Production of biogas from water hyacinth (Eichhornia crassipes) (Mart) (Solms) in a two-stage bioreactor A.K. Kivaisi* and M. Mtila A two-stage rumen-derived anaerobic digestion process was tested for the conversion of water hyacinth shoots and a mixture of the shoots with cowdung (7:3) into biogas. Under conditions similar to those of the rumen and loading rates (LR) in the range of 11.6���19.3 g volatile solids (VS) l)1 d)1 in the rumen reactor, the degradation efficiencies were 38% for the shoots and 43% for the mixture. The major fermentation products were volatile fatty acids (VFA) with a maximum yield of 7.92 mmol g)1 VS digested, and biogas with a yield of 0.2 l g)1 VS digested. The effect of varying LR, solid retention time (SRT) and dilution rates on the extent of degradation of the water hyacinth��� cowdung mixture was examined. Overall conversion of the substrate was highest at the loading rate of 15.4 g VS. l)1 d)1. Varying the retention times between 60 and 120 h had no effect on the degradation efficiency, but a decrease was observed at retention times below 60 h. The overall performance of the reactor was depressed by changing the dilution rate from 0.5 to 0.34 h)1. By applying a LR of 15.4 VS. l)1d)1, a SRT of 90 h and a dilution rate of 0.5 h)1 in the rumen reactor, and connecting it to a methanogenic reactor of the upflow anaerobic sludge blanket type, 100% conversion efficiency of the VFA into biogas with a methane content of 80% was achieved. The average methane gas yield was 0.44 l g)1 VS digested. Key words: Two-stage anaerobic digestion, rumen microorganisms, water hyacinth. In sub-Saharan Africa, the increased use of synthetic agricultural fertilizers which end up in rivers, lakes and dams, and the discharge of untreated industrial waste- water into surface-water bodies have resulted in eutro- phication over the years (Rogers & Davis 1972). This has resulted in extensive infestations of a prolific water weed, the water hyacinth. In East Africa water hyacinth has been growing in large rivers and dams for most of the century, and during more recent years it has heavily infested most of the shores of Lake Victoria (Johansson 1976). The weed is obstructing navigation, interfering with fishing and re- ducing the recreational values of the beaches. It is choking the lake biota and providing suitable habitats for the breeding of disease vectors (Mshigeni & Kivaisi 1995). Despite the detrimental effects of the water hyacinth infestation, the weed has several economic uses which can be part of its management (Pinto et al. 1987 Tripathi & Shukla 1991). Among the possible uses, anaerobic di- gestion of water hyacinth to methane gas appears to be attractive (Shilapour & Smith 1984 Shankar & Tondon 1986 Teherruzan & Kushani 1989). However, previous experiments on anaerobic digestion of water hyacinth obtained low degradation and methane yields at long retention times (Mwadamwar et al. 1990). Due to its lignocellulosic nature, its conversion into biogas in the one-phase reactors used previously, was most probably limited by the hydrolysis step as reported earlier (Noike et al. 1985). In order to enhance degradation of lignocellulosic materials, various methods have been applied. These include physical and chemical pretreatments (Han et al. 1975 Partos et al. 1983 Muller & Trosch 1986), operating at thermophilic temperatures, and the use of highly cel- lulolytic microorganisms (Gijzen 1987 Ahring 1995). World Journal of Microbiology & Biotechnology, Vol 14, 1998 125 World Journal of Microbiology & Biotechnology 14, 125���131 �� 1998 Rapid Science Publishers The authors are with the Applied Microbiology Unit, Botany Department, Uni- versity of Dar es Salaam, POBox 35060, Dar es salaam, Tanzania fax: +255 51 43764. E-mail address: amu@udsm.ac.tz. *Corresponding author.

Since anaerobic digestion is essentially a two-stage pro- cess involving two major metabolic groups of microor- ganism ��� acidogens and methanogens ��� further efforts to optimize the overall anaerobic process have been focus- ing on separate stages. This has been based on the con- cept that in a single reactor which harbours both the acidogenic and methanogenic bacteria, it is not possible to provide optimal conditions for both groups (Pohland & Ghosh 1971). The use of two stages in anaerobic di- gestion of organic matter in waste-water has proved to have several advantages over only one stage. These in- clude higher conversion rates and increased stability of the overall process, protection of methanogens against shock loads, removal (at least in part) of compounds toxic to methanogens, and enrichment of the methane content of biogas since most of the diluting CO2 is pro- duced during the first phase (Hanaki et al. 1987 Mac- Dougall 1993 Alexiou & Anderson 1994). Phase separation has also been applied in the degra- dation of solid organic matter. A two-phase rumen-de- rived reactor which employs rumen microorganisms (RUDAD process) for an efficient acidogenesis step has been shown to enhance degradation of various lignocel- lulosic materials (Kivaisi et al. 1992 Kivaisi & Eliapenda 1995). The present study was conducted to establish condi- tions for optimal conversion of water hyacinth into bio- gas in the RUDAD process. Materials and Methods Preparation of the Substrates Water hyacinth plants were obtained from five sites along River Pangani (North East Tanzania). After harvesting, all the plants were put in one heap and mixed thoroughly. Roots were re- moved by cutting and the shoots were collected together and labelled as water hyacinth shoots (WHS). The shoots were sun- dried for 5 days and milled to pass through a 2 mm mesh sieve. A portion of the ground WHS was mixed with dry cowdung (CD) in the ratio of 7:3 according to Mwadamwar et al. (1990) as an additional nitrogen source and was labelled WHS���CD. Both materials (WHS and WHS���CD) were stored at room tempera- ture (25 ��C). Their chemical compositions are shown in Table 1. Fermentation Systems The Bioreactor. A 3-l fermenter (1.5 l working volume), operated with differential removal rates of solids and liquids as previ- ously described (Gijzen et al. 1986) was used for efficient acid- ogenesis of the WHS and WHS���CD. The fermenter was inoculated with strained fresh rumen fluid obtained from a slaughter house. After inoculation, the fermenter was filled with pre-warmed (39 ��C) fermentation medium according to Rufener et al. (1963), which was modified by the addition of 0.2 ml l)1 trace elements (Vishniac & Santer 1957) and NH4Cl (28 mM). M Desired dilution rate and solid retention times (SRT) were es- tablished by adjusting peristaltic pumps which continuously supplied fresh fermentation medium and removed filtered (30 lm pore size) liquid effluent, while homogenous reactor contents were removed once every day shortly before substrate addition. The difference adjusted between the rate of buffer supply and filtered effluent removal determined the amount of homogenous reactor contents, thus leading to the desired SRT. The reactor contents were mixed every 15 min for a period of 45 s by means of a laboratory rotary shaker. The Rudad-process. The rumen reactor was placed in serial con- nection with an upflow anaerobic sludge blanket (UASB)-type methanogenic reactor with a total working volume of 2.5 l. A detailed description of the features and mode of operation has been given earlier (Gijzen et al. 1986 Kivaisi et al. 1992). Experimental Conditions The rumen-derived reactor was tested for anaerobic degrada- tion of WHS and WHS���CD as substrates with an arbitrarily chosen loading rate (LR) of 11.6 g of volatile solids per litre per day (g VS l)1 d)1) at standard conditions (solid retention time of 60 h and a dilution rate of 0.5 h)1 fermenter volumes per day). All experiments lasted 14 days. In order to establish optimal conditions for the degradation of the WHS���CD, vari- ous combinations of LR, SRT and dilution rate were tested. Loading rates were varied between 11.6 and 19.3 g VS l)1 d)1 at standard conditions. The effect of varying SRT was tested in the range of 40 and 120 h at a dilution rate of 0.5 h)1 and an arbitrarily chosen loading rate (LR) of 15.4 g VS l)1 d)1. Using a constant SRT of 60 h and a LR of 15.4 g VS l)1 d)1 the effect of varying the dilution rate in the range of 0.9���0.34 h)1 was examined. Anaerobic conversion of the substrate into biogas in the RUDAD process was examined in a duplicate experiment for 21 days. The substrate loading rate was 15.4 g VS l)1 d)1 throughout the experimental period, except on the day of com- mencing the experiment when the rumen reactor loading was doubled. The acidogenic reactor was run at a constant SRT of 90 h and D 0.5 h)1. Coupling of the reactors started from day 7 of the experimental period. Liquid effluent from the acidogenic reactor containing the organic acids, was continuously fed to the methanogenic reactor (UASB) at a rate of 3 l d)1. The effluent of the UASB-reactor was analysed for volatile fatty acids (VFA) and was discharged. Analytical Procedures Neutral detergent fibre (NDF), acid detergent fibre (ADF), cel- lulose, hemicellulose and lignin (permanganate method) were Table 1. Composition of the substrates (% dry weight, mean* % standard deviation). Determination WHS WHS���CD Dry weight 92:6 0:1 95:1 0:2 Volatile solids 84:9 0:1 77:1 0:1 Ash 15:1 0:1 22:8 0:1 NDF 58:0 0:2 52:4 1:4 ADF 34:9 0:3 33:6 1:2 Hemicellulose 23:1 0:3 18:8 0:2 Cellulose 25:2 0:4 22:5 0:4 Lignin 9:7 0:2 11:1 0:5 Cell solublesF 26:9 0:1 24:7 0:3 C:N ratio 20:1 16.8:1 * n = 6. F Volatile solids minus NDF. 126 World Journal of Microbiology & Biotechnology, Vol 14, 1998 A.K. Kivaisi and M. Mtila A.K. Kivaisi and M. Mtila