Kinetics, diffusional limitation and microscale distribution of chemistry and organisms in a CANON reactor

Abstract

In the Completely Autotrophic Nitrogen removal Over Nitrite (CANON) process, aerobic and anaerobic ammonia oxidizing bacteria cooperate to remove ammonia in one oxygen-limited reactor. Kinetic studies, microsensor analysis, and fluorescence in situ hybridization on CANON biomass showed a partial differentiation of processes and organisms within and among aggregates. Under normal oxygen-limited conditions (∼5 μM O₂), aerobic ammonia oxidation (nitrification) was restricted to an outer shell (<100 μm) while anaerobic ammonia oxidation (anammox) was found in the central anoxic parts. Larger type aggregates (>500 μm) accounted for 68% of the anammox potential whereas 65% of the nitrification potential was found in the smaller aggregates (<500 μm). Analysis with ∼5 μM O₂ microsensors showed that the thickness of the activity zones varied as a function of bulk O₂ and NO⁻₂ concentrations and flow rate.

1 Introduction

Nitrogen removal is one of the most important and costly elements in wastewater treatment and is normally carried out by sequential nitrification–denitrification processes. Ammonium NH⁺₃ is oxidized to nitrate NO⁻₃ followed by a reduction of NO⁻₃ to free nitrogen (N₂). In the mid-1990s, an alternative process of nitrogen removal was described, the so-called anammox process [1,2]. Under completely anoxic conditions NH⁺₄ is oxidized with nitrite NO⁻₂ as electron acceptor to N₂ and small amounts of NO⁻₃[3]. The anammox process was first discovered in a wastewater treatment plant in Delft, The Netherlands, and today anammox activity has been reported from several other treatment plants [4–8]. Recently the process has also been shown to occur in nature in marine sediments and anoxic water columns [9–11]. Anammox is carried out by autotrophic bacteria belonging to the order Planctomycetales, like Candidatus Brocadia anammoxidans and Candidatus Kuenenia stuttgartiensis[4,12].

Since the very first description the focus has been on optimization of the anammox process for wastewater treatment. Compared to alternative treatment methods the anammox process offers several advantages and is especially suited for the treatment of high ammonia wastewater [13,14]. However, to work optimally the anammox process has to be supplied with an approximately 1:1 mixture of NH⁺₄ and NO⁻₂. One approach is partial nitrification of NH⁺₄ to NO⁻₂ in an aerated reactor operated as a chemostat by the so-called SHARON process [15,16]. In this reactor a combination of short residence time and high temperature prevents the growth of NO⁻₂ oxidizing bacteria, and the effluent containing NH⁺₄ and NO⁻₂ is subsequently fed to an anoxic anammox reactor [13,17]. Recently, a new approach has been described, the CANON reactor, where aerobic ammonia oxidation and anammox are occurring within the physical constrains of the same reactor [18–20]. This reactor is operated at oxygen-limited conditions allowing both aerobic and anaerobic processes at the same time. Further oxidation of NO⁻₂ to NO⁻₃ is prevented by reactor operation at high NH⁺₄ concentration (5 mM) and oxygen-limited conditions. A nitrogen removal efficiency of up to 85% has been reported[18]. However, in terms of wastewater treatment, the actual nitrogen removal rate is crucial and dependent on O₂ mass transfer efficiency from the gas to the liquid phase. Using a gas-lift reactor with more efficient mass transfer than a sequencing batch reactor (SBR) it is possible to increase nitrogen removal rate from 0.07 to 1.5 kg N/m3/day[19].

For future optimization of the CANON process, it is important to obtain detailed information about kinetics, diffusional limitation, and microscale distribution of chemistry and organisms. We have obtained such information by analysis of the reactor processes with micro- and macroscale sensors for O² and NO⁻₂. Microbial aggregates were furthermore analyzed with fluorescence in situ hybridization (FISH) so that the microdistribution of organisms could be compared with the microdistribution of chemistry and processes.

2 Materials and methods

2.1 Mineral salt medium

For culturing and experiments a mineral salt medium supplemented with different amounts of NH⁺₄ and NO⁻₂ has been used: 1.25 g/l KHCO₃, 0.025 g/l KH₂PO₄, 0.3 g/l CaCl₂× 2H₂O, 0.2 g/l MgSO₄× 7H₂O, 6.25 mg/l FeSO₄, 6.25 mg/l EDTA and 1.25 ml/l trace elements according to Sliekers et al.[18].

2.2 CANON reactor

The CANON reactor has been described in detail by Sliekers et al.[18]. Briefly, it is a SBR with a working volume of between 1 and 2 L, operated at 12 h cycles including a 11.5 h filling period, a 0.25 h settling period, and a 0.25 h draining period. Biomass in the reactor primarily consisted of aggregates up to a size of ∼1.5–2 mm in diameter and some biofilm on reactor walls. The reactor was kept at a temperature of 30 °C. During the filling period the system was stirred to keep the aggregates in suspension. The reactor was fed with mineral salt medium containing 9.4 nM NH⁺₄, and oxygen was supplied from a 91.8% Ar/8.2% O₂ mixture with a bubbling rate of 30.5 ml/min. During normal operation typical concentrations of O² NH⁺₄ and NO⁻₂ in the reactor were about 5 μM, 5 mM, and 45 μM, respectively.

2.3 NO⁻₂ and O² microsensors

NO⁻₂ concentrations were measured with a new NO⁻₂ biosensor consisting of an immobilized culture of a NO⁻₄reducing/N² O⁻ producing bacterial strain (Stenotrophomonas nitritireducens) coupled to a Clark type N₂O-microsensor[21]. The sensor was constructed in micro- and macroscale versions. At 30 °C the macroscale sensors had 90% response time of about 1–2 min, a detection limit of about 1μM NO⁻₂, and a linear range up to 0.5–1.5 mM NO⁻₂. Microscale sensors were constructed with a tip diameter of about 30–35 μm and were operated with a positive tip potential to enhance sensitivity [22,23]. Operated at +0.4 V tip potential the detection limit was about 1μM NOSO⁻₂ and the range for linear response was 0–0.25 mM NOSO⁻₂. The 90% response time was 30–50 s. Two point in situ calibrations were used for both types of biosensors. During our investigations of the CANON reactor we had unexpected problems with the long-term stability of the NO⁻₂ biosensors as they became sensitive also to NO⁻₃ due to rapid bacterial contamination. This problem that limited both the duration of continuous measurements and the number of replicates has subsequently been solved by addition of tungstate to the biosensor interior (M. Nielsen, unpublished).

Clark-type O₂-microsensors equipped with guard cathode[24] were used for microscale analysis of O₂. The sensors were constructed with tip diameters of about 10 μm and had 90% response time <1 s. Bulk O₂ concentrations in the CANON reactor were measured with STOX (Switch able Trace amount OXygen) sensors (Unisense, Aarhus, Denmark) with 90% response time of about 5 s. The STOX sensor is a modified Clark type O₂ sensor with two front cathodes designed for measurements of very low O₂ concentrations down to 0.01–0.02 μM. An electrically mediated activation/deactivation of the cathode closest to the tip gives a very precise in situ zero calibration and explains the very high sensitivity of the sensor. The O₂ sensors were two point calibrated in air saturated reactor water and anoxic alkaline ascorbic acid solution.

2.4 CANON reactor studies

On-line monitoring of NO⁻₂ and O₂ concentrations in the CANON reactor was performed with macroscale NO⁻₂ biosensors and STOX sensors connected to a data logger (ADC-16, Pico Technology). A sampling frequency of 30 data points per hour was used. Experiments with the CANON reactor were performed during two experimental periods: 2001 and 2002 campaigns. Identical reactors were used but with some variation in biomass composition possibly due to a difference in reactor operational lifetime. The first set of measurements were performed on a reactor characterized by relatively large aggregates and with significant wall-growth of biofilm while the second reactor had been in operation for a shorter time and had relatively smaller aggregates and less biofilm growth.

Oxygen perturbation experiments were performed on the CANON reactor when operated as a SBR-reactor (2001 campaign) and when operated as a 1.6 L batch system (2002 campaign). A partial pressure of O₂ in the inflowing gas ranging between 5% and 90% was used. To maintain a high ammonia concentration during batch operation where the flow feed into the reactor was stopped, NH⁺₄ was supplied manually at regular intervals. At different oxygen supply rates complete 12 h cycles for O² and NO⁻₂ were recorded during SBR operation, while steady state values for O² and NO⁻₂ were obtained during batch operation.

2.5 Mini-reactor studies

A mini reactor was used to determine the kinetic characteristics of aerobic and anaerobic ammonia oxidation of the CANON biomass. The mini reactor was constructed as a closed system and consisted of a stoppered conical flask with a liquid phase of 110 ml and a gas phase of ∼20 ml. Stirring of the system was accomplished with a magnetic stirring bar (4 cm long, 100 rpm) and temperature was kept at 30 °C (±1 °C). The gas composition of the headspace was controlled through a valve. The concentrations of NO⁻₂ and O₂ in the bulk liquid were monitored by insertion of a macroscale NO⁻₂ biosensor and a microscale O₂ sensor into the reactor. Liquid with aggregates (110 ml) was retrieved from the operating CANON reactor (2002 campaign) and separated by sieving through a screen into aggregates <500 μm and >500 μm. Each of the two size fractions were subsequently re-suspended into 110 ml of fresh mineral salts medium.

Aerobic and anaerobic ammonia oxidizing activities were determined from O² and NO⁻₂ consumption rates during oxic and anoxic incubations, assuming that nitrification and anammox were the only O² and NO⁻₂ consuming processes in the reactor. Aerobic and anaerobic ammonia oxidizing bacteria have been shown to constitute about 85% of the CANON biomass[18]. Furthermore, general lack of organic electron donors in the reactor limits alternative O² and NO⁻₂ consuming processes. Oxygen was supplied to the reactor by bubbling the liquid phase with pure O₂, and the headspace was subsequently flushed with N₂. A correction factor to compensate measured O₂ consumption rates for O₂ loss at the gas/water interface was determined in a separate experiment where there was no biomass in the mini reactor. Nitrite was added to the anoxic reactor after bubbling the liquid phase with N₂.

2.6 Microscale analysis

Analysis of the spatial distribution of metabolic processes in CANON aggregates was performed with O² and NO⁻₂ microscale sensors in a flow chamber (Fig. 1). The flow chamber was flushed with mineral medium supplemented with 5 mM NH⁺₄ and either 30 or 100 μM NO⁻₂. The flow inside the chamber was controlled with a peristaltic pump and the effluent from the flow chamber was re-circulated to a 10 L medium reservoir. The temperature was kept at 30 °C (±1 °C) and the O₂ concentration in the system was adjusted with a mass flow controller (Model 5878, Brooks Instrument B.V) mixing air with N₂.

Larger aggregates (1–1.3 mm) were taken directly from the CANON-reactor (2002 campaign) and placed on a metal grid where they were exposed to a downward flow. Two different flow settings were applied: ∼0.2 and 1 cm s⁻¹. The flow field was not totally homogeneous, so we can only indicate the approximate flow at the location of the aggregate. With the sensor penetrating vertically from above, profiles were made on the flow side of the aggregate. A pre-incubation period of 1 h was used to create (semi-) steady state conditions. Concentration profiles were recorded by introducing the sensors into the aggregates at 50 μm steps using a micromanipulator. The sensor signal was continuously recorded on a strip-chart recorder. A dissection microscope was used to determine the position of the aggregate/water interface.

Rates of O² and NO⁻₂ uptake by the granules were calculated from the concentration gradients in the diffusive boundary layer using Fick's first law of diffusion, J=−D(dC/dx), where J is the flux, D is the diffusion coefficient for O₂ (2.73 × 10⁻⁵ cm² s⁻¹,[25]) or NO⁻₂ (2.02 × 10⁻⁵ cm² s⁻¹,[26]) in water at a temperature of 30 °C, and dC/dx is the concentration gradient.

2.7 Fluorescence in situ hybridization

Aggregates from the CANON reactor were fixed in paraformaldehyde solution and cryosectioned as described previously[27]. Fluorescence in situ hybridizations were performed according to Schmid et al.[4]. The probes used in this study were S-∗-Neu-0653-a-A-18 (NEU) for halophilic and halotolerant Nitrosomonas spp. (N. europaea, N. eutropha, N. mobilis, N. halophila), S-P-Betao-1225-a-A-20 for most betaproteobacterial aerobic ammonia oxidizers and S-∗-Amx-0820-a-A-22 for anammox bacteria (Candidatus “Kuenenia”, Candidatus “Brocadia”). Information about all probes can be found in probeBase.net[28]. After hybridization slides were washed briefly with ddH₂O, air-dried and embedded in Vectashield. For image acquisitions a Zeiss Axioplan microscope (Zeiss, Jena, Germany) was used together with the standard software packages delivered with the instruments.

3 Results

3.1 CANON reactor studies

Recorded cycles of NO⁻₂2 and O₂ during 12 h cycles in the CANON reactor (Fig. 2) showed a consistent pattern that was present at all oxygen supply rates examined. The O₂ concentration in the reactor remained at a stable level during the complete cycle whereas the NO⁻₂ concentration showed an initial increase followed by a leveling off and a subsequent slow decrease. Increasing the oxygen supply rate led to higher levels of both O² and NO⁻₂ in the reactor. Compared to normal operational setting the highest supply rate tested resulted in a 2.5-fold increase in bulk O₂ concentration and an almost 3-fold increase in NO⁻₂ concentration (Table 1). However, a saturation of either aerobic or anaerobic ammonia oxidation was not observed, as such saturation would have caused rising O² or NO⁻₂ concentrations during the cycle. Steady state concentrations for O² and NO⁻₂ obtained during batch operation of the reactor (Table 1) showed the same effect of enhanced oxygenation. At steady state the NO⁻₂ concentration was about 7–10 times higher than the O₂ concentration.

3.2 Mini-reactor studies

The O₂ depletion rate was most rapid for the flasks containing small aggregates (<500 μm) with initial rates of 15–16 μmol L⁻¹ min⁻¹, while the rate for the large aggregates (>500 μm) was about 8 μmol L⁻¹ min⁻¹ (Fig. 3(a)+(c), Table 2). It was difficult to estimate an apparent K_m value for the fraction consisting of the larger aggregates due to the rather heterogeneous rate of decrease, but it was close (about 12 μM) to the apparent K_m value of 14 μM that could be estimated for the small aggregates. The O₂ concentrations observed in the CANON reactor were always below or near these apparent K_m values (Table 1) and calculated from the slopes in Fig. 3(c) the turnover time for oxygen in the reactor was found to be about 1 min. The initial NO⁻₂ depletion rate under anoxic conditions was higher for the large aggregates (5.1 μmol L⁻¹ min⁻¹) than for the small aggregates (2.4 μmol L⁻¹ min⁻¹), and the apparent K_m value was also higher (23 μM) for the large aggregates than for the small aggregates (14 μM) (Fig. 3(b)–(d), Table 2). The larger difference in apparent K_m for NO⁻₂ uptake indicates a distribution where annamox bacteria are found within the anoxic central parts of the aggregate, and where nitrite thus has a long diffusion path before it reaches all active zones in large aggregates. The largest (65%) aerobic ammonia oxidizing capacity (O₂ uptake) was located in the aggregates <500 μm, whereas the largest (68%) anaerobic ammonia oxidizing capacity (NO⁻₂) was found for the aggregates >500 μm (Table 2). This result shows a size dependent skewed distribution of processes among aggregate. Calculations of weight specific activity rates yield an almost three times higher O₂ uptake rate for the smaller aggregates while NO⁻₂ uptake rates where about 40% higher for the larger sized aggregates as compared to the small aggregates.

3.3 Microscale analysis

The microprofiles were performed under as realistic conditions in terms of water flow, ambient chemistry, and temperature as we could obtain. The applied flow system made it possible to perform microscale analysis at very low O₂ concentrations (2–12 μM) comparable to in situ reactor levels.

O² and NO⁻₂ profiles were determined in large (1.0–1.3 mm diameter) aggregates with variation in external O² and NO⁻₂ concentrations (Fig. 4). The profiles obtained at O₂ concentrations of 2 and 9 μM (Fig. 4(a) and (b)) represent “extremes” as compared to a concentration of about 5 μM O₂ during normal reactor operation. In both cases there was a spatial separation of processes with the O₂ reduction zone found in the upper <100 μm and the NO⁻₂ reduction zone starting at the oxic/anoxic interface and penetrating to a depth of 250–300 μm. Elevation of the external NO⁻₂ concentration from 30 to 100 μM (Fig. 4(c)) led to a very significant increase in NO⁻₂ penetration depth to about 550 μm. We did several replicate measurements at this high NO⁻₂ concentration, and some profiles showed complete NO⁻₂ penetration of the aggregates. There were, however, pronounced difficulties with these measurements, as the CANON aggregates turned out to be extremely tough to penetrate to greater depth with the relatively thick (about 30–35 μm tip diameter) and relatively conical NO⁻₂ biosensors, and the insertion thus caused a variable compression of the aggregate. The profiles (Fig. 4(d)) measured at a lower flow rate (about 0.2 cm s⁻¹ as compared to 1 cm s⁻¹) deviate from the other profiles by a 100 μm thicker (∼250 μm) diffusive boundary layer.

O² and NO⁻₂ uptake rates (Table 3) were strongly influenced by bulk water concentrations and boundary layer conditions. Significantly higher uptake rates were found at elevated O₂ concentrations with a 5-fold increase in O₂ uptake rate at 9 μM as compared to 2 μM. At similar high O₂ concentration the NO⁻₂ uptake rate at 100 μ NO⁻₂ was 180% of the uptake rate at 30 μ NO⁻₂. Increasing the boundary layer thickness from about 150 μm to about 250 μm caused 20–40% reduction in O² and NO⁻₂ uptake rates.

At the high O₂ concentrations of 9 μM (Fig. 4(b)) the O₂ flux was about 1.28 × 10⁻⁵μmol cm⁻² s⁻¹ (Table 3). This flux corresponds to a NO⁻₂ production of 0.85 × 10⁻⁵μmol cm⁻² s⁻¹, assuming that all O₂ was used for aerobic ammonia oxidation. The NO⁻₂ consumption rate by anaerobic ammonia oxidation calculated from the NO⁻₂ profile (Fig. 4(b)) was 2.33 × 10⁻⁵μmol cm² s¹. Nitrification in the oxic region of the aggregate thus supplied about 36% of the NO⁻₂ used for the anammox process in the deeper anoxic layers. At the low O₂ of 2 μM (Fig. 4(a)) the rate of O₂ uptake decreased to 2.3 × 10⁻⁶μmol cm⁻² s⁻¹ and nitrification within the aggregate could thus only supply about 11% of the NO⁻₂ uptake of 2.16 × 10⁻⁵μmol cm⁻² s⁻¹. It should be stressed that these experiments were done in a flow-cell with fixed concentrations of O² and NO⁻₂. In the CANON reactor, the concentration of NO⁻₂ is strongly affected by the O₂ concentration as shown in Table 1. The conditions applied for the data shown in Fig. 4(b) are, however, not far from actual reactor conditions, and the data thus show that large aggregates such as the one analyzed consume more NO⁻₂ than they produce.

3.4 Fluorescence in situ hybridization

Fluorescence in situ hybridisation indicated a distinct distribution of aerobic ammonia oxidizing bacteria and anaerobic ammonium oxidizing bacteria within the aggregate (Fig. 5). While aerobic ammonia oxidizers were located at the surface layer of the granule, anaerobic ammonia oxidizers occupied most of the interior parts.

4 Discussion

4.1 Distribution of nitrification and anammox within aggregates

The combination of reactor (Fig. 3), microsensor (Fig. 4) and, FISH (Fig. 5) data show that aerobic ammonia oxidation by N. europaea affiliated bacteria was limited to an oxic <0.1 mm thin surface layer of the CANON reactor aggregates, and that anaerobic ammonia oxidation by anammox bacteria occurred in the deeper anoxic layers. The CANON reactor granules thus functioned according to the “magic bead concept”[29], where one aggregate can mediate two types of reactions based on differences in chemical conditions at the periphery and in the aggregate center. This pronounced stratification of organisms and reactions is different from the spatial distribution of organisms in a rotating biological contractor operated at oxygen-limited conditions analyzed by Pynaert et al.[7] where both aerobic and anaerobic ammonia oxidizing bacteria were found throughout the biofilm. The exact thickness of the nitrifying layer in the CANON reactor was governed by the O₂ concentration in the bulk liquid and by the degree of turbulence (Fig. 4). O₂ concentrations up to 11 μM, which is about double the concentration measured in the reactor under normal operational conditions, did not saturate the O₂ uptake of the nitrifying bacteria so that oxygen only penetrated to about 0.1 mm depth (Fig. 4). There thus seemed to be a large over-capacity by the aerobic ammonia oxidizing bacteria under normal operational conditions that was also verified by the O₂ perturbation experiments with the CANON reactor (Fig. 2). The microscale analysis performed at 100 μM NO⁻₂ caused very deep NO⁻₂ penetration in the aggregate (Fig. 4(c)) approaching full penetration, and the maximum possible anaerobic ammonia oxidizing activity was apparently approached at this concentration. This conclusion is also supported by the experiment with small and large aggregates in mini reactors (Fig. 3), where even the large aggregates exhibited saturation at 100 μ NO⁻₂. The NO⁻₂ uptake kinetics experiments conducted during the 2002 campaign (Fig. 3, Table 2) actually indicate that the high NO⁻₂ concentrations of >100 μM measured in the CANON reactor at 14 μM O₂ during the 2001 campaign (Fig. 2(b)) should have caused reactor failure due to an inability of the anammox biomass to reduce all the nitrite produced by aerobic ammonia oxidation. This absence of reactor failure was probably due to an average larger aggregate size and extensive wall growth of anammox biomass in the intact reactor during 2001, as both situations would result in higher apparent K_m values for NO⁻₂ than the ones shown in Table 2.

Concentration profiles such as those shown in Fig. 4 may seem more defined than they actually are and the aggregate surface is less well defined. Microscopy of the aggregates showed a somewhat fluffy surface layer where a visual definition of “surface” within ±50 μm was difficult. It is consequently not possible to tell whether the active zone is 50 or 100 μm thick. What can be seen from the profiles is, however, that the diffusive boundary layer is extremely important for the mass transport of chemical species to the aggregate, and that the chemical conditions in the active layers of the aggregate are very different from the bulk conditions. We, unfortunately, do not have comparable O² and NO⁻₂ microsensor data from the normal operational conditions of 5–6 μM O₂ and 45 μM NO⁻₂, but the shown data for 9 μM O₂ and 30 μM NO⁻₂ (Fig. 4(b)) are sufficiently close to illustrate the point. At these bulk concentrations, O₂ within the nitrifying zone varied between 0% and 50% of the bulk concentration, and NO⁻₂ in the zone with anammox activity varied between 0% and 25% of the bulk concentration. Microprofiles as shown in Fig. 4(b) are measured in an immobilized aggregate exposed to a high flow rate of about 1 cm s⁻¹, and the free-floating aggregates in the reactor are presumably exposed to less shear as exemplified with the flow at about 0.2 cm s⁻¹ (Fig. 4(d)). At low flow almost all the diffusion limitation of O₂ supply occurred in the diffusive boundary layer, and even at the applied bulk O₂ concentration of 12 μM, the O₂ concentration experienced by the bacteria in the outermost layers of the aggregate was only about 2–3 μM. It should be realized that the microgradients within aggregates in a reactor are dynamic due to random turbulences, collisions with other aggregates, etc. However, the extremes in flow rate applied in Fig. 4(b) and (d) probably give a realistic representation of the range in chemical gradients experienced within the reactor. There were no O₂ gradients in the bulk liquid as shown by the constant signal from the only 50 μm thick O₂ microsensor (Fig. 2).

The low O² and NO⁻₂ concentrations experienced by the bacteria within the aggregates favor bacteria with low K_m values. Even for the diffusion-limited intact aggregates, the experiments with mini reactors resulted in apparent K_m values of 12 μM for O₂, and 14 μM for NO⁻₂ for the small aggregates (Table 2). For non-diffusion-limited cells, the real K_m values must then be far lower. In another study with CANON reactor biomass (O. Sliekers et al., unpublished results), a K_m value as low as 2.3 μM O₂ was found, which is in line with our data. Other authors have also found relatively low K_m values for N. europaea[30,31]. The apparent K_m value of 14 μM NO⁻₂ for the small aggregates, which were less diffusion-limited than the large aggregates (apparent K_m of 23 μM), indicate a real Km < 10 μM NO⁻₂ for anammox, which is comparable with the K_m of <7 μM reported by Strous et al.[32].

4.2 Distribution of nitrification and anammox among aggregates

Even at the highest oxygen concentration there was no visible peak of nitrite within the nitrifying layers of the analyzed large aggregates (Fig. 4). The concentration gradients thus indicated a NO⁻₂ flux to the aggregate from the bulk liquid, and the rate of aerobic ammonia oxidation within these aggregates was consequently lower than the rate of anaerobic ammonia oxidation. Fluxes of O² and NO⁻₂ into the aggregate as calculated from the data (Fig. 4, Table 3) show an almost 10-fold lower O₂ as compared to NO⁻₂ uptake at 2 μM O₂. Even at an elevated O₂ concentration of 9 μM, the NO⁻₂ uptake was almost double that of the O₂ uptake. The theoretical ratio is an uptake of 1.5 O₂ for each NO⁻₂ produced and the imbalance must be due to our selection of large aggregates only for microsensor studies. The biological reactions in the small aggregates were, however, investigated by the mini-reactor studies (Table 2), where reactors were operated with biomass <500 and >500 μm, respectively. This experiment showed that the small aggregates, that only constituted 40% of the biomass, were responsible for roughly two thirds of the overall aerobic ammonia oxidation capacity (V_max), while the opposite was the case for anaerobic ammonia oxidation. A skewed distribution of processes with respect to aggregate size was expected because of the extreme O₂ sensitivity of the anammox bacteria[3]. Even the low bulk O₂ concentration (about 5 μM) during normal reactor operation prevents anammox activity in newly formed small aggregates, which become purely aerobic ammonia oxidizing compartments and thus only sources for nitrite. At a certain size anoxic zones develop and the aggregate will support anaerobic ammonia oxidation and growth of anammox bacteria. As the aggregate gradually increases in size the lower surface to volume ratio yields a progressively higher anoxic fraction and the aggregate will eventually transform to a net sink for nitrite. However, depending on prevailing conditions the inner parts of the aggregate will eventually become nitrite-limited as evidenced by the microscale analysis (Fig. 4), and this diffusion limitation probably prevents the formation of very large aggregates. The distinct organization in the aggregates of these immotile and slow growing bacteria with doubling times of >7 h and >10 days for aerobic and anaerobic ammonia oxidizing bacteria, respectively [14,33], also suggests that the aggregates must be stable over very long time.

Partial differentiation of organisms and processes among different sized aggregates may be of relevance in the context of full-scale reactor control and stability. Adjusting the efficiency of the sludge retention can be used to control the relative ratios of aerobic and anaerobic ammonium oxidizing bacteria within the reactor. Low biomass retention efficiency will select for larger aggregates and promote anammox activity whereas a high nitrification potential can be obtained by efficient settling so that only very small aggregates leave with the reactor effluent. By always keeping the anammox potential higher than the nitrification potential the occurrence of fatal nitrite build-up and following irreversible nitrite inhibition[32] of the anammox biomass can be reduced.

4.3 Feed-back control of CANON reactors

This study has confirmed that significantly elevated process rates in the sequencing batch CANON reactor can be obtained by elevation of the oxygen supply. Very high rates of up to 1.5 kg N/m3/day have thus been obtained in a CANON gas lift reactor[19], where the mass transfer between gas and liquid is more efficient. Recent studies[20] showed that the CANON biomass is very resilient against disturbances in wastewater composition and by proper control the CANON process may be a good alternative to existing nitrification–denitrification systems for treatment of liquid waste rich in ammonia. However, it may be critical to have on-line information about the chemical conditions within the reactor. A recent simulation study of a CANON type biofilm system performed by Hao et al.[34] showed that oxygen regulation according to variations in ammonia load was essential for optimal reactor performance. The continuous threat of reactor failure due to oxygen overloading and subsequent nitrite poisoning of the anammox biomass probably necessitate information about NO⁻₂ status. It may also be relevant to have on-line information about nitrate (e.g.,[21]) as a warning system about increasing NO⁻₃ concentration due to growth of NO⁻₂ oxidizing bacteria. This study also showed that size distribution and amount of aggregates greatly affects steady state concentrations of O² and NO⁻₂, and monitoring of both parameters may be used to manage sludge retention as discussed above.

Acknowledgements

We thank Preben Soerensen for technical assistance. This study was performed as part of the Icon project under the Fifth Framework Programme of the European Commission (EVK1-CT-2000-00054).

References

Author notes