VS is measured by burning dried materials for at least 2 hours at 525 °C, where the residues are defined as ash and the volatile fraction as VS. As each VS component has different stoichiometric methane potentials (TBMP) and different digestibility, knowing the composition of the VS component could be used to assess BMP alternatively instead of performing a fermentation test. Table 3 presents the TBMP of each organic component, where it shows that lipid and lignin is only preferable in respect to TBMP.

Whereas stoichiometric methane potential of each organic component is known relatively well, BD of it in animal slurry is poorly researched except VFA and Lignin. VFA is the intermediate during the procedure of digestion and the presence of VFA in animal slurry

indicates the previous occurrence of hydrolysis. As hydrolysis decides degradation rate, we may hypothesise that the concentration of VFA in animal slurry may significantly correlate with digestibility, and that can further be correlated to BMP. For the lignin, Triolo et al. [10] confirmed that BD is significantly related to lignin concentration. Using the VFA results from the animal slurry used as independent variables against BMP, a reasonable correlation between VFA concentration and BMP was found (Figure 8). Furthermore, a fine correlation between lignin and BMP was also found.

Statistical analysis showed that BMP significantly correlated with VFA, lignin and celluloses, though the correlation level of cellulose to BMP was quite weak. (p<0.05). On the other hand, it was not possible to find any correlation from other protein, hemicellulose, lipid, etc. The result of a simple linear regression test between BMP and organic components is given in Table 4, only showing significant models. Furthermore, multiple regression tests were performed using the significant variables, but excluding cellulose, since the model was not improved significantly including cellulose.

materials (10, 37-43). Therefore we tested the precision of the algorithms obtained to test if the model could be used to predict BMP well enough.

The precision of the model was evaluated by employing the relative root mean square error (RRMSE), which represents relative errors. As can be seen in Table 4, relative errors of the BMP model were similar for lignin and VFA, being 17% approximately, while relative error decreased to 11.8% when both of the variables were used for multiple regression tests.

Measured BMP versus predicted BMP using the model from multiple linear regression tests is plotted in figure 9, where it shows a good linear correlation. The slope of the best regression line and linear trend obtained was also very similar. The results indicate that the model predicted by cellulose is not preferable, whereas the BMP model using VFA and lignin could be useful for BMP assessment instead of time demanding fermentation tests.