Integration of anammox into the MBR process for main stream sewage treatment

Anaerobic ammonia oxidation (Anammox) is a new nitrogen removal technology that has attracted increasing attention for treating ammonia-rich wastewater (Park et al. 2017; Liu et al. 2019; Ma et al. 2019a; Wen et al. 2020). Under anaerobic conditions, anammox bacteria, as autotrophic microorganisms, can use nitrite as an electron acceptor and ammonia nitrogen as an electron donor and can convert them into N₂ (Mulder et al. 1995), which offers the advantages of decreasing operating costs and producing less sludge (Kuenen 2008; Abma et al. 2010; Feng et al. 2019). However, the growth rate of anammox bacteria is extremely low, with a doubling time of approximately 11 days (Strous et al. 1998; Tang et al. 2017). Thus, retaining enough anammox sludge in the reactor is a challenge for operation (Li et al. 2010, 2011a). In addition, the anammox reaction produces nitrate, so about 10% of the total nitrogen (TN) in the influent remains in the effluent, which might exceed stringent sewage discharge standards when influent nitrogen concentrations are high. Therefore, the practical application of the anammox nitrogen removal process is limited (Joss et al. 2009; Ma et al. 2019b; Wang et al. 2019). Anammox plays a significant role in the nitrogen cycle of the natural environment and nitrogen removal in engineering systems. Environmental conditions are important parameters that directly affect the reaction rate and kinetics of the anammox process and determine the community structure and metabolic pathway.

Meanwhile, the MBR process can fully retain the sludge biomass, thereby providing a practical and effective method for enriching the sludge with slow-growth bacteria (Star et al. 2010; Tao et al. 2012; Sun et al. 2018). The aerobic MBR for treating mainstream sewage is well developed and is increasingly applied in full-scale projects. However, nitrogen removal by MBR still relies on traditional nitrification and denitrification. The integration of anammox is expected to effectively increase the nitrogen removal of the traditional nitrification and denitrification process and reduce energy consumption and operating costs.

In this study, the integration of anammox into the MBR process (AX-MBR) is proposed for treating mainstream sewage. Anammox sludge was directly added into the MBR process, and the treatment performance of AX-MBR was studied. The anammox, denitrification, nitrification, and COD removal were evaluated during the study. The microbial diversity of the sludge was observed, and the mechanism of nitrogen removal was discussed.

In this study, an AX-MBR made of plexiglass was established, as shown in Figure 1. The AX-MBR was wrapped with a black cloth to protect it from light, thereby preventing negative effects on the anammox bacteria (Jetten et al. 2005). As shown in Figure 1, the anoxic tank had an effective volume of 4 L and overall dimensions of 9 × 9 × 74 cm, while the aerobic tank had an effective volume of 6 L and overall dimensions of 18 × 10 × 50 cm. The oxygen reactor is equipped with a ceramic plate film having a size of 20 cm × 10 cm, a thickness of 7 mm, a pore diameter of 0.1 μm. It mainly intercepts activated sludge to ensure the biomass of the reaction system, with an effective interception rate of up to 95%. The aerobic reactor was also equipped with an aeration device and dissolved oxygen (DO) control system. Moreover, a backwashing device was installed, and the filtrate pump and backwashing pump were intermittently operated to alleviate pollution of the membrane module. Portable pH and temperature online monitoring devices were used during system operation to analyze the operating status and stability of the reaction system.

The daily treatment capacity of AX-MBR was 70 L/d with a hydraulic retention time of 3.4 h, and no temperature control was applied. During the study, the hydraulic retention time was maintained. The AX-MBR was operated for 210 days, which could be divided into three stages, as indicated in Table 1.

During stage I, the AX-MBR was initiated using sewage as the influent, and the circulation ratio was set at 1:2. During stage II, the DO of the aerobic tank was gradually reduced to less than 0.5 mg/L by reducing the blower aeration to provide a suitable living environment for the anammox bacteria. During stage III, 1 L of anammox sludge was added into the anoxic tank, and the other operating conditions remained unchanged.

The inoculated sludge used in the experiment was obtained from a sewage treatment plant in Guilin City, China. The anammox sludge was cultivated in a pilot-scale reactor using granulated active carbon, and it appeared red (Zhang et al. 2014, 2015, 2016a, 2018a). To start the AX-MBR, 4 L of activated sludge with a mixed liquor suspended solids (MLSS) of 3,500 mg/L was used as seed sludge. Next, 1 L of anammox sludge (Liu et al. 2020a, 2020b; Jin et al. 2021; Wen et al. 2021a; Wei et al. 2020a) was added into the anoxic tank after the start-up of the AX-MBR. The inoculated sludge was presented in Figure 2.

The wastewater treated in the experiment was collected from the mainstream sewage of Guilin University of Technology (Zhang et al. 2017a). The main components were as follows: NH₄⁺-N: 40–90 mg/L, NO₂⁻-N: 0.1–5 mg/L, TN: 60–130 mg/L, COD: 90–170 mg/L, turbidity: 1.9–6.8 NTU, and pH: 6.19–6.8. The wastewater was first deoxidized using a nitrogen generator to facilitate the subsequent adjustment and control of the DO in the reaction system. The pH index of the anoxic reactor was adjusted to be in the range of 7.45–7.55 using dilute H₂SO₄ and NaHCO₃ solutions.

With the aid of a suction/backwashing pump, the wastewater passed through the flat-sheet ceramic membrane for drug washing. An aeration blower was used for the continuous aeration of the aerobic tank. The aeration provided oxygen for the growth of the microorganisms and helped clean the flat-sheet ceramic membrane surface through airflow oscillation to prevent membrane clogging. The trans-membrane pressure (TMP) was monitored using a digital pressure sensor and recorded in the controlling computer.

Membrane fouling of the FCMBR (Zhang et al. 2018b, 2018c; Jin et al. 2019) was controlled using the filtration-backwash mode (filtration: 540 s, backwash: 60 s). When TMP surpassed 25 kPa, the operation of the membrane units was stopped. The flat-sheet ceramic membranes were then cleaned using chemical pumps, which pumped 1,000 mg/L NaClO into the inner space of the membranes (Wen et al. 2021b). The membranes were soaked for 1–2 h to recover the filtration performance of the flat-sheet membranes.

After the reactor was started, water samples were collected daily. NH₄⁺-N and NO₂⁻-N were measured according to the Standard Methods protocol (APHA 1985). The TN was determined using the persulfate method (APHA 1985), and NO₃⁻-N was calculated as the difference between the TN and the sum of NH₄⁺-N and NO₂⁻-N (Zhang et al. 2018d; Li et al. 2019; Wei et al. 2020b). The filtered COD (1 μm) was measured using the closed reflux colorimetric method (APHA 1985). The MLSS and mixed liquid volatile suspended solids (MLVSS) of the sludge samples were measured in accordance with the Standard Methods protocol (APHA 1985). The pH was measured using a pH meter (9010; Jenco Instruments, San Diego, CA, USA), and the DO was measured using a DO meter (9010; Jenco Instruments, San Diego, CA, USA). The water temperature was measured by online monitoring DEC digital pH meter (DPH10AC; Tianjian Innovation Monitoring, Beijing, China). Pearson correlation analysis was done based on SPSS 23.0 software, the nitrogen loading rate (NLR) and the nitrogen removal rate (NRR) were obtained by referring to the formulae in the literature (Karasuta et al. 2021).

The water samples were measured before and after adding anammox sludge using luminescence spectrometry (F-4500 FL spectrophotometer, Hitachi, Japan). Excitation–emission matrix (EEM) fluorescence spectroscopy was used to describe the relationship between the fluorescence intensity, excitation wavelength, and emission wavelength. It can directly reveal changes in the protein conformation and microenvironment tryptophan residues. With the help of contour spectroscopy, it was possible to see how the emission wavelength migrates with such changes clearly. Different kinds of organic substances have unique fluorescence characteristics. According to the relevant literature, the excitation and emission wavelengths (Ex and Em, respectively) of the fluorescence peaks in the EEM fluorescence spectra can be divided into five regions, namely Region I: Ex 200–250 nm, Em 260–320 nm; Region II: Ex 200–250 nm, Em 320–380 nm; Region III: Ex 200–250 nm, Em 380–550 nm; Region IV: Ex 250–450 nm, Em 260–380 nm; and Region V, Ex 250–450 nm, Em 380–550 nm.

The operational taxonomic unit (OTU) refers to the same mark artificially set for each taxon to facilitate analysis in phylogenetic or population genetics studies. To understand information regarding items such as strains and bacteria in a sample sequencing result, the sequence must be classified. This method involves clustering all the sample sequences according to the distance between the sequences and then dividing the sequence into different OTUs based on the similarity between the sequences. A Venn diagram can be used to count the number of common and unique OTUs in the sample, thus visually illustrating the similarities and overlapping of the number of OTUs in the environmental sample.

Sludge samples were collected from various stages of the reaction system. Following dehydration, the sludge samples were sealed in a centrifuge tube and placed in a refrigerator at −20 °C. The PowerSoil DNA Isolation Kit (MO BIO, Carlsbad, CA, USA) was used to extract DNA from the sludge samples. After verifying the purity (OD260/OD280: 1.6–1.8), the DNA was amplified by a polymerase chain reaction (PCR). Specifically, the target fragment was amplified using PCR while cutting the target DNA fragment. The amplified DNA fragment in this experiment was the V6 region of the bacterial 16S rDNA. The desired sequence was amplified using primers 968F, GC-968F, and 1401R, and the target DNA was separated and purified by denaturing gradient gel electrophoresis. DNA fragments of different microorganisms were immobilized in different locations on the gel. Following every gelation step, each band was recovered and redissolved to purify the target DNA fragment. The PCR product was sequenced by Shanghai Shenggong Biotechnology Co., Ltd (Shanghai, China), and the sequenced genes were compared with the National Center for Biotechnology Information GenBank database using BLAST analysis.

As shown in Figure 3(c), during the first 7 d of stage I, the influent NH₄⁺-N concentration is around 50 mg/L, the treatment performance is unstable, and the effluent NH₄⁺-N concentration is relatively high, reaching a maximum of 15.79 mg/L. On day 8, the influent NH₄⁺-N concentration increased, the treatment efficiency gradually increased, and the effluent NH₄⁺-N concentration could be maintained below 10 mg/L. On day 43, the effluent NH₄⁺-N concentration reached 2.02 mg/L, and most of the NH₄⁺-N in the influent water is removed. The average removal rate during the first 8 days is only 66.91%, while 87.61% could be achieved after the start-up period. During stage II, with decreasing DO, the effluent NH₄⁺-N increased significantly, reaching a maximum of 34.23 mg/L, with an average removal rate of 53.37%. The results showed that the low DO concentration inhibited the nitrifying bacteria. During stage III, about 1 L of anammox sludge was added into the anoxic tank, the influent NH₄⁺-N concentration is maintained at about 73 mg/L. As shown in Figure 3(c), the effluent NH₄⁺-N was immediately reduced to 14.90 mg/L. This might be due to the anammox reaction, which can reduce the effluent NH₄⁺-N. After that, in the following days, the anammox bacteria survived in the AX-MBR, and the NH₄⁺-N removal rate stabilized at about 87.87%.

As shown in Figure 3(b), the trend in the TN removal was consistent with that of NH₄⁺-N. After the start-up period, the TN removal rate reached approximately 66.42%. During stage II, limited by the DO, the effluent TN concentration increased significantly, and the TN removal rate decreased. During stage III, with the anammox reaction, a significant decrease in the effluent TN was observed. Figure 3(b) shows that the TN removal was stable, and the effluent TN could be lower than 30 mg/L, which improved, even compared to that of stage I.

Figure 3(a) demonstrates that the COD removal remained stable throughout the whole study period, with the effluent COD remaining below 40.32 mg/L. The results showed that DO and temperature did not affect COD removal during the study. The remaining COD was difficult for the microorganisms to decompose; nevertheless, the effluent COD could meet Chinese discharge standards.

The changes of NLR, NRR, and water temperature in AX-MBR are recorded in Figure 3(d). As shown in the figure, the trend of NLR variation is consistent with that of feedwater TN variation, and the increase of feedwater NH₄⁺-N and TN concentration in phase I increases the NLR, which ensures the smooth start-up operation of the system. The average NLR in phase III is maintained at about 0.51 kg N m⁻³ d⁻¹ and the NRR is stabilized at about 0.44 kg N m⁻³ d⁻¹. The system operation was stable, TN, NH₄⁺-N, and COD removal rates remained stable, and the effluent TN, NH₄⁺-N, and COD all met the Chinese discharge standards.

In the AX-MBR process, NH₄⁺-N is removed by nitrification and anaerobic ammonia oxidation reactions, while TN is removed by anaerobic ammonia oxidation and denitrification. Compared to other ANAMMOX studies (Li et al. 2018; Cui et al. 2021), the water temperature was not controlled during this study, and the water temperature varied mainly with seasonal temperature. Since the temperature in Guilin is generally above 20 °C, only the changes in the average removal rates of TN, NH₄⁺-N, and COD were studied when the water temperature varied in the range of 20 °C–32 °C (Figure 4). From Figure 4, it can be seen that the average removal rates of TN and NH₄⁺-N gradually increased with the increase of temperature, with TN removal rate rising from 57% to 72% and NH₄⁺-N removal rate rising from 83.7% to 86.9%, and the highest average removal rates of TN, NH₄⁺-N, and COD were observed when the water temperature is in the range of 30 °C–32 °C. According to the literature, the optimum temperature for anaerobic ammonia oxidation reaction is at 30 ± 2 °C (Chen et al. 2016) when microbial activity is higher and nitrogen removal is high, which is consistent with the results of this experiment. In the correlation analysis of the stage III process listed in Table 2, there is an extremely strong correlation between the change of water temperature in stage III on the removal of TN, NH₄⁺-N, and NRR from the effluent, and combined with Figures 3(d) and 4, it can also be seen that the temperature affects the nitrogen removal performance, and since the temperature is a necessary condition for microbial growth, it will directly affect the microbial activity, and thus will affect the nitrogen removal.

C_A is the nitrogen removal by anammox, and C_III and C_II are the average nitrogen removals during stage II and III, respectively.

According to the AX-MBR flow chart, wastewater entered into the aerobic reactor after passing through the anoxic reactor, and nitrification occurs in the aerobic reactor to form NO₂⁻ and NO₃⁻. The anoxic reactor was fed with mainstream sewage and circulating water to provide a substrate for denitrification and anammox. With a circulation ratio of 1:2, the maximum denitrification contribution rate is 33.3%, and the average removal rates of NH₄⁺-N, TN, and COD in the three stages are illustrated in Figure 5.

Stage I is the traditional nitrification and denitrification process. Average removal rates of NH₄⁺-N and TN are 85.4% and 61.88%, respectively. The denitrification effect is efficient. However, the process requires a higher aeration cost to maintain DO in the range of 1–1.5 mg/L, notwithstanding the cost of additional carbon sources. In stage II, though reducing DO saved some energy consumption costs, the denitrification effect is significantly reduced, and the TN removal rate is only 53.42%. In stage III, anammox bacteria were added to ensure a good denitrification effect, while reducing the energy consumption of aeration and the cost of additional carbon sources. The average removal rates of NH₄⁺-N, TN, and COD were 86.58%, 71.92%, and 73.07%, respectively. Comparing stage II and stage III, the denitrification performance of stage III is better under the same aeration energy consumption, and the denitrification rate of stage III is increased by about 20%. Comparing stage I and III, there is no apparent difference in the efficiency of denitrification, but the energy consumption of aeration in stage III reduced by 20%.

Traditional sewage treatment plants mainly rely on nitrification and denitrification to achieve the purpose of denitrification. Ignoring the different types of process conditions, the processes of biological denitrification through nitrification and denitrification have disadvantages such as high energy consumption, high sludge production, and poor denitrification effect for low carbon to nitrogen ratio wastewater (Wei et al. 2020a). There are also individual defects in different processes. For example, the oxidation ditch process occupies a large area compared to other processes. It was only suitable for small and medium-sized urban wastewater treatment plants with sufficient land resources (Shen et al. 2011). The Sequencing Batch Reactor (SBR) process combines homogenization, primary sedimentation tank, biodegradation tank, and secondary sedimentation tank, which has the disadvantages of low volume utilization and high oxygen demand and unsuitable for large water volumes (Li et al. 2011b). The Anaerobic/Oxic (A/O) process does not have an independent sludge reflux system, so it cannot cultivate sludge with unique functions. To improve the denitrification efficiency of the A/O process, the internal circulation ratio must be increased, which makes the operation cost increase. In addition, after increasing the internal circulation ratio, the dissolved oxygen (DO) concentration of the internal circulation fluid from the aeration tank will increase, making it difficult to maintain the ideal anoxic state in the anoxic section, which affects the denitrification effect and makes it difficult to reach 90% denitrification rate (Wang et al. 2015). Anaerobic/Anoxic/Oxic (A2/O) process in the biological denitrification and phosphorus removal process has conflicting sludge ages of nitrifying bacteria and phosphorus-polymerizing bacteria, while in terms of carbon sources, denitrifying bacteria and phosphorus-polymerizing bacteria will compete with each other, leading to the efficiency of nitrogen removal and phosphorus removal was not high (Sorm et al. 1996; Kapagiannidis et al. 2011). In conclusion, compared with traditional denitrification processes, the AX-MBR process can ensure the quality of effluent and stable operation and save 20% of the aeration cost.

With the treatment of wastewater, part of the organic material was used by microorganisms to synthesize new cytoplasm, microorganisms in the metabolism, accompanied by part of the death of microorganisms, so the sludge gradually increased, nitrification-denitrification and anaerobic ammonia oxidation are faced with the problem of sludge discharge and disposal, which invariably increases the cost of wastewater treatment (Zhang et al. 2007, 2008, 2009, 2011; Khanh et al. 2011; Jin et al. 2016a, 2016b). AX-MBR process operates for 210 days without sludge discharge. Moreover, the filter interception of FCMBR can retain the complete biomass and extend the hydraulic retention time indefinitely and improve the treatment efficiency. As one of the domestic tourist cities, Guilin has large population mobility, and its domestic sewage contains more catering wastewater than ordinary cities. Therefore, the content of NH₄⁺-N and TN in Guilin domestic sewage is higher than that in ordinary domestic sewage. In addition, it is based on the pollutant standards of municipal wastewater treatment plants in China. The TN in the effluent of this experiment did not meet the first A standard (<15 mg/L), mainly because the effluent of this experiment is used for reclaimed water reuse and campus greening. The standard of greening water quality is primarily aimed at the NH₄⁺-N index, and the NH₄⁺-N concentration in the effluent of this experiment can meet the application requirements.

The coupled anammox process is a new type of biological denitrification technology that has existed for only a short time. Its engineering remains far from mature, and the compositions of industrial wastewater and domestic sewage are complex, which presents significant challenges for the engineering promotion and stable operation of AX-MBR. Combined with the advantages of MBR, the potential of the AX-MBR process to save energy will inevitably result in huge returns; therefore, future research should focus on optimizing the operating conditions and measures for dealing with environmental changes (Jin et al. 2015a, 2015b, 2016c, 2016d; Zhang et al. 2016b, Zhang & Jin 2017b). The direct denitrification of the AX-MBR process applied to actual wastewater at normal or low temperatures is an important developmental direction. The denitrification level of the AX-MBR process can be maintained by increasing the bacteria dosage or the low-temperature acclimation to screen out dominant strains that can adapt to the environment. However, in actual projects, improving the activity of the bacteria under low temperatures and expanding the bacteria remain drawbacks to be overcome.

Different types of organic matter exhibit unique fluorescent properties. EEM fluorescence spectra can reflect the characteristics of the organic matter content, classification, and composition. The fluorescence peak in the EEM fluorescence spectra is divided into five peaks, representing five different classes of organic matter (Baker 2002; Chen et al. 2003). This approach could be applied to investigate differences and relationships between the dissolved organic groups, molecular constituents, and their counterparts in the effluent. Therefore, two samples were obtained during stages II and III, and EEM fluorescence spectroscopy was performed. The results of the effluent samples were compared and analyzed using the spectra of mainstream sewage, and the results are presented in Table 3 and Figure 6.

According to Figure 6 and Table 3, a total of five fluorescent peaks appeared in the EEM fluorescence spectra of the effluents. The fluorescence peaks of the two samples were similar, indicating that the ability of the microorganisms to treat the dissolved organic matter in the sewage is relatively stable. The excitation wavelength/emission wavelengths (Ex/Em) corresponding to the maximum fluorescence peak position of the sample before adding anammox sludge were Ex/Em = 246 nm/272 nm (Peak A), Ex/Em = 206 nm/358 nm (Peak B), Ex/Em = 248 nm/420 nm (Peak C), Ex/Em = 276 nm/310 nm (Peak D), and Ex/Em = 254 nm/426 nm (Peak E). The Ex/Em values corresponding to the maximum fluorescence peak position of the sample after adding anammox sludge were Ex/Em = 248 nm/276 nm (Peak A), Ex/Em = 208 nm/378 nm (Peak B), Ex/Em = 248 nm/418 nm (Peak C), Ex/Em = 278 nm/312 nm (Peak D), and Ex/Em = 252 nm/420 nm (Peak E).

Peak A and Peak B are protein-like fluorescent peaks, mainly derived from detergents, food residues, and human or animal waste. According to Baker (2002), Peak A (Ex/Em = 200 to 250 nm/260 to 320 nm) represents tyrosine-like organic matter, while Peak B (Ex/Em = 200 to 250 nm/320 to 380 nm) represents tryptophan-like organic matter. Peak C belongs to the fulvic-acid-like fluorescent peak. At the same time, Peak C is located in the ultraviolet region (Ex/Em = 200 to 250 nm/380 to 550 nm), representing organic matter with a high fluorescence efficiency and low molecular weight. Peak E belongs to the fulvic acid fluorescent peak in the visible region (Ex/Em = 250 to 450 nm/380 to 550 nm) produced by high-molecular-weight and relatively stable aromatic fulvic acid organic matter. In the spectrum, Peak A, Peak C, Peak D, and Peak E exhibited high fluorescence intensity. Peak E's peak band is the widest, indicating the sample's highest level of humic-acid-like dissolved organic matter.

In total, four sludge samples were collected in the study. Except during stage I, sludge samples were taken when the removal performance stabilized. A1 and A2 were taken from the aerobic tank, and B1, B2 were taken from the anoxic tank. A1 and B1 were collected during stage II, A2 and B2 were during stage III.

Figure 7 presents the Venn diagram of the OTU sample distribution. Clearly, the microbial diversity in the three samples changed significantly. Sample A1 had 15,860 OTUs, while B1 had 15,746 OTUs. The structure of the microbial community was complicated. Sample A2 had 2623 OTUs, B2 had 2388 OTUs, and thus, the number of microbial communities was substantially reduced. A large number of strains was reduced or even eliminated during the long-term operation of the process, and the diversity and abundance of the microbial community structure decreased significantly. During stage III, a microbial community increased by adapting to this environment. Therefore, the number of OTUs in samples A2 and B2 was less than that in A1 and B1.

A heatmap uses color to reflect the abundance information of community distributions and can intuitively display the community distribution abundance value with a defined color depth. Moreover, the sample and community distribution information were clustered and rearranged, and the results following clustering were displayed in the heatmap. Therefore, it can reflect the similarities and differences between the community distributions at each classification level. Figure 8 presents a heatmap of the species abundance at the genus level.

According to the genus heatmap, the species of samples A1 and B1 are mainly composed of Nitrospira, Candidatus Kuenenia, Nitrosomonas, and Aridibacter. Moreover, the abundance of the different species changed with the operating conditions. The Aridibacter associated with denitrification in samples A2 and B2 is the dominant species. Owing to the limitations of operating conditions such as the DO, the species abundance of Nitrospira, Candidatus Kuenenia, and Nitrosomonas is reduced. As Candidatus Kuenenia is the most sensitive to environmental conditions, the abundance of this species is the lowest.

The collinearity between samples and species was shown in Figure 9. The right semicircle represents the species abundance composition of the samples, and the left semicircle represents the distribution proportion of species in different samples at the classification level. In the outermost first and second color circles, the right half-circles represent the species composition corresponding to different samples. The various colors represent different species. The length represents the abundance ratio of a species in the sample (the percentage shown in the second circle). The left half-circle represents the distribution proportion of different samples in the dominant species. Various colors represent different samples. The length represents the distribution proportion of the samples in a species (the percentage shown in the second circle). In the third outermost circle, one end of the color band connects the sample (right half circle), and the other end connects the species (left semicircle). The width of the band endpoint represents the distribution proportion of the sample in the corresponding species, and the outside value represents the abundance of the corresponding species.

As can be observed from Figure 9, at the genus level, the highest proportion in the A1 sample was Nitrospira at 15.77%, followed by Candidatus Kuenenia at 5.8%, and Nitrosomonas at 2.59%. This result indicates that the nitrification reaction remained in the short-cut nitrification stage, which provides a theoretical basis for the enrichment of the reactor with nitrite. The structural composition of B1 was similar to that of A1, and Candidatus Kuenenia accounted for 8.39%, creating appropriate conditions for anammox in the anaerobic reactor. Acinetobacter and Aridibacter accounted for 17.29% and 3.5%, respectively, in the A2 sample. Aridibacter is a type of bacteria that can synergize with other bacteria to increase the activity of denitrifying bacteria (Wang et al. 2017). Moreover, owing to the limitation of DO, Nitrospira exhibited a more significant reduction compared to 15.77% in A1, but it still played a role in the sewage treatment process. However, in the B2 sample, Acinetobacter accounted for the highest proportion at 16.19%, which was the dominant strain adapted to the living environment and operating conditions in the system, while Candidatus Kuenenia accounted for 1.74%.

In summary, after the reaction system was operated in stage III, the dominant strain Nitrospira in the two reactors was finally reduced to approximately 2.24%. The anammox bacteria decreased from the initial 7.095% to 1.53% for various reasons, such as the slow growth rate. However, during the long-term operation of the process, anammox bacteria play an important role in denitrification, regardless of their amount, together with the nitrifying bacteria. Moreover, the Nitrosomonas present in each sample indicated that microorganisms that provide nitrite as an electron acceptor for anammox are always present in the system.

The suspended matter intercepted by the membrane module will stay in the reactor for some time because it cannot be completely decomposed by the microorganisms immediately. Coupled with the stirring and reflux mechanism in the reactor, the mud-water mixture is mixed very uniformly. As a result, membrane contamination is sure to occur when membrane modules are used for long periods of time. (Yu et al. 2017).

The main cause of membrane fouling is the adsorption and deposition of sludge flocs, colloidal particles, dissolved organic matter, or inorganic salts on the membrane surface, which blocks the membrane pores, which leads to a decrease in membrane flux or an increase in TMP (Zuthi et al. 2013). Therefore, we need to record the real-time TMP changes through the LCD control cabinet (Figure 10) and analyze the membrane pollution status. When it is observed that the membrane pressure of the membrane module reaches about 25 kPa, replace or clean the flat membrane in time to ensure a good membrane flux. It can be seen from Figure 10 that during the first period of operation of the flat membrane, the TMP rises slowly. When the TMP reaches 15 kPa or more, the TMP will rapidly rise to about 25 kPa in about 5 days. It is necessary to take out the flat membrane to clean up the membrane fouling and restore the membrane flux. The TMP of the cleaned membrane will drop to about 10 kPa again.

It can be seen from Figure 10 that after the membrane is used for some time, the TMP rises, and the membrane flux decreases, which increases the power consumption of the filter pump. The ceramic flat membrane is a brittle material, and excessive pressure will also cause harm to the membrane (Wen et al. 2021b). Therefore, we need to clean up membrane fouling and restore membrane flux. At present, the main cleaning of the ceramic mold comprises a physical washing (backwashing, low pressure, high-velocity cleaning, vacuum cleaning, and mechanical scraping, etc.), chemical cleaning (acids, bases, oxidants, and surfactants), ultrasonic cleaning, the cleaning electric field with the combined cleaning method (Wen et al. 2021b). To ensure the normal operation of the membrane treatment equipment, this study uses a combined cleaning method to clean the ceramic flat membrane. The suction/backwash time is set to 9 min/1 min during the use of the ceramic flat membrane when the TMP further increases to about 25 KPa, the TMP is further reduced by mechanical cleaning or sodium hypochlorite cleaning.

During the operation of AX-MBR, the combined cleaning method used to clean the ceramic flat membrane can effectively solve the decline of membrane flux. The membrane module can restore the previous good membrane flux level after treatment. Following the stable operation of AX-MBR, the NLR is maintained at 0.51 kg N m⁻³ d⁻¹, and the NRR is stabilized at about 0.44 kg N m⁻³ d⁻¹. The nitrogen removal rate of stage III is the highest, the NH₄⁺-N removal rate could reach 87.87%, and the average TN removal rate is approximately 72.9%. The effluent COD is maintained at about 35.87 mg/L with an average removal rate of 72.81%. The effluent COD is less affected by a change in environmental conditions.

The microorganisms in AX-MBR exhibited a relatively stable ability to treat the dissolved organic matter in the sewage. In the five parts of the dissolved organic matter in the effluent, the content of humic acid-like substances is the highest, and the average proportion of the two effluent samples reached approximately 57.8%.

The heatmap of the species abundance demonstrated that the main species of AX-MBR are Nitrospira, Candidatus Kuenenia, Nitrosomonas, and Aridibacter. From the perspective of microbial diversity, the microbial community structure in the samples underwent major changes. The main manifestation is that the microbial community structure changed from complex to relatively simple, and certain strains are reduced or even eliminated. At the genus level, Nitrospira, Nitrosomonas, and Aridibacter, involved in microbial nitrification and denitrification, are present and predominant in all samples. Candidatus Kuenenia, which is significantly affected by the environment, is gradually reduced owing to the slow growth of anammox bacteria and harsh growth conditions.

This research was supported by the National Natural Science Foundation of China (grant numbers 51638006); the Guangxi Natural Science Foundation (grant number 2019GXNSFFA245017); Special Funding for Guangxi ‘BaGui Scholar’ Construction Projects.

Conflict of Interest: The authors declare no conflict of interest.

All relevant data are included in the paper or its Supplementary Information.