Effects of C/N ratio on the performance of a hybrid sponge-assisted aerobic moving bed-anaerobic granular membrane bioreactor for municipal wastewater treatment

This study aimed to evaluate the impact of C/N ratio on the performance of a hybrid sponge-assisted aerobic moving bed-anaerobic granular membrane bioreactor (SAAMB-AnGMBR) in municipal wastewater treatment. The results showed that organic removal efficiencies were above 94% at all C/N conditions. Nutrient removal was over 91% at C/N ratio of 100/5 but was negatively affected when decreasing C/N ratio to 100/10. At lower C/N ratio (100/10), more noticeable membrane fouling was caused by aggravated cake formation and pore clogging, and accumulation of extracellular polymeric substances (EPS) in the mixed liquor and sludge cake as a result of deteriorated granular quality. Foulant analysis suggested significant difference existed in the foulant organic compositions under different C/N ratios, and humic substances were dominant when the fastest fouling rate was observed. The performance of the hybrid system was found to recover when gradually increasing C/N ratio from 100/10 to 100/5.

membrane fouling, thus exacerbating the long-term performance of G-AnMBRs (Ozgun et al., 2015;Chen et al., 2017a). The low-cost polyutrethane sponge, an ideal attached growth mobile carrier, has been successfully applied in many aerobic membrane bioreactors (AMBRs) studies to enhance the overall performance of AMBRs due to its high internal porosity and specific surface area, high stability to hydrolyse and light weight (Guo et al., 2010). Chen et al. (2017a) worked on a sponge-assisted G-AnMBR (SG-AnMBR), and indicated that sponge addition into G-AnMBR could enhance organic and nutrient removal, and maintain superior granular quality. Additionally, sponge media could not only positively affect the concentration and properties of microbial products (e.g. SMP and extracellular polymeric substances (EPS)) in granular sludge, cake layer as well as settling zone mixed liquor, but also reduce fouling resistance by 50.7%, thereby alleviating membrane fouling.
Although studies have proved that the sponge addition could improve nutrient removal (Nguyen et al., 2011), nutrient removal efficiencies were still considered quite low in the SG-AnMBR (Chen et al., 2017a), limiting its universal appeal for municipal wastewater treatment (Smith et al., 2012). Additionally, adopting conventional biological nutrient removal technologies at the downstream of SG-AnMBRs was also not feasible since low C/N ratio in SG-AnMBR effluents inhibited denitrification and phosphorus removal processes due to insufficient organic electron donor presented. Thus, C/N ratio is one of the most influential parameters affecting nutrient removal process as it affects the population and biodiversity of functional microorganisms (Lin et al., 2016). Moreover, membrane fouling can be significantly influenced by C/N ratio because C/N ratio profoundly affects the physiological property of microorganisms and chemical composition of biomass, and influences the concentrations of EPS and SMP and their protein and polysaccharides contents (Hao et al., 2016).
In this study, a new hybrid sponge-assisted aerobic moving bed-anaerobic granular membrane bioreactor (SAAMB-AnGMBR) was developed to overcome the two major issues (i.e. fouling and low nutrient removal) impeding the progress of G-AnMBRs. Based on the literature, it is the first development of the hybrid configuration for enhancing nutrient removal and fouling control of G-AnMBR during municipal wastewater treatment. The main aim of this study was to evaluate the effects of C/N ratio on the performance of such a hybrid system in terms of pollutants removal (particularly for nutrient removal) and membrane fouling. The system recovery after the overloaded nitrogen event was also evaluated in the study.

Experimental set-up and operation conditions
The hybrid SAAMB-AnGMBR, consisting of a sponge-assisted aerobic moving bed reactor (SAAMBR) and a submerged sponge-assisted anaerobic granular membrane bioreactor (SS-AnGMBR), was continuously operated for 282 days in a temperaturecontrolled room (20 ± 0.5 ºC). Each of the SAAMBR and the SS-AnGMBR had effective working volume of 3 L, and sponge fraction was 20% of working volume. At the bottom of the SAAMBR, fine bubble diffuser was set to supply air in order to provide complete liquid-solid mixing and moderate sponge up/down motion, and maintain dissolved oxygen (DO) concentration of 3.5 -4.8 mg/L. Prior to continuous operation, the SAAMBR with fresh sponge was acclimatized to synthetic wastewater for 30 days at HRT of 12 h until the system reached relatively stable treatment performance. The attached growth on the sponge also reached steady state at 1.02 ± 0.04 g MLVSS/g sponge. The sponges and anaerobic granular sludge were acclimatised to synthetic wastewater for 30 days until a stable treatment performance was reached.
The SS-AnGMBR was seeded with anaerobic granular sludge with initial MLSS concentration of 20.12 ± 1.21 g/L, and biomass grown on sponge cubes after acclimatization was 1.78 ± 0.09 g MLVSS/g sponge. A polyvinylidence (PVDF) hollow fiber membrane module with a pore size of 0.22 μm and surface area of 0.06 m 2 was immersed in the settling zone of the SS-AnGMBR.
The SS-AnGMBR was continuously fed with synthetic wastewater at a flow rate of 4.17 mL/min while wastewater from the SS-AnGMBR was continually transferred into the SAAMBR at the same flow rate. The SAAMBR effluent was recirculated back to the SS-AnGMBR through a nitrogen gas sparged buffer tank. The permeate pump in the SS-AnGMBR was operated in an intermittent mode with relaxation (8 min on and 2 min off) to acquire permeate from the membrane module with a constant filtration flux of 5.21 LMH. Both SAAMBR and SS-AnGMBR had HRT of 12 h, and upflow velocity in the SS-AnGMBR was maintained at 3.2 m/h using internal recirculation. The membrane fouling propensity was indicated by normalized trans-membrane pressure (TMP), which was recorded by a pressure transmitter. Operation was terminated when TMP exceeded 30 kPa, and fouled membrane was taken out for ex situ cleaning (Deng et al., 2016a).
The entire study period was divided in 2 phases according to the research objectives.
In phase 1, the hybrid SAAMB-AnGMBR was fed with wastewater having C/N/P ratio = 100/5/1 (0 -75 day), 100/6/1 (76 -126 day), 100/8/1 (126 -151 day) and 100/10/1 (151 -166 day), respectively, with the aim to investigate the impact of C/N ratios on the performance of the hybrid system. In phase 2, after overloaded nitrogen events, the hybrid system was operated with C/N/P ratios of 100/6/1 (167 -210 day) and then 100/5/1 (211 -282 days) to investigate the extent of system recovery after the overloading nitrogen event. The detailed measurement protocol of membrane fouling resistances including total resistance (R T ), intrinsic membrane resistance (R M ), cake layer resistance (R C ) and pore blocking resistance (R P ), respectively, was followed by methods reported in Deng et al. (2015). Membrane fouling resistances were determined using the resistance-inseries model (Choo and Lee, 1996). The extraction and analysis of EPS and SMP in the granular sludge, mixed liquor and cake layer were performed based on the protocol of were obtained using a Varian Cary Eclipse Fluorescence Spectrophotometer, USA, according to methods suggested by Hong et al. (2012). EEM spectra were plotted as the elliptical shape of contours using software OriginPro 9.1. All the liquid, gas and sludge samples were tested in triplicate, with an average value and standard deviation for discussion.

Overall treatment performance
The hybrid SAAMB-AnGMBR system showed sound organic removal even though influent C/N ratio varied. COD and DOC removal efficiencies were kept over 94.8% and 94.5% regardless of C/N ratios ( Table 1). The results suggested that decreasing C/N ratio from 100/5 to 100/10 did not adversely influence the organic removal since complete retention of all particulate COD and macromolecular COD components was achieved by the membrane and most of the readily biodegradable COD was removed by aerobic and anaerobic biodegradation (Martinez-Sosa et al., 2011).

Table 2
When C/N ratio was maintained at 100/6 or higher, more than 84% of TN removal was achieved (Table 1). Due to effective simultaneous nitrification and denitrification (SND) process in the hybrid system, effluent NO 3 -N concentrations were maintained at low levels with 0.39 ± 0.15 and 1.63 ± 0.54 mg/L at C/N ratios of 100/5 and 100/6. Table 2 showed the SDR values for each type of functioning biomass in the SAAMB-AnGMBR system. Anaerobic granular biomass exhibited the highest SDR, hence contributing the most to the denitrification process. SDR of aerobic spongeattached biomass were found comparable to that of anaerobic sponge-attached biomass, although DO in the aerobic compartment was maintained at 3. However, lower C/N ratios of 100/8 and 100/10 showed significant reduction on TN removal efficiencies to 68.9% and 44.1%, respectively. Biological nitrogen removal is achieved by nitrification (an autotrophic bioprocess), followed by denitrification (a heterotrophic bioprocess) (Kim et al., 2016). Since the NH 4 -N removal efficiencies were kept more than 92.5% at all designated C/N ratios, the low TN reduction was mainly attributed to the poor denitrification when insufficient carbon was present for heterotrophic denitrifier for cell growth and nitrate reduction .
Insufficient carbon source restrained the denitrification process, caused relatively higher NO 3 -N concentrations in the effluent (5.33 ± 1.08 and 12.46 ± 1.19 mg/L for C/N ratios of 100/8 and 100/10, respectively) and broke the balance between nitrification and denitrification processes (Lin et al., 2016). Low carbon source not only negatively affected the abundance of heterotrophic population but also significantly influenced vital parameters of denitrification process such as SDR. SDR of granular biomass was significantly reduced when decreasing C/N ratio to 100/8 and 100/10 ( Table 2). This finding might be related to the deteriorated granule quality (details discussed in Section It is interesting to observe that more VFAs were accumulated with increased nitrogen content in the feed, and total VFAs concentration for each C/N ratio was 2.77 ± 0.56 (100/5), 5.72 ± 1.12 (100/6), 9.38 ± 1.02 (100/8) and 11.51 ± 1.28 (100/8) mg/L. Acetic acid was found the most sensitive to the change of C/N ratio, whose concentration at C/N ratio of 100/10 (8.92 ± 0.95) was nearly 4 times higher than that at C/N ratio of 100/5 (1.82 ± 0.35 mg/L). Higher levels of residual VFAs, especially for acetic acid at lower C/N ratio, might indicate the inhibition of denitrification and phosphate removal processes. In addition, decreasing methane yields were found at 137.40 ± 3.83, 126.58 ± 3.07, 115.25 ± 2.29 and 102.12 ± 1.92 mL CH 4 (STP) /g COD removed (STP: volume of methane produced at and 0 °C Standard Temperature and 1 atm Pressure) at C/N ratios of 100/5, 100/6, 100/8 and 100/10, respectively.

EPS and SMP of the mixed liquor
In the SS-AnGMBR, the membrane module was only challenged by the mixed liquor in the settling zone. Hence, the EPS and SMP concentrations of the mixed liquor and their polysaccharide (EPS C , SMP C ) and protein contents (EPS P , SMP P ) were analysed in order to explain the relationship between the mixed liquor properties and fouling propensity, and the results were shown in Table 3. EPS at C/N ratio of 100/10 (48.41 mg/L) were found more than 4.27, 2.04, 1.29 times to the corresponding values obtained at C/N ratios of 100/5, 100/6, 100/8 (11.33, 23.66, 37.40 mg/L), respectively.
At the lowest C/N ratio (100/10), the SS-AnGMBR contained the highest EPS content, which was mainly responsible for the highest fouling rate. Studies have reported that EPS were biopolymers attached on flocs or cells surface, and were known to SMP, known as biopolymers released from microorganisms into solution, has the strongest relationship with fouling rate . In this study, despite the increase in N concentration in the feed, SMP showed the relatively low concentrations (5.58 -8.08 mg/L) at all four C/N ratios. This might be due to the fact that sponge addition into hybrid system could significantly reduce the mixed liquor SMP values (Deng et al., 2014). The highest SMP value was observed at C/N ratio of 100/5 (8.08 mg/L), followed by those at C/N ratios of 100/6 (7.53 mg/L), 100/8 (6.85 mg/L) and 100/10 (5.58 mg/L). The higher SMP levels at higher C/N ratios in the hybrid system might be due to the SMP accumulation with evolution of time in the mixed liquor since prolonged operational time was recorded at higher C/N ratios (i.e. the time slots were 75, 50, 25, 15 days for C/N ratios of 100/5, 100/6, 100/8, 100/10, respectively). In this case, the total SMP concentration was not responsible for faster fouling at lower C/N ratios. Nevertheless, an increase in SMP C /SMP P was positively correlated with faster fouling rate at lower C/N ratio. The averaged values of SMP C /SMP P increased from 0.54 to 0.79, 1.14, and 1.44 when lowering C/N ratios from 100/5 to 100/6, 100/8 and 100/10, respectively, and it suggested that higher amount of hydrophilic SMP C than SMP P were presented in the mixed liquor at lower C/N ratios. SMP C were more susceptible to membrane fouling since they could penetrate into the cake layer and membrane pore and lodge insides, thus inducing severe pore clogging and gel layer formation, and causing significant membrane filterability loss (Deng et al., 2016a).

EPS and SMP of the anaerobic granular sludge
According to Table 3, both EPS P and EPS C concentrations of the granular sludge were noticeably reduced when influent N was increased. The lowest EPS concentration (EPS P and EPS C of 4.92 and 2.18 mg/L, respectively) was observed at C/N ratio of 100/10. Since EPS played an vital role on integrating cells into granules, the dramatic decrease in EPS content at lower C/N ratio might indicate weaker, scattered and looser structures of flocs and granules, contributing to granules breakage (reduced sludge particle size) and to the increase in EPS of the mixed liquor in settling zone (Chen et al., 2017a). The granular growth rates (ΔMLSS/Δt) were found at 0.052, 0.045, 0.038 and 0.035 g/L d, at C/N ratios of 100/5, 100/6, 100/8 and 100/10, respectively, suggesting that lower C/N ratio discouraged the growth of retained sludge agglomerates in the granular sludge bed. PSD analysis confirmed that smaller anaerobic granular sludge particles existed when operating at lower C/N ratios. Granular sludge with the biggest median particle size of 355 μm was observed at C/N ratio of 100/5, while lower values of 308 μm, 223 μm and 165 μm were found at C/N ratios of 100/6, 100/8 and 100/10, respectively. As a result, the disintegration of granules decreased settling capacity of the granular sludge. AnGMBR reached up to 293.8 mg/L at C/N ratio of 100/10, which was nearly 1.5, 2.7 and 3.9 times higher than those at C/N ratios of 100/8, 100/6 and 100/5, respectively. Moreover, the highest SMP level in the granular sludge (SMP P and SMP C of 14.21 and 2.12 mg/g VSS) was observed at C/N ratio of 100/10 (Table 3), which confirmed that the major fraction of the proteins and polysaccharides existed in the soluble form rather than being the part of the anaerobic granules in the form of EPS. These results highlighted that severe fouling at lower C/N ratios was not only associated with larger amounts of EPS, higher SMP C /SMP P ratio of mixed liquor but also due to the deteriorated granule quality.

Fouling resistance and cake layer analysis
Membrane fouling resistances were obtained for each of C/N regimes when cleaning of membrane was performed (Fig. 1). With increasing N dose in the feed, R T increased from 2.61 × 10 12 to 3.65 × 10 12 m -1 at C/N ratios of 100/5 and 100/6, which further rose to 5.62 × 10 12 and 6.56 × 10 12 m -1 at C/N ratios of 100/8 and 100/10, respectively. R C and R P values followed the similar trends as shown in Fig. 1 in which R C (2.03 × 10 12 to 5.19 × 10 12 m -1 ) increased with a greater magnitude than R P (3.9 × 10 11 to 1.18 × 10 12 m -1 ), while R M (1.89 × 10 11 to 1.92 × 10 11 m -1 ) remained stable for all C/N ratios. These results suggested that higher N dose in the feed deteriorated membrane permeability due to more cake formation and pore clogging. Although R p contributed to declined permeability, the values of R C /R T (over 77% at all C/N ratios) indicated that cake formation was the predominant fouling mechanism of the hybrid system, contributing more significantly to membrane fouling. As discussed in Section 3.2.2, the mixed liquor containing SMP with higher SMP C /SMP P ratio and higher amounts of EPS was responsible for the elevated R C and R P at lower C/N ratios. More EPS resulted in cake formation whist SMP with higher SMP C /SMP P ratio could modify the surface properties of membrane and outer cake layer to promote the self-accelerating fouling phenomena (accelerate cake formation rate) (Hao et al., 2016). Cake layer as the dominant mechanism of R T was further investigated and characterized by the composition of EPS and SMP. As shown in Fig. 2, the cake layer at C/N ratio of 100/5 contained the lowest concentration of EPS P and EPS C . The reduction of C/N ratio to 100/10 induced a noticeable rise of EPS P and EPS C to 10.95 mg/g MLSS and 25.59 mg/g MLSS, respectively , which were 6.0 and 7.8 times to the corresponding values obtained at C/N ratio of 100/5. The results revealed that higher nitrogen dose increased R C by the accumulation of EPS within the sludge cake on the membrane surface. On the other hand, in spite of varying C/N ratios, SMP presented significantly low concentrations and minor variations, with SMP P of 0.47 -0.52 mg/g MLSS and SMP C of 0.87 -1.08 mg/g MLSS, which further confirmed that SMP were not the primary fouling factor for this study as discussed in Section 3.2.2.

Foulant characterization
LC-OCD provides critical information regarding the different components of the foulant organics by dividing the total organics into two groups: hydrophobic organics (HPO) and hydrophilic organics (HPI). The hydrophilic group can be further characterized into subgroups such as biopolymers, humic substances, building blocks, and lower molecular weight (LMW) neutrals and acids (Johir et al., 2012). Fig. 3 showed that larger HPO occurred at lower C/N ratios, but they might not be the main foulants due to their much lower concentrations as compared to those of HPI. Within HPI, lower nitrogen dose favoured the accumulation of biopolymers in the foulant while humic acids were dominant at higher N dosage, contributing to the faster fouling rate.
Building blocks, as the breakup of the humics, were found decreasing in response to the increase in influent nitrogen content. Though building blocks and LMW neutrals and acids were found in relatively lower concentrations, Aryal et al. (2009) suggested that they were the significant factors affecting fouling as their assemblage could promote biopolymers formation on the membrane surface, exacerbating fouling propensity. nm, respectively, which suggested that foulant formed at such C/N ratios contained higher content of humics but less biopolymers. The EEM results were found in agreement with the LC-OCD results in the sense that the higher N dosage at lower C/N ratio favoured the relative dominance of the humic-like substances in the extracted foulant, resulting in faster fouling. Similar results were reported by Hong et al. (2012) that humics were prevalent when the fastest fouling was observed.

System recovery from high nitrogen loading
As mentioned above, the performance of hybrid system including nutrient removal and membrane filtration was deteriorated in phase 1 as a result of the decrease in C/N ratios from 100/5 to 100/10. Therefore, the hybrid system was further investigated in phase 2 by gradually increasing C/N ratio back to 100/5 to observe the extent of system recovery after the overloaded nitrogen event.

Treatment performance of the hybrid system
The system performed reasonably well in term of organic and NH 4 -N removal in phase 2, with the comparable removal efficiencies of DOC and COD over 95.2% and NH 4 -N over 93.5%. On the other hand, TN and PO 4 -P removal efficiencies (82.3 ± 8.8% and 85.7 ± 6.7%, respectively) were found significantly improved after C/N ratio was increased from 100/10 to 100/6. When further increasing C/N ratio to 100/5, the hybrid system could achieve TN and PO 4 -P removal efficiencies of 90.1 ± 6.8% and 92.1 ± 8.2%, respectively, indicating that the inhibitory effects on denitrification and phosphate removal processes imposed by overloaded nitrogen were not permanent. The system recovery in terms of nutrient removal was mainly attributed to efficient SND process and effective elimination of residual NO 3 -N in the system (Deng et al., 2016b). Table 4 showed that the fouling rate was significantly reduced when increasing C/N ratio from 100/10 to 100/6 (0.69 kPa/d) and then to 100/5 (0.47 kPa/d) in phase 2.

Membrane fouling behaviour analysis
As compared to fouling resistance at C/N ratio of 100/10, the hybrid system in phase 2 exhibited much lower R T values at C/N ratios of 100/6 and 100/5. During the recovery stage, EPS contents in the mixed liquor were maintained at 28.38 and 13.21 mg/L at C/N ratios of 100/6 and 100/5, respectively, which were about only 59% and 27% of the corresponding values obtained at C/N ratio of 100/10. The analysis of the cake layer also revealed that the hybrid system in phase 2 contained less EPS P and EPS C at C/N ratios of 100/6 and 100/5. Thus, increasing C/N ratio helped reduce membrane fouling mainly through reducing EPS in the mixed liquor and cake layer, thereby alleviating fouling (Chen et al., 2017a;Hao et al., 2016). Furthermore, lower SMP C /SMP P (0.88 and 0.59, respectively) under C/N ratios of 100/6 and 100/5 revealed less severe fouling in the hybrid system, compared with that at C/N ratio of 100/10. The major foulant organics in phase 2 were found to be biopolymers. The results also indicated that the impact of overloaded nitrogen on fast fouling propensity was temporary.

Conclusion
This study examined the overall performance of the SAAMB-AnGMBR operated at different C/N ratios. Decreased C/N ratio deteriorated TN and PO 4 -P removal, worsened the granular quality, and exacerbated cake layer formation and pore blocking, thereby aggravating membrane fouling. R C and R P increased as a result of increased EPS and SMP C /SMP P ratio in the mixed liquor. This study also revealed humics were the dominant organic foulant at lower C/N ratios. Additionally, the hybrid system could be recovered by increasing C/N ratio from 100/10 to 100/5 through improving SND process and the properties of mixed liquor and cake layer.

Supplementary information
Summary of sludge properties of seed sludge and granular sludge in the SS-AnGMBR at different C/N ratios, and EEM fluorescence spectra of membrane foulants at different C/N ratios can be found in the supplementary information.        Table 2.  Table 3.

Specific Nitrification rates and denitrification rates (SNR and SDR
The concentration and properties of EPS and SMP in the mixed liquor of the SS-AnGMBR settling zone and anaerobic granular sludge.