- Email: [email protected]
A new algorithm to characterize biodegradability of biomass during anaerobic digestion: Influence of lignin concentration on methane production potential
Bioresource Technology 102 (2011) 9395–9402
Contents lists available at SciVerse ScienceDirect
Bioresource Technology journal homepage: www.elsevier.com/locate/biortech
A new algorithm to characterize biodegradability of biomass during anaerobic digestion: Influence of lignin concentration on methane production potential Jin M. Triolo a,⇑, Sven G. Sommer a, Henrik B. Møller b, Martin R. Weisbjerg c, Xin Y. Jiang d a University of Southern Denmark, Faculty of Engineering, Institute of Chemical Engineering, Biotechnology and Environmental Technology, Niels Bohrs Allé 1, DK-5230 Odense M, Denmark b Aarhus University, Faculty of Agricultural Sciences, Department of Biosystems Engineering, Research Centre Foulum, P.O. Box 50, DK-8830 Tjele, Denmark c Aarhus University, Faculty of Agricultural Sciences, Department of Animal Health and Bioscience, P.O. Box 50, DK-8830 Tjele, Denmark d Central South University of Forestry and Technology, College of Materials Science and Technology, Changsha 410004, Hunan, People’s Republic of China
a r t i c l e
i n f o
Article history: Received 14 March 2011 Received in revised form 29 June 2011 Accepted 11 July 2011 Available online 20 July 2011 Keywords: Biogas Digestibility Lignin Lignocellulose Predicting methane production potential
a b s t r a c t We examined the influence of fibrous fractions of biomass on biochemical methane potential (BMP) with the objective of developing an economical and easy-to-use statistical model to predict BMP, and hence the biodegradability of organic material (BD) for biogas production. The model was developed either for energy crops (grass, maize, and straw) or for animal manures, or as a combined model for these two biomass groups. It was found that lignin concentration in volatile solids (VS) was the strongest predictor of BMP for all the biomass samples. The square of the sample correlation coefficient (R2) from the BMP versus lignin was 0.908 (p < 0.0001), 0.763 (p < 0.001) and 0.883 (p < 0.001) for animal manure, energy crops and the combined model, respectively. Validation of the combined model was carried out using 65 datasets from the literature. Ó 2011 Elsevier Ltd. All rights reserved.
1. Introduction Biogas has been used as a source of renewable energy for more than 100 years. Its production has also played a valuable role in the sustainable management of agricultural byproducts, including animal manure. In Denmark, approximately 5% of animal manure is used for biogas production, and the aim is to use 40% of animal slurry in feedstock-to-biogas digesters by 2020 (Green Growth, 2009). Therefore it is necessary to improve profitability by enhancing methane yield in biogas production. Animal manure contains more readily degradable organic materials, such as proteins and lipids, than other agricultural byproducts, but it also has a high content of lignocellulose biofibers (40–50% of the total solids; Bruni et al., 2010). Lignocellulose consists mainly of three biopolymers: cellulose, hemicelluloses, and lignin. In lignocellulosic materials, cellulose is physically associated with hemicelluloses, and physically and chemically associated with lignin (Mussatto et al., 2008). Lignin and hemicelluloses are intermeshed and chemically bound through covalent cross-linkages such as ester or ether linkages (Jeffries, 1994). The low biodegradability (BD) of lignocellulose in biogas reactors is due to lignin being nondegradable in anaerobic environments (Mauseth, 1988) because the extracellular
⇑ Corresponding author. Tel.: +45 4117 8867; fax: +45 6550 7354. E-mail address: [email protected] (J.M. Triolo). 0960-8524/$ - see front matter Ó 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.biortech.2011.07.026
enzymes require oxygen to depolymerize. Furthermore, hydrolysis of cellulose in lignocellulosic materials is reduced by lignin and hemicelluloses, since these components act as a protective coat, making the cellulose resistant to enzymatic digestion (Mussatto et al., 2008). Animal manure contains high concentrations of lignin because it consists of residues from feed, where the easily degradable compounds have been taken up by the animals; therefore the biomass excreted contains mainly the slowly degradable components including lignocellulose. Ruminants are particularly efficient in using the carbon components in feed, and therefore excreta from ruminants contain high concentrations of slowly digestible organic matter. Biochemical methane potential (BMP) has been used as the most relevant indicator for assessing BD (Lesteur et al., 2010). BMP cannot be directly related to BD, since BMP is the methane yield, reflecting the destruction of organic materials, and the methane potential of each organic component in the volatile solids (VS) pool varies widely. For example, theoretically, the methane potential of lipid is 1018 L/kg, while the methane potential of cellulose is only 415 L/kg, based on the anaerobic degradation equation suggested by Symons and Buswell (1933). BMP assessed by different researchers and institutes is usually not comparable, due to differences in equipment used, environmental conditions, and experimental protocols (Angelidaki et al., 2009). As BMP data may vary depending on the batch method used, the application of a standardized method is needed. Furthermore, since the current
9396
J.M. Triolo et al. / Bioresource Technology 102 (2011) 9395–9402
methods of determining BMP are costly and time-consuming, innovative techniques for predicting BMP or BD are of great importance. Several alternate methods of estimating BD have been suggested (Lesteur et al., 2010) and numerous studies have attempted to predict BD or BMP by measuring the content of organic components (Tong et al., 1990; Gunaseelan, 2007). Near infrared spectroscopy (NIRS) has been used to predict organic matter components (Bruun et al., 2005; Michel and Ludwig, 2010). Recently the BMP of municipal solid waste has been shown to correlate well with NIRS results (Lesteur et al., 2011). Few studies have focused on lignin as the predominant variable to predict BD or BMP for biogas production. Chandler et al. (1980) observed a strong correlation between biodegradable fraction and lignin content in VS (R2 = 0.94) from diverse organic wastes, and Gunaseelan (2007) reported a weak relationship between the BMP and lignin content of several fractions of fruit and vegetable solids (R2 = 0.49). It was suggested that the low correlation could be due to the narrow range of lignin content in the fruits that were tested, making it impossible to reveal a significant effect of lignin. To our knowledge there is no satisfactory model for predicting BMP using lignin as a predominant variable that can be used for animal slurries, energy crops, and a mixture of both. The objective of this study was to examine the influence of lignin on BMP in energy crops and manure and to construct a statistical model to predict BMP and BD. The hypothesis was that lignin concentration can be used to assess BMP, and further to predict BD. The model, and the analytical procedures needed to provide data for the model, must be a cheap and fast alternative to existing methods.
2. Methods 2.1. Substrates and inoculum used Both animal slurries and energy crops were included in the study. Fresh pig and cattle slurry samples were taken from the pre-storage tanks of 10 farms in Horsens, Denmark. Grass, maize and straw were collected from farms in central Jutland. Table 1 shows the energy crops included in the study, with harvest date, pretreatment, and particle size. Representative subsamples for characterization and for the fermentation study were stored at 18 °C. The inocula used for the BMP assay were collected from two biogas plants at Fangel and Foulum. The Fangel biogas plant is operated under mesophilic conditions (37 °C), processing pig manure from 26 animal farms mixed with industrial organic waste. The Foulum biogas plant processes cattle manure mixed with crop residues under thermophilic conditions (55 °C). Inoculum from Foulum was taken from the post storage which is running under mesophilic conditions. The inocula were degassed at 37 °C for
14 days before application. The average pH and volatile solid of inocula were 8.0 and 66 (% of dry matter), respectively. The average methane concentration of biogas released from inocula was 68.2%. The pH of animal manure ranged between 7.3 and 7.7.
2.2. BMP assay The BMP of the animal manure and energy crops were determined using a batch technique based on methods described by Møller et al. (2004). 1100-mL infusion bottles were used as the batch digesters. Headspace was set to 30%. Inoculum to substrate was at unity or very close to unity on a VS basis (1:1). Blanks were tested using 770 g of inoculum to correct gas production. After addition of the substrate inoculum mixture, the digesters were closed with butyl rubber stoppers, sealed with aluminum crimps, flushed with N2 atmosphere and incubated at 37(±0.5) °C. All assays were performed in triplicate. Gas volume was measured either by replacing water or using a large syringe, as described by Steed and Hashimoto (1994). Digestion was continued until no further gas production was observed (90 days). Each batch digester was mixed thoroughly by shaking to prevent dry layers and to encourage degassing on workday. Gas volumes were measured every day at the beginning of fermentation and then gradually at larger time intervals. Methane and carbon dioxide were determined simultaneously once a week by a gas chromatograph (HP 6890 series), equipped with a thermal conductivity detector and a 30 m 0.320 mm column (J&W 113-4332). The carrier gas was helium (30 cm/s), and injection volume was 0.4 mL. Injector temperature was 110 °C, and detector and oven temperature was 250 °C. The split rate was 1:100. Methane quantification was evaluated according to VDI 4630 (2006). In detail, gas volume as read off was corrected as dry gas flow and as STP conditions (273 K, 1.013 bar). For quantification of methane concentration, simultaneously measured methane and carbon dioxide concentrations were multiplied by the same factor, being the sum of the corrected measured values as 100%, assuming that the fractions of ammonia and hydrogen sulfide are insignificant quantities (Eq. (1)).
C dry Cor ¼ C CH4
100 ðC CH4 þ C CO2 Þ
ð1Þ
where C dry Cor is corrected concentration of methane in the dry gas, (%), C CH4 is measured concentration of methane in the gas (%) and C CO2 is measured concentration of carbon dioxide in the gas (%). Methane volumes were calculated using corrected dry gas volume and corrected methane concentration. The methane volumes only from the substrate were calculated by subtracting the mean value of the inoculum control.
Table 1 The energy crops included in the study, with harvest date, pretreatment, and particle size. No.
Crop
Cultivar
Pretreatment
Harvest date
Particle size (mm)
1 2 3 4 5 6 7 8 9 10
Perennial grass Perennial grass Dried grass Grass Grass Grass Maize Maize Maize Dried straw
Mixed wild types Mixed wild types Mixed wild types Festulolium + 20% red clover, Hykor and Amos Festulolium, Achilles Festulolium + 20% red clover, Achilles and Amos Anvil Patrick Aurelia Wheat straw
Coarse cut Coarse cut Coarse cut Coarse cut Coarse cut Coarse cut None, cut during harvest None, cut during harvest None, cut during harvest Coarse cut
14 June 14 September NA 22 October 7 August 8 October 15 October 15 October 15 October NA
10–15 10–15 10–15 10–15 10–15 10–15 5 5 5 10–15
NA: not available.
9397
J.M. Triolo et al. / Bioresource Technology 102 (2011) 9395–9402
where Oi is the measured value, Pi is the predicted value, and O is the mean of the measured values.
2.3. Substrate analysis Prior to the BMP test, biochemical and physiochemical analyses were carried out. Dry matter (DM), volatile solids (VS), crude lipid, total ammoniacal nitrogen (TAN), and total Kjeldahl nitrogen (TKN) were determined according to standard procedures (APHA Standard Method, 2005). Volatile fatty acids (VFA) were determined according to the method of Lahav et al. (2002). Neutral detergent fibers (aNDF) were determined by a-amylase neutral detergent extraction (Mertens et al., 2002). Acid detergent fiber (ADF) and acid detergent lignin (ADL) were determined ash free by acid detergent extraction as described in the ISO Standard (ISO 13906; ISO, 2009). 2.4. Data analysis 2.4.1. Experimental data analysis Organic nitrogen (Norg) was calculated as the difference between TKN and TAN. Crude protein was determined by multiplying Norg by 6.25. Different fibrous fractions were determined in accordance with Van Soest’s characterization for fiber analysis (Van Soest, 1963; Goering and Van Soest, 1970), which enables the differentiation of fiber fractions using specific detergents. The aNDF treatment was used to determine total cell wall components, including hemicelluloses, cellulose, lignin, and fiber-bound proteins. ADF mainly consists of cellulose, lignin, and insoluble proteins that are components in the cell walls. The amounts of fibrous fractions were assessed as follows: lignocelluloses were assumed to be the aNDF fraction. Cellulose was determined by calculating the difference between ADF and ADL, and hemicelluloses as the difference between ADF and aNDF. Lignin is identified with ADL, with the assumption that the fraction of lignin-bound nitrogen is insignificant. 2.4.2. Statistical analysis Statistical data analysis was performed using the SAS software package (SAS Institute, 1992). BMP was tested as the dependent variable against all the fiber fractions. Simple linear regression and successive stepwise regression analyses were performed with statistically significant variables. In the statistical analysis, regression models to predict BMP for both energy crops and animal manure were developed. Furthermore regression models encompassing both types of biomass were developed. Relative root mean square error (RRMSE) was used to assess the accuracy of the model (Eq. (3)). The models were validated using scatter plots of predicted BMP versus measured BMP. Statistical indices proposed by Loague and Green (1991) were incorporated into the model. The final model included RRMSE, modeling efficiency (EF), and coefficient of residual mass (CRM), according to the following equations.
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Pn 2 i¼1 ðP i Oi Þ RMSE ¼ n
ð2Þ
where Oi is the measured value, Pi is the predicted value, and n is the number of data points.
RRMSE ¼
RMSE O
100
ð3Þ
where O is the mean of the measured values.
Pn
i¼1 Oi
CRM ¼
Pn
Pn
i¼1 P i
i¼1 Oi
ð4Þ
where Oi is the measured value and Pi is the predicted value.
Pn EF ¼
i¼1 ðOi
P OÞ2 ni¼1 ðPi Oi Þ2 Pn 2 i¼1 ðOi OÞ
ð5Þ
2.4.3. Theoretical BMP Theoretical BMP (TBMP) has been used to assess methane potential in several studies (Møller et al., 2004; Chae et al., 2008). BMP will never account for all of the TBMP, even if we include the 7% of VS that is used for bacterial growth (i.e., accounted as C5H9O3N; see Kalyuzhnyi, 1997). In the present study, we calculated TBMP with the method used by Møller et al. (2004) with some slight modifications. Lignin was included with the following empirical formula C10H13O3. Theoretical methane potential from lignin was calculated using the equation provided by Symons and Buswell (1933), which showed a theoretical production of 727.1 (CH4 NL (kg lignin)1) following Eq. (6):
Lignin C10 H13 O3 þ 5:25H2 O ! 5:875CH4 þ 4:125CO2
ð6Þ
Based on the component composition presented by Møller et al. (2004) and the one for lignin in Eq. (6), we propose that TBMP is calculated as follows:
TBMP ¼ ðVFA 373 þ Lipid 1014 þ Protein 496 þ Carbohydrate 415 þ Lignin 727Þ 0:001
ð7Þ
1
with TBMP as CH4 NL (kg VS) , and VFA, lipid, protein, carbohydrate, and lignin as g (kg VS)1. For TBMP calculation, all the data were used from our experiment, except crude fat from energy crop which was referenced by the Nordic standard feed table (NORFOR, 2011). Following the table, crude fat for the grasses (Nos. 1–6) were assumed as 39 g kg1 DM (feed code, 006-0082), for the maize as 22 g kg1 DM (feed code, 006-0307) and straw as 19 g kg1 DM (feed code, 006-0413) for the calculations. 3. Results and discussion The slurries included in this study had a wide range of water contents and VS concentrations (Table 2). On average, pig slurry contained 92.8% water and cattle slurry 80.8%. Some of the pig manures were very dilute and contained up to 99% water. The higher water content in pig manure was probably caused by spillage of cleaning and drinking water in the animal houses. On the other hand, the calf manure contained little water and had been collected in the form of semi-solids. The low water content of calf manure was due to a large fraction of straw used as bedding materials in the calf houses. VS concentrations were on average 74.9% of DM for pig manure and 80.2% of DM for cow manure. TAN in the cow manure ranged from 1.65 to 2.19 g kg1, while TAN in the pig manure varied more, from 1.13 to 5.63 g kg1. TKN ranged from 3.07 to 4.68 g kg1 for the cow manure, and from 1.33 to 6.87 g kg1 for the pig manure. In most studies, organic nitrogen in cattle manure is considerably higher than in pig manure, and most of the nitrogen in pig manure is in the form of TAN. The higher organic nitrogen in cattle manure is because most nitrogen is excreted in feces and this nitrogen is mainly of bacterial origin (70– 85%; Mosenthin et al., 1994). Nitrogen excreted in urine is mainly in the form of urea, which is more susceptible to decomposition than bacterial nitrogen excreted in feces (Mroz et al., 1993; Sommer et al., 2006). The crops tested were maize (Zea spp.), a mixture of perennial grasses and festulolium (a cross between Festuca spp. (meadow fescue/tall fescue) and Lolium multiflorum (Italian ryegrass)). Details of the crops are given in Table 3. TKN ranged from 0.99 to 2.02 g kg1 DM in the maize, 1.96 to 2.56 g kg1 DM in the festulolium, and from 1.92 to 2.78 g kg1 DM in the perennial grasses. The TAN of the crops was below the detection limit of the measuring method.
No.
1 2 3 4 5 6 7 8 9 10
Manure type
Piglet manure Sow manure Sow manure Pig fatteners Pig fatteners Calf manure Calf manure Cattle manure Cattle manure Inoculum
9398
Table 2 Composition, BMP and TBMP of the manure samples used in the study. DM
VS
VFA
XL
TKN
TAN
Lignocellulose
Hemicelluloses
Cellulose
Lignin
BMP
TBMP
CH4 content
(g kg1 w/w)
(g kg1 w/w)
(mM)
(g kg1 DM)
(g kg1 w/w)
(g kg1 w/w)
% of DM
% of DM
% of DM
% of DM
CH4 NL (kg VS1)
CH4 NL (kg VS1)
% of dry gas
16.2(0.3) 49.0(1.2) 109.4(6.5) 119.6(1.0) 9.4(0.1) 154.6(3.3) 322.6(2.7) 115.6(0.9) 173.6(0.5) 34.7(0.5)
11.8(0.1) 38.2(0.4) 90.2(0.3) 100.6(0.5) 5.2(0.1) 142.3(0.1) 295.2(1.0) 94.3(0.6) 96.8(0.6) 22.9(0.4)
109.5(3.8) 75.1(1.4) 170.6(3.7) 275.8(2.2) 28.6(0.7) 115.6(4.1) 453.6(17.8) 180.6(2.5) 166.4(2.5) 12.7(1.0)
60.4(3.1) 107.4(17.5) 120.5(12.0) 121.5(10.5) 62.5(6.9) 10.89(0.25) 15.54(2.05) 74.7(2.3) 57.6(3.4) 8.1(2.4)
1.33(0.24) 3.94(0.03) 4.87(0.28) 6.87(0.28) 1.64(0.10) 3.07(0.24) 4.57(0.30) 4.68(0.11) 4.38(0.13) 4.74(0.20)
1.13(0.04) 3.71(0.04) 3.56(0.03) 5.63(0.04) 1.51(0.03) 1.65(0.01) 2.19(0.25) 1.90(0.06) 1.82(0.03) 3.90(0.07)
5.11 47.11(2.11) 47.59(0.62) 47.10(0.96) NA 71.18(0.75) 72.92(0.04) 36.46(0.48) 25.63(0.42) 25.27(0.38)
0.97(0.22) 15.36(1.09) 14.79(2.68) 15.65(0.55) NA 17.74(2.58) 22.24(1.65) 7.97(1.55) 3.56(0.59) 3.41(1.17)
3.76(0.02) 21.41(0.94) 23.22(1.79) 24.16(4.47) 1.15(0.26) 42.78(1.76) 42.97(1.64) 16.58(1.36) 15.58(0.20) 10.65(3.32)
0.38(0.25) 10.34(0.08) 9.58(0.18) 7.30(4.88) 0.51(0.13) 10.65(0.07) 7.76(1.05) 11.91(2.43) 6.49(0.37) 11.21(2.39)
417.2(11.8) 213.8(9.2) 248.8(16.3) 284.8(3.8) 345.3(3.7) 198.7(3.6) 237.0(7.8) 197.0(3.3) 223.6(3.7) 142.1(2.5)
449.6 537.5 531.0 527.8 482.2 496.8 442.7 525.1 522.7 492.7
66.4(1.3) 62.8(0.8) 60.8(0.2) 63.9(0.3) 69.1(3.6) 57.9(1.5) 58.4(0.5) 61.1(0.3) 61.7(0.3) 67.7(0.9)
Figures in parentheses are standard deviations. XL: crude lipid; NL: norm liter (273 K, 1.013 bar); NA: not available.
No.
Biomass type
DM (g kg1 w/w)
VS (g kg1 w/w)
TKN (g kg1 w/w)
Lignocellulose % of DM
Hemicelluloses % of DM
Cellulose % of DM
Lignin % of DM
BMP CH4 NL (kg VS1)
TBMP CH4 NL (kg VS1)
CH4 content % of dry gas
1 2 3 4 5 6 7 8 9 10
Perennial grass Perennial grass Dried wild grass Grass Grass Grass Maize Maize Maize Straw
288.2 235 934.1(0.7) 187.4 180.2 175.3 387.5 299.2 278.9 936.9(0.1)
77.2 52 880.2(1.1) 32.3 28.9 27.7 145.5 85.6 74.5 906.3(1.9)
5.53 6.54 10.82(1.2) 4.8 3.55 3.53 4.84 2.96 5.62 3.45(0.7)
60.79(2.08) 58.33(0.54) 64.85(0.17) 59.54(0.30) 47.72(0.18) 51.73(0.57) 63.82(1.61) 65.77(1.58) 60.86(2.31) 83.76(0.14)
25.99(1.95) 28.25(0.64) 22.89(0.44) 27.30(0.56) 20.93(0.74) 24.39(0.66) 37.68(1.70) 35.92(2.51) 36.73(2.85) 28.38(1.32)
29.05(0.81) 26.86(0.08) 35.99(0.02) 29.97(0.72) 25.56(0.09) 25.38(0.33) 23.74(2.14) 26.86(3.91) 22.31(0.51) 47.97(1.64)
5.75(0.68) 3.22(0.03) 5.98(0.59) 2.27(0.44) 1.23(0.47) 1.96(0.25) 2.40(2.21) 2.98(0.18) 1.81(1.05) 6.41(0.18)
271.0(8.5) 410.5(67.8) 306.3(7.9) 373.5(35.7) 420.8(22.2) 438.9(21.8) 399.4(21.3) 405.3(20.3) 360.5(24.8) 289.5(17.5)
466.3 462.4 461.4 460.7 453.9 455.7 452.3 443.8 445.2 447.7
53.8 56.0 62.5(0.1) 58.6 55.0 54.2 59.0 57.9 59.0 62.2(0.0)
Figures in parentheses are standard deviations. NL: norm liter (273 K, 1.013 bar).
J.M. Triolo et al. / Bioresource Technology 102 (2011) 9395–9402
Table 3 Composition, BMP and TBMP of the energy crops used in the study.
J.M. Triolo et al. / Bioresource Technology 102 (2011) 9395–9402
9399
Fig. 1. Lignocellulose fraction in VS in animal manures and energy crops.
Fig. 2. Distribution of hemicelluloses, cellulose, and lignin in lignocellulose.
For most of the animal manure samples, the lignocellulose content was between 40% and 80% of VS, except for the piglet manure, where lignocellulose content was 7.0% of VS (Fig. 1). The low lignocellulose content of piglet manure was due to the piglet diet having a very low fiber content. The highest lignocellulose contents were found in calf manure, which was due to calf bedding materials. For adult animals, the concentration of lignocellulose was approximately 50% of VS for both cow and pig manure. In manure the content of hemicelluloses was 26.5(±8.1)% of VS, of cellulose 55.6(±12.5)% of VS, and of lignin 17.9(±8.7)% of VS (Fig. 2). For the energy crops, the content of hemicelluloses, cellulose, and lignin were 45.3(±11.9)% of VS, 48.9(±9.9)% of VS, and 5.8(±2.0)% of VS, respectively. Comparing the two types of biomass, the hemicelluloses content was lower in manure than in the crops, while the lignin content was higher. During digestion, the relatively easily degradable hemicelluloses seem to be transformed and utilized by the animals more readily than cellulose. Lignin is a larger fraction in manure than energy crops, as lignin is upconcentrated by animal digestion. Cumulative methane production of pig manure, cattle manure and crop as a function of time are shown respectively in Fig. 3. For the first 2 weeks the great majority of methane gas was produced, and thereafter only small amounts of gas were released.
Fig. 3. Cumulative methane production from pig manure, cattle manure and energy crops.
The methane concentrations in the dry gas of each substrate were ranged from 53.8% to 62.5% for energy crops. For animal manure it was slightly higher, ranging from 57.9% to 69.1%. Cumulative methane production (90 days) used as BMP ranged from 197 to 417 CH4 NL (kg VS)1 for animal manure, and for energy crops 271–439 CH4 NL (kg VS)1. The results of TBMP are presented in Tables 2 and 3. As shown in the tables, TBMP was less variable than BMP, ranging from 443 to 538 CH4 NL (kg VS)1 for animal manure, 444–466 CH4 NL (kg VS)1 for crop. 3.1. Stepwise regression analysis 3.1.1. Lignin Using the single variables, the most significant BMP models were predicted by lignin for all the categories (Table 4). The square of the sample correlation coefficient (R2) between BMP and lignin
9400
J.M. Triolo et al. / Bioresource Technology 102 (2011) 9395–9402
Table 4 Summary statistics of statistical analyses for energy crops and animal manure. Variable
R2
p
RRMSE (%)
Equation
Energy crop
Lignin Cellulose ADF NDF Lignin, cellulose
0.763 0.407 0.511 0.435 0.766
<0.001 <0.05 <0.05 <0.05 <0.01
7.4 11.7 10.7 11.5 6.1
BMP = 2.58 lignin + 460.6 BMP = 0.49 cellulose + 521.5 BMP = 0.45 ADF + 668.1 BMP = 0.46 NDF + 447.1 BMP = 2.77 lignin 0.07 cellulose + 447.1
Manure
Lignin Cellulose ADF NDF Lignin, cellulose
0.908 0.485 0.695 0.491 0.941
<0.0001 <0.05 <0.01 <0.05 <0.0001
9.1 21.7 16.7 20.7 7.4
BMP = 1.40 lignin + 388.9 BMP = 0.36 cellulose + 341.8 BMP = 0.34 ADF + 373.3 BMP = 0.25 NDF + 371.3 BMP = 1.22 lignin 0.11 cellulose + 399.4
Combined model
Lignin Cellulose ADF Lignin, cellulose
0.883 0.106 0.358 0.885
<0.001 >0.05 <0.01 <0.001
9.8 NA 22.9 7.4
BMP = 1.675 lignin + 421.7 NA BMP = 0.373 ADF + 439.6 BMP = 1.649 lignin 0.035 cellulose + 430.0
NA: not available.
was 0.908 for animal manure and 0.763 for the energy crops. Relative error (RRMSE) of the model was below 10% when lignin was chosen as the independent variable, indicating that the models relating BMP to lignin predicted BMP well. RRMSE shows slightly better model accuracy for energy crops (RRMSE = 7.4%) than for animal manure (RRMSE = 9.1%). Moreover, the slope of the regression models using lignin was more negative for energy crops (2.58) than for animal manure (1.40), which shows that lignin affects BMP more significantly in energy crops than in animal manure. The slope of BMP to lignin, which we call lignin dependency, is much higher for crop residues than for manure, because the cell wall is protected by an intact lignocellulose matrix in crops, whereas in animal manure part of this matrix is broken down inside the animal during digestion. The higher statistical significance in contrast to the lower model accuracy for the manure model compared with the crop model is probably caused by the wider range of values of the independent variables of manure data, which improve the correlation level. For example, the lignin concentration of samples used ranged from 5.2 to 169.8 g (kg VS)1 in animal manure, whereas lignin in crop residue only varied between 13.6 and 66.2 g (kg VS)1. In the combined model, the correlation between lignin concentration and BMP was good (R2 = 0.883). For both crops and animal manure, the lignin concentrations were found in clusters of high and low lignin content. Two of the animal manures were so different from the rest that they may be defined as outliers. However, when these two outliers are included in the statistical analysis they contribute positively to the correlation between BMP and lignin and improve the regression models. However, the accuracy of the model decreased (RRMSE = 9.8%) when the individual models were combined. The results show that the influence of lignin concentration in the biomass is similar between the two categories of biomass, but the best model for prediction of BMP is the one that accounts for the different origin of the biomass. This indicate a need to define a variable and parameters that provide a method to develop one model; i.e., an indicator of the destruction of the physical lignocellulose matrix during the digestion of feed in animals.
show a significant relationship between cellulose and BMP. Furthermore, no significant relation was found between hemicelluloses and BMP in any of the tested models. It is interesting that BMP seems to be related only to the content of lignin, and the relations between BMP and other fibers such as cellulose and hemicelluloses are insignificant, depending on the model category. This suggests that cellulose and hemicelluloses are greatly affected by the lignin barrier, and consequently the destruction rate of these fibers differs between sample categories as affected by the treatment before BMP measurements. The result confirms that lignin controls VS destruction most significantly, which makes it possible to provide fine correlations between BMP and lignin. Since lignin is not degradable, BMP will be negatively correlated with the fraction of lignin in VS. Moreover, the significant relation of BMP to cellulose indicates that degradation of cellulose, which is suppressed by the lignocellulose matrix, affects the destruction rate of VS. Consequently a large amount of cellulose was not transformed to methane during the batch experiment and this is reflected by the BMP. This is also the case with lignin, which links to cellulose and creates barriers for microbial degradation not only of lignin but also of cellulose. Therefore, cellulose in lignocellulosic materials may not significantly contribute to BMP.
3.1.2. Cellulose and hemicelluloses BMP was significantly (p < 0.05) related to cellulose for the two individual models, but the correlation was inferior to the relation between BMP and lignin. R2 values for cellulose were much lower than those for the lignin models (R2 = 0.485 for animal manure, R2 = 0.407 for energy crops). The relatively high RRMSE values indicate low model efficiency when cellulose was used as a single variable. Moreover, a combined model for crops and manure does not
A BD model assessed as the ratio between BMP to TBMP versus lignin together with a BMP model is shown in Fig. 4. R2 between BMP/TBMP was slightly higher than the BMP model as 0.775 (RRMSE = 7.4) for energy crop, 0.922 (RRMSE = 9.8) for animal manure and 0.893 (RRMSE = 10.7) for the combined model. The intercept of the regression line was very close to 1(1.0197) for energy crop and for animal manure lower as 0.8322. This result indicates that only lignin is a nondegradable fraction for energy crop,
3.1.3. Multiple regression tests In the multiple regression tests, lignin and cellulose were applied as two independent variables and R2 was slightly improved from 0.763 to 0.766 (energy crops), 0.908 to 0.941 (animal manure), and 0.883 to 0.885 (combined model) compared with the models where only lignin was used as the independent variable (Table 4). Relative errors were somewhat reduced, when lignin and cellulose were employed; that is, RRMSE was 6.1% (energy crops), 7.4% (animal manure), and 7.4% (combined model). This unsatisfactory improvement from including cellulose in the regression equation was due to the interrelated effect of lignin and cellulose on BMP. 3.2. Biodegradability influenced by lignin
9401
J.M. Triolo et al. / Bioresource Technology 102 (2011) 9395–9402
Table 5 The datasets used for validation of the models and their regression linear trend of the measured BMP versus lignin content in VS. References
R2
Slope
Intercept
Biomass used
Amon et al. (2007) Gunaseelan (2007) Møller et al. (2004) Oslaj et al. (2010)
0.7715
1.19
373.7
0.5002
1.38
493.1
0.6800
3.26
489.1
0.0021
NA
NA
n = 16 (maize 10; dairy cow manure 6) n = 24 (fruit and vegetable waste 17; sorghum and Napier grass 7) n = 10 (cattle manure 2; pig fattener manure 4; sow manure 3; straw 1) n = 10 (maize 10)
NA: not available.
Fig. 4. Effect of lignin concentration on BMP and BD (BMP/TBMP): A, as regression line of BMP for energy crops (y = 2.58x + 460.6); B, for animal manure (y = 1.40x + 388.9); C, for the two categories combined (y = 1.675x + 421.7); A0 , as regression line of BD for energy crop (y = 0.0059x + 1.0197); B0 , for animal manure (y = 0.0033x + 0.8322); C0 , for the two categories combined (y = 0.0041x + 0.9331).
while the lower intercept for animal manure probably shows that BD of animal manure was affected to some extent by recalcitrant fractions, which reduce BD of animal manure. Therefore, we may conclude BMP or BD could be more effectively predictable using lignin for energy crop, since in energy crop crude protein and crude fat are low, and that is critical for BMP.
3.3. Model validation We tested 65 datasets from the literature to validate the combined model (see Table 5). Most of the datasets did not include the effect of lignocelluloses on BMP and statistical analyses of BMP versus lignin had not been performed. Gunaseelan (2007) carried out a statistical analysis, but found no significant relation of BMP to lignin for the two biomass groups, manure and crops. The reason for the non significant results was probably that lignin concentration of the biomass did not have a wide enough range to test the effect of lignin concentration as variables. However, when including all the datasets from Gunaseelan (2007) in the statistical analysis, BMP can be shown to be significantly related to the lignin content of the biomass. The study of Oslaj et al. (2010) indicated a tendency, but no significant relation, of BMP to lignin. This was probably also due to the limited range of lignin concentration and BMP in the data.
Fig. 5. Validation of the suggested model: measured BMP versus predicted BMP and the linear trend.
The predicted values versus measured values of BMP from the various studies were plotted in Fig. 5. The BMP values from Gunaseelan (2007) generally lie below the trend line of predicted BMP, while those of Møller et al. (2004) are higher. This indicates that heavy scattering does not occur only because of model weakness, but also because BMP has been determined using different batch methods. In the BMP results from Amon et al. (2007), we can clearly distinguish between two groups (lower BMP from dairy cow manure, higher BMP from maize biomass); if the model was tested separately for the individual types of biomass, there would probably be no statistical significance. The slop and intercept of the linear regression line was 0.45 and 125, respectively, despite the perfect fit for 1 and 0, respectively. The low slope of the regression line indicates that the suggested model has a tendency to underestimate BMP at high BMP values. Moreover, the weakness of the model was that its intercept is considerably lower (421.7 CH4 NL (kg VS)1) than the theoretical intercept (around 500 CH4 NL (kg VS)1) that is expected for TBMP after subtracting 7% for bacterial growth. For the combined model, RRMSE, CRM, and EF were 26.5%, 0.11%, and 0.21%, respectively. The higher statistical significance in contrast to the lower model accuracy for the manure model compared with the crop model is probably caused by the wide range of concentrations of independent variables of manure data, which improve the correlation level.
9402
J.M. Triolo et al. / Bioresource Technology 102 (2011) 9395–9402
4. Conclusions This study confirmed that the lignin fraction could be used to predict BMP for the combined model for energy crops and animal manures. The BMP model was slightly improved when cellulose was included in the model together with lignin as an independent variable. The validation procedure shows that the model needs to be improved to increase precision of prediction of BMP. Nevertheless, the study shows that BMP had the greatest dependency on lignin concentration, and information on fiber fractions in the biomass could also be used to predict the BMP and BD of biomass. Acknowledgement This study has been supported with a grant from Energi Fyns Udviklingsfond. References Amon, T., Amon, B., Kryvoruchko, V., Zollitsch, W., Mayer, K., Gruber, L., 2007. Biogas production from maize and dairy cattle manure: influence of biomass composition on the methane yield. Bioresour. Technol. 100, 5777–5782. Angelidaki, I., Alves, M., Bolzonella, D., Borzacconi, L., Campos, J.L., Guwy, A.J., Kalyuzhnyi, S., Jenicek, P., van Lier, J.B., 2009. Defining the biomethane potential (BMP) of solid organic wastes and energy crops: a proposed protocol for batch assays. Water Sci. Technol. 59 (5), 927–934. APHA, 2005. Standard Methods for the Examination of Water and Wastewater, 21st ed. American Public Health Association, Washington, DC. Bruni, E., Jensen, A.P., Angelidaki, I., 2010. Steam treatment of digested biofibers for increasing biogas production. Bioresour. Technol. 101, 7668–7671. Bruun, S., Stenberg, B., Breland, T.A., Gudmundsson, J., Henriksen, T.M., Jensen, L.S., Korsæth, A., Luxhøi, J., Pálmason, F., Pedersen, A., Salo, T., 2005. Empirical predictions of C and N mineralization patterns from near infrared spectroscopy, stepwise chemical digestion or C/N ratios. Soil Biol. Biochem. 37, 2283–2296. Chae, K.J., Jang, A., Yim, S.K., Kim, I.S., 2008. The effect of digestion temperature and temperature shock on the biogas yields from the mesophilic anaerobic digestion of swine manure. Bioresour. Technol. 99, 1–6. Chandler, J.A., Jewell, W.J., Gossett, J.M., Van Soest, P.J., Robertson, J.B., 1980. Predicting methane fermentation biodegradability. Biotechnol. Bioeng. Symp. 10, 93–107. Goering, H.K., Van Soest, P.J., 1970. Forage Fiber Analysis (Apparatus, Reagents, Procedures and Some Applications). Agriculture Handbook No. 379. Agriculture Research Service, United States Department of Agriculture, Washington, DC. Green Growth, 2009. Grøn Vækst. Danish Environmental Ministry, April 2009 (in Danish). Available from:. Gunaseelan, V.N., 2007. Regression models of ultimate methane yields of fruits and vegetable solid wastes, sorghum and napiergrass on chemical composition. Bioresour. Technol. 98, 1270–1277. ISO (International Organization for Standardization), 2009. ISO 13906:2008. Animal Feeding Stuffs: Determination of Acid Detergent Fibre (ADF) and Acid Detergent Lignin (ADL) Contents. Available from: . Jeffries, T.W., 1994. Biodegradation of Lignin and Hemicelluloses. Institute for Microbial and Biochemical Technology, USDA Forest Service, Forest Products Laboratory, Madison, WI.
Kalyuzhnyi, S.V., 1997. Batch anaerobic digestion of glucose and its mathematical modeling. II. Description, verification and application of model. Bioresour. Technol. 59, 249–258. Lahav, O., Morgan, B.E., Loewenthal, R.E., 2002. Rapid, simple, and accurate method for measurement of VFA and carbonate alkalinity in anaerobic reactors. Environ. Sci. Technol. 36 (12), 2736–2741. Lesteur, M., Bellon-Maurel, V., Gonzalez, C., Latrille, E., Roger, J.M., Junqua, G., Steyer, J.P., 2010. Alternative methods for determining anaerobic biodegradability: a review. Process Biochem. 45 (4), 431–440. Lesteur, M., Latrille, E., Maurel, V.B., Roger, J.M., Gonzalez, C., Junqua, G., Steyer, J.P., 2011. First step towards a fast analytical method for the determination of Biochemical Methane Potential of solid wastes by near infrared spectroscopy. Bioresour. Technol. 102 (3), 2280–2288. Loague, K., Green, R.E., 1991. Statistical and graphical methods for evaluating solute transport models: overview and application. J. Contam. Hydrol. 7, 51–73. Mauseth, J.D., 1988. Plant Anatomy. The Benjamin/Cummings Publishing Company, CA. Mertens, D. et al., 2002. Gravimetric determination of amylase-treated neutral detergent fiber in feeds with refluxing in beakers or crucibles: collaborative study. J. AOAC Int. 85 (6), 1217–1240. Michel, K., Ludwig, B., 2010. Prediction of model pools for a long-term experiment using near-infrared spectroscopy. J. Plant Nutr. Soil Sci. 173, 55–60. Møller, H.B., Sommer, S.G., Ahring, B.K., 2004. Methane productivity of manure, straw and solid fractions of manure. Biomass Bioenerg. 26, 485–495. Mosenthin, R., Sauer, W.C., Ahrens, F., 1994. Dietary pectin’s effect on ileal and fecal amino acid digestibility and exocrine pancreatic secretions in growing pigs. J. Nutr. 124, 1222–1229. Mroz, Z., Jongbloed, A.W., Beers, S., Kemme, P.A., de Jonge, L., van Berkum, A.K., van der Lee, R.A., 1993. Preliminary studies on excretory patterns of nitrogen and anaerobic deterioration of fecal protein from pigs fed various carbohydrates. In: Verstegen, M.W.A., den Hartog, L.A., van Kempen, G.J.M., Metz, J.H.M. (Eds.), Nitrogen Flow in Pig Production and Environmental Consequences. EAAP Publication No. 69. PUDOC, Wageningen, The Netherlands, pp. 247–252. Mussatto, S.I., Fernandes, M., Milagres, A.M.F., Roberto, I.C., 2008. Effect of hemicellulose and lignin on enzymatic hydrolysis of cellulose from brewer’s spent grain. Enzyme Microb. Technol. 43, 124–129. NORFOR (Nordic Feed Evaluation System) 2011. In: Harald Volden (Ed.), EAAP Scientific Series, ISSN 0071-2477, 130, Wageningen Academic. Available from:
Contents lists available at SciVerse ScienceDirect
Bioresource Technology journal homepage: www.elsevier.com/locate/biortech
A new algorithm to characterize biodegradability of biomass during anaerobic digestion: Influence of lignin concentration on methane production potential Jin M. Triolo a,⇑, Sven G. Sommer a, Henrik B. Møller b, Martin R. Weisbjerg c, Xin Y. Jiang d a University of Southern Denmark, Faculty of Engineering, Institute of Chemical Engineering, Biotechnology and Environmental Technology, Niels Bohrs Allé 1, DK-5230 Odense M, Denmark b Aarhus University, Faculty of Agricultural Sciences, Department of Biosystems Engineering, Research Centre Foulum, P.O. Box 50, DK-8830 Tjele, Denmark c Aarhus University, Faculty of Agricultural Sciences, Department of Animal Health and Bioscience, P.O. Box 50, DK-8830 Tjele, Denmark d Central South University of Forestry and Technology, College of Materials Science and Technology, Changsha 410004, Hunan, People’s Republic of China
a r t i c l e
i n f o
Article history: Received 14 March 2011 Received in revised form 29 June 2011 Accepted 11 July 2011 Available online 20 July 2011 Keywords: Biogas Digestibility Lignin Lignocellulose Predicting methane production potential
a b s t r a c t We examined the influence of fibrous fractions of biomass on biochemical methane potential (BMP) with the objective of developing an economical and easy-to-use statistical model to predict BMP, and hence the biodegradability of organic material (BD) for biogas production. The model was developed either for energy crops (grass, maize, and straw) or for animal manures, or as a combined model for these two biomass groups. It was found that lignin concentration in volatile solids (VS) was the strongest predictor of BMP for all the biomass samples. The square of the sample correlation coefficient (R2) from the BMP versus lignin was 0.908 (p < 0.0001), 0.763 (p < 0.001) and 0.883 (p < 0.001) for animal manure, energy crops and the combined model, respectively. Validation of the combined model was carried out using 65 datasets from the literature. Ó 2011 Elsevier Ltd. All rights reserved.
1. Introduction Biogas has been used as a source of renewable energy for more than 100 years. Its production has also played a valuable role in the sustainable management of agricultural byproducts, including animal manure. In Denmark, approximately 5% of animal manure is used for biogas production, and the aim is to use 40% of animal slurry in feedstock-to-biogas digesters by 2020 (Green Growth, 2009). Therefore it is necessary to improve profitability by enhancing methane yield in biogas production. Animal manure contains more readily degradable organic materials, such as proteins and lipids, than other agricultural byproducts, but it also has a high content of lignocellulose biofibers (40–50% of the total solids; Bruni et al., 2010). Lignocellulose consists mainly of three biopolymers: cellulose, hemicelluloses, and lignin. In lignocellulosic materials, cellulose is physically associated with hemicelluloses, and physically and chemically associated with lignin (Mussatto et al., 2008). Lignin and hemicelluloses are intermeshed and chemically bound through covalent cross-linkages such as ester or ether linkages (Jeffries, 1994). The low biodegradability (BD) of lignocellulose in biogas reactors is due to lignin being nondegradable in anaerobic environments (Mauseth, 1988) because the extracellular
⇑ Corresponding author. Tel.: +45 4117 8867; fax: +45 6550 7354. E-mail address: [email protected] (J.M. Triolo). 0960-8524/$ - see front matter Ó 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.biortech.2011.07.026
enzymes require oxygen to depolymerize. Furthermore, hydrolysis of cellulose in lignocellulosic materials is reduced by lignin and hemicelluloses, since these components act as a protective coat, making the cellulose resistant to enzymatic digestion (Mussatto et al., 2008). Animal manure contains high concentrations of lignin because it consists of residues from feed, where the easily degradable compounds have been taken up by the animals; therefore the biomass excreted contains mainly the slowly degradable components including lignocellulose. Ruminants are particularly efficient in using the carbon components in feed, and therefore excreta from ruminants contain high concentrations of slowly digestible organic matter. Biochemical methane potential (BMP) has been used as the most relevant indicator for assessing BD (Lesteur et al., 2010). BMP cannot be directly related to BD, since BMP is the methane yield, reflecting the destruction of organic materials, and the methane potential of each organic component in the volatile solids (VS) pool varies widely. For example, theoretically, the methane potential of lipid is 1018 L/kg, while the methane potential of cellulose is only 415 L/kg, based on the anaerobic degradation equation suggested by Symons and Buswell (1933). BMP assessed by different researchers and institutes is usually not comparable, due to differences in equipment used, environmental conditions, and experimental protocols (Angelidaki et al., 2009). As BMP data may vary depending on the batch method used, the application of a standardized method is needed. Furthermore, since the current
9396
J.M. Triolo et al. / Bioresource Technology 102 (2011) 9395–9402
methods of determining BMP are costly and time-consuming, innovative techniques for predicting BMP or BD are of great importance. Several alternate methods of estimating BD have been suggested (Lesteur et al., 2010) and numerous studies have attempted to predict BD or BMP by measuring the content of organic components (Tong et al., 1990; Gunaseelan, 2007). Near infrared spectroscopy (NIRS) has been used to predict organic matter components (Bruun et al., 2005; Michel and Ludwig, 2010). Recently the BMP of municipal solid waste has been shown to correlate well with NIRS results (Lesteur et al., 2011). Few studies have focused on lignin as the predominant variable to predict BD or BMP for biogas production. Chandler et al. (1980) observed a strong correlation between biodegradable fraction and lignin content in VS (R2 = 0.94) from diverse organic wastes, and Gunaseelan (2007) reported a weak relationship between the BMP and lignin content of several fractions of fruit and vegetable solids (R2 = 0.49). It was suggested that the low correlation could be due to the narrow range of lignin content in the fruits that were tested, making it impossible to reveal a significant effect of lignin. To our knowledge there is no satisfactory model for predicting BMP using lignin as a predominant variable that can be used for animal slurries, energy crops, and a mixture of both. The objective of this study was to examine the influence of lignin on BMP in energy crops and manure and to construct a statistical model to predict BMP and BD. The hypothesis was that lignin concentration can be used to assess BMP, and further to predict BD. The model, and the analytical procedures needed to provide data for the model, must be a cheap and fast alternative to existing methods.
2. Methods 2.1. Substrates and inoculum used Both animal slurries and energy crops were included in the study. Fresh pig and cattle slurry samples were taken from the pre-storage tanks of 10 farms in Horsens, Denmark. Grass, maize and straw were collected from farms in central Jutland. Table 1 shows the energy crops included in the study, with harvest date, pretreatment, and particle size. Representative subsamples for characterization and for the fermentation study were stored at 18 °C. The inocula used for the BMP assay were collected from two biogas plants at Fangel and Foulum. The Fangel biogas plant is operated under mesophilic conditions (37 °C), processing pig manure from 26 animal farms mixed with industrial organic waste. The Foulum biogas plant processes cattle manure mixed with crop residues under thermophilic conditions (55 °C). Inoculum from Foulum was taken from the post storage which is running under mesophilic conditions. The inocula were degassed at 37 °C for
14 days before application. The average pH and volatile solid of inocula were 8.0 and 66 (% of dry matter), respectively. The average methane concentration of biogas released from inocula was 68.2%. The pH of animal manure ranged between 7.3 and 7.7.
2.2. BMP assay The BMP of the animal manure and energy crops were determined using a batch technique based on methods described by Møller et al. (2004). 1100-mL infusion bottles were used as the batch digesters. Headspace was set to 30%. Inoculum to substrate was at unity or very close to unity on a VS basis (1:1). Blanks were tested using 770 g of inoculum to correct gas production. After addition of the substrate inoculum mixture, the digesters were closed with butyl rubber stoppers, sealed with aluminum crimps, flushed with N2 atmosphere and incubated at 37(±0.5) °C. All assays were performed in triplicate. Gas volume was measured either by replacing water or using a large syringe, as described by Steed and Hashimoto (1994). Digestion was continued until no further gas production was observed (90 days). Each batch digester was mixed thoroughly by shaking to prevent dry layers and to encourage degassing on workday. Gas volumes were measured every day at the beginning of fermentation and then gradually at larger time intervals. Methane and carbon dioxide were determined simultaneously once a week by a gas chromatograph (HP 6890 series), equipped with a thermal conductivity detector and a 30 m 0.320 mm column (J&W 113-4332). The carrier gas was helium (30 cm/s), and injection volume was 0.4 mL. Injector temperature was 110 °C, and detector and oven temperature was 250 °C. The split rate was 1:100. Methane quantification was evaluated according to VDI 4630 (2006). In detail, gas volume as read off was corrected as dry gas flow and as STP conditions (273 K, 1.013 bar). For quantification of methane concentration, simultaneously measured methane and carbon dioxide concentrations were multiplied by the same factor, being the sum of the corrected measured values as 100%, assuming that the fractions of ammonia and hydrogen sulfide are insignificant quantities (Eq. (1)).
C dry Cor ¼ C CH4
100 ðC CH4 þ C CO2 Þ
ð1Þ
where C dry Cor is corrected concentration of methane in the dry gas, (%), C CH4 is measured concentration of methane in the gas (%) and C CO2 is measured concentration of carbon dioxide in the gas (%). Methane volumes were calculated using corrected dry gas volume and corrected methane concentration. The methane volumes only from the substrate were calculated by subtracting the mean value of the inoculum control.
Table 1 The energy crops included in the study, with harvest date, pretreatment, and particle size. No.
Crop
Cultivar
Pretreatment
Harvest date
Particle size (mm)
1 2 3 4 5 6 7 8 9 10
Perennial grass Perennial grass Dried grass Grass Grass Grass Maize Maize Maize Dried straw
Mixed wild types Mixed wild types Mixed wild types Festulolium + 20% red clover, Hykor and Amos Festulolium, Achilles Festulolium + 20% red clover, Achilles and Amos Anvil Patrick Aurelia Wheat straw
Coarse cut Coarse cut Coarse cut Coarse cut Coarse cut Coarse cut None, cut during harvest None, cut during harvest None, cut during harvest Coarse cut
14 June 14 September NA 22 October 7 August 8 October 15 October 15 October 15 October NA
10–15 10–15 10–15 10–15 10–15 10–15 5 5 5 10–15
NA: not available.
9397
J.M. Triolo et al. / Bioresource Technology 102 (2011) 9395–9402
where Oi is the measured value, Pi is the predicted value, and O is the mean of the measured values.
2.3. Substrate analysis Prior to the BMP test, biochemical and physiochemical analyses were carried out. Dry matter (DM), volatile solids (VS), crude lipid, total ammoniacal nitrogen (TAN), and total Kjeldahl nitrogen (TKN) were determined according to standard procedures (APHA Standard Method, 2005). Volatile fatty acids (VFA) were determined according to the method of Lahav et al. (2002). Neutral detergent fibers (aNDF) were determined by a-amylase neutral detergent extraction (Mertens et al., 2002). Acid detergent fiber (ADF) and acid detergent lignin (ADL) were determined ash free by acid detergent extraction as described in the ISO Standard (ISO 13906; ISO, 2009). 2.4. Data analysis 2.4.1. Experimental data analysis Organic nitrogen (Norg) was calculated as the difference between TKN and TAN. Crude protein was determined by multiplying Norg by 6.25. Different fibrous fractions were determined in accordance with Van Soest’s characterization for fiber analysis (Van Soest, 1963; Goering and Van Soest, 1970), which enables the differentiation of fiber fractions using specific detergents. The aNDF treatment was used to determine total cell wall components, including hemicelluloses, cellulose, lignin, and fiber-bound proteins. ADF mainly consists of cellulose, lignin, and insoluble proteins that are components in the cell walls. The amounts of fibrous fractions were assessed as follows: lignocelluloses were assumed to be the aNDF fraction. Cellulose was determined by calculating the difference between ADF and ADL, and hemicelluloses as the difference between ADF and aNDF. Lignin is identified with ADL, with the assumption that the fraction of lignin-bound nitrogen is insignificant. 2.4.2. Statistical analysis Statistical data analysis was performed using the SAS software package (SAS Institute, 1992). BMP was tested as the dependent variable against all the fiber fractions. Simple linear regression and successive stepwise regression analyses were performed with statistically significant variables. In the statistical analysis, regression models to predict BMP for both energy crops and animal manure were developed. Furthermore regression models encompassing both types of biomass were developed. Relative root mean square error (RRMSE) was used to assess the accuracy of the model (Eq. (3)). The models were validated using scatter plots of predicted BMP versus measured BMP. Statistical indices proposed by Loague and Green (1991) were incorporated into the model. The final model included RRMSE, modeling efficiency (EF), and coefficient of residual mass (CRM), according to the following equations.
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Pn 2 i¼1 ðP i Oi Þ RMSE ¼ n
ð2Þ
where Oi is the measured value, Pi is the predicted value, and n is the number of data points.
RRMSE ¼
RMSE O
100
ð3Þ
where O is the mean of the measured values.
Pn
i¼1 Oi
CRM ¼
Pn
Pn
i¼1 P i
i¼1 Oi
ð4Þ
where Oi is the measured value and Pi is the predicted value.
Pn EF ¼
i¼1 ðOi
P OÞ2 ni¼1 ðPi Oi Þ2 Pn 2 i¼1 ðOi OÞ
ð5Þ
2.4.3. Theoretical BMP Theoretical BMP (TBMP) has been used to assess methane potential in several studies (Møller et al., 2004; Chae et al., 2008). BMP will never account for all of the TBMP, even if we include the 7% of VS that is used for bacterial growth (i.e., accounted as C5H9O3N; see Kalyuzhnyi, 1997). In the present study, we calculated TBMP with the method used by Møller et al. (2004) with some slight modifications. Lignin was included with the following empirical formula C10H13O3. Theoretical methane potential from lignin was calculated using the equation provided by Symons and Buswell (1933), which showed a theoretical production of 727.1 (CH4 NL (kg lignin)1) following Eq. (6):
Lignin C10 H13 O3 þ 5:25H2 O ! 5:875CH4 þ 4:125CO2
ð6Þ
Based on the component composition presented by Møller et al. (2004) and the one for lignin in Eq. (6), we propose that TBMP is calculated as follows:
TBMP ¼ ðVFA 373 þ Lipid 1014 þ Protein 496 þ Carbohydrate 415 þ Lignin 727Þ 0:001
ð7Þ
1
with TBMP as CH4 NL (kg VS) , and VFA, lipid, protein, carbohydrate, and lignin as g (kg VS)1. For TBMP calculation, all the data were used from our experiment, except crude fat from energy crop which was referenced by the Nordic standard feed table (NORFOR, 2011). Following the table, crude fat for the grasses (Nos. 1–6) were assumed as 39 g kg1 DM (feed code, 006-0082), for the maize as 22 g kg1 DM (feed code, 006-0307) and straw as 19 g kg1 DM (feed code, 006-0413) for the calculations. 3. Results and discussion The slurries included in this study had a wide range of water contents and VS concentrations (Table 2). On average, pig slurry contained 92.8% water and cattle slurry 80.8%. Some of the pig manures were very dilute and contained up to 99% water. The higher water content in pig manure was probably caused by spillage of cleaning and drinking water in the animal houses. On the other hand, the calf manure contained little water and had been collected in the form of semi-solids. The low water content of calf manure was due to a large fraction of straw used as bedding materials in the calf houses. VS concentrations were on average 74.9% of DM for pig manure and 80.2% of DM for cow manure. TAN in the cow manure ranged from 1.65 to 2.19 g kg1, while TAN in the pig manure varied more, from 1.13 to 5.63 g kg1. TKN ranged from 3.07 to 4.68 g kg1 for the cow manure, and from 1.33 to 6.87 g kg1 for the pig manure. In most studies, organic nitrogen in cattle manure is considerably higher than in pig manure, and most of the nitrogen in pig manure is in the form of TAN. The higher organic nitrogen in cattle manure is because most nitrogen is excreted in feces and this nitrogen is mainly of bacterial origin (70– 85%; Mosenthin et al., 1994). Nitrogen excreted in urine is mainly in the form of urea, which is more susceptible to decomposition than bacterial nitrogen excreted in feces (Mroz et al., 1993; Sommer et al., 2006). The crops tested were maize (Zea spp.), a mixture of perennial grasses and festulolium (a cross between Festuca spp. (meadow fescue/tall fescue) and Lolium multiflorum (Italian ryegrass)). Details of the crops are given in Table 3. TKN ranged from 0.99 to 2.02 g kg1 DM in the maize, 1.96 to 2.56 g kg1 DM in the festulolium, and from 1.92 to 2.78 g kg1 DM in the perennial grasses. The TAN of the crops was below the detection limit of the measuring method.
No.
1 2 3 4 5 6 7 8 9 10
Manure type
Piglet manure Sow manure Sow manure Pig fatteners Pig fatteners Calf manure Calf manure Cattle manure Cattle manure Inoculum
9398
Table 2 Composition, BMP and TBMP of the manure samples used in the study. DM
VS
VFA
XL
TKN
TAN
Lignocellulose
Hemicelluloses
Cellulose
Lignin
BMP
TBMP
CH4 content
(g kg1 w/w)
(g kg1 w/w)
(mM)
(g kg1 DM)
(g kg1 w/w)
(g kg1 w/w)
% of DM
% of DM
% of DM
% of DM
CH4 NL (kg VS1)
CH4 NL (kg VS1)
% of dry gas
16.2(0.3) 49.0(1.2) 109.4(6.5) 119.6(1.0) 9.4(0.1) 154.6(3.3) 322.6(2.7) 115.6(0.9) 173.6(0.5) 34.7(0.5)
11.8(0.1) 38.2(0.4) 90.2(0.3) 100.6(0.5) 5.2(0.1) 142.3(0.1) 295.2(1.0) 94.3(0.6) 96.8(0.6) 22.9(0.4)
109.5(3.8) 75.1(1.4) 170.6(3.7) 275.8(2.2) 28.6(0.7) 115.6(4.1) 453.6(17.8) 180.6(2.5) 166.4(2.5) 12.7(1.0)
60.4(3.1) 107.4(17.5) 120.5(12.0) 121.5(10.5) 62.5(6.9) 10.89(0.25) 15.54(2.05) 74.7(2.3) 57.6(3.4) 8.1(2.4)
1.33(0.24) 3.94(0.03) 4.87(0.28) 6.87(0.28) 1.64(0.10) 3.07(0.24) 4.57(0.30) 4.68(0.11) 4.38(0.13) 4.74(0.20)
1.13(0.04) 3.71(0.04) 3.56(0.03) 5.63(0.04) 1.51(0.03) 1.65(0.01) 2.19(0.25) 1.90(0.06) 1.82(0.03) 3.90(0.07)
5.11 47.11(2.11) 47.59(0.62) 47.10(0.96) NA 71.18(0.75) 72.92(0.04) 36.46(0.48) 25.63(0.42) 25.27(0.38)
0.97(0.22) 15.36(1.09) 14.79(2.68) 15.65(0.55) NA 17.74(2.58) 22.24(1.65) 7.97(1.55) 3.56(0.59) 3.41(1.17)
3.76(0.02) 21.41(0.94) 23.22(1.79) 24.16(4.47) 1.15(0.26) 42.78(1.76) 42.97(1.64) 16.58(1.36) 15.58(0.20) 10.65(3.32)
0.38(0.25) 10.34(0.08) 9.58(0.18) 7.30(4.88) 0.51(0.13) 10.65(0.07) 7.76(1.05) 11.91(2.43) 6.49(0.37) 11.21(2.39)
417.2(11.8) 213.8(9.2) 248.8(16.3) 284.8(3.8) 345.3(3.7) 198.7(3.6) 237.0(7.8) 197.0(3.3) 223.6(3.7) 142.1(2.5)
449.6 537.5 531.0 527.8 482.2 496.8 442.7 525.1 522.7 492.7
66.4(1.3) 62.8(0.8) 60.8(0.2) 63.9(0.3) 69.1(3.6) 57.9(1.5) 58.4(0.5) 61.1(0.3) 61.7(0.3) 67.7(0.9)
Figures in parentheses are standard deviations. XL: crude lipid; NL: norm liter (273 K, 1.013 bar); NA: not available.
No.
Biomass type
DM (g kg1 w/w)
VS (g kg1 w/w)
TKN (g kg1 w/w)
Lignocellulose % of DM
Hemicelluloses % of DM
Cellulose % of DM
Lignin % of DM
BMP CH4 NL (kg VS1)
TBMP CH4 NL (kg VS1)
CH4 content % of dry gas
1 2 3 4 5 6 7 8 9 10
Perennial grass Perennial grass Dried wild grass Grass Grass Grass Maize Maize Maize Straw
288.2 235 934.1(0.7) 187.4 180.2 175.3 387.5 299.2 278.9 936.9(0.1)
77.2 52 880.2(1.1) 32.3 28.9 27.7 145.5 85.6 74.5 906.3(1.9)
5.53 6.54 10.82(1.2) 4.8 3.55 3.53 4.84 2.96 5.62 3.45(0.7)
60.79(2.08) 58.33(0.54) 64.85(0.17) 59.54(0.30) 47.72(0.18) 51.73(0.57) 63.82(1.61) 65.77(1.58) 60.86(2.31) 83.76(0.14)
25.99(1.95) 28.25(0.64) 22.89(0.44) 27.30(0.56) 20.93(0.74) 24.39(0.66) 37.68(1.70) 35.92(2.51) 36.73(2.85) 28.38(1.32)
29.05(0.81) 26.86(0.08) 35.99(0.02) 29.97(0.72) 25.56(0.09) 25.38(0.33) 23.74(2.14) 26.86(3.91) 22.31(0.51) 47.97(1.64)
5.75(0.68) 3.22(0.03) 5.98(0.59) 2.27(0.44) 1.23(0.47) 1.96(0.25) 2.40(2.21) 2.98(0.18) 1.81(1.05) 6.41(0.18)
271.0(8.5) 410.5(67.8) 306.3(7.9) 373.5(35.7) 420.8(22.2) 438.9(21.8) 399.4(21.3) 405.3(20.3) 360.5(24.8) 289.5(17.5)
466.3 462.4 461.4 460.7 453.9 455.7 452.3 443.8 445.2 447.7
53.8 56.0 62.5(0.1) 58.6 55.0 54.2 59.0 57.9 59.0 62.2(0.0)
Figures in parentheses are standard deviations. NL: norm liter (273 K, 1.013 bar).
J.M. Triolo et al. / Bioresource Technology 102 (2011) 9395–9402
Table 3 Composition, BMP and TBMP of the energy crops used in the study.
J.M. Triolo et al. / Bioresource Technology 102 (2011) 9395–9402
9399
Fig. 1. Lignocellulose fraction in VS in animal manures and energy crops.
Fig. 2. Distribution of hemicelluloses, cellulose, and lignin in lignocellulose.
For most of the animal manure samples, the lignocellulose content was between 40% and 80% of VS, except for the piglet manure, where lignocellulose content was 7.0% of VS (Fig. 1). The low lignocellulose content of piglet manure was due to the piglet diet having a very low fiber content. The highest lignocellulose contents were found in calf manure, which was due to calf bedding materials. For adult animals, the concentration of lignocellulose was approximately 50% of VS for both cow and pig manure. In manure the content of hemicelluloses was 26.5(±8.1)% of VS, of cellulose 55.6(±12.5)% of VS, and of lignin 17.9(±8.7)% of VS (Fig. 2). For the energy crops, the content of hemicelluloses, cellulose, and lignin were 45.3(±11.9)% of VS, 48.9(±9.9)% of VS, and 5.8(±2.0)% of VS, respectively. Comparing the two types of biomass, the hemicelluloses content was lower in manure than in the crops, while the lignin content was higher. During digestion, the relatively easily degradable hemicelluloses seem to be transformed and utilized by the animals more readily than cellulose. Lignin is a larger fraction in manure than energy crops, as lignin is upconcentrated by animal digestion. Cumulative methane production of pig manure, cattle manure and crop as a function of time are shown respectively in Fig. 3. For the first 2 weeks the great majority of methane gas was produced, and thereafter only small amounts of gas were released.
Fig. 3. Cumulative methane production from pig manure, cattle manure and energy crops.
The methane concentrations in the dry gas of each substrate were ranged from 53.8% to 62.5% for energy crops. For animal manure it was slightly higher, ranging from 57.9% to 69.1%. Cumulative methane production (90 days) used as BMP ranged from 197 to 417 CH4 NL (kg VS)1 for animal manure, and for energy crops 271–439 CH4 NL (kg VS)1. The results of TBMP are presented in Tables 2 and 3. As shown in the tables, TBMP was less variable than BMP, ranging from 443 to 538 CH4 NL (kg VS)1 for animal manure, 444–466 CH4 NL (kg VS)1 for crop. 3.1. Stepwise regression analysis 3.1.1. Lignin Using the single variables, the most significant BMP models were predicted by lignin for all the categories (Table 4). The square of the sample correlation coefficient (R2) between BMP and lignin
9400
J.M. Triolo et al. / Bioresource Technology 102 (2011) 9395–9402
Table 4 Summary statistics of statistical analyses for energy crops and animal manure. Variable
R2
p
RRMSE (%)
Equation
Energy crop
Lignin Cellulose ADF NDF Lignin, cellulose
0.763 0.407 0.511 0.435 0.766
<0.001 <0.05 <0.05 <0.05 <0.01
7.4 11.7 10.7 11.5 6.1
BMP = 2.58 lignin + 460.6 BMP = 0.49 cellulose + 521.5 BMP = 0.45 ADF + 668.1 BMP = 0.46 NDF + 447.1 BMP = 2.77 lignin 0.07 cellulose + 447.1
Manure
Lignin Cellulose ADF NDF Lignin, cellulose
0.908 0.485 0.695 0.491 0.941
<0.0001 <0.05 <0.01 <0.05 <0.0001
9.1 21.7 16.7 20.7 7.4
BMP = 1.40 lignin + 388.9 BMP = 0.36 cellulose + 341.8 BMP = 0.34 ADF + 373.3 BMP = 0.25 NDF + 371.3 BMP = 1.22 lignin 0.11 cellulose + 399.4
Combined model
Lignin Cellulose ADF Lignin, cellulose
0.883 0.106 0.358 0.885
<0.001 >0.05 <0.01 <0.001
9.8 NA 22.9 7.4
BMP = 1.675 lignin + 421.7 NA BMP = 0.373 ADF + 439.6 BMP = 1.649 lignin 0.035 cellulose + 430.0
NA: not available.
was 0.908 for animal manure and 0.763 for the energy crops. Relative error (RRMSE) of the model was below 10% when lignin was chosen as the independent variable, indicating that the models relating BMP to lignin predicted BMP well. RRMSE shows slightly better model accuracy for energy crops (RRMSE = 7.4%) than for animal manure (RRMSE = 9.1%). Moreover, the slope of the regression models using lignin was more negative for energy crops (2.58) than for animal manure (1.40), which shows that lignin affects BMP more significantly in energy crops than in animal manure. The slope of BMP to lignin, which we call lignin dependency, is much higher for crop residues than for manure, because the cell wall is protected by an intact lignocellulose matrix in crops, whereas in animal manure part of this matrix is broken down inside the animal during digestion. The higher statistical significance in contrast to the lower model accuracy for the manure model compared with the crop model is probably caused by the wider range of values of the independent variables of manure data, which improve the correlation level. For example, the lignin concentration of samples used ranged from 5.2 to 169.8 g (kg VS)1 in animal manure, whereas lignin in crop residue only varied between 13.6 and 66.2 g (kg VS)1. In the combined model, the correlation between lignin concentration and BMP was good (R2 = 0.883). For both crops and animal manure, the lignin concentrations were found in clusters of high and low lignin content. Two of the animal manures were so different from the rest that they may be defined as outliers. However, when these two outliers are included in the statistical analysis they contribute positively to the correlation between BMP and lignin and improve the regression models. However, the accuracy of the model decreased (RRMSE = 9.8%) when the individual models were combined. The results show that the influence of lignin concentration in the biomass is similar between the two categories of biomass, but the best model for prediction of BMP is the one that accounts for the different origin of the biomass. This indicate a need to define a variable and parameters that provide a method to develop one model; i.e., an indicator of the destruction of the physical lignocellulose matrix during the digestion of feed in animals.
show a significant relationship between cellulose and BMP. Furthermore, no significant relation was found between hemicelluloses and BMP in any of the tested models. It is interesting that BMP seems to be related only to the content of lignin, and the relations between BMP and other fibers such as cellulose and hemicelluloses are insignificant, depending on the model category. This suggests that cellulose and hemicelluloses are greatly affected by the lignin barrier, and consequently the destruction rate of these fibers differs between sample categories as affected by the treatment before BMP measurements. The result confirms that lignin controls VS destruction most significantly, which makes it possible to provide fine correlations between BMP and lignin. Since lignin is not degradable, BMP will be negatively correlated with the fraction of lignin in VS. Moreover, the significant relation of BMP to cellulose indicates that degradation of cellulose, which is suppressed by the lignocellulose matrix, affects the destruction rate of VS. Consequently a large amount of cellulose was not transformed to methane during the batch experiment and this is reflected by the BMP. This is also the case with lignin, which links to cellulose and creates barriers for microbial degradation not only of lignin but also of cellulose. Therefore, cellulose in lignocellulosic materials may not significantly contribute to BMP.
3.1.2. Cellulose and hemicelluloses BMP was significantly (p < 0.05) related to cellulose for the two individual models, but the correlation was inferior to the relation between BMP and lignin. R2 values for cellulose were much lower than those for the lignin models (R2 = 0.485 for animal manure, R2 = 0.407 for energy crops). The relatively high RRMSE values indicate low model efficiency when cellulose was used as a single variable. Moreover, a combined model for crops and manure does not
A BD model assessed as the ratio between BMP to TBMP versus lignin together with a BMP model is shown in Fig. 4. R2 between BMP/TBMP was slightly higher than the BMP model as 0.775 (RRMSE = 7.4) for energy crop, 0.922 (RRMSE = 9.8) for animal manure and 0.893 (RRMSE = 10.7) for the combined model. The intercept of the regression line was very close to 1(1.0197) for energy crop and for animal manure lower as 0.8322. This result indicates that only lignin is a nondegradable fraction for energy crop,
3.1.3. Multiple regression tests In the multiple regression tests, lignin and cellulose were applied as two independent variables and R2 was slightly improved from 0.763 to 0.766 (energy crops), 0.908 to 0.941 (animal manure), and 0.883 to 0.885 (combined model) compared with the models where only lignin was used as the independent variable (Table 4). Relative errors were somewhat reduced, when lignin and cellulose were employed; that is, RRMSE was 6.1% (energy crops), 7.4% (animal manure), and 7.4% (combined model). This unsatisfactory improvement from including cellulose in the regression equation was due to the interrelated effect of lignin and cellulose on BMP. 3.2. Biodegradability influenced by lignin
9401
J.M. Triolo et al. / Bioresource Technology 102 (2011) 9395–9402
Table 5 The datasets used for validation of the models and their regression linear trend of the measured BMP versus lignin content in VS. References
R2
Slope
Intercept
Biomass used
Amon et al. (2007) Gunaseelan (2007) Møller et al. (2004) Oslaj et al. (2010)
0.7715
1.19
373.7
0.5002
1.38
493.1
0.6800
3.26
489.1
0.0021
NA
NA
n = 16 (maize 10; dairy cow manure 6) n = 24 (fruit and vegetable waste 17; sorghum and Napier grass 7) n = 10 (cattle manure 2; pig fattener manure 4; sow manure 3; straw 1) n = 10 (maize 10)
NA: not available.
Fig. 4. Effect of lignin concentration on BMP and BD (BMP/TBMP): A, as regression line of BMP for energy crops (y = 2.58x + 460.6); B, for animal manure (y = 1.40x + 388.9); C, for the two categories combined (y = 1.675x + 421.7); A0 , as regression line of BD for energy crop (y = 0.0059x + 1.0197); B0 , for animal manure (y = 0.0033x + 0.8322); C0 , for the two categories combined (y = 0.0041x + 0.9331).
while the lower intercept for animal manure probably shows that BD of animal manure was affected to some extent by recalcitrant fractions, which reduce BD of animal manure. Therefore, we may conclude BMP or BD could be more effectively predictable using lignin for energy crop, since in energy crop crude protein and crude fat are low, and that is critical for BMP.
3.3. Model validation We tested 65 datasets from the literature to validate the combined model (see Table 5). Most of the datasets did not include the effect of lignocelluloses on BMP and statistical analyses of BMP versus lignin had not been performed. Gunaseelan (2007) carried out a statistical analysis, but found no significant relation of BMP to lignin for the two biomass groups, manure and crops. The reason for the non significant results was probably that lignin concentration of the biomass did not have a wide enough range to test the effect of lignin concentration as variables. However, when including all the datasets from Gunaseelan (2007) in the statistical analysis, BMP can be shown to be significantly related to the lignin content of the biomass. The study of Oslaj et al. (2010) indicated a tendency, but no significant relation, of BMP to lignin. This was probably also due to the limited range of lignin concentration and BMP in the data.
Fig. 5. Validation of the suggested model: measured BMP versus predicted BMP and the linear trend.
The predicted values versus measured values of BMP from the various studies were plotted in Fig. 5. The BMP values from Gunaseelan (2007) generally lie below the trend line of predicted BMP, while those of Møller et al. (2004) are higher. This indicates that heavy scattering does not occur only because of model weakness, but also because BMP has been determined using different batch methods. In the BMP results from Amon et al. (2007), we can clearly distinguish between two groups (lower BMP from dairy cow manure, higher BMP from maize biomass); if the model was tested separately for the individual types of biomass, there would probably be no statistical significance. The slop and intercept of the linear regression line was 0.45 and 125, respectively, despite the perfect fit for 1 and 0, respectively. The low slope of the regression line indicates that the suggested model has a tendency to underestimate BMP at high BMP values. Moreover, the weakness of the model was that its intercept is considerably lower (421.7 CH4 NL (kg VS)1) than the theoretical intercept (around 500 CH4 NL (kg VS)1) that is expected for TBMP after subtracting 7% for bacterial growth. For the combined model, RRMSE, CRM, and EF were 26.5%, 0.11%, and 0.21%, respectively. The higher statistical significance in contrast to the lower model accuracy for the manure model compared with the crop model is probably caused by the wide range of concentrations of independent variables of manure data, which improve the correlation level.
9402
J.M. Triolo et al. / Bioresource Technology 102 (2011) 9395–9402
4. Conclusions This study confirmed that the lignin fraction could be used to predict BMP for the combined model for energy crops and animal manures. The BMP model was slightly improved when cellulose was included in the model together with lignin as an independent variable. The validation procedure shows that the model needs to be improved to increase precision of prediction of BMP. Nevertheless, the study shows that BMP had the greatest dependency on lignin concentration, and information on fiber fractions in the biomass could also be used to predict the BMP and BD of biomass. Acknowledgement This study has been supported with a grant from Energi Fyns Udviklingsfond. References Amon, T., Amon, B., Kryvoruchko, V., Zollitsch, W., Mayer, K., Gruber, L., 2007. Biogas production from maize and dairy cattle manure: influence of biomass composition on the methane yield. Bioresour. Technol. 100, 5777–5782. Angelidaki, I., Alves, M., Bolzonella, D., Borzacconi, L., Campos, J.L., Guwy, A.J., Kalyuzhnyi, S., Jenicek, P., van Lier, J.B., 2009. Defining the biomethane potential (BMP) of solid organic wastes and energy crops: a proposed protocol for batch assays. Water Sci. Technol. 59 (5), 927–934. APHA, 2005. Standard Methods for the Examination of Water and Wastewater, 21st ed. American Public Health Association, Washington, DC. Bruni, E., Jensen, A.P., Angelidaki, I., 2010. Steam treatment of digested biofibers for increasing biogas production. Bioresour. Technol. 101, 7668–7671. Bruun, S., Stenberg, B., Breland, T.A., Gudmundsson, J., Henriksen, T.M., Jensen, L.S., Korsæth, A., Luxhøi, J., Pálmason, F., Pedersen, A., Salo, T., 2005. Empirical predictions of C and N mineralization patterns from near infrared spectroscopy, stepwise chemical digestion or C/N ratios. Soil Biol. Biochem. 37, 2283–2296. Chae, K.J., Jang, A., Yim, S.K., Kim, I.S., 2008. The effect of digestion temperature and temperature shock on the biogas yields from the mesophilic anaerobic digestion of swine manure. Bioresour. Technol. 99, 1–6. Chandler, J.A., Jewell, W.J., Gossett, J.M., Van Soest, P.J., Robertson, J.B., 1980. Predicting methane fermentation biodegradability. Biotechnol. Bioeng. Symp. 10, 93–107. Goering, H.K., Van Soest, P.J., 1970. Forage Fiber Analysis (Apparatus, Reagents, Procedures and Some Applications). Agriculture Handbook No. 379. Agriculture Research Service, United States Department of Agriculture, Washington, DC. Green Growth, 2009. Grøn Vækst. Danish Environmental Ministry, April 2009 (in Danish). Available from:
Kalyuzhnyi, S.V., 1997. Batch anaerobic digestion of glucose and its mathematical modeling. II. Description, verification and application of model. Bioresour. Technol. 59, 249–258. Lahav, O., Morgan, B.E., Loewenthal, R.E., 2002. Rapid, simple, and accurate method for measurement of VFA and carbonate alkalinity in anaerobic reactors. Environ. Sci. Technol. 36 (12), 2736–2741. Lesteur, M., Bellon-Maurel, V., Gonzalez, C., Latrille, E., Roger, J.M., Junqua, G., Steyer, J.P., 2010. Alternative methods for determining anaerobic biodegradability: a review. Process Biochem. 45 (4), 431–440. Lesteur, M., Latrille, E., Maurel, V.B., Roger, J.M., Gonzalez, C., Junqua, G., Steyer, J.P., 2011. First step towards a fast analytical method for the determination of Biochemical Methane Potential of solid wastes by near infrared spectroscopy. Bioresour. Technol. 102 (3), 2280–2288. Loague, K., Green, R.E., 1991. Statistical and graphical methods for evaluating solute transport models: overview and application. J. Contam. Hydrol. 7, 51–73. Mauseth, J.D., 1988. Plant Anatomy. The Benjamin/Cummings Publishing Company, CA. Mertens, D. et al., 2002. Gravimetric determination of amylase-treated neutral detergent fiber in feeds with refluxing in beakers or crucibles: collaborative study. J. AOAC Int. 85 (6), 1217–1240. Michel, K., Ludwig, B., 2010. Prediction of model pools for a long-term experiment using near-infrared spectroscopy. J. Plant Nutr. Soil Sci. 173, 55–60. Møller, H.B., Sommer, S.G., Ahring, B.K., 2004. Methane productivity of manure, straw and solid fractions of manure. Biomass Bioenerg. 26, 485–495. Mosenthin, R., Sauer, W.C., Ahrens, F., 1994. Dietary pectin’s effect on ileal and fecal amino acid digestibility and exocrine pancreatic secretions in growing pigs. J. Nutr. 124, 1222–1229. Mroz, Z., Jongbloed, A.W., Beers, S., Kemme, P.A., de Jonge, L., van Berkum, A.K., van der Lee, R.A., 1993. Preliminary studies on excretory patterns of nitrogen and anaerobic deterioration of fecal protein from pigs fed various carbohydrates. In: Verstegen, M.W.A., den Hartog, L.A., van Kempen, G.J.M., Metz, J.H.M. (Eds.), Nitrogen Flow in Pig Production and Environmental Consequences. EAAP Publication No. 69. PUDOC, Wageningen, The Netherlands, pp. 247–252. Mussatto, S.I., Fernandes, M., Milagres, A.M.F., Roberto, I.C., 2008. Effect of hemicellulose and lignin on enzymatic hydrolysis of cellulose from brewer’s spent grain. Enzyme Microb. Technol. 43, 124–129. NORFOR (Nordic Feed Evaluation System) 2011. In: Harald Volden (Ed.), EAAP Scientific Series, ISSN 0071-2477, 130, Wageningen Academic. Available from: