The objectives of this pilot-scale study were to optimize backwash frequency and empty bed contact time (EBCT) of biofilters treating ozonated surface water from Lake Ontario. Performance was benchmarked in terms of the reduction of turbidity, dissolved organic carbon (DOC), disinfection by-product (DBP) precursors, and ultrafiltration foulants (biopolymers). Increasing the EBCT from 4 to 8 min resulted in a higher reduction of DOC (5%), trihalomethane (THM4) and haloacetic acid (HAA9) precursors (∼12%) without negatively impacting effluent turbidity (consistently below 0.4 NTU), while biopolymer removal remained unaffected (2%). The impact of varying backwash frequency (5, 10, and 25 day intervals) was also compared for biofilters operated at an EBCT of 4 min. Results showed no impact of extended run times (up to 25 days) on DOC or DBP precursor removal; however turbidity removal was affected beyond 15 days of operation. Backwashing biofilters at 10 vs 5 day intervals would result in a reduction of backwash water, energy consumption and amount to nearly $17,000 in savings for the utility.

INTRODUCTION

Natural organic matter (NOM) represents a heterogeneous mixture of organic compounds present in surface waters. A specific fraction, biopolymers, have been identified as the primary ultrafiltration (UF) membrane foulants (Tian et al. 2013). Biopolymers consist of proteins, protein-like substances, and polysaccharides, with a molecular weight (MW) of about 20,000 Da (Huber et al. 2011). Ozonation typically breaks down large NOM molecules (MW > 10,000 Da) into smaller compounds (MW < 300 Da) such as ketones, aldehydes and carboxylic acids, which are amenable to biodegradation (Treguer et al. 2010). Biofiltration is a treatment system whereby indigenous microbes are allowed to colonize granular filters, when operated without exposure to chlorinated water, and effectively consume biodegradable organic matter that passes through the filter (Huck et al. 1998; Hammes et al. 2011).

Biofiltration, without pretreatment, has been shown to effectively degrade biopolymers and reduce UF membrane fouling. Peldszus et al. (2012) reported a decrease in the hydraulically irreversible UF fouling rate from 6.33 kpa/h to 0.58 kpa/h when treating river water (dissolved organic carbon, DOC0 = 5.8–8.1 mg/L, biopolymer C0 = 0.41 ± 0.16 mg/L) using pilot-scale sand-anthracite biofilters (empty bed contact time, EBCT = 5 min). Other studies have also reported similar benefits of direct biofiltration as a pretreatment to UF membranes. Hallè et al. (2009) observed that direct biofiltration using sand-anthracite biofilters (EBCT = 5 min) reduced the maximum transmembrane pressure from 9 psi to 1.5 psi across UF membranes treating river water (total organic carbon, TOC0 = 5–7 mg/L).

An increase in biofilter EBCT has been shown to enhance overall NOM removal due to a longer period of interaction of the biofilm with organic matter. Hu et al. (2005) reported a 30% increase in DOC removal across pilot-scale biofilters when the EBCT was increased from 5 to 30 min (DOC0 = 5.1–10.7 mg/L), beyond which no improvement was observed. Other studies have suggested an optimal EBCT of 15 min for DOC and ozonation by-product removal (Yavich et al. 2004). In terms of specific NOM components, higher EBCTs (14 min vs 5 min) have been shown to enhance biopolymer degradation by up to 40% (0.09 mg/L) across pilot sand-anthracite biofilters and to reduce associated UF fouling (Hallè et al. 2009). Peldszus et al. (2012) reported an additional 35% biopolymer reduction when increasing the EBCT of a sand-anthracite biofilter from 5 to 15 min, and a corresponding decrease in hydraulically irreversible UF fouling by 50% (DOC0 = 5.8–8.1 mg/L, biopolymer C0 = 0.41 ± 0.16 mg/L). However, there is limited literature on the impact of EBCT for DOC and biopolymer removal across ozone-biofiltration systems. Siembida-Lösch et al. (2014) reported no impact of a range of EBCTs (4–19 min) on biopolymer removal across an ozone-biofiltration facility treating surface water, based on observations of full-scale biofilters (where EBCT varied with consumer demand). To complement NOM characterization by liquid chromatography-organic carbon detection (LC-OCD), fluorescence spectroscopy was also utilized to estimate the impact of biofiltration on protein-like and humic-like fractions. Fluorescence spectroscopy has shown promise as a rapid NOM characterization technique with potential for real time monitoring (Hudson et al. 2007).

Filter run time is often dictated by effluent turbidity when biofilters are used for particulate removal; in practice a backwash is initiated such that the effluent turbidity values do not exceed 0.3 NTU (typically at turbidities of 0.2 NTU; Emelko et al. 2006). Many studies have evaluated the impact of different backwash procedures including air scour, bed fluidization and collapse-pulsing conditions on biologically active filter performance and reported no impact on DOC removal (Ahmad et al. 1998; Liu et al. 2001; Emelko et al. 2006). However, Liu et al. (2012) observed that dissolved organic nitrogen (DON), normalized to the influent concentration, decreased from 0.98 to 0.90 post backwash, and increased to 1.5 following 8 h of operation for a full-scale granular activated carbon (GAC)-sand drinking water biofilter (EBCT = 15 min, DOC0 = 8.31 m/L, DON0 = 0.69 mg/L). This has implications for increased disinfection by-product (DBP) formation, as reported by Delatolla et al. (2015), where extending filter run time (24 to 69 h) increased THM4 and HAA9 formation potential by 25–30% (4–5 μg/L) across full-scale sand-anthracite biofilters (EBCT = 20 min, DOC0 = 3.1–3.3 mg/L). The above studies examined biofiltration following conventional treatment (sedimentation-flocculation), and prior to chlorination. Limited pilot or full-scale studies have examined optimal filter run times in terms of turbidity and NOM removal when biofiltration is considered alone, or in combination with ozone pretreatment.

The objectives of this pilot-scale study were to examine the impact of varying biofiltration EBCT on process performance, and also to investigate the effect of extending biofilter run times between backwashes for an ozone-biofiltration system. Operational parameters: backwash frequency and EBCT, were examined independently. Performance was monitored in terms of turbidity, DOC, UV254 absorbance (UVA), and DBP (THM4 and HAA9) precursor reduction; NOM fractionation using LC-OCD was used to measure biopolymer removal. EBCTs of 4 min and 8 min were compared, while the effect of backwash frequency (every 5, 10, and 25 days) was evaluated for biofilters operated at a 4 min EBCT, mimicking full-scale design operations at the site.

MATERIALS AND METHODS

Full-scale plant configuration

The Lakeview Water Treatment Plant is an Ozone-Biofiltration-Ultraviolet-UF Membrane (OBM) facility treating surface water from Lake Ontario, Canada. Intake prechlorination for zebra mussel control is practiced year round; any residual chlorine (0.5–1.0 mg/L) is quenched by the addition of sodium bisulfite. The plant utilizes biofiltration following ozonation (O3 dose = 1.0 ± 0.2 mg/L) to reduce biodegradable NOM (measured as DOC), as well as a pretreatment to UF. Full-scale biologically active carbon contactors (BACCs) have been designed to operate at an EBCT of 4 min at maximum plant capacity (600 MLD) and backwashed every 5 days with their unchlorinated effluents. The filtration velocity varied daily based on demand (7,000–55,600 L/min) and averaged 36,675 L/min during the study.

Pilot configuration and experimental design

The pilot consisted of four parallel filter columns made of acrylic (D = 7.62 cm (3 in)). Each filter contained 100 cm of media harvested from the BACCs of the full-scale facility. The media (Calgon Filtrasorb-300, effective size 1.3–1.5 mm, uniformity coefficient 1.4) had been in service for 20 months prior to commencement of the study, and the adsorptive capacity was assumed to be exhausted. The filters were fed from a constant head tank receiving ozonated water (post quenching of any residual). The filtration velocity of the pilot biofilters was adjusted daily to 1 L/min to maintain a 4 min EBCT and to 0.5 L/min for a 8 min EBCT. Pilot influent DOC varied between 2.3 and 2.7 mg/L during the study, while turbidity, pH and temperature ranged between 0.1–1.8 NTU, 7.4–8.2, and 1.7–16.8°C, respectively.

Backwash (BW) of the pilot biofilters included slow bed fluidization (30%) for 2 min followed by 8 min of collapse-pulsing conditions (50% bed fluidization), and gradually terminated with 2 min of slow backwash (30% bed fluidization), mimicking full-scale backwash procedure (3 min of low wash, 7 min of high wash, followed by 3 min of low wash). Collapse-pulsing conditions are effective for particle removal, while the slow backwash sequence towards the end has been shown to reduce peak turbidity and particle breakthrough during filter ripening (Slavik et al. 2013). Filters operated at EBCTs of 4 and 8 min during phase 1 were backwashed every 5 days (Figure 1). Subsequent phases (2 and 3) compared filters operated at an EBCT of 4 min and backwashed every 5 or 25 days. This ensured that the impacts of varying EBCT and backwash frequency were evaluated independently. Based on these initial results, it was identified that filter run times beyond 15 days were detrimental in terms of turbidity removal and subsequently, backwash frequencies were limited to a comparison of 5 vs 10 days.
Figure 1

Schematic of pilot. Filter configuration during each phase is shown. indicates sampling location.

Figure 1

Schematic of pilot. Filter configuration during each phase is shown. indicates sampling location.

Water quality parameters including pH, temperature, dissolved oxygen, UVA, DOC, and DBPs (THM4 and HAA9) under uniform formation condition (UFC) were measured on grab samples collected from the influent (post ozone and quenching), pilot and full-scale effluents (as shown in Figure 1), while flowrate and headloss were monitored using gauges affixed to the pilot biofilters. All samples were collected immediately prior to a backwash. As backwash typically restores filter performance, the period just before a backwash is generally that of least/limiting performance. Therefore, sampling right before a backwash indicates the maximum extent of deterioration during a filter run. LC-OCD analyses were performed on samples collected at the end of phases 1, 2, and 3. BACC effluent from the full-scale plant was also monitored. Samples were collected singly and all analyses were performed in duplicate. Active biomass was quantified using adenosine triphosphate (ATP) analysis performed prior to commencement and at the end of each phase on media obtained from the top 5 cm of the pilot and full-scale biofilters.

Analytical methods

DOC was measured using a wet oxidation method based on Standard Method 5310 D (Standard Methods 2012). The analysis was carried out using O-I Corporation Model 1010 Analytical TOC Analyzer with a Model 1051 Vial Multi-Sampler. Ultraviolet absorbance at 254 nm (UV254) was measured using a CE 3055 Single Beam Cecil UV/Visible Spectrophotometer (Cambridge, England) equipped with 1 cm quartz cells (Hewlett Packard, Mississauga). ATP analyses were conducted using a LuminUltra Deposit Surface Analysis kit (DSA-100C, Fredericton, New Brunswick) as per the manufacturer's instructions. LC-OCD analyses were conducted at the University of Waterloo (Waterloo, ON) according to the method described by Huber et al. (2011).

Fluorescence excitation-emission matrices (EEMs) were collected using an Agilent Cary Eclipse spectrofluorometer. Excitation wavelengths were iterated at 10 nm increments in a range of 200–380 nm and emission wavelengths at 1 nm increments between 250 and 600 nm. Instrument settings and sample processing were set based on in-house testing and previous studies (Peiris et al. 2009). Samples were collected at approximately 1 hour intervals for select sampling days for a total of 59 samples. A regional integration (RI) approach was taken to determine changes to protein-like and humic-like organic fractions, following the methodology described by Chen et al. (2003) using ±5 nm regions.

DBP formation (THM4, HAA9) was examined by chlorinating the pilot influent, biofilter and full-scale filter effluents under modified UFC. The target 24 h free chlorine residual was 1.5 ± 0.5 mg/L. Samples were incubated at 20 ± 2 °C for 24 h after which chlorine residuals were measured as described in Standard Method 4500-CI G (Standard Methods 2012) and quenched with ascorbic acid (100 mg/L). The pH was not adjusted to mimic full-scale operating conditions. The difference in the DBPs formed between the influent and biofilter effluents indicates the extent of precursor removal. Trihalomethanes (THMs) including chloroform (trichloromethane), bromodichloromethane, dibromochloromethane, and bromoform (tribromomethane) were analyzed using a liquid-liquid extraction gas chromatographic method based on Standard Method 6232 B (Standard Methods 2012). Haloacetic acids (HAAs) including monochloroacetic acid, monobromoacetic acid, dichloroacetic acid, trichloroacetic acid, bromochloroacetic acid, dibromoacetic acid, bromodichloroacetic acid, dibromochloroacetic acid, and tribromoacetic acid were analyzed using a liquid-liquid extraction gas chromatographic method based on Standard Method 6251 B (Standard Methods 2012). Both analyses were performed using a Hewlett Packard 5890 Series II Plus Gas Chromatograph (Agilent Technologies Canada Inc., Mississauga, ON) equipped with an electron capture detector (GC-ECD) and a DB 5.625 capillary column. Quality control was ensured by analyzing a sequentially prepared check standard after every ten samples according to Standard Method 1020 (Standard Methods 2012).

Statistical analysis

A two tailed paired t-test was used to evaluate the impact of varying the EBCT or filter run time on pilot biofilter performance, while all comparisons with respect to the full-scale BACCs were performed using a t-test (unpaired). All analyses were evaluated at a 95% significance level.

RESULTS AND DISCUSSION

Effect of EBCT

Significant (P < 0.05) improvements in DOC (4 ± 3%), THM4 (11 ± 9%) and HAA9 (11 ± 7%) precursor removal were observed (influent concentrations C0 = 2.6 ± 0.1 mg/L, 56.7 ± 8.6 μg/L and 16.0 ± 0.1 μg/L, respectively) at an EBCT of 8 min when compared to biofilters operated at a 4 min EBCT (Figure 2, Table 1). Longer retention time within biofilters enhanced the biodegradation of NOM, which contributed to lower DOC and DBP formation potentials. LeChevallier et al. (1992) reported a 6% increase in TOC removal (TOC0 = 3.9 mg/L) when the EBCT of an ozone-biofiltration system increased from 5 to 10 min, as well as an improvement in effluent turbidity (0.34 to 0.17 NTU, influent turbidity = 2.4–12.2 NTU), which was attributed to greater particle retention. No significant impact (P < 0.05) on turbidity was observed in this study, perhaps due to the consistently low influent values (0.35–1.21 NTU) to the pilot (Figure 2, Table 1).
Figure 2

Average organic carbon and turbidity removal (%) across pilot and full-scale biofilters (n = 5).

Figure 2

Average organic carbon and turbidity removal (%) across pilot and full-scale biofilters (n = 5).

Table 1

Impact of operational variables on pilot- and full-scale biofilter effluent quality

PhaseConfigurationTurbidity (NTU)ATP (ng/g)DOC (mg/L)UVA (cm−1)THM4 UFC (μg/L)HAA9 UFC (μg/L)
1. EBCT evaluation
(4 vs 8 min)
Backwash every 5 d 
EBCT 4 min 0.36 (0.07) 488 (40) 2.4 (0.1) 0.016 (0.002) 45.6 (6.7) 15.2 (1.1) 
EBCT 8 min 0.32 (0.09) 534 (31) 2.3 (0.1) 0.015 (0.002) 38.9 (4.7) 12.5 (1.2) 
Full-scale BACC 0.27 (0.04) 311 (11) 2.3 (0.1) 0.015 (0.001) 46.3 (8.5) 12.7 (1.4) 
2. Backwash frequency (5 d vs 25 d)
4 min EBCT 
Backwash every 5 d 0.17 (0.05) 334 (64) 2.4 (0.1) 0.014 (0.002) 31.2 (3.3) 7.0 (1.3) 
Backwash every 25 d 0.23 (0.02) 352 (78) 2.4 (0.1) 0.014 (0.001) 31.2 (3.3) 7.0 (1.3) 
Full-scale BACC 0.14 (0.02) 310 (28) 2.3 (0.1) 0.013 (0.001) 30.6 (2.6) 6.3 (1.0) 
3. Backwash frequency (5 d vs 10 d)
4 min EBCT 
Backwash every 5 d 0.28 (0.03) 324 (145) 2.5 (0.1) 0.014 (0.001) 38.4 (6.6) ̀13.4 (2.4) 
Backwash every 10 d 0.30 (0.04) 453 (185) 2.5 (0.1) 0.015 (0.002) 41.2 (6.6) 13.8 (1.8) 
Full-scale BACC 0.24 (0.03) 355 (28) 2.4 (0.1) 0.013 (0.001) 39.4 (8.1) 12.7 (1.8) 
PhaseConfigurationTurbidity (NTU)ATP (ng/g)DOC (mg/L)UVA (cm−1)THM4 UFC (μg/L)HAA9 UFC (μg/L)
1. EBCT evaluation
(4 vs 8 min)
Backwash every 5 d 
EBCT 4 min 0.36 (0.07) 488 (40) 2.4 (0.1) 0.016 (0.002) 45.6 (6.7) 15.2 (1.1) 
EBCT 8 min 0.32 (0.09) 534 (31) 2.3 (0.1) 0.015 (0.002) 38.9 (4.7) 12.5 (1.2) 
Full-scale BACC 0.27 (0.04) 311 (11) 2.3 (0.1) 0.015 (0.001) 46.3 (8.5) 12.7 (1.4) 
2. Backwash frequency (5 d vs 25 d)
4 min EBCT 
Backwash every 5 d 0.17 (0.05) 334 (64) 2.4 (0.1) 0.014 (0.002) 31.2 (3.3) 7.0 (1.3) 
Backwash every 25 d 0.23 (0.02) 352 (78) 2.4 (0.1) 0.014 (0.001) 31.2 (3.3) 7.0 (1.3) 
Full-scale BACC 0.14 (0.02) 310 (28) 2.3 (0.1) 0.013 (0.001) 30.6 (2.6) 6.3 (1.0) 
3. Backwash frequency (5 d vs 10 d)
4 min EBCT 
Backwash every 5 d 0.28 (0.03) 324 (145) 2.5 (0.1) 0.014 (0.001) 38.4 (6.6) ̀13.4 (2.4) 
Backwash every 10 d 0.30 (0.04) 453 (185) 2.5 (0.1) 0.015 (0.002) 41.2 (6.6) 13.8 (1.8) 
Full-scale BACC 0.24 (0.03) 355 (28) 2.4 (0.1) 0.013 (0.001) 39.4 (8.1) 12.7 (1.8) 

The values indicate actual concentrations in the effluent of filters. Standard deviation is shown in brackets. Bold values indicate statistically significant difference at a 5% significance level (α = 0.05) with respect to the control; n = 2 for ATP analysis; n = 5 for all other parameters.

No impact of EBCT on ATP was evident, as all pilot biofilters had similar concentrations throughout the study (439–534 ng/g), suggesting that these measurements are not suitable for predicting NOM degradation rates across biofilters (Table 1). Recent studies by Siembida-Lösch et al. (2014) and Pharand et al. (2014) have respectively shown ATP is a poor surrogate for biopolymer and DOC biodegradation, but rather that ATP is a useful method for estimating the density of the biomass; this highlights the need for a measurement that can reflect biofilter performance in terms of metabolic activity.

As a result of the varying daily flow rates across the full-scale BACCs due to consumer demand, EBCTs typically ranged between 4 and 10 min during the study (Figure 3). Full-scale filters performed similarly (P < 0.05) to the pilot biofilters operated at 8 min EBCT (Figure 2, Table 1), indicating that performance enhancement observed at full-scale resulted from the longer EBCTs.
Figure 3

EBCT of a full-scale biofilter (BACC#3) at the Lakeview WTP.

Figure 3

EBCT of a full-scale biofilter (BACC#3) at the Lakeview WTP.

NOM characterization using LC-OCD showed that biopolymers constituted approximately 16% of the DOC in the influent water to the biofilters (DOC0 = 2.3 mg/L; Table 2). Low biopolymer removal was observed (2%, C0 = 0.370 mg/L) across all biofilters, irrespective of the EBCT (Table 2). Siembida-Lösch et al. (2014) similarly reported no impact of EBCT (4–19 min) on biopolymer removal. McKie et al. (2015) also observed no biopolymer degradation in pilot GAC biofilters (EBCT = 16 min) treating Lake Ontario water; however significant removals (α = 0.05, 31–37%, biopolymer C0 = 0.14 mg/L) were achieved when in-line polyaluminum chloride (PACl) was added (0.8 mg/L) immediately prior to biofiltration. The biopolymer fraction in our study can be hypothesized to be refractory and not easily biodegradable under the current operating conditions. Others have reported that increasing the EBCT from 5 to 14–15 min may enhance biopolymer degradation by 35–40% (biopolymer C0 = 0.41 ± 0.16 mg/L) in GAC and sand-anthracite biofilters when treating river waters (Hallè et al. 2009; Peldszus et al. 2012). Inconsistent biopolymer degradation in the literature indicates that it is source water dependent and underscores the need for site-specific evaluations.

Table 2

Impact of operational variables on individual NOM fractions in pilot- and full-scale biofilter effluent

PhaseConfigurationBiopolymers (mg/L)Humic substances (mg/L)Building blocks (mg/L)LMW acids (mg/L)LMW neutrals (mg/L)Total DOC (mg/L)
1. EBCT evaluation
(4 vs 8 min)
BW every 5 d 
Influent 0.38 (16.2) 1.10 (46.9) 0.43 (18.3) 0.29 (12.5) 0.14 (6.1) 2.34 
EBCT 4 min 0.38 (16.2) 1.00 (43.5) 0.53 (23.1) 0.26 (11.4) 0.13 (5.6) 2.31 
EBCT 8 min 0.37 (16.6) 0.95 (41.6) 0.55 (23.9) 0.27 (11.7) 0.15 (6.5) 2.29 
Full-scale BACC NS NS NS NS NS NS 
2. Backwash frequency (5 d vs 25 d)
4 min EBCT 
Influent 0.36 (13.8) 1.18 (44.7) 0.59 (22.6) 0.31 (11.7) 0.19 (7.3) 2.63 
Backwash every 5 d 0.34 (13.5) 1.14 (45.6) 0.54 (21.7) 0.27 (10.7) 0.21 (8.4) 2.49 
Backwash every 25 d 0.36 (14.2) 1.19 (46.5) 0.64 (24.9) 0.26 (10.0) 0.14 (5.3) 2.56 
Full-scale BACC 0.35 (14.2) 1.10 (44.7) 0.59 (23.9) 0.25 (10.3) 0.17 (6.8) 2.45 
3. Backwash frequency (5 d vs 10 d)
4 min EBCT 
Influent 0.36 (13.8) 1.18 (44.7) 0.59 (22.6) 0.31 (11.7) 0.19 (7.3) 2.63 
Backwash every 5 d 0.34 (13.5) 1.14 (45.6) 0.54 (21.7) 0.27 (10.7) 0.21 (8.4) 2.49 
Backwash every 10 d 0.35 (13.7) 1.17 (46.1) 0.54 (21.3) 0.27 (10.7) 0.21 (8.1) 2.53 
Full-scale BACC 0.35 (14.2) 1.10 (44.7) 0.59 (23.9) 0.25 (10.3) 0.17 (6.8) 2.45 
PhaseConfigurationBiopolymers (mg/L)Humic substances (mg/L)Building blocks (mg/L)LMW acids (mg/L)LMW neutrals (mg/L)Total DOC (mg/L)
1. EBCT evaluation
(4 vs 8 min)
BW every 5 d 
Influent 0.38 (16.2) 1.10 (46.9) 0.43 (18.3) 0.29 (12.5) 0.14 (6.1) 2.34 
EBCT 4 min 0.38 (16.2) 1.00 (43.5) 0.53 (23.1) 0.26 (11.4) 0.13 (5.6) 2.31 
EBCT 8 min 0.37 (16.6) 0.95 (41.6) 0.55 (23.9) 0.27 (11.7) 0.15 (6.5) 2.29 
Full-scale BACC NS NS NS NS NS NS 
2. Backwash frequency (5 d vs 25 d)
4 min EBCT 
Influent 0.36 (13.8) 1.18 (44.7) 0.59 (22.6) 0.31 (11.7) 0.19 (7.3) 2.63 
Backwash every 5 d 0.34 (13.5) 1.14 (45.6) 0.54 (21.7) 0.27 (10.7) 0.21 (8.4) 2.49 
Backwash every 25 d 0.36 (14.2) 1.19 (46.5) 0.64 (24.9) 0.26 (10.0) 0.14 (5.3) 2.56 
Full-scale BACC 0.35 (14.2) 1.10 (44.7) 0.59 (23.9) 0.25 (10.3) 0.17 (6.8) 2.45 
3. Backwash frequency (5 d vs 10 d)
4 min EBCT 
Influent 0.36 (13.8) 1.18 (44.7) 0.59 (22.6) 0.31 (11.7) 0.19 (7.3) 2.63 
Backwash every 5 d 0.34 (13.5) 1.14 (45.6) 0.54 (21.7) 0.27 (10.7) 0.21 (8.4) 2.49 
Backwash every 10 d 0.35 (13.7) 1.17 (46.1) 0.54 (21.3) 0.27 (10.7) 0.21 (8.1) 2.53 
Full-scale BACC 0.35 (14.2) 1.10 (44.7) 0.59 (23.9) 0.25 (10.3) 0.17 (6.8) 2.45 

Fraction (%) of total DOC is indicated in brackets.

NS, not sampled.

Complementary to LC-OCD, organic matter characterization was also carried out using fluorescence spectroscopy. With EEMs, specific excitation/emission (ex/em) regions can be associated with fluorescence of protein-like substances (tryptophan at ex/em 280/340 nm) and humic-like substances (ex/em 340/450 nm) (Chen et al. 2003). Biofilters operated at an 8 min EBCT on average decreased both protein-like (26.9 ± 8.3%) and humic-like (32.3 ± 3.0%) fluorescence when compared to 4 min EBCT biofilters (n = 59). RI results are shown in Figure 4, demonstrating the increased reduction at higher EBCTs. Fluorescence results show an increased sensitivity to monitoring biofilter performance. It was expected that biopolymers, as analyzed by LC-OCD, and protein-like removals would be similar since the biopolymer fraction is thought to be comprised of proteins and polysaccharides (Huber et al. 2011). Probably the inclusion of polysaccharides in the LC-OCD biopolymer fraction obfuscates the relatively small changes to the protein fraction. Furthermore, it should be noted that protein-like fluorescence changes are inferred from the aromatic amino acid tryptophan. Therefore, it is possible that increased protein-like removal reflects the reduction of total proteins which contain tryptophan. Similarly, preferential removal of humic-like fluorescence was observed (Figure 4(b)), compared to overall DOC differences, due to EBCT (4 ± 3%). In combination with increased DBP precursor removal due to biofiltration, this is interpreted as a reduction in total humic substances, as well as the degree of aromaticity associated with this fraction. The role of aromatic moieties in the formation of THMs and HAAs has been reported by several authors (Ates et al. 2007; Lu et al. 2009).
Figure 4

Comparison of reduction of organic fluorophores post 4 and 8 min EBCT biofilters. Results in arbitrary units (AU) from RI of (a) protein-like (ex/em 280/340 nm) and (b) humic-like (ex/em 340/450 nm) fluorescence regions.

Figure 4

Comparison of reduction of organic fluorophores post 4 and 8 min EBCT biofilters. Results in arbitrary units (AU) from RI of (a) protein-like (ex/em 280/340 nm) and (b) humic-like (ex/em 340/450 nm) fluorescence regions.

It was observed that the variations among the significantly different parameters (DOC, THM4 and HAA9 precursors) were small (5–10%). In terms of concentration, the difference in DOC was 0.1 mg/L, while differences in DBP precursors were between 6 and 10 μg/L. This observation was consistent for all samples analyzed. Reductions that were significant when using a paired t-test are shown in Table 1.

Optimal filter run time - backwash every 5 vs 25 days

In phase 2, filters operated at a 4 min EBCT and backwashed every 5 days were directly compared to filters backwashed after 25 days (Figure 5). Turbidity and organic carbon removal were measured every 5 days immediately prior to a backwash. A decrease in backwash frequency from every 5 to 25 days did not impact DOC, UVA or DBP (THM4 and HAA9) precursor removal, as there was no significant difference (P > 0.05, Table 1). However, effluent turbidity increased following 15 days of operation and was similar (0.23 NTU) to the influent (0.22 NTU). Towards the end of the 25 day filter run, biofilter effluent turbidity exceeded the influent by 50% (0.14 NTU), indicative of particle breakthrough (Figure 5). As such, operation of biofilters beyond 15 days may have a detrimental impact on downstream UF membranes due poor turbidity removal and particle breakthrough.
Figure 5

Average headloss profile and effluent turbidity of filters backwashed every 5 and 25 days (d). Vertical bars represent standard deviation (n = 5).

Figure 5

Average headloss profile and effluent turbidity of filters backwashed every 5 and 25 days (d). Vertical bars represent standard deviation (n = 5).

Backwash every 5 vs 10 days

No significant impact (P > 0.05) on ATP concentrations at the beginning and end of phase 3 (50 days) was observed for biofilters backwashed at 5 or 10 day intervals (151–253 ng/g). Liao et al. (2015) reported a biomass density reduction of 54% (C0 = 33 nmol-P/g media) immediately following backwash due to microbial detachment, which recovered following two days of operation. Short term effects of backwash were not evaluated in this study; however backwash frequencies of 5 and 10 days were not observed to have long term impacts on ATP concentrations.

Average headloss between backwash cycles was higher (24 cm) at the end of a 10 day backwash cycle versus 5 days (13 cm, Figure 6). Ahmad et al. (1998) reported a similar relationship, where extending biofilter run times from 12 to 24 h increased headloss from 75 cm to 95 cm for a pilot-scale sand-anthracite biofilter (EBCT = 6 min), an observation attributed to greater particle retention. Interestingly, the rate of headloss development was similar (3 cm/d) for all the pilot biofilters, indicating that headloss development was gradual. Extending backwash frequency from 5 to 10 days did not present an operational challenge as the maximum headloss across any filter never exceeded 30 cm during the study (Figure 6).
Figure 6

Headloss profile of pilot biofilters backwashed every 5 and 10 days.

Figure 6

Headloss profile of pilot biofilters backwashed every 5 and 10 days.

Water quality impacts and biopolymer removal

DOC removal across the biofilters consistently ranged from 4 to 8% (0.1–0.2 mg/L), irrespective of backwash frequency. THMs formed under UFC (THM4 UFC) for filters backwashed every 5 vs 10 days were 38.4 ± 6.6 μg/L and 41.2 ± 6.6 μg/L respectively, while the HAA9 UFC was lower (13.8 ± 1.8 μg/L and 13.4 ± 2.4 μg/L respectively). Although Delatolla et al. (2015) reported a decrease in THM4 and HAA5 precursor removal (4–5 μg/L) on extending filter run times from 48 to 69 h in a full-scale sand-anthracite biofilter (EBCT = 20 min) receiving conventionally treated river water (DOC = 3.1 ± 0.1 mg/L), longer filter run times (10 days) did not significantly impact (P > 0.05) THM4 or HAA9 precursor removal across the pilot-scale biofilters in our study (Table 1). As such, biofilter run times for this system could be extended to 10 days without adversely affecting organic carbon removal.

DBPs formed under UFC were correlated to DOC, UVA, and specific ultraviolet absorbance (SUVA) to determine whether a surrogate to predict DBP formation for different backwash strategies could be identified. Poor correlations were observed between DOC and THM4 UFC (R2 = 0.52), HAA9 UFC (R2 < 0.1), similar to observations by Delatolla et al. (2015) and McKie et al. (2015). UVA was not an ideal surrogate, as poor correlations were observed with both THM4 and HAA9 UFC (R2 = 0.48 and 0.01, respectively). Although SUVA has been reported to be a better surrogate for the estimation of DBPs formed (Hua et al. 2015), similarly poor correlations were observed with THM4 and HAA9 in this study (R2 = 0.33 and 0.03, respectively). The magnitude of SUVA values associated with the influent and biofilter effluents (0.5–0.7 L/mg m) could explain the lack of relationship, as waters with low SUVA values (<2 L/mg m) have been shown to exhibit poor correlations due to the predominance of non UV-absorbing NOM molecules when considering DBP formation (Chow et al. 2006; Ates et al. 2007).

NOM fractionation using LC-OCD showed that biopolymers comprised approximately 14% of the pilot influent DOC (DOC0 = 2.6 mg/L); biofilters backwashed every 10 days had a similar biopolymer removal as those backwashed every 5 days (6%, biopolymer C0 = 0.36 mg/L; Table 2). Microbial degradation has been reported as the primary removal mechanism for biopolymer reduction, rather than physical filtration (Huang et al. 2011). Since there was no difference in either ATP or organic carbon removal, biopolymer concentrations were not expected to be impacted by backwash frequency; extending biofilter run times from 5 to 10 days had no impact on biopolymer removal. Graphical representation of LC-OCD fractionation has been provided in the supplementary information (Figure S1, available with the online version of this paper).

Economic benefits

Backwash costs estimations comprise three main elements: water utilized for backwash, operation and maintenance of equipment, and treatment of spent backwash water. The full-scale biofilters are currently backwashed with water treated with ozone, and the spent backwash water is subjected to coagulation-flocculation to remove particles before discharge. The costs associated with the above procedures were calculated from data obtained from the utility, and are attached in the supplementary information (available with the online version of this paper). The operation and maintenance costs were determined using design charts (USEPA 1979). The cost of performing a backwash every 5 days was calculated to be approximately $33,500 a year. Thus, a reduction of backwash frequency from every 5 to 10 days would result in annual savings of nearly $16,750.

CONCLUSIONS

This study examined the effects of EBCT and backwash frequency on organic carbon and turbidity removal in pilot-scale biofilters treating ozonated surface water from a large surface water source. Increasing EBCT improved DOC and DBP precursor removal, while there was no observed impact on biopolymer reduction, as such higher EBCTs may be required. Site specific evaluations are recommended.

Extending backwash frequency from 5 to 10 days did not have a detrimental impact on DOC, DBP precursor, turbidity or biopolymer removal. However, filter effluent turbidities increased significantly (0.1 NTU) when extending filter run times beyond 15 days, indicating a point of diminishing return. Increasing filter run time from 5 to 10 days would result in a 50% reduction in backwash costs, without any apparent impact on finished water quality or potential UF fouling.

ACKNOWLEDGEMENTS

This work was funded by the Natural Sciences and Engineering Research Council of Canada (NSERC) Chair in Drinking Water Research at the University of Toronto, and the Ontario Research Fund (ORF). The authors are grateful to Dr Monica Tudorancea and Dr Sigrid Peldszus (University of Waterloo) for performing LC-OCD analyses, and Jane Bonsteel, Iman Hashemi and colleagues at the Lakeview Water Treatment Plant for their assistance with the pilot plant operation and maintenance.

REFERENCES

REFERENCES
Ahmad
R.
Amirtharajah
A.
Al-Shawwa
A.
Huck
P. M.
1998
Effects of backwashing on biological filters
.
Journal American Water Works Association
90
(
12
),
62
73
.
Chen
W.
Westerhoff
P.
Leenheer
J. A.
Booksh
K.
2003
Fluorescence excitation-emission matrix regional integration to quantify spectra for dissolved organic matter
.
Environmental Science & Technology
37
,
5701
5710
.
Delatolla
R.
Séguin
C.
Springthorpe
S.
Gorman
E.
Campbell
A.
Douglas
I.
2015
Disinfection byproduct formation during biofiltration cycle: implications for drinking water production
.
Chemosphere
136
,
190
197
.
Emelko
M.
Huck
P.
Coffey
B.
Smith
F.
2006
Effects of media, backwash, and temperature on full-scale biological filtration
.
Journal American Water Works Association
98
(
12
),
61
73
.
Hallè
C.
Huck
P. M.
Peldszus
S.
Haberkamp
J.
Jekel
M.
2009
Assessing the performance of biological filtration as pretreatment to low pressure membranes for drinking water
.
Environmental Science & Technology
43
(
10
),
3878
3884
.
Hammes
F.
Velten
S.
Egli
T.
Juhna
T.
2011
Biotreatment of drinking water
. In:
Comprehensive Biotechnology
, (
Murray
M.
, ed.),
Academic Press
,
Burlington
, pp.
517
530
.
Hu
J. Y.
Song
L. F.
Ong
S. L.
Phua
E. T.
Ng
W. J.
2005
Biofiltration pretreatment for reverse osmosis (RO) membrane in a water reclamation system
.
Chemosphere
59
(
1
),
127
133
.
Huck
P. M.
Finch
G. R.
Hrudey
S. E.
Peppler
M. S.
1998
Design of biological processes for organics control
.
AWWARF report 90722
,
AWWA
,
USA
.
LeChevallier
M. W.
Becker
W. C.
Schorr
P.
Lee
R. G.
1992
Evaluating the performance of biologically active rapid filters
.
Journal American Water Works Association
84
(
4
),
136
146
.
Liao
X.
Chen
C.
Zhang
J.
Dai
Y.
Zhang
X.
Xie
S.
2015
Operational performance, biomass and microbial community structure: impacts of backwashing on drinking water biofilter
.
Environmental Science and Pollution Research
22
(
1
),
546
554
.
Liu
X.
Huck
P. M.
Slawson
R. M.
2001
Factors affecting drinking water biofiltration
.
Journal American Water Works Association
93
(
12
),
90
101
.
Liu
B.
Gu
L.
Yu
X.
Yu
G.
Zhang
H.
Xu
J.
2012
Dissolved organic nitrogen (DON) profile during backwashing cycle of drinking water biofiltration
.
Science of the Total Environment
414
,
508
514
.
McKie
M. J.
Taylor-Edmonds
L.
Andrews
R. C.
Andrews
S. A.
2015
Engineered biofiltration for the removal of disinfection by-product precursors and genotoxicity
.
Water Research
81
,
196
207
.
Peiris
R. H.
Budman
H.
Moresoli
C.
Legge
R. L.
2009
Acquiring reproducible fluorescence spectra of dissolved organic matter at very low concentrations
.
Water Science & Technology
60
,
1385
1392
.
Peldszus
S.
Benecke
J.
Jekel
M.
Huck
P. M.
2012
Direct biofiltration as pretreatment for fouling control of ultrafiltration membranes
.
Journal American Water Works Association
104
(
7
),
E430
E445
.
Pharand
L.
Dyke
Van M. I.
Anderson
W. B.
Huck
P. M.
2014
Assessment of biomass in drinking water biofilters by adenosine triphosphate
.
Journal American Water Works Association
106
(
10
),
433
444
.
Siembida-Lösch
B.
Anderson
W. B.
Wang
Y. M.
Bonsteel
J.
Huck
P. M.
2014
Effect of ozone on biopolymers in biofiltration and ultrafiltration processes
.
Water Research
70
,
224
234
.
Standard Methods for the Examination of Water and Wastewater
2012
22nd edn.
American Public Health Association/American Water Works Association/Water Environment Federation
,
Washington, DC, USA
.
United States Environmental Protection Agency
1979
Estimating water treatment costs Volume 2-Cost curves applicable to 1 to 200 mgd treatment plants
.
EPA-600/2-79-16b
.
Yavich
A. A.
Lee
K. H.
Chen
K.-C.
Pape
L.
Masten
S. J.
2004
Evaluation of biodegradability of NOM after ozonation
.
Water Research
38
(
12
),
2839
2846
.

Supplementary data