filtration
C
o
n
c
e
n
t
r
a
t
i
o
n

(
m
g
/
L
)
Ammonia Nitrite Nitrate
Without ozone With ozone
Fig. 5. Removal of ammonia during treatments with or without ozonation.
0
30
60
90
120
150
Raw water After
flocculation
After
Ozonation/UF
After GAC
filtration
S
m
e
l
l
y

s
u
b
s
t
a
n
c
e
s

(
n
g
/
L
)
Geosmin
2-MIB
Fig. 6. Removal of geosmin and 2-MIB during treatment processes.
52 X. Fan et al. / Desalination 335 (2014) 4754
molecules frombeing adsorbed on the membrane pore interface or may
help to desorb already adsorbed foulants. Since these activities take
place in the nano-scale channels, where mass transfer is in nanometer
and reaction in nanoseconds, thus the porous channel can be called as
nano-reactor [25,26]. In this way, the integration of ozonation with
ceramic membrane can be simplied as a process in numerous nano-
reactors. Some previous research demonstrated that catalytic ozonation
on ceramic membrane surface was important for the control of mem-
brane fouling and the removal of organic matters [12]. The catalytic
ozonation inside membrane pores also should be taken into account.
Thus, much higher reaction efciency could be achieved in the mem-
brane pores. The further research is very meaningful to both academics
and engineering.
4. Conclusions
A hybrid process combining coagulation, ozonation, ceramic mem-
brane ultra-ltration and GAC ltration has been demonstrated to
have full capability to deal with the micro-polluted surface water. A va-
riety of contaminants can be removed effectively including turbidity
(99%), particle counts (99.9%), Coliform bacteria (100%), DOC (64%),
THMFPs (73%), HAAFPs (75%), ammonia (98%), geosmin (96%), 2-MIB
(88%), EDCs (98%) and PPCPs (98%). The integration of ozonation with
ceramic membrane has been demonstrated to play a critical role to en-
hance the removal efciencies of multiple contaminants and to alleviate
membrane fouling. Maintaining the ozone dosage at 25 mg/L can get
ultraltration working circle extended under low TMP of 20 kPa. The
energy consumption is 0.51 kW/m
3
. It is believed that the hybrid pro-
cess developed in this paper is of substantial signicance for the modern
upgrading of conventional water treatment plants in developing coun-
tries to deal with micro-polluted surface water.
Acknowledgments
This work has been supported by the National Grand Project Water
Pollution Control and Treatment Technology (2008ZX07423-002) as
well as the Guangdong Science Foundation (2012B030800001).
References
[1] X. Yu, Z.H. Qi, X.J. Zhang, P. Yu, B. Liu, L.M. Zhang, L. Fu, Nitrogen loss and oxygen
paradox in full-scale bioltration for drinking water treatment, Water Res. 41
(2007) 14551464.
[2] J.Y. Tian, Z.L. Chen, H. Liang, X. Li, Z.Z. Wang, G.B. Li, Comparison of biological activat-
edcarbon (BAC) and membrane bioreactor (MBR) for pollutants removal indrinking
water treatment, Water Sci. Technol. 60 (2009) 15151523.
[3] H. Du, Y. Xu, Determination of the microbial origin of geosmin in Chinese liquor,
J. Agric. Food. Chem. 60 (2012) 22882292.
0
200
400
600
800
1000
1200
1400
1600
Raw water After
flocculation
After UF After GAC
filtration
Raw water After
flocculation
After
Ozonation/UF
After GAC
filtration
A
v
e
r
a
g
e

c
o
n
c
e
n
t
r
a
t
i
o
n

(
n
g
/
L
)
E3 EE2 E2 E1 17a-E2 BPA OP NP
Without ozone With ozone
EDCs:
0
50
100
150
200
250
300
Raw water After
flocculation
After UF After GAC
filtration
Raw water After
flocculation
After
Ozonation/UF
After GAC
filtration
A
v
e
r
a
g
e

c
o
n
c
e
n
t
r
a
t
i
o
n

(
n
g
/
L
)
trimethoprim
Griseofulvin
sulfapyridine
sulfamerazine
2-QCA
Indomethacin
sulpiride
Sulfamethoxazole
roxithromycin
Metoprolol
Lincomycin
erythromycin
DEET
Benzafibrate
Without ozone With ozone PPCPs:
Fig. 7. The variation of EDC and PPCP concentration during treatment processes.
-30
-25
-20
-15
0 20 40 60 80 100 120
T
M
P

(
K
P
a
)
Time (hour)
without O3
O3 2mg/L
O3 2-5mg/L
Fig. 8. Effect of ozone dosage on trans-membrane pressure.
53 X. Fan et al. / Desalination 335 (2014) 4754
[4] L.Y. Wang, N.F.Y. Tam, X.H. Zhang, Analysis and occurrence of seven typical endo-
crine disrupting chemicals in three sewage treatment plants, Water Sci. Technol.
60 (2010) 25012509.
[5] T.J. Qiao, Z.R. Yu, X.H. Zhang, W.T.A. Doris, Occurrence and fate of pharmaceuticals
and personal care products in drinking water in southern China, J. Environ. Monit.
13 (2011) 30973103.
[6] S.G. Iris, H. David, S. Raphael, Humic substances in ultraltration processes, Desali-
nation 216 (2010) 218231.
[7] W. Gao, H. Liang, J. Ma, M. Han, Z.L. Chen, Z.S. Han, G.B. Li, Membrane fouling control
in ultraltration technology for drinking water production: a review, Desalination
272 (2011) 18.
[8] M. Hashino, Y. Mori, Y. Fujii, N. Motoyama, N. Kadokawa, H. Hoshikawa, W.
Nishijima, M. Okada, Pilot plant evaluation of an ozone-microltration system for
drinking water treatment, Water Sci. Technol. 41 (2000) 1723.
[9] B.S. Karnik, M.J. Baumann, S.J. Masten, M.J. Davies, AFM and SEM characterization of
iron oxide coated ceramic membranes, J. Mater. Sci. 41 (2006) 68616870.
[10] B. Schlichter, V. Mavrov, H. Chmiel, Study of a hybrid process combining ozonation
and microltration/ultraltration for the drinking water production from surface
water, Desalination 168 (2004) 307317.
[11] J. Kim, S.H.R. Davies, M.J. Baumann, V.V. Tarabara, S.J. Masten, Effect of ozone dosage
and hydrodynamic conditions on the permeate ux in a hybrid ozonationceramic
ultraltration system treating natural waters, J. Membr. Sci. 311 (2008) 165172.
[12] B.S. Karnik, S.H.R. Davies, K.C. Chen, D.R. Jaglowski, M.J. Baumann, S.J. Masten, Effect
of ozonation on the permeate ux of nanocrystalline ceramic membranes, Water
Res. 39 (2005) 728734.
[13] B.S. Karnik, S.H.R. Davies, M.J. Baumann, S.J. Masten, Fabrication of catalytic mem-
branes for the treatment of drinking water using combined ozonation and ultral-
tration, Environ. Sci. Technol. 39 (2005) 76567661.
[14] B.S. Karnik, M.J. Baumann, L.M. Corneal, S.J. Masten, S.H.R. Davies, TEMcharacteriza-
tion of iron-oxide-coated ceramic membranes, J. Mater. Sci. 44 (2008) 41484154.
[15] B. Schlichter, V. Mavrov, H. Chmiel, Study of a hybrid process combining ozonation
and membrane ltrationltration of model solutions, Desalination 156 (2003)
257265.
[16] S. Byun, S.H.R. Davies, A.L. Alpatova, L.M. Corneal, M.J. Baumann, V.V. Tarabara, S.J.
Masten, Mn oxide coated catalytic membranes for a hybrid ozonationmembrane
ltration: comparison of Ti, Fe and Mn oxide coated membranes for water quality,
Water Res. 45 (2011) 163170.
[17] L.M. Corneal, M.J. Baumann, S.J. Masten, S.H.R. Davies, V.V. Tarabara, S. Byun, Mn
oxide coated catalytic membranes for hybrid ozonationmembrane ltration:
membrane microstructural characterization, J. Membr. Sci. 369 (2011)
182187.
[18] M. Sartor, B. Schlichter, H. Gatjal, V. Mavrov, Demonstration of a newhybrid process
for the decentralised drinking and service water production from surface water in
Thailand, Desalination 222 (2008) 528540.
[19] S.B. Watson, B. Brownlee, T. Satchwill, E.E. Hargesheimer, Quantitative analysis of
trace levels of geosmin and MIB in source and drinking water using headspace
SPME, Water Res. 34 (2000) 28182828.
[20] C. Volk, P. Renner, H. Roche, H. Paillard, J.C. Joret, Effects of ozone on the production
of biodegradable dissolved organic carbon (BDOC) during water treatment, Ozone
Sci. Eng. 15 (1993) 389404.
[21] H. Ahn, S. Chae, S. Kim, C. Wang, R.S. Summers, Efcient taste and odour removal by
water treatment plants around the Han River water supply system, Water Sci.
Technol. 55 (2007) 103109.
[22] A. Bruchet, J.M. Laine, Efciency of membrane processes for taste and odor removal,
Water Sci. Technol. 51 (2005) 257265.
[23] G.G. Ying, R.S. Kookana, A. Kumar, M. Mortimer, Occurrence and implications of
estrogens and xenoestrogens in sewage efuents and receiving waters from South
East Queensland, Sci. Total Environ. 407 (2009) 51475155.
[24] T. Basile, A. Petrella, M. Petrella, G. Boghetich, V. Petruzzelli, S. Colasuonno, D.
Petruzzelli, Review of endocrine-disrupting-compound removal technologies in
water and wastewater treatment plants: an EU perspective, Ind. Eng. Chem. Res.
50 (2011) 83898401.
[25] L. Qiao, Y. Liu, S.P. Hudson, P.Y. Yang, E. Magner, B. Liu, A nanoporous reactor for
efcient proteolysis, Chem. Eur. J. 14 (2008) 151157.
[26] M.J. Monteiro, Nanoreactors for polymerizations and organic reactions, Macromole-
cules 43 (2010) 11591168.
54 X. Fan et al. / Desalination 335 (2014) 4754