Elephant grass genotypes for bioenergy production by direct biomass combustion

The objective of this work was to evaluate elephant grass (Pennisetum purpureum Schum.) genotypes for bioenergy production by direct biomass combustion. Five elephant grass genotypes grown in two different soil types, both of low fertility, were evaluated. The experiment was carried out at Embrapa Agrobiologia fi eld station in Seropédica, RJ, Brazil. The design was in randomized complete blocks, with split plots and four replicates. The genotypes studied were Cameroon, Bag 02, Gramafante, Roxo and CNPGL F06-3. Evaluations were made for biomass production, total biomass nitrogen, biomass nitrogen from biological fi xation, carbon/ nitrogen and stem/leaf ratios, and contents of fi ber, lignin, cellulose and ash. The dry matter yields ranged from 45 to 67 Mg ha. Genotype Roxo had the lowest yield and genotypes Bag 02 and Cameroon had the highest ones. The biomass nitrogen accumulation varied from 240 to 343 kg ha. The plant nitrogen from biological fi xation was 51% in average. The carbon/nitrogen and stem/leaf ratios and the contents of fi ber, lignin, cellulose and ash did not vary among the genotypes. The fi ve genotypes are suitable for energy production through


Introduction
In recent decades, the search for alternatives to the use of fossil fuels is increasing globally.High international prices of oil and its derivates and concerns about the environment motivate this process.Moreover, there is great expectation on the possible economic benefi ts from the clean development mechanism projects coming from the use of renewable sources in the agricultural sector, such as bioethanol from sugarcane (Boddey et al., 2008) and eucalyptus timber for coal substitution (Kraxner et al., 2003).There is also a possibility of obtaining energy from plant biomass that could be transformed to charcoal Pesq. agropec. bras., Brasília, v.44, n.2, p.133-140, fev. 2009 or even directly burned for heat production (Samson et al., 2005).
Elephant grass (Pennisetum purpureum Schum.) is highly effi cient in fi xing atmospheric CO 2 , being capable of accumulating more than 60 Mg ha -1 of dry matter per year, when under optimal conditions for growth, at high fertilization rates, especially with N, and good water availability (Andrade et al., 2005).Botrel et al. (2000) tested several clones of elephant grass, among which Cameroon, CNPGL 91 F27-01 and CNPGL 91 F06-3.These authors found an annual average productivity of 31, 43 and 37 Mg ha -1 of dry matter per year, respectively.Traditionally, this crop has been managed for forage production for dairy cattle due to its high yield and protein content.However, for bioenergy purposes, apart from high dry matter production, the plant material should suit quality parameters for reaching the optimal energetic effi ciency, which means high levels of fi bre and lignin, and low levels of water, N and ash (Lemus et al., 2002;McKendry, 2002).
For elephant grass growing in poor N soils, the average of dry matter production after two cuts per year were about 30 Mg ha -1 , and fi bre and lignin contents matched the desirable parameters for energy production from direct burning (Quesada, 2005).Protein content was remarkably reduced in comparison to common levels observed in elephant grass genotypes, destined for forage production.Biomass yield levels were not drastically reduced, in comparison to the ones in fertilised systems (Quesada, 2005).The suitability of the biomass for energy production, along with the possibility of two cuts a year, increases the potential of using this crop for energy production purposes (Samson et al., 2005).
The occurrence of diazotrophic bacteria associated with elephant grass, as reported in previous studies (Baldani et al., 1998;Kirchhof et al., 2001), indicates that biological nitrogen fi xation (BNF) might play an important role in N supply to this crop in poor soils.
The objective of this study was to evaluate the production and quality of biomass and the biological nitrogen fi xation in elephant grass genotypes, for energy production by direct combustion.

Materials and Methods
This study was performed in the experimental area of Embrapa Agrobiologia, in Seropédica, RJ, Brazil (22°49'22"S, 43°38'42"W, at an altitude of 43 m).The experiment was composed of two factors in a split plot array.The main plot was the soil type and the subplot consisted of elephant grass genotypes, in randomized complete block design, with four replicates.The soils under evaluation were Argissolo Vermelho-Amarelo (Typic Hapludult -Acrisol) and a Planosolo Háplico (Typic Albaqult -Planosolo) (Santos et al., 2006), both of low fertility and low plant N availability (Table 1).The following genotypes of elephant grass were evaluated: Cameroon, Bag 02, Gramafante, Roxo and CNPGL F06-3, all of them from the collection of Embrapa Gado de Leite.Plots were composed of 6 lines with 7-m length, with each row 1 m spaced apart, giving a total plot area of 42 m 2 .
At planting, P and K fertilizer were applied in the furrows at a rate of 150 kg ha -1 of K 2 O (KCl) and 100 kg ha -1 of P 2 O 5 (single superphosphate), and 60 kg ha -1 of FTE BR-12.The doses of P, K and micronutrients were applied based on a mean accumulation of these nutrients in elephant grass plants (Andrade et al., 2005;Moreira et al., 2006).No lime application was required (Table 1).Hence, in theory, N was the only factor limiting plant development in the experiment.
The experiment was set up in October 2005.The fi rst cut was made in June 2006, the second in January 2007 and the third in July 2007.Manual weeding was performed as needed.
At harvesting, plant biomass of a 20 m -2 central area of each plot was cut at ground level and weighed fresh, without separating stem and leaves.Samples were taken to oven dry at 65°C, until constant weight, for the determination of accumulated dry matter.The proportion of stems and leaves, in each harvested plot, was calculated from ten plants taken at random within each plot, at the time of the cuts.Oven-dried plant samples were ground to 2 mm in a Wiley mill and, then, powdered in a roller mill similar to that of Smith & Myung (1990).The N content and the 15 N natural abundance of each sample were analysed (1) Semi-micro-Kjeldahl method.Pesq. agropec. bras., Brasília, v.44, n.2, p.133-140, fev. 2009 according to Boddey et al. (2001).Total dry matter and N accumulated by the plants and the proportional BNF contribution were calculated.For the latter, the value of 15 N natural abundance of the whole plant (stems plus leaves) was considered (Yoneyama et al., 1997).Non-N 2 -fi xing weeds were collected from the plots to serve as control plants for the BNF estimation based on the 15 N natural abundance technique (Unkovich et al., 2008).For the fi rst and second biomass cuts the control species considered were Emilia sonchifolia, Tradescantia fl uminensis, and Arnica montana, and for the third, Tradescantia fl uminensis, Richardia brasiliensis, and Arnica montana.For these species, a composite sample per block for each species was taken, with single plant samples being collected randomly within the plots, along the lines of elephant grass.
The quality analysis of the elephant grass biomass consisted of the determination of fi bre content (acid detergent fi bre, ADF), cellulose, lignin and ash contents, following the procedure of Van Soest & Wine (1968).The C:N ratio was calculated considering 45% C in plant biomass.The stem/leaf ratio was also calculated considering the dry mass of each part.
The statistical procedures were those of the SAEG 9.1 (Universidade Federal de Viçosa, 2007).Normality and homogeneity of variance of errors were analysed using the Lilliefors and Cochran & Bartley tests, respectively.The required conditions were met in all cases.The analysis of variance was, then, performed with the application of the F test.Differences between means were separated by the use of Tukey's test at 5% probability.

Results and Discussion
A signifi cant interaction between genotype and soil type was observed regarding total dry matter accumulation after the three cuts (Table 2).At the fi rst cut, after 9 months from planting, dry matter accumulation by genotypes Roxo and Cameroon, when grown on the Acrisol, were 20.3 and 29.5 Mg ha -1 , respectively, and between 21.1 and 32.0 Mg ha -1 , when grown on Planosol.At the second cut, after further 7 months, the greatest dry matter accumulation were observed on the Acrisol, with a signifi cant difference between genotypes Bag 02 and Gramafante, with a production of 18.1 Mg ha -1 and 27.9 Mg ha -1 , respectively.For the genotypes grown on the Planosol, there was a signifi cant difference between Roxo and CNPGL F 06-3, with total dry matter production of 18.8 and 22.8 Mg ha -1 , respectively.At the last cut, after the further 6 months of growth, the dry matter accumulation ranged from 6.7 to 14.5 Mg ha -1 on the Acrisol, with the lowest value observed for genotype Roxo and the largest for genotype Bag 02.On the Planosol, dry matter varied from 5.1 to 10.9 Mg ha -1 for the genotypes Cameroon and Roxo, respectively.The high productive potential for biomass production observed with elephant grass in the present study is consistent with those already reported in literature (Queiroz Filho et al., 2000;Magalhães et al., 2006).Still, the positive results hereby shown are of great relevance for the agroenergy sector, as no N fertilizer was applied and the soils studied were N depleted.
The high performance in dry matter accumulation was consequently accompanied by a high N accumulation by plants (Table 3).However, at the fi rst cut, there were no statistical differences among genotypes, regarding N accumulation, when growing on the Acrisol, with values reaching over 100 kg ha -1 N. Nevertheless, when growing on the Planosol, differences between the genotypes Gramafante and Cameroon were observed, with values of 92 and 162 kg ha -1 of accumulated N, respectively.For the second cut, the genotypes Roxo and Gramafante showed the lowest N accumulation for the two soils, signifi cantly lower than the other genotypes.At the Table 2. Dry-matter production of elephant grass genotypes, grown in two soils and subjected to three cuts, over a 22-month period (1) . ( Means followed by the same letters, in the columns, do not differ by Tukey's test at 5% probability.Pesq. agropec. bras., Brasília, v.44, n.2, p.133-140, fev. 2009 third cut, there was no statistical signifi cance between genotypes in either soil.Considering the three cuts, on average, 296 and 292 kg ha -1 N were accumulated by the plants in 22 months, when grown in the Acrisol and Planosol, respectively.Cameroon, Bag 02 and CNPGL F06-3 accumulated more N after the three cuts, in both Acrisol and Planosol.Quesada (2005) worked with various genotypes of elephant grass, including those used in the present study, obtained similar results and recorded N accumulations of 270 kg ha -1 N, in 15 months, without N fertilization.
These results are very promising, especially the high dry matter yield produced without the application of N fertilizer, which can represent more than 20% of all energy expenditure in biomass production of elephant grass, for direct burning purposes (Samson et al., 2005).
The high N accumulation indicates that BNF must be a key process for elephant grass cropping systems.The 15 N abundance (δ 15 N) found in elephant grass were lower than in non-N 2 fi xing weeds, in all cases (Table 4), which confi rms that the supply of N from BNF was appreciable.No signifi cant differences between the 15 N natural abundance of the different weed reference species were recorded.Thus, an average of the δ 15 N values from the three control species sampled was considered to estimate the BNF to the elephant grass genotypes.The averaged δ 15 N values of the reference species for the Acrisol were 6.53, 6.99 and 7.43 deltas for the 1 st , 2 nd and 3 rd cuts, respectively.For the Planosol, the δ 15 N values were 7.89, 7.85 and 6.25 for the three cuts, respectively.
For the three cuts, there were statistically signifi cant differences in BNF contributions for the plants.On the Acrisol, at the fi rst cut, Roxo genotype presented 21% of its total N derived from BNF, the lowest contribution when compared to the others (Table 5).On Planosol, the reliance on BNF by Roxo was 30%, again the lowest observed values, but only statistically different from CNPGL F06-3, which was of 52%.For the second cut, the results changed slightly on the Acrisol, where Roxo presented an increased BNF reliance (43%), compared to Bag 02 (29%), but not different from Cameroon (43%).However, on Planosol, which is usually more N defi cient due to its sandy texture, genotype Roxo fi gured again among the genotypes with the lowest BNF reliance (35%), being signifi cantly different from genotype Cameroon (47%).In the last cut, this same trend was observed for the two soils, with the lowest BNF reliance registered for genotype Roxo and the highest for genotype Cameroon.Table 3. Nitrogen accumulation of elephant grass genotypes, grown in two soils and subjected to three cuts, over a 22-month period (1) .
(1) Means followed by the same letters, in the columns, do not differ by Tukey's test at 5% probability.
(1) Means followed by the same letters, in the columns, do not differ by Tukey's test at 5% probability. . agropec. bras., Brasília, v.44, n.2, p.133-140, fev. 2009 Considering the results of all the three cuts, genotype Roxo stood out as the one that was least benefi ted from BNF, and genotype Cameroon as the most.Bag 02 and CNPGL F06-3 genotypes were intermediate.The largest contributions of BNF were observed in the Acrisol.These results agree with Quesada (2005) and Samson et al. (2005), who found BNF contributions of 58, 70 and 68% for the genotypes Cameroon, Bag 02, and Gramafante, respectively.They concluded that BNF presented the lower potential in Roxo genotype.There are few fi eld studies showing that BNF can contribute to the N nutrition of elephant grass.However, the presence of N 2 fi xing bacteria of the genus Herbaspirillum inside this grass, detected by Kirchhof et al. (2001), reinforces the possibility that BNF is an important process to the survival and productivity of this species.The importance of BNF to other grass species was already shown, with contributions of BNF of 30 to 40% registered for Brachiaria (Boddey & Victoria, 1986) and ecotypes of Panicum maximum (Miranda & Boddey, 1987), respectively.In sugarcane, Urquiaga et al. (1992) found BNF could contribute up to 70% of the plant N needs, but the expected contribution of this process for sugarcane growing in farmer fi elds are of the order of 30%, ranging from 0 to 60% (Boddey et al., 2001).The importance of this natural process must be highlighted as each 100 kg ha -1 of N fertilizer that is substituted by BNF means a mitigation of about 450 kg ha -1 CO 2 , considering the phases of processing, transportation and fertiliser application (Robertson & Grace, 2004).

Pesq
Signifi cant interaction between elephant grass genotypes and soil types was also observed for the qualitative parameters C:N and stem:leaf ratios, and fi bre, cellulose, lignin and ash contents (Table 6).All genotypes studied showed a high C:N ratio (above 70), especially the genotype Gramafante, which presented a C:N ratio 37% above average, which could be explained by its high stem:leaf ratio and low N tissue content.In both soils, the C:N ratio of genotype Gramafante (118 to 121) was signifi cantly higher than in the other genotypes, except for genotype CNPGL F06-3.The stem:leaf ratio was not statistically different among genotypes, when plants were grown on the Acrisol, and the observed values ranged from 2.5, for genotype Gramafante, to 1.6, for genotype Roxo.Again, in the Planosol, the genotype Roxo presented the lowest stem:leaf ratio, on average 45% lower than that of the other genotypes.According to Ferrari Júnior & Lavezzo (2001) and Queiroz Filho (1998, 2000), the shorter the time between cuts, the greater is the proportion of leaves in the produced biomass.In the case of animal feeding forage, silage or grazing, the recommendation is the short-time cutting intervals (1 to 2 months), which also end up with a plant material with increased protein levels (Andrade et al., 2005;Fagundes et al., 2007).For energy purposes, however, the strategy should be the opposite, as fi brous and low protein plant materials promote the most effi cient heat production.
Few studies have reported the biomass quality of elephant grass for energy purposes.Quesada (2005)  5. Nitrogen-15 natural abundance and the estimates of the proportional contribution of the biological nitrogen fi xation (BNF) to the nutrition of elephant grass genotypes grown in two soils and subjected to three cuts, over a 22-month period (1) .
(1) Means followed by the same letters, in the columns, do not differ by Tukey's test at 5% probability.
found C:N ratio of about 95, when genotypes were grown on a Planosol, in the rainy season, and 55, when grown during the dry season.
The contents of acid detergent fi bre (ADF) were not infl uenced by soil type, and all genotypes showed levels above 40% (Table 6).In studies with elephant grass grown for fodder purposes, the fi bre content of the whole plant increases and protein content decreases, the longer the plant remains in the fi eld (Andrade et al., 2005).The levels of fi bre in the present study were similar to those reported by Savioli et al. (2000) and Campos et al. (2002), which recorded values close to 42%, and by Queiroz Filho et al. (2000), which reported values up to 48%, for plants growing for 100 days in the fi eld.When grown on the Acrisol, all genotypes showed similar lignin content, except for genotype Cameroon, which was signifi cantly lower (9.7%) in this parameter than the genotype Bag 02 (13.5%).When grown on the Planosol, no statistical difference was detected.According to McKendry et al. (2002), the levels of lignin and fi bre observed in the genotypes used in the present study are considered satisfactory for energy production by direct combustion.
There was no statistically difference for cellulose contents in genotypes grown on the Acrisol, with values ranging from 22.9% (genotype Roxo) to 33% (genotype Bag 02).A small difference was detected on plants grown on the Planosol, with genotype Cameroon presenting the highest content (26.2%) and genotype BAG 02 (29.6%) the lowest.The contents of ash were more variable among genotypes, ranging from 1.9 to 3% when grown on Acrisol, the lowest value observed in the genotype CNPGL F06-3 and the largest in genotype Cameroon.On Planosol, ash contents varied from 2%, for genotype Roxo, to 3.3%, measured in genotype Cameroon.
Although signifi cant differences were observed among treatments, the levels of fi bre, lignin, cellulose and ash were still adequate for heat production, since the maximum ash level considered critical for the calorifi c value of the biomass is 5% (McKendry, 2002;Kauter et al., 2003), high above the results found in the present work.
The results obtained so far, along with the work of Quesada (2001,2005), indicate that this species has good adaptability to soils with low fertility, which is partially sustained by the signifi cant contribution of BNF observed.Even though the BNF process could result in an apparent elephant grass independence of N fertilizer, more than 50% of the N required by the plant growth came from the soil.Thereby, the high CO 2 mitigation potential of elephant grass production systems, shown in the present study, is only sustainable, in the long term, if some other nitrogen source is introduced.

Conclusions
1.The genotypes Cameroon, CNPGL F 06-3 and Bag 02 are the most promising for bioenergy production purposes.
2. The genotypes present lignin, cellulose and ash contents suitable for use in direct burning.
3. The reliance in biological nitrogen fi xation of Pennisetum purpureum varies with soil and genotypes, and can reach up to 50% of this species N needs.Table 6.Carbon:nitrogen (C:N) ratio, stem:leaf ratio, concentrations of acid detergent fi bre (ADF), cellulose, lignin, and ash contents of elephant grass genotypes grown in two soils and subjected to three cuts, over a 22-month period (1) .
(1) Means followed by the same letters, in the columns, do not differ by Tukey's test at 5% probability.

Table 1 .
Soil chemical fertility parameters of the two soils.