Escherichia coli removal in copper-zeolite-integrated stormwater biofilters: Effect of vegetation, operational time, intermittent drying weather

Ecological Engineering 90 (2016) 234–243

Contents lists available at ScienceDirect

Ecological Engineering journal homepage: www.elsevier.com/locate/ecoleng

Escherichia coli removal in copper-zeolite-integrated stormwater biofilters: Effect of vegetation, operational time, intermittent drying weather Yali Li a,b , David T. McCarthy a,b,∗ , Ana Deletic a,b,∗ a Environmental and Public Health Microbiology Lab (EPHM LAB), Monash Water for Liveability, Department of Civil Engineering, Monash University, Melbourne, Victoria 3800, Australia b Cooperative Research Centre for Water Sensitive Cities, Melbourne, Victoria 3800, Australia

a r t i c l e

i n f o

Article history: Received 2 March 2015 Received in revised form 17 December 2015 Accepted 26 January 2016 Keywords: Stormwater treatment Biofiltration Bacteria treatment Antimicrobial media Copper ion-exchanged zeolite

a b s t r a c t Existing biofiltration systems have shown variable and often inadequate bacterial removal efficacy. Previous work has shown antimicrobial media copper-zeolite as a promising alternative to reduce the variability and excessive discharge of faecal indicator bacteria such as Escherichia coli. A large-scale biofilter column study was conducted over eight months to investigate the benefits of incorporating copper-zeolite into biofilters on E. coli removal. The incorporation of copper-zeolite into biofilters improved E. coli log removal rate by 53% reducing E. coli concentration from 21,800 MPN/100 mL (median inflow) to 126 MPN/100 mL (median outflow) comparable to international primary contact recreational water quality. In addition, the E. coli removal performance of copper-zeolite amended biofilters increased after intermittent dry weather periods; this is notable, especially considering biofilter performance usually decreases after drying. Furthermore, these designs reduced inflow copper concentration by 91% (comparable to the metal removal performance of traditional biofilters) and provided a median effluent copper concentration of 8 ␮g/L. The vegetation in copper-zeolite filters survived. These results validate the use of copper-zeolite as bioretention media, particularly for sites requiring microbial reduction. Future research will include systematic investigation of the processes involved in reduction of bacteria in copper-zeolite filters and optimise filter design to augment the system performance to meet more stringent stormwater reuse requirements. © 2016 Elsevier B.V. All rights reserved.

1. Introduction Stormwater biofilters are gravity-fed filter beds, often located within an urban environment while creating an amenity feature (Dietz and Clausen, 2006; FAWB, 2009; Zinger et al., 2013). They remove faecal microbes mainly through sedimentation, straining, adsorption and natural die-off; however leaching of faecal microbes occurs through, for example, survival or regrowth in the media, resuspension, and desorption (Stevik et al., 2004; Zhang et al., 2011; Chandrasena et al., 2014a). Indeed, monitored systems range from providing net leaching to reasonable faecal microbe removal

∗ Corresponding authors at: Environmental and Public Health Microbiology Lab (EPHM LAB), Monash Water for Liveability, Department of Civil Engineering, Monash University, Melbourne, Victoria 3800, Australia. E-mail addresses: [email protected] (Y. Li), [email protected] (D.T. McCarthy), [email protected] (A. Deletic). http://dx.doi.org/10.1016/j.ecoleng.2016.01.066 0925-8574/© 2016 Elsevier B.V. All rights reserved.

capability (over 90%, with most being trapped in the top media layer); regardless of such, the effluent water quality rarely meets stormwater harvesting requirements (Rusciano and Obropta, 2007; Hathaway et al., 2009; NRMMC et al., 2009; Zhang et al., 2011; Chandrasena et al., 2012; Li et al., 2012). These findings suggest that both inadequate microbial removal capacity of traditional filter media and low die-off rate of bacteria trapped in the media contribute to the low and variable performance of existing biofilters. Therefore, recent research has been to investigate alternative filter media with more favourable surface properties for microbial removal (Mohanty et al., 2013; Mohanty and Boehm, 2014). Antimicrobial materials exert microbiocidal effects through contact with microbial solution or slow release of antimicrobial agents (Milan et al., 2001; Hrenovic et al., 2012). The utilisation of antimicrobial filter media to replace inert filter media has the potential to enhance microbial removal through improved adsorption and inactivation during filtration, and accelerated inactivation of trapped bacteria during dry periods. A recent study, which

Y. Li et al. / Ecological Engineering 90 (2016) 234–243

developed and evaluated a wide range of antimicrobial filter media, demonstrated that, Cu2+ - and Cu(OH)2 -treated media showed consistently effective 2 log removal of Escherichia coli from natural stormwater for over five months under typical stormwater operational conditions (Li et al., 2014a). The Cu2+ -exchanged zeolite was further calcined at 400 ◦ C or in situ Cu(OH)2 coated to prepare stable copper-zeolite (ZCu400 and ZCuCuO180, respectively) showing over 95% reduction in copper leaching (Li et al., 2014b). ZCu400 showed 2 log bacterial inactivation during a 24 h drying period, and ZCuCuO180 showed consistent 2 log E. coli removal during filtration. In the same study, laboratory trials using sand filter columns integrated with different layered configurations of copper-zeolite, showed that ZCu400 at the top and ZCuCuO180 in the middle provided the most effective E. coli removal (1.7 log). The research by Li et al. (2014a,b) was conducted in small, nonvegetated columns (diameter of 30 mm), under highly controlled experimental conditions, making it hardly possible to transfer these findings to field-scale biofilters. For example, water turbidity levels were not representative of natural stormwater, which could overestimate the true performance of the system; indeed, these experiments were unable to account for the effect that sediments have on shielding of the microbes, fouling of the filter media or the unsaturated conditions caused by surface clogging. Furthermore, the testing duration was just 8 days using a daily dosing regime, limiting our understanding of the long-term system stability and performance and whether the systems were resilient against strenuous drying conditions or seasonality effects. This paper presents, for the first time, a large-scale laboratory study on the development of stormwater biofilters that integrate copper-zeolite for effective E. coli treatment. Since this is the first attempt to develop such biofilters, the study focuses on understanding how key design components (i.e. vegetation, filter media type, saturated zone) and stormwater operational conditions (i.e. intermittent wetting/drying, seasonal variation, operational time) impact on E. coli removal. E. coli was selected as it is a commonly applied indicator for assessing overall water quality and is a conservative indicator for evaluating soil filters (VGG, 2003; NRMMC et al., 2009; Li et al., 2012). 2. Materials and methods 2.1. Preparation of copper zeolites Natural zeolite (0.1–0.6 mm; Escott Zeolite, Zeolite Australia) was used as the base medium. It was washed 10 times with 10 volumes of tap water and air dried before use. Preparation of copper-zeolite was conducted following procedures specified by Li et al. (2014b). In brief, natural zeolite was treated with NaCl to produce Na-zeolite, followed by treatment with CuSO4 solution to produce Cu2+ -exchanged zeolite. The Cu2+ -exchanged zeolite was then calcined at 400 ◦ C to prepare ZCu400, or in situ coated with Cu(OH)2 followed by heat treatment at 180 ◦ C to prepare ZCuCuO180. The modified media ZCu400 and ZCuCuO180 were analysed using inductively coupled plasma mass spectrometry (ICP-MS) to contain 10 mg Cu/g media, and 13 mg Cu/g media, respectively. 2.2. Experimental set-up Biofilters were constructed from 240 mm diameter PVC pipes 860 mm in depth, with a transparent Perspex top section 280 mm in depth allowing for plant growth and ponding of water (Table 1). The inner walls of the filters were sand blasted to minimise preferential flow effects. The filters were placed in a greenhouse constructed with a clear, impermeable roof.

235

The biofilter construction followed industry standards (FAWB, 2009) and previous biofilter studies (Blecken et al., 2009a; Bratieres et al., 2009; Li et al., 2012). In brief, the filters had 70 mm of coarse gravel at the base, including a slotted drainage pipe connected to a vertical riser pipe (Table 1b). The riser created a saturated zone (SZ) in the system 440 mm high with an outlet at the end. 70 mm of coarse sand lay on top of the gravel (as a transition layer), topped by 300 mm of triple-washed sand (0.075–0.6 mm, Daisy Garden Supplies, Melbourne) mixed with 270 g of carbon source (1/4 sugarcane mulch and 3/4 pine wood chips), and 400 mm of filter media layer which included the copper-zeolite layers (ZCu400 and ZCuCuO180, prepared as explained above). Two configurations of the 400 mm top filter media layer were tested (Table 1b): • ‘Layered’ – 50 mm ZCu400 at the top (found to be effective for inactivation while systems were resting during dry weather periods, and was therefore placed at the top where the majority of bacteria are concentrated; 100 mm washed sand for supporting vegetation (in which the top 50 mm was ameliorated with appropriate organic matter, fertiliser and trace elements as per Australian biofiltration design guidelines (Bratieres et al., 2009)); 50 mm of ZCuCuO180 (found to be efficient at bacteria removal during rain events); 50 mm of natural zeolite to adsorb heavy metals leached from the copper-zeolite media, and 150 mm of washed sand at the base to protect the underlying SZ. • ‘Mixed’ – 100 mm of ZCu400/ZCuCuO180 mixture in 1:1 ratio at the top; 50 mm of natural zeolite; and 250 mm of washed sand just above the SZ. Different vegetation designs were adopted in the cases of “layered” and “mixed” configurations, because the former was preferred configuration due to its superior E. coli removal performance in a feasibility study (Li et al., 2014b) while the latter was to assess the potential of retrofitting existing biofilters with antimicrobial media (by simply adding a new top layer). The ‘layered’ configuration was investigated (1) without vegetation (termed SCu) to assess the ‘reference characteristics’ of copper-zeolite media, (2) with Palmetto soft leaf buffalo (termed PBCu) and (3) with Leptospermum continentale (termed LCCu, one plant each column) to assess performance of antimicrobial media in the presence of vegetation. The same designs, replacing copper-zeolite with untreated natural zeolite, served as controls (termed S, PB, and LC). The ‘mixed’ configuration was planted with Leptospermum continentale (termed LCCu-T). The two selected plant types have shown effective removal of both nutrients (Payne et al., 2014) and E. coli (Chandrasena et al., 2014b) in traditional biofilters. In total, 7 types of biofilter designs (5 replicates each) were constructed. During construction, the filter media were added in segments of 75 mm, and each layer was smoothed and compacted by dropping a 3 kg weight once from 100 mm. The filters were then vegetated with plants established over 4 months within 220 mm diameter sand planter bags. Once constructed, all filters were subjected to 6 weeks of twice-weekly watering using dechlorinated tap water to establish the plants and achieve hydraulic compaction. 2.3. Experimental procedure 2.3.1. Dosing Synthetic stormwater was prepared in a continuously mixed, 1500 L tank using dechlorinated tap water and raw sewage collected from Pakenham Treatment Plant, sediment from a stormwater wetland (sieved through a 1000 ␮m sieve), and the chemicals listed in Table 2. The target concentrations described in Table 2 were matched to ‘typical’ worldwide and Australian urban stormwater quality characteristics (Duncan, 1999; NRMMC et al., 2009; McCarthy et al., 2012).

236

Y. Li et al. / Ecological Engineering 90 (2016) 234–243 Table 1 Experimental set-up and operational conditions.

Table 2 Synthetic stormwater quality and pollutant sources. Parameter

Median (min, max)

Temperature (◦ C) TSS (mg/L) TP (mg/L) TN (mg/L) pH EC (␮S/cm) Pb (␮g/L) Zn (␮g/L) Cu (␮g/L) Mn (␮g/L) Cd (␮g/L) Cr (␮g/L) E. coli (MPN/100 mL)

15 (9, 30) 74 (47, 190) 0.40 (0.31, 0.48) 2.5 (2.2, 3.4) 250 (200-269) 180 (80, 420) 270 (180, 350) 76 (62, 180) 185 (99, 230) 9.8 (5.5, 13) 39 (15, 60) 21,800 (10,100, 41,000)

Targeta – median (5th, 95th)

Source

77 (19, 254) 0.37 (0.08, 1.26) 2.5 (0.6, 7.5) 6.3 (5.5, 7.3) 250 60 (20, 160) 250 (80, 570) 40 (10, 140) 103 (50, 200) 13 (1, 60) 8 (2, 17) 38,000 (3500, 185,000)

Stormwater wetland sediment KH2 PO4 Raw sewage, sediment, KNO3 , NH4 Cl, C6 H5 O2 N NaCl Pb(NO3 )2 ZnCl2 CuSO4 ·5H2 O Mn(NO3 )2 ·4H2 O Standard solution from Merck Cr(NO3 )3 ·9H2 O Raw sewage from Pakenham Treatment Plant, Melbourne

a Urban stormwater quality taken from (NRMMC et al., 2009); EC of 250 ␮S/cm represented 75th percentile of EC in urban runoff monitored in 45 samples at 6 catchments in Melbourne (unpublished data).

The filters were dosed with the synthetic stormwater following the frequency and volume specified in Table 1c. Under normal operations, the filters were dosed twice weekly to mimic Melbourne’s long-term climatic pattern (Bratieres et al., 2008). During a ‘dosing event’, each filter received 13 L from the continuously mixed tank; equated with that typically seen for a stormwater biofilter sized to 2% of its impervious catchment area in Melbourne (as per best design practice (FAWB, 2009)). To achieve consistent input concentrations for each filter, the target volume was applied in ‘pulses’; i.e. during a ‘dosing event’, each filter received roughly three ‘pulses’ of water from the tank. To simulate the extreme dry weather periods that Melbourne experiences, the filters were left without any water for periods of between 1 and 4 weeks (drying weather periods – see Table 1c).

2.3.2. Sampling 42 dosings of synthetic stormwater were applied to each filter over the course of eight months; 19 of which were sampled and analysed (Table 1c). Four types of sampling were conducted to test the impact of different biofilter operational conditions, as per previous studies (Li et al., 2012; Chandrasena et al., 2014b): (1) 10 ‘typical sampling events’, wherein, on one of the two weekly dosing days during normal operation, the influent was increased to 20 L of synthetic stormwater (i.e. a 10 mm rainfall event), (2) four ‘dry weather sampling events’, wherein after a period of drying the filters were dosed with 20 L of synthetic stormwater, (3) Three ‘challenging events’, wherein, on one of the two weekly dosing days during normal operation, the influent was increased to 40 L of synthetic stormwater (i.e. a 19 mm rainfall event, or 1

Y. Li et al. / Ecological Engineering 90 (2016) 234–243

in 3 month return period), and (4) two ‘rinsing events’, wherein 20 L of sewage-free stormwater was dosed into each filter 24 h after the last two ‘challenging events’. The latter two sampling types fall outside the scope of this study; indeed, they focus on analysing the effluent of the biofilters for many other parameters than E. coli (including other microorganisms, nutrients, heavy metals, etc.). As such, this paper focuses on the first two sampling types. During each sampling event, duplicate composite inflow samples were collected in 5 L jars, while the entire effluent from each filter was collected into 20 L barrels. When the effluent volume reached the 10 L mark, the barrel was mixed and a 100 mL subsample was immediately collected and labelled ‘SZ water’ (antecedent water in the SZ zone which was estimated to be approximately 10 L based on a tracer study (Chandrasena, 2014)). When filters ceased draining, the barrel was again mixed and another subsample was taken and labelled ‘composite water’. Starting from week 14, three LC filters had very low infiltration rates and hence their entire outflow could not be fully sampled. In these cases, samples were collected at the end of the day while recording the total volume drained. To study E. coli effluent concentration changes during events, discrete samples (at 2 L intervals) were taken from designs SCu, LC, LCCu (only two replicates of each design were chosen due to limited resources) over four sampling events on sampling days, immediately before and after the fortnightly and the last monthly drying periods, and analysed for E. coli. All inflow, outflow and discrete samples were analysed using the ColilertTM method (IDEXX-Laboratories, 2007) for E. coli concentration. Inflow samples and some outflow samples from copper-zeolite filters were also analysed for total copper concentration using ICP-MS in a NATA-accredited lab. To monitor temporal change in the infiltration rate, infiltration rate measurements were conducted during every sampling event. As documented earlier (Chandrasena et al., 2014b), this was conducted after applying all the stormwater to the systems by measuring the average drop in pond water level over approximately 30 min at nine time points. The recorded ponding depth was plotted against time and the gradient of the graph was taken as the average infiltration rate through the filter.

2.4. Data analysis For analysis, the sampling events were divided into three stages based on the maturity of the systems, which also spanned multiple seasons: Stage 1: establishment stage (average temperature 26 ◦ C); Stage 2: established stage (average temperature 18 ◦ C); and Stage 3: mature stage (average temperature 10 ◦ C) (Table 1c). Where outflow concentration was lower than the detection limit, half of the detection limit was taken as the concentration for statistical analysis. To assess E. coli removal performance, log removal rates were calculated for all designs and operating regimes based on log concentration differences between inflow and outflow samples. Both log removal rate and outflow concentrations were compared with Australian stormwater harvesting requirements for unrestricted irrigation (E. coli median removal >1.5 log; median effluent concentration <10 MPN/100 mL) (NRMMC et al., 2009) and recreational water use (primary contact median ≤150 E. coli/100 mL and secondary contact median ≤1000 E. coli/100 mL) (VGG, 2003). All E. coli log removal data were checked for normality using the Shapiro–Wilk test. Univariate analysis of variance, along with post-hoc tests (Tukey HSD), were performed to test the significant impact of filter media type,

237

vegetation, intermittent drying periods and operational time on E. coli removal.

3. Results and discussion 3.1. Overall performance and comparison to targets Table 3 summarises the performance of the tested designs regarding E. coli and copper removal and hydraulic permeability, as well as relevant Australian stormwater reuse guidelines. Noncopper-zeolite amended filters had median log removals above the Australian unrestricted irrigation targets (1.5 log; NRMMC et al., 2009), yet these designs failed to meet this target consistently, with 5th percentiles being below 1.3 log removal. The median effluent E. coli concentrations met the Victoria Government Gazette S107 specified secondary contact recreational water quality targets (median ≤1000 E. coli/100 mL) (VGG, 2003), but failed to meet the primary contact recreation and the unrestricted irrigation targets. These findings are consistent with conclusions from recent traditional biofilter studies (Rusciano and Obropta, 2007; Hathaway et al., 2009; Zhang et al., 2011; Chandrasena et al., 2012; Li et al., 2012). In contrast, the copper-zeolite amended filters met the Australian unrestricted irrigation standard for E. coli removal (>1.5 log), with all filters meeting this target more than 95% of the time (Table 3). However, these systems were unable to consistently meet the concentration guideline for unrestricted irrigation (<10 E. coli/100 mL) (NRMMC et al., 2009). Nonetheless, median E. coli effluent concentrations from designs SCu and LCCu were below the requirement for both primary and secondary contact recreation (≤150 E. coli/100 mL) (VGG, 2003). Copper removal by copper-zeolite filters ranged between 76% and 91% (Table 3), which is consistent with the reported removal by traditional biofilters (Hunt et al., 2008; Feng et al., 2012). The effluent copper concentrations (median of 8–16 ␮g/L) were well below Australian drinking water guidelines and long term irrigation guidelines (NRMMC–EPHC–NHMRC, 2008; NRMMC et al., 2009), as expected since raw stormwater concentrations (and hence inflow concentrations) are typically below these guidelines. The standards for aquatic ecosystem protection (ANZECC/ARMCANZ, 2000) were not met by any of the tested systems, as expected, since these standards are not currently achievable by any Water Sensitive Urban Design system in use (Blecken et al., 2009b). These results confirmed that the copper coating was stable (i.e. removal was observed with no leaching) over the operational conditions tested. Since the effluent copper concentrations from copper-zeolite filters were very low, post bacteria inactivation in effluent was not expected (an experimental problem discussed by Li et al. (2014a)). This hypothesis was reinforced by the absence of a positive correlation between E. coli removal and effluent copper concentrations (linear regression coefficient < 0). It was also confirmed by a separate batch test, where the effluent from LCCu filters was spiked with raw sewage with and without ethylenediaminetetraacetic acid disodium salt, and the E. coil concentration was continuously monitored over 4 h. In all cases, no obvious change in E. coli concentration was observed (data not shown). Marked reductions in infiltration rates were observed (Table 3), with some designs showing greater reduction than others: 88% reduction for LC filters, 47% reduction for PBCu filters and 59–71% reduction for other designs. Nevertheless, with the exception of three LC filters, infiltration rates were within biofilter guidelines and comparable to field measurements (FAWB, 2009; Le Coustumer et al., 2009). The observed change in infiltration rate may affect the

238

Table 3 Performance of all biofilter designs over eight months presented as median value with 5th, 95th percentiles in parentheses. Inflow

E. coli 21,800 (10,100, 41,000) (MPN/100 mL)

Outflow

No of samples

a

Media a None None None Cu-Z Cu-Z Cu-Z Cu-Z

Plant b None Buffalo Lepto None Buffalo Lepto Lepto

Layout layered layered layered layered layered layered mixed

59 57 51c 60 58 60 60

Composite water Concentration (MPN/100 mL)

210 (13, 1306) 105 (22, 579) 79 (19, 313) 6 (1, 84) 11 (1, 53) 15 (2, 106) 20 (6, 148)

809 (228, 1860) 624 (225, 1516) 331 (69, 981) 126 (14, 599) 183 (62, 1050) 117 (20, 593) 198 (40, 907)

<10 (median)

≤150 (median)

Log removal

1.5 (1.1, 2.0) 1.6 (1.2, 1.9) 1.9 (1.3, 2.3) 2.3 (1.6, 3.2) 2.1 (1.5, 2.5) 2.3 (1.6, 3.1) 2.1 (1.5, 2.8)

>1.5

No of Samples

46 42 46 45

Composite water Concentration (␮g/L)

8 (2,47) 14 (3, 54) 16 (4, 59) 17 (6, 59) 2000

Infiltration rate (IR) – Stages 2 and 3 Removal (%)

IR (mm/h) Week 7

IR (mm/h) Week 35

Reduction (%)

91 (52, 98) 84 (30, 97) 81 (41, 97) 76 (30, 95)

438 354 349 425 296 353 484

170 116 42 149 156 146 140

61 67 88 65 47 59 71

200 (95th)

1000

≤1000 (median)

1.4 (95th)

None, non-copper-zeolite; Cu-Z, copper-zeolite. None, non-vegetated; Buffalo, Palmetto soft leaf buffalo; Lepto, Leptospermum continentale. c Less LC samples than other designs due to slow infiltration rates of some replicates. By end of the sampling day, if the samples were below 14 L (targeting for 18 L), they were analysed for E. coli concentrations but were excluded from statistical analysis. b

Y. Li et al. / Ecological Engineering 90 (2016) 234–243

Design S PB LC SCu PBCu LCCu LCCu-T Drinking water guideline (NRMMC–EPHC–NHMRC, 2008) Unrestricted irrigation – Public, open space irrigation (no access control) guideline (NRMMC et al., 2009) Primary contact recreational water use guideline (ANZECC/ARMCANZ, 2000; VGG, 2003) Secondary contact recreational water use guideline (ANZECC/ARMCANZ, 2000; VGG, 2003) Fresh water ecosystems protection guideline (ANZECC/ARMCANZ, 2000)

SZ water Concentration (MPN/100 mL)

Cu 76 (62, 180) (␮g/L)

Y. Li et al. / Ecological Engineering 90 (2016) 234–243

239

Table 4 Post-hoc test of 7 filter designs for E. coli removal (if p < 0.05, the difference between two groups is considered significant). p value S S PB LC SCu PBCu LCCu LCCu-T

–

PB 0.127 –

LC

SCu

PBCu

LCCu

LCCu-T

<0.001 <0.001 –

<0.001 <0.001 <0.001 –

<0.001 <0.001 0.001 <0.001 –

<0.001 <0.001 <0.001 0.998 <0.001 –

<0.001 <0.001 <0.001 <0.001 1.000 <0.001 –

Fig. 1. Median and 95% confidence interval of E. coli logarithmic reduction rate over time (each point represents 5 measurements; shaded area indicates drying periods; WD, weeks drying).

removal performance of the systems, as found by Chandrasena et al. (2014b) and discussed below. 3.2. Effect of design factors on treatment performance Significant differences between designs (combination of vegetation and filter media type) were analysed using one-way ANOVA post-hoc tests (Table 4). The impact of filter media type was examined by comparing non-vegetated filters SCu and S through overall performance (Tables 3 and 4) as well as temporal performance evolution (Fig. 1). The impact of vegetation was analysed for noncopper-zeolite filters and copper-zeolite filters separately. 3.2.1. Filter media type As highlighted above, filter media type (copper-zeolite, noncopper-zeolite) was found to be a significant factor for E. coli removal (p < 0.001; univariate). Filters with copper-zeolite (SCu) performed significantly better than non-copper-zeolite filters (S) (Tables 3 and 4). The E. coli removal performance of both SCu and S differed over various operational regimes and time (Fig. 1), yet a significant difference persisted between SCu and S over the entire experimental duration. It was speculated that improved removal by SCu filters was achieved by enhancing contact-based inactivation, adsorption and survival/growth inhibition by copperzeolite. Specifically, ZCu400 placed at the top of SCu filters inactivated the majority of trapped bacteria between events, and ZCuCuO180 placed in the middle of the filters removed released/desorbed bacteria from the top layers during dosing events. In comparison of LCCu-T with LCCu designs, the ‘layered’ design LCCu performed better than the ‘mixed’ design LCCu-T (Tables 3 and 4). This observation highlighted the importance of having ZCuCuO180 in the middle as post-treatment during events. However, the ‘mixed’ system LCCu-T as retrofitting design, should

not be dismissed, partly because of its significantly improved performance in comparison with the traditional non-copper-zeolite design (LC) (p < 0.001, Table 4) and more so because it is relatively simple and practical compared to the ‘layered’ system. 3.2.2. Vegetation Vegetation was a significant factor (p < 0.001; univariate) in bacterial removal. As shown in Tables 3 and 4, Leptospermum continentale in non-copper-zeolite filters (LC) effected higher removal than filters without vegetation (S) (p < 0.001). However, Palmetto soft leaf buffalo did not exert such positive effect, and PB filters showed similar removal performance to S filters (p = 0.127, post-hoc). The fine and dense roots of Leptospermum continentale may explain or contribute to these observations. For example, the roots may provide extra surface area for adsorption, contribute to decreased porosity along with increased dispersion, thus enhanced straining, and benefit uptake of E. coli through colonisation or predation around the rhizosphere (Bitton and Marshall, 1980; Mukerji et al., 2006; Rusciano and Obropta, 2007). It is also noted that the superior removal performance of LC filters was accompanied by the lowest infiltration rate (IR) (Table 3). The correlation between IR and removal rate was performed for events during Stages 2 and 3 only, when systems were established (Fig. 2). A significant negative linear correlation was identified for LC filters (p < 0.001; linear regression analysis). In systems with low filtration rate, extended detention occurs, benefiting straining, adsorption and die-off, while minimising remobilisation and desorption (Bitton and Marshall, 1980; Stevik et al., 2004; Hathaway et al., 2011). However, systems with low IR are not always desirable due to their low hydraulic efficiency and therefore overall low load reduction rates (Le Coustumer et al., 2009). Nevertheless, the low infiltration rate might not be the only factor contributing to performance. Sand filters (S filters), for example, also showed a wide range of IR, but no significant change in performance (Fig. 2). The root exudates, particularly

240

Y. Li et al. / Ecological Engineering 90 (2016) 234–243

Fig. 2. Linear correlation between logarithmic reduction rates and infiltration rates (IRs) during Stages 2 and 3.

by Leptospermum continentale which is known to have antimicrobial attributes (Brix, 1997), may contribute to E. coli die-off, thus improving removal rates. The advantages of Leptospermum continentale were not as prominent in copper-zeolite filters (LCCu) which performed in a similar manner to SCu filters (p = 0.998, post-hoc). On the other hand, the presence of Palmetto soft leaf buffalo in PBCu filters induced reduced performance in comparison with LCCu filters and SCu filters (Tables 3 and 4). These results accorded with the significant interactive effect detected between media type and plants (p < 0.001; univariate), and confirmed that vegetation was not an important factor for all media types. Although Leptospermum continentale and Palmetto soft leaf buffalo survived in copper-zeolite, their growth was quite stressed (Fig. 3).

3.2.3. Saturated zone The E. coli concentration in SZ water was significantly lower than in the composite outflow water for all designs (Table 3), indicating the role of die-off and inactivation of E. coli in this SZ water (which remained in the bottom of the filters for 2–3 days during normal operational conditions and 7–28 days during dry weather periods). This aligns with findings from studies of traditional biofiltration systems (Chandrasena et al., 2014b), and indicates that SZ is an important design element (Dietz and Clausen, 2006; Blecken et al., 2009a). However, SZ E. coli concentrations varied significantly between designs, with the presence of copperzeolite playing an important role in reducing E. coli concentrations in SZ zone (Table 3). For example, median E. coli concentration in SZ water of SCu filters was 6 MPN/100 mL whereas that of S filters was 210 MPN/100 mL. This is an important indicator of the benefits of using antimicrobial media, and confirms the above speculation regarding the role of the antibacterial media. Importantly, the SZ was initially designed for nutrient removal, particularly for denitrification (Zinger et al., 2013), and our tests indicated that the level of copper in SZ did not influence the beneficial microbial communities for denitrification, as nitrogen was still well removed (average > 50%, data not shown).

3.3. Effect of operational conditions on treatment performance Fig. 4 shows the E. coli log removal rates as well as infiltration rates of seven biofilter designs, for three stages of monitoring and two types of sampling event. Both operational time and intermittent dry weather periods affected E. coli removal significantly (p < 0.001, univariate).

3.3.1. Operational time – normal operational conditions S filters showed unexpectedly reduced E. coli removal over time (Fig. 4) accompanied by an increase in E. coli concentrations in both SZ water and new water (data not shown). This observation seems contrary to previous studies of a mature system for bacteria removal (Rusciano and Obropta, 2007). Speculations on the difference observed in this study are mixed effects by decreases in temperature over time thus reduced E. coli die-off rate (Table 1c; Chandrasena et al. (2014a), and subsequent remobilisation during rain events (Auset et al., 2005; Mohanty et al., 2013). The E. coli removal rates by PB filters remained unchanged over time, while LC filters increased over time, possibly due to plant establishment (Bratieres et al., 2008) or lower infiltration rates, as demonstrated in Fig. 2. While their removal was mostly better than non-copper systems, all copper-zeolite filters (SCu, PBCu, LCCu and LCCu-T) showed decreased performance over time (Fig. 4). SCu, for example, removed E. coli by 2.6 log in Stage 1 but only by 1.8 log in Stage 3. As with the S filters, it is very likely that cold operational temperatures at the end of Stage 2 and during Stage 3 played a critical role in the reduced performance of SCu. In addition, the accumulation of sediments on the top of the filters formed a surface-clogging layer, a common problem with biofilters (Le Coustumer et al., 2009). Although this surface clogging may benefit pollutant removal by traditional biofilters through reduced flow rates, it may also impose a negative impact on the antimicrobial efficiency of ZCuCuO180 by way of (1) unsaturated flow within copper-zeolite media, thus limiting contact time between microbes and the media and (2) deaeration of the media (Borkow and Gabbay, 2005; Li et al., 2014a).

Y. Li et al. / Ecological Engineering 90 (2016) 234–243

241

Fig. 3. Plants in various designs (photos taken in Week 28).

Fig. 4. E. coli logarithmic reduction rates and infiltration rates over three stages of experiment.

3.3.2. Intermittent dry weather periods The effect of dry weather periods on E. coli log removal was twofold: (1) those configurations which showed little or negative effects (i.e. PB and LC) and (2) those which showed positive effects (i.e. S, SCu, PBCu, LCCu and LCCu-T). Importantly, these observations were not consistently linked to infiltration rate changes after drying periods (Fig. 4). The impact of drying periods is further examined below using the breakthrough curves in Fig. 5: ‘before’ graphs were sampled after roughly 3 days of drying, while the ‘after’ graphs were sampled after 2 weeks or 4 weeks of drying. The breakthrough curves of SCu and LCCu filters during the sampling in Week 20 (before 2 weeks of drying) were slightly lower than that of LC, while such difference disappeared during the sampling in Week 29. This is consistent with the earlier discussion about decreased performance of copper-zeolite filters coinciding with decreased temperature and surface clogging over time. After 2–4 weeks of drying weather, the breakthrough curves of all three designs (LC, SCu, LCCu) started with E. coli concentration at or below detection limit. Compared with copper-media filters, LC filters showed earlier breakthrough after drying period. Such earlier breakthrough in LC filters is caused by the uptake of the ‘cleaned’

residual pore and submerged zone water during extended drying, hence the earlier arrival of the less treated new water. This is confirmed by water balances of the effluent (Fig. 5): LC filters had 1–2 L less effluent than SCu and LCCu filters. Furthermore, LC filters also showed increased magnitude of breakthrough concentration particularly after 4 weeks drying. SCu and LC filters, on the contrary, showed significantly lower breakthrough concentration, indicating the recovery of antimicrobial efficiency of copper-zeolite media after drying period. As discussed in earlier section, the accumulated sediment layer at the top surface can act as a negative factor for copper-zeolite work efficiently; and extensive drying can break this layer leading to recovered performance. As such, it is hypothesised that dry weather periods cause multiple other effects in biofilters. On the one hand, small dry weather periods (<1 week) can help the removal rates of microbes through cleansing of the residual water contained in the pores and saturated zone via die-off processes. However, longer drying periods in systems with extensive vegetation (like PB and LC) will also deplete the water reservoirs in the biofilter, meaning less of the cleansed residual water is available to dilute the outflow.

242

Y. Li et al. / Ecological Engineering 90 (2016) 234–243

Fig. 5. E. coli outflow concentration over sampling volume for three biofilter designs (two replicates each design): SCu, solid line with shaded circles; LC, dashed line with open squares; and LCCu, dashed line with shaded diamonds (Note: WD, weeks drying).

4. Conclusions The performance of antimicrobial media copper-zeolite amended biofilters was monitored over 8 months. E. coli removal by the copper-zeolite biofilters met well the Australian unrestricted irrigation guideline and effluent E. coli concentrations met the primary contact recreation use. Conversely, E. coli removal and effluent concentrations from non-copper-zeolite biofilters did not meet such guidelines. Filter media type was found to play a key role in determining the bacterial treatment performance, where the improved removal by copper-zeolite amended filters was due to enhancing contactbased inactivation, adsorption and survival/growth inhibition by copper-zeolite. Vegetation was not a significant factor for bacterial removal by copper-zeolite filters, which could be due to the stressed growth of vegetation in these designs. Due to the potential benefits of using vegetation in biofilters, future studies should investigate how to design the systems to support vegetation, and hence improve the overall performance of the systems by optimising the synergistic effects of plants and the antimicrobial media on faecal bacteria removal. Copper-zeolite filters showed decreased performance over time, which was speculated to be due to (but not limited to) mixed effects of decreased temperature and surface clogging reducing the inactivation efficiency of antimicrobial media. In fact, intermittent drying period (2 weeks or longer) was found to positively affect the performance of copper-zeolite filters, which was believed to be due to dryness induced breaking surface clogging layer thus eliminating its negative impact. Future longevity test particularly covering cycles of cold-warm seasons may prove the speculation about temperature effect.

This study provides unique evidence that stormwater biofitlration systems can be engineered to successfully meet Australian guidelines for stormwater harvesting in a consistent manner. Future work should be conducted to ensure that these systems are recognised as an important part of a stormwater harvesting system. Indeed, treating stormwater to levels acceptable for harvesting using this single low-energy, low-maintenance treatment device, will enable widespread implementation of distributed alternate water supplies, thereby creating resilient cities of the future. Acknowledgements The support of the Commonwealth of Australia through the Cooperative Research Centre program is acknowledged. Louisa John-Krol is warmly thanked for editing grammatical errors in English. Active support from Gayani Chandrasena, Anthony Brosinsky, Minna Tom, Rebekah Henry, Christelle Schang, Richard Williamson, Emily Payne, Bonnie Glaister, Zhenmin Huang, Josh J. Kamil, Peter Kolotelo, Kan Bu and Rebecca Coulthard, is acknowledged with much gratitude. Stewart Crowley from the School of Biological Sciences is gratefully acknowledged for coordinating usage of the greenhouse. Ashley Connelly and Ben Evans from Pakenham Treatment Plant are acknowledged for their sincere support for raw sewage sampling. References ANZECC/ARMCANZ, 2000. Australian and New Zealand Guidelines for Fresh and Marine Water Quality. Australian and New Zealand Environmental Conservation Council and Agriculture and Resource Management Council of Australia and New Zealand.

Y. Li et al. / Ecological Engineering 90 (2016) 234–243 Auset, M., Keller, A.A., Brissaud, F., Lazarova, V., 2005. Intermittent filtration of bacteria and colloids in porous media. Water Resour. Res. 41, W09408, http://dx.doi. org/10.01029/2004WR003611. Bitton, G., Marshall, K. (Eds.), 1980. Adsorption of Microorganisms to Surfaces. Wiley, New York. Blecken, G.T., Zinger, Y., Deletic, A., Fletcher, T.D., Viklander, M., 2009a. Impact of a submerged zone and a carbon source on heavy metal removal in stormwater biofilters. Ecol. Eng. 35, 769–778. Blecken, G.T., Zinger, Y., Deletic, A., Fletcher, T.D., Viklander, M., 2009b. Influence of intermittent wetting and drying conditions on heavy metal removal by stormwater biofilters. Water Res. 43, 4590–4598. Borkow, G., Gabbay, J., 2005. Copper as a biocidal tool. Curr. Med. Chem. 12, 2163–2175. Bratieres, K., Fletcher, T.D., Deletic, A., Somes, N., Woodcock, T., 2009. The advantages and disadvantages of a sand based biofilter medium; results of a new laboratory trial. In: 6th International WSUD Conference and Hydropolis, Perth, Australia. Bratieres, K., Fletcher, T.D., Deletic, A., Zinger, Y., 2008. Nutrient and sediment removal by stormwater biofilters: a large-scale design optimisation study. Water Res. 42, 3930–3940. Brix, H., 1997. Do macrophytes play a role in constructed treatment wetlands? Water Sci. Technol. 35, 11–17. Chandrasena, G.I., Deletic, A., Ellerton, J., McCarthy, D.T., 2012. Evaluating Escherichia coli removal performance in stormwater biofilters: a laboratory-scale study. Water Sci. Technol. 66, 1132–1138. Chandrasena, G.I., Deletic, A., McCarthy, D.T., 2014a. Survival of Escherichia coli in stormwater biofilters. Environ. Sci. Pollut. Res. 21, 5391–5401. Chandrasena, G.I., 2014. Faecal Microorganism removal in stormwater biofilters. PhD Thesis. Monash University, Clayton, Victoria, Australia, http://arrow. monash.edu.au/hdl/1959.1/965653. Chandrasena, G.I., Pham, T., Payne, E.G., Deletic, A., McCarthy, D.T., 2014b. E. coli removal in laboratory scale stormwater biofilters: influence of design and operational conditions. J. Hydrol. 519, 814–822. Dietz, M.E., Clausen, J.C., 2006. Saturation to improve pollutant retention in a rain garden. Environ. Sci. Technol. 40, 1335–1340. Duncan, H.P., 1999. Urban Stormwater Quality: A Statistical Overview. Cooperative Research Centre for Catchment Hydrology, Melbourne. FAWB, 2009. Guidelines for Soil Filter Media in Bioretention Systems. Facility for Advancing Water Biofiltration Monash University, Melbourne, Australia. Feng, W., Hatt, B.E., McCarthy, D.T., Fletcher, T.D., Deletic, A., 2012. Biofilters for stormwater harvesting: understanding the treatment performance of key metals that pose a risk for water use. Environ. Sci. Technol. 46, 5100–5108. Hathaway, J.M., Hunt, W.F., Graves, A.K., Wright, J.D., 2011. Field evaluation of bioretention indicator bacteria sequestration in Wilmington, North Carolina. J. Environ. Eng. 137, 1103–1113. Hathaway, J.M., Hunt, W.F., Jadlocki, S., 2009. Indicator bacteria removal in stormwater best management practices in Charlotte, North Carolina. J. Environ. Eng. 135, 1275–1285. Hrenovic, J., Milenkovic, J., Ivankovic, T., Rajic, N., 2012. Antibacterial activity of heavy metal-loaded natural zeolite. J. Hazard. Mater. 201, 260–264. Hunt, W.F., Smith, J.T., Jadlocki, S.J., Hathaway, J.M., Eubanks, P.R., 2008. Pollutant removal and peak flow mitigation by a bioretention cell in urban Charlotte, NC. J. Environ. Eng. 134, 403–408.

243

Le Coustumer, S., Fletcher, T.D., Deletic, A., Barraud, S., Lewis, J.F., 2009. Hydraulic performance of biofilter systems for stormwater management: influences of design and operation. J. Hydrol. 376, 16–23. Li, Y.L., Deletic, A., Alcazar, L., Bratieres, K., Fletcher, T.D., McCarthy, D.T., 2012. Removal of Clostridium perfringens, Escherichia coli and F-RNA coliphages by stormwater biofilters. Ecol. Eng. 49, 137–145. Li, Y.L., Deletic, A., McCarthy, D.T., 2014a. Removal of E. coli from urban stormwater using antimicrobial-modified filter media. J. Hazard. Mater. 271, 73–81. Li, Y.L., McCarthy, D.T., Deletic, A., 2014b. Stable copper-zeolite filter media for bacteria removal in stormwater. J. Hazard. Mater. 273, 222–230. McCarthy, D.T., Hathaway, J.M., Hunt, W.F., Deletic, A., 2012. Intra-event variability of Escherichia coli and total suspended solids in urban stormwater runoff. Water Res. 46, 6661–6670. Milan, Z., de las Pozas, C., Cruz, M., Borja, R., Sanchez, E., Ilangovan, K., Espinosa, Y., Luna, B., 2001. The removal of bacteria by modified natural zeolites. J. Environ. Sci. Health A: Toxic/Hazard. Subst. Environ. Eng. 36, 1073–1087. Mohanty, S.K., Boehm, A.B., 2014. Escherichia coli removal in biocharaugmented biofilter: effect of infiltration rate, initial bacterial concentration, biochar particle size, and presence of compost. Environ. Sci. Technol. 48, 11535–11542. Mohanty, S.K., Torkelson, A.A., Dodd, H., Nelson, K.L., Boehm, A.B., 2013. Engineering solutions to improve the removal of fecal indicator bacteria by bioinfiltration systems during intermittent flow of stormwater. Environ. Sci. Technol. 47, 10791–10798. Mukerji, K.G., Manoharachary, C., Singh, J., 2006. Microbial Activity in the Rhizosphere. Springer, Berlin. NRMMC–EPHC–NHMRC, 2008. Australian Guidelines for Water Recycling: Augmentation of Drinking Water Supplies. NRMMC, EPHC, NHMRC, 2009. Australian Guidelines for Water Recycling (Phase 2): Stormwater Harvesting and Reuse. Natural Resource Management Ministerial Council, Environment Protection and Heritage Council, National Health and Medical Research Council, Canberra, Australia. Payne, E., Pham, T., Cook, P.L.M., Fletcher, T.D., Hatt, B.E., Deletic, A., 2014. Biofilter design for effective nitrogen removal from stormwater: influence of plant species, inflow hydrology and use of a saturated zone. Water Sci. Technol. 69, 1312–1319. Rusciano, G.M., Obropta, C.C., 2007. Bioretention column study: fecal coliform and total suspended solids reductions. Trans. ASABE 50, 1261–1269. Stevik, T.K., Aa, K., Ausland, G., Hanssen, J.F., 2004. Retention and removal of pathogenic bacteria in wastewater percolating through porous media: a review. Water Res. 38, 1355–1367. VGG, 2003. Victoria Government Gazette S107 – Variation to State Environment Protection Policies, (Waters of Victoria). Craftsman Press Pty. Ltd., Victoria, Australia. Zhang, L., Seagren, E.A., Davis, A.P., Karns, J.S., 2011. Long-term sustainability of Escherichia Coli removal in conventional bioretention media. J. Environ. Eng. 137, 669–677. Zinger, Y., Blecken, G.T., Fletcher, T.D., Viklander, M., Deletic, A., 2013. Optimising nitrogen removal in existing stormwater biofilters: benefits and tradeoffs of a retrofitted saturated zone. Ecol. Eng. 51, 75–82.