- Email: [email protected]
Biological ion exchange as an alternative to biological activated carbon for drinking water treatment
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 rube d, Zhen Liu a, b, *, Kim Maren Lompe b, Madjid Mohseni c, Pierre R. Be bastien Sauve a, Benoit Barbeau b Se Department of Chemistry, University of Montr eal, Montr eal, QC, H3T 1J4, Canada 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
b
a r t i c l e i n f o
a b s t r a c t
Article history: Received 11 August 2019 Received in revised form 24 September 2019 Accepted 1 October 2019 Available online 5 October 2019
Biological ion exchange (BIEX) has proved to remove natural organic matter (NOM) better than biological activated carbon (BAC). This raises the question if BIEX can be integrated into a full-scale drinking water treatment plant to remove NOM and ammonia. In this study, a pilot plant consisting of one BIEX filter, three GAC filters and one BAC filter was set up as second-stage filtration at the Sainte-Rose drinking water treatment plant (Laval, Canada). The pilot plant was operated for a period of nine months without 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 achieved a nearly constant removal of 29e36% over the whole study period. The DOC removals of GAC were similar to BIEX at < 8000 BV but then stabilized at 13e24% after 8 000 BV. Most DOC removal in the BIEX filter was achieved at the top 30 cm layer (81%) compared to 62e66% removal in the GAC/BAC filters in the same layer. After the rapid exhaustion of the primary ion exchange capacity (<1 000 BV), sulfate 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.
Keywords: Biological ion exchange Biological activated carbon Natural organic matter Anion exchange resin Nitrification
1. Introduction 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 byproducts (DBPs) (Krasner et al., 2006; Richardson et al., 2007; Brezinski et al., 2019) as well as biofilm growth in the distribution
al, * Corresponding author. Department of Chemistry, University of Montre al, QC, H3T 1J4, Canada. Montre E-mail addresses: [email protected] (Z. Liu), [email protected] (K.M. Lompe), [email protected] (M. Mohseni), [email protected] rube ), [email protected] (S. Sauve ), benoit.barbeau@polymtl. (P.R. Be ca (B. Barbeau). https://doi.org/10.1016/j.watres.2019.115148 0043-1354/© 2019 Elsevier Ltd. All rights reserved.
system (Hijnen et al., 2018). Water treatment plants (WTPs) 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
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, biofiltration 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 counterion (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 development of biofilm was proposed as the reason for this increased performance and the process was referred to as biological ion exchange 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 SO2 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 calculated 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 performance 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 phenomena. Regeneration produces a brine containing high concentrations of sodium chloride and NOM which has to be disposed of (Levchuk et al., 2018). Direct disposal of brine to the aquatic environment 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 processes 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 applications 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 alternative 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 SainteRose 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 chlorination. We set up pilot filters after ozonation, in parallel to the fullscale BAC filtration. The influent to the pilot showed low turbidity (z0.5 NTU) and low DOC (z2.5 mg C/L) (Table 1). The concentration 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 Turbidity DOC UVA254 SUVA pH Chloride Sulfate Alkalinity Ammonia
C NTU mg C/L cm1 L/mg,m e mg/L mg/L CaCO3 mg/L mg N/L
14.1 ± 9.2 ~0.5 2.5 ± 0.2 0.019 ± 0.007 ~0.76 6.5 ± 0.4 7.7 ± 4.4 28.2 ± 4.5 12.1 ± 7.7 78.0 ± 26.0
a
Average ± standard deviation.
Z. Liu et al. / Water Research 168 (2020) 115148
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 performed 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 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 accumulation. 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 residual 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 unchlorinated 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 backwash 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
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 (accumulation 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 accumulation 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
Table 2 Media characteristics for IEX, three different GAC and BAC used in this study. Abbreviations
Media
Diameter (mm)
D10 (mm)
Cu
Surface area (m2/g)
BIEX GAC1 GAC2 GAC3 BAC
Purolite® A860 Calgon Acticarbone® BGX Nuchar® WV-B30 Jacobi PICABIOL-HP120 Jacobi PICABIOL-2
0.3e1.2 0.5e0.7 0.8e1.1 1.2e1.4 1.2e1.4
0.5 0.6 1.0 1.3 1.3
1.7 1.8 1.8 1.4 1.4
2.19 1550e1650 1400e1600 1400 1600
Cu: Uniformity coefficient ¼ d60/d10; D10: measured with a MASTERSIZER 3000 (Malvern, UK). Other information was provided by suppliers.
4
Z. Liu et al. / Water Research 168 (2020) 115148
Fig. 2. Head loss accumulation for the BIEX, three GAC and the BAC filter after (i) 30 min of operation following a backwash (D1), (ii) seven days of operation (D7) and (iii) fourteen days of operation (D14).
UVA254 of the influent were 2.5 ± 0.2 mg C/L and 0.019 ± 0.007 cm1 (average ± standard deviation), respectively. We illustrated the normalized DOC concentration (Ceffluent/Cinfluent) and the normalized UVA254 absorbance (UVeffluent/UVinfluent) as a function of the number of bed volumes (BV) of filtered water (Fig. 3 & S1). Each day of operation was equivalent to about 133 BV. The
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 breakthrough, BIEX removed DOC at an average of 29% from 1000 to 8000 BV. During this period, all GAC filters achieved higher performances (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 performances 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 BAC filter achieved 13% removal. In summary, apart from the initial DOC release observed at about 1 000 BV, BIEX offered equivalent 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 showed no significant difference with regards to DOC removal (p > 0.05). With respect to UV254 absorbance (Fig. S1), removals fluctuated as they were directly impacted by the ozonation conditions preceding the pilot filters. The ozone residual concentration in 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
5
concentrations (±standard deviation) of sulfate, chloride, bicarbonate 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):
qði; jÞ ¼
j¼3 i¼34 X X i¼1 j¼1
Cin;i;j Cout;i;j Q 7d Vresin
(1)
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 describes 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 concentration 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 concentrations 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 concentration 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 chloride 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
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.
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.
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 backwash 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
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 reacclimating 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 treatment capacity after 32 weeks or 4 weeks of operation below 10 C. The GAC1, 2 and 3 filters also suffered from the decline of temperature 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. Overall, 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.
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 concentration 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 conventional 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 exchange dynamics occurring during the long-term operation of the
Z. Liu et al. / Water Research 168 (2020) 115148
7
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
hand, the concentration of anions can also affect their competitiveness, 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 fraction) 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 unexchangeable 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
LNOM ¼ DOC PNOM3 Vwater CD
Z. Liu et al. / Water Research 168 (2020) 115148
(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 concentration 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 estimated 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 contribution 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 proportion 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 hypothesize 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 phenomenon. In addition, ozonation also enhances NOM biodegradability (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 retention 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 agglomerates, a phenomenon mainly due to the insufficient air injection 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 nitrification 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 exchange 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 environmental 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. Longterm 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 Formation 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 treatment 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 anionexchange resins. J. AWWA (Am. Water Works Assoc.) 82 (10), 70e77. vre, B., Fern Gibert, O., Lefe andez, M., Bernat, X., Paraira, M., Pons, M., 2013. Fractionation 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, Chichester, 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 precursors. 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. €la €inen, M., Sillanp€ €, M.J.A.i.c., science, i., 2010. Natural Organic Matilainen, A., Vepsa aa 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, regeneration, 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 biofilter 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.
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 rube d, Zhen Liu a, b, *, Kim Maren Lompe b, Madjid Mohseni c, Pierre R. Be bastien Sauve a, Benoit Barbeau b Se Department of Chemistry, University of Montr eal, Montr eal, QC, H3T 1J4, Canada 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
b
a r t i c l e i n f o
a b s t r a c t
Article history: Received 11 August 2019 Received in revised form 24 September 2019 Accepted 1 October 2019 Available online 5 October 2019
Biological ion exchange (BIEX) has proved to remove natural organic matter (NOM) better than biological activated carbon (BAC). This raises the question if BIEX can be integrated into a full-scale drinking water treatment plant to remove NOM and ammonia. In this study, a pilot plant consisting of one BIEX filter, three GAC filters and one BAC filter was set up as second-stage filtration at the Sainte-Rose drinking water treatment plant (Laval, Canada). The pilot plant was operated for a period of nine months without 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 achieved a nearly constant removal of 29e36% over the whole study period. The DOC removals of GAC were similar to BIEX at < 8000 BV but then stabilized at 13e24% after 8 000 BV. Most DOC removal in the BIEX filter was achieved at the top 30 cm layer (81%) compared to 62e66% removal in the GAC/BAC filters in the same layer. After the rapid exhaustion of the primary ion exchange capacity (<1 000 BV), sulfate 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.
Keywords: Biological ion exchange Biological activated carbon Natural organic matter Anion exchange resin Nitrification
1. Introduction 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 byproducts (DBPs) (Krasner et al., 2006; Richardson et al., 2007; Brezinski et al., 2019) as well as biofilm growth in the distribution
al, * Corresponding author. Department of Chemistry, University of Montre al, QC, H3T 1J4, Canada. Montre E-mail addresses: [email protected] (Z. Liu), [email protected] (K.M. Lompe), [email protected] (M. Mohseni), [email protected] rube ), [email protected] (S. Sauve ), benoit.barbeau@polymtl. (P.R. Be ca (B. Barbeau). https://doi.org/10.1016/j.watres.2019.115148 0043-1354/© 2019 Elsevier Ltd. All rights reserved.
system (Hijnen et al., 2018). Water treatment plants (WTPs) 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
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, biofiltration 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 counterion (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 development of biofilm was proposed as the reason for this increased performance and the process was referred to as biological ion exchange 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 SO2 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 calculated 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 performance 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 phenomena. Regeneration produces a brine containing high concentrations of sodium chloride and NOM which has to be disposed of (Levchuk et al., 2018). Direct disposal of brine to the aquatic environment 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 processes 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 applications 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 alternative 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 SainteRose 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 chlorination. We set up pilot filters after ozonation, in parallel to the fullscale BAC filtration. The influent to the pilot showed low turbidity (z0.5 NTU) and low DOC (z2.5 mg C/L) (Table 1). The concentration 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 Turbidity DOC UVA254 SUVA pH Chloride Sulfate Alkalinity Ammonia
C NTU mg C/L cm1 L/mg,m e mg/L mg/L CaCO3 mg/L mg N/L
14.1 ± 9.2 ~0.5 2.5 ± 0.2 0.019 ± 0.007 ~0.76 6.5 ± 0.4 7.7 ± 4.4 28.2 ± 4.5 12.1 ± 7.7 78.0 ± 26.0
a
Average ± standard deviation.
Z. Liu et al. / Water Research 168 (2020) 115148
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 performed 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 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 accumulation. 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 residual 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 unchlorinated 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 backwash 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
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 (accumulation 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 accumulation 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
Table 2 Media characteristics for IEX, three different GAC and BAC used in this study. Abbreviations
Media
Diameter (mm)
D10 (mm)
Cu
Surface area (m2/g)
BIEX GAC1 GAC2 GAC3 BAC
Purolite® A860 Calgon Acticarbone® BGX Nuchar® WV-B30 Jacobi PICABIOL-HP120 Jacobi PICABIOL-2
0.3e1.2 0.5e0.7 0.8e1.1 1.2e1.4 1.2e1.4
0.5 0.6 1.0 1.3 1.3
1.7 1.8 1.8 1.4 1.4
2.19 1550e1650 1400e1600 1400 1600
Cu: Uniformity coefficient ¼ d60/d10; D10: measured with a MASTERSIZER 3000 (Malvern, UK). Other information was provided by suppliers.
4
Z. Liu et al. / Water Research 168 (2020) 115148
Fig. 2. Head loss accumulation for the BIEX, three GAC and the BAC filter after (i) 30 min of operation following a backwash (D1), (ii) seven days of operation (D7) and (iii) fourteen days of operation (D14).
UVA254 of the influent were 2.5 ± 0.2 mg C/L and 0.019 ± 0.007 cm1 (average ± standard deviation), respectively. We illustrated the normalized DOC concentration (Ceffluent/Cinfluent) and the normalized UVA254 absorbance (UVeffluent/UVinfluent) as a function of the number of bed volumes (BV) of filtered water (Fig. 3 & S1). Each day of operation was equivalent to about 133 BV. The
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 breakthrough, BIEX removed DOC at an average of 29% from 1000 to 8000 BV. During this period, all GAC filters achieved higher performances (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 performances 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 BAC filter achieved 13% removal. In summary, apart from the initial DOC release observed at about 1 000 BV, BIEX offered equivalent 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 showed no significant difference with regards to DOC removal (p > 0.05). With respect to UV254 absorbance (Fig. S1), removals fluctuated as they were directly impacted by the ozonation conditions preceding the pilot filters. The ozone residual concentration in 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
5
concentrations (±standard deviation) of sulfate, chloride, bicarbonate 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):
qði; jÞ ¼
j¼3 i¼34 X X i¼1 j¼1
Cin;i;j Cout;i;j Q 7d Vresin
(1)
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 describes 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 concentration 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 concentrations 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 concentration 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 chloride 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
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.
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.
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 backwash 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
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 reacclimating 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 treatment capacity after 32 weeks or 4 weeks of operation below 10 C. The GAC1, 2 and 3 filters also suffered from the decline of temperature 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. Overall, 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.
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 concentration 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 conventional 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 exchange dynamics occurring during the long-term operation of the
Z. Liu et al. / Water Research 168 (2020) 115148
7
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
hand, the concentration of anions can also affect their competitiveness, 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 fraction) 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 unexchangeable 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
LNOM ¼ DOC PNOM3 Vwater CD
Z. Liu et al. / Water Research 168 (2020) 115148
(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 concentration 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 estimated 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 contribution 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 proportion 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 hypothesize 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 phenomenon. In addition, ozonation also enhances NOM biodegradability (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 retention 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 agglomerates, a phenomenon mainly due to the insufficient air injection 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 nitrification 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 exchange 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 environmental 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. Longterm 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 Formation 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 treatment 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 anionexchange resins. J. AWWA (Am. Water Works Assoc.) 82 (10), 70e77. vre, B., Fern Gibert, O., Lefe andez, M., Bernat, X., Paraira, M., Pons, M., 2013. Fractionation 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, Chichester, 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 precursors. 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. €la €inen, M., Sillanp€ €, M.J.A.i.c., science, i., 2010. Natural Organic Matilainen, A., Vepsa aa 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, regeneration, 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 biofilter 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.