3.1 Impact of aeration mode on ammonia and total nitrogen removal: HRT of 2.5 h with membrane scouring
Two MABRs were operated for 230 days, under the normally-scoured condition to minimize ammonia diffusion limitation to AOB. Scouring is a typical membrane operation strategy to keep biofilm thickness within an optimum range and minimize mass transfer resistance. Using scouring, ammonia diffusion within the biofilm layers will face less resistance, and consequently, higher ammonia removal efficiency is supposed to occur. In this work, the normal mode of scouring refers to 1 min long of scouring (2000 mL/min) with four times per day frequency. The ammonia loading rate in previous MABR mainstream wastewater studies ranged between 0.3-7 g/m2.d (Gilmore et al., 2013; Semmens, 2005). In this experiment, the ammonia loading rate was selected as 2.7 g/m2.d resulting in a short HRT of 2.5 h. Although the loading rate of 2.7 is within the typical range for nitrification systems, no studies have been found at a shorter HRT of 2.5 h. The shortest HRT used in the previous co-diffusional or counter-diffusional biofilm studies was 7.6 h, while the typical range was 12–28 h (Table 1). Two different aeration modes, intermittent (5 min On/12 min Off) and continuous aeration, were applied to R1 and R2.
Figure 3a represents the removal and accumulation of nitrogen compounds in intermittently- (R1) and continuously- (R2) aerated MABRs. Ammonia removal of 62% (1.7 g/m2.d) has been achieved in the continuously-aerated MABR (R2) that is significantly higher than the 40% (1.1 g/m2.d) removal achieved in R1 (P < 0.05). Having used an average C:N ratio of 2.2, TN removal efficiencies of 64% and 47% were achieved (removal rates of 0.7 and 0.8 g/m2.d) in R1 and R2, respectively. However, at lower C:N ratios (< 4–5), unstable and low denitrification efficiency has been reported in conventional (co-diffusional) biofilm systems like moving bed biofilm reactor (MBBR), and integrated fixed-film activated sludge (IFAS) (Gu et al., 2018; Iannacone et al., 2019; Liao, 2017). It is important to point out that the ammonia and total nitrogen were both removed using the same biofilm. Thus, the biofilm's overall capacity can be estimated by adding the ammonia oxidation and the nitrate or nitrite reduction rate that yields 1.9 and 2.5 gN/m2.d for the intermittently and continuously-aerated MABRs, respectively. Previous studies have looked at biofilm processes for total nitrogen removal from mainstream wastewater using IFAS mode. Nitrification rates of 0.3–1.3 g/m2.d were reported within the biofilm while the suspended growth bacteria supported the denitrification (Liao, 2017; Onnis-Hayden et al., 2007; Regmi et al., 2011). Regmi et al. (2011) reported a 50% denitrification rate while using C:N ratio of 7; however, the denitrification occurred in bulk liquid by suspended growth, not by attached growth bacteria (Regmi et al., 2011). A few numbers of investigations using hybrid biofilm and activated sludge systems observed denitrification activity and hypothesized that part of denitrifying activity might belong to the biofilm, but the report lacked further analysis to confirm the hypothesis (Germain et al., 2018; Ito et al., 2019). A study by Peeters et al. (2017) reported relatively higher nitrification rates of 1.2 and 2.6 g/m2.d in a hybrid MABR-activated sludge system where denitrification occurred in the suspended sludge zone rather than the biofilm zone (Peeters et al., 2017).
One of the challenges reported regarding TN removal efficiency is outcompeting denitrifiers by aerobic heterotrophs when C:N ratio of 5 and higher is applied (Iannacone et al., 2019). Using intermittent aeration mode in MABR, it is hypothesized that developing an environment for the denitrifiers to outcompete the aerobic heterotrophic bacteria for the limited available carbon is a possible counter-diffusional pathway that would limit oxygen availability to the most exterior biofilm layer where the aerobic carbon oxidizing bacteria typically grows. Therefore, the significance of combining counter-diffusional biofilm and supplying intermittent aeration comes to play when dramatically decreased nitrification and denitrification rates can be seen in an MBBR (co-diffusional biofilm) under C:N ratio of 5.6, which caused aerobic heterotrophs overgrowth and consequently out-competing both nitrifiers for oxygen and denitrifiers for carbon (Iannacone et al., 2019). Table 2 summarizes the recent co-diffusional biofilm studies conducted to remove total nitrogen from municipal wastewater.
Table 2
Summary of recent biofilm studies in nitrogen removal
Process
|
Reactor
|
HRT (h)
|
Inf. NH4 (mg/L)
|
Nitrification rate
|
Denit. in biofilm
|
Ref.
|
Nit/denit.
|
IFAS train
|
30
|
40
|
0.35 g/m2.d
|
None
|
(Regmi et al., 2011)
|
Nit/denit
|
IFAS
|
N.A.
|
35
|
1.12 g/m2.d
|
None
|
(Onnis-Hayden et al., 2007)
|
SND
|
Hybrid RBCa -ASb
|
8
12
|
38
|
1.63 g/m2.d
1.13 g/m2.d
(in AS)
|
1.2 g/m2.d
0.86 g/m2.d
(in biofilm)
|
(Ito et al., 2019)
|
Nitrification
|
IFAS
|
10
|
25
|
0.095 g/m2.d
|
None
|
(Liao, 2017)
|
Nitrification
|
RSc
|
16
|
35
|
0.6 kg/m3/d
|
None
|
(Hewawasam et al., 2017)
|
Nit/denit.
|
IFAS
|
6
|
27
|
0.8 g/m2.d
|
0.35 g/m2.d
|
(Germain et al., 2018)
|
SND
|
MBBR
|
48
|
62
|
0.16 g/m2.d
|
0.05 g/m2.d
|
(Iannacone et al., 2019)
|
SND
|
MABR
|
2.5
4
|
16.9
|
1.1–1.7 g/m2.d
0.9–1.4 g/m2.d
|
0.7–0.8 g/m2.d
0.4–0.6 g/m2.d
|
This study
|
a Rotating Biological Contactor
b Activated Sludge
c Rotational Sponge
In this study, the average oxygen transfer rate (OTR) and average bulk liquid DO was measured as 5.2 g/m2.d and 0.6 mg/L, respectively. The OTR values were 50% lower than the one observed in continuous aeration (11.7 g/m2.d); however, the bulk liquid DO was slightly lower, where the latter had an average DO of 0.8 mg/L. As a pathway indicator, the ratio of g O2 consumed/g NH4-N removed was calculated as 4.7 and 6.9 g/m2.d in R1 and R2, respectively. In continuously-aerated MABR, the ratio of 6.9 is relatively higher than the stoichiometric oxygen demand for nitrification (4.57 g O2 consumed/g NH4 removed). As a result, although the amount of available carbon was relatively low, the carbon removal pathway included aerobic oxidation and denitrification in continuously-aerated MABR.
It is also worth noting that, compared to intermittent aeration, a significant proportion (P<0.05) of the oxidized ammonia remained as nitrate in the continuously-aerated MABR (0.4 and 1 g/m2.d for R1 and R2, respectively) (Figure 3a). Figures 3b and 3c also present ammonia and TN removal efficiency profiles over time. The results demonstrate that 27% more of the oxidized ammonia was removed as total nitrogen in R1 compared to R2. Remarkably, this correlation between ammonia removal and nitrate production explains the trade-off between intermittent vs. continuous operation: (i) with intermittent aeration: less ammonia removal and having nitrogen gas as the dominant end-product with a saving of 70% aeration intensity or (ii) with continuous aeration: higher ammonia removal rate with more nitrate as the end-product. This result has further strengthened the conviction that MABR can be used as an adaptable technology depending on the WWTP's treatment goal(s). In WWTPs, which consider ammonia removal as the primary target, and there is no restriction to meet effluent nitrate criteria, continuous aeration will be a wise choice since it can lead to higher ammonia removal efficiency. However, when total nitrogen removal is emphasized in a treatment plant, intermittent aeration can be a suitable strategy to achieve higher total nitrogen removal efficiency.
3.2 Impact of aeration mode on ammonia and total nitrogen removal: HRT of 4 h associated with no membrane scouring
The ammonia and total nitrogen removal performance under a lower loading rate of 1.6 g/m2.d, resulting in HRT of 4 h, were characterized using a second pair of MABR pilots (R3 and R4). The MABRs were continued from previous research, where they had been running under non-scoured and intermittent aeration mode for 365 d (Mehrabi et al., 2020). In this study, both MABRs were continued under non-scoured mode; however, continuous aeration was supplied to R4. Since the ammonia loading rate has been a parameter of interest in characterizing nitrification efficiency, a lower ammonia loading rate was selected in this experiment, comparable to the loading rate used in previous studies that reported higher ammonia removal rates (Lin et al., 2016; Martin et al., 2012). At ammonia loading rate of 1 g/m2.d, nitrification efficiency was reported as 98% in MBBR; however, at higher loadings, decreased nitrification rates were reported (Andreottola et al., 2000). In agreement with the high ammonia removal efficiency at lower ammonia loadings, Lin et al. (2016) reported 0.27 g/m2.d of ammonia removal rate in MABR when loading was 0.29 g/m2.d (Lin et al., 2016). On the other hand, Peeters et al. (2017) observed 2.6 and 1.2 g/m2.d of ammonia removal rates in MABR when ammonia loading was 7 and 3.1 g/m2.d, respectively (Peeters et al., 2017). In this study, it was hypothesized that using a combination of lower ammonia loading rate and a lower C:N (< 3) wastewater, the MABR effluent that does not require downstream polishing can be generated.
Figure 4a presents the removal and accumulation rates for different nitrogen compounds in R3 (intermittent aeration) and R4 (continuous aeration) as well as Figs. 4b and 4c which demonstrate ammonia and TN removal efficiency profiles. On day 40 of the 230-day operation period, R4 faced a slough-off phenomenon, and consequently, a significant portion of sludge/biomass was washed out from the reactor, and denitrifying activity reached almost zero (Fig. 4c). The collapsed denitrification efficiency due to slough-off proved that the denitrifying population was developed on the biofilm's outer layers. In addition, the fact that the slough-off occurred in the continuously-aerated reactor confirms the overgrowth of heterotrophs and the significance of scouring to control the biofilm thickness. Under the loading of 1.6 g/m2.d, continuously-aerated non-scoured MABR (R4) achieved ammonia removal rate of 1.4 g/m2.d (86%), which resulted in an average effluent ammonia concentration of 2.5 mg/L that is aligned with discharge limits. While continuously-aerated non-scoured MABR showed the full capability of removing ammonia with no need for downstream processes, intermittently-aerated MABR (R3) achieved 0.9 g/m2.d (56%) ammonia removal rate. In terms of total nitrogen removal, R3 showed a higher TN removal capability of 70% with 30% nitrate accumulation, while a much higher nitrate accumulation (87%) was observed in R4. It should be noted that all the R4 data discussed so far include the whole experiment containing before and after the slough-off phenomenon.
Regarding the performance of R4, ammonia removal efficiency can be divided into two different periods, before and after the slough-off phenomenon. Before slough-off occurred, the ammonia removal efficiency was 75% in R4, whereas 90% efficiency was observed after the slough-off. Regarding denitrifying activity, TN removal efficiency dropped from 65–14%, and consequently, residual nitrate increased from 35–85%. These findings indicated that although denitrifying activity was destroyed due to slough-off, a 20% increase in ammonia removal efficiency was seen, which might be the result of decreased diffusion resistance for ammonia into a thinner biofilm after the slough-off. Slough-off is a common issue in any biofilm process. It is typically accompanied by loss of nitrification activity; however, this study's finding showed that it does not reduce nitrification activity and enhanced ammonia oxidation rate. Increased nitrification rate after the slough-off can confirm the reliability of MABR in terms of ammonia removal.
As discussed earlier, although extensive research has been carried out on SND in conventional biofilm systems, very few numbers of MABR studies explored the possibility of implementing a biofilm-based SND process, particularly for carbon-captured municipal wastewater. Among such MABR studies, some used hybrid systems (which do not support a biofilm-based process), and others have been restricted to apply very long HRTs, higher influent ammonia concentration, and higher C:N ratios (Table 1). For example, Lin et al. (2016) conducted a study that examined the potential of MABR in establishing SND. They reported a 62% TN removal efficiency at C:N of 3 while HRT was 24 h. However, the same TN removal efficiency was achieved in this study by the HRT of 2.5 h, which translates into a faster process by almost ten times. Hence, we hope that this study's results could eliminate the research gap in using MABR to establish a robust one-stage TN removal system.
Table 3 and Fig. 5 summarized the whole performance of the four MABRs operated under different aeration modes, scouring modes, and HRTs. Comparing ammonia and total nitrogen removal rates and efficiencies in two continuously-aerated MABRs (R2 and R4) but under normal and non-scoured conditions showed a 20% more ammonia removal in R2 compared to R4. This agrees with the finding obtained in (Mehrabi et al., 2020), where normal scouring led to a higher ammonia removal efficiency than the non-scoured MABRs. The higher ammonia removal corresponding with normal scouring is expected to relate with the scouring of the outer heterotrophic carbon oxidizing biofilm layer resulting in a reduced mass transfer resistance for ammonia diffusion through biofilm layers. Another interesting point that came off this experiment is scouring versus non-scouring in TN removal efficiency of the continuously-aerated MABRs. Under continuous aeration condition, the TN removal efficiency remained comparable despite the difference in scouring modes. This observation is quite different from the results reported in the intermittently-aerated reactors (R1 and R3) and also reported by Mehrabi et al. (2020) that indicated a higher TN removal rate of non-scoured MABRs compared to normally-scoured MABRs when intermittent aeration was supplied. That might be justified as when continuous air is introduced to the process, the scouring has no impact on total nitrogen removal efficiency due to the excess air available within the biofilm layers. This unlimited oxygen availability challenges developing an extensive anoxic biofilm layer on the MABR. Typically, the oxygen transfer rate is controlled by the substrate (ammonia and carbon) availability and biomass demand. However, is the observation in this study indicates that the nature of aeration supply mode (intermittent versus continuous) through the membrane also dictates the biochemical environment. Continuous aeration encourages the nitrifying organisms and limits the anoxic zone's availability as a desirable environment for denitrifying activity; however, further biofilm micro-level profiling is required to confirm this observation.
Table 3
Summary of operating conditions and obtained results
Operating condition
|
NH4-N (g/m2.d)
|
OTR (g/m2.d)
|
TN removal (g/m2.d)
|
NOx-N residue (g/m2.d)
|
Effluent NH4-N (mg/L)
|
Effluent TN (mg/L)
|
Aeration mode
|
HRT (h)
|
Loading
|
Removal
|
Intermittent
|
2.5
|
2.7
|
1.1
|
5.2
|
0.7
|
0.4
|
10.3
|
13
|
4
|
1.6
|
0.9
|
5.6
|
0.6
|
0.37
|
7.4
|
11.5
|
Continuous
|
2.5
|
2.7
|
1.7
|
11.7
|
0.8
|
1
|
6.3
|
12.4
|
4
|
1.6
|
1.4
|
10.6
|
0.4
|
1.08
|
2.5
|
14.2
|
3.3. Determination of maximum specific substrate removal rate
Batch experiments were carried out to characterize the performance of the MABRs to remove ammonia and total nitrogen and identify the maximum removal potential of the system. The ammonia oxidation and denitrification rate studies were carried out using CEPT wastewater and spiked CEPT wastewater, respectively (as described in Sect. 2.3). Figures 6a and b present the ammonia profile data obtained during the nitrification batch tests. The batch test showed ammonia removal rates of 3.5 and 5.4 g/m2.d in R1 and R2 (intermittent and continuously-aerated reactors at HRT of 2.5 h), respectively. As can be seen, the maximum potential of both MABRs is 3.2 times higher than the rates observed in the long-term experiment, representing the substantial nitrification capability of the MABRs if they were run at higher loading rates in a batch mode using a similar HRT of 2.5 h. In non-scoured MABRs (Fig. 6b) where the biofilm was developed on a long-term ammonia loading of 1.6 g/m2.d, ammonia removal rate obtained in batch tests were 2.3 and 2.9 g/m2.d in intermittently and continuously-aerated MABRs, respectively. Again, the maximum capability of the non-scoured MABRs is higher than the long-term results by 2-2.5 times; however, it is relatively lower than the maximum capability of the normally-scoured MABRs.
Denitrification rates were also identified in R1 and R2. Both nitrite and nitrate removal rates have been identified as the denitrification capacity of the system. The nitrite removal rate (Fig. 6c) was measured as 5 and 3.5 g/m2.d under intermittent and continuous aeration. However, the corresponding nitrate removal rates were lower (2 and 1.8 g/m2.d, respectively) (Fig. 6d). The exciting finding is the substantially higher potential (two times) of the system to remove nitrite compared to the nitrate in both reactors. This observation might be due to the concept of biofilm-based SND, which grows both nitrifiers and denitrifiers on the same medium. When denitrifiers grow adjacent to the nitrifiers, they might have the better capability to compete for the NOB for nitrite compared to the systems that denitrifiers mostly grow in suspended activated sludge. As a result, in long-term treatment, the denitrifying population might selectively improve its nitrite reduction capability rather than nitrate. However, due to the high amount of available air in R2, the denitrification rate (including both nitrite and nitrate) is relatively lower than R1, which shows the potential of intermittent aeration in MABR to enhance denitrification activity via nitrite over NOB.
3.4. Impact of aeration on the microbial community of biofilm
To further characterize the biological process, the microbial population grown on the surface of membrane fibres were analyzed. The bacterial population are presented in the phylum, class, and species-level in Fig. 7. At the phylum level (Fig. 7a), the dominant phylum, which accounted for more than 50% of the population, is Proteobacteria, including Betaproteobacteria, Gammaproteobacteria, Alphaproteobacteria, Deltaproteobacteria and Epsilonproteobacteria classes, which was reported as the dominant phylum in many WWTPs involved with biological nitrogen and carbon removal processes. Betaproteobacteria and Gammaproteobacteria classes were identified as related bacteria to nitrogen and organic matter removal. Besides, the AOB genera of Nitrosomonas belongs to Betaproteobacteria, which is the dominant AOB genera in both R1 and R2. Bacteroidetes and Chloroflexi, which were identified as two of the most popular phyla in both reactors, were observed as subdominant phyla in several WWTPs (Cydzik-Kwiatkowska and Zielińska, 2016; Gong et al., 2020; Tchobanoglous et al., 2014). The aerobic phylum of Nitrospirae grew in a significantly higher amount under continuous aeration rather than intermittent, which is consistent with the species result reporting Nitrospira sp as the most dominant species in R2 (Fig. 7c). Since unlimited oxygen was supplied to the reaction zone in continuous aeration, it justifies the higher ammonia removal efficiency and higher residual nitrate in R2. Therefore, at the species level, respective AOB and NOB relative abundance (RA) were 10.3% and 20.7% under continuous aeration (R2); however, the relative abundance of AOB and NOB were 1.3% and 1.8% under intermittent aeration (R1). This finding highlights the fact intermittent aeration significantly limits nitrifiers' growth, which might have simultaneous positive and negative impacts on the process efficiency, as previously discussed. The lower NOB relative abundance in intermittently-aerated MABR was in agreement with previous works that reported a decreased nitrite-oxidizing bacterial (NOB) activity and growth rate under intermittent aeration. The authors reported Nitrospira as the dominant NOB species rather than the Nitrobacter that is typical for continuous aeration systems (Mota et al., 2005; Yang and Yang, 2011). Similarly, another study indicated that in the biofilm regions with oxygen scarcity, Nitrospira was found as the dominant species, which has a lower substrate utilization rate than Nitrobacter (Downing and Nerenberg, 2008).
The other striking result of the microbial analysis data is the significantly different relative abundance of nitrifiers vs. denitrifiers in intermittently-aerated MABR vs. the continuously-aerated one. The relative abundance of denitrifying bacteria was measured as 40.5% in R1, which is two times higher than R2 (20.5%). It is worth mentioning that 16% and 6% of denitrifying bacteria comprised of sulphur-oxidizing bacteria such as Sulfurisoma sediminicola in R1 and R2, respectively, can use nitrate as an electron acceptor and convert it to nitrogen gas. In general, R2 hosted a narrower range of microbial populations, mostly aerobic microorganisms, while R1 grew a vast range of different species, mostly anoxic/anaerobic with denitrifying capability as Sulfurisoma sediminicola, Trichococcus pasteurii, Thiothrix disciformis, and Methyloversatilis. The higher percentage of denitrifiers and higher total nitrogen removal efficiency observed in R1 confirms the hypothesis of a well-developed anoxic region within the intermittently-aerated biofilm. In other words, denitrifiers could outcompete the NOB population for nitrite due to the favourable anoxic condition; however, NOB and denitrifiers relative abundances were almost equal in R2. Besides, while none of the reactors were seeded with anammox bacteria, it was interesting to find 0.67% and 0.9% of anammox bacteria in R1 and R2, respectively. Figure 8 shows the relative abundance of different functional populations in each MABR.
3.4.1. Impact of biofilm sample extraction method
For microbial analysis, three biofilm extraction methods from the membrane were utilized. The methods include Qiagen Powersoil DNA extraction kit, Ultrasonic bath, and Vortex Mixer were examined. In both ultrasonic bath (30 min, 70 kHz) and vortex mixer (3 min, 2500 rpm), distilled water was used as the solvent. According to the analysis results, significant differences were reported in the relative abundance of microbial populations using three different extraction methods. The relative abundance of AOB, NOB, and denitrifying bacteria for ultrasonic bath and vortex mixer methods was detected as approximately 50% lower than using the Qiagen Powersoil DNA extraction kit RA of denitrifying bacteria using vortex mixer in R1 (Fig. 9). Regardless of the negative or positive error percentage, both ultrasonic bath and vortex mixers were found highly unreliable methods to extract the biofilm sample from the media comparing to the Qiagen Powersoil DNA extraction kit.