The occurrence of enhanced biological phosphorus removal in a 200,000 m³/day partial nitration and Anammox activated sludge process at the Changi water reclamation plant, Singapore

Changi water reclamation plant (WRP) is the largest WRP (800,000 m³/d) in Singapore (Daigger et al. 2008). There are four step-feed activated sludge (SFAS) process trains in the plant, each with five step feed points and each train treating 200,000 m³/day wastewater. Of the four trains, three are operated for chemical oxygen demand (COD) and biological nutrient removal (BNR), and one for COD removal only. For the BNR process, the total sludge retention time (SRT) is five days; two and a half days each for the anoxic and aerobic zones. The sewage temperature is 28 to 32°C all year around. The yearly average removal efficiency of total nitrogen (TN) has historically been 86%. Despite the conventional BNR step-feed design, it was observed that partial nitritation occurred in the aerobic zones and Anammox in the anoxic zones of the BNR activated sludge process in Changi WRP (Cao et al. 2013,, 2016). Quantitative polymerase chain reaction (qPCR) tests (using modified primers from the primer pair of Amx694F2/Amx960R) found the potential Anammox bacteria concentration of (16S rRNA gene) 3.6 ± 3.2 × 10⁶ copies/mL, and the abundance in 16 S rRNA gene was 0.83 ± 0.18% (He 2015). The main species identified was a close relative of Candidatus Brocadia sp. 40, with 98% sequence similarity (Cao et al. 2014a; Lee et al. 2014; He 2015). It is the same organism with high growth rate cultured at Delft University of Technology (TU Delft) (Lotti et al. 2015). Enhanced biological phosphorus removal (EBPR) co-existing with partial nitritation and Anammox (PN/A) in Changi WRP was reported (Cao et al. 2013, 2016), but no quantitative results and detailed analyses were provided. The popular opinion is that EBPR is not possible or difficult in warm climates (Erdal et al. 2003; Panswad et al. 2003; Rabinowitz et al. 2004; Gu et al. 2005; Lopez-Vazquez et al. 2008; Cao 2011) due to strong competition between glycogen accumulation organism (GAO) and phosphorus accumulation organism (PAO) for volatile fatty acid (VFA) at tropical temperature (25°C − 30°C) (Lopez-Vazquez et al. 2008, 2009a). Short SRT (Whang & Park 2006), influent with constant and appropriate carbon (acetate or propionate) supply (Ong et al. 2013, 2014) as well as a higher pH (7.0 − 8.0) (Filipe et al. 2001) or segregation of different sizes of granular sludge (Winkler et al. 2011) were proposed for PAO competitive advantages, but few full-scale cases were reported to verify the ideas, until the reporting of several full-scale EBPR processes under warm climates in recent years (Ginestet et al. 2014; Jimenez et al. 2014; Sayi-Ucar et al. 2015).

This paper presents the results of a study focusing on EBPR in the mainstream PN/A process in Changi WRP. The scope included: process (in-situ) performance and kinetics, ex-situ specific activities measurement and mechanism study, microbial community analysis, and the effects of EBPR on the fate of COD, etc. The main objectives were: (i) to understand the EBPR kinetics and mechanisms, (ii) to understand the potential interactions of EBPR and PN/A through analysis of carbon balance and fate, and (iii) to explore the feasibility to accommodate EBPR in mainstream PN/A, especially EBPR in warm climates.

Ex-situ batch tests to measure specific P release and P uptake activity, and to study denitrifying phosphorus accumulation organism (DPAO) activity, were carried out mainly according to Neethling et al. (2005). The 5 L working volume beaker was used as reactor. A heater controlled the temperature inside the reactor at 30 ± 2°C. In ex-situ specific P release and uptake batch test, the anaerobic phase lasted 45 min and the aerobic phase 140 to 260 min. For DPAO study, anaerobic phase lasted 180 min with PE as carbon sources and 75 min with final effluent amended by HAc as carbon sources; anoxic phase lasted 90 min and NO₃⁻, NO₂⁻ and both NO₃⁻ and NO₂⁻ as electron acceptor were added at the beginning of the anoxic phase with an initial concentration of either 5 mg NO₂-N/L or 5 mg NO₃-N/L or both 5 mg NO₂⁻-N/L and 5 NO₃⁻-N/L; aerobic phase lasted 30 to 260 min. During the anaerobic and anoxic phase, dissolved oxygen (DO) was controlled to below 0.05 mg/L by bubbling nitrogen gas into mixed liquor; during the aerobic phase, DO was controlled to above 1.5 mg/L by using air diffusers to bubble air into the liquid. NaOH and H₂SO₄ were used to control the pH similar to the site conditions (pH: 6.5–7.2). Sludge, including RAS and MLSS at the end of the last aerobic zone, was taken from the site prior to testing in the laboratory. The volume ratio of sludge and liquor was 1.5:2.5 and MLSS in the reactor was maintained around 2,000 mg/L. A sample was taken from the reactor regularly (every 10–30 min). Liner regression on the first 30 min data was used to calculate the specific activity of P release and VFA-COD uptake during the anaerobic phase. Similarly, liner regression on the first 60 min data was used to calculated specific PO₄⁻, NO₂⁻ and NO₃⁻ uptake activity during the anoxic phase, and the data from the first 30 or 60 min were used for calculating specific P uptake activity during the aerobic phase.

Standard methods (APHA 1998) were adopted in the analyses of NH₄⁺, NO₃⁻_, COD, TP PO₄⁻-P_, ALK, and MLSS. NO₂⁻ was analysed according to USEPA (1987). DO and temperature were measured by portable meter (YSI 85D). For all the liquid composition analyses the samples were filtered (0.45 μm, Whatman) immediately after sampling on site, then stored within 1 h at a temperature below 4°C prior to analysis. The samples for VFA analysis were filtered prior to acetate, propionate, butyrate and valerate analyses using gas chromatography (Prominence, Shimadzu) equipped with a flame ionization detector fitted with a DB-FFAP (30 m length, 0.25 m diameter, and 0.25 μm film) column. Specific oxygen uptake rate (SOUR) and nitrate uptake rate (NUR) measurements were according to Melcer et al. (2003). For fluorescent in-situ hybridization (FISH) analysis of PAO and GAO including the probes and procedures, etc.; see Winkler et al. (2011).

Three factors contributed to the higher P release and uptake in the initial step feed pass and the subsequent decrease in the following passes. (i) The real anaerobic environment of the first ‘anoxic’ zone because of little NO₃⁻ and NO₂⁻ (<1.0 N mg/L) present in the PE and RAS. This resulted in little competition for carbon between PAO and heterotrophic denitrifiers, while the other four anoxic zones received higher NO₂⁻ and NO₃⁻ (>4 mg N/L) from the preceding aerobic zones, resulting in less carbon for PAO due to the competition with heterotrophic denitrifiers. As a consequence, the intracellular stored PHA/glycogen for PAO decreased, impairing P uptake activity in the aerobic zone (Neethling et al. 2005). (ii) P taken up by DPAO (see section Denitrifying PAO) occurring in the second to fifth anoxic zones, which reduced the apparent P release activity observed. (iii) The intracellular PHA/PHV storage of PAO increased after FST as reported by (Yang et al. 2016). The batch tests also showed that following a pre-anoxic phase with little sCOD in the liquid, the P release and sCOD uptake rate increased during the anaerobic phase (data not shown). More investigations on the mechanism(s) involved are needed.

The ratio of P release/sCOD uptake in the first anoxic zone varied from 0.28 to 0.60, with an average of 0.45, which was close to 0.50 mg P/mg VFA-COD, the stoichiometric ratio for PAO at a temperature of 25°C and pH 7 (Smolders et al. 1994) and was in the range (>0.2 mg P/mg VFA-COD) indicating PAO dominance according to Schuler & Jenkins (2003).

In addition to stable supply of VFA-COD of the feed, plug-flow anoxic-aerobic reactor configuration, short aerobic SRT (2.5 d) and effective PN/A process contribute to the successful EBPR co-existence with PN/A in the SFAS process at the Changi WRP. The effective PN/A process plays a determining role in little NO₃⁻/NO₂⁻ in RAS, thus generating an anaerobic environment in the first anoxic zone favouring P release and VFA-COD uptake. The essential role of PN/A process in providing ‘additional’ VFA-COD utilization to PAO compared to heterotrophic denitrification is illustrated in the following section of fate of COD. The short aerobic SRT favours proliferation of PAO and wash-out of GAO (Daigger et al. 1988; Whang & Park 2006; Whang et al. 2007; Barnard & Comeau 2014). As a consequence, efficient EBPR under warm climates can be achieved under limited supply of VFA-COD as demonstrated in this study and other full-scale processes (Onnis-Hayden et al. 2013; Ginestet et al. 2014). Furthermore, recent reports (Cao et al. 2013; Ong et al. 2013, 2014; Sayi-Ucar et al. 2015) showed even under a longer SRT, high efficient EBPR can be achieved under warm climates when a high influent bCOD/N ratio (e.g. >10) is maintained or additional carbon input is provided. In this situation, PAO are able to share VFA-COD, co-exist with GAO, and to be active in the process. In summary, effective EBPR is achievable in warm climates under several conditions. This paper also illustrates stable EBPR with dominance of PAO population can be achieved and maintained in warm climates when acetate, rather than propionate, is predominantly in the feed. This may provide an opportunity to re-think the popular opinion, which forms the basis of modelling EBPR in high temperatures, that propionate supports sustainable PAO population, while acetate favours GAO (Whang et al. 2007; Lopez-Vazqueza et al. 2009b; Oehmen et al. 2010).

To reduce the final effluent P concentration, it is necessary to explore alternate operational strategy for the FST to reduce the high returning PO₄⁻-P loading from the RAS to the activated sludge process. Also, the centrifuge thickening operation needs attention.

For an MLE process with a similar influent C/N ratio as in this study in Singapore, 20 mg NO₃⁻-N/L was denitrified (Cao et al. 2014b). Assuming 5.7 mg COD/L used per mg NO₃⁻-N/L heterotrophically denitrified (Henze et al. 1997), the COD used for conventional denitrification dissimilation will be 57 mg COD/L (accounting for 17% of influent COD). This suggests that PN/A provided an additional 21 mg COD/L for EBPR dissimilation, which can remove about 4 mg PO₄⁻-P/L. Kinetically, fast uptake of VFA-COD by PAO in the anaerobic/anoxic zone largely limited the carbon available for heterotrophic denitrification (Onnis-Hayden et al. 2013). It has the same effect as a reduction in the influent BOD₅/N ratio from 3.4 to 1.6 favouring the Anammox process. The case reported in this paper and the EBPR in full-scale nitrogen short cut process in the city of St Petersburg Southwest WRF, USA (Jimenez et al. 2014) demonstrate the feasibility to accommodate EBPR into the mainstream PN and PN/A process. The local conditions, typically temperature, etc., should be taken into account when considering this feasibility.

High efficiency EBPR was achieved concomitantly with excellent nitrogen removal in a 200,000 m³/d mainstream PN/A activated sludge process in Singapore. Ex-situ specific P release and uptake activities were comparable with those of effective EBPR in the USA and Europe. PAO populations were dominant and able to use NO₃⁻ and NO₂⁻ as electron acceptors to perform PO₄⁻-P uptake in anoxic conditions. Few GAO were observed as compared to PAO. COD oxidation (11% of input COD) by EBPR dissimilation was the same as that of heterotrophic denitrification. Compared to conventional biological nitrogen removal, much lower carbon demand for nitrogen removal owing to autotrophic PN/A conversions made effective EBPR possible. This paper demonstrated for the first time the occurrence of EBPR co-existing with PN/A in mainstream process, the feasibility of accommodating EBPR into mainstream PN/A, and the interactions between C, N- and P Cycle forming an energy and resource – efficient BNR in a large full-scale activated sludge process. It also showed that EBPR can work under warm climates.

Singapore Centre for Life and Environment Sciences Engineering (SCELSE) helped on the VFA measurement.