ANAEROBIC MEMBRANE BIOREACTOR TREATING MUNICIPAL WASTEWATER: BIOSOLIDS CHARACTERISTICS AND MEMBRANE FOULING

Transcription

1 ANAEROBIC MEMBRANE BIOREACTOR TREATING MUNICIPAL WASTEWATER: BIOSOLIDS CHARACTERISTICS AND MEMBRANE FOULING Dong, Q. 1,3*, Dagnew M. 2, Cumin J. 2, Parker W. 1 and Dold, P. 3 1 Department of Civil and Environmental Engineering, University of Waterloo, Ontario, Canada 2 GE Water and Process Technologies, Ontario, Canada 3 EnviroSim Associates Ltd., Ontario, Canada *Corresponding Author qirong.dong@envirosim.com Abstract A pilot-scale anaerobic membrane bioreactor (AnMBR) treating raw municipal wastewater was operated for approximately 600 days. This study assessed the characteristics of biosolids and membrane performance. The production of total solids (TS) and volatile solids (VS) was low, and comparable to that reported for an extended aeration system at solids residence time (SRT) longer than 40 days. The yields of TS and VS were reduced as SRT increased from 40 to 100 days. Ferric chloride (FeCl3) addition of 26 mg/l influent increased both TS and VS. The VS loading in the influent wastewater to the AnMBR was reduced by 60 82% and it was concluded the biosolids met the requirements for vector attraction reduction for land application. The nutrient content in terms of total Kjeldahl nitrogen (TKN) and total phosphorus (TP) was similar to that of anaerobically digested municipal sludges. The dewaterability of the biosolids was poorer than that reported for sludges from conventional aerobic treatment and anaerobically digested sludges. The biosolids met standards for land application with regards to heavy metals but would need further treatment to meet Class B pathogen indicator criteria. The long-term impacts of SRT and the addition of FeCl3 on membrane performance were assessed. At lower SRT (40 days) mixed liquor TSS concentration and colloidal COD (ccod) concentration were lowest, and capillary suction time (CST) was lowest, indicating the lowest fouling propensity. Recovery cleaning resulted in substantial reduction of resistance as indicated by both pilot plant operation and clean water flux tests. The long-term fouling rate was observed to be higher with cleaned membranes as compared to virgin membranes. The lower membrane fouling with virgin membranes suggested that accumulation of foulants resistant to cleaning caused the higher fouling rates for the cleaned membranes. The addition of 26 mg/l of FeCl3 significantly enhanced the membrane performance. For operation started with a virgin membrane the trans-membrane pressure (TMP) was maintained at values lower than 5 kpa for the first 75 days. The superior membrane performance was attributed to the reduced colloidal COD concentration and improved dewaterability. Keywords AnMBR; Biosolids; SRT; Recovery cleaning; FeCl3 Dosing 1. Introduction Anaerobic membrane bioreactors (AnMBRs) are recognized as a sustainable technology for wastewater treatment because the anaerobic process has low sludge production, low energy requirements and can generate methane as an alternative energy source (Metcalf & Eddy Inc., 2003; Smith et al., 2012). The use of membranes for biomass separation allows long solids retention time (SRT) operation which offsets the low growth rates of anaerobic organisms while also producing a solids-free effluent (Smith et al., 2013; Gao et al., 2014). In addition, low hydraulic retention times (HRTs) can be achieved to ensure the economical viability of the AnMBRs. Therefore the AnMBR is a promising technology for municipal wastewater treatment. The application of AnMBRs to actual municipal wastewaters has been reported relatively recently (Gao et al., 2014; Huang et al., 2011; Gimenez et al., 2011; Smith et al; 2012; Lew et al., 2009; Martinez-

2 Sosa et al., 2011). In these studies, it has been found that extending the SRT results in improved removal of COD with a concurrent increase in biogas production. Further, mixed liquor suspended solids (MLSS) concentrations increases. Long SRT operation results in a reduction in the generation of volatile suspended solids in the waste biosolids stream; however, the extent of reduction in biosolids production has not been quantified. Other AnMBR operating strategies may also influence the properties and amount of biosolids generated in AnMBRs. For example, the addition of ferric chloride (FeCl3) to an AnMBR treating municipal wastewater (Dong et al., 2015) enhanced the removal efficiencies of COD and BOD5 because both colloidal and some soluble organic material is coagulated. The addition of FeCl3 increases the quantity of fixed suspended solids produced in the AnMBR; however, the extent of the increase has not been quantified. Changing MLSS concentration with changing SRT, and changes in particle size distribution and the reduction of colloidal material with FeCl3 addition, likely will influence the dewaterability of the biosolids and the propensity for membrane fouling (Liao et al., 2006). MLSS concentration has been identified as a key factor in membrane fouling (Liao et al., 2006; Huang et al., 2011). It is believed that an increase in the suspended solids concentration, as measured by TSS, increases the convective flow of solids towards the membrane surface and enhances cake formation and fouling. Colloidal COD also is reported to contribute to the formation of a strongly-attached fouling layer on/into the membrane that resists physical cleaning (Choo and Lee, 1996; Fan et al., 2006; Liao et al., 2006; Dagnew et al., 2012). In the current study it was hypothesized that optimized SRT and addition of FeCl3 should improve membrane performance by reducing fouling because both of these factors change MLSS and the amount of colloidal COD in the mixed liquor. On a related AnMBR issue, there is a lack of information in the literature on waste biosolids disposal and potential use in land application. An assessment of the feasibility for land application requires information on nutrient content (concentrations of TKN and TP), volatile solids reduction, and the concentrations of pathogens and heavy metals (McFarland, 2001; Metcalf and Eddy, 2003). Little information on these properties of the biosolids generated in AnMBRs treating municipal sewage is available. The paper presents the impact of SRT and addition of FeCl3 on the quality and quantity of biosolids in a pilot scale AnMBR treating municipal wastewater. Membrane performance and the characteristics of mixed liquor with regard to membrane fouling also are discussed. 2. Materials and Methods A pilot scale AnMBR receiving raw municipal wastewater (after 3 mm screening) from the Burlington Skyway Wastewater Treatment Plant (Ontario, Canada) was employed in this study (Figure 1). A description of the design and operating parameters of the system is presented in Table 1. The AnMBR consisted of a completely mixed anaerobic digester (AD) and a separate membrane tank (MT). The MT held a polyvinylidine fluoride (PVDF) hollow fibre membrane module (GE ZeeWeed 500). The AD contents were mixed by recirculation with a positive displacement pump. In addition, the AD mixed liquor was circulated through the MT using a centrifugal pump that withdrew mixed liquor from the bottom of the AD and pumped it to the bottom of the MT after which it overflowed from the top of the MT back to the AD. This circulation mixed the MT contents and generated a cross flow velocity (CFV) that would enhance surface shear for membrane fouling control. Biogas produced in the AD was released from the head space of the AD and its production was measured by a gas flow meter (Aalborg GFM17). Biogas was circulated through the MT with a blower (KNF NEUBERGER, PM ) to reduce membrane fouling.

3 Figure 1 Schematic of AnMBR System Throughout the study the temperature of the AD was controlled using a heat tape and a temperature sensor in the digester. Digester ph was monitored and controlled through NaHCO3 addition. The operation and data acquisition were conducted using a programmable logic controller (PLC). The feed of raw sewage and the wasting of anaerobic digester contents for SRT control were controlled on the basis of the weight of the pilot digester monitored by load cells installed at the base of the pilot digester. Biosolids were wasted directly from the AD for SRT control, but calculation of SRT accounted for solids in both the AD and MT. Concentrations of particulate species in the biosolids were the same as that of the AD. The pressures of the headspace in the MT and the permeate line were recorded using digital pressure gauges. The trans-membrane pressure (TMP) was calculated as the difference between the pressure in the MT and that of the permeate line. To reduce membrane fouling, discontinuous filtration and weekly membrane chemical cleaning were incorporated in the pilot operation. The discontinuous filtration mode consisted of a repeating cycle with filtration for 8 minutes and relaxation for 2 minutes. Weekly membrane cleaning in place (CIP) consisted of backpulsing with 2000 mg/l citric acid at 2.9 L/min for 4 minutes through the membrane. Membrane performance was characterized by monitoring the TMP after relaxation when filtration re-started. These TMP values were considered to be indicative of fouling that could not be removed by mixed liquor recirculation and biogas sparging. Table 1 AnMBR Operational Parameters Parameter Pilot AnMBR AD Volume (L) 550 MT Volume (L) 80 Membrane Surface Area (m 2 ) 5.4 Membrane Pore Size (µm) 0.04 Temperature ( C) 23±1 ph 6.7~6.8 AD Mixing Flow (L/h) 3,600 AD-MT Recirculation Flow (L/h) 918 Biogas Sparging Rate 0.786

4 (m 3 /h at 20 ºC and 1 atm) Influent flow (L/d) 1,700 HRT (hr) 8.5 Recovery cleaning of the membranes was conducted in Phase 2 (once), 3 (once) and 4 (twice) for TMP control. In the first 3 recovery cleanings the membrane module was sequentially soaked in solutions of 2000 mg/l of citric acid and 2000 mg/l of NaOCl for 16 hours respectively. The fourth recovery cleaning was inadvertently conducted with a reversed soaking sequence. The concentrations of citric acid and NaOCl and its soaking duration were within the membrane vendor s operational limit for the membrane. Therefore, the recovery cleaning was assumed to have minimal impact on the physiochemical properties of the membrane. In all recovery cleanings, clean water flux (CWF) tests were carried out on the virgin, fouled and cleaned membranes. The CWF tests involved permeating tap water through the membrane with a series of flux values each of a fixed duration while monitoring the TMP at each flux value. The flux values were set at 5, 10, 15, 20, 25, 30 and 35 LMH, and each step was conducted for 8 minutes, with a 2-minute relaxation. Linear regression of the TMP and flux values from each CWF test was employed to estimate the resistance of the membrane. Influent raw wastewater 24-hour composite samples were collected by an auto-sampler. The mixed liquor samples were grab samples. Wastewater and mixed liquor samples were collected twice a week from the outlets of the feed pump (upstream of the FeCl3 dosing) and anaerobic reactor, respectively. The number of the samples is shown in Table 2. The samples were analyzed for concentrations of total COD, total solids (TS), total suspended solids (TSS), volatile solids (VS), volatile suspended solids (VSS), total Kjeldahl nitrogen (TKN) and total phosphorus (TP). All analyses were conducted according to Standard Methods (APHA, 2005). The dewaterability of the sludge was assessed using the Capillary Suction Time test and was measured using a GENEQ model 304B device. The concentrations of pathogen indicators (Fecal coliforms and Escherichia coli (E.coli)) were assessed by (APHA, 2005: 9222D). The concentrations of heavy metals (arsenic, cadmium, copper, lead, mercury, molybdenum, nickel, selenium and zinc) were analyzed to further assess the biosolids with respect to qualities that might impact on disposal. The heavy metals were measured by ICP-MS using an Agilent ICP-MS 7500ce equipped with a CETAC ASX510 Autosampler. Various COD analyses were performed on permeate and mixed liquor samples once per week. As the membrane had a nominal pore size of 0.04µm, the permeate COD was regarded as soluble COD. Mixed liquor total COD was measured. Also mixed liquor was centrifuged at 4000 rpm for 12 minutes and the supernatant was filtered through a 1.5µm glassfiber filter and the filtrate was analyzed for COD. The difference between 1.5µm-filtered COD and permeate COD was regarded as colloidal COD (ccod) according to the modified method reported by Fan et al., (2006). The test plan facilitated an assessment of the impact of SRT and FeCl3 addition on biosolids production and characteristics and on the membrane performance. Before starting Phase 1 of the study the pilot AnMBR was operated at an HRT of 8.5 hours and an SRT of 70 days for 5 months without FeCl3 addition. After this the research was conducted in 4 phases at a constant HRT of 8.5 hours, corresponding to a membrane flux of 17 LMH. During Phase 1 the pilot AnMBR was operated at an SRT of 70 days and fed with non-fe dosed sewage to establish a base condition. Subsequent testing was conducted with the AnMBR operated at SRTs of 100, 70 and 40 days in Phases 2, 3 and 4, respectively, and with addition of FeCl3 (26mg/L) to the influent (Table 2). The impact of FeCl3 was assessed by comparing the data from Phases 1 and 3 and the impact of SRT was assessed by analyzing the data from Phases 2, 3 and 4. Each phase was operated to achieve both biological steady state and to assess long-term membrane performance.

5 Table 2 Phase Operating Conditions for Test Phases FeCl 3 Dosage SRT (mg/l (days) influent) Duration (days) Number of Samples Data collected during each phase was employed to evaluate the influence of SRT and addition of FeCl3 on the responses. The influence was assessed statistically using ANOVA tests. In the subsequent discussion of results the significance of the statistical analysis is presented in brackets (i.e. p< xxx) whenever a statistical assessment was conducted to determine if a comparison was statistically significant. The treatment performance is affected by the characteristics of the influent sewage. Hence, the sewage employed in this study was characterized with respect to total and soluble COD, TSS, VSS, TP and TKN. The average (±SD) concentrations of these parameters are presented in Table 3. Due to the seasonal variability, the average concentrations of total COD in the sewage in Phases 3 and 4 were generally higher than in Phases 1 and 2. However, no significant difference was found between Phases 3 and 4 (p<0.67). Other parameters such as TSS, VSS, TKN and TP followed the trend of total COD in the 4 phases. Table 3 Sewage Characteristics Phase SRT (d) FeCl 3 (mg/l influent) Total COD (mg/l) 251 ± ± ± ± 48 TSS (mg/l) 118 ± ± ± ± 29 VSS (mg/l) 95 ± ± ± ± 26 TKN (mg/l) 27.4 ± ± ± ± 10.3 TP (mg/l) 3.6 ± ± ± ± Results and Discussion 3.1 Treatment Performance The treatment performance was evaluated at steady state in the pilot AnMBR on the basis of the concentrations of COD and BOD5 in the permeate and the production of methane. From Table 4 it can be seen that the average COD concentrations in the permeates were less than 50mg/L and the average BOD5 concentrations were less than 15 mg/l in all tests (the VFA concentration ranged from 1.7 to 10 mg/l). The superior permeate quality may be attributed to complete retention of solids and colloids by the membrane (pore size: 0.04 µm). The membrane also enhances the accumulation of biomass, which can result in enhanced hydrolysis and subsequent methanogenesis. These results indicate that all the investigated operational conditions achieved good effluent quality in terms of COD and BOD5.

6 Table 4 Treatment Performance Phase SRT (d) FeCl 3 (mg/l influent) Permeate COD (mg/l) 45.8 ± ± ± ± 8.5 COD Removal Efficiency (%) 79.7 ± ± ± ± 1.9 Permeate BOD5 (mg/l) 14.7 ± ± ± ± 3.0 BOD5 Removal Efficiency (%) 84.5 ± ± ± ± 1.5 Methane yield (ml/gcod fed) (0ºC and 101.3kPa) 102 ± ± ± ± 14 Due to the seasonal variation in sewage concentration, the removal efficiency was employed to characterize the impact of SRT and FeCl3 addition on the treatment performance. From Table 4 it is evident that the average removal efficiencies of COD and BOD5 were higher than 88% and 93% respectively in Phases 2, 3 and 4. Statistical analyses showed there was no significant effect of SRT on either removal of COD (p < 0.42) or BOD5 (p < 0.36). Therefore, it was concluded that SRT had no influence on the removal of COD and BOD5. The impact of FeCl3 dosing was evaluated by comparing the results of Phase1 and 3. From Table 4 it can be seen that the average COD removal efficiencies were 79.9% and 93.7% in Phases 1 and 3 respectively while the average BOD5 removal efficiencies were 84.2% and 95.0% respectively. Hence, the removal efficiencies of COD and BOD5 in Phase 3 were 13.4% (p<2.86*10-11 ) and 12.8% (p<5.88*10-7 ) higher than in Phase 1. The results suggest that the addition of FeCl3 had a significant influence on the removal of COD and BOD5. It was hypothesized that some of the soluble organic matter that was coagulated to form floc which was retained in the reactor by the membrane. The production of methane can contribute to the sustainability of AnMBRs as it might be employed to compensate for the energy consumed in AnMBR operation (Smith et al., 2014). Therefore, the methane measured in the biogas and the estimated methane dissolved in the permeate (Henry s Law) were employed to calculate the methane yield based on the influent COD. From Table 4 it can be seen average methane yield in Phase 2 was 26.4% (p < 0.003) and 51.3% (p < 1.3*10-7 ) higher than in Phase 3 and 4 respectively. The decreasing methane yields with decreasing SRT were attributed to less hydrolysis and subsequent methanogenisis at shorter retention times. The decreased methane yields were consistent with increased VSS yields at the shorter SRTs. The impact of FeCl3 additon on methane yield was investigated by comparing the steady state data in Phase 1 and 3. The methane yields were 102 ± 30 mlch4/gcodfed in Phase 1 and 91 ± 19 mlch4/gcodfed in Phase 3 and were not statistically different (p<0.31). Hence it was concluded that FeCl3 dosing did not inhibit the activity of methanogens or reduce the availability of organic matter in the sewage. 3.2 Biosolids Characteristics The production of biosolids in the AnMBR was first evaluated in terms of the yields of volatile solids (VS) and total solids (TS) normalized on the basis of the volume of treated sewage (Figure 2). [That basis is often used in references]. From Figure 2 it can be seen the generation of VS in the AnMBR ranged between 30 and 91 g/m 3 sewage and differed between the phases of the study. The values were lower than the typical VS yield that would be expected from a combination of primary and

7 secondary aerobic treatment reported in the range 138 to 190 g/m 3 sewage (Metcalf and Eddy 2003). Hence, it was concluded that direct treatment of municipal sewage in an AnMBR only produces approximately 25-50% of the VS that are generated by traditional wastewater treatment. Figure 2 Yields of TS and VS Anaerobic digestion of the sludges produced in conventional wastewater treatment is often employed to reduce VS production. Hence, the previously referenced combination of yields of primary and secondary treatment sludges were combined with typical values for their anaerobic digestibility to provide a point of comparison with biosolids production from direct AnMBR treatment of wastewater. Assuming equivalent yields of primary and secondary VS and typical VS destruction efficiencies of 60% for primary sludges and 40% for secondary sludges (Jones et al., 2007) a combined system could be expected to generate gvs/m 3 wastewater. The VS production observed in this study was less than these typical values for 3 of the 4 scenarios tested. The yield of VS in the AnMBR per unit mass of COD in the raw sewage should depend on SRT of the AnMBR and potentially on the use of FeCl3 as a flux enhancer. As shown in Table 3 the COD of the raw sewage differed substantially between the phases due to seasonal variability. To remove the impact of feed variability from the analysis, the VS yield was calculated on the basis of the mass loading of COD to the AnMBR (Table 5). Table 5 VS Yield based on COD Loading Phase SRT (d) FeCl 3 (mg/l influent) VS Yield (g VS/g CODfed) 0.12 ± ± ± ± 0.01 It was anticipated that decreasing the SRT of the AnMBR would increase the VS yield as extended SRTs would allow more time in the reactor for hydrolysis of the influent VS. Further, reduced yields of biomass would also result from increased endogenous decay of biomass at the extended SRTs. The impact of SRT on VS yields was assessed by examining the values estimated for Phases 2, 3 and 4 (all with FeCl3 addition) where the SRT decreased from 100 to 70 and then to 40 days. From Table 5 it can be seen that the average yield in Phase 4 was 22.2% (p<1.7*10-5 ) higher than in Phase 3 and 37.5% (p<4.3*10-7 ) higher than in Phase 2. Hence, it was apparent that operation at extended SRTs could substantially reduce the production of VS in the AnMBR and this was consistent with increased

8 hydrolysis and endogenous processes. This observation is consistent with COD removal and biogas production discussed earlier. It was hypothesized that the addition of FeCl3 to the AnMBR feed may increase the VS yield due to coagulation of colloidal and some soluble organics into a particulate form. In this study, the impact of FeCl3 addition was assessed by comparing the yields observed in Phases 1 and 3 where the SRTs were the same but operating without and with FeCl3 addition, respectively. From Table 5 it can be seen that the average yield of VS increased from 0.12 to 0.18 g VS/gCODfed upon FeCl3 addition (p<5.8*10-8 ). Hence, the addition of FeCl3 substantially increased the production of VS in the AnMBR. The production of TS is an important characteristic of biosolids production in AnMBRs as it reflects the quantity of solids that may require further handling. From Figure 2 it can be seen that the yield of TS ranged from 41 to 140 g/m 3 and varied substantially between the phases of the study. Despite the variability, the TS yields in the current study were considerably less than the yields that have been reported from the combination of primary and secondary aerobic treatment (180~270g/m 3 sewage) (Metcalf and Eddy 2003). With the exception of Phase 4, the TS yields were lower or comparable to those reported for an extended aeration process without primary treatment and without chemical addition (80~120g/m 3 sewage). It was concluded that treatment with AnMBR could substantially reduce TS production when compared to traditional wastewater treatment. The addition of FeCl3 as a flux enhancer was expected to have a greater impact on the yield of TS in the AnMBR than on VS production due to the production of inorganic precipitates. The impact of FeCl3 addition on TS production was assessed by comparing the yields in Phases 1 and 3 that had the same SRT but were operated without and with FeCl3 addition respectively (Figure 2). From Figure 2 it can be seen that the average TS yield in Phase 3 was 165% higher than Phase 1 (p<2.6*10-22 ). The increase in TS yield was partly due to the higher sewage COD loading (Table 3) and also to the addition of FeCl3. The relative impact of FeCl3 addition on TS and VS production was assessed by examining the volatile fraction of the produced solids in Phases 1 and 2-4 (Figure 2). From Figure 2 it can be seen that the volatile fraction was approximately 92% in Phase 1 and decreased to 72~76% in Phases 2-4. Hence, it was apparent that the addition of FeCl3 resulted in the production of more inorganic solids than organic solids thereby increasing the fixed solids fraction of the produced sludge. Volatile solids reduction is an important factor when land application is considered as a disposal option. This property is reflective of the biological stability of the organic matter in the biosolids. In traditional digestion the requirements for vector attraction reduction are considered to be satisfied if 38% of the VS mass is reduced in the treatment (US EPA 1999). In the current study, it was assumed that VAR requirements would be met if 38% of the VS loading in the feed wastewater to the AnMBR were destroyed. Hence, VS destruction was calculated on the basis of a mass balance of VS around the AnMBR. The VS mass flows decreased from185 ± 55 to 31 ± 13 g/d in Phase 1, from 300 ± 104 to 75 ± 18 g/d in Phase 2, 326 ± 97 to 114 ± 29 g/d in Phase 3 and 317 ± 55 to 127 ± 17 g/d in Phase 4. Correspondingly, from Figure 3 it can be seen that the VS destruction through the AnMBR ranged from 60% to 83%. Hence it was concluded that the biosolids produced in AnMBR would satisfy VAR requirements for land application.

9 Figure 3 VS Mass Reduction The concentrations of TSS and VSS in the waste stream from the AnMBR (Figure 4) were characterized as these values may influence decisions on downstream processing (storage, dewatering, etc.). From Figure 4 it can be seen that the mean concentration of TSS ranged from 5.8 g/l in Phase 1 to 17.5 g/l in Phase 3. The corresponding range of VSS concentrations was 5.5 g/l to 12.0 g/l. Hence, the concentrations of suspended solids in the waste stream were somewhat less than that which would typically be observed with anaerobic digestion of conventional primary and secondary sludge mixtures that range from 1.5-4% TSS (Metcalf and Eddy, 2003). It is therefore likely that thickening or dewatering of the waste stream may be required prior to offsite transport. Figure 4 reveals substantial variations in the TSS and VSS concentrations in the mixed liquor between the different phases. It was anticipated that these concentrations would increase with SRT due to the accumulation of non-biodegradable components. Further, it was anticipated that addition of FeCl3 and increases in the feed sewage strength, would also increase the concentrations in the waste stream. In the current study the feed sewage concentrations differed between phases and hence the impact of SRT on the mixed liquor TSS and VSS concentrations could not be isolated in all Phases. However Phases 3 and 4 had similar feed concentrations with SRTs of 70 days and 40 days respectively. Figure 4 reveals that the TSS and VSS decreased by 21.1% (p<1.7*10-6 ) and 17.7% (p<2.1*10-5 ) respectively when then SRT was reduced. Further, the impact of FeCl3 addition and feed concentration variation on mixed liquor concentrations could not be isolated in this study. A comparison of Phase 1 versus Phase 3 in Figure 4 where both of these varied while the SRT remained constant reveals that the concentrations of TSS and VSS increased by 201% (p<4.3*10-11 ) and 118% (p<3.4*10-8 ). The greater increase in TSS relative to VSS was attributed to the formation of inorganic precipitates due to FeCl3 addition. Hence, the trends in the mixed liquor suspended solids concentrations followed those that were anticipated based upon SRT, feed concentration and FeCl3 addition.

10 Figure 4 Concentrations of TSS and VSS in Mixed Liquor The presence of TKN and TP in biosolids is an important property when considering application to agricultural land as a soil amendment. In the current study the presence of these components was normalized on the basis of TS concentrations and Figure5 shows that there was relatively little variation in the fractions between the study phases. Further, the concentrations of TKN and TP were comparable to those reported for anaerobically digested municipal biosolids which in the range of 0.5~17.6% and 0.5~14.3% for TKN and TP respectively (US EPA 1984). The AnMBR biosolids were deemed to be equivalent to anaerobically digested municipal biosolids in terms of TKN and TP content for land application purposes. Figure 5 Concentrations of TKN and TP in Biosolids The dewaterability of the waste stream is important when designing downstream dewatering processes (e.g., centrifugation, belt filter press). In the current study the dewaterability of the waste stream was evaluated using the capillary suction time (CST) test (Figure 6). From Figure 6 it can be seen that the CST values ranged from 345s to 2,265s through the various Phases of the study. It was anticipated that the CST values would be affected by SRT and the addition of FeCl3. Figure 6 shows that the CST generally decreased from Phases 2-4 as the SRT decreased and hence the suspended solids concentrations in the waste stream decreased. By contrast the CST values in Phase 3 were

11 substantially less than those observed in Phase 1 despite the significantly higher TSS concentrations. The reduced CST values in Phase 3 were attributed to the addition of FeCl3 that would have coagulated and precipitated soluble and/or colloidal matter to form larger floc. Overall it was concluded that reduced SRT and the addition of FeCl3 enhanced the dewaterability of the waste stream from the AnMBR. Figure 6 CST Values for AnMBR Waste Stream To further assess the dewaterability of the mixed liquor from the AnMBR, the CST values in this study were compared with those reported for aerobic processes and anaerobically digested biosolids. The CST values in all phases in this study were significantly higher than the values reported for aerobic sludges (5~13s) (Smollen 1986; Ge et al., 2011). In addition, the CST values in Phase 1-3 were generally higher than those reported for anaerobically digested biosolids (200~800s) (Smollen 1986; Vesilind et al., 1988; Krishnamurthy and Viraraghavan, 2005). The results suggest that while the AnMBR process operating conditions may be manipulated to enhance dewaterability special attention may be required in the design of downstream dewatering processes. It was anticipated that the operational conditions in Phase 4 may result in the least membrane fouling as the MLSS exhibited the highest dewaterability, and may hold the most potential for full scale implementation. Therefore, the pathogen indicator and heavy metal concentrations were tested in Phase 4 to evaluate the feasibility of biosolids for agricultural application. The potential presence of pathogens in municipal biosolids is an important property when considering land application of biosolids. Hence in this study the presence of fecal coliforms and E. coli. were assessed in Phase 4 to assess the feasibility of direct agricultural application of the biosolids (Table 6). From Table 6 it can be seen that the concentrations of fecal coliforms in the AnMBR biosolids were less than that reported for typical untreated sludges (10 10 CFU/L) and were comparable to that reported for anaerobically digested biosolids (3*10 5 ~6*10 7 CFU/L) (US EPA, 1979). However, when normalized on the basis of TS concentrations, the AnMBR biosolids failed to meet Class B pathogen-reduction criteria (2*10 6 CFU/gTS) (US EPA, 1999). The E. coli. concentrations (1.7~8.0*10 6 CFU/gTS) in the AnMBR waste stream also failed to meet reported standards for agricultural land application in Ontario (2*10 6 CFU/gTS) (Nutrient Management Act, 2002). The results suggest that additional treatment of the biosolids would be required to satisfy requirements for direct land application of biosolids. The concentration of heavy metals (arsenic, cadmium, copper, lead, mercury, molybdenum, nickel, selenium and zinc) was evaluated to further assess the feasibility of biosolids for agricultural application. From Table 6 it can be seen that all the measured metals concentrations met standards for land application in Ontario (Nutrient Management Act, 2002). It was concluded that heavy metal concentrations were not a concern for land application of biosolids from AnMBR treating municipal wastewater.

12 Table 6 Concentration In this study (mg/kg) a Standards for Land Application(mg/k g) a a: Dry weight basis Concentration of Metals and Pathogens in AnMBR Biosolids M As Cd Cu Hg Ni Pb Se Zn Fecal Coliforms E.Coli o ~9.8* CFU/g ~ ~ 8 ~ TS ~ ~ ~ ~ ~ ~ 843 or ~16.6* CFU/ L ~8.0*10 6 CFU/g TS or 3.4~13.6*10 7 CFU/ L Membrane Performance Impact of SRT SRT has been reported as a controllable parameter that can be employed to modify the mixed liquor characteristics (Huang et al., 2011) and the membrane performance. The membrane performance was assessed by examining the TMP profiles after relaxation. Figure 7 presents the TMP values ( kpa) that were observed over the testing period. The results presented in Figure 7 show that the pilot plant was operated for an extended period (536 days) where the TMP consistently fell within the range of acceptable operation as defined by the membrane vendor (maximum TMP of 80 kpa). The overall testing period could be divided into a number of sub-intervals with discernible patterns of TMP increase that were interpreted as being indicative of membrane fouling. The sub-intervals in the TMP dataset were bounded by interventions that included recovery cleaning (four times) and installation of virgin membranes (twice) and these were made to reduce TMP values. The sub-intervals varied in length and the longest period of operation without intervention was 178 days (from day ). From Figure 7 it can be seen that TMP values declined substantially after the recovery cleaning in each of the 3 phases and this was interpreted to indicate an effective removal of foulants from the membrane surface. The fouling rates after the first 3 recovery cleanings were consistently in the range of 0.27 kpa/d. However, the fouling rate after the fourth recovery cleaning, where a reversed cleaning sequence was inadvertently employed, increased to 0.61 kpa/d. The significantly higher fouling rate after the fourth recovery cleaning may have resulted from the inadvertent use of the reversed cleaning sequence. The installation of virgin membranes resulted in lower initial TMP values ( kpa) than those of the recovery cleaned membranes ( kpa) indicating the accumulation of foulants that were not removed during cleaning. Further, with virgin membranes the fouling rates of kpa/d and kpa/d observed in the first 60 days of operation in Phases 2 and 3 respectively were lower than those observed for fouled membranes that were recovery cleaned. The superior performance of the virgin membranes as compared to that of the cleaned membranes suggested the mitigated accumulation of foulants on the membrane. As it was believed that the recovery cleaning had no impact on the membrane chemistry, the subsequent discussion examines the relationship between fouling trends and mixed liquor properties to obtain insight into the nature of the fouling. In the current study, the SRT was reduced from 100 to 40 days from Phases 2 to 4. Therefore, it was hypothesized that the mixed liquor characteristics would vary from phase to phase and that this would impact upon membrane performance. Hence, the mixed liquor characteristics, including the concentration of TSS and the dewaterability of the mixed liquor as indicated by the ccod concentration and CST, were investigated to evaluate this hypothesis.

13 The dataset on mixed liquor properties that was generated in this study represented the results obtained from the pilot under both pseudo-steady state conditions for each SRT and under transient conditions as the operation was transitioned between steady states. The pseudo-steady state conditions reflected operation with authentic domestic wastewater but with seasonal variation in the wastewater properties that subsequently influenced the mixed liquor. Viewed collectively, a broad range of mixed liquor properties were attained in the pilot over the test period. The subsequent discussion examines the collective datasets to assess the correspondence between membrane fouling and the mixed liquor properties.

14 Figure 7 TMP Profile and Mixed Liquor Characteristics

15 The TSS concentration of the mixed liquor has been reported as a key factor affecting fouling (Liao et al., 2006; Huang et al., 2011; Gimenez et al., 2011). It is believed that an increase in mixed liquor solids concentration results in an increase in the convective flow of solids towards the membrane surface and enhances cake formation and fouling. From Figure 7 it can be observed that during Phase 2 the TSS concentrations increased from 11.3 g/l to an average pseudo-steady state value of 16.2 g/l. Upon reduction of the SRT from 100 to 70 days in Phase 3, the TSS concentration initially decreased with time. However, after Day 158 the TSS increased to 21.3 g/l at the end of Phase 3. This increase was attributed to a 24% increase in the sewage COD concentration in this period. When the SRT was subsequently reduced to 40 days in Phase 4, the TSS concentration declined with time and stabilized at approximately 11.5 g/l due to the reduced SRT and a small decline in the sewage COD concentration. Overall, it was anticipated that membrane performance would vary in concert with variations in the TSS concentration in the mixed liquor. The trends in TMP were compared with that of TSS concentrations to investigate the potential effects of TSS on membrane fouling. From Figure 7 it can be observed that in the period between days 1-35 a fouling rate of kpa/d was observed with a cleaned membrane and TSS concentrations in the range of g/L. This performance was similar to that which was observed immediately after installation of a virgin membrane. The fouling rate subsequently increased to 0.42 kpa/d (Days 36-50) and this corresponded to an increase in the TSS concentration to approximately 16.0 g/l. The results for this short period suggest that the increased TSS concentrations enhanced membrane fouling. After the first 50 days of operation the correspondence between the TMP and TSS values became less apparent. Elevated fouling rates of kpa/d were observed under conditions of both higher ( g/l for Days ) and lower ( g/l for Days ) TSS concentrations. Further, extended periods of operation at lower fouling rates ( kpa/d) were observed at both extremes of the TSS concentrations (19-23 g/l for Day and g/l for Day ). The lack of a relationship between membrane fouling rate and TSS concentration after the first 50 days of operations suggested that fouling in this period was impacted by factors other than TSS concentration. It was hypothesized that the dewaterability of the mixed liquor or the presence of irrecoverable foulants that resisted recovery cleaning (Judd, 2011; Resosudarmo et al., 2013; Wang et al., 2014) may have influenced the fouling rates and hence this was explored in detail. The presence of colloidal matter in the mixed liquor has been reported to cause fouling as it can contribute to the formation of a strongly attached cake layer and contribute to pore blocking (Fan et al., 2007; Meng et al., 2007). Therefore, the ccod concentration was employed to evaluate the impact of this mixed liquor property on fouling. From Figure 7 it can be seen that the concentrations of ccod essentially can be separated into three ranges. The lowest range of concentrations ( mg/l) was observed at the end of the testing (days ) while intermediate concentrations ( mg/l) were observed in two intervals (days and ). The highest concentrations ( mg/l) were observed in the days There was some consistency in the trends in ccod and TSS concentrations; however, notably, the ccod concentrations in the first 50 days of operation were in the middle range of concentrations whereas the TSS concentrations were relatively low. The relationship between ccod concentrations and membrane fouling rates was evaluated by examining the fouling rates that were observed when the pilot plant was operating in the three ccod concentration ranges (Table 7). From Table 7 it can be seen that there was little correspondence between the ccod concentrations and the fouling rates. Both high (>0.4 kpa/day) and low (<0.1 kpa/day) fouling rates were observed during low and intermediate ccod operation. By contrast, the fouling rates that were observed during the period of highest ccod concentration were amongst the lowest observed in the study. Hence, it was concluded that there was little relationship between long term fouling rates and ccod concentration.

16 Table 7 Fouling rate (kpa/d) Fouling rates categorized by ccod concentration range ccod Range (mg/l) < <cCOD<600 > ; ; 0.42; 0.028; 0.29; 0.10; ; 0.15; 0.10 Membrane fouling has been correlated to the dewaterability of the mixed liquor (Huang et al., 2013). Dewaterability typically is easier to measure than direct measurement of potential foulants such as ccod. In the current study the dewaterability of the mixed liquor was measured by the CST test and the results are presented in Figure 7. From Figure 7 it can be seen that the trend of CST values closely followed that of the ccod concentrations. Linear regression of the CST values against the ccod concentrations demonstrated a strong linear relationship (r 2 =0.81) between these parameters. As previously discussed the trends in ccod concentrations were not consistent with the trends in long term fouling rates. Hence there was a similar lack of consistency between the CST values as a measure of the dewaterability of the mixed liquor and the fouling responses. With the exception of the apparent relationship between fouling rate and TSS concentration in the first 50 days of the study, the apparent lack of relationship between mixed liquor properties and long term membrane fouling was inconsistent with previously reported results (Lin et al., 2009). The previously described results indicated that long term fouling rates did not correspond to the mixed liquor properties. Hence, it was hypothesized that the higher fouling rates that were observed after recovery cleaning, when compared to that of virgin membranes, was caused by the presence of irrecoverable foulants that resisted recovery cleaning. The presence of these irrecoverable foulants appeared to result in more rapid accumulation of foulants on the cleaned membrane when compared to that of the virgin membranes. To further explore the presence of irrecoverable fouling after cleaning, clean water flux (CWF) tests were conducted before and after each recovery cleaning. The flux values and the corresponding TMPs were analyzed by linear regression to estimate the resistances of the fouled and cleaned membrane according to Darcy s Law (Equation 1). R= (ΔP/J)/µ (Equation 1) where R is the resistance, ΔP is TMP, µ is the dynamic viscosity and J is the membrane flux. The resistances estimated from the clean water tests of the fouled, the recovery cleaned, and the virgin membranes are presented in Figure 8. From Figure 8 it can be seen that, in all cases, recovery cleaning resulted in a substantial reduction in membrane resistance after cleaning. This was consistent with the TMPs that were initially observed in the pilot plant after recovery cleaning. It is noteworthy that the resistance of the cleaned membrane after the fourth recovery cleaning (1.54*10 9 m -1 ), that employed a reversed cleaning sequence, was higher than that of the other recovery cleanings (1.00*10 9 m -1 ~ 1.15*10 9 m -1 ) indicating more irrecoverable foulants were left on membrane. This was consistent with the higher fouling rate that was observed in the pilot plant after the fourth recovery cleaning (0.61 kpa/d) as compared to the other cleanings (0.26~0.29 kpa/d). Although based on only this one episode, the results suggest the importance of following the recovery cleaning protocol to maintain membrane performance. To further quantify the effectiveness of the cleaning, a resistance model was employed.

17 Figure 8 Cleaning Efficiencies and Membrane Resistances The total resistance of the fouled membrane (RT) was fractionated into the virgin membrane resistance (RM), the fouling resistance removed by recovery cleaning (Rre) and the fouling resistance not removed by the recovery cleaning (Rir) (Equation 2). RM and RT values were estimated from the CWF tests on the virgin and fouled membrane respectively. Rre was calculated as the difference in resistance between the fouled membrane and the cleaned membrane. Rir was calculated as the difference in resistances between the cleaned membrane and the virgin membrane. RT =RM + Rre+Rir (Equation 2) where RT= Total fouled membrane resistance (m -1 ), RM= Virgin membrane resistance (m -1 ), Rre = Fouling resistances removed by recovery cleaning (m -1 ), Rir = Fouling resistance not removed by the recovery cleaning (m -1 ). The total cleaning efficiency (CETotal) provides evidence of the effectiveness of recovery cleaning. CETotal values were calculated according to Equation 3 (Figure 4) and they indicated over 80% of the fouling resistance was removed by the recovery cleaning despite having considerably different initial resistances. The high cleaning efficiencies suggested most of the foulants were removed from the membrane, but there was still approximately 10%~20% of the fouling resistance caused by irrecoverable foulants. These irrecoverable foulants caused low resistance as indicated by the low initial TMPs when either mixed liquor or clean water was filtered suggesting that a majority of the membrane pore spaces were recovered in cleaning. However, the irrecoverable foulants appear to have modified the conditions at the membrane surface (i.e. surface charge) such that foulants accumulated to produce fouling more rapidly on the cleaned membranes than on the virgin membranes. CETotal= Rre/(Rre+Rir)* 100% (Equation 3) Similar cleaning efficiency values and fouling patterns were reported by Kim et al., (2011) after recovery cleaning that involved sequentially soaking a membrane in NaOCl and NaOH when a bench scale AnMBR was operated at infinite SRT. In the study of Kim et al., fouling was significantly mitigated by

18 reducing the SRT from an infinite value to 25 days to reduce the foulant concentrations in the mixed liquor. The higher fouling rates after recovery cleaning in the current study may have been due to the use of SRTs in the range of 100 to 40 days that retained higher foulant concentrations. Viewed collectively, the results of this study indicate the potential to operate AnMBRs on municipal wastewaters for extended periods of time at fluxes that are typical of industry practice in aerobic MBRs. This was achieved with MLSS concentrations that consistently exceeded the vendors recommended operating values. Operation at reduced SRT (i.e. 40 days) that reduced the concentrations of TSS and ccod in the mixed liquor appeared to reduce the mixed liquor fouling propensity. However, the accumulation of irrecoverable foulants after recovery cleaning operations can be challenging for long term operation. This accumulation appeared to result in substantial fouling even when the indicators of fouling propensity of the mixed liquor were positive. Additional characterization of the nature of the irrecoverable foulants and the development of strategies to minimize their accumulation would further the advancement of the AnMBR technology for this application Impact of FeCl 3 Dosing Prior to Phase 1 the pilot AnMBR was operated at an HRT of 8.5 hours and SRT of 70 days for 5 months without FeCl3 addition. At this time a virgin membrane was installed and all of Phase 1 was assumed to be at steady state as indicated by near constant VSS concentrations (Mean±SD: 5.6 ± 0.6 g/l). During Phase 3 the AnMBR system was operated with FeCl3 addition to the influent at 26mg/L for 3 months. The membrane was then replaced with a virgin membrane 40 days after Phase 3 was initiated. The VSS in the first 75 days after the virgin membrane was installed in Phase 3 stabilized at 12.8 ± 2.0 g/l, indicating that reasonable steady-state operation was achieved. However, from Day 76 to 90, the VSS increased because the sewage COD increased by approximately 24% (p<0.009). Therefore, the data collected in the first 75 days with virgin membrane was deemed to represent steady state and was employed in the subsequent analysis. The membrane performance was assessed by examining the TMP profiles after relaxation (Figure 9). The TMP values may be employed to characterize fouling since the AnMBR was operated at a constant flux of 17LMH. Changes in TMP were assumed to be indicative of the accumulation of foulants on the membrane surface that withstood scouring stress and was attributed to the formation of either a strongly-attached cake layer or pore blocking. The impact of FeCl3 dosing on membrane performance is shown in Figure 9. From Figure 9 it can be seen that without FeCl3 addition TMP values were in the range of 5.7~6.3 kpa for the first 45 days but then increased to 21.5 kpa by day 91. In contrast, with FeCl3 addition TMP values were in the range of 1.5~5.1 kpa for the first 75 days and then increased to 8.8 kpa by day 90. The results indicate that the addition of FeCl3 mitigated the development of foulants that withstood scouring stress and hence low fouling rates were maintained over an extended period. The mitigated fouling with FeCl3 addition may have been due to the coagulation of colloidal and soluble substances in the mixed liquor as these have been reported to directly contribute to foulants that withstood scouring stresses (Lee et al., 2001; Wu et al., 2008; Meng et al., 2009). The results of the investigation of these foulants will be discussed subsequently.