Pollutant Removal and Energy Recovery from Swine Wastewater Using Anaerobic Membrane Bioreactor: A Comparative Study with Up-Flow Anaerobic Sludge Blanket

: Due to its high content of organics and nutrients, swine wastewater has become one of the main environment pollution sources. Exploring high-efﬁcient technologies for swine wastewater treatment is urgent and becoming a hot topic in the recent years. The present study introduces anaerobic membrane bioreactor (AnMBR) for efﬁcient treatment of swine wastewater, compared with up-ﬂow anaerobic sludge blanket (UASB) as a traditional system. Pollutant removal performance, methanogenic properties, and microbial community structures were investigated in both reactors. Results showed that by intercepting particulate organics, AnMBR achieved stable and much higher chemical oxygen demand (COD) removal rate (approximately 90%) than UASB (around 60%). Due to higher methanogenic activity of anaerobic sludge, methane yield of AnMBR (0.23 L/g-COD) was higher than that of UASB. Microbial community structure analysis showed enrichment of functional bacteria that can remove refractory organic matter in the AnMBR, which promoted the organics conversion processes. In addition, obvious accumulation of acetotrophic and hydrotrophic methanogens in AnMBR system was recorded, which could broaden the organic matter degradation pathways and the methanogenesis processes, ensuring a higher methane yield. Through energy balance analysis, results concluded that the net energy recovery efﬁciency of AnMBR was higher than that of UASB system, indicating that applying AnMBR for livestock wastewater treatment could not only efﬁciently remove pollutants, but also signiﬁcantly enhance the energy recovery.


Introduction
Livestock breeding is an important growing industry to supply the world with the increased requirements of meat and nutrition. In 2020, approximately 850 million pigs were bred in China [1], which discharged a large amount of wastewater and solid wastes into the surroundings. Swine wastewater (SW) contains high content of organics, nutrients, antibiotics, and heavy metals [2][3][4][5], which result in serious water pollution and environmental deterioration if not properly managed. To reduce the negative impacts of SW, some physico-chemical methods have been adopted to recover organics and nutrients [6,7]. In addition, various microbial treatment systems such as aerobic, anaerobic, and even ecological systems were designed and applied into practice [8][9][10]. However, the high Although AnMBR systems have been extensively used in the treatment of wastewater with high organic matter concentration, research about SW treatment using AnMBR has been less conducted. Despite their advantages, comparing the potential of AnMBR in pollutant removal and methane production from SW with other conventional systems such as UASB was not systematically investigated. In addition, evaluation of energy production from SW in AnMBR and UASB is of great importance. In addition, comparing the microbial communities in both reactors has not been analyzed. Thus, in this study, AnMBR and UASB system were operated and compared. Pollutant removal efficiencies and energy recovery properties were investigated. The relationships between microbial community and reactor performance for pollutants removal and biogas production were also explored. The results of this study provide more information for future application of AnMBR in SW treatment and energy recovery.

Swine Wastewater and Inoculum Sludge
Swine wastewater (SW) was collected from a local pig farm, then particles bigger than 5 mm were removed from the SW using a mesh screen. The filtrate was stored at 4 • C until further use. pH value of the SW was 7.8, and total solid (TS) and volatile solid (VS) content were 7-8 and 4-5 g/L, respectively. Chemical oxygen demand (COD) and volatile fatty acids (VFAs) content were approximately 20 g/L and 5 g/L, respectively. The inoculum sludge was obtained from a full-scale mesophilic UASB reactor, which was used to treat the brewery wastewater. Mixed liquor suspended solids (MLSS) and mixed liquor volatile suspended solids (MLVSS) concentrations of the sludge were approximately 10.3 g/L and 6.3 g/L, respectively.

Reactors Design and Operation
The AnMBR and UASB were simultaneously conducted under the same operating conditions (Figure 1), with a total volume of 3.0 L and working volume of 2.6 L. AnMBR was a modified UASB with a flat-sheet microfiltration membrane module at the top of the reactor. The length and width of the membrane module was 12.8 cm and 7.8 cm, respectively. Membrane on both sides of the module frame was made of chlorinated polyethylene with a normal pore size of 0.22 µm and a total effective area of 0.02 m 2 . SW was fed into the bioreactor using a peristaltic pump (Longer BT-100, Longer, Baoding, China), which was connected with a water level sensor to maintain a constant water level in the reactors. Digestate was discharged using another peristaltic pump ( Figure 1). An online pressure sensor (SIN-P400, Sinomeasure, Hangzhou, China) located on the permeate line of AnMBR was used to monitor the trans-membrane pressure (TMP). The volume of biogas was measured by wet-type gas flow meters (LML-2, Changchun Automobile Filter Co., Ltd., Changchun, China). A water bath was conducted to maintain a mesophilic temperature (37 ± 2 • C) in both reactors. The feedstock was stored at 4 • C in a substrate tank continuously stirred and connected to a cooling water bath. To control the membrane fouling in AnMBR, membrane filtration was intermittently operated (8 min-on, 2 min-off) during the whole operating period [29,30]. In addition, backwash using the effluent was conducted to remove membrane foulants and recover its permeability when the TMP exceeded 40 kPa. Before starting this study, the reactors were operated for more than 110 days using SW. Hydraulic retention time in both reactors was approximately 10 days. During the operation, the influent, mixed liquor, and permeate were periodically sampled and analyzed.

General Parameters
Samples of influent, mixed liquor, and effluent were collected and centrifuged at 5000 rpm for 10 min at 4 • C. The supernatants were used to measure the pH, soluble COD (SCOD), and VFAs. Total solids (TS), VS, pH, and COD were analyzed in accordance with the American Public Health Association (APHA) standard methods [31]. Total and soluble carbohydrate and protein were detected using the methods described in our previous study [32]. Prior to determining VFAs and lactic acid, liquid samples were filtered through 0.22-µm filter membranes. The filtrate was analyzed by high performance liquid chromatography (LC-10AD, Shimadzu, Kyoto, Japan) equipped with an ultraviolet detector (210 nm) using sulfuric acid (5 mmol/L) as eluent. Biogas was collected in airbags, then its composition was regularly monitored using gas chromatograph (GC2010, Shimadzu, Kyoto, Japan) equipped with a packed column (TDX-01, Haohan, Tengzhou, China) and thermal conductivity detector (TCD). All analyses of individual samples were performed in triplicates, and methane production was analyzed by kinetic analysis using the modified Gompertz model [33].

Specific Methanogenic Activity (SMA) Analysis
Measurement of SMA for anaerobic sludge was carried out in 120 mL serum bottles using sodium acetate as a substrate (COD = 0.5 g/L) [34]. In SMA test, the feed/microorganisms (F/M) ratio was set at 0.5. The total working volume of each serum bottle was 80 mL, containing 20 mL sludge from the reactor, 50 mL substrate, and 10 mL nutrient solution. The nutrient solution was boiled for 0.5 h to remove dissolved oxygen and then cooled down to room temperature before use. Each bottle was purged with nitrogen gas for 2 min to remove the oxygen, sealed with rubber stoppers, and rapidly secured using an aluminum crimp. Finally, the above bottles were incubated on a shaker at 115 rpm and 25 ± 1 • C. After the set temperature was reached, the headspace of each bottle was vented using a syringe to release the pressure caused by thermal expansion. The production and composition of biogas were measured every 3-5 h and expressed as the value at the standard state. Each experiment was conducted in triplicates to ensure reliability.

Microbial Community Analysis
Sludge samples were collected from each reactor for microbial community analysis. The samples were centrifuged at 12,000× g for 15 min and then the supernatant was decanted. For each sludge sample, 0.45 g of the solid phase was collected and transferred into a commercial soil DNA isolation kit for DNA extraction. Thereafter, DNA for bacteria was amplified with universal forward primer 341F: CCTACGGGNGGCWGCAG and 805R: GACTACHVGGGTATCTAATCC. The DNA for Archaea was amplified with 340F: CCCTAYGGGGYGCASCAG and 1000R: GGCCATGCACYWCYTCTC in the first round and 349F: GYGCASCAGKCGMGAAW and 806R: GGACTACVSGGGTATCTAAT in the second round. After PCR amplification, minimization of random sequencing errors was performed by eliminating the low-quality sequences and those having a length shorter than 500 nucleotides. After the primers and barcodes were trimmed, the high-quality sequences were clustered into operational taxonomic units (OTUs) at a 0.03 cut-off. PCR amplicon sequencing analysis was executed at Sangon Inc. (Shanghai, China) based on the Illumina MiSeq platform.

Energy Balance Analysis
Energy consumption was calculated based on the equivalent electric power used for SW treatment, while energy production was calculating by the energy of produced biomethane. Energy consumption included the energy for influent pump and effluent pump during per ton wastewater treatment. In addition, the energy used for substrate mixing and water bath was calculated based on the power of the equipment. Energy recovery can be obtained according to the methane volume when one ton of wastewater was treated. The net energy was the result of energy produced minus the energy consumed. Energy consumption for both influent and effluent pumps during the operation was calculated based on a previous study [35]; In which P represents the energy needed for the pump (kW), Q is the flow rate of the pump (m 3 /s), γ is the water density (9800 N/m 3 ), and h is the length of water column (m); η represents the energy transfer efficiency from electrical energy to pump energy (65%).
Based on methane production for per volume SW treatment, the energy recovery was calculated using the following equation [36]; where V represents methane volume (m 3 ), α is the low calorific value of methane (0.222 kWh/mol-CH 4 ), η is the cogeneration efficiency (33%), and Q is the volume of rerated wastewater (m 3 ).

COD Removal
COD removal properties in both reactors during the operation are shown in Figure 2a. It can be clearly noted that although high COD concentration (approximately 20 g/L) was observed in the influent, it was relatively stable and lower than 5 g/L in the effluent of the AnMBR at the beginning and gradually decreased to less than 2 g/L, showing a removal rate around 80-90%. However, COD content in the effluent of UASB was much higher and retained at approximately 10 g/L, achieving COD removal efficiency at 40-50%. Therefore, with the assistance of membrane separation, AnMBR exhibited better COD removal capacity, which ensured higher pollutant removal and more energy production [37]. The lower pollutant removal rate and higher COD content in the effluent of UASB further verified the advantages of membrane filtration in anaerobic bioreactors. As shown in Figure 2b, the components of COD varied along the reactors. In the influent, SCOD accounted for approximately 68.4% of the total COD, which decreased to 41.6% in the mixed liquor, indicating that soluble organics were mainly utilized by microorganisms in the anaerobic sludge. However, particulate organic matter was not effectively biodegraded, but retained in the reactor and accounted for 58.4% of the total COD in the mixed liquor. After membrane filtration, with a sharp decrease of COD content from 15.1 g/L to 2.7 g/L, SCOD became the dominant component (accounting for 90.9% of the total COD in the effluent), which verified that most of particulate organics were intercepted by the membrane. It has been reported that membrane and cake layer can intercept most of the particulate or macromolecular organics in the mixed liquor and guarantee low COD concentration in effluent [38]. In addition, with the assistance of membrane separation, some functional microorganisms can be easily accumulated in the reactor and contribute to degrade pollutants easily and rapidly [30,39]. Unsurprisingly, the proportion of SCOD in the mixed liquor and effluent of UASB was almost the same, and COD content retained constant, which further verifies that the high COD removal rate of AnMBR was mainly due to the interception of particulate organics.

Volatile Fatty Acids
The changes in VFAs content in the reactors are shown in Figure 3a. In the influent, VFAs content was approximately 6 g/L, while it sharply decreased to 2.3 g/L in the mixed liquor of AnMBR, showing that VFAs in the influent was effectively utilized by microorganisms during the operation. At the early stages of operation, VFAs in the effluent of AnMBR was relatively higher, which may be due to the fact that the microorganisms did not fully adapt to the operation conditions. However, with the domestication of microorganisms, VFAs content in the effluent of AnMBR gradually decreased and maintained below 1 g/L. Although membrane module had a lower ability to intercept VFAs due to their low molecular weight, concentration of VFAs in the effluent was much lower than that in the mixed liquor, which can be attributed to the microbial degradation. The decrease of VFAs content after membrane filtration ensured satisfactory methane production and high-quality effluent. In UASB, VFAs content in the mixed liquor and effluent was very close (approximately 1.5 g/L) and was much higher than that of AnMBR, indicating that the implantation of membrane was also favorable for VFAs degradation due to functional microorganisms enriched in the reactor through membrane interception. In VFAs, acetate accounting for 39.9% of the total VFAs was the dominant components in the influent (Figure 3b), which may be produced from metabolisms of indigenous bacteria in SW. In addition, propionate with a proportion of 30.3% was also a dominant acid. Other VFAs, such as butyrate and valerate, were also detected, even with low proportions. However, in the mixed liquor and effluent of AnMBR, propionate with a proportion of 44.9% became the dominant VFA and was higher than that of acetate, which might be attributed to two reasons: firstly, propionate was largely produced during the operation and accumulated in the reactor. Secondly, degradation rate of propionate was lower than that of acetate, which is in consistent with the previous studies [40,41]. The variations in VFAs of UASB were similar to those of AnMBR, which might result from similar microbial community structures in the reactors, as discussed in the next section.

Proteins and Carbohydrates
Proteins and carbohydrates are the dominant organics in SW, which are favorable substrates for methane production. It was found that protein in the influent was about 4 g/L, but it sharply decreased to 1.5 g/L in the mixed liquor (Figure 4a), indicating that protein was effectively utilized by microorganisms in the reactors. Due to the fact that the membrane can intercept protein, the protein content in the effluent of AnMBR further decreased to less than 0.5 g/L. In addition, microorganisms in the cake layer may also contribute to protein degradation. Thus, protein in the effluent of AnMBR was much lower than that of UASB. Carbohydrate content along the reactor showed a similar trend to that of protein (Figure 4b). However, carbohydrate content in the mixed liquor of AnMBR was much higher than that of UASB, which might be due to the fact that microorganisms in AnMBR had higher hydrolysis rate than those in the UASB and/or membrane exhibited better interception for carbohydrates. The effective interception of proteins and carbohydrates provided favorable conditions for methane production and further guaranteed lower COD content in the effluent.

Methane Yield
Changes in methane production and content during the operation are shown in Figure 5. It can be noted that daily methane production in AnMBR was approximately 1.3-1.4 L, while it was approximately 1.0-1.1 L in the UASB, indicating that AnMBR could obviously enhance the methane yield. It was reported that membrane can intercept organics in the reactor, which lengthens the retention time of organics in the reactor and promotes the microbial degradation processes [42]. In addition, with the assistance of membrane filtration, functional microorganisms can be easily enriched in the anaerobic sludge and contribute to enhance the methane processes. As shown in Figure 5, methane content in the biogas during the operation of AnMBR was around 83%, which was slightly higher than that of UASB (82%). In AnMBR, the average methane yield was approximately 0.23 L/g-COD during the operation, while methane yield of UASB was only 0.19 L/g-COD. It was deduced that macromolecule or refractory organic matter can be effectively trapped by the membrane [42,43], which can prolong the residence time and provide favorable conditions for efficient conversion of organic matter. In addition, membrane module contributes to the enrichment of functional microorganisms, which increases microbial activity and thus promotes methane yield.

Specific Methanogenic Activity
It was found that methane content of AnMBR was higher than that of UASB, and the methane yield of AnMBR was also higher during the operation, which might be due to the differences in the activity of microbial metabolism. Therefore, specific methanogenic activity (SMA) of anaerobic sludge was analyzed. As shown in Table 1, the maximum methane yield and SMA value of the sludge from AnMBR were 1.02 mL-CH 4 /h and 5.71 mg-COD/g-VSS.h, respectively, which were much higher than the maximum methane yield of UASB sludge (0.83 mL-CH 4 /h) and SMA value (5.04 mg-COD/g-VSS.h). In addition, the retardation time (R) of methane production of sludge from AnMBR was approximately 6.9 h, which was significantly shorter than that of UASB sludge (7.3 h). These results indicate that AnMBR sludge had higher methane production activity, which was further confirmed by the higher methane yield in AnMBR.

Microbial Community Analysis
To reveal the relationships between methane production and microbial communities, high throughput sequencing was conducted. A total of 108,768 raw 16S rRNA sequences (longer than 500 bp) were obtained from the two sludge samples. As shown in Figure 6, samples from UASB contained reads of 52,226, while samples from AnMBR showed more reads (56,542) for bacteria. The rarefaction curves tended to approach the saturation plateau. It was found that the 711 OTUs were found in the samples from AnMBR, which was higher than that from UASB (612), indicating the higher microbial diversity in AnMBR. The rarefaction curves of Achaea in both samples showed a similar trend, and 53 OTUs were found in the sludge from AnMBR, which also exhibited richer microbial communities than that from the UASB. Microbial diversity can also be verified by the Shannon diversity indices, ACE, and Chao index as shown in Table 2. The coverage estimator of all samples was higher than 0.99, indicating that these libraries well-represent the majority of bacterial or archaeal 16S rRNA in each sample. Shannon index can reflect the diversity of microbial communities [44], where the higher recorded Shannon index AnMBR samples represents more complexity of the microbial community. In addition, Chao and ACE indices indicate the bacterial and archaeal richness of the sample [45]. Chao and ACE indices of sludge samples from AnMBR were also higher than that of UASB, showing richer microbial communities in the AnMBR system. Even though some bacteria might be still missed as a result of methodological limitations, the size of reads is sufficient to reveal most of the dominant microbial species in the sludge. The higher microbial diversity in AnMBR was beneficial for system stability and high pollutant removal efficiencies, which explained the higher COD removal efficiency and methane yield during the operation.  In order to explore the biological mechanisms of higher pollutant removal rate and methane yield in AnMBR than those of UASB, bacteria in the anaerobic sludge were analyzed. It was found that Firmicutes, Proteobacteria, Bacteroidetes, Actinobacteria, and Atribacteria are the dominant bacterial phyla in the sludge of both AnMBR and UASB (Figure 7). Firmicutes with a relative abundance of 80.7% in UASB was much higher than that in AnMBR (56.9%), while Proteobacteria and Bacteroidetes in the samples from AnMBR accounted for 30.3% and 5.4%, respectively, which were higher than those in UASB (11.2% and 2.3%, respectively). These results indicate that with the participation of membrane interception, microbial community structure in the anaerobic sludge was significantly different. It was reported that Firmicutes, Bacteroidetes, Actinobacteria, and Proteobacteria are commonly recognized as hydrolytic fermentative bacteria in AD process. These bacteria are involved in hydrolysis phase of AD process and convert complex macromolecular organics into biodegradable molecules [46,47], and thus provide favorable substrates for methanogenic processes. Synergistetes are able to digest monocarboxylic and long chain fatty acids (butyrate, isoheptanoate, oleate, etc.), and produce hydrogen, acetate, and CO 2 for methane production [48]. The higher relative abundance of this phylum in AnMBR is beneficial for enhanced digestion of macromolecules, which resulted in higher efficiency of pollutant removal and methane production.
On the genus level, Tissierella accounted for 24.1% and 25.0% in AnMBR and UASB, respectively, which can efficiently degrade protein, glucose, and produce VFAs with enhanced methane production [49]. In UASB, Sporosarcina with a relative abundance of 36.6% was the dominant genus and can secrete urease and degrade urea in the SW, which may result in high content of ammonia accumulation in the reactor and pose risks to methanogenic microbes [50]. Clostridium_sensu_stricto plays an essential role in hydrolyzing proteins and carbohydrates [51], and was more abundant in UASB (11.0%) than AnMBR (4.7%), indicating that microbial communities in UASB inclined to consume easily biodegradable organics. Psychrobacter was effectively enriched in the AnMBR (21.9%) during the operation, which is favorable for complicated organics degradation and widened the utilizable substrates. It was reported that Psychrobacters can utilize many organic compounds as sole carbon and energy sources for growth and can survive in a wide temperature range and organic loading conditions [52], which is beneficial for system stabilization. Other genera such as Terrisporobacter, Syntrophomonas, and Clostridium_III were also obviously detected in AnMBR, which provided favorable conditions for acid transformation for methane production and further explained the higher methane yield. Interestingly, Ignatzschineria with a relative abundance of 3.6% was highly detected in AnMBR, but was not detected in UASB, indicating that membrane interception contributes to enrich the microbial communities. In addition, other unclassified genera were also found in the reactors and more abundant in AnMBR, which is beneficial for high pollutant removal and methane production. Overall, the presents study confirmed that accumulation of bacteria for complicated organics degradation in AnMBR provides favorable conditions for high pollutant removal and methane production.

Archaea
Euryarchaeota was the sole archaeal phylum recorded in both reactors. As shown in Figure 8a, Methanosarcinales with a relative abundance of 91.1% in AnMBR and 95.5% in UASB was the dominant order in both reactors. However, Methanomicrobiales accounting for 6.2% was more abundant in AnMBR than that in UASB. In addition, Methanomassiliicoccales and Methanobacteriales were also detected in both reactors. At the genus level (Figure 8b), Methanosarcina, acetoclastic methanogens in AD process dominate in AnMBR (90.2%) and UASB (94.8%), indicating that the microorganisms in UASB were inclined to utilize acetate as substrate for methane production [48,53]. Methanoculleus is the major hydrogenotrophic methanogen and can utilize hydrogen as a substrate for methane production [48]. The higher abundance of this genus in AnMBR (4.4%) than that in UASB (1.9%) indicated that methane production from hydrogen also existed in the AnMBR system, which further explains the higher methane content and yield in AnMBR. Methanomassiliicoccus only produces methane with hydrogen and methanol, even with a relative abundance of 1.3% in AnMBR, but plays key supporting roles and functions to maintain the stability of the AD process [48]. In addition, Methanocorpusculum and Methanosphaera were more abundant in AnMBR, which widened methane production pathways and benefited for system stability. Therefore, it can be concluded that the higher microbial diversity and efficient methanogenisis pathways are the main reasons for the recorded higher methane yield in AnMBR than UASB.

Energy Balance Analysis
Energy consumption and production in both reactors during the operation were analyzed ( Table 3). The energy required for influent was 5.04 × 10 −3 kWh/m 3 in both reactors, while energy needed for effluent discharge from the AnMBR to overcome the trans-membrane pressure was 8.4 × 10 −3 kWh/m 3 , which is almost consistent with previous studies [19,35,36]. In addition, energy for the water bath and substrate mixing was approximately 10.0 and 1.0 kWh/m 3 , respectively. However, energy recovery through methane production from AnMBR was 14.05 kWh/m 3 , meaning that 3.03 kWh/m 3 can be produced as net energy from AnMBR. The net energy produced from UASB was only 0.92 kWh/m 3 , which indicates that more net energy can be obtained using AnMBR for SW treatment. Previously, Xiong et al. [35] recorded a net energy production (0.0884 kWh/m 3 ) from municipal wastewater using a dynamic membrane filtration and AD combined system. Although energy production and requirement are influenced by substrate properties, reactor configurations, and operation parameters [19,54], the net energy production obtained in this study confirms that AnMBR is a promising technique for SW treatment coupled with efficient bioenergy recovery.

Conclusions
Pollutant removal performance and efficiency of methane recovery from SW using AnMBR and UASB were compared. Results showed that AnMBR can intercept particulate organic matter and enhance metabolic processes, achieving much higher COD removal efficiency than UASB. In addition, the higher methanogenic activity of the activated sludge, richer microbial communities, and significant accumulation of functional bacteria which