Effect of carbon-to-nitrogen ratio on high-rate nitrate removal in an upflow sludge blanket reactor for polluted raw water pre-treatment application

The drinking water treatment plants (DWTPs) in the developing countries urgently need an efficient pre-treatment for nitrate (NO3 ) removal to cope with the increasing NO3 − pollution in raw water. An upflow sludge blanket (USB) reactor applied for NO3 − removal from domestic wastewater may be adopted by the DWTPs. However, studies on the optimal carbon-to-nitrogen ratio (C/N) and operation of USB reactor at short hydraulic retention times (HRT) for high-rate polluted raw water pre-treatment are lacking. In this study, we first investigated the optimal C/N for biological NO3 − removal in a sequencing batch reactor (SBR). An USB reactor was then operated with the optimal C/N for pre-treating synthetic raw water contaminated with NO3 − (40 mg N L ) to monitor the NO3 − removal performance and to examine opportunities for reducing the HRT. After operating the SBR with designed C/N of 4, 3 and 2 g C g 1 N, we selected C/N of 3 g C g 1 N as the optimal ratio due to the lower carbon breakthrough and nitrite (NO2 ) accumulation in the SBR. The USB reactor achieved complete NO3 − and NO2 − removal with a lower designed C/N of 2 g C g 1 N due to the longer sludge retention time when compared with that of SBR (10 d). The high specific denitrification rate (18.7 ± 3.6 mg N g 1 mixed liquor volatile suspended solids h ) suggested a possible HRT reduction to 36 min. We successfully demonstrated an USB reactor for high-rate NO3 − removal, which could be a promising technology for DWTPs to pre-treat raw water sources polluted with NO3 .


Introduction
Nitrate (NO 3 − ) pollution is increasingly threatening the water quality in many developing countries undergoing rapid urbanization, mainly due to agricultural runoffs, livestock wastewater and discharge of inadequately treated domestic wastewater. For example, Malaysia is one these developing countries experiencing temporal elevated nitrate nitrogen (NO 3 − -N) up to 35 mg L − 1 in rivers [1,2]. To prevent elevated NO 3 − pollution from causing public health concerns, World Health Organization [3] recommended that NO 3 − -N in drinking water must not exceed 10 mg L − 1 . Most of the drinking water treatment plants (DWTPs) in the developing countries lack a treatment technology for NO 3 − removal from raw water because most of these plants are still using conventional process [4]. Thus, research study on a robust raw water pre-treatment technology for NO 3 − removal is urgently needed to protect the public health.
In recent years, biological denitrification treatment is gaining attention in NO 3 − removal rather than ion exchange, catalytic reduction, or reverse osmosis due to its high efficiency, low cost, and capability of converting NO 3 − into inert nitrogen gas [5,6]. Biological denitrification was commonly reported for NO 3 − removal from domestic wastewater in activated sludge process [7][8][9]. Besides using activated sludge, using an upflow sludge blanket (USB) reactor for biological denitrification could encourage the agglomeration of sludge into granules. The sludge granulation occurred through the washout of light flocs, while sludge with good settleability remained in the USB reactor could agglomerate into granules [10,11]. The granular sludge in the USB reactor may retain very high biomass concentration (7-130 g L − 1 ) to promote high-rate NO 3 − removal from the domestic wastewater [11][12][13]. For example, Watari et al. [11] reported granular sludge 3.8 mm in diameter with a biomass concentration between 7 and 8 g L − 1 in an USB reactor after 15 d of operation using real domestic wastewater. The USB reactor influent was supplemented with sodium nitrate and operated at a NO 3 − -N loading rate of 360 ± 90 mg L − 1 d − 1 [11]. Du et al. [12] applied USB reactor coupled with anaerobic ammonia oxidation (anammox) process to achieve 89% total nitrogen removal in treating synthetic wastewater rich in NO 3 − at a high NO 3 − -N loading rate of 720 mg L − 1 d − 1 . Jin et al. [13] has also successfully operated an USB reactor with high biomass concentration (130 g L − 1 ) and granular sludge (1.5-3.5 mm in diameter) to achieve high denitrification efficiency (98%) at a nitrite nitrogen (NO 2 − -N) loading rate of 1200 mg L − 1 d − 1 .
However, the use of USB reactor configuration for drinking water pre-treatment is still not well understood. A potential challenge of applying USB reactor for pretreating raw water polluted with NO 3 − is the treatment capacity. USB reactor was typically operated with a hydraulic retention time (HRT) between 8 h and 24 h for domestic wastewater treatment [14,15]. To meet the high treatment capacity of DWTPs, USB reactor must operate efficiently at lower HRTs, but the effect of HRT on the USB reactor performance for polluted raw water pre-treatment is still lacking in the literature. In contrast with wastewater, raw water contained very low concentration (< 5 mg L − 1 ) of organic carbon [2]. Biological denitrification requires sufficient organic carbon as electron donor, in theory 1.5 g of carbon is required to remove 1 g of NO 3 − -N by denitrification [16,17]. Thus, external carbon dosage was needed to enhance the denitrification performance. The common carbon sources commercially used include acetate, glucose, and methanol [18]. Methanol is the most widely used carbon source due to its high availability, biodegradability, and efficiency. However, methanol usage is hazardous as it is flammable, toxic, and highly reactive [19]. As an alternative solution, Peng et al. [20] suggested the use of acetate because it has low toxicity, high biodegradability and high NO 3 − utilization rate. An optimum dosage requirement expressed as carbon-to-nitrogen ratio (C/N) for raw water pre-treatment in the USB reactor is required to provide sufficient amount of carbon source for biological NO 3 − removal, at the same time over-dosage must be avoided to prevent carbon breakthrough in the effluent.
This study aimed to investigate the effect of C/N on the NO 3 − removal performance of an USB reactor pretreating synthetic raw water contaminated with NO 3 − . A sequencing batch reactor (SBR) was first operated to study the optimal C/N for biological denitrification pretreating the synthetic raw water. Subsequently, we applied the optimal C/N in an USB reactor to monitor its denitrification performance and to examine opportunities to reduce the HRT of the reactor.

SBR operation
A SBR of 1-L working volume was set up (Fig. 1). The SBR was inoculated with activated sludge obtained from a domestic wastewater treatment plant in Kuala Lumpur, Malaysia operating in SBR configuration with a preanoxic selector. Synthetic raw water contaminated with NO 3 − was prepared according to the composition listed in Table 1. Sodium acetate and sodium nitrate were used as the dissolved organic carbon (DOC) and NO 3 − sources, respectively. The synthetic raw water was autoclaved at 121°C for 20 min immediately after preparation to prevent microbial consumption of DOC and NO 3 − for their growth. Each SBR cycle consisted of five phases, including 15 min of filling phase, 300 min of reaction phase, 30 min of settling phase, 14 min of decanting phase and 1 min  of idling phase [21,22]. Anoxic condition in the SBR was achieved through the addition of 0.1 g of sodium sulfite per liter of synthetic raw water [23]. Sodium sulfite was added to ensure a controlled anoxic environment in the lab-scale SBR. The addition of sodium sulfite will not be required in full-scale DWTPs operation due to lower dissolved oxygen (DO) concentrations in real raw water. Mixed liquor within the reactor was agitated using a single impeller operating at 90 rpm throughout the reaction phase. The sludge retention time (SRT) and HRT were set at 10 d and 10 h, respectively. The SBR was operated at ambient temperature (28 ± 2°C). The DO and pH of the reactor was monitored on-line using a M300 Process 2-channel ¼ DIN transmitter coupled with an InPro6850i DO probe and a 405-DPAS-SC-K851200 combination pH/temperature probe (Mettler-Toledo, US). We operated the SBR for 22 d in three operating phases (SP1, SP2 and SP3 in Table 2). The C/N of each operating phase was increased stepwise and operated for a week to monitor the denitrification performance at different C/N. Three sampling campaigns were carried out each week. In each sampling campaign, the mixed liquor samples were collected every 2 h for chemical analyses to monitor the evolution of DOC, nitrite (NO 2 − ) and NO 3 − in the SBR.

USB reactor operation
We set up an USB reactor with a 2-L working volume (Fig. 2). The total height of the reactor was 100 cm with an internal diameter of 10 cm, the spacing between each sampling port was 15 cm. A purge gas recirculation line was installed to prevent floating sludge (Fig. 2). The nitrogen gas produced was recycled to the bottom of the USB reactor intermittently for 1 min every hour to purge the trapped nitrogen gas in the sludge flocs. We seeded the USB reactor with the same source of activated sludge for SBR as shown above. The synthetic raw water fed into the USB reactor was prepared by diluting the concentrated feedstock solution with tap water at a ratio of 1:5. Sodium acetate and potassium nitrate were used as the source of DOC and NO 3 − in the synthetic raw water, respectively. The macronutrient, micronutrient and phosphate feedstock solutions were mixed at a ratio of 1: 0.0004:1 in the feed tank. The compositions of the USB reactor influent were listed in Table 3. The synthetic raw water was fed into the USB reactor continuously using a peristaltic pump.
The HRT of the reactor was 3 h. We operated the USB reactor at ambient temperature (28 ± 2°C). We started up the reactor for 22 d to reach a stable denitrification performance with a C/N of 3 g C g − 1 N. The reactor was subsequently operated in two consecutive operating phases (UP1 and UP2) to monitor the NO 3 − removal performance of the USB reactor (Table 4). Three sampling campaigns were performed per week to collect the influent and effluent samples at the bottom and top of the USB reactor, respectively. Additionally, detailed sampling campaign were conducted biweekly to show the evolution of carbon and nitrogen species along the vertical flow length of the reactor by collecting influent sample, effluent sample and mixed liquor samples from each of the sampling port (Fig. 2).

Chemical analyses
The pH of mixed liquor samples obtained from the USB reactor was measured using an 827 pH Lab with Primatode (Metrohm, Switzerland). We filtered all the mixed liquor samples from the SBR and the USB reactor through a 0.2-μm membrane filter for anion analysis (NO 2 − and NO 3 − ) using an 861 Advanced Compact Ion Chromatograph (Metrohm, Switzerland). The DOC of the samples was measured using a TOC-V CSN total organic carbon analyzer (Shimadzu, Japan) after filtration through a 0.45-μm membrane filter. The concentrations of mixed liquor suspended solids (MLSS) and mixed liquor volatile suspended solids (MLVSS) were determined according to the Standard Methods [24].   Specific denitrification rate (SDNR) calculation The SDNR was calculated using Eq. (1).
where ΔNO x − -N (mg L − 1 ) was the sum of NO 3 − -N and 0.6 times NO 2 − -N removed in the reactor to account for the lower oxygen equivalent of NO 2 − when compared with NO 3 − [25]. MLVSS and HRT were in the unit of g L − 1 and h, respectively.

Effect of C/N on denitrification performance in SBR
The SBR was operated for 22 d ( Table 2). The pH, DOC, NO 3 − and NO 2 − profiles under different C/N in SP1, SP2 and SP3 are shown in Fig. 3. The C/N was calculated from DOC, which was a normalized parameter for the measurement of organic carbon concentration from sodium acetate. The use of DOC allows fair comparison of reactor performance with other studies using different types of carbon source, such as methanol and ethanol, under similar C/N. In SP1 operated with a designed C/N of 4 g C g − 1 N, the pH of the SBR (Fig. 3a)  increased from 8.9-9.3 in the influent to 9.4 in the effluent due to the production of alkalinity during the denitrification process [26]. The initial DOC concentration in the SBR was 118.2 ± 4.8 mg L − 1 , which was higher than the designed initial DOC concentration of 80 mg L − 1 ( Table 2). The higher measured initial DOC concentration in the SBR was caused by DOC carryover from the previous SBR cycle. Based  , we may reason that the higher initial DOC concentration in the SBR was caused by the carryover of unremoved DOC from the previous SBR cycle. The NO x − -N removed, which was defined as the sum of NO 3 − -N and 0.6 times NO 2 − -N, during SP1 was 21.7 ± 0.4 mg L − 1 , with negligible NO 3 − -N and NO 2 − -N at the end of SBR cycle (Fig. 3c  and d). A small amount of NO 2 − -N (1.3 ± 0.5 mg L − 1 ) was present at the start of the SBR reaction phase (Fig.  3d) because some NO 3 − could be reduced to NO 2 − by the heterotrophs during the anoxic filling phase. Based on the difference between the DOC at the beginning and at the end of the SBR reaction phase in SP1 (Fig.  3b), approximately 52.1 ± 0.6 mg L − 1 of DOC was consumed for both denitrification and cell synthesis. The theoretical concentration of DOC required to remove the NO x − -N present at the start of the SBR reaction phase (21.7 ± 0.4 mg L − 1 ) was only 32.6 mg L − 1 using a C/N (expressed as a ratio between carbon and nitrogen) of 1.5 g C g − 1 N [16]. The remaining DOC consumed (19.5 mg L − 1 ) may be attributed to the cell synthesis. Thus, the high DOC concentration at the beginning of SBR cycle in SP1 (118 ± 5 mg L − 1 , Fig. 3b) resulted in excess DOC supplied and significant carbon breakthrough in the effluent.
To reduce the carbon breakthrough in the effluent, we decreased the designed C/N to 3 g C g − 1 N in SP2  ( Table 2). The pH increased from 9.2 to 9.4 (Fig. 3a), indicating active denitrification in the SBR. The DOC concentration at the start of cycle was 68.8 ± 1.1 mg L − 1 , which was slightly higher than the designed concentration of 60 mg L − 1 due to the carryover of unremoved DOC from the previous cycle. The DOC at the end of cycle in SP2 (16.7 ± 1.3 mg L − 1 ; Fig. 3b) was 4 times lower than that in SP1 (66.1 ± 7.1 mg L − 1 ; Fig. 3b), thus indicating a significantly lower carbon breakthrough in the effluent during SP2 when compared with that during SP1. This implied that a designed C/N of 3 g C g − 1 N may be a suitable C/N to be adopted for NO 3 − removal from polluted raw water, while ensuring minimal carbon breakthrough in the effluent. Similar to SP1, the effluent NO 3 − -N and NO 2 − -N were negligible and the NO x − -N removed was 22.1 ± 0.1 mg L − 1 (Fig. 3c and d). From the DOC, NO 3 − and NO 2 − profiles in SP1 and SP2, the ratio of DOC to NO x − -N consumed (DOC/NO x − -N) to achieve complete denitrification was 2.4 ± 0.1 g DOC g − 1 NO x − -N. The calculated DOC/NO x − -N agreed with the C/N range of 1.7 to 2.9 g C g − 1 N typically applied for denitrification systems [17,20,27], which should be slightly higher than the theoretical C/N of 1.5 g C g − 1 NO x − -N removed reported in the literature [16]. To validate that DOC/NO x − -N of 2.4 g C g − 1 N was required for complete denitrification, we operated the SBR with a slightly lower designed C/N of 2 g C g − 1 N in SP3 ( Table 2). The pH of the reaction mixture did not increase (Fig. 3a), which indicated a reduction in denitrification activity. The DOC concentration at the start of cycle in SP3 was 39.9 ± 0.7 mg L − 1 , while its concentration reduced to 5.8 ± 0.9 mg L − 1 at the end of cycle (Fig. 3b). The NO 3 − -N and NO 2 − -N at the end of cycle were 4.4 ± 0.1 and 7.0 ± 0.7 mg L − 1 , respectively ( Fig. 3c  and d), which was comparatively higher than the negligible NO 3 − -N and NO 2 − -N at the end of cycle during SP1 and SP2. The high NO 2 − concentration in the reactor exceeded the limit for drinking water (3.0 mg L − 1 ) recommended by World Health Organization [3], which may cause lethal methemoglobinemia amongst infants. Thus, higher C/N (> 2.4 g C g − 1 N) should be applied to prevent NO 2 − accumulation. From the investigation of different C/N (2, 3 and 4 g C g − 1 N) in the SBR, we deduced that a designed C/N of 3 g C g − 1 N was the optimal ratio that should be applied for the subsequent USB reactor operation. By using a designed C/N of 3 g C g − 1 N, the carbon breakthrough could be minimized while ensuring complete NO 3 − and NO 2 − removal from the synthetic raw water.

Denitrification efficiency and SDNR in USB reactor
After a stabilization period of 22 d, the USB reactor was operated in UP1 (designed C/N = 3 g C g − 1 N) for 6 days (Table 4). Figure 4 shows the pH, DOC, NO 3 − and NO 2 − profiles of the USB reactor. The pH of the reaction mixture in USB reactor increased from 7.3-8.6 in the influent to 8.7-9.5 in the effluent during UP1 (Fig. 4a). The larger magnitude of pH increases in the USB reactor when compared with that in the SBR was caused by the higher designed influent NO 3 − -N (40 mg L − 1 ) in the USB reactor in comparison to 20 mg NO 3 − -N L − 1 in the SBR. The influent pH in the USB reactor (7.3-8.6) was also lower than that in the SBR after influent addition (9.1) because potassium hydrogen phosphate monobasic was used as the phosphate buffer during the USB reactor operation, while potassium hydrogen phosphate dibasic was used for the SBR operation. In UP1, the DOC concentration reduced from 98.9 ± 23.5 mg L − 1 in the influent to 62.1 ± 12.9 mg L − 1 (Fig. 4b). The influent DOC on day 6 was lower due to reactor operation error that caused denitrification along the transport tube from the feed tank to the USB reactor. Based on the ratio of the difference between the designed and measured influent DOC and NO x − -N on day 6, the C/N for denitrification along the transport tube was 2.3 g DOC g − 1 NO x − -N. The C/N was similar to the ratio calculated from the SBR operation, which explained the partial denitrification of NO 3 − in the influent into NO 2 − at the entrance of the USB reactor. During UP1, Fig. 4c showed that only 15.1 ± 2.4 mg L − 1 of NO 3 − -N was present in the influent, while the influent NO 2 − -N was 17.8 ± 6.3 mg L − 1 (Fig. 4d). Thus, the sum of influent NO 2 − -N and NO 3 − -N was 32.9 ± 8.4 mg L − 1 , which was close to the designed influent NO 3 − -N (40 mg L − 1 ). Both NO 3 − and NO 2 − was completely removed from the synthetic raw water with a low effluent NO 3 − -N (0.5 ± 0.1 mg L − 1 ) and NO 2 − -N (≈ 0). Based on the concentrations of DOC (36.9 ± 17.2 mg L − 1 ) and NO x − -N (25.3 ± 6.0 mg L − 1 ) removed in UP1, the calculated DOC/NO x − -N was 1.4 ± 0.4 g DOC g − 1 NO x − -N. The ratio was lower than the calculated ratio in the SBR (2.4 ± 0.1 g DOC g − 1 NO x − -N) and along the transport tube on day 6 during UP1 (2.3 g DOC g − 1 NO x − -N). A probable reason for the lower DOC/NO x − -N was the long SRT operation of USB reactor because the sludge in the reactor was not wasted regularly, when compared with a SRT of 10 d in the SBR. Nonetheless, the true SRT of the USB reactor should be evaluated in the future works considering small amount of the floating sludge that discharged together with the effluent. The longer SRT of USB reactor could promote sludge hydrolysis that supplemented the carbon source [28,29], thus the DOC supplied to the reactor may not be fully utilized for denitrification. We may further reduce the C/N of the USB reactor influent due to the lower required DOC/NO x − -N. The reduction in the influent C/N could lower the carbon breakthrough observed in UP1 (62.1 ± 12.9 mg L − 1 ). Subsequently, we operated the USB reactor for 28 d in UP2 (designed C/N = 2 g C g − 1 N; Table 4). In UP2, the pH of the USB reactor increased from 8. 1-8.7 in the influent to 9.4-9.5 in the effluent (Fig. 4a). Figure 4b shows that the carbon breakthrough in UP2 significantly improved with an effluent DOC (4.5 ± 3.2 mg L − 1 ), or 12 times lower than that in UP1 (62.1 ± 12.9 mg L − 1 ), which may serve as a practical guideline for the DWTPs' operators to adopt a lower designed C/N of 2.0 g C g − 1 N when pre-treating raw water contaminated with NO 3 − using an USB reactor. At the same time, the USB reactor achieved complete denitrification with effluent NO 3 − -N and NO 2 − -N of 0.2 ± 0.1 mg L − 1 and close to 0, respectively ( Fig. 4c and d). The corresponding NO 3 − and NO 2 − removal efficiencies were 99 ± 1% and near 100%, respectively. The USB reactor operation in UP2 implied a designed C/N of 2 g C g − 1 N was sufficient for an USB reactor system pre-treating synthetic raw water with NO 3 − -N loading rate of 320 mg L − 1 d − 1 to achieve a high denitrification efficiency with a HRT of 3 h. The HRT applied in this study was relatively shorter than other denitrification reactors with a similar or lower NO 3 − -N loading rate and C/N (all expressed in g C g − 1 N) in Table 5 [5,11,16,30,31]. For instance, Her et al. [30] reported a NO 3 − removal efficiency of 97-99% in a batch denitrification system treating synthetic wastewater (C/N = 1.9 g C g − 1 N, HRT = 12 h, NO 3 − -N loading rate = 100 mg L − 1 d − 1 ). Chen et al. [5] needed a HRT of 24 h for complete NO 3 − removal in a freshwater batch denitrification system operated with a C/ N of 3.0-5.0 g C g − 1 N and a NO 3 − -N loading rate of 5 mg L − 1 d − 1 . Similarly, a batch biofilm reactor with a C/N of 1.1 g C g − 1 N and a NO 3 − -N loading rate of 4 mg L − 1 d − 1 required a HRT of 24 h to remove 82% of the total nitrogen from the synthetic polluted water [31]. Dahab et al. [16] also reported an upflow packed bed reactor operated with a C/N and a NO 3 − -N loading rate of 1.5 g C g − 1 N and 267 mg L − 1 d − 1 , respectively, but needed a longer HRT of 9 h to achieve near complete NO 3 − removal from their synthetic raw water. In contrast, Watari et al. [11] reported that a HRT of 3 h in an USB reactor operated at a C/N and a NO 3 − -N loading rate of 1.4 g C g − 1 N and 360 ± 90 mg L − 1 d − 1 , respectively, attained a poor NO 3 − removal efficiency (35%). The poor denitrification performance was attributed to the decreasing wastewater temperature from 27 to 10°C during the winter months in Japan [11]. To explore opportunities for HRT reduction to achieve the high treatment capacity required for raw water pretreatment, we examined the detailed pH, NO x − and DOC profiles of the USB reactor during UP2 (Fig. 5). The detailed profile indicated that the denitrification reaction was completed by the first sampling port (flow length = 0.15 m). The pH of the reaction mixture increased from 8.2 to 9.4, after which the pH remained constant from the first sampling port to the effluent outlet (flow length = 0.75 m). In line with the increasing pH trend, we observed concomitant removal in DOC and NO x − in the USB reactor from the influent inlet to the first sampling port. The DOC and NO x − -N at the first sampling port were 16.2 ± 3.1 and 0.4 ± 0.1 mg L − 1 (Fig.   5). The concentrations of DOC and NO x − remained relatively constant from the first sampling port to the effluent outlet. The effluent DOC and NO x − -N were 6.9 ± 0.5 and 0.3 ± 0.2 mg L − 1 , respectively (Fig. 5). Thus, the HRT may be reduced to 36 min, which was one-fifth of the HRT of our USB reactor. By using Eq. (1) and a HRT of 36 min, the SDNR of the USB reactor was 18.7 ± 3.6 mg N g − 1 MLVSS h − 1 . The calculated SDNR for our USB reactor was within the range reported in the literature (10-24 mg N g − 1 MLVSS h − 1 ) [12,13,32]. The high SDNR of USB reactor could lead to potential reduction in HRT, which could be a promising technology for DWTPs to pre-treat raw water contaminated with NO 3 − .