Water Research 168 (2020) 115148

Contents lists available at ScienceDirect

Water Research
journal homepage: www.elsevier.com/locate/watres

Biological ion exchange as an alternative to biological activated carbon

for drinking water treatment
Zhen Liu a, b, *, Kim Maren Lompe b, Madjid Mohseni c, Pierre R. Be

rube

d,

bastien Sauve
Se
a, Benoit Barbeau b
a
Department of Chemistry, University of Montr
eal, Montr
eal, QC, H3T 1J4, Canada
b
NSERC-Industrial Chair on Drinking Water, Department of Civil, Mining and Geological Engineering, Polytechnique Montreal, Montr

eal, QC, H3T 1J4,
Canada
c
Department of Chemical and Biological Engineering, University of British Columbia, Vancouver, BC, V6T 1Z4, Canada
d
Department of Civil Engineering, University of British Columbia, Vancouver, BC, V6T 1Z4, Canada

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

Article history: Biological ion exchange (BIEX) has proved to remove natural organic matter (NOM) better than biological
Received 11 August 2019 activated carbon (BAC). This raises the question if BIEX can be integrated into a full-scale drinking water
Received in revised form treatment plant to remove NOM and ammonia. In this study, a pilot plant consisting of one BIEX filter,
24 September 2019
three GAC filters and one BAC filter was set up as second-stage filtration at the Sainte-Rose drinking
Accepted 1 October 2019
water treatment plant (Laval, Canada). The pilot plant was operated for a period of nine months without
Available online 5 October 2019
regeneration of the ion exchange resins. The influent water showed low DOC (2.5 mg/L) and high sulfate
concentrations (28.2 mg/L). Except of a short peak of DOC released at about 1 000 BV, the BIEX filter
Keywords:
Biological ion exchange
achieved a nearly constant removal of 29e36% over the whole study period. The DOC removals of GAC
Biological activated carbon were similar to BIEX at < 8000 BV but then stabilized at 13e24% after 8 000 BV. Most DOC removal in the
Natural organic matter BIEX filter was achieved at the top 30 cm layer (81%) compared to 62e66% removal in the GAC/BAC filters
Anion exchange resin in the same layer. After the rapid exhaustion of the primary ion exchange capacity (<1 000 BV), sulfate
Nitrification displaced the fraction of NOM with lower affinity than sulfate, corresponding to the initial DOC release in
the BIEX filter. The fraction of NOM with higher affinity than sulfate can still replace sulfate, which
explains the good long-term performance of the BIEX filter. ́ BIEX released ammonia with an average of
15% in warm water condition, probably related to the small diameter of the column which limited
backwash effectiveness.
© 2019 Elsevier Ltd. All rights reserved.

1. Introduction system (Hijnen et al., 2018). Water treatment plants (WTPs)

Removing natural organic matter (NOM) is one of the main
objectives during drinking water treatment as NOM may cause
taste, odor and color issues (Edzwald, 2010; Thurman, 2012),
membrane and filter fouling (Gibert et al., 2013; Kennedy et al.,
2008; Kim and Dempsey, 2013), formation of disinfection by-
products (DBPs) (Krasner et al., 2006; Richardson et al., 2007;
Brezinski et al., 2019) as well as biofilm growth in the distribution
* Corresponding author. Department of Chemistry, University of Montre
al,
Montre
al, QC, H3T 1J4, Canada.
E-mail addresses: zhen.liu.1@umontreal.ca (Z. Liu), kim.lompe@polymtl.ca
(K.M. Lompe), madjid.mohseni@ubc.ca (M. Mohseni), berube@civil.ubc.ca

rube
(P.R. Be
), sebastien.sauve@umontreal.ca (S. Sauve

), benoit.barbeau@polymtl.
ca (B. Barbeau).
currently employ different techniques for NOM removal, including
coagulation/flocculation (Matilainen et al., 2010), activated carbon
adsorption (Velten et al., 2011), and biofiltration (Korth et al., 2001).
Among these options, NOM removal through biofiltration has
received considerable attention due to its low maintenance cost
and ease of operation. Rapid-rate biofilters can be designed using
various filtration media to act as a support for the development of
biomass. Among them, granular activated carbon (GAC) has been
shown to offer better performance compared to inert medias (e.g.,
sand, anthracite), an advantage related to their sorptive capacity.
After the exhaustion of this capacity, the GAC transitions into the
so-called biological mode (i.e. biological activated carbon or BAC)
which can also offer high NOM removal in cold water due to their
higher surface area or porosity available for biomass growth and
the potential bio-regeneration of the adsorption capacity (Basu

https://doi.org/10.1016/j.watres.2019.115148
0043-1354/© 2019 Elsevier Ltd. All rights reserved.
2 Z. Liu et al. / Water Research 168 (2020) 115148
et al., 2016). Nevertheless, NOM removal with BAC filtration is fairly
limited (5%e20% DOC removal) and the kinetic is highly impacted
by temperature (Terry and Summers, 2018). For this reason, bio-
filtration is not typically used as the sole NOM removal process
within a treatment train unless the source water naturally offers a
low NOM content.
Recently, we have reported that ion exchange (IEX) resins can be
used as an alternative media for biofiltration (Schulz et al., 2017).
Firstly, IEX resins remove NOM by anion exchange with a counter-
ion (usually chloride). Then, similar to a GAC filter, a progressive
depletion of exchange sites is observed while the surface of the
resin is colonised by biofilm which participates to NOM reduction
via biodegradation. During a lab-scale study (Schulz et al., 2017),
IEX filters were fed with a 5 mg C/L surface water with low mineral
anion content for a period of one year. We observed an average DOC
removal of 60% for an IEX filter and 40% for an IEX filter made
abiotic through continuous sodium azide injection. The develop-
ment of biofilm was proposed as the reason for this increased
performance and the process was referred to as biological ion ex-
change or BIEX. In a follow-up lab-scale study using the same
source water, BIEX was shown to provide superior DOC removals
(56 ± 7%) compared to a BAC filter (15 ± 5%) (Winter et al., 2018).
This performance was confirmed in a pilot study (Amini et al.,
2018), where four types of filter (IEX, BIEX, GAC and BAC) were
fed for 331 days (without regeneration of BIEX) using a river water
of high DOC (7e8 mg C/L) and low anionic mineral content (5e8 mg
SO
2
4 /L). During summer conditions after five months of operation
(z 840 bed volumes), BIEX achieved 62% DOC removal whereas the
BAC filter achieved only 7%. Nitrification in the BIEX filter was as
efficient as in the BAC filter, confirming that the BIEX media was
biologically active. IEX was exhausted significantly later than GAC
(90 d vs. < 7 d, respectively). While a carbon mass balance calcu-
lated after 331 days of operation showed that about 31% of DOC
removal in the BIEX filter was due to biodegradation, the high
performance of BIEX could not be explained by biodegradation
alone. Secondary ion displacements (e.g. NOM displacing sulfate)
was proposed as an important mechanism to explain BIEX perfor-
mance after initial anion exchange process with chloride was
exhausted. However, more studies are needed to elucidate the
competition of NOM with sorbed anions after the exhaustion of the
initial IEX capacity.
Resin suppliers consider biomass growth on resin a nuisance as
it may cause excessive fouling in the IEX filter (Flemming, 1987).
Hence, conventional IEX are regularly regenerated using a highly
concentrated salt solution (usually 10e12% W/V NaCl solution) to
regain the ion exchange capacity and alleviate biofouling phe-
nomena. Regeneration produces a brine containing high concen-
trations of sodium chloride and NOM which has to be disposed of
(Levchuk et al., 2018). Direct disposal of brine to the aquatic envi-
ronment or to the sewer is prohibited in many jurisdictions, since
high concentrations of chloride can cause corrosion of plumbing,
have detrimental effects on biological wastewater treatment pro-
cesses and can be toxic to the aquatic fauna and flora (Rokicki and
Boyer, 2011). BIEX provides a novel ion exchange operation strategy
since it implies greatly reducing the frequency of regeneration.
Operating IEX in BIEX mode could also potentially lengthen the
service life of the resin, which is strongly dependent on the number
of regeneration cycles (Hochmuller, 1984).
Up to now, we have only evaluated BIEX as a standalone process
using surface waters with low mineral content as we targeted ap-
plications in small rural systems given the simplicity of operation.
The high performance achieved so far has raised the question as
whether BIEX could also be an attractive option for large-scale
WTPs. For such application, it is anticipated that BIEX filters
would be located after coagulation and settling. Therefore, a lower
influent DOC would be expected compared to the conditions tested
so far. If alum is used as coagulant, high sulfate concentrations in
the feed water are expected which may be in competition with
NOM for IEX sites. Thus, the general objective of this study was to
investigate the potential of a BIEX filter to replace a second-stage
BAC filter as a polishing process for simultaneous NOM removal
and nitrification. Specifically, we compared BIEX with four alter-
native GAC (three fresh and one exhausted). NOM, DBP precursors,
ammonia removals as well as head loss rates served as metrics to
compare their performances.
2. Materials and methods
2.1. Pilot location and source water characteristics
This study was conducted for a period of 9 months at the Sainte-
Rose drinking water treatment plant (Laval, Canada). With a daily
production capacity of 110 000 m3, the plant is fed by the Mille-Iles
River and employs coagulation (alum), flocculation, sedimentation,
sand-anthracite filtration, ozonation, BAC filtration and chlorina-
tion. We set up pilot filters after ozonation, in parallel to the full-
scale BAC filtration. The influent to the pilot showed low turbidity
(z0.5 NTU) and low DOC (z2.5 mg C/L) (Table 1). The concentra-
tion of sulfate (z28.2 mg/L) was high due to the use of alum as a
coagulant (average dosage z 40 mg/L). This condition of operation
corresponds to a sulfate/DOC ratio of 11.2. These characteristics
contrasted strongly with our previous assay (Amini et al., 2018)
where a BIEX filter was fed with raw waters with a DOC of
7e8 mg C/L and 5e8 mg/L of sulfate which translates into a sulfate/
DOC ratio roughly ten times lower (0.6e1.1).
2.2. Pilot plant design and operation
The pilot plant consisted of five parallel columns (CPVC, 5.08 cm
diameter), each being filled with a 30 cm sand sublayer (0.6 L) and a
150 cm top layer (3.0 L) of either GAC, BAC or IEX (Fig. 1). The five
top layer media investigated in this study (Table 2) were: (1) virgin,
type-I macroporous strong base anion exchange resin with an
acrylic quaternary amine backbone in chloride form (Purolite®
A860, IEX capacity: 0.8 eq/L), (2) three virgin wood-based activated
carbons (Calgon Acticarbone® BGX, Nuchar® WV-B30, Jacobi
PICABIOL-HP120) and (3) exhausted activated carbon (Jacobi
PICABIOL-2). The latter was recovered from a full-scale BAC filter at
the Chomedey drinking water treatment plant (Laval, Canada),
where it had been in operation for 2 years. Only wood-based or
lignite GAC media were tested during this study as past research
conducting at this location had shown their superiority over
mineral-based GAC with respect to nitrification.
Four sampling nozzles were located on each of the columns at
Table 1
Influent water characteristics (average ± standard deviation) during the pilot study
period (April 11, 2018 to January 07, 2019).
Parameters Unit Valuea
Temperature 
C 14.1 ± 9.2
Turbidity NTU ~0.5
DOC mg C/L 2.5 ± 0.2
UVA254 cm
1 0.019 ± 0.007
SUVA L/mg,m ~0.76
pH e 6.5 ± 0.4
Chloride mg/L 7.7 ± 4.4
Sulfate mg/L 28.2 ± 4.5
Alkalinity CaCO3 mg/L 12.1 ± 7.7
Ammonia mg N/L 78.0 ± 26.0
a
Average ± standard deviation.
 Z. Liu et al. / Water Research 168 (2020) 115148
Fig. 1. Schematic overview of the pilot plant consisting of five parallel columns: BIEX,
three GAC filters and one BAC filter. The pilot is fed with coagulated/settled/filtered/
ozonated surface waters from the Sainte-Rose drinking water treatment plant.
different depths (10 cm, 30 cm, 60 cm, 150 cm) to profile the columns
performance (Fig. 1). The columns were operated under pressure and
were equipped with manometers to monitor head loss accumula-
tion. Given that IEX resins have low resistance to ozone and that the
columns were fed with post-ozonated waters, we added a pre-filter
containing 5 cm of fresh GAC (Jacobi PICABIOL-HP120) with an
empty bed contact time (EBCT) of 10 s in order to destroy any re-
sidual ozone that may be present in the influent. The interference on
influent water characteristics was tested and found to be negligible
(average DOC removal during study  3%, average ammonia removal
during study  1%, data not shown).
The filtration rate was kept constant at 10 m/h (10.8 min EBCT)
which is a typical filtration design rate for a second stage filter. The
columns (including the GAC pre-filter) were backwashed every two
weeks using first air injection (3 min at 10 m/h) and then unchlo-
rinated BAC filtered water from the plant. Backwash water flowrate
was adjusted to yield a 50% bed expansion. The duration of the
backwash was media specific as it was conducted until the back-
wash waters turbidity was lower than 10 NTU. Typically, backwash
lasted 40 min for the BIEX filter and 20 min for the GAC and BAC
filters.
2.3. Analytical methods
Influent and effluent water characteristics were sampled weekly
for the study period. Temperature, pH, and dissolved oxygen were
measured on-site with a multimeter (HACH HQ40D). DOC and
Table 2
Media characteristics for IEX, three different GAC and BAC used in this study.
Abbreviations Media Diameter (mm)
BIEX Purolite® A860 0.3e1.2
GAC1 Calgon Acticarbone® BGX 0.5e0.7
GAC2 Nuchar® WV-B30 0.8e1.1
GAC3 Jacobi PICABIOL-HP120 1.2e1.4
BAC Jacobi PICABIOL-2 1.2e1.4
Cu: Uniformity coefficient ¼ d60/d10; D10: measured with a MASTERSIZER 3000 (Malvern, UK).
Other information was provided by suppliers.
3
UVA254 samples were first prefiltered with a syringe filter (0.45 mm
Merck Millex®-HV) prior to analysis on a TOC-meter (Sievers
5310C, GE water, USA) or a spectrophotometer (Ultrospec 3100pro,
GE Healthcare, USA). Anions (nitrate, chloride, sulfate) were
measured with an ion chromatograph (ICS 5000 AS-DP DIONEX,
USA) equipped with an AS18 column. Bicarbonate concentration
was evaluated from alkalinity measurements which were per-
formed using the acid titration method 5.10 (USEPA 2012). THM
and HAA5 precursors were measured after chlorination according
to the Uniform Formation Conditions technique (UFC) (Summers
et al., 1996), i.e. by maintaining a free chlorine residual of
1.0 ± 0.5 mg Cl2/L after a contact time of 24 h at pH 8.0 and 20  C.
They were analyzed by gas chromatography (7890B GC system
from Agilent Technologies, USA) according to methods 524.2 (THM)
and 552.3 (HAA) (USEPA 2003). Ammonia concentrations (in mg N/
L) were quantified in triplicate using the indophenol colorimetric
method NF T90-015 (AFNOR 2000) with a 5 cm spectrophotometric
cell in order to achieve a detection limit of 5 mg N/L.
3. Results
3.1. Head loss accumulation
Head losses were recorded after (i) 30 min of operation
following a backwash (D1), (ii) seven days of operation (D7) and
(iii) fourteen days of operation (D14) (Fig. 2). The BIEX filter was
composed of a finer media which explains why it generated the
highest head loss compared to GAC and BAC filters. The median
BIEX head losses were 1.48 m for D1, 3.44 m for D7 and 3.60 m for
D14, which translates into a head loss accumulation of 2.12 m after
fourteen days. As a comparison, GAC1, which had higher head loss
than the other GAC, exhibited a median head loss of only 0.72, 0.73
and 0.81 m for D1, D7 and D14 conditions, respectively (accumu-
lation of 0.09 m after fourteen days). The other filters GAC2, GAC3
and BAC had lower head losses with values between 0.36 and
0.59 m for all the monitored conditions with an accumulation of
only 0.11e0.16 m after fourteen days. Overall, the head loss accu-
mulation after 14 days was low and very similar for all GAC/BAC
media tested. On the other hand, the head loss of BIEX filter after
fourteen days of operation fluctuated between 2.19 m (minimum)
and 5.63 m (maximum) during the study. Such head losses are high
compared to the typical maximum allowable head loss (2.4 m) in a
gravity-fed granular media filter according to (Kawamura, 2000).
This also explains why we decided to operate the filters under
pressure rather than by gravity. Finally, we observed that most of
the head loss accumulation in the BIEX filter occurred during the
first week of operation, a phenomenon which differed from the
other GAC/BAC filters. We suspect that the BIEX media was not fully
compacted after 30 min of operation following the backwash.
3.2. NOM removal
During the study period, the average DOC concentration and
D10 (mm) Cu Surface area (m2/g)
0.5 1.7 2.19
0.6 1.8 1550e1650
1.0 1.8 1400e1600
1.3 1.4 1400
1.3 1.4 1600
4 Z. Liu et al. / Water Research 168 (2020) 115148

performance of the filters was divided into three phases. (1) NOM
release from BIEX and initial GAC exhaustion (0 to 8 000 BV):
BIEX first showed a 50% DOC release compared to the influent after
one week of operation (z1 000 BV). After this initial DOC break-
through, BIEX removed DOC at an average of 29% from 1000 to
8000 BV. During this period, all GAC filters achieved higher per-
formances (GAC1: 48%, GAC2: 39%, GAC3: 32%, on average) while
the BAC filter (i.e. exhausted GAC) achieved 10% DOC removal. (2)
From 8 000 to 25 000 BV (mid-June to mid-October): During this
period, the water temperature rose progressively from 18  C to
26  C in JulyeAugust before it dropped to 13  C in October. The BIEX
filter achieved a higher DOC removal than GAC/BAC filters during
this period with an average of 36% DOC removal. All GAC filters
exhibited a similar performance with average DOC removals of 24%,
22%, 21% for GAC1, GAC2 and GAC3, respectively. These perfor-
mances were slightly superior to the BAC filter which achieved 17%
removal during this period, an indication that the GAC filters still
exhibited a small adsorption capacity. (3) From 25 000 to 37 000
BV (October-January): During this period, the temperature
declined from 13 to 1  C. The BIEX filter maintained an average DOC
removal of 30%. In contrast, the average DOC removals declined to
16%, 13% and 15% for GAC1, GAC2 and GAC3, respectively, while the
Fig. 2. Head loss accumulation for the BIEX, three GAC and the BAC filter after (i)
BAC filter achieved 13% removal. In summary, apart from the initial
30 min of operation following a backwash (D1), (ii) seven days of operation (D7) and DOC release observed at about 1 000 BV, BIEX offered equivalent
(iii) fourteen days of operation (D14). performance to GAC (<8 000 BV) or superior (>8 000 BV) to GAC or
BAC filtration. Among the three tested GAC, GAC1 showed higher
performance than GAC2 and GAC3 (p < 0.05) while GAC2 and GAC3
UVA254 of the influent were 2.5 ± 0.2 mg C/L and showed no significant difference with regards to DOC removal
0.019 ± 0.007 cm
1 (average ± standard deviation), respectively. (p > 0.05).
We illustrated the normalized DOC concentration (Ceffluent/Cinfluent) With respect to UV254 absorbance (Fig. S1), removals fluctuated
and the normalized UVA254 absorbance (UVeffluent/UVinfluent) as a as they were directly impacted by the ozonation conditions pre-
function of the number of bed volumes (BV) of filtered water (Fig. 3 ceding the pilot filters. The ozone residual concentration in the
& S1). Each day of operation was equivalent to about 133 BV. The influent varied from a minimum of 0.07 mg O3/L to a maximum of

Fig. 3. Normalized DOC concentrations (Ceffluent/Cinfluent) in the effluents of the BIEX filter, three GAC filters and one BAC filter for the study period from April 11, 2018 to January 07,
2019.
 Z. Liu et al. / Water Research 168 (2020) 115148
0.6 mg O3/L (according to the on-line monitor of the plant).
Nevertheless, the trend in UV254 removal in the BIEX filter exhibited
a similar pattern to the one of DOC removal. For example, BIEX
surpassed the other filters at about 8000 BV and became the most
efficient media until the end of the study. For the entire study
period, BIEX provided the best UVA254 removal with an average of
48% compared to GAC1 (31%), GAC2 (28%), GAC3 (25%) and BAC
(15%). The fact that the UVA254 in BIEX effluent was sustained for a
very long period, independent of water temperature, suggests that
removal of aromatic NOM through ion exchange was an active
mechanism throughout the study.
We realized a profile study to monitor DOC removal across the
filter beds at about 21 000 BV ¼ 154 days of operation (Fig. 4). In the
BIEX filter, the upper 30 cm layer (1.8 min EBCT) made up for 81% of
the DOC removal while the other 150 cm (9 min EBCT) accounted
for the remaining 19%. However, in the case of the activated carbon
filters, all filters showed similar patterns where 62e66% of the total
DOC removal was achieved in the top 30 cm while the remainder
(34e38%) was removed in the lower 150 cm layer. The rapid
removal kinetic of BIEX is more consistent with ion exchange than
biodegradation. This finding also indicates that the BIEX filter
would be a more robust process than BAC filtration given that NOM
is removed more effectively inside the filter.
3.3. Removal of THM and HAA precursors
From August 22, 2018 to January 7, 2019 (18 000 BV to 37 000
BV), we monitored THM and HAA5 precursors biweekly in the
influent and effluents (Fig. 5). The average influent THM-UFC and
HAA5-UFC concentrations were 60 mg/L and 68 mg/L, respectively.
BIEX achieved the highest THM and HAA precursors reductions
with an average of 48% and 66%, respectively. Meanwhile, all GAC
filters had similar performance with average reductions of 28e33%
for THM-UFC and 45e48% for HAA5-UFC. Removal of THM and HAA
precursors averaged 28% and 44%, respectively in the BAC. The
precursors removal performance was consistent with the observed
DOC and UVA254 removals for the five filters, confirming the higher
NOM removal performance in the BIEX filter.
3.4. Impact of BIEX on inorganic anions
The major inorganic anions concentrations were monitored
weekly in the influent and BIEX effluent. The average influent
Fig. 4. Distribution of total DOC removal (0.87, 0.76, 0.73, 0.71, 0.52 mg C/L) in the
upper (30 cm) vs. lower layer (150 cm) for BIEX, GAC1, GAC2, GAC3 and BAC, realized at
about 21 000 BV.
5
concentrations (±standard deviation) of sulfate, chloride, bicar-
bonate and nitrate for the study period were: 28.2 ± 4.5 mg/L,
7.7 ± 4.4 mg/L, 14.8 ± 9.4 mg/L and 1.1 ± 0.6 mg/L, respectively.
Given that ion exchange resin was initially charged with chloride,
the average concentration of chloride in the BIEX effluent was
higher than in the influent (9.1 ± 11.7 mg/L). In order to understand
the dynamics of ion exchange occurring on the resins, the loading
of anions on the resin was calculated using a weekly cumulative
charge balance (expressed as eq/L) as described by equation (1):
 
X X
i¼34 j¼3
Cin;i;j
Cout;i;j  Q  7d
qði; jÞ ¼ (1)
i¼1 j¼1
Vresin
where q(i,j) is the total cumulative loading of the anion sorbed on
the resin after week i (expressed as eq/L of resin). The index j de-
scribes the three solutes considered in the mass balance (DOC,
sulfate and chloride). Cin,i,j and Cout,i,j describe the concentrations of
the solute j during week i in the influent and effluent, respectively.
Q is the flow (486 L/d) while V is the volume of resin in the column
(3 L). Bicarbonate and nitrate were neglected from the calculation
as they were found not to be significantly removed in the BIEX filter.
DOC charge density was assumed to be 10 meq/g C, a value
representative of a low SUVA (0.7 L/mg,m) NOM at pH 6.5 based on
Boyer et al. (2008).
Fig. 6 presents the result of the cumulative loading of the three
studied solutes on the resin during the study. The initial capacity of
the fresh resin was 0.80 eq/L present under the form of chloride.
During the first week of operation, this chloride was replaced by
sulfate which reached 0.76 eq/L. In other words, the high con-
centration of sulfate in the influent almost “regenerated” the IEX
resin from chloride-form into a sulfate-form within 1 000 BV. At
this point, the resin was therefore fully exhausted with respect to
chloride release, a period (z1 000 BV) which corresponded to the
significant DOC release (3.31 mg C/L) measured in the effluent. We
postulate that the incoming sulfate displaced the fraction of NOM
on the resin with a lower affinity than sulfate. Interestingly, the
sulfate concentration on the resin started to decline after 1 000 BV
whereas the DOC and chloride concentrations on the resin started
to rise. At the end of the study, the DOC and chloride concentra-
tions on the resin were about 0.33 and 0.10 eq/L, respectively while
the sulfate concentration had decreased to about 0.19 eq/L resin.
The total loading of the three studied solutes equals to 0.62 eq/L
resin (lower than IEX capacity 0.80 eq/L resin), indicating that
other anions excluded from this study also occupied a portion of
IEX position on the resin phase. The variation in sulfate concen-
tration was not monotonous and most likely reflects the impact of
variations in influent water characteristics. Finally, we do not know
if the DOC would have eventually broken through once all the
sulfate had been released from the resin. However, Fig. 6 suggests
that the resin would have been fully loaded with NOM and chlo-
ride at approx. 40 000 BV. Considering that IEX filters are typically
regenerated after 48e72 h of operation (500-1 000 BV), the BIEX
mode of operation would translate into very low regeneration
costs.
3.5. Removal of ammonia
During the study period, the average concentration (±standard
deviation) of ammonia in the influent was 78 ± 26 mg N/L. We
monitored the effluent ammonia concentration of the five filters
(Fig. 7). Before 11 weeks (z10 000 BV), BIEX, GAC1, GAC2 and GAC3
showed negligible ammonia removals with averages of 0%, 9%, 13%
and 10%, respectively. Meanwhile, the BAC filter released ammonia
with an average of 19% during this period. We postulate that this
6 Z. Liu et al. / Water Research 168 (2020) 115148
Fig. 5. Distribution of (A) THM-UFC and (B) HAA5-UFC precursors in the influent and effluents of the filter media under investigation. Period: 18 000 to 32 000 BV. Numbers indicate
the average concentrations.
Fig. 6. Evolution of the anion concentrations on the resin during 34 000 BV.
may have been due to a period of the nitrifying biomass reac-
climating to its new environment given that it was collected from a
first stage biofilter. As water temperature continued to rise, GAC1
and GAC2 were the first filters to nitrify (11 weeks z10 000 BV)
while GAC3 and BAC followed the same trend one week later (12
weeks z11 000 BV). From 15 to 25 weeks (14 000e24 000 BV), all
GAC and BAC filters provided excellent ammonia removals
(98e99%) under warm water condition (17e26  C). On the other
hand, the BIEX filter released an average 15% ammonia from weeks
11e25, i.e. during summer conditions. After 25 weeks (z24 000
BV), as temperature declined below 15  C, ammonia removal first
declined in the BAC filter which eventually completely lost treat-
ment capacity after 32 weeks or 4 weeks of operation below 10  C.
The GAC1, 2 and 3 filters also suffered from the decline of tem-
perature but were still nitrifying 14%, 49% and 32% respectively
after 35 weeks (or 7 weeks below 10  C). Meanwhile, the BIEX
continued to show no ammonia removal during this period. Over-
all, for the entire study period, GAC2 had the best performance
with an average removal of 68% while GAC1 and GAC3 showed
an equivalent (p > 0.05) performance, with an average of 59%
removal.
4. Discussion
4.1. Head loss in the BIEX filter
The BIEX filter had the highest head loss compared to GAC and
BAC filters mainly due to the smaller resin bead size. Operation in a
pressurized vessel is thus advisable. Alternatively, the use of larger
IEX bead size could be an option to mitigate head loss, although
using larger beads may extend the mass transfer zone inside the
resin particles (Ball and Harries, 1989). Future research will also
need to address the design conditions required to properly back-
wash BIEX filters. The development of a crust in the upper layer is a
phenomenon that will need to be managed with proper air/water
backwash and probably more frequent backwash than what was
done during this study (i.e., once every 2 weeks) to reduce the head
loss accumulation in BIEX filters.
4.2. NOM removal mechanisms in the BIEX filter
In our previous study (Amini et al., 2018), we have evaluated
that the BIEX filter operated under summer conditions removed as
much as 62% of NOM, using a high DOC/low inorganic anions
surface water. The observed performance was much higher
compared to a BAC filter (7%). In this study, DOC removal in the
BIEX was once again higher than what was achieved in the GAC/
BAC filters investigated, despite being fed with a low DOC/high
sulfate pre-treated water. The sulfate concentration in the influent
is an important factor for IEX processes, as (i) high sulfate con-
centration can lead to a reduction in DOC removal (Ates and
Incetan, 2013; Dixit et al., 2018), and (ii) they can reverse the
NOM preference from low molecular weight (MW) species to high
MW ones (Tan and Kilduff, 2007). However, Verdickt et al. (2012)
found that the DOC removal efficiency did not reduce significantly
using an IEX regenerated with sulfate (33%) compared to a con-
ventional IEX with chloride as counter-ion (42%). Also, using IEX
in sulfate form can avoid unnecessary anions exchange, since
anions having lower affinities towards IEX compared to sulfate
(chloride, bicarbonate, nitrate), cannot be further exchanged. This
finding indicates that IEX can still be used for NOM removal even
with a high concentration of sulfate in the influent due to the
exchange between sulfate and the fraction of NOM with higher
affinity than sulfate.
Fig. 8 presents a simplified schematic overview of the ion ex-
change dynamics occurring during the long-term operation of the
 Z. Liu et al. / Water Research 168 (2020) 115148
Fig. 7. Normalized ammonia concentration in the effluents of the BIEX filter, three GAC filters and one BAC filter for the study period from April 11, 2018 to January 7, 2019.
Fig. 8. Displacement of NOM fractions in the BIEX filter a) virgin IEX; b) NOM3, sulfate
and NOM2 replace chloride while NOM1 is nonexchangeable; c) NOM3 and sulfate
replace NOM2 leading to the DOC release in the BIEX filter; d) NOM3 replaces sulfate,
which explains the long-term performance of NOM removal in the BIEX filter. N.B. The
anion on each band presents the dominant species but not the only one.
BIEX media (assuming that only NOM, sulfate and chloride are the
dominant solutes to consider). The affinity of each anionic species
towards IEX depends on two factors. On the one hand, the intrinsic
property of the anion molecule will affect its affinity for the anion
exchange resin. For example, several authors found that charge
density (Boyer et al., 2008; Finkbeiner et al., 2019), hydrophobicity
(Li and SenGupta, 1998) and molecular size (Bazri and Mohseni,
2016) can affect the affinity of NOM molecules. On the other
7
hand, the concentration of anions can also affect their competi-
tiveness, since the more molecules are present, the higher are the
chances that the species would occupy the ion exchange site. Given
that NOM is an assemblage of molecules with variable charge,
molecular weight and hydrophobicity, we divided NOM into three
different groups based on their affinities for IEX. The first fraction
NOM1 has no affinity for the resin (e.g., the unexchangeable frac-
tion) and is therefore expected to breakthrough at time 0. The
remaining NOM fractions are referred to as NOM2 and NOM3
which respectively describes the NOM fractions with either a lower
or higher affinity than sulfate. Initially, the resin is loaded with
chloride ions (Fig. 8a) which are exchanged for NOM3, sulfate and
NOM2 (Fig. 8b). Once the chloride capacity is exhausted (z1 000
BV in this study), NOM2 is displaced by sulfate (Fig. 8c). Meanwhile,
NOM3 continues to displace sulfate which explains the long-term
performance of the resin for NOM removal (Fig. 8d). The unex-
changeable NOM (NOM1) was 0.48 mg C/L (22% of the NOM in the
influent. The removal efficiency stabilized at about 29% (NOM3)
after the DOC release peak, which means that NOM2 accounted for
about 49% of the overall NOM in the influent. Even though NOM1
and NOM2 were no longer removable after the DOC breakthrough,
BIEX filter can still yield 29e36% DOC removal until the end of the
study. Such phenomenon had already been reported by Fu and
Symons (1990) who observed that an anion exchange filter was
still removing DOC after breakthrough due to displacement of
sulfate. Typically, IEX filters are regenerated near the sulfate
breakthrough to avoid organic matter leakage (Kim and Symons,
1991), an operation strategy requiring frequent regeneration and
thus producing a concentrated brine. In this study, the IEX filter
achieved a prolonged NOM removal after the sulfate breakthrough
mainly due to the exchange between sulfate and the NOM fraction
with higher affinity than sulfate. This performance was similar or
better than those of GAC/BAC filters for a period of at least 35 000
BV. Hence, we suggest (1) setting IEX filters off-line during a short
period (e.g. 800e900 BV) to avoid the NOM leakage; and (2)
operating IEX filters beyond the sulfate breakthrough to remove
NOM by secondary ion exchange as well as biodegradation to avoid
frequent regeneration (i.e. BIEX mode).
A mass balance was performed at 34 000 BV to confirm the
proposed mechanism of NOM removal according to equation (2).
8 Z. Liu et al. / Water Research 168 (2020) 115148
LNOM ¼ DOC  PNOM3  Vwater  CD (2)
where LNOM is the cumulative loading of NOM on the resin (eq/L
resin) due to the IEX mechanism. DOC is the average DOC con-
centration in the influent during the study (2.5 mg C/L). PNOM3 is
the percentage of NOM3 in the influent NOM (about 29%), a
fraction that can continually accumulate on the resin. Vwater is the
volume of water considered in the mass balance (34 000 BV). CD is
the charge density of NOM3 (9.0e10.6 meq/g C), a range esti-
mated based on Boyer et al. (2008) at pH 6.5 (transphilic acids
excluded). Consequently, the cumulative loading of NOM due to
the IEX mechanism at 34 000 BV is estimated to be 0.22e0.26 eq/L
resin. A similar calculation was performed for sulfate (28.2 mg/L),
which showed that the breakthrough of sulfate is expected at 1
362 BV, similar to the observed sulfate breakthrough during this
study (approx. 1 000 BV). However, we obtained a higher loading
value of NOM (0.33 eq/L resin) according to equation (1), a
calculation that includes all mechanisms of NOM removal in the
BIEX filter. Therefore, NOM removal in addition to IEX mechanism
is estimated to be 0.07e0.11 eq/L resin (i.e. 21e33% of the total
NOM removal), which is speculated to be due to the biomass
contribution (e.g. biosorption, biodegradation, bioregeneration).
However, more studies are still needed to elucidate the contri-
bution of biodegradation during the operation of IEX filters in
BIEX mode.
From our previous study (Amini et al., 2018), NOM1, NOM2
and NOM3 are respectively estimated to be 20%, 3% and 77% of
the overall influent NOM, a different breakdown compared to the
current study (22%, 49%, 29%) which is mainly due to the sulfate/
DOC concentration in the influents. In this study, we had a higher
sulfate concentration in the influent compared to Amini et al.
(2018) (28.2 mg/L vs. 5e8 mg/L), which leads to a higher pro-
portion of NOM2 and lower proportion of NOM3. This higher
proportion of NOM2 favored a net DOC release whereas the
lower proportion of NOM3 results in a lower NOM removal
performance after DOC breakthrough compared to our previous
study (29% vs. 62%). To conclude, higher sulfate concentrations in
the influent can increase the NOM2 proportion and decrease
NOM3 proportion to the overall influent NOM, which then
translates into a higher DOC breakthrough and lower NOM
removal efficiency after DOC breakthrough. Although we hy-
pothesize that NOM2/NOM3 ratios are mostly determined by the
competition (i.e. sulfate/DOC ratio), the NOM characteristics can
also play an important role. For example, Kim and Symons (1991)
found that the NOM fraction with <0.5 K MW was badly removed
throughout the column test (NOM1) while 0.5e1 K and
1e5 K MW fractions were the major organic fractions that surged
during DOC breakthrough (NOM2). However, more studies are
needed to elucidate to what extent other NOM characteristics
(charge density, hydrophobicity) also play a role in the sulfate/
NOM competition.
Evaluating the coupling of ozonation with BIEX was of interest
in our study. Firstly, ozonation fractionates higher MW organic
matter with a corresponding increase in lower MW organic matter
(Amy et al., 1988). This shall facilitate the diffusion of organic
matter into the resin pores and reduce the size-exclusion phe-
nomenon. In addition, ozonation also enhances NOM biodegrad-
ability (Hozalski et al., 1999), and thus should enhance
biodegradation on the BIEX media. However, IEX resins are
incompatible with dissolved ozone. In this study, we employed a
GAC filter with a short EBCT as a preventive measure to protect
BIEX. In a full-scale application, it would be more cost-effective to
quench the ozone residual chemically or plan for a longer reten-
tion time before entering into a BIEX filter in order to let the ozone
residual decay naturally.
4.3. Ammonia release in the BIEX filter
In our previous study (Amini et al., 2018), we found that BIEX
had a similar ammonia removal efficiency compared to BAC while
in this study we obtained results where the BIEX filter released 15%
ammonia in warm waters. Our results also demonstrate that this
ammonia release originated from the top 10 cm layer where most
NOM was removed (Fig. S2), and a thick biomass layer was
observed (Fig. S3). Also, during the biweekly backwash of the BIEX
filter, air injection could barely break down the solid crust of the
thick biomass layer. The crust broke into smaller IEX resin ag-
glomerates, a phenomenon mainly due to the insufficient air in-
jection rate and small column diameter. Further, the influent water
characteristics in the former study (untreated surface waters) may
also explain the successful nitrification obtained in 2018, since pH,
alkalinity, micronutrients (P, Cu etc.) were more favorable to nitri-
fication than in our present study.
5. Conclusion
This was the first pilot study evaluating the application of BIEX
as a second-stage filter within a drinking water treatment plant fed
with low DOC/high sulfate pretreated surface water. We highlight
the following findings:
 Despite the 50% DOC release observed at about 1 000 BV, BIEX
achieved similar (<8 000 BV) or higher performance (>8 000
BV) compared to GAC/BAC filters with 29e36% DOC removal.
 In the BIEX filter, after the rapid exhaustion of primary ion ex-
change capacity (<1 000 BV), sulfate replaced the fraction of
NOM with lower affinity than sulfate (NOM2) leading to the DOC
release.
 In the BIEX filter, the exchange between the fraction of NOM
with higher affinity than sulfate (NOM3) and sulfate is the
dominant mechanism to explain the long-term performance of
NOM removal.
 Within the top 30 cm layer (1.8 min EBCT), BIEX filter achieved
81% of the total DOC removal whereas GAC/BAC filters realized
only 62e66% at the same depth.
 BIEX released ammonia in warm water conditions with an
average of 15%, a phenomenon may due to the small diameter of
the column and the improper backwash.
Future studies will need to address the design conditions as well
as backwash strategies to properly wash a BIEX filter.
Declaration of competing interest
The authors declare that they have no known competing
financial interests or personal relationships that could have
appeared to influence the work reported in this paper.
Acknowledgement
The authors would like to thank Yves Fontaine and Mireille
Blais for their support in pilot plant installation. We also
acknowledge Julie Philibert, Jacinthe Mailly, Gabriel St-Jean for
the assistance of chemical analysis. We appreciate Sainte-Rose
drinking water treatment plant for their site support to the pilot
plant. Finally, we acknowledge the CREATE Program in environ-
mental decontamination technologies and integrated water and
wastewater management (TEDGIEER) for the Ph.D. Scholarship
awarded to Zhen Liu.
 Z. Liu et al. / Water Research 168 (2020) 115148
Appendix A. Supplementary data
Supplementary data to this article can be found online at
https://doi.org/10.1016/j.watres.2019.115148.
References
Amini, N., Papineau, I., Storck, V., Berube, P.R., Mohseni, M., Barbeau, B., 2018. Long-
term performance of biological ion exchange for the removal of natural organic
matter and ammonia from surface waters. Water Res. 146, 1e9.
Amy, G.L., Kuo, C., Sierka, R., 1988. Ozonation of Humic Substances: Effects on
Molecular Weight Distributions of Organic Carbon and Trihalomethane For-
mation Potential.
Ates, N., Incetan, F.B., 2013. Competition impact of sulfate on NOM removal by
anion-exchange resins in high-sulfate and low-SUVA waters. Ind. Eng. Chem.
Res. 52 (39), 14261e14269.
Ball, M., Harries, R.R., 1989. Resins for high-purity water production. J. Chem.
Technol. Biotechnol. 45 (2), 97e107.
Basu, O.D., Dhawan, S., Black, K., 2016. Applications of biofiltration in drinking water
treatmentea review. J. Chem. Technol. Biotechnol. 91 (3), 585e595.
Bazri, M.M., Mohseni, M., 2016. Impact of natural organic matter properties on the
kinetics of suspended ion exchange process. Water Res. 91, 147e155.
Boyer, T.H., Singer, P.C., Aiken, G.R., 2008. Removal of dissolved organic matter by
anion exchange: effect of dissolved organic matter properties. Environ. Sci.
Technol. 42 (19), 7431e7437.
Brezinski, K., Gorczyca, B., Sadrnourmohammadi, M., 2019. Ion-exchange for
trihalomethane control in potable water treatmentea municipal water treat-
ment case study in Rainy River, Ontario, Canada. Water Qual. Res. J. Can. 54 (2),
142e160.
Dixit, F., Barbeau, B., Mohseni, M., 2018. Simultaneous uptake of NOM and
Microcystin-LR by anion exchange resins: effect of inorganic ions and resin
regeneration. Chemosphere 192, 113e121.
Edzwald, J.K., 2010. Water Quality and Treatment A Handbook on Drinking Water.
McGrawHill.
Finkbeiner, P., Moore, G., Tseka, T., Nkambule, T., DeKock, L., Jefferson, B., Jarvis, P.,
2019. Interactions between organic model compounds and ion exchange resins.
Environ. Sci. Technol. 53 (16), 9734e9743.
Flemming, H.C., 1987. Microbial-growth on ion-exchangers. Water Res. 21 (7),
745e756.
Fu, P.L.K., Symons, J.M., 1990. Removing aquatic organic-substances by anion-
exchange resins. J. AWWA (Am. Water Works Assoc.) 82 (10), 70e77.
Gibert, O., Lefevre, B., Fern

andez, M., Bernat, X., Paraira, M., Pons, M., 2013. Frac-
tionation and removal of dissolved organic carbon in a full-scale granular
activated carbon filter used for drinking water production. Water Res. 47 (8),
2821e2829.
Hijnen, W., Schurer, R., Bahlman, J., Ketelaars, H., Italiaander, R., van der Wal, A., van
der Wielen, P.J.W.r., 2018. Slowly Biodegradable Organic Compounds Impact the
Biostability of Non-chlorinated Drinking Water Produced from Surface Water,
vol.129, pp. 240e251.
Hochmuller, K., 1984. Shock Test for the Determination of the Resistance of Ion
Exchange Resins to Osmotic and Hydromechanical Stress. Ellis Horwood, Chi-
chester, p. 472.
9
Hozalski, R.M., Bouwer, E.J., Goel, S., 1999. Removal of natural organic matter (NOM)
from drinking water supplies by ozone-biofiltration. Water Sci. Technol. 40 (9),
157e163.
Kawamura, S., 2000. Integrated Design and Operation of Water Treatment Facilities.
John Wiley & Sons.
Kennedy, M.D., Kamanyi, J., Heijman, B.G.J., Amy, G., 2008. Colloidal organic matter
fouling of UF membranes: role of NOM composition & size. Desalination 220
(1), 200e213.
Kim, H.-C., Dempsey, B.A., 2013. Membrane fouling due to alginate, SMP, EfOM,
humic acid, and NOM. J. Membr. Sci. 428, 190e197.
Kim, P.H.S., Symons, J.M., 1991. Using anion-exchange resins to remove thm pre-
cursors. J. AWWA (Am. Water Works Assoc.) 83 (12), 61e68.
Korth, A., Bendinger, B., Czekalla, C., Wichmann, K., 2001. Biodegradation of NOM in
Rapid Sand Filters for Removing Iron and Manganese, vol.29, pp. 289e295, 5.
Krasner, S.W., Weinberg, H.S., Richardson, S.D., Pastor, S.J., Chinn, R., Sclimenti, M.J.,
Onstad, G.D., Thruston Jr., A.D., 2006. Occurrence of a new generation of
disinfection byproducts. Environ. Sci. Technol. 40 (23), 7175e7185.
Levchuk, I., Rueda Marquez, J.J., Sillanpaa, M., 2018. Removal of natural organic
matter (NOM) from water by ion exchange - a review. Chemosphere 192,
90e104.
Li, P., SenGupta, A.K., 1998. Genesis of selectivity and reversibility for sorption of
synthetic aromatic anions onto polymeric sorbents. Environ. Sci. Technol. 32
(23), 3756e3766.
Matilainen, A., Vepsa €la
€inen, M., Sillanp€
aa€, M.J.A.i.c., science, i., 2010. Natural Organic
Matter Removal by Coagulation during Drinking Water Treatment: a Review,
vol.159, pp. 189e197, 2.
Richardson, S.D., Plewa, M.J., Wagner, E.D., Schoeny, R., DeMarini, D.M., 2007.
Occurrence, genotoxicity, and carcinogenicity of regulated and emerging
disinfection by-products in drinking water: a review and roadmap for research.
Mutat. Res. Rev. Mutat. Res. 636 (1), 178e242.
Rokicki, C.A., Boyer, T.H., 2011. Bicarbonate-form anion exchange: affinity, regen-
eration, and stoichiometry. Water Res. 45 (3), 1329e1337.
Schulz, M., Winter, J., Wray, H., Barbeau, B., Berube, P., 2017. Biologically active ion
exchange (BIEX) for NOM removal and membrane fouling prevention. Water
Sci. Technol. Water Supply 17 (4), 1178e1184.
Summers, R.S., Hooper, S.M., Shukairy, H.M., Solarik, G., Owen, D., 1996. Assessing
DBP yield: uniform formation conditions. J. Am. Water Work. Assoc. 88 (6),
80e93.
Tan, Y., Kilduff, J.E., 2007. Factors affecting selectivity during dissolved organic
matter removal by anion-exchange resins. Water Res. 41 (18), 4211e4221.
Terry, L.G., Summers, R.S., 2018. Biodegradable organic matter and rapid-rate bio-
filter performance: a review. Water Res. 128, 234e245.
Thurman, E.M., 2012. Organic Geochemistry of Natural Waters. Springer Science &
Business Media.
Velten, S., Knappe, D.R., Traber, J., Kaiser, H.-P., Von Gunten, U., Boller, M.,
Meylan, S.J.W.R., 2011. Characterization of Natural Organic Matter Adsorption in
Granular Activated Carbon Adsorbers, vol.45, pp. 3951e3959, 13.
Verdickt, L., Closset, W., D’Haeseleer, V., Cromphout, J., 2012. Applicability of ion
exchange for NOM removal from a sulfate-rich surface water incorporating full
reuse of the brine. Water Sci. Technol. Water Supply 12 (6), 878e887.
Winter, J., Wray, H.E., Schulz, M., Vortisch, R., Barbeau, B., Berube, P.R., 2018. The
impact of loading approach and biological activity on NOM removal by ion
exchange resins. Water Res. 134, 301e310.