PII: S0043-1354(00)00365-1

Wat. Res. Vol. 35, No. 5, pp. 11691178, 2001 # 2001 Elsevier Science Ltd. All rights reserved Printed in Great Britain 0043-1354/01/$ - see front matter

NITRIFICATION OF HIGH STRENGTH AMMONIA WASTEWATERS: COMPARATIVE STUDY OF IMMOBILISATION MEDIA

WENDY M. ROSTRON1, DAVID C. STUCKEY2* and ANDREW A. YOUNG3
2 1 Anglian Water, Broadholme STW, Ditchford Lane, Welllingborough, Northants. NN8 1RR, UK; Department of Chemical Engineering, Imperial College, London, SW7 2AZ, UK and 3 Ozone Industries Ltd., Unit B15, Armstrong Mall, Southwood Summit Centre, Farnborough, Hants. GU14 0NR, UK

(First received 4 November 1999; accepted in revised form 17 July 2000) Abstract}Due to legislative pressures, sludge production and processing in the UK will increase substantially in the future resulting in a supernatant liquid high in ammonia (5001000 mg 1 1) and hard COD ( $ 500 mg 1 1). A small footprint reactor is required to eectively nitrify this euent, and the aim of this work was to compare a number of immobilisation media under a variety of conditions in order to determine which media held the most promise for future development. Laboratory-scale continuously stirred tank reactors containing freely suspended and immobilised biomass were operated with a high-strength synthetic ammonia wastewater (500 mg N l 1) to determine the nitrication rates at various temperatures, and ammonia and COD loadings. COD : NH3 ratios in sludge liquors vary widely depending on the treatment processes employed, and therefore ratios of 1 : 1 and 2 : 1 were tested as being fairly typical. The freely suspended nitriers were washed out of the reactors at a 1 d hydraulic retention time (HRT), whereas the reactors containing adsorption particles (Linpor and Kaldnes) and PVAencapsulated nitriers continued partially nitrifying down to 12 h, and oxygen addition enhanced nitrication. A decrease in temperature from 25 to 168C only caused a small (10%) decrease in nitrication in the immobilised cell reactors, demonstrating that nitrication was mass transfer rather than kinetically controlled. A reduction in nitrication occurred when glucose (500 mg l 1) was added to the feed due to the growth of a heterotrophic population. The adsorbed biomass reactors lost 35% of nitrication compared to only 7% with PVA, and it appears that the colonisation of PVA by heterotrophs is more dicult than for Linpor and Kaldnes. Respiration rates for all particles increased with time in the reactors, and nitriers immobilised in PVA retained approximately 40% of their viability after immobilisation. Volumetric nitrication rates were generally higher for the PVA reactor than for Linpor and Kaldnes, and were: suspended biomass reactor: 0.36; Linpor: 0.57; Kaldnes: 0.53 and PVA: 0.70 kg N m 3-reactor d 1 for a 25% reactor ll. These equate to 2.28, 4.24 and 3.97 g N m 2-media d 1 for Linpor, Kaldnes and PVA respectively, hence other reactor ll rates for Kaldnes warrant further investigation. However, the PVA particles with the highest nitrication rates under all conditions showed promise as an immobilisation medium, and are amenable to further optimisation for the nitrication of high-strength ammonia wastewaters. # 2001 Elsevier Science Ltd. All rights reserved Key words}nitrication, immobilisation, ammonia removal, supernatant treatment, polyvinyl alcohol

INTRODUCTION

In the UK an increase in the production of sewage sludge is expected due to the Urban Wastewater Treatment Directive (UWWTD) (European Community, 1991). Sludge will need to be processed in order to reduce its volume and to produce a product which provides less of a disposal problem. Processing can include thickening, anaerobic digestion, dewatering, drying and incineration. All of these steps produce a supernatant liquor which is usually

*Author to whom all correspondence should be addressed. Tel.: +44-0207-594-5591; fax: +44-0207-594-5629; e-mail: d.stuckey@ic.ac.uk

returned to the head of the sewage treatment works (STW) for treatment. These liquors often contain high levels of ammonia and oxidisable organic compounds, and can have a signicant eect on the STW (Grulois et al., 1993). Ammonia can be removed from return liquors by nitrication involving bacteria (nitriers) which use ammonia as an electron donor to provide energy for growth. The ammonia is oxidised to nitrate which may then be biologically denitried, i.e. reduced to molecular nitrogen. Biological treatment is relatively cheap and produces no unwanted side-products. Nevertheless, nitriers are slow growing and have a low yield, and hence without long retention times they will be washed out of a continuous reactor

1169

1170

Wendy M. Rostron et al.

unless they are kept there by immobilisation. Immobilising the micro-organisms either on a support media, or by encapsulating the organisms within a matrix, or combining both methods, would ensure that the nitriers are retained within the treatment system. A high cell concentration is possible with immobilisation, and therefore the volumetric eciency is greatly increased. This can lead to relatively small reactors, and may aord protection from toxic shocks and adverse temperatures which would help maintain year round treatment. Adsorption has generally been the preferred method of immobilisation in wastewater treatment. Polyurethane foam (Linpor, Captor) has been widely used (Morper, 1994; Golla et al., 1994); however, some particles can become anaerobic in the centre and settle to the reactor oor (Lessel, 1993). For this reason the Captor process (Golla et al., 1994) incorporated a pad cleaning system to periodically remove some of the biomass, but this added complexity to the system. Literature results for Kaldnes (Rusten et al., 1995) claim that high loading rates can be achieved. However, this media has been used mainly with domestic sewage, and therefore its performance in the nitrication of high ammonia feeds is unknown at present. The encapsulation of biomass in natural polymers retains high cell viability, but they are not strong enough for use in wastewater treatment (van Ginkel et al., 1983). Considerable research has been conducted using alginate and carrageenan despite their drawbacks because they are easy to work with. Such research has focused mainly on pure bacterial cultures and synthetic media, although the validity of these results for mixed cultures and real wastewaters has yet to be determined. Many techniques using stronger synthetic polymers involve chemical polymerisation initiators which cause loss of cell viability; this factor also gives rise to safety concerns, and can complicate the immobilisation procedure. Encapsulation in polyvinyl alcohol (PVA) has been successfully demonstrated, and the freezing method is a simple technique which does not involve chemical initiation (Asano et al., 1992). The advantages of adsorption over encapsulation noted by McLoughlin (1994) include the fact that no chemical additions were required, and that the particles are often reusable. PVA encapsulation by the freezing method meets both these requirements, and appears from the literature to have considerable potential. It has been reported that over time encapsulated biomass becomes concentrated in the outer regions of the particles (Tanaka et al., 1994, Wijels et al., 1993). With k-carrageenan the biomass colonies eventually break through the particle surface and erupt, emptying their contents into the medium (Husken et al., 1996). Adhesion of other bacteria to the particles has also been mentioned (Asano et al., 1992), but it is not clear whether these organisms would out-compete the immobilised ones. Immobilis-

ing bacteria may provide some protection from adverse conditions (Wijels et al., 1995); this is important because nitrication has been shown to be sensitive to temperature, substrate inhibition, dissolved oxygen levels, and also chemical oxygen demand (COD) (Barnes and Bliss, 1983). This work focused on the biological treatment of high-strength ammonia wastewaters, and presents the results from the nitrication of a synthetic ammonia wastewater using both freely suspended and immobilised bacteria. The aim of the work was to determine if immobilisation conferred an advantage on reactor operation, and which of the immobilisation media tested performed best under the conditions of ammonia loading, temperature and COD loading applied.

MATERIALS AND METHODS

Biological nitrication was studied in ve continuously stirred tank reactors (CSTRs), with 4 l operating volumes. Immobilisation media The immobilisation media chosen were; Linpor (polyurethane foam cut into 0.6 0.6 0.6 cm cubes) and Kaldnes (polyethylene pasta shapes of 1 cm diameter), both commercial adsorption particles, and encapsulation in PVA (0.8 0.8 0.8 cm cubes). One bulk litre of each immobilisation material was added to the reactors (25% reactor ll). Preparation of PVA particles A nitrifying biomass (2.5 g VSS 1) was added to an aqueous solution of PVA (20% w/v) to give a total volume of 1 litre. The mixture was frozen for 39 h at 208C, and after thawing for 24 h at room temperature was cut into 0.8 cm cubes, and washed in water before use. The PVA ([CH2CH(OH)]n) (molecular weight of 7200), was bought from BDH. CSTR description and operation The reactors were made of PVC, and fed by variable speed peristaltic pumps (Watson Marlow) using marprene tubing (Fig. 1). Aeration at the base of the reactors kept the contents well-mixed and aerated (DO>4 mg l 1). A mesh (1 mm 1 mm) at the reactor outlet was used to retain the immobilisation particles. The reactors were maintained at 258C for 135 d using a water bath, and then at 168C for the remaining 50 d. The hydraulic retention time (HRT) was varied from 8 d to 12 h by adjusting the feed owrate. Initially, the reactors were inoculated with nitrifying activated sludge and fed with a synthetic ammonia feed at an 8 d HRT. One litre of feed with an ammonia-nitrogen concentration of 500 mg l 1 contained: (NH4)2SO4, 2357 mg; NaHCO3, 6000 mg; K2HPO4, 250 mg; MgSO4 7H2O, 300 mg; CaCl2 2H2O, 120 mg; and 0.5 ml of a 1 L nutrient stock solution containing: Na2MoO4 2H2O, 274 mg; MnCl2 4H2O, 1268 mg; ZnSO4 7H2O, 2742 mg; CuSO4 5H2O, 36 mg; CoCl2 6H2O, 30 mg; FeSO4 7H2O, 7000 mg (Wiesmann, 1994; Hanaki et al., 1990; Staley, 1989). COD was added in the form of glucose at concentrations of 500 and 1000 mg l 1 (equivalent to 533 and 1067 mg l 1 as COD). At the start the ve reactors were operated as identical freely suspended reactors for 2 months. One bulk litre of immobilisation media was added to three of the reactors

Nitrication with immobilized biomass

RESULTS AND DISCUSSION

1171

Nitrication in the CSTRs Figures 25 present the results, and the stages of the tests correspond to the operating conditions given in Table 2. The results for only one of the suspended biomass reactors are shown in Fig. 2 as both reactors had similar results. Full nitrication was maintained during stages I, II and III. In stage IV a build-up of nitrite indicated washout of Nitrobacter and/or that there was not enough biomass to cope with the increased loading. The lower cell yield (Wiesmann, 1994) of Nitrobacter (0.042 mg cells mg N 1) compared to Nitrosomonas (0.142 mg cells mg N 1) indicates that a lack of Nitrobacter biomass was probably the reason for the build-up of nitrite, particularly as the maximum specic growth rate of Nitrobacter (1.08 d 1) is higher than that for Nitrosomonas (0.77 d 1), and therefore washout of Nitrobacter prior to Nitrosomonas was unlikely. When the HRT was further decreased to 1 d, the concentration of ammonia steadily increased, indicating washout of Nitrosomonas; the low maximum specic growth rates of Nitrosomonas and Nitrobacter mean that washout was expected at an HRT of about 1 d (doubling times of 0.9 and 0.64 d, respectively). The decrease in MLVSS around this time supports this conclusion (Table 3). During these trials the sides of the reactors were regularly scraped to remove biolm growth. However, it was impossible to remove all the biolm on the reactor walls and euent tubing, and this could explain why complete loss of nitrication and VSS was not seen despite the reactor being operated at dilution rates above washout. A lack of sucient biomass to nitrify

Fig. 1. Diagram of the continuously stirred tank reactor (CSTR). (day 0), while the other two reactors remained as freely suspended controls. The volume and number of particles in the one bulk litre of media added are reported in Table 1, while a summary of the reactor operating conditions over time is given in Table 2.

Monitoring and analysis Reactor pH, temperature and ow rate were monitored daily, while the euent was analysed for ammonia, nitrate, nitrite and COD using Dr. Lange test kits at least twice a week. The suspended solids (SS) and volatile suspended solids (VSS) of the mixed liquor were measured weekly using standard methods (APHA et al., 1992). The respiration rates of the immobilisation particles were measured periodically in a 0.1 M phosphate buer (pH 7.5) using a Rank oxygen cell with an ammonia-nitrogen concentration of 50 mg l 1. The particles were also regularly examined with a scanning electron microscope.

Table 1. Details of immobilisation media Media Linpor Kaldnes PVA

a

Volumea (ml) 340 220 900

% of Reactor volume 9 6 23

Number of particles 2128 1106 2012

Specic surface area (m2 m 3) 1000 500 706

Actual volume of mixed liquor displaced when 1 bulk litre of media was added to reactor.

Table 2. CSTR operating conditions Stage 0 I II III IV V VI VII VIII IX

a b

Days 011 1133 3359 5974 7483 83108 108135 (128129)b 135151 151178 178185

HRT (d) 8.0 6.0 3.4 2.2 1.5 1.0 0.5 0.5 0.5 0.5

NH3-N loadinga (gN m 3 d 1) 63 83 147 227 333 500 1000 1000 1000 1000

Temperature (8C) 25 25 25 25 25 25 25 16 16 16

COD loadinga (g m 3 d 1) } } } } } } } } 1067 2134

Loadings are based on reactor volume. Oxygen sparging.

1172

Wendy M. Rostron et al.

Fig. 2. Conditions in the freely suspended control over time.

Fig. 4. Conditions in the CSTR with Kaldnes media over time.

Fig. 3. Conditions in the CSTR with Linpor media over time.

Fig. 5. Conditions in the CSTR with PVA media over time.

the inuent ammonia led to an increase in ammonia concentration, and this became increasingly inhibitory to the micro-organisms involved, particularly as loss of nitrication caused an increase in the mixed liquor pH leading to a greater percentage of the ammonia being present in the inhibitory unionised form (Anthonisen et al., 1976). From these data it is evident that pH control is vital for nitrication of high-strength feeds. Inhibition of nitriers by unionised ammonia (due to an increase in pH when nitrication fails), or by unionised nitrous acid (when buering capacity is inadequate and the pH falls due to nitrication), can be minimised by controlling the pH at a suitable level. The synthetic feed used in this work contained sucient alkalinity to buer nitrication, but when full nitrication faltered the pH in the reactors rose above the ideal.

Immobilised cell reactors The immobilised cell reactors continued nitrifying at a 1 d HRT (stage V) showing a denite advantage to using immobilised bacteria when the HRT was less than 2 d. However, the adsorbed nitrier reactors experienced problems with nitrite accumulation; the increase in nitrite concentration showed that there were not enough Nitrobacter present to cope with the increased loading. In addition, the reactor pH rose slightly increasing the percentage of ammonia present as free ammonia, which is inhibitory to Nitrobacter at very low concentrations (0.11.0 mg l 1). The nitrite in the Linpor and Kaldnes reactors reached a maximum, and then started to decrease as the reactors recovered. The Nitrobacter had time to increase and also inhibition by free ammonia was reduced as the pH decreased due to acid production

Table 3. Mixed liquor VSS in the CSTRs Stage I II III IV V VI VII VIII
a

HRT (d) 6.0 3.4 2.2 1.5 1.0 0.5 0.5 0.5

Linpor (mg l 157 43 0 1 4 15 18 227

Kaldnes (mg l 1) 201 130 46 56 76 72 24 340

PVA (mg l 1) 165 107 199 88 81 77 24 261

SS-1 (mg l 1) 156 37 64 64 37 }a } }

SS-2 (mg l 1) 174 117 108 95 39 18 }a }

Reactor operation ceased.

Nitrication with immobilized biomass

1173

by Nitrosomonas. After 3 weeks at a 1 d HRT the Linpor reactor recovered complete nitrication, while the reactor containing Kaldnes had recovered 90% nitrication. Alternative explanations for nitrite build-up including the presence of toxins or lack of sucient micronutrients for Nitrobacter were discounted due to the reactors recovery and also that in contrast to the adsorbed biomass reactors, the reactor containing the PVA cubes maintained full nitrication at a 1 d HRT. This suggests that this reactor was supporting a higher biomass concentration. When the HRT was reduced to 12 h (stage VI) all the reactors suered a loss of nitrication. Some degree of recovery did occur in the adsorption reactors, but the second oxidation step to nitrate remained inhibited, and very little nitrate was produced. The PVA reactor maintained approximately 30% full nitrication to nitrate, although about 30% of the inuent ammonia was untreated. The loading rate changes during this work were quite severe, particularly the change from 500 to 1000 gN m 3 d 1 when the HRT was reduced from 1 d to 12 h, i.e. below the doubling time of the nitriers. More gradual changes in the ow rates may have led to more successful nitrication at the shorter HRTs, as the biomass would have had more time to grow in response to the increased loading, thereby keeping the substrate concentration low. The complications of the substrates being inhibitory in the unionised form, and the dependence of reactor pH on the nitrication reaction itself, mean that once full nitrication is lost, it is very dicult to recover. The long time periods required to reach steady state in the immobilisation reactors led to some operating conditions being changed before steady-state operation had been achieved due to time constraints. However, more substantive conclusions could have been drawn if steady-state operation had been achieved prior to changing, although the fact that the trials were performed in parallel reactors and that the response times of the dierent media were similar, means that the media comparison is still valid. As the ammonia loading increased, the biomass concentration, and hence the total oxygen demand, also increased. DO levels were not routinely monitored during the work, however, the air ow rates to all reactors were kept constant, and the DO was maintained above 4 mg l 1. The existence of oxygen limitations within the biolms/particles was demonstrated by an immediate improvement in nitrication in the Kaldnes and PVA reactors when the aeration supply was supplemented with oxygen sparging (dissolved oxygen concentrations rose from 5 to 10 mg l 1). This occurred whilst maintaining a constant total gas volumetric ow rate to ensure that any treatment improvement was not due to increased turbulence in the reactors, and hence a reduction in liquid lm diusion limitations. The eect of increased reactor DO in the Linpor reactor was less clear due to problems with reactor operation during

oxygen sparging. However, a slight improvement in nitrication was apparent, and therefore there may have been some degree of oxygen limitation. There are several solutions to the problem of oxygen limitation, including: increasing the dissolved oxygen concentration in the reactors; increasing turbulence in the reactors; and in the case of biomass immobilised in PVA, increasing the surface area of the particles in order to reduce diusional limitations. Although a more turbulent ow regime would have decreased the liquid lm resistance in all the reactors, it may have resulted in excessive loss of biomass from the Linpor and Kaldnes support particles. This would not have been a problem for the PVA particles, as their activity did not depend on adsorbed biomass. Liquid lm diusion limitations in Kaldnes reactors have been reported in the literature, mainly due to the growth of biomass on the inner protected surface (degaard et al., 1993). Contribution of VSS to nitrication in immobilised biomass reactors The VSS in the reactors are shown in Table 3. Freely suspended biomass was adsorbed onto the surface of the Kaldnes particles and onto and into the Linpor particles. As adsorption occurred the freely suspended bacteria (VSS) in the reactor decreased, and this is most apparent in the Linpor reactor. The VSS also quickly decreased in the Kaldnes reactor. The same decrease in VSS was not noted in the PVA reactor as these particles were produced as encapsulated biomass. The VSS values reported in Table 3 suggest that some biomass may have leaked out of the PVA particles as this reactor had more freely suspended VSS than the control reactors. Wash-out of the suspended biomass in the control reactors occurred in stages V and VI (1 d and 12 h HRTs), and hence they were stopped. This indicates that the VSS detected after this time in the immobilised cell reactors was probably due to growth of biomass on the particles, and shearing and sloughing. However, the low values of VSS measured throughout these trials were subject to large errors due to sampling diculties. From Table 3 the VSS appeared to make a signicant contribution to nitrication rates in the rst stages of reactor operation, particularly in the Kaldnes and PVA reactors. Despite the unreliable nature of the VSS results, it can be concluded that the immobilisation particles accounted for the majority of the nitrication after stage IV. The three immobilisation media were used in exactly the same way, and therefore their relative success can be measured in terms of a comparison of the nitrication results. These are presented in Fig. 6, calculated using total euent concentrations of nitrite and nitrate. Eect of temperature on nitrication Nitrifying bacteria are known to be sensitive to temperature (Barnes and Bliss, 1983), and a tem-

1174

Wendy M. Rostron et al.

Fig. 6. Comparison of CSTR nitrication rates over time.

Fig. 7. COD removal in the immobilisation reactors.

perature of 168C as well as 258C was assessed as the former was thought to be a more realistic operating temperature. A decrease in nitrication of approximately 10% was seen in stage VII, but this was considerably less than the 58% decrease calculated using a temperature dependence equation from the literature (US EPA, 1991). If nitrication was diusion limited, then the reduction in reaction rate would vary with the diusion coecient. Diusion is proportional to temperature (K) (Lide, 1994), and hence a reduction from 25 to 168C should give a 3% reduction in rate. This value seems to be closer to what was measured than the estimate based on kinetics, indicating that mass transfer rather than reaction kinetics was the rate limiting step in the reactors, and demonstrating the protective eect of immobilisation. Nevertheless, it is important to note that there are several parameters which change with temperature, including diusivity, KL a, the kinetics of nitrication, and equilibrium DO concentration. All of these will inuence the reaction rates measured, particularly in immobilised systems, and all should be considered when attempting to explain temperature eects. During the temperature change a small increase in reactor DO from approximately 7 to 8 mg l 1 was noted. There are several reports in the literature of immobilised biomass being protected from adverse temperatures (Wijels et al., 1995; Asano et al., 1992; Itoh et al., 1989), and substrate diusion limitation was the probable reason. It seems likely that the mechanism which appears to limit immobilised cells, i.e. diusion through the matrix or biolm, can help protect them from adverse conditions by creating micro-environments within the gel matrix, or, with diusion often being the limiting step, a decrease in reaction rate is often masked by diusion, and the ndings of Okey and Albertson (1989) conrm this. Eect of COD addition on nitrication At the minimum operational HRT of 12 h glucose was added to the feed in order to investigate if heterotrophs would grow and inhibit the nitriers (Fig. 7). This was of interest because supernatant

liquors can have high CODs which may need to be reduced before nitrication can occur. The COD loadings applied during stages VIII and IX exceeded those recommended in order to achieve nitrication in trickling lters and activated sludge plants (Metcalf and Eddy, 1991). Heterotrophs grew rapidly reducing the inuent COD by 90% within 10 d of adding 500 mg l 1 glucose to the feed. The ammonia in the Linpor and Kaldnes reactors increased by approximately 40%, while in the PVA reactor the increase was less, and all three reactors showed a loss of nitrate production. This may have been due to oxygen limitations preventing the oxidation of nitrite (nitratication), particularly as heterotrophs grow quickly and exert a large demand for oxygen. However, the DO in the reactors remained above 7 mg l 1, and therefore oxygen limitations were unlikely. Nevertheless, colonisation of the biomass support particles by heterotrophs may have blanketed the nitriers, reducing their oxygen supply. When glucose was added to the feeds, it can be seen that the PVA reactor was initially slower to degrade the COD. This may mean that colonisation of PVA by heterotrophs was more dicult than Linpor and Kaldnes due to the closed structure of the PVA particles. In addition, the PVA reactor had a shorter actual HRT due to the volume of the PVA particles (Table 1), and therefore quicker growth would have been required by the freely suspended heterotrophs. Despite the further loss of nitrication the reactors did maintain some degree of ammonia oxidation, and therefore it may be possible to nitrify high ammonia liquors with high COD concentrations in a one-stage process. Nitrogen mass balance in the CSTRs Nitrogen mass balances over the CSTRs showed some discrepancies between the inuent ammonia and the total euent nitrogen levels, although these were mostly within analytical error ( 6%). These small nitrogen losses could have been due to nitrogen incorporated into biomass during growth, ammonia stripping at high pH, the release of gaseous forms of nitrogen (N2O and NO), and biological denitrica-

Nitrication with immobilized biomass

1175

Fig. 8. Respiration rates of the immobilisation particles.

tion. Nevertheless, high nitrogen losses (818%) in the immobilised reactors were found for stages VIII and IX when glucose was added to more closely mimic return liquors and encourage the growth of heterotrophs. The VSS in the reactors increased substantially (Table 3), and together with good levels of COD reduction (Fig. 7) show that the heterotrophs grew rapidly. 300 mg VSS (approximate value in CSTRs during stage VIII) is made up of 37 mg nitrogen (7.4% of the inuent nitrogen concentration), and therefore heterotrophic biomass growth seemed to account for a signicant proportion of these nitrogen losses. Respiration rates All the immobilisation particles showed increased activity with time, as shown by the increase in respiration rates (Fig. 8), and for the Linpor and Kaldnes particles this was due to the growth of biomass on the particles. The activity of the Linpor particles increased more rapidly than Kaldnes, probably due to the large surface area for microbial attachment, and the porous nature of the sponge. The volatile suspended solids measured in the mixed liquor of the Linpor reactor decreased to virtually zero by stage III (Table 3), demonstrating excellent microbial adsorption by the particles. The respiration rate of Linpor increased by about 400% to 1213 mg O2 l particles 1 h 1 when the HRT of the reactor was reduced from 1.5 (stage IV) to 1.0 d (stage V). As the MLVSS was very low from stage III onwards (approximately zero), this large increase must have been due to the growth of adsorbed biomass in response to the increase in reactor loading. The result appears reliable as it was calculated from duplicate measurements with a coecient of variation of 8%. However, the rate decreased to 468 mg O2 l particles 1 h 1 when the HRT was lowered to 0.5 d (stage VI), possibly due to an increase in shear forces at the greater ow rate. The fragility of the Linpor immobilised biomass compared to the more stable respiration rates observed for both Kaldnes and PVA, demonstrates the weakness of the bacterial attachment, which may be attributed to the pseudo-

encapsulation of biomass within the porous structure. Biomass growth during stage VI increased the activity by the end of the stage to 745 mg O2 l particles 1 h 1. The Kaldnes particles showed a steady increase in activity with time in the reactor, but the activity remained lower than Linpor and PVA throughout the trials. However, a comparison on a particle basis shows that the three media had much more similar activities, as one bulk litre of PVA and Linpor particles contained approximately twice the number of particles as 1 l of Kaldnes, and a larger surface area (Table 1). Encapsulation of nitrifying bacteria in PVA by the freezing method retained approximately 40% cell viability, and reached an equivalent level to that originally immobilised after 60 d. Growth occurred mainly in the outer layers of the PVA particles as seen by scanning electron microscopy SEM (see later). This is in agreement with the literature (Myoga et al., 1991; Tanaka et al., 1994), and was probably due to diusional limitations of substrates into the particles, leading to higher growth rates of the nitriers on the particle surface compared to its interior. Therefore, greater treatment potential may be achieved by using smaller particles than the 8 mm cubes used in this work. For immobilised cell systems steady state operation is usually determined by a constant euent quality. In terms of treatability a steady state can be achieved, however, a true steady state is unlikely to exist due to the continual growth and lysis of cells within the matrix, or in the biolm. Immobilisation coupled with the complexity of nitrication means that long time periods can be required to establish constant euent quality. The results from this preliminary work illustrate this point well as the reactors did not have automatic pH control, and hence when loading changes led to ammonia buildup, the pH changed and substrate inhibition occurred.

Scanning electron microscopy (SEM) SEM was carried out on the initial particles, and after 33, 46 and 108 d in the reactors. The SEM clearly showed the adsorption of biomass onto Linpor and Kaldnes, and the bacteria encapsulated in PVA. Biomass built up on both the inner and outer surfaces of the hollow Kaldnes shapes, and was found within the Linpor foam cubes, as well as on the surface. In the PVA particles the majority of the bacteria were found near the outer edges of the particles when a cross-section was examined. The outer surface of the PVA cubes became covered in globules which appeared to be colonies of bacteria either close to the PVA surface, or breaking through the surface. The available surface area is therefore an important aspect of treatment capacity.

1176

Wendy M. Rostron et al.

Table 4. Nitrication rates with immobilised biomass reported in the literature

Nitrication rate (kg N m 3 d 1) 0.29 0.58 0.2 0.25 0.73 1.5

Wastewater

Methodology

Author and year

Domestic sewage (50 mg NH3-N l 1) Synthetic feed (27 mg NH3-N l 1) Synthetic feed (35 mg NH3-N l 1) Synthetic feed (50 mg NH3-N l 1) Synthetic feed (up to 200 mg NH3-N l 1) Synthetic feed (50 mg NH3-N l 1)

Linpor}adsorption full-scale Captor}adsorption Lab. scale Membrane aerator}biolm Lab. scale Calcium alginate}encapsulation Lab. scale Carrageenan}encapsulation Lab. scale PVA}3 mm cubes}encapsulation Lab. scale

Morper and Wildmoser (1990) Banerji and Liu (1993) Brindle and Stephenson (1996) van Ginkel et al. (1983) Wijels and Tramper (1989) Myoga et al. (1991)

Reactor performance Reactor performance can be evaluated based on the rates of nitrication achieved per unit of reactor volume per unit time. The nitrication rates achieved in the CSTRs were: suspended biomass reactor, 0.36; Linpor, 0.57; Kaldnes, 0.53; and PVA, 0.70 kg N m 3-reactor d 1. These results give a clear indication of the potential for each reactor, demonstrating the much greater nitrication rates achieved with immobilised nitriers than with the freely suspended biomass reactors. The higher treatment rates achievable with PVA compared to Linpor and Kaldnes is also apparent in this reactor comparison. Consideration can also be given to the reaction rate based on the surface area of particles. Using the specic surface areas for the immobilisation particles (Table 1), the nitrication rates achieved were 2.28, 4.24 and 3.97 g N m 2-media d 1 for Linpor, Kaldnes and PVA, respectively. Table 4 summarises the literature results for nitrication using immobilised biomass in order to compare them with the nitrication rates achieved during this work. The rates achieved in our work compare favourably with those reported in Table 4, despite the far higher feed concentrations used here. The higher rates with nitriers in carrageenan (Wijels and Tramper, 1989) and PVA (Myoga et al., 1991) were probably due to the use of pure cultures (in carrageenan), and smaller particles which would have been less aected by diusional limitations. This later contention is supported by the data of Myoga et al. who reported a rate of 1.0 kg N m 3 d 1 for 5 mm particles, and a higher rate of 2.2 kg N m 3 d 1 for 1 mm particles. At present PEG is the only encapsulation immobilisation media to be used at full-scale. Tanaka et al. (1994) reported maximum rates of 0.54 kg N m 3reactor d 1 using PEG pellets (3.6 kg N m 3media d 1) to treat wastewater from a sludge drying plant. The PVA reactor attained a maximum nitrication rate of 0.70 kg N m 3-reactor d 1, equivalent to 3.04 kg N m 3-PVA d 1. The rates achieved in the PVA reactor could be further

improved with particle and reactor optimisation to improve oxygen mass transfer. These results show that the application of nitriers immobilised in PVA warrants further research. In summary, the reactor containing PVA-encapsulated bacteria maintained full nitrication even at an HRT of 1 d, and consistently oxidised a greater percentage of the inuent ammonia than the reactors with adsorption particles. This was probably due to this reactor supporting more biomass and/or encapsulated bacteria having greater protection from substrate inhibition eects. The Kaldnes particles showed a slightly higher nitrication rate per unit area of particles than PVA, hence varying the particle ll amounts from the 25% used here would be worth further investigation, particularly as adding Kaldnes particles has relatively little impact on the actual HRT. However, considering the actual HRT in the adsorption reactors was over 18% higher than that in the PVA reactor, due to the volume taken up by the PVA particles (Table 1), the use of biomass immobilised in PVA appears very promising. This form of immobilisation also has a great potential for optimisation by varying the particle production methodology, for example, by reducing the particle size.

CONCLUSIONS

The following conclusions can be drawn from this work: 1. The freely suspended reactors fully nitried a 500 mg N l 1 synthetic feed at an HRT of 2.2 d, but started to fail at 1.5 d and were washed out at 1 d. In contrast, at a 1 d HRT both the Linpor and Kaldnes reactors experienced problems with nitrite accumulation, but after 3 weeks nitrication in both had recovered to >90%. Finally, with the PVA cubes at 1 d full nitrication was maintained. At 12 h all reactors suered a loss of nitrication, with the PVA reactor maintaining the highest nitrication rate and 30% full nitrication to nitrate.

Nitrication with immobilized biomass

1177

2. Increased nitrication rates during oxygen sparging indicated that both the Kaldnes and PVA reactors were oxygen limited, and most likely the Linpor one was also. 3. A decrease in operating temperature of the immobilised biomass reactors from 25 to 168C caused an approximate 10% decrease in nitrication, i.e. considerably less than predicted for freely suspended biomass, and therefore it appears that nitrication was mass transfer rather than kinetically limited. 4. When 500 mg l 1 of glucose was introduced into the feed a reduction in nitrication occurred due to a heterotrophic population developing in the reactors, particularly in the adsorbed biomass reactors which lost 35% of nitrication in comparison with only a 7% reduction in nitrication in the PVA reactor. It appears that the colonisation of PVA by heterotrophs is more dicult than either Linpor or Kaldnes. 5. Respiration rates for all particles increased with time in the reactors. Nitriers immobilised in PVA retained approximately 40% of their viability based on particle respiration rates after immobilisation. After prolonged operation biomass became more prevalent on the outer surface of the PVA particles indicating diusion limitations. 6. Nitrication rates were generally higher for PVA particles than for the same bulk volume of Linpor and Kaldnes, and compared favourably with literature values for mixed cultures. 7. PVA particles showed excellent promise as an immobilisation medium. They not only achieved the highest volumetric nitrication rates, but are also amenable to being optimised in terms of both manufacture and size, and therefore warrant further research.
Acknowledgements}Wendy Rostron would like to acknowledge the support of EPSRC and WRc for this work, which was done within the WRc/Imperial College Postgraduate Training Partnership scheme.

REFERENCES

Anthonisen A. C., Loehr R. C., Prakasam T. B. S. and Srinath E. G. (1976) Inhibition of nitrication by ammonia and nitrous acid. J. Water Pollut. Contol Fed. 48, 835852. APHA, AWWA and WEF (1992) Standard Methods for the Examination of Water and Wastewater, 18th ed. American Public Health Association, American Water Works Association and Water Environment Federation. Asano H., Myoga H., Asano M. and Toyao M. (1992) A study of nitrication utilizing whole micro-organisms immobilised by the PVA-freezing method. Water Sci. Technol. 26, 10371046. Banerji S. K. and Liu M.-H. (1993) Comparison of captor system with other xed-lm media nitrication systems. Water Pollut. Res. J. Canada 28, 289310. Barnes D. and Bliss P. J. (1983) Biological Control of Nitrogen in Wastewater Treatment. University Press, Cambridge.

Brindle K. and Stephenson T. (1996) Nitrication in a bubbleless oxygen mass transfer membrane bioreactor. Water Sci. Technol. 34, 261267. European Community (1991) Council directive of 21 May 1991 concerning urban waste water treatment. Ocial Journal of the European Communities, L135/40-52. Golla P. S., Reddy M. P., Simms M. K. and Laken T. J. (1994) Three years of full-scale captor process operation at Moundsville WWTP. Water Sci. Technol. 29, 175181. Grulois P., Bousseau A., Blin E. and Fayoux C. (1993) Evaluation of the impact of return ows on the operation of a wastewater treatment plant. Water Sci. Technol. 28, 273281. Hanaki K., Wantawin C. and Ohgaki S. (1990) Eects of the activity of heterotrophs on nitrication in a suspended-growth reactor. Water Res. 24, 289296. Husken L. E., Tramper J. and Wijels R. H. (1996) Growth and eruption of gel-entrapped microcolonies. In R. H. Wijels, R. M. Buitelaar, C. Bucke and J. Tramper, eds Immobilised Cells: Basics and Applications. Progress in Biotechnology, Vol. 11 pp. 336340. Elsevier, Amsterdam. Itoh K., Itadani T., Yosimura H. and Shinoda S. (1989) Immobilisation of nitrifying bacteria and its application for wastewater treatment. Eisei Kagaku 35, 125133. Lessel T. H. (1993) Upgrading and nitrication by submerged biolm reactors}experiences from a large scale plant. IAWQ Second International Specialised Conference on Biolm Reactors, Paris, September 29 October 1, pp. 231238. Lide D. R. (1994) CRC Handbook of Chemistry and Physics, 75th ed. CRC Press, Boca Raton, FL. McLoughlin A. J. (1994) Immobilisation as a strategy to regulate ecological competence of inocula. Adv. Biochem. Eng./Biotech. 51, 145. Metcalf and Eddy (1991) Wastewater Engineering: Treatment, Disposal, and Reuse, 3rd ed. McGraw-Hill, New York. Morper M. (1994) Upgrading of activated sludge systems for nitrogen removal by application of the Linpor CN process. Water Sci. Technol. 29, 167176. Morper M. and Wildmoser A. (1990) Improvement of existing wastewater treatment plants eciencies without enlargement of tankage by application of the linpor process}case studies. Water Sci. Technol. 22, 207215. Myoga H., Asano H., Nomur Y. and Yoshida H. (1991) Eects of immobilisation conditions on the nitrication treatability of entrapped cell reactors using the PVA freezing method. Water Sci. Technol. 23, 11171124. degaard H., Rusten B. and Westrum T. (1993) A New Moving Bed Biolm Reactor}Applications and results. IAWQ Second International Conference on Biolm Reactors, September 29October 1, Paris, pp. 221229. Okey R. W. and Albertson O. E. (1989) Diusions role in regulating rate and masking temperature eects in xedlm nitrication. J. Water Pollut. Contol Fed. 61, 500509. Rusten B., Hem L. J. and degaard H. (1995) Nitrication of municipal wastewater in moving-bed biolm reactors. Water Environ. Res. 67, 7586. Staley J. ed. (1989) Bergeys Manual of Systematic Bacteriology, Vol. 3. Williams & Wilkins, Baltimore, MD. Tanaka K., Nakao M., Mori N., Emori H., Sumino T. and Nakamura Y. (1994) Application of immobilised nitriers gel to removal of high ammonium nitrogen. Water Sci. Technol. 29, 241250. US Environmental Protection Agency (1991) Nitrogen Control. Technomic Publishing, Lancaster. van Ginkel C. G., Tramper J., Luyben K. Ch. A. M. and Klapwijk A. (1983) Characterisation of Nitrosomonas

1178

Wendy M. Rostron et al. Wijels R. H., Leenan E. J. T. M. and Tramper J. (1993) Possibilities of nitrication with immobilised cells in waste-water treatment: model or practical system? Water Sci. Technol. 27, 233240. Wijels R. H., Englund G., Hunik J. H., Leenan E. J. T. M., Bakketun A., Gunther A., Obon de Castro J. M. and Tramper J. (1995) Eects of diusion limitation on immobilised nitrifying micro-organisms at low temperatures. Biotechnol. Bioengng. 45, 19.

europaea immobilised in calcium alginate. Enzyme Microb. Tech. 5, 297303. Wiesmann U. (1994) Biological nitrogen removal from wastewater. In Engineering/Biotechnology, Vol. 51, ed A. Fiechter, pp. 113155. Biotechnics/Wastewater Springer, Berlin. Wijels R. H. and Tramper J. (1989) Performance of growing Nitrosomonas europaea cells immobilised in k-carrageenan. Appl. Microb. Biotech. 32, 108112.