Partial denitrification with glycerol as external carbon in moving bed biofilm reactors applied for anaerobic ammonia oxidation of low nitrogen concentration secondary effluent

Introduction
Partial denitrification (PD) has been considered to be a promising alternative to partial nitrification (PN) for NO2--N generation in view of the challenge in suppressing nitrite oxidizing bacteria (NOB) under the low mainstream temperature and nitrogen level (Zhang et al. 2019). The PD prepares NO2--N for enriching anaerobic ammonium oxidizing bacteria (AnAOB) required for successful annamox (AMX). Glycerol has been regarded as a preferred carbon source for PD which has been reported to achieve PD efficiency as high as 56% to 94% (Baideme 2019, Campolong 2019, Le et al. 2019, Sharp et al. 2017, Zhang et al. 2020). However, all of these studies were performed in high strength wastewater with NO3--N > 10 mg/L and COD/NO3--N ranging from 2.5 to 4.5, which raises the question whether the same PD approach can be applied in low strength tertiary effluent. The moving bed biofilm reactors (MBBRs) have often been used as a tertiary effluent polishing process for biological nitrogen removal (BNR) to meet the permissible total nitrogen (TN), e.g., 3 mg TN/L level, in the plant discharge. Although PdNA/PANDA has been pursued in recent years to reduce carbon and aeration demand in mainstream biological nitrogen removal, the applicability of glycerol as an external carbon source for PdNA/PANDA in tertiary effluent with NO3--N and NO2--N as low as 6.3 ± 1.6 and 0.07 ± 0.10 mg N/L in the context of this study has not been explored. Such low substrate concentration makes the enrichment of denitratating culture and AnAOB especially difficult. These being said, the objectives of this study are to: i) verify the feasibility of performing nitrogen removal in the tertiary effluent via the PdNA/PANDA with glycerol as an external carbon source; ii) evaluate the system performance and unravel the mechanism behind it; and iii) make recommendations for full-scale applications of the technique.
Materials and methods Reactor setup. A MBBR treatment train dosed with glycerol were operated on site at Noman M. Cole Jr., Pollution Control Plant (NCPCP) (Figure 1). Briefly, two 14 L working volume anoxic MBBRs filled with K1 media at a volumetric fraction of 45% were connected in sequence and then followed by an 8 L working volume reaeration reactor filled with K1 media at a volumetric fraction of 32% to simulate the full-scale MBBR trains at NCPCP (Figure 1). Realtime feed-forward control. Real-time NO3--N, NO2--N, NH4+-N and DO data of tertiary effluent from Chemscan analyzer and DO probe (Figure 1) were sent to Spyder (python 3.8) in a PC. A programmable pump receiving signals from the PC can precisely dose chemicals to the system.
Activity Testing. In-situ batch tests for denitrification activity were conducted by dosing nitrate sodium at 10 mg NO3--N /L and sCOD at 50mg/L (20% glycerol) after isolating reactors and suspending COD dosing for one hours prior to the test. NO3--N, NO2--N, NH4+-N, and sCOD were evaluated with Hach tubes in 0.45μm-filtered grab samples for a period of 1 to 2.5 hours.
qPCR analysis. Quantitative polymerase chain reaction (qPCR) was performed to quantify the copy numbers of total bacterial 16S ribosomal RNA (rRNA) gene, the 16S rRNA of Candidatus Kuenenia and hzsB gene on a QuantStudio 3 thermocycler (Applied Biosystems, Waltham, MA). Each 20-μL qPCR reaction mixture contained 10 μL of 2× SYBR™ Green PCR Master Mix with ROX (Life Technologies, Carlsbad, CA), 1 μL each of forward and reverse primers (0.2 μM), 7 μL molecular-grade water (Sigma-Aldrich, St. Louis, MO), and 1 μL DNA template. Model development and calculations of kinetic parameters using Eadie–Hofstee linearization method in denitrification batch tests. In Eq. 1, rNO3- represents nitrate reduction rate, rm1 represents the maximum rate of nitrate reduction, KNO3- and KCOD1 represent the Monod half saturation constants for nitrate and organic matters in the denitratation, respectively. When nitrate is sufficient (SNO3- » KNO3-), Eq. 1 can be simplified as Eq. 2, where rm1' is the maximum rate of nitrate reduction when nitrate is sufficient. This hyperbolic function of Monod in Eq. 2 can be converted into a linear form in Eq. 2 using the Eadie–Hofstee linearization method. When we plot rNO3- against rNO3-/ SCOD, the absolute value of the slope is KCOD1 and the intercept is rm1' according to Eq. 3.
Results and discussion PD performance. By adjusting glycerol dosage, NO3--N removal could be easily achieved (Figure 2a) but PD efficiency in the Cell A dramatically fluctuated between 3 and 79% during the 330-day operation (Figure 2b). Given this unstable PD, AMX activity also dramatically fluctuated according to the %TIN removal by AMX in Figure 2b. It is noteworthy that most of the previous studies have reported the opposite results, i.e., glycerol gave stable and good PD performance. We would like to point out that almost all of these previous studies were performed in secondary or tertiary processes where NO3--N concentration was above 10 mg/L (Du et al. 2019, You et al. 2020). PdNA/PANDA in tertiary effluent with NO3--N concentration as low as 6.3 ± 1.6 mg/L used in this study has rarely been studied. Figures 2 just indicate glycerol might not be a preferred carbon source for PdNA/PANDA in low nitrogen environment.
R-strategist in glycerol-driven denitratation and denitritation. Monod kinetic parameters of denitratation and denitritation can be calculated according to the model developed in Eqs. (1) to (8). The results in Tables 1 and 2 show that ratios of the half-saturation constants (KCOD1 / KCOD2 and KNO3- / KNO2-) and the maximum rates (rm1' /rm2' and rm1'' / rm2'') determined from this study and reported in all the literature we can find are all greater than 1. That being said, the glycerol-driven denitratation has lower COD and nitrogen affinities but higher maximum reaction rates than denitritation process, i.e., R-strategy is taken by glycerol-driven PD. This could potentially explain why glycerol performs better PD at high NO3--N concentrations but otherwise in the low-strength tertiary effluent herein. Thus, it is recommended to perform glycerol-driven PdNA/PANDA in primary or secondary processes with high substrate concentrations.
qPCR results and Carbon savings. The abundance of Anammox 16S rRNA and hzsB measured by qPCR in Figure 3a evidenced the existence of AMX in Cell A but not Cell B in Figure 1. Owing to the existing of these AMX, ΔsCOD / ΔTIN profiles in Figure 3b showed that this glycerol-driven PdNA/PANDA pilot still consumed less COD to removal a unit of TIN than the full-scale MBBR. However, due to the unstable PD efficiency and AMX activity as a result of the R-strategy in glycerol-driven denitratation and denitritation (Figure 2b), glycerol-driven PdNA/PANDA might not be an optimal choice for tertiary effluent polishing.