CHEESE WHEY AND CHEESE FACTORY WASTEWATER TREATMENT WITH A BIOLOGICAL ANAEROBIC–AEROBIC PROCESS

F. Malaspina, L. Stante, C. M. Cellamare and A. Tilche

ENEA - Divisione Tecnologie di Depurazione e Trattamento Reflui - Via Martiri di Monte Sole, 4
40129 Bologna - Italy

ABSTRACT

A research on the anaerobic treatment of raw cheese whey started in 1990 with the objective of developing a technology suitable for medium size cheese factories that have growing disposal problems and cannot afford high investment costs for whey valorisation technologies (such as whey protein and lactose recovery, spray drying, etc.). In order to couple process stability and high loads, a new downflow-upflow hybrid reactor (DUHR) has been designed. The reactor was able to reach B_v values around 10 g COD·L^–1·d^–1, with 98% COD converted to gas and effluent soluble COD values close to 1,000 ppm; no external addition of alkalinity is required to maintain a stable pH that was constantly around 6.5-6.7 in the downflow pre-acidification chamber and around 7.5 in the bio-methanation upflow chamber. The high strength of the cheese whey treated gives an effluent that still contains high amounts of COD, ammonia nitrogen and phosphorus and therefore a post treatment is required in order to meet standard limits. Tests of post treatment were carried out during two years with a Sequencing Batch Reactor (SBR). The SBR was tested at various F/M values with different durations of anoxic-anaerobic-oxic cycles, obtaining, under certain conditions, more than 90% removal of COD, Nitrogen and Phosphorus.

Keywords

Anaerobic digestion, cheese whey, downflow-upflow hybrid reactor, two-phase anaerobic digestion, Sequencing Batch Reactor, nutrient removal.

INTRODUCTION

Milk whey is a byproduct that the dairy industry can reprocess or transform in other valuable products. Small and medium size cheese factories, like most of those in Italy produce high quality typical cheeses, do not have the economic convenience or the market dimension to afford most of the reuse technologies available. Traditionally, these cheese factories are strictly connected to pig farms where milk whey is directly used as the liquid basis for animal feed.

If for any reason (environmental, sanitary, economic) the downstream animal farm is lacking, a solution to whey disposal must be found. This is the case of most cheese factories of the Sardinian Pecorino cheese; due to epidemic disease, pig farming in Sardinia has been drastically reduced a few years ago.
From the analysis of the problem that ENEA was called to do, anaerobic digestion - followed by aerobic post-treatment together with process wastewater - came out theoretically as the most convenient solution. Biogas produced could be economically reused to substitute most or all fossil fuels used for process thermal energy generation; moreover, anaerobic digestion should destroy the greatest part of COD, allowing the effluent to be post-treated together with the process wastewater in the existing treatment plants of the factories with only little modifications to the plant structure. Anaerobic treatment of whey is not a new process. The literature about it is very rich, but most of the authors have worked with very diluted whey, a much simpler waste to treat. Because of its very high biodegradability (close to 99%) and concentration (~70 g·COD·L^–1) and the very low bicarbonate alkalinity (~50 meq·l^-1), raw whey is a particularly difficult substrate to treat in high loaded anaerobic digesters. Moreover, granulation is not supposed to occur using whey as substrate (Hickey et al., 1991), making difficult the use of UASB reactors.

In the present paper, part of the results of laboratory experiences started in 1990 are reported. The research was first carried out on a hybrid upflow reactor, later on a two-phase system and finally on a downflow-upflow hybrid reactor, a new concept that was specifically developed for raw whey treatment.
The main concepts behind this design are that phases are separated within the same reactor and that, in order to obtain more stable process conditions and higher resistance to shock loadings, a downflow completely mixed pre-acidification zone is followed by a plug-flow upflow biomethanation zone. The CSTR zone acts as a shock buffer. In fact, the influent is introduced at the top of the downflow chamber, where mixing is very high and bacterial activity should be correspondingly high. This design could reduce the risk of pH drop in case of recycle pump failure. Moreover, the DSFF design of the pre-acidification chamber should reduce the passage of acidifying biomass to the upflow part.

The effluent from the anaerobic reactor cannot be discharged into surface waters before a post-treatment for the removal of residual organic matter and nutrients. Its composition is highly unbalanced, because most of the organic matter has been destroyed in the anaerobic step, while nitrogen and phosphorus are quantitatively conserved as soluble inorganic compounds. A post-treatment has been therefore designed and tested. The process is based on a Sequencing Batch Reactor, that from previous experiences has demonstrated to be very suited to treat wastes containing high nutrient concentrations. The organic matter needed for the biological removal of nutrients is provided by the addition of a syntetic stream of cheese factory process wastewater that presents a very high Carbon/Nitrogen ratio.

MATERIALS AND METHODS

All experiments were carried out in mesophilic conditions, utilizing raw whey as substrate, obtained from a cheese factory nearby Bologna, derived from the manufacturing of cheese from 90% cow and 10% sheep milk. Its main characteristics are reported in Table 1.

Table 1 raw cheese whey Characteristics

Parameter	Unit	Average	St. Dev.
Total COD	mg·l^–1	68814	11518
Soluble COD	mg·l^–1	57876	11272
TSS	g·kg^–1	1.3	1.14
VSS	g·kg^–1	0.94	0.74
TKN	mg·l^–1	1462	263
NH4^+–N	mg·l^–1	64	31
Total P	mg·l^–1	379	49
PO4–P	mg·l^–1	326	64

Data from previous experiences (F. Malaspina et al., 1994) were used for the design of a downflow-upflow hybrid reactor (DUHR) that is still in operation in our laboratory (Fig. 1); results from the first nine months are presented here. The reactor, of the total volume of 51 L is composed of a downflow pre-acidification chamber where influent is pumped, filled 2/3 with a channelled polyurethane filter; on the bottom of it, this chamber is connected with an upflow one - arranged as a hybrid reactor - that contains a similar filter located in the top 2/5 of it. The volumetric ratio between the two parts of the reactor is 1:5. A recycle from the top of the second chamber (r≥2.5) is applied to provide alkalinity and dilution to the influent.

An SBR of 5 dm³ of maximum capacity (Fig. 2) operated whit anoxic, anaerobic and aerobic phases for each cycle in order to obtain biological removal of organic substances and nutrients. Several tests were carried out at different loadings and operating cycles, in order to evaluate the best operational condition for COD, nitrogen and phosphorus removal.

In order to decrease the very high TKN/COD ratio of the anaerobic effluent, a syntetic stream of cleaning waters of cheese factory, rich in organic substances and poor of nutrients, was mixed with the anaerobic effluent in a ratio similar to the true wastewater/whey volume ratio of the Parmigiano Reggiano cheese manufacturing process: two parts of cleaning water and one part of digested effluent.

Fig. 1. The DUHR reactor	Fig. 2. The SBR

Analyses of COD, TS, VS, TSS, VSS, NTK, total P and sulphides were performed following the Standard Methods (APHA, 1989). Monovalent anions and cations were analysed using a HPIC (Dionex 4000i). Total alkalinity (TA) was measured titrometically at pH 3.8. Bicarbonate alkalinity (BA) was calculated from VFA and TA values. VFA were determined gaschromatographically, using a DANI 8510 GC equipped with a 25 m · 0.53 mm 1.2 mm capillary wide bore column (Alltech SO FA bonded FSOT) and a FID detector; hydrogen was used as carrier gas and 2, 2, dimethylbutyric acid was employed as internal standard.
Biogas was metered with wet tip gas meters and analysed gaschromatographically with a DANI 3865 GC equipped with a TC detector; a custom made 6m · 3.18 mm teflon packed column filled with Chrompack Hayesep C allowed good separation of N₂, CH₄, CO₂ and H₂S.

RESULTS AND DISCUSSION

Results of the test carried out on DUHR are presented in Fig 3 to 15. After 40 days needed for the adaptation of the seed sludge taken from a pig waste digester, B_v was gradually increased up to the design loading rate of 10 g COD·L^–1·d^–1.

Fig. 3. Bv applied to DUHR	Fig. 4. VFA in acidogenic(H) and methanogenic (E) section	Fig. 5. pH of acidogenic (H) and methanogenic (E) section

After three weeks at high load, some methanogenic biomass washout due to the very high mixing caused by gas production and by the passage of acidogenic biomass from the first into the second chamber resulted in growth of VFA concentration and particularly of acetic acid, without causing therefore any unbalance of pH in the upflow chamber. On the other hand, the increase of loading rate resulted sometimes in pH decrease in the downflow chamber. Phase separation allowed to maintain stable pH in the methanogenic section also after an accidental recycle breakdown happened during a weekend around day 215; during this accident, pH in the first chamber fell down to 4.9, while in the second chamber it didn’t show any significant decrease. The process recovered very rapidly after recycle restore and a very little alkalinity supply (5.25 meq·L^–1 Na₂CO₃ + 2 meq·L^–1 CaCO₃). Recycle ratios in the range 1:4 to 1:2.5 were usually sufficient to provide alkalinity and dilution to the influent. No external alkalinity was used for pH control, except than after the accident described.
COD removal was always well over 90%; within the experimental B_v range, no dependence of COD removal from B_v is shown. During a 36 days period of stable operation at 10 g COD·L^–1·d^–1, 98.4% COD removal (Fig. 13) and methane production of 0.33 nL·g COD_in^–1 (Fig. 14) have been calculated. Peak biogas productions close to 10 vol·vol^–1·d^–1 were achieved at the maximum loading. Methane content in biogas averaged 53%, with no significant dependence on B_v. The high gas flow rates resulted in upflow velocities 2 orders of magnitude higher respect to the liquid upflow velocity (fig. 15). Due to the little sludge yield and to the absence of granulation, the correct design of a g/l/s separation device is critical in order to maintain the reactor biomass content at high loading rates.

Fig. 6. pH vs. Bv in acidogenic(H) and methanogenic (E) section of DUHR	Fig. 7. VFA vs. Bv in acidogenic(H) and methanogenic (E) section of DUHR		Fig. 8. Biogas (F) and methane (E) production of DUHR (normalised volumes)
Fig. 9. Concentration of main VFA in acidogenic section of DUHR (acetic acid (Z), propionic acid (J), n–butyric acid (C), n–valeric acid (F) and n–capronic acid (G)		Fig. 10. Concentration of main VFA in methanogenic section of DUHR (acetic acid (Z), propionic acid (J), n–butyric acid (C), n–valeric acid (F) and n–capronic acid (G)

During some steady loading periods at different B_v, intensive profile sampling allowed to calculate some basic process parameters, reported in Table 2. Calculations were done making the assumption, derived from literature data on the polyurethane support material used (Van Rompu and Verstraete, 1988), that the filter contained 15 g VSS·L^–1. Data are in very good agreement with the ones obtained in previous experiences with a two-phase CSTRs.

During the whole experiment the internal reactor liquor always showed a viscous consistence. This phenomenon was particularly relevant during periods of overloading, like for instance in the load increase period of the start-up phase, when severe washout can occur because the biomass settling capacity is highly reduced. This phenomenon is probably due to the over-production of ExoPolySaccharides (EPS) of bacterial origin; their synthesis uses mono- and di-saccharides, glucosamines and uronic acids as substrates, and is favoured by high concentrations of sugars and by high C/N ratios (Sutherland, 1985; Guiot et al., 1988).

table 2 duhr kinetic parameters

Period	Bv	F/M	Uspec.	Yobs.	µ
(d)	(gCOD·L^–1·d^–1)	(gCOD·gVSS^–1·d^–1)	(gCODr·gVSS^–1·d^–1)	(gVSS·gCODr^–1)	(d^–1)
7	7.5	0.58	0.56	0.079	0.044
7	9.5	0.63	0.56	0.032	0.019
9	10	0.66	0.55	0.060	0.033
7	8	0.66	0.51	0.120	0.066
average		0.63	0.55	0.075	0.04
St. Dev.		0.04	0.03	0.04	0.01

		Fig. 11. COD removal during DUHR test			Fig. 12. COD removal of DUHR as function of applied Bv
	Fig. 13. Mass of CODrem versus mass of CODin during a period of 36 d at Bv of 10 gCOD^.L^-1.d^-1 in DUHR.			Fig. 14. Volume of methane produced from inlet COD during the same period.
			Fig. 15. Liquid flow in acidogenic (F) and methanogenic (B) section of reactor and gas flow in both section (E) of DUHR during the experiment.

Effluent characteristics are summarized in Table 3. The high total COD values are to be ascribed to the relevant bacterial washout at high loads - also witnessed by the high TSS and VSS figures - due to the absence, in the laboratory reactor, of a well designed g/l/s device. The effluent soluble COD was very low respect to the influent values but not enough to approach effluent standards, that was overpassed also by nitrogen and phosphorus. It is therefore necessary to provide a post-treatment before effluent discharge.

table 3 duhr effluent characteristics

Parameter	Unit	Average	St. Dev.	sample n.
Total COD	mg·l^–1	5839	3117	25
Soluble COD	mg·l^–1	1085	1263	25
TSS	g·kg^–1	5.19	3.30	18
VSS	g·kg^–1	3.02	1.69	14
NH4^+–N	mg·l^–1	816	122	6
PO4–P	mg·l^–1	35	21	3

The addition of process wastewater deriving from dairy machinery washing, that presents high C/N ratios, allows to compose a wastewater that can be biologically treated. Laboratory experiences carried out using a Sequencing Batch aerobic reactor have already demonstrated the feasibility of combined carbon, nitrogen and phosphorus biological removal. SBR operated with different cycles in order to test the best operational conditions for removing both organics and nutrients. The cycles are summarized in Fig. 16 and the average main characteristics of influent for SBR treatment, made of one parts of effluent from DUHR and two parts of process wastewater, are reported in Table 4.

TabLE 4 Main characteristics of SBR influent during the tests.

Parameter	Unit	Average value
Total COD	mg·l^–1	4048
Soluble COD	mg·l^–1	2611
TKN	mg·l^–1	435
NH4^+-N	mg·l^–1	165.27
TSS	g·kg^–1	1.45
VSS	g·kg^–1	1.08
Total P	mg·l^–1	82.42
PO4-P	mg·l^–1	27.76

Fig. 16. SBR cycles for the five test periods.

The first test lasted three months with 4 Nitrification/Denitrification(N/DN) cycles/24h at loading condition as showed in Fig. 17 and Fig. 18 and a average F/M of 0.215 gCOD^.gVSS^–1.d^–1 (corresponding to HRT of 2.5 days and B_v conditions of 0.93 gCOD^.dm^–3.d^–1).

The reactor gave very good performance both in COD and nutrient removal as shown in Fig. 19 and in Fig. 20. Overall performances are summarised in Table 5 (concentrations are in mg·l^–1, Relative Standard Deviation -RSD- and removal rate are in percent).

Table 5 Inlet and outlet average value during test #1

	Parameter	inlet value	RSD		outlet value	RSD	removal
	Total COD	3841	19		311	42	90.7
	Soluble COD	2672	22		137	40	94.1
	TKN	410	16		16	141	93.2
	NH4^+-N	166	42		0.49	172	99.6
	NO3^--N	0.83	80		33	88	–
	Total P	66	46		9,63	23	93.2
	PO4-P	18.6	66		4	34	78.79
Fig. 17. COD conc. in SBR during test #1				Fig. 18. F/M applied during test #1
Fig. 19. COD removal in SBR during test #1				Fig. 20. N and P removal during test #1

The averaged values show quite high COD and Nitrate effluent values, but they are due to some peak that don’t represent the normal performance, when values very close to the Italian effluent standards were reached. Very good removal of phosphorus was obtained particularly at the end of the period, when stable biological P removal was established. The sludge had an excellent settleability.

In order to try to improve reactor performance, a second test was carried out with shorter and more frequent N/DN periods (6 cycles/24h) with the following operating conditions: B_v of 0.86 gCOD^.dm^–3.d^–1, HRT of 3.6 days and F/M of 0.128 gCOD^.gVSS^–1.d^–1. Results were not satisfactory, as summarised in Table 6 (concentrations are in mg·l^–1, RSD and removal in %), and the test was stopped after 48 days.

Table 6 Inlet and outlet average value during test#2

Parameter	inlet value	RSD	outlet value	RSD	removal
Total COD	3241	42	1259	62	61.1
Soluble COD	2015	74	210	54	89.5
TKN	385	10	65	141	83.1
NH4^+-N	121	79	17.6	156	85.4
NO3^--N	0.72	75	60.7	62	–
Total P	28.8	–	6.45	–	76
PO4-P	13.24	79	5.5	43	58.4

Outlet data showed high value of ammonia and nitrate due to low rates of nitrification and denitrification, whereas the very high value of COD was caused by washout of sludge due to the very consistent presence of filamentous bacteria.

The third test was run with 5 N/DN cycles/24h with organic loading as reported in Fig. 22 (averge B_v of 0.94 gCOD^.dm^–3.d^–1, HRT of 4 days and F/M of 0.188 gCOD^.gVSS^–1.d^–1). Under these operating conditions and cycles (Fig. 16) good nitrogen removal with low concentration of ammonia and nitrate in the effluent was obtained, while P, TKN and CODt removal were poor due to wash–out of high SVI sludge (Fig. 23 and 24). Summary reactor performance of this test are reported in Table 7 (concentrations are in mg·l^–1, RSD and removal are in %).

Fig. 21. COD concentration in SBR in test #3 Fig. 22. F/M applied during test #3

Table 7 Inlet and outlet average value during test #3

Parameter	inlet value	RSD	outlet value	RSD	removal
Total COD	3542	21	800	106	77.4
Soluble COD	2271	62	194	113	91.43
TKN	434	51	148	55	65.7
NH4^+-N	136	68	4.21	186	96.8
NO3^--N	0.43	140	5.4	100	–
Total P	141	45	49	70	65
PO4-P	23.4	84	7	34	70

Fig. 23. COD removal in SBR during test #3 Fig. 24. N and P removal during test #3

The fourth test lasted 210 days with 4 N/DN cycles/24h at the operating conditions reported by Fig. 25 and Fig. 26 (average value of B_v 1.36 gCOD^.dm^–3.d^–1, HRT 3.68 days and F/M 0.27 gCOD^.gVSS^–1.d^–1). Reactor performance is reported in Fig. 27 and 28, and summarised in Table 8 (concentrations are in mg·l^–1, removal and RSD are in %).

Table 8 Inlet and outlet average value during test #4

Parameter	inlet value	RSD	outlet value	RSD	removal
Total COD	4480	44	1082	99	77.8
Soluble COD	3031	67	201	105	93.3
TKN	484	12	153	111	68.3
NH4^+-N	267	46	11	239	95.8
NO3^--N	1.19	–	1.56	–	–
Total P	109	85	44.4	113	59.3
PO4-P	59	103	7.49	90	87.3

Also in this test some bulking problems (Fig. 29) caused wash–out and high values of total COD, TKN and total Phosphorous in the effluent (low removal as shown in picture); despite the high load, good performance was obtained for nitrification and denitrification (see graph of Fig. 28).

Fig. 25. COD concentration in test #4 Fig. 26. F/M applied during test #4

Fig. 27. COD removal in SBR during test #4	Fig. 28. N and P removal during test #4
	Fig. 29. Outlet suspended solids during test #4

In order to remove filamentous bacteria, a fifth test was carried out with unmixed fill at high load (B_v of 1.28 gCOD^.dm^–3.d^–1, HRT of 3.66 days and F/M of 0.266 gCOD^.gVSS^–1.d^–1) and 3 N/DN cycles/24h (see Fig. 16) according to “K_s strategy” in bulking control (Chudoba, 1985).

The test lasted 76 days and gave very good results in filamentous bacteria control: after 60 days only few filaments were found in the sludge that, on the other hand, was very rich in big sarcinas similar to G–Bacteria. Good performance on COD and nitrogen removal was obtained (Fig. 32), while phosphorous removal was poor (Table 9), probably due to conditions that promoted the growth of G-bacteria instead of polyphosphatic ones (Cech and Hartman, 1993).

Table. 9. Inlet and outlet average value during test #5

Parameter	inlet value	RSD	outlet value	RSD	removal
Total COD	4737	30	1960	61	58.6
Soluble COD	3064	41	375	72	87.7
TKN	463	6	157	61	66.1
NH4^+-N	136	54	6.13	122	95
NO3^--N	1.56	57	0.78	80	–
Total P	65.65	36	43	96	34.8
PO4-P	24.42	87	8.41	115	65.5

Fig. 30. COD in SBR during test #5 Fig. 31. F/M applied during test #5

Fig. 32. COD removal in SBR during test #5

In summary, the SBR demonstrated to have a very good potential in nutrient and organic removal. In order to obtain good nutrient removal together with high sludge settleability (low SVI), an equilibrium between number of N/DN cycles, their duration and feeding modalities has to be found. As a general remark, short and frequent N/DN cycles theoretically allow to reach lower nutrient concentrations, but too short cycles can negatively affect nitrification and denitrification. Short cycles probably favour filamentous bulking; fast and unmixed feeding and longer oxic/anoxic–anaerobic time ratios can be promising strategies of bulking control.

Phosphorus removal was checked during tests by fractionation of cellular P in samples taken at the end of anaerobic and aerobic phases. During the aerobic phases the content of polyphosphates in sludge (15–20 mg/gVSS) was about four fold respect to the samples taken during anaerobic phases (3–5 mg/gVSS) and therefore it is possible to state that biological phosphorous removal is involved in the process. Nevertheless a high fraction of P was removed by precipitation as insoluble “metal-bound” compounds (20–50 mg/gTSS).

Organic matter removal was very good but it was reduced by wash–out due to filamentous bulking and by a background of unbiodegradable COD. Moreover the tests shown that also in presence of very clear supernatant, a relatively high amount of COD was found in effluent. This condition is caused by a background of unbiodegradable very fine suspended-colloidal COD coming from anaerobic digestion. For this reason the effluent from SBR needed a tertiary treatment to remove the unbiodegradable COD. This condition occurs only when the cheese factory uses very low amount of water (and therefore discharges low amount of cleaning–waters) like in our simulation; in many cases however, the volume ratio between cleaning water and cheese whey is from 4 to 6, with a dilution effect that has a positive effect on effluent quality.

Most of residual colloidal and soluble organic substances, that are not biologically degraded, can therefore be easily removed by chemical flocculation and physical separation (flotation or settling).
Coagulation and flocculation tests were carried out dosing FeCl₃ (200-400 mg Fe^.g COD_in^–1) and anionic polyelectrolite (7-20 mg polyelectrolite^.g COD_in^–1) with very good results: 80% COD removal and 90% total P removal.

In conclusion the biological anoxic/anaerobic/oxic process is able to treat successfully effluent from anaerobic digestion of cheese whey when this is mixed with factory cleaning–waters. Physical-chemical treatments can be required in order to reach standard values when low amount of cleaning waters are discharged from the cheese factory.

CONCLUSIONS

Despite of its very high biodegradability, raw cheese whey is a quite problematic substrate to treat anaerobically, because of the lack of alkalinity, the high COD concentration, the tendency to acidify very rapidly, the difficulty to obtain granulation and the tendency to produce excess of viscous exopolymeric materials of probable bacterial origin that severely reduces sludge settleability and can be cause of biomass washout.

The DUHR design demonstrated COD removal rates of 98% at considerably high B_v (10 g COD·L^–1·d^–1) without any need of external alkalinity supply. The particular design allows to obtain phase separation in the same reactor, therefore reducing investment costs respect to separated reactors.

The SBR is a very flexible reactor for both organic matter and nurient removal from wastewater and is particulary suitable for post–treatment of digested whey.

The residual unbiodegradable colloidal and soluble organic substances, that are not degraded biologically, can be easily removed by chemical flocculation and physical separation (flotation or settling).

Acknowledgements

Authors wish to acknowledge Emanuele Sbaffi and Chiara Parmigiani for their precious collaboration.

REFERENCES

APHA- AWWA-WPCF (1989). Standard methods for the examination of water and wastewater. 17th Edition Edited by Clesceri L.S., Greenberg A.E., Trussel R.R.

Cech J. S. & Hartman P. (1993). Competition between polyphosphate and pomysaccharide accumulating bacteria in enhanced biological phosphate removal systems. Water Res., 27, 1219--1225.

Chudoba J. (1985). Control of activated sludge filamentous bulking – VI: Formulation of basic principles. Water. Res., 19, 8, 1017–1022

Guiot S.R., Gorur S.S. and Kennedy K.J. (1988). Nutritional and environmental factors contributing to microbial aggregation during upflow anaerobic sludge bed-filter (UBF) reactor start-up. In: Anaerobic digestion 1988. E.R. Hall and P.N. Hobson editors, 47-53, Pergamon Press.

Hickey R.F., Wu W.-M., Veiga M.C. and Jones R. (1991). Start-up, operation, monitoring and control of high-rate anaerobic treatment systems. Wat. Sci. Tech., 24 (8), 207-255.

Malaspina F., Stante L. & Tilche A. (1994). Anaerobic treatment of cheese whey with a downflow–upflow hibrid reactor. Oral paper preprints of the Seventh International Symposium on Anaerobic Digestion, Cape Town, SA, 23--27 January, pp. 658--667.

Schink B. (1988). Principles and limits of anaerobic degradation: environmental and technological aspects. In: Biology of anaerobic microorganisms, Wiley Interscience Publication, A.J.B. Zehnder ed., John Wiley and Sons, New York.

Sutherland I.W. (1985). Biosynthesis and composition of gram–negative bacteria extracellular and wall polysaccharides. Ann. Rev. Micr. 39, 243–270.

Van Rompu K. and Verstraete W. (1988). The polyurethane parallel plate separator. In: Poster papers, 5th Int. Symp. on Anaerobic Digestion, A. Tilche and A. Rozzi editors, Monduzzi, Bologna, Italy, 777-779.