Nitrate increases the capacity of an aerobic moving-bed biofilm reactor (MBBR) for winery wastewater treatment Open Access

A challenge with many industrial wastewaters, such as that from the winery industry, is the highly variable nature of flow rates and loadings (Chapman et al. 2001; Beck et al. 2005). Winery wastewater has high concentrations of readily biodegradable chemical oxygen demand (COD), primarily ethanol, sugars and organic acids (Chapman et al. 2001; Andreottola et al. 2005; Colin et al. 2005). Also, winery wastewaters are typically limited in nitrogen (N) and phosphorous (P), which are required for biological treatment (Storm 2001; Jobbágy et al. 2002). Ammonium or organic nitrogen is often added as the N source, due to their relatively low cost (Andreottola et al. 2002; Artiga et al. 2007).

Biofilm processes are increasingly used for winery wastewater treatment, as they can treat high COD loadings in compact reactors and can tolerate significant fluctuations in flows and loadings (Andreottola et al. 2005). Examples of biofilm processes for winery wastewater treatment include air-bubble column reactors, packed-bed reactors (Petruccioli et al. 2000), moving-bed biofilm reactors (MBBRs) (Andreottola et al. 2002), rotating biological contactors (Malandra et al. 2003), fixed-bed biofilm reactors (Andreottola et al. 2005), and attached growth systems based on granular sludge (Bolzonella et al. 2019).

Due to the low solubility of oxygen, dissolved oxygen (DO) often becomes depleted in the outer portions of the biofilm, limiting the carbon removal rates (Yu & Bishop 2001). Although aerobic processes can provide high removal rates for organics, they require costly aeration infrastructure and energy to provide the needed aeration supply rates, as well as nutrients to support biomass growth. These can result in high capital and operational costs (Bolzonella et al. 2019; Gupta et al. 2022). Supplementation with a more soluble electron acceptor, such as nitrate, can increase fluxes by allowing COD oxidation in the deeper, oxygen-limited regions of the biofilm. This was previously shown in a pilot-scale trickling filter for municipal wastewater treatment, where a nitrate spike considerably increased the substrate removal capacity (Vanhooren et al. 2003).

Some researchers have explored nitrate addition in suspended-growth winery treatment processes for odor control (Bories et al. 2007). Others have shown that winery wastewater can be used as an exogenous carbon source to drive denitrification (Rodriguez et al. 2007). However, no previous research has addressed nitrate addition to increase the capacity of an MBBR.

The objective of this research was to investigate the effectiveness of an aerobic MBBR for winery wastewater treatment, and the potential for enhancing COD removal by supplementing with nitrate. A bench-scale MBBR fed with synthetic winery wastewater and supplemented with urea or nitrate was operated in a continuous flow mode for over 139 days. A one-dimensional biofilm process model was developed, calibrated, and used to assess the reactor behavior.

Synthetic winery wastewater was used, following Chapman et al. (2001). The influent contained glucose (694 mgCOD/L), sucrose (853 mgCOD/L), tartaric acid (122 mgCOD/L), lactic acid (91 mgCOD/L), acetic acid (43 mgCOD/L), and diluted wine (0.01 ppv, 2,548 mgCOD/L). The theoretical COD was 4,351 mg/L. Nitrogen concentrations and organic loadings are shown in Table 1.

The reactor initially was run with a hydraulic retention time (HRT) of 25 h. Subsequently, the HRT was decreased to 9 h (Table 1). The nitrogen source was varied as shown in Table 1. The typical phosphorous requirement for heterotrophic growth is 1 mgP/gCOD. The phosphorous dosage was 0.6 mgP/gCOD through day 70, resulting in phosphorus limitation. After day 70, it was increased to above 3 mgP/gCOD.

A mathematical model was developed based on the Activated Sludge Model No. 1 (ASM1) (Henze et al. 1987). The model's processes and stoichiometry are summarized in Table 2. Model components are readily biodegradable substrate (S_S1), slowly biodegradable substrate (X_S), active heterotrophic biomass (X_H), products from biomass decay (X_P), DO (S_O) and nitrogen (ammonium or nitrate; S_N). The model included a single microbial species, heterotrophic bacteria, carrying out aerobic or anoxic respiration, growth, and decay. Hydrolysis of entrapped particulate organic compounds was also included. Since ASM1 does not include nutrient limitation nor the use of nitrate as nitrogen source, the following modifications were made: (i) the same state variable, S_N, was used to represent ammonium or nitrate, as they were never used concurrently; (ii) microorganisms could use either nitrate or ammonium as the nitrogen source with the same stoichiometry; (iii) an extra Monod term was added for nitrogen, to prevent growth when nitrogen was limiting.

Table 2

Model process matrix

.	SS1 .	XS .	XH .	XP .	SO .	SN .	Process rate .
Process
Aerobic growth of heterotrophs on SS1			1
Anoxic growth of heterotrophs on SS1			1
Decay of heterotrophs			−1
Hydrolysis of entrapped organics	1	−1

.	SS1 .	XS .	XH .	XP .	SO .	SN .	Process rate .
Process
Aerobic growth of heterotrophs on SS1			1
Anoxic growth of heterotrophs on SS1			1
Decay of heterotrophs			−1
Hydrolysis of entrapped organics	1	−1

If ammonium or urea is added, ⁠; if nitrate is added, _.

View Large

The model described in Table 2 was implemented in the computer program AQUASIM 2.1 (Reichert 1998). AQUASIM includes a one-dimensional biofilm model (Wanner & Morgenroth 2004). It allows for multiple substrates and microbial species to be modeled, and also provides a parameter estimation routine (Wanner & Gujer 1986; Wanner & Reichert 1996; Reichert & Wanner 1997). The parameters for the model were determined from measured data, parameter estimation from respirometry tests, reactor performance, and the literature (Table 3). It was assumed that the same kinetic parameters applied to bacteria in suspension and in the biofilm. The biofilm model was set for a confined reactor, a rigid biofilm matrix, and with biofilm pores filled by liquid only. The biofilm area was the internal area of carriers in the experimental MBBR. A global surface detachment rate was assumed, proportional to biofilm growth, until the observed biofilm thickness of 1.2 mm was achieved. At that point, the detachment velocity was set to the growth velocity, keeping the biofilm thickness constant.

Table 3

Parameters used for the simulation of the MBBR

Symbol .	Parameter .	Value .	Unit .	Source .
Stoichiometry
Y1	Yield for readily biodegradable substrate SS1	0.67	gcell COD/gCOD oxidized	Respirometry experiments
fP	Fraction of biomass yielding decay	0.08	_	Henze et al. (1987)
iXB	Mass of nitrogen in biomass	0.043	gN/gCOD in biomass	Calibration with reactor data
Kinetics
	Specific growth rate	1.77	1/day	Respirometry experiments
KS1	SS1 half-rate constant	4.38	gCOD/m3	Respirometry experiments
KOH	Oxygen half-rate constant	0.20	gO2/m3	Henze et al. (1987)
KN	Nitrogen (nitrate) half-rate constant	0.50	gN/m3	Henze et al. (1987)
b	Decay rate	0.16	1/day	Calibration with reactor data
ng	Correction factor for anoxic growth	0.80	_	Henze et al. (1987)
KH	Hydrolysis of entrapped organics	3.00	gslow COD/gcellCOD-day	Henze et al. (1987)
KX	Hydrolysis of entrapped organics	0.03	gslow COD/gcellCOD-day	Henze et al. (1987)
nh	Correction factor for anoxic hydrolysis	0.40	_	Henze et al. (1987)
Biofilm
DS	Soluble substrate diffusion in water (ethanol)	1.1 × 10−4	m2/day	Perry & Green (1997)
DO	Oxygen diffusion in pore water	1.7 × 10−4	m2/day	Perry & Green (1997)
DN	Nitrogen (nitrate) diffusion in pore water	2.6 × 10−4	m2/day	Perry & Green (1997)
ρ	Bacterial density	128	kg/m3	Calibration with reactor data

Symbol .	Parameter .	Value .	Unit .	Source .
Stoichiometry
Y1	Yield for readily biodegradable substrate SS1	0.67	gcell COD/gCOD oxidized	Respirometry experiments
fP	Fraction of biomass yielding decay	0.08	_	Henze et al. (1987)
iXB	Mass of nitrogen in biomass	0.043	gN/gCOD in biomass	Calibration with reactor data
Kinetics
	Specific growth rate	1.77	1/day	Respirometry experiments
KS1	SS1 half-rate constant	4.38	gCOD/m3	Respirometry experiments
KOH	Oxygen half-rate constant	0.20	gO2/m3	Henze et al. (1987)
KN	Nitrogen (nitrate) half-rate constant	0.50	gN/m3	Henze et al. (1987)
b	Decay rate	0.16	1/day	Calibration with reactor data
ng	Correction factor for anoxic growth	0.80	_	Henze et al. (1987)
KH	Hydrolysis of entrapped organics	3.00	gslow COD/gcellCOD-day	Henze et al. (1987)
KX	Hydrolysis of entrapped organics	0.03	gslow COD/gcellCOD-day	Henze et al. (1987)
nh	Correction factor for anoxic hydrolysis	0.40	_	Henze et al. (1987)
Biofilm
DS	Soluble substrate diffusion in water (ethanol)	1.1 × 10−4	m2/day	Perry & Green (1997)
DO	Oxygen diffusion in pore water	1.7 × 10−4	m2/day	Perry & Green (1997)
DN	Nitrogen (nitrate) diffusion in pore water	2.6 × 10−4	m2/day	Perry & Green (1997)
ρ	Bacterial density	128	kg/m3	Calibration with reactor data

View Large

Respirometry tests were conducted, following the static gas, static liquid principle (Spanjers et al. 1998), in a 300 mL respirometer. Temperature was kept at 20 ± 1 °C with a water bath, and a magnetic stirrer provided complete mixing. Aeration was provided by an air pump connected to aeration stones. During tests, the system was first saturated with oxygen, and then DO depletion was monitored to 1.5–2 mg/L. This process was repeated until an endogenous baseline was reached. DO was measured in the bulk liquid with an online fluorescence oxygen probe (LDO101, Hach Lange, Loveland, CO, USA), connected to a digital meter (HQ40d, Hach Lange, Loveland, CO, USA).

Three types of experiments were conducted, depending on the characteristics of the biomass. The objective was to independently test the response of suspended biomass only (Experiment 1), attached biomass only (Experiment 2), and attached biomass plus suspended biomass (Experiment 3). In all cases, biomass was aerated for at least 16 h before starting the respirometry tests, to achieve the endogenous phase. Biomass samples were taken from the bioreactor at the end of the reactor operation (day 139).

Experiment 1: A 290-mL sample of mixed liquor suspended solids (MLSS) was taken from the MBBR (818 mgVSS/L) and then tested with successive spikes of 15, 30, 45, and 60 mgCOD/L of synthetic wastewater.

Experiment 2: A sample of 100 carriers was withdrawn from the MBBR (5.2 g of solids) and washed carefully with deionized water to avoid detachment. The carriers with attached biofilm were then transferred to a respirometer with enough filtered mixed liquor to fill a total of 290 mL. Tests were conducted with successive spikes of 15, 30, and 45 mgCOD/L of synthetic wastewater.

Experiment 3: The same 100 carriers of Experiment 2 were transferred to a volume of unfiltered MLSS (1,100 mgVSS/L) to fill 290 mL in the respirometer. Tests were conducted with successive spikes of 15, 30, 45, 60 and 75 mgCOD/L of synthetic wastewater.

COD (influent and effluent), TSS, and TAS were analyzed every 2–3 days. The analyses were performed on the soluble COD fraction (sCOD), filtered with a 0.45-μm Millipore filter, using a commercial colorimetric method (Hach COD 0–1500 ppm digestion vials and Hach DR/2010 Spectrophotometer, Loveland, CO, USA). Solids were estimated as described above. Wastewater flow rates and pH (9157BN electrode, Thermo Orion, Beverly, MA, USA) were measured twice a day.

Stoichiometry (yield for readily biodegradable substrate Y₁) was obtained from each respirogram and resulted in 0.63, 0.72 and 0.67 [gcellCOD/gCOD oxidized] for Experiments 1, 2, and 3, respectively. The value obtained from Experiment 3 was further used to model the MBBR because it represented the closest condition to actual MBBR environment, which includes both suspended and attached biomass.

Kinetic parameters (specific growth and half-rate constant K_S1) were estimated using the AQUASIM parameter estimation routine in Experiment 1. Results are shown in Table 3. Experiments 2 and 3 were used to validate kinetic parameters obtained from Experiment 1 and to assess the effect of oxygen limitation within the biofilm.

The fractionation of synthetic wastewater COD was also evaluated. The mean influent sCOD was 4,338 mg/L (see Table 4), which is 99.7% of the total theoretical COD. Given that the influent COD was almost 100% soluble, the inert particulate fraction (X_I) and active biomass (X_H) were assumed negligible and a value of 0 was used. The readily biodegradable substrate, S_S1, was estimated from respirometry tests as 85% of the total sCOD. The inert soluble fraction (S_I) was estimated as 3% of the influent sCOD, given that the maximum sCOD removal observed for the MBBR was 97%. Finally, the slowly biodegradable substrate fraction (X_S) was 12%, by difference. This fractionation was similar to that obtained in previous research on winery wastewater (Beck et al. 2005).

The model provided a good fit to all three experiments (Figure 3), supporting the assumption that one set of kinetic parameters can be used to model suspended and attached biomass. As expected, the results from Experiment 1 suggested that oxygen limitation was not important when only suspended biomass was present, since oxygen was always above 1.5 mg/L. This can be seen in Figure 3(a), where the fit is the same with or without modelling oxygen limitation. However, Experiments 2 and 3 suggest that oxygen is limiting within the biofilm. This can be observed in Figure 3(b) and 3(c), where the fit is slightly better when oxygen limitation is included in the model.

A summary of the performance of the reactor and the organic loads applied during the experimental evaluation are shown in Table 4.

Following supplementary nutrient addition at day 70 (COD/N/P = 100/3.15/3.25), the viscous bulking ceased and the removal efficiency increased to an average of 95%, or approximately 23 gCOD/m²-d, between days 85 and 97 (Figure 4). At this point, achievement of the steady state was assumed and the HRT was reduced from 25 to 9 h. Another episode of viscous bulking occurred, probably due to the shock loading caused by decreased HRT. This diminished the removal efficiencies to 70%, although the removal fluxes increased to 46 gCOD/m²-d due to the higher loading. The nutrient concentration was increased on day 108 (COD/N/P = 100/7.9/4.9). The bulking ceased and the removal fluxes increased to around 60 gCOD/m²-d. From day 115 to 125, nitrate was replaced with an equivalent amount of urea (Table 1 and Figure 4). While no bulking was observed, the removals decreased to 50–63%, or 34–45 gCOD/m²-d. When returning to nitrate, between days 128 and 136, the removal efficiencies increased again to an average of 96% with no bulking episodes, and the removal fluxes increased near to 65 gCOD/m²-d. At this point, achievement of the steady state was assumed again and the experiment concluded.

Maximum removal rates for MBBRs are typically 25–35 gCOD/m²-d, and the recommended design flux is 30 gCOD/m²-d (Ødegaard et al. 2000). Aygun et al. (2008) tested loadings up to 96 gCOD/m²-d and obtained a maximum removal rate of 43 gCOD/m²-d (45% efficiency). In the case of winery wastewater, loads around 30 gCOD/m²-d have been tested in biofilm treatment systems yielding maximum removal efficiencies near 100% (Andreottola et al. 2009). In our experiments, the highest loading was 67 gCOD/m²-d. When urea was added as N source, the removal rate was 34–45 gCOD/m²-d, similar to the previously reported values. However, when nitrate was added, fluxes increased up to 65 gCOD/m²-d. As explained by Vanhooren et al. (2003), denitrification can be induced by adding nitrate, increasing considerably substrate removal capacity. As discussed in the following sections, the enhancement of COD removals with nitrate addition is further supported by our modelling.

A dynamic simulation was carried out using the model described in Section 2.4. Results showed that simulations overestimated the effluent COD when there was nutrient limitation (Figure 4). The model assumes that no growth or COD removal occurs when nutrients are deficient. However, it has been proposed that microorganisms react to a nutrient deficiency by shifting their metabolism to produce extracellular polysaccharides (EPS), which has low concentrations of nitrogen and phosphorus, instead of other cell materials (Jenkins et al. 2003). This means that under nutrient limitation, substrate oxidation is not completely inhibited and there is a certain degree of solids production (EPS), which can be also the cause of viscous bulking. This difference between the model and the actual behavior presumably leads simulations to overestimate COD (Figure 4) and underestimate solids due to EPS formation (see Supplementary Material, Figures S1 and S2).

Following the initial nutrient deficiency period (days 0–70), the model correctly reproduced the COD profile in the reactor, even accounting for the effects resulting from the change in HRT (from 25 to 9 h) and the use of urea instead of nitrate (Figure 4). In particular, when HRT was set to 9 h and nutrient dosage was COD/N = 100/7.9 (from days 108 to 139), the model was able to reproduce the decrease in COD removal due to switching from nitrate to urea as N source. This is explained by denitrification in the deeper portions of the biofilm, leading to an increase in COD oxidation activity, as is shown in Section 3.4.

During the viscous bulking episode, simulations underestimated TSS as resulted in calculated values 35% lower than measured, in average. When no bulking occurred, the model reproduced the TSS trends, even accounting for the change in HRT (from 25 to 9 h), and the use of urea instead of nitrate. TAS were typically underestimated by the model, but for HRT = 25 h they were in the correct order of magnitude. However, for HRT = 9 h, TAS reached values 15–20 times higher than predicted, and finally values over 30,000 mg/L, were produced. This suggests that the biofilm density increased when denitrification took place, since no substantial changes were observed in biofilm thickness. It is possible that changes in the biofilm internal structure occurred due to the high denitrification activity (see Section 3.4), which allowed the anoxic growth of bacteria in deeper zones of the biofilm, thus generating a denser biofilm. Another possibility is that the biofilm accumulated a significant amount of inert material.

A sensitivity analysis was carried out with the reactor model using soluble COD input and output data. The most sensitive parameters were specific growth rate and the mass of nitrogen in biomass (i_xb). Decay rate (b) and bacterial density (rho) had a small effect, meanwhile the yield for readily biodegradable substrate (Y₁) and the half-rate constant for readily substrate (K_S1) had almost no effect (see Supplementary Material, Figures S3–S5). In every scenario tested, the model showed a contribution of denitrification to COD removal, the difference was only its magnitude compared to aerobic oxidation. A better fit can be obtained if is assumed to be 20% higher than that was found in respirometry tests (R² from 0.73 to 0.74). In respirometry tests COD was present in lower concentrations than in the MBBR, thus it is possible that was underestimated.

The calibrated model was used to predict the reactor's performance when ammonium (urea) or nitrate is added at different doses. A hypothetical operation scenario was set with an HRT of 9 h and influent COD spikes of 1,000, 2,000, 3,000, 4,000, 5,000, 6,000, 7,000 and 8,000 mgCOD/L (Table 5).

At a loading rate of 20 gCOD/m²-d, there are no differences between urea and nitrate because the process is COD limited, not oxygen limited. At higher COD loadings, oxygen limitation occurs and differences in COD removal rate can be observed depending on the nitrogen source and quantity added. If ammonium is added in excess, the COD removal rate reaches a maximum of around 25 gCOD/m²-d. With nitrate, the model predicts a maximum removal rate of 63 gCOD/m²-d. If the process efficiency is set at 90%, the maximum design loading rate would be near 60 gCOD/m²-d according to the model. This is approximately double what has been observed for aerobic biofilm systems for winery wastewater treatment (Andreottola et al. 2009).

The above assessment suggests that high-rate aerobic biofilm systems can experience major capacity improvements if nitrate is properly dosed. As pointed out by Vanhooren et al. (2003), the continuous use of nitrate may not be economically feasible. However, a more cost-effective solution may be to add nitrate during peak loads only. If peak loads can be ‘shaved’ by nitrate addition, the reactor size can be minimized. While valid for any type of process, this is particularly interesting for winery wastewater treatment, where there are substantial peak loading events during the wine season (Bolzonella et al. 2019).

When choosing the nitrate dose, there is a tradeoff between improving COD removal and creating excess effluent nitrate. Modelling suggests that a COD/N = 100/6 dose may be the best choice. For future research, it is important to assess the effects of biofilm thickness on the effectiveness of this approach. For example, if a thin biofilm allows full penetration of oxygen, the nitrate addition strategy will not be effective. Alternative and more cost-effective acceptors should be studied. For example, chlorate may be a potential candidate. It is a waste product from a variety of industries, a highly energetic electron acceptor, and commonly used by bacteria in activated sludge (Bryan. & Rohlich 1954). Finally, further research should confirm that nitrate addition will not lead to formation of undesirable denitrification intermediates, such as nitric or nitrous oxide (Rassamee et al. 2011). Nevertheless, from sustainability assessments, it has been observed that MBBRs shown relatively fewer environmental burdens compared to other existing biological treatment techniques (Gupta et al. 2022).

Our research suggests the MBBR is an effective option for treating winery wastewater. In particular, the addition of nitrate as both a nitrogen source and electron acceptor can substantially increase COD removal fluxes. Experimental and modelling results suggest that nitrate addition enhances COD degradation in the deeper, anoxic regions of the biofilm. Thus, nitrate addition may be cost-effective for ‘peak shaving’ during seasonal peak loadings.

We greatly appreciate the financial support provided by Corfo, Vinnova S.A. and the Chilean Government through ANID research grants FONDECYT 1200984 and PIA/BASAL FB0002.

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.