Influence of dissolved oxygen on the nitrification kinetics in a circulating bed biofilm reactor

Influence of dissolved oxygen on the nitrification kinetics in a circulating bed biofilm reactor R. Nogueira, V. Lazarova, J. Manem, L.F. Melo Abstract The in¯uence of dissolved oxygen concentration on the nitri®cation kinetics was studied in the circulating bed reactor (CBR). The study was partly performed at laboratory scale with synthetic water, and partly at pilot scale with secondary ef¯uent as feed water. The nitri®- cation kinetics of the laboratory CBR as a function of the oxygen concentration can be described according to the half order and zero order rate equations of the diffusionreaction model applied to porous catalysts. When oxygen was the rate limiting substrate, the nitri®cation rate was close to a half order function of the oxygen concentration. The average oxygen diffusion coef®cient estimated by ®tting the diffusion-reaction model to the experimental results was around 66% of the respective value in water. The experimental results showed that either the ammonia or the oxygen concentration could be limiting for the nitri®cation kinetics. The latter occurred for an oxygen to ammonia concentration ratio below 1.5±2 gO2/gN-NH 4 for both laboratory and pilot scale reactors. The volumetric oxygen mass transfer coef®cient (k L a) determined in the laboratory scale reactor was 0.017 s )1 for a super- ®cial air velocity of 0.02 m s )1 , and the one determined in the pilot scale reactor was 0.040 s )1 for a super®cial air velocity of 0.031 m s )1 . The k L afor the pilot scale reactor did not change signi®cantly after bio®lm development, compared to the value measured without bio®lm. List of symbols am 2 m )3 speci®c mass transfer surface area of the gas bubbles CL;O2mgO2lÿ1actual dissolved oxygen concentration C L;O2mgO2lÿ1dissolved oxygen concentration at saturation conditions DfO2m 2 s )1 oxygen diffusion coef®- cient in the bio®lm DM kgDM m )3reactor dry matter concentration DO mgO2lÿ1dissolved oxygen concentration DW;O2m 2 s )1 molecular oxygen diffusion in water L f m bio®lm thickness k1=2aO2gO0:5 2m )0.5 d )1 apparent half order oxygen rate constant k0aNgN-NH 4m )2 d )1 apparent zero order ammonia rate constant k0fNgN-NH 4m )3 d )1 intrinsic ammonia rate constant k L as )1 volumetric oxygen mass transfer coef®cient raNgN-NH 4m )2 d )1 nitri®cation rate rN-NOÿ 2kgN-NOÿ 2m )3 d )1 nitrite conversion rate rN-NOÿ 3kgN-NOÿ 3m )3 d )1 nitrate conversion rate rO2kgO 2 m )3 d )1 oxygen conversion rate TP kgPT m )3reactor total protein concentration U g ms )1 super®cial air velocity mO2=NgO2/gN-NH 4stoichiometric constant for ammonia oxidation (taken as 4.57) 1 Introduction Nitrogen compounds like ammonia and nitrate can be found in many wastewaters and need to be removed in order to prevent oxygen depletion and eutrophication of surface waters. Often, ammonia biological removal from wastewater is problematic because of the low growth rate and growth yield of the bacteria involved. Due to the increasing volumes of wastewater combined with limited space availability and progressively tightening international standards, new intensive biotechnologies for wastewater treatment are being developed. Immobilisation of biomass in the form of bio®lms is an ef®cient method to retain slow growing microorganisms in continuous ¯ow reactors. Fixed biomass processes offer several advantages compared with conventional biological treatments: higher volumetric load, increased process stability and compactness of the reactors. However, thick bio®lm development poses a problem with respect to oxygen transport to the microorganisms. Generally the average oxygen penetration depth in bio®lms is around 100 to 200 lm [1]. For this reason the bio®lm thickness should be preferably in the same range and a high bio®lm surface Bioprocess Engineering 19 (1998) 441±449 ÓSpringer-Verlag 1998 441 Received: 5 January 1998 R. Nogueira, L.F. Melo University of Minho, Dept. Bioengineering, 4700 Braga, Portugal V. Lazarova, J. Manem Centre of International Research for Water and Environment (CIRSEE), Lyonnaise des Eaux, 38 rue du pre Âsident Wilson, 78230 Le Pecq, France Correspondence to: L.F. Melo Financial support of program PRAXIS XXI through Project 2/2.1/ Bio/37/94 program INTERREG and Lyonnaise des Eaux is gratefully acknowledged. area is required in order to obtain maximal volumetric conversion rates. It is generally agreed that, at least as a ®rst approximation, the bio®lm diffusion-reaction process can be modeled in a manner which is analogous to the description of a chemical reaction taking place in a porous catalyst [2±6]. According to this basic theory, bio®lm reactor performance depends on the diffusional resistance of the bio®lm to the penetration of substrates (ammonia and oxygen for nitri®cation processes). If the bio®lm is fully penetrated by both substrates, then the overall process is reaction rate limited and the nitri®cation kinetics is described by a zero order equation. Otherwise, the diffusional resistance to the penetration of either the substrates is controlling the overall reaction and the nitri®cation kinetics is a half order function of the substrate concentration which penetrates the bio®lm the least. In bio®lm systems the maximum volumetric conversion rate is usually limited by the liquid-bio®lm [7] or the gasliquid oxygen mass transfer rate [8]. This study is aimed at investigating the in¯uence of dissolved oxygen in the kinetics of ammonia removal in a circulating bed reactor at laboratory and pilot scales. 2 Experimental 2.1 Experimental set-up The circulating bed reactor (CBR), as shown in Fig. 1, is a three phase-bioreactor. It has a rectangular geometry and is divided into two sections: an up ¯ow aerated section (riser) and a down ¯ow non aerated section (downcomer). The support for the biomass (24% v/v) is an irregular plastic granular product with a diameter between 0.5 and 3 mm and an average density of 0.87 g cm )3 . A homogeneous three phase circulation (liquid-gas-solid) is induced by the injection of air in the riser. The geometry and operating conditions of the laboratory and pilot scale reactors used in the present work are summarised in Table 1. Laboratory scale reactor operation A series of preliminary batch experiments were carried out to promote microbial adhesion and initial bio®lm formation. The support medium was seeded in a batch fermenter with a nitrifying bacterial suspension using an enrichment medium [9]. The bacterial suspension was obtained from the sedimentation tank downstream of a nitrifying bio®lm reactor. Afterwards, the support medium was introduced in the circulating bed reactor which was continuously fed with a solution of 50 mgN-NH 4l )1 in a synthetic medium composed of tap water, ammonium sulfate, sodium hydrogenocarbonate and potassium dihydrogenophosphate. The ratios HCOÿ 3/N-NHÿ 4and PO3ÿ 4/N-NH 4were mainted at 8.68 and 0.22, respectively. Three trace elements solutions, 0.75 ml l )1 of solution A (20 mg l )1 ZnSO 4 .7H 2 O, 20 mg l )1 CuSO 4 .5H 2 O and 20 mg l )1 Na 2 MoO 4 .2H 2 O), 0.75 ml l )1 of solution B (2 mg l )1 MgSO 4 .7H 2 O) and 3.75 ml l )1 of solution C (1.44 g l )1 FeSO 4 .7H 2 O and 2.06 g l )1 EDTA) were supplied in order to stimulate nitrifying bacteria activity [10]. The temperature was controlled at 25  1 °C for the ®rst 10 months of operation and afterwards to 29  1 °C. The pH average value of the reactor during the ®rst 180 days was within the range of 6.5±7.6 and after that it was controlled at 7.5  0.2 with sodium hydrogenocarbonate 50 g l )1 NaHCO 3 . Pilot scale reactor operation The reactor was seeded in batch operation with activated sludge according to a speci®c start-up procedure [11]. Municipal wastewater after secondary treatment was used as reactor in¯uent in continuous operation. The reactor was operated under conditions of step-wise increase of the secondary ef¯uent ¯ow rate. Average in¯uent ammonia and chemical oxygen demand were 50 mgN-NH 4l )1 and 53 mgO 2 l )1 , respectively. The temperature was within the range of 16.7±22.5 °C and the pH had an average value between 6.9 and 7.8. 2.2 Analytical procedures Ammonia, nitrite and nitrate were evaluated photometrically, immediately after sampling and ®ltration through 0.22 lm membranes, by spectrophotometer HITACHI U100 according to Standard Methods [12], procedures 4500NH 3 C, 4500-NO 2 B and 4500-NO 3 B, respectively. The dissolved oxygen concentration was measured by an oxymeter WTW, model OXI 538. The quantity of washed out biomass in the ef¯uent was determined as total suspended solids after ®ltration by 0.22 lm ®lter, according to Standard Methods [12], procedure 2540 B. The bio®lm mass was estimated by means of total protein (TP) and by dry matter (DM) measurements. The bio®lm was removed from the support medium after 7.5 minutes ultrasound treatment in 30 s cycles. The biomass TP was measured according to the method of Lowry [11]. Biomass DM was determined by the gravimetric method. Fig. 1. The Circulating Bed Reactor (CBR) con®guration 442 Bioprocess Engineering 19 (1998) Biomass density represents the weight of biomass expressed as DM per unit volume of bio®lm. Bio®lm volume was measured using a pycnometer as follows: the difference in the mass of acetone weighted in a pycnometer containing support particles with and without bio®lm, divided by the acetone density. Bio®lm scanning electron microscopy (SEM) observations were made after adequate preparation of the bioparticles. After sampling, the bio®lm cells were ®xed with a solution of 3% (v/v) glutaraldehyde prepared in cacodylate buffer pH 7.2 for 2 h. Subsequently they were submitted to dehydration in a graded ethanol series (30, 50, 70, 90 and 100%). The samples were gold coated before they were observed in a Leica Cambridge scanning electron microscope. 3.3 Effect of dissolved oxygen concentration on the nitrification rate Laboratory scale reactor The effect of dissolved oxygen concentration (DO) on the nitri®cation rate was investigated in short term experiments (10±12 hours), with constant biomass concentration. The DO was manipulated by changing the oxygen content of the gas, while keeping the super®cial gas velocity constant at 0.02 m s )1 . This was obtained by mixing pressurized air and nitrogen or oxygen gas. The in¯uence of the DO on the nitri®cation rate was also evaluated in long term reactor operation at constant air ¯ow rate and ammonia loading rate. Pilot scale reactor The effect of DO on the nitri®cation rate was investigated in short term experiments (2±3 hours), with constant biomass concentration. The DO in the reactor was manipulated by changing the air ¯ow rate. The in¯uence of the DO on the nitri®cation rate was also evaluated in long term reactor operation at constant bulk oxygen concentration and varying ammonia loading rate. 3.4 Oxygen transfer coefficient (k L a) measurement The overall oxygen transfer coef®cient (k L a) can be calculated by means of a direct method [13]. Under steady state conditions the overall oxygen balance over the reactor yields: dCL;O2=dtkLaC  L;2ÿCL;O2  ÿrO2 where dCL;O2=dt0;1 i.e.: kLarO2=C L;O2ÿCL;O2  ;2 where C L;O2is the dissolved oxygen concentration at saturation conditions. The actual dissolved oxygen concentration CL;O2was measured with an oxygen electrode, whereas the oxygen conversion rate rO2was calculated from the measured nitrite rN-NOÿ 2and nitrate rN-NOÿ 3 conversion rates using the stoichiometric constants for ammonia (taken as 4.57 gO2/gN-NH 4) and nitrite (taken as 3.43 gO2/gN-NHÿ 2) oxidation: rO23:43rN-NOÿ 24:57rN-NOÿ 3:3 3 Results and discussion 3.1 Biofilm development Bio®lm development, observed by light and scanning electron microscopes can be described as a three phase process of (i) cell attachment, (ii) bio®lm formation on the concavities of the support medium (non homogeneous coverage of the support by the bio®lm), and (iii) bio®lm spreading from the colonized surfaces areas to the remaining surface until all the support medium is completely covered by the bio®lm (homogeneous coverage). Phases (i), (ii) and (iii) are presented in Fig. 2. This process was similar to the one described by Heijnen [14]. Bio®lm growth curves expressed by total protein (TP) and chemical oxygen demand (COD) are shown in Fig. 3, for the laboratory and pilot scale reactors. The bio®lm accumulation rate was higher during the start-up period of the laboratory scale reactor. After 100 days of reactor operation, bio®lm total protein steady-state concentration was reached and had an average value of 0.91 kgTP m )3reactor . In the pilot scale reactor, bio®lm steady state composition was not attained because the reactor operation was changed to secondary nitri®cation after day 90. It can be observed an increase in the bio®lm accumulation rate expressed by COD compared to the one expressed by total protein after 60 days of pilot reactor operation. This fact can be related to the accumulation of inert materials in the bio®lm [15]. 3.2 The effect of oxygen concentration on the nitrification rate At laboratory scale, experiments were carried out during continuous reactor operation at conditions of pH not controlled (®rst 180 days) and pH controlled, to study the Table 1. Geometry and operating conditions of the circulating bed reactors Parameters Laboratory scale reactor Pilot scale reactor Liquid volume (m 3 ) 0.0046 4.6 Width = Length (m) 0.09 0.9 Height (m) 0.6 6 Super®cial air velocity (m s )1 ) 0.02 0.024±0.031 Hydraulic residence time (h) 1.0 from 2.3 to 0.5 Loading rate (KgN-NH 4m )3 d )1 ) 1.19 0.5±2.5 443 R. Nogueira et al.: In¯uence of dissolved oxygen on nitri®cation kinetics in¯uence of dissolved oxygen on nitri®cation kinetics. Steady state results from the laboratory scale reactor, for both conditions, and from the pilot scale reactor are summarized in Table 2. Figure 4 shows the speci®c ammonia removal rate as a function of dissolved oxygen concentration on logarithm scale for the two operating conditions of the laboratory scale reactor: with and without pH control. It is evident from the ®gure that for both operating conditions, the speci®c ammonia removal rate depends on the bulk oxygen concentration, and a linear dependency is found until it reaches a constant value. The slope of both curves was calculated as 0.5. The theory put forward by Harremo Èes [2] based on the diffusion-reaction model in porous catalysts is used here. Due to the highly turbulent conditions, liquid phase mass transfer is not considered to be the rate limiting step. When oxygen is partly penetrating the bio®lm, the removal rate is governed by a half order kinetics according to Eq. (4a). The oxygen removal rate is expected to increase proportionally to the square root of the bulk concentration, until the maximum removal rate is achieved, the latter being given by Eq. (5): raN1=mO2=N ÿ k1=2aO2 CL;O2 p;4a where: a b c 20 µm Fig. 2a±c. Scanning electron microscopy photographs for support medium with: aadhered individuals cells, bnon homogeneous coverage and chomogeneous coverage by the bio®lm Fig. 3a,b. Nitrifying bio®lm growth: alaboratory scale reactor, b pilot scale reactor. Symbols: d, Total Protein; ¨, COD 444 Bioprocess Engineering 19 (1998) k1=2aO2 2DfO2mO2=Nk0fN q:4b raNk0aNk0fNLf;5 where raNnitri®cation rate (gN-NH 4mÿ2dÿ1); mO2=N stoichiometric constant for ammonia oxidation (taken as 4.57 mgO2/mgN-NH 4); CL;O2bulk oxygen concentration (mgO21ÿ1); k1=2aO2apparent half order oxygen rate constant (gO0:5 2mÿ0:5dÿ1); k0aNapparent zero order ammonia rate constant (gN-NH 4mÿ2dÿ1); k0Nintrinsic rate constant (gN-NH 4mÿ3dÿ1); DfO2oxygen diffusion coef®cient in the bio®lm (m 2 s )1 ); Lfbio®lm thickness (m). Values of the bio®lm apparent constants obtained by ®tting Eqs. (4a) and (5) to the experimental results are summarized in Table 3. As can be seen in Table 3, two different half order oxygen rate constants were determined. The half order oxygen rate constant for the case where the pH was controlled is almost 100% higher than for the case where the pH was not controlled. This discrepancy can be related with the increase of the ammonia removal rate during continuous reactor operation due to pH control [16]. Other authors using the half order reaction concept reported apparent constants for several types of bio®lm reactors. These values are summarized in Table 4. The apparent zero order and half order rate constants determined in this study are within the range of values reported in literature (Table 4). Some authors presented different values of the half order rate constant, depending on the dissolved oxygen concentration in the reactor [6] and on the fraction of autotrophs in the bio®lm [3]. The half order rate constant determined in this study is in the upper limit of the range of values reported in the literature. The reason for this could be related to the hydrodynamic characteristics of the CBR, that ensures a good mass transfer and effective control of bio®lm thickness [17], enhancing the oxygen transport into the bio®lm. Hem et al. [5] studied nitri®cation in a moving bed bio- ®lm reactor (MBBR) as a function of the oxygen concentration and the kinetics obtained was close to a ®rst order reaction, indicating a possible signi®cant in¯uence of the liquid ®lm resistance on the bio®lm kinetics. Rusten et al. [18] suggested that the reason why liquid ®lm diffusion was important in the MBBR was related with the carrier used. They demonstrated that only the protected surfaces inside the carrier were available for bio®lm growth. The oxygen diffusion coef®cient determined in the present work using Eq. (6), 1:75 10ÿ9m2sÿ1, was obtained by ®tting the diffusion-reaction model, Eqs. (4b) and (5), to the rate of ammonia consumption as a function of oxygen concentration for the case where the pH was controlled: Table 2. Steady-state operating conditions of laboratory and industrial scale reactors Laboratory scale Pilot scale pH not controlled pH controlled Removal rate (kgN-NH 4m )3 d )1 0.63 1.14 1.63 N-NH 4removal ef®ciency 64% >95% >95% Biomass concentration (kgPT mÿ3 reactor) 0.72 0.91 0.89 (kgDM mÿ3 reactor) 1.47 2.85 1.57 Biomass residence time (d) 10.8 20.1 ± Speci®c activity (kgN-NH 4kgPT )1 d )1 ) 0.93 1.29 1.90 (kgN-NH 4kgDM )1 d )1 ) 0.43 0.40 0.60 Fig. 4. Speci®c ammonia removal rate as a function of dissolved oxygen concentration in the laboratory scale reactor for the two operating conditions. pH not controlled and pH controlled Table 3. Bio®lm apparent constants for the laboratory scale reactor Kinetic constants Laboratory scale pH not controlled pH controlled Zero order constant K0aN (kgN-NH 4kgPT )1 d )1 ) 0.83 1.5 (gN-NH 4m )2 d )1 ) 0.85 2.1 Half order constant k1=2aO2 gO0:05 2mÿ0:5dÿ12.2 4.0 Oxygen transition value (mgO 2 l )1 ) 3.4 5.2 Oxygen/Ammonia ratio at transition (gO 2 g )1 N-NH 4) ± 1.5±2 445 R. Nogueira et al.: In¯uence of dissolved oxygen on nitri®cation kinetics DfO2k2 1=2aO2=2k0aO2  Lf:6 The average bio®lm thickness used in the determination of the oxygen diffusion coef®cient, 179 lm, was calculated from the quotient between the bio®lm dry matter per unit surface area of the support and the bio®lm dry density. The bio®lm dry density was experimentally measured as 28.3 kgDM m )3wet bio®lm and the bio®lm dry matter per unit surface area of the support was obtained from the quotient between the bio®lm dry matter per unit support volume, which was experimentally measured as 13.5 kgDM m )3support , and the speci®c surface area of the support (2670 m 2 m )3 ). The oxygen diffusion coef®cient calculated (1:75 10ÿ9m2sÿ1) was around 66% of the respective value in water. Several authors reported values of diffusion coef®cients of compounds in bio®lms, in microbial cell aggregates and in gels with immobilized cells [19±21]. However, literature on the determination of the oxygen diffusion coef®cient in nitrifying bio®lms is scarce. Directly measured values of the oxygen diffusion coef®cient reported in literature for mixed nitrifying bio®lms are presented in Table 5, together with the values obtained in this study. The oxygen diffusion coef®cient determined in the present work (Table 4) is lower than the values published by other authors, probably due to the different methods that have been used. In the diffusion cell method a microbial layer was made by ®ltering a dispersed nitrifying culture onto the support ®lter, and oxygen was allowed to diffuse through the prepared biomass layer from a chamber with a high oxygen concentration to another one which was previously deoxygenated [22, 23]. However, inherent inaccuracies in the experimental method can be mentioned. Filtered biomass would not have the same structure as a true bio®lm and the mass transfer effect of the rough surface of the grown bio®lm may be lost. The method used in the present work requires the determination of the speci®c surface area of the bio®lm particles. Since the bio®lm was developed in an irregular granular support, the experimental measurement of this parameter was dif®cult and a simpli®cation was made by considering that the bio®lm particle had a spherical shape. In order to compare the quotient of effective oxygen diffusivity in the bio®lm to molecular oxygen diffusivity in water obtained in this work with those reported in literature (Table 5), the same correlation was used to determine the molecular oxygen diffusivity in water (Wilke-Change correlation, [24]). The quotient value obtained by Williamson and McCarty [22], indicating that the oxygen diffusivity in the bio®lm is higher than the molecular oxygen diffusivity in water, seems to indicate that another mechanism was involved in the transport of oxygen in the ``bio®lm''. However, no comment was made by the authors about this value. Nitrate accumulation was not detected in the ef¯uent from the laboratory scale reactor for both operating conditions: pH not controlled and pH controlled. For the phase where the pH was not controlled, the ef¯uent nitrite was 0.47 mgN-NOÿ 2l )1 and ammonia removal ef®ciency was 64%. An ef¯uent nitrite of 0.81 mgN-NOÿ 2l )1 and ammonia removal ef®ciency higher than 95% was observed for the phase where the pH was controlled. Dissolved oxygen in the reactor was always higher than 3.5 mgO 2 l )1 . This implies that oxygen did not limit the nitri®cation rate. Table 4. Nitrifying bio®lm kinetic constants Bio®lm Reactor k0aN (gN-NH 4m )2 d )1 )k1=2aO2 (gO0:5 2m )0.5 d )1 ) References Rotating Disc System [3] (1±30 mgO 2 l )1 ) 0.8±4 2.08±3.76 Floating Filter [4] (1±7 mgO 2 l )1 ) 1.7 2.8±2.9 Moving Bed Bio®lm Reactor [5] (1±15 mgO 2 l )1 ) 0.7±1 ± Submerged Aerated Filter [6] (4±5 mgO 2 l )1 ) 1.82 2.7 (2±3 mgO 2 l )1 ) 0.47 1.4 Circulating Bed Reactor 2.1 4.0 This study (1±7.2 mgO 2 l )1 ) Table 5. Measured oxygen diffusion coef®cient within bio®lms Method DfO210ÿ9DfO2=DW;O2Biomass type References Diffusion cell 2.5 1.21 Filtered [22] (20 °C, pH = 7) microbial layer Diffusion cell 2.05 0.97 Filtered [23] (20 °C, pH = 7.8) microbial layer Diffusion reaction 1.75 0.66 Nitrifying This study model (29 °C, pH = 7.5) bio®lm The diffusivity of oxygen in water Dw;O2was obtained from the Wilke-Change correlation [24] 446 Bioprocess Engineering 19 (1998) 3.3 The phenomenon of either the ammonia or the oxygen concentration being limiting for the nitrification rate According to the experimental results obtained in the laboratory scale reactor, the dissolved oxygen concentration for which a change in the order of reaction was observed, i.e. half order reaction to zero order reaction, depended on the pH control (Fig. 4). The transition values were 3.4 and 5.2 mgO 2 l )1 (Table 3), for the cases of pH not controlled and pH controlled, respectively. Figure 5 presents the ammonia removal ef®ciency as a function of bulk oxygen to ammonia ratio for the laboratory and pilot scale reactors. Results were obtained from two sets of experiments: long term (several months) and short term (some hours) experiments. For both reactors, the transition from ammonia limitation to oxygen limitation was found to occur when the bulk oxygen to ammonia ratio was about 1.5±2 gO2/gN-NH 4. It seems that reactor scale up did not affect the transition value from ammonia limitation to oxygen limitation. Values reported in the literature for the transition from ammonia rate limited to oxygen limited nitri®cation are within the range 2.5±4 gO2/gN-NH 4[3, 5, 6, 22]. The value obtained in this work for the laboratory and pilot scale reactors is lower than the values obtained by other authors. Again it can be suggested that due to the hydrodynamic characteristics of the CBR and the resulting thin bio®lms, the resistance to oxygen mass transfer within the bio®lm is lower than the observed in other bio®lm reactors. 3.4 Oxygen mass transfer The volumetric oxygen mass transfer coef®cient, kLa,in the laboratory scale reactor presented a constant value of 0.017 sÿ1for both phases of reactor operation, with and without pH control, corresponding to different ef®ciencies of ammonia nitrogen removal in the reactor, 64% and >95%, respectively. This can be explained according to Eq. (2) by the increase on the ammonia nitrogen removal rate due to pH control that resulted in a decrease on the dissolved oxygen concentration in the reactor from 7.1 mgO 2 l )1 to 4.7 mgO 2 l )1 . Fig. 6 provides kLaas a function of Fig. 5. Ammonia removal as a function of bulk oxygen to ammonia ef¯uent concentration ratio in the laboratory and pilot scale reactors. Symbols: à, lab scale (short term); ¨, lab scale (long term); j, pilot scale (short term); h, pilot scale (long term) Fig. 6. k L aas a function of super®cial air velocity (U g )inthe laboratory scale reactor Fig. 7.k L aand temperature as a function of oxygen conversion rate (r O2 ) in the pilot scale reactor. Symbols: jk L a, ±temperature 447 R. Nogueira et al.: In¯uence of dissolved oxygen on nitri®cation kinetics super®cial air velocity (Ug). It was found that kLais a power function of Ugin the range of 0.004±0.023 m s )1 . Figure 7 shows the kLaand the dissolved oxygen concentration as a function of the oxygen conversion rate (rO2) in the pilot scale reactor. For rO2values in the range of 0.12 to 4.8 kgO2mÿ3dÿ1,kLaincreased with rO2, because the dissolved oxygen concentration in the reactor was kept constant. For rO2values higher than 4.8 kgO2mÿ3dÿ1,kLapresented a constant value (0.040 s )1 ) because the temperature dropped and therefore, the dissolved oxygen concentration at saturation increased. In another work performed on this reactor operating in tertiary nitri®cation, kLawas measured by means of the gas-off method [11]. The average value obtained 0.043 s )1 , agreed with the one obtained in the present work, 0.040 s )1 . Also, a kLavalue was calculated by means of an equation relating the kLawith the super®cial air velocity in the absence of bio®lm on the support particles, proposed by Lazarova et al. [11]: kLahÿ11:7106 Ugms ÿ12:66. For a super®cial air velocity of 0.031 m s )1 the kLacalculated was 0.046 m s )1 . This value is similar to the ones obtained by the two experimental methods and it can be suggested that the mass transfer coef®cient did not change signi®cantly after bio®lm development compared to the situation without bio®lm. The growth of bio®lm on the support particles did not change its volume fraction in the reactor. Furthermore, the previous equation adequately describes oxygen mass transfer in the laboratory scale reactor. The average kLavalue obtained in this work for pilot scale reactor, 0.040 s )1 , is higher than the kLavalue obtained in the laboratory scale reactor, 0.017 s )1 . This is due to the different super®cial air velocities at which the two reactors were operated, 0.031 m s )1 and 0.02 m s )1 for the pilot and laboratory scale reactors, respectively. 4 Conclusion The following conclusions can be drawn from the present study: 1. The bio®lm development can be described as a three phase process of cell attachment, bio®lm formation on the concavities of the support particles (non homogeneous coverage of the support by the bio®lm), and bio®lm spreading from the colonized surface areas to the remaining surface area until all the support medium is completely covered by the bio®lm (homogeneous coverage). 2. The nitri®cation kinetics of the laboratory CBR as a function of the oxygen concentration follows the half and zero order rate equations of the diffusion reaction model. When oxygen was the rate limiting substrate, the nitri®cation rate was close to a half order function of the oxygen concentration. The oxygen diffusion coef®cient estimated by ®tting the diffusion-reaction model to the experimental results was around 66% of the respective value in water. 3. The transition from ammonia limitation to oxygen limitation was found to occur when the bulk oxygen/ ammonia ratio was within the range 1.5±2 gO2/gN-NH 4 for the laboratory and pilot scale reactors. 4. The volumetric oxygen mass transfer coef®cient (kLa) determined in the laboratory scale reactor was 0.017 s )1 for a super®cial air velocity of 0.02 m s )1 , and determined in the pilot scale reactor was 0.040 s )1 for a super®cial air velocity of 0.031 m s )1 . The kLafor the pilot scale reactor did not change signi®cantly after bio®lm development compared to the value measured without bio®lm. References 1. De Beer, B.; Van Den Heuvel, J.C.; Ottengraf, S.P.P.: Microelectrode measurements of the activity distribution in nitrifying bacteria aggreagates. Appl. Environ. Microbiol. 59 (1993) 573±579 2. Harremo Èes P.: Bio®lm Kinetics. In: Mitchell, R. (Ed.): Water Pollution Microbiology, vol. 2, New York: John Wiley & sons 1978 pp. 82±109 3. Go ÈnencË I.E.; Harremo Èes P.: Nitri®cation in rotating disk systems ± I, Criteria for transition from oxygen to ammonia rate limitation. Wat. Res. 19 (1985) 1119±1127 4. Toettrup, H.; Rogalla, F.; Vidal, A.; Harremo Èes, P.: The treatment trilogy of ¯oating ®lters: from pilot to prototype to plant. Wat. Sci. Tech. 29 (1994) 23±32 5. Hem, L.J.; Rusten, B.; édegaard,H.: Nitri®cation in a moving bed bio®lm reactor. Wat. Sci. Tech. 28 (1994) 1425±1433 6. CËecËen, F.; Go ÈnencË, I.E.: Nitrogen removal characteristics of nitri®cation and denitri®cation ®lters. Wat. Sci. Tech. 29 (1994) 409±416 7. Tanaka, H.; Dunn, I.J.: Kinetics of bio®lm nitri®cation. Bitotech Bioeng. 24 (1982) 669±689 8. Tijhuis, L.; Huisman, J.L.; Hekkelman, H.D.; van Loosdrecht, M.C.M.; Heijnen, J.J.: Formation of nitrifying bio®lms on small suspended particles in airlift reactors. Biotech. Bioeng. 47 (1995) 585±595 9. Schmidt, J.E.; Molina, J.A.E.; Chiang, C.: Isolation of chemoautotrophic nitri®ers from Maroccan soils. Bull. Ecol. Comm. (Stockholm) 17 (1973) 166±167 10. Sharma, B.; Ahlert, R.: Nitri®cation and nitrogen removal. Wat. Res. 11 (1977) 897±925 11. Lazarova, V.; Pierzo, V.; Fontvieille, D.; Manem, J.: Integrated approach for bio®lm characterization and biomass activity control. Wat. Sci. Tech. 29 (1994) 345±354 12. Clesceri, L.S.; Greenberg, A.E.; Trussel, R.R. Washington: Standard Methods for the Examination of Water and Wastewater, 17th Edition. In: APHA, AWWA, WPCF 1989 13. Kargi, F.; Moo-Young, M.: Transport phenomena in bioprocesses. In: Moo-Young, M.; Blanch, H.W.; Drew. S; Wang, D. I. C. (Eds.): Comprehensive biotechnology, vol. 3, pp. 5±56. Oxford: Pergamon Press 1985 14. Heijnen, J.J.; van Loosdrecht, M.C.M.; Mulder, A.; Tijhuis, L.: Formation of bio®lms in a bio®lm air-lift suspension reactor. Wat. Sci. Tech. 26 (1992) 647±654 15. Liu, Y.; Capdeville, B.: Dynamics of nitrifying bio®lm growth in biological nitrogen removal process. Wat. Sci. Tech. 29 (1994) 377±380 16. van Ginkel, C.G.; Tramper, J.; Luyben, K.Ch. A.M.; Klapwijk A.: Characterization of Nitrosomonas europaea immobilized in calcium alginate. Enzyme Microb. Technol. 5 (1983) 297±303 17. Lazarova, V.; Manem, J.: An innovative process for waste water treatment: the circulating ¯oating bed reactor. Wat. Sci. Tech. 34 (1997) 89±99 18. Rusten, B.; Siljudalen, J.G.; Nordeidet, B.: Upgrading to nitrogen removal with the KMT moving bed bio®lm process. Wat. Sci. Tech. 29 (1994) 185±195 19. Karel, S.F.; Libicki, S.B.; Robertson, C.R.: The immobilization of whole cells: Engineering principles. Chemical Engineering Science 40 (1985) 1321±1354 448 Bioprocess Engineering 19 (1998) 20 Libicki, S.B.; Salmon, P.M.; Robertson, C.R.: The effective diffusive permeability of a nonreacting solute in microbial cell aggregates. Biotech. Bioeng. 32 (1988) 68±85 21. Fan, L.-S.; Leyva-Ramos, R.; Wisecarver, K.D.; Zehner, B.J.: Diffusion of phenol in a bio®lm grown on activated carbon particles in a draft-tube three phase ¯uidized-bed bioreactor. Biotechnol.Bioeng. 35 (1990) 279±286 22. William, K.; McCarty, P.L.: Veri®cation studies of the bio®lm model for bacterial substrate utilization. Journal WPCF 48 (1976) 281±296 23. Onuma, M.; Omura T.: Mass transfer characteristics within microbiol systems. Wat. Sci. Tech. 14 (1982) 553±569 24. Geankoplis, C.J.: Transport processes and unit operations, 3rd Edition. New Jersey: Prentice-Hall, Inc 1993 449 R. Nogueira et al.: In¯uence of dissolved oxygen on nitri®cation kinetics