Effect of COD:N ratio on biological nitrogen removal using full-scale step-feed in municipal wastewater treatment plants

This study investigated the effect of low and high chemical oxygen demand (COD):N ratios on biological nitrogen removal and microbial distributions in full-scale step-feed (SF) municipal wastewater treatment plants (WWTPs) in Thailand (SF1) and Taiwan (SF2). The SF1 WWTP had a low COD:N (4:1) ratio, a long solids retention time (SRT) (> 60 d), and low dissolved oxygen (DO) conditions (0.2 mg L 1 in anoxic tank and 0.9 mg L 1 in aerobic tank). The total nitrogen (TN) removal efficiency was 48%. The SF2 WWTP had a high COD:N (10:1) ratio, a short SRT (7 d), and high DO (0.6 mg L 1 in anoxic tank and 1.8 mg L 1 in aerobic tank). The TN removal efficiency was 61%. The nitrification and denitrification rates from these two plants were inadequate. Using a quantitative polymerase chain reaction (qPCR) technique, the populations of ammonium oxidizing bacteria (AOB) and ammonium oxidizing archaea were quantified. Measurement of ammonia monooxygenase (amoA) gene abundances identified these AOB: Nitrosomonas sp., Nitrosospira sp., Nitrosoccus sp. and Zoogloea sp. Higher amounts of the archaeal-amoA gene were found with long SRT, lower DO and COD:N ratios. Abundance of Nitrobacter sp. was slightly higher than Nitrospira sp. at the SF1, while abundance of Nitrobacter sp. was two orders of magnitude greater than Nitrospira sp. at the SF2. More denitrifying bacteria were of the nirS-type than the nirK-type, especially at higher COD:N ratio. Most bacteria belong to the phyla Acidobacteria, Actinobacteria Bacteroidetes, Chloroflexi, Proteobacteria. The results from this work showed that insufficient carbon sources at the SF1 and high DO concentration in anoxic tank of SF2 adversely affected nitrogen removal efficiencies. In further research work, advanced techniques on the next generation sequencing with different variable regions should be recommended in full-scale WWTPs.


Introduction
Increases in water pollution are usually related to growing urban populations. Efficient removal of nitrogen in wastewater treatment plants (WWTPs) is essential to avoid downstream eutrophication which adversely affects not only animal but also human health globally. Nitrogen is removed from wastewaters with physical methods (air stripping), chemical methods (ion exchange), biological treatment (nitrification and denitrification processes), and/ or combinations of these. Biological treatment processes are dominant over all other physical and chemical methods and are attractive because of relative low costs [1].
The most popular domestic wastewater treatment system for large communities is activated sludge process with plug flow configuration. However, with some sitespecific conditions, existing processes or equipment and demand for high biological nitrogen removal efficiency, a modification of plug flow with step-feed is recommended. Dividing a reactor tank into anoxic and aerobic zones and/or using step-feed configuration are commonly recommended for improving nitrogen removal [2]. However, not all step-feed configurations require pre-anoxic process. The step-feed process has many advantages over conventional activated sludge processes, including more uniform distribution of oxygen demand, superior ability to handle peak wet-weather flows, and flexible operation.
Step-feed systems can often achieve treatment objectives with smaller bioreactor volumes [3], and the process will often achieve low effluent total inorganic nitrogen concentrations [4].
Information on TN removal for full-scale step-feed municipal WWTP specifically for low and high chemical oxygen demand (COD):N ratios and various DO concentrations is rare in the literature. For this reason, our work was focused on two full-scale step-feed WWTPs in Bangkok, Thailand and Taipei, Taiwan. These two fullscale WWTPs with similar configurations were selected because of low and high COD:N ratios in influent. The definitions for high and low COD:N ratios of wastewaters are > 4.3:1 and ≤ 4.3:1, respectively. The study compared the efficiencies of nitrogen removal from these step-feed WWTPs, and different observations due to design parameters and operating conditions were explained. In addition, the abundance of microbial communities in these full-scale WWTPs were investigated and discussed. The results from this work could be applied to step-feed WWTPs in either country to solve carbon limitation when treating low COD:N wastewater and/or reduce aeration energy by using low-DO processes for improving biological nitrogen removal efficiencies.

Wastewater treatment systems
Two underground full-scale municipal step-feed WWTPs were selected from the downtown area of two capital cities, Bangkok, Thailand, (SF 1 ) and Taipei, Taiwan, (SF 2 ). All wastewaters samples were collected and analyzed over an entire year (2018-2019). Both plants were designed for removal of both organic matter and nitrogen with reaction tanks consisting of anoxicaerobic zones. The SF 1 (see Fig. 1a) had four feed points to four anoxic and four aerobic tanks. However, due to low flow conditions into the SF 1 system, only two feed points were operated and rotated with another two feed points. The SF 2 (see Fig. 1b) had three feed points to three anoxic tanks and three large aeration tanks (each large aeration tank was divided to four small aeration tanks) due to high flow of the system. These two WWTPs were built underground because of land limitations in these dense capital cities. The above ground areas of these two plants were used as recreation and education centers.

Analytical methods
Influent and effluent samples from each full-scale stepfeed WWTP in this work were collected monthly during 2018-2019. Characteristics of these samples were measured by using the method described in Standard Methods for the Examination of Water and Wastewater (2005). Mixed liquor suspended solids (MLSS) and mixed liquor volatile suspended solids (MLVSS) from each reactor tank were also analyzed by following the method described in Standard Methods. Temperature and pH were immediately measured in the field.

Nitrification and denitrification rates
To calculate the nitrification rates with various COD:N ratios, the concentration of ammonium-nitrogen (NH 4 + -N) in influent and effluent was determined. The nitrification rate was defined based on the NH 4 + -N removal as shown in Eq. (1) [8], where γ nitrification is the nitrification rate (d − 1 ), Q in is flow rate (m 3 d − 1 ), V reactors is volume of reactors (m 3 ), and VSS nitrifying in reactor is MLVSS of nitrifying organisms in reactor (mg L − 1 ). The nitrifying organisms in the reactor is calculated based on MLVSS using Eqs. (2) and (3), where f N is the fraction of nitrifying organisms presenting in the mixed liquor of a step-feed system. This fraction of nitrifying organisms can be estimated using Eq. (3). BOD influent and BOD effluent are the concentrations of biochemical oxygen demand in influents and effluents (mg L − 1 ), respectively. The denitrification rate was defined as Eq. (4), where γ denitrification is the denitrification rate (d − 1 ) and [TKN influent -(NH 4 + -N) influent ] can be substituted with organic nitrogen in the influent. The TN removal (%) was calculated using Eq. (5).
Microbial communities analysis Sludge samples from SF 1 and SF 2 and were taken from both anoxic and aerobic tanks for the analysis of microbial communities. The nitrifying bacterial communities were identified through analysis of ammonia monooxygenase (amoA) gene abundances of ammonium oxidizing bacteria (AOB) and ammonium oxidizing archaea (AOA). The 16S rDNA target gene of Nitrospira (NSR) and Nitrobacter (Nitro) was used to determine nitrite oxidizing bacteria (NOB) abundance. The functional targeted gene of nirK and nirS genes were used as molecular markers for denitrifying bacterial (DNB) abundances.

DNA extraction and polymerase chain reaction (PCR) amplification
1 mL of the samples were taken for DNA extraction following the manufacturer's method using FavorPrep™ soil DNA isolation mini kit (Favogen® Biotech Corp, Taiwan). The PCR protocol and oligonucleotide primers for quantitative PCR (qPCR) and denaturing gradient gel electrophoresis are shown in Table S1 in Supplemental Materials.
qPCR of functional and 16S rDNA genes

Denaturing gradient gel electrophoresis (DGGE) fingerprints
The PCR mixture contained 10X Ex Taq™ buffer, 5 units μL − 1 TaKaRa Ex Taq™, 2.5 mM dNTP Mixture, 10 pmol of each primer, 1 μL of DNA template (~10-20 ng μL − 1 ) and nuclease-free water up to 25 μL per reaction. Each sample was completed on T100™ Thermal cycler (Bio-Rad Laboratories, CA and USA). 15 μL of each PCR product was loaded into individual lanes of a DGGE gel of 8% (W/V) acrylamide gel with 35-60% (EUB) and 35-50% (AOB), and 6% (W/V) acrylamide gel with 20-50% (AOA) denaturing gradients. Electrophoresis was performed for 16 h at 58°C with a constant voltage at 80 V in 1X TAE buffer. Each DGGE band was excised with a scalpel, DNA fragment was eluted from the band by milli-Q water overnight in a refrigerator, followed by PCR with the same primer without attached CG-clamp. Representative sequences were aligned against the National Center for Biotechnology Information database using Basic Local Alignment Search Tool.
In this work, nirK and nirS genes were used because they are typically contained in denitrifying bacteria, but are structurally different from nitrite reductase The key average operating parameters of the SF 1 and SF 2 WWTPs are shown in Table 1. The characteristics of influent and effluent of each SF are shown in Table 2. The BOD:TN and COD:N ratios of SF 1 were 2:1 and 4:1, respectively. The BOD:TN and COD:N ratios of SF 2 were 4.6:1 and 10:1, respectively. The wastewater treatment loading rate, BOD:TN and COD:N ratios of SF 1 were significantly lower than those in SF 2 . Both SF 1 and SF 2 were able to remove SS, COD, and BOD well, but not nitrogen and phosphorus. At SF 1 , there was not enough carbon source (low BOD in the influent) for denitrifying bacteria as electron donor. For this reason, this insufficient carbon source would affect on denitrification process. The process performances on COD, NH 4 + -N, NO 2 − -N, and NO 3 − -N concentrations profile of SF 1 and SF 2 WWTPs are shown in Fig. 2. Although BOD:N and COD:N ratios of SF 1 were significantly lower than SF 2 , the NH 4 + -N removal efficiency of SF 1 (> 88%) was higher than that of SF 2 (59%). Average temperature of wastewater at SF 1 was higher than average temperature of wastewater at SF 2 ( Table 2). Higher average temperature and longer SRT for SF 1 could be significant factors promoting AOB activities. Although the DO concentration in the aerobic tank of SF 1 was quite low, the DO was sufficient for adequate nitrification. It is also noted that the concentration of nitrifier communities at SF 1 was significantly higher than that for SF 2 (see subsequent Section microbial communities AOB and NOB populations and communities).

Nitrification and denitrification rates
Overall biological nitrogen removal in SF 1 and SF 2 was determined by calculating nitrification and denitrification rates in the aerobic and anoxic tanks. These two rates should not be the same value [9]. In this work the nitrification and denitrification rates of SF 1 and SF 2 were significantly different (Table 3). There are several possible explanations for this inequality. First, Thai sewage piping combines wastewater and rainwater that occurs all seasons, diluting the Thai influent BOD and SS to very low levels (< 50 mg L − 1 and < 90 mg L − 1 , respectively) [10]. Second, the MLVSS:MLSS ratio of SF 1 was only 0.45-0.55 compared with that of SF 2 (0.8-0.82) due to longer SRT in both anoxic and aerobic tanks in SF 1 . For this reason, when MLVSS is used to calculate biomass, inaccurate higher estimations of microorganisms would result. The significant difference of MLVSS:MLSS ratio between SF 1 and SF 2 might also be due to the absence of primary clarifier in SF1. The main purpose of a primary clarifier is to remove solids and particulates. Third, other factors affecting the growth of nitrifiers and denitrifiers would include DO concentration, SRT duration and temperature. Especially important was maintenance of appropriate DO concentration (< 0.2 mg L − 1 ) in the anoxic phase [6].
In this study, it was shown that the longer SRT (> 60 d) of SF 1 promoted TN removal efficiency (48%) although the COD:N at this plant was quite low. Davies et al. [11] reported that longer SRTs improved nitrification and denitrification, resulting in high TN removal efficiency. At SF 2 , the operation was normal with sufficient carbon presence (high COD:N ratio) but TN removal efficiency (only 61%) was not much better than that of SF 1 . Moreover, the nitrification and denitrification rates were only 1.23 and 0.12 g NH 4 + -N g − 1 VSS d − 1 , respectively. The main reason for the low nitrogen removal efficiency and differing microbial processes was excessive DO in anoxic tank (0.6 mg L − 1 ). Other investigators [9,12,13] have stated that high DO concentration enhances nitrification rates in aerobic tank while not increasing denitrification rates in anoxic tank. Maintaining lower DO concentrations (< 1.0 mg L − 1 ) throughout an entire year in aerobic tank adversely affected the nitrification process (TN removal efficiency only 55%). Meng et al. [14] reported that TN removal was increased to 78% by increasing DO concentrations to > 1.0 mg L − 1 . Wang and Chen [6] demonstrated that a simultaneous nitrification-denitrification (anoxic-aerobic) process could result in TN removal efficiency of 57%. They also suggested that DO concentration in anoxic tank should be < 0.25 mg L − 1 and aeration should be reduced when the DO concentration exceeds 2 mg L − 1 in the aerobic tank. This current study showed that at both SF 1 and SF 2 , DO levels between 0.2 and 0.6 mg L − 1 in anoxic tank could be postulated to impact on the denitrification rate.

Microbial communities by using qPCR and DGGE technique AOB and archaea (AOA) populations and communities
The different COD:N ratios at SF 1 and SF 2 affected the sizes of archaeal-amoA (AOA) and bacterial-amoA (AOB) populations. Figure 3 shows that a large amount of AOA (1.0 × 10 5 copies mL − 1 sludge) was found with the low COD:N ratio of SF 1 . However, very low amounts of AOA (1.0 × 10 0 copies mL − 1 sludge) were found with the higher COD:N ratio of SF 2 . The very low amount of AOA found in SF 2 , but not in SF 1 , is due to the high DO concentration (> 1.8 mg L − 1 ) maintained in the aerobic tank. The quantitative results for AOA and AOB in this study were similar to that of Kayee et al. [15] who found an abundance of AOA in municipal full-scale anoxic and aerobic tanks at the Bangkok WWTP, which had low COD:N ratio (4.3:1), still higher than the COD: N ratio of SF 1 in this work. In Kayee's work, it was shown that the low DO concentration was maintained in  aerobic tank of these WWTPs. Low DO concentration and longer SRT in full-scale WWTPs promoted large populations of AOA [16][17][18]. Gao et al. [16] found that a high ratio of COD:N (10.7:1) in full-scale activated sludge plants in Beijing, China, along with low DO concentration (0.5 mg L − 1 ) in the aerobic tank, led to large populations of AOA. It is noted that the relatively high temperature of wastewater in this work (only for SF 1 ) could be a contributing factor for the abundance of AOA and AOB populations. Several studies reported the effects of warm climate on AOA and nitrifying community. For example, Limpiyakorn et al. [17] found high AOA and other nitrifying communities in domestic and industrial WWTPs in Thailand. Sinthusith et al. [18] reported that long SRT with high temperature (30°C) and pH > 7 at the WWTP in Thailand was associated with the dominance of AOA amoA genes over AOB amoA genes. As indicated in Fig. 4, AOB species in the SF 1 were the same as in SF 2 . AOB species included Nitrosomonas europaea, Nitrosomonas halophile, Nitrosospira multiformis, Nitrosospira tenuis and Zoogloea caeni via 16S rRNA of CTO primer pairs and Nitrosoccus halophilus via 16S rRNA of EUB primer pairs (see Table S2). Moreover, AOB communities present in this study were similar to those found by Shen et al. [19], who investigated the microbial community in a full-scale domestic WWTP (anoxic/oxic process). The main AOA communities at SF 1 were Crenarchaeotal sp. and uncultured Thaumarchaeote. Thaumarchaeota are autotrophic and capable of performing the oxidation of NH 4 + to NO 2 − [20,21]. Generally, Crenarchaeotes have been found in extreme environments, such as low oxygen concentrations in aquatic systems, hot springs, and full-scale anaerobic digester systems [22][23][24]. For full scale WWTP applications, it would be advantageous to maintain conditions which support AOA and AOB communities to improve biological nitrogen removal.

NOB populations and communities
In the second step of nitrification, Nitrobacter sp. and Nitrospira sp. are classically acknowledged as the most relevant NOB group in WWTPs. In Fig. 3, the copies number of Nitrospira via NSR gene were found 1.0 × 10 4 copies mL − 1 sludge at SF 1 (low COD:N ratio), but they were present at less than 1.0 × 10 2 copies mL − 1 sludge at SF 2 (high COD:N ratio). For Nitrobacter via Nitro gene 1.0 × 10 5 copies mL − 1 sludge) no significant difference was found between SF 1 and SF 2 . Yu et al. [25] reported on fluorescence in situ hybridization (FISH) technique results in their submerged membrane bioreactors under two SRTs (30 and 90 d). The fast-growing Nitrobacter with long SRT (> 60 d) was significantly higher than the amount of Nitrospira in SF 2 with short SRT (7 d). Moreover, Huang et al. [26] studied distribution of NOB communities in the full-scale WWTP by controlling DO concentration. They found that Nitrospira was dominant when low DO (< 0.9 mg L − 1 ) concentration was controlled, while Nitrobacter increased when DO concentrations were increased at higher than 0.9 mg L − 1 . This DO fact could explain why Nitrospira was dominant at SF 1 (operated with DO concentrations of 0.2-0.9 mg L − 1 ) and Nitrobacter was dominant at SF 2 (operated with DO concentrations of 0.6-1.8 mg L − 1 . Consequently, low DO conditions and long SRT would be the major operating conditions that contributed to high Nitrospira population in WWTP. It should be noted that in this work other NOB communities were not analyzed because the qPCR technique for analysis of 16S RNA would only reveal Nitrospira and Nitrobacter.

Effects of COD:N ratios on populations of DNB
The abundance of the denitrifying bacteria in this work is shown in Fig. 3. The denitrifiers are found in both anoxic and aerobic tanks of SF 1 and SF 2 . Two gene types (nirK and nirS) were used to characterize denitrifiers. The nirK-type bacteria in both SF 1 and in SF 2 were found to be 10 4 copies mL − 1 sludge in anoxic tanks and 10 5 copies mL − 1 sludge in aerobic tanks. For the nirStype denitrifiers, in SF 1 anoxic and aerobic tanks, the bacteria were present in the same order of magnitude (10 5 ). However, in SF 2 the nirS-type denitrifiers existed at two orders of magnitude higher (10 6 in anoxic tank and 10 7 in aerobic tank) than the nirK-type denitrifiers (10 4 in anoxic tank and 10 5 in aerobic tank). The higher population of nirS-type denitrifiers is attributed to the high COD:N ratio SF 2 . This observation is similar to the results from Wang et al. [27]. They found the number of nirS-type denitrifiers (10 4 to 10 5 ) was higher than that of nirK-type denitrifiers (10 3 to 10 4 ) in two full-scale WWTPs (upflow anaerobic sludge blanket and anaerobic/aerobic). Geets et al. [28] reported that nirK-type denitrifiers (10 5 ) were lower than nirS-type denitrifiers (10 6 ) in the sludge of industrial influent of anaerobic digester. They also found that nirK-type denitrifiers (10 6 ) were lower than nirS-type denitrifiers (10 7 ) in the sludge of domestic wastewater influent from hospital wastewater. From the present work and the cited studies, it is recommended that the nirS-type denitrifier growth be encouraged in full-scale WWTPs.