Optimization of Wastewater Phosphorus Removal in Winter Temperatures Using an Anaerobic–Critical Aerobic Strategy in a Pilot-Scale Sequencing Batch Reactor

Biological phosphorus removal using an anaerobic–aerobic sequencing batch reactor (SBR) in a low temperature can be difficult to remove, and aeration always accounts for nearly half of the total electricity costs at many wastewater treatment plants. In this study, a pilot-scale anaerobic–critical aerobic SBR (A–CA SBR) was developed for synthetic domestic wastewater. More importantly, the phase, whose concentration of diffused oxygen was controlled at 1.0–1.5 mg/L, was defined as a critical aerobic phase, which reduced expenses during the operation. To be specific, half of the ammonia was removed within 10 days and no NO3−–N was accumulated during the process. From the SEM and metagenome analysis, Rhodocyclus, Zooglea, Dechloromonas, and Simplicispira had the ability to remove phosphorus and NO3−–N simultaneously, which proved the existence of a potential double-layer sludge structure under an A–CA operational condition. All of the results disclose that the pilot-scale A–CA SBR is a reliable manipulation strategy for phosphorus removal under low temperatures, which can hopefully apply to practical wastewater remediation.


Introduction
Discharging phosphorus into the river and sea without treatment will deteriorate the aquatic environment, and causes eutrophication, which aggravates the situation for the algae competing for oxygen and nutrition with aquatic organisms [1][2][3]. In order to avoid this situation happening, various wastewater treatment plants are used, like enhanced biological phosphorus removal (EBPR), which can effectively remove phosphorus and nitrogen [4]. Among EBPRs, a sequencing batch reactor (SBR) has been widely used because of its effective removal efficiency, good adaptation for influent water fluctuations, and space-saving characteristic [5]. There are some illustrated examples for different operation modes of SBRs. Using an anaerobic-aerobic SBR, Li et al. [2] realized a phosphorus removal efficiency of over 80% under three levels of temperature (5,15, and 20 • C), and Panswad et al. [6] realized a soluble phosphorus of the effluent as low as 0.4 mg/L. for long anoxic-short aerobic SBR, and 4.3 mg/L of soluble phosphorus was left in high-strength wastewater of 4500 mg/L Chemical Oxygen Demand (COD) and 35 mg/L PO 4 3− -P under an ambient temperature [7]. Most of the operation modes mentioned above needed to supply enough oxygen for a good performance. However, supplying a long period of air could account for approximately 50% of the total energy consumption in typical wastewater treatment [8,9], which is not in accordance with the current mainstream concept of building an energy-saving society, and why a lower concentration of oxygen should be considered more in processes.
China) in the bottom of SBR at a concentration of dissolved oxygen (DO) of 2.0 mg/L approximately, to guarantee the aerobic organism living properly. For another 73 days, each cycle consisted of the following five phases: filling (0.25 h), anaerobic (2 h), critical aerobic phase (length is always changing by the results), settling (0.5 h), and withdrawing (0.25 h). During the experiments, the duration of the critical aerobic phase changed from 2 h to 1 h, and eventually to 0.5 h, due to the treatment results, and the concentration of DO was always controlled at 1.0-1.5 mg/L.  The specific operation mode is shown in Table 1. Moreover, the pH was uncontrolled, except that the initial pH was adjusted to 7.0-7.5 at the beginning of each cycle. Sludge was not discharged during the whole manipulation of SBR. However, when necessary, manual sludge disposal was performed for the subsequent treatment, such as sludge dewatering. The relevant components were periodically measured to monitor the reactor performance during the whole operation.  The specific operation mode is shown in Table 1. Moreover, the pH was uncontrolled, except that the initial pH was adjusted to 7.0-7.5 at the beginning of each cycle. Sludge was not discharged during the whole manipulation of SBR. However, when necessary, manual sludge disposal was performed for the subsequent treatment, such as sludge dewatering. The relevant components were periodically measured to monitor the reactor performance during the whole operation.

Physicochemical Analysis
The concentrations of COD, NH 4 + -N, PO 4 3− -P, NO 3 − -N, and NO 2 − -N were analyzed according to the Standard Methods for the Examination of Water and Wastewater from the American Public Health Association before the removal of treated wastewater and after every different phase [16]. The mixed liquor suspected solids (MLSS) and settling velocity (SV) were also determined periodically using a gravimetric method. The pH was measured using a pH analyzer (PB-10, Sartorius, Gottingen, Germany), and the dissolved oxygen (DO) was monitored using a DO quick-analysis apparatus (SC200, Hitachi, Washington, DC, America). The temperature was detected by the sensor installed in the SBR reactor.

Morphological Observation
All of the selective sludge samples were observed by the scanning electron microscope (SEM) for the morphological observation of sludge so as to analyze the sludge structure. The specific procedure was the same as for Liu et al. [17].

Microbial Community Analysis
The metagenomics method manufactured by Sangon Biological Technology Company (Sangon Biological Engineering Co., Shanghai, China) was utilized to unravel the microbial community contained in the sludge, using the polymerase chain reaction followed by high-throughput sequencing (Hiseq 2500, Illumina, CA, USA). The specific procedure was the same as for Xiong et al. [18].
Sludge samples from different phases were taken at day 38 to represent the acclimation results of the PAOs in the stage I, while days 86 and 112 were selected to represent the condition of the sludge in this reactor, especially the microbial community change in the critical aerobic circumstance.

Phosphorus Removals
As shown in Figure 2, the phosphorus release efficiency increased all the time, up until almost 200%, and the phosphorus removal efficiency was maintained at 85%, where the PO4 3− -P concentration in the effluent was below 4 mg/L at the end of stage I. These results indicated the acclimation of the PAOs was successful [19][20][21].
For stage II, the aerobic phase was altered into the critical aerobic phase, whose DO concentrations was 1.0-1.5 mg/L [22,23] so as to reduce energy consumption and, depending on the effluent, the length of the critical aerobic phase changed from 2 h to 1 h to 0.5 h. Then, a surprising phenomenon occurred, that the PO4 3− -P concentration of the effluent could be as low as below the detective line within 10 days, regardless of how many times we changed the length of the critical aerobic phase (the regular change pattern of phosphorus is in Figure 2a in blue dots). During this

Phosphorus Removals
As shown in Figure 2, the phosphorus release efficiency increased all the time, up until almost 200%, and the phosphorus removal efficiency was maintained at 85%, where the PO 4 3− -P concentration in the effluent was below 4 mg/L at the end of stage I. These results indicated the acclimation of the PAOs was successful [19][20][21]. For stage II, the aerobic phase was altered into the critical aerobic phase, whose DO concentrations was 1.0-1.5 mg/L [22,23] so as to reduce energy consumption and, depending on the effluent, the length of the critical aerobic phase changed from 2 h to 1 h to 0.5 h. Then, a surprising phenomenon occurred, that the PO 4 3− -P concentration of the effluent could be as low as below the detective line within 10 days, regardless of how many times we changed the length of the critical aerobic phase (the regular change pattern of phosphorus is in Figure 2a in blue dots). During this period, almost 100% of PO 4 3− -P was removed. The PAOs in this system could be responsible, because PAOs are more suitable for a lower temperature (<20 • C) than GAOs, which deteriorated the phosphorus removal and competed with the nutrients with PAOs. However, 0.5 h was not suitable for this experiment (seen in stage IV in Figure 2), as the special phenomenon did not meet our expectations, while only less than 5% PO 4 3− -P was eliminated. The typical cycle profile for the various stages exhibited in Figure 3 could validate that as well. It can be known that phase II and III achieved a good performance for the COD and NH 4 + -N, and removed almost all of the total PO 4 3− -P. However, when the length of the critical aerobic was cut to 0.5 h (phase IV in Figure 2), a worse result was elucidated, that the residual concentrations of COD, PO 4 3− -P, and NH 4 + -N were higher than before, proving that 0.5 h was not a sufficient time-length for PAOs to remove phosphorus. From a previous report, PAOs require a lag period for synthesizing enzymes for phosphorus uptake, especially when switched to a different oxygen condition (changed to critical aerobic phase in this study). Thus, under this condition, 1 h was more suitable for the PAOs to develop enzymes for phosphorus removal, rather than 0.5 h [24].

Nitrogen Removals
From Figures 2 and 3, the NH 4 + -N removal efficiencies in stage I for the anaerobic phase and aerobic phase were 20% and 50%, respectively, which were contributed to the demands of normal microorganism metabolism. Simultaneously, it was also revealed that stable nitrifying bacteria existed in the reactor, without obvious denitrifying bacterial. Thus, the concentrations of NO 3 − -N and NO 2 − -N were not measured during this stage.
Stage II to IV witnessed a relatively poor nitrification with a lower removal efficiency of NH 4 + -N than stage I. Additionally, the NO 3 − -N in effluent was accumulated because of the effects of denitrification. The partly NH 4 + -N deceleration was insensitive to the length of the critical aerobic phase and the products, and NO 3 − -N had a slightly higher concentration (stage II and III: <2.5 mg/L, stage IV: <3 mg/L) than the influent (1 mg/L). Futhermore, because of the low concentration (around detection limits), there was no observation of NO 2 − -N during these three stages. Poor nitrification would be caused by a low operating DO concentration, because of the inhibition of the growth rate of the nitrifiers [25]. Additionally, the highest dissimilatory nitrate-to-ammonia process (DNRA) genes in glucose supported the denitrification system [26], which would dissimilate NO 3 − -N to generate ammonia, was favorable for a high ammonia concentration in th effluent [27][28][29], and the dominated genus in the glucose denitrification supported system were Thauera, Dechloromonas, and Zoogloea [26], which was in accordance with the metagenome results (seen in discussions of chapter 3.2). However, the previous literature from Fan et al. [30] indicated that long SRTs reactors could improve the nitrification efficiencies even in a low DO concentration. Coincidentally, this A-CA SBR system did not generate much sludge during operation, which meant that the system had long SRTs. As a whole, a certain amount of NH 4 + -N was changed into NO 3 -N, because of the nitrification, causing an increasing concentration of NO 3 − -N ( Figure 2). Simultaneously, compared with stage II, part of NO 3 − -N was clearly removed, which could be denoted in stage III. Generally speaking, denitrifying phosphate accumulating organisms (DPAOs) could realize the synchronous removal of NO 3 − -N and PO 4 3− -P [31] by using NO 3 − -N for electron donation instead of O 2 for excess PO 4 3− -P absorption in an anoxic environment [32][33][34]. Therefore, from this perspective, the existence of DPAOs in the system was demonstrated. In addition, the dominant genus of Dechloromonas and Zoogloea in the subsequent analysis also validated it properly (seen in two critical aerobic phases of Figure 5). Conventionally, NH 4 + -N is achieved by the nitrification of ammonia to nitrate or nitrite under aerobic conditions, and NO 3 − -N and NO 2 − -N were removed by denitrification under anoxic conditions [35]. In our results, the DO concentration of 1.0-1.5 mg/L in our critical aerobic phase did not exert much of an impact on the NO 3 − -N removal. It could be noted that the activated sludge flocs were larger, irregular, and porous under low DO conditions, which could contain an anoxic area inside the flocs, as seen in Fan et al. [30]. Inspired by this result, a hypothesis formed that the sludge might consist of two different conditions in order to meet the various oxygen demands of denitrification and phosphorus denitrifying removal, associating with various nutrients' capture in the same process. An SEM and metagenome analysis were utilized to prove this, which can be seen in Figures 4 and 5.
The relevant discussions are shown in Chapter 3.2.
denitrification and phosphorus denitrifying removal, associating with various nutrients' capture in the same process. An SEM and metagenome analysis were utilized to prove this, which can be seen in Figure 4 and Figure 5. The relevant discussions are shown in Chapter 3.2.

COD Removals
For the COD, the influent concentration was approximately 200-350 mg/L, while it was 250-440 mg/L for the synthesized solution. The wide range of COD concentrations were used to simulate the fluctuations of the influent wastewater in the real situation. During the whole experiment, the removal efficiencies of the anaerobic phase and critical aerobic phase were moderately impacted after the configuration changes. Of the COD, 54% and 61% were accounted for by the aerobic phase and critical aerobic phase, respectively. Figure 4 shows the observations from the SEM. From Figure 4a, it was obvious that the different morphology of the bacteria embedded together and cohered tightly. Figure 4b denoted that the co-existence of the rod-shaped and small coccus-shaped microorganisms. Figure 4c,d showed that filamentous bacteria existed in the entire process, and created an inner space to transfer the NO 3 − -N and NO 2 − -N for DAPOs, utilized with the help of the rod-shaped bacteria. Additionally, it was easy to distinguish them, and it was found that this was typical bacteria for the different functioning bacteria. For the phosphorus removal bacteria, they were short and rod-shaped, which could be seen directly from the SEM observations, which was similar to Wang et al.'s experiment [36]. It was observed that the abundances of Betaproteobacteria and Alphaproteobacteria were the highest in the bacteria community (the same as our results), and their SEM observations had large amounts of rod-shaped bacteria. Nevertheless, the SEM observation results could not fully explain the double-layer structure of sludge. Moreover, the metagenome analysis was an effective technology to unravel the predominant species in various levels and their abundance in the microbial community at these stages [37,38]. The relative abundance of the microbial diversity analyzed is presented in Figure 5. The figure shows the data from a class and genus level based on the relative abundance of the microbial population. Only the top 12 taxonomic categories at a class level and representative of 17 at a genus level are summarized here. The sequences with a low similarity or no similarity, or where the V3-V4 region did not have any alignment hits against the taxonomic database and the others could not find out had been categorized as "others".

Sludge Structure and Microbial Community Composition
At a genus level, it was obvious that Rhodocyclus, Zoogloea, Dechloromonas, Thauera, Propionivibrio, and Simplicispira took up a great proportion in three different sludge samples. After setting a critical aerobic phase in the process, the abundance of Rhodocyclus, which was the major and typical PAO in our common wastewater treatment plant [39,40], increased from less than 1% to 8.19%, and decreased to 5.69% in 1 h and 0.5 h length for critical aerobic phase, respectively. The reason was estimated to be nutrition competence. Dechloromonas (3.74% and 2.48%) and Simplicispira (4.47% and 3.21%) were the most common DPAOs for the biological wastewater treatment, which could use NO 3 − -N as the electron acceptor in an anoxic environment for oxidized organic matter [41][42][43], and their abundances both decreased after 0.5 h of the critical aerobic phase. Zoogloea was an easily adapted bacteria because they could adjust themselves either in an aerobic environment or anoxic environment for nitrifying and denitrifying [36,44]. Although the unique phenomenon had not occurred after 0.5 h, the abundance of Zoogloea was mildly enhanced (from 4.87% to 5.77%), which guaranteed the DNRA for NO 3 − -N removal [26]. However, the concentration of NO 3 − -N in the effluent was higher than previously, which demonstrated the denitrifying phosphorus removal was the main contributor for NO 3 − -N removal.
Thauera was a facultative anaerobic denitrifying bacteria that could compete with the electron acceptor of the PAOs [45][46][47][48], and its abundance was always lower than 0.5%. Propionivibrio (2.79% and 9.75%) was one genus of the GAOs in the daily wastewater treatment plants that impeded the EBPRs efficiency in high abundance and competed for nutrition with the PAOs [49]. The decrease of its abundance evidenced the former research, that GAOs were not dominant at a low temperature compared with PAOs [6,[50][51][52]. All of these changes elucidated the impact on the deterioration of the removal of both PO 4 3− -P and NO 3 − -N. On the basis of the literature, the critical aerobic phase of this experiment needed a low concentration of O 2 , in order to ensure the bacterial regularly metabolism [53], which was in accordance with our results.

Conclusions
In this study, we contrived a novel operation strategy of a pilot-scale anaerobic-critical aerobic SBR (A-CA SBR) for capturing the phosphorus of wastewater in cold regions. Accurately, almost the total PO 4 3− -P was captured and no NO 3 − -N was accumulated during the process, which were contributed to the proper critical aerobic phase length of 1 h. Moreover, from the SEM results and the metagenome analysis, Rhodocyclus, Dechloromonas, Zoogloea, and Simplicispira, which belonged to the phosphorus removal bacteria or denitrifying bacteria, co-existed in the different spaces of sludge. In essence, it was helpful for acquiring the aforementioned dominating bacteria in the subsequent artificial sludge regulation. Furthermore, the results signified that A-CA SBR was a feasible, high-efficient, and easy-handled operation strategy for domestic wastewater treatment in low-temperature conditions. Thus, this experiment not only solves the practical domestic wastewater treatment problems in winter temperatures, but also carves out an appropriate operating strategy for SBR with lower expenses. It could strongly believe that this operation mode of a pilot-scale SBR has a promising future for wastewater treatment in cold temperatures.

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