Prediction of nutritive values in grass silages: I. Nutrient digestibility and energy concentrations using nutrient compositions and fermentation characteristics

Grass silages (n = 136) were selected from commercial farms across Northern Ireland according to their pH, ammonia nitrogen, DM, and predicted ME concentration. Each silage was offered to four sheep as a sole feed at maintenance feeding level to determine nutrient digestibility and urinary energy output. Dry matter concentration was determined as alcohol-corrected toluene DM and was subsequently used as the basis for all nutrient concentrations. The objectives were to use these data to examine relationships between nutritive value and nutrient concentration or fermentation characteristics in silages and then develop prediction equations for silage nutritive values using stepwise multiple regression techniques. The silages had a large range in quality (DM = 15.5 to 41.3%, ME = 7.7 to 12.9 MJ/kg of DM, pH = 3.5 to 5.5) and a relatively even distribution over the range. There was a positive relationship (P < 0.001) between silage GE and DE or ME concentration. Digestible OM in total DM (DOMD); ME/GE; and digestibility of DM, OM, and GE were positively related (P < 0.05) to CP, soluble CP, ether extract, lactic acid concentration, and lactic acid/ total VFA, whereas they were negatively related (P < 0.05) to ADF, NDF, lignin, individual VFA concentration, pH, and ammonia N/total N. Concentrations of DE and ME and digestibility of CP and NDF had similar relationships with those variables, although some relationships were not significant. Three sets of multiple prediction equations for DE and ME concentration; ME/ GE; DOMD; and digestibility of DM, OM, GE, CP, and NDF were therefore developed using three sets of predictors. The first set included GE, CP, soluble N/total N, DM, ash, NDF, lignin, lactic acid/total VFA, and ammonia N/total N; the second set excluded soluble N/ total N and lignin because they are not typically measured; the third set further excluded the fermentation data. The R2 values generally decreased with exclusion of predictors. The second and third sets of equations, except for NDF digestibility, were validated using the mean-square-prediction-error model and an independent grass silage data set published since 1977 (n = 17 DM digestibility] to 28 DOMD and OM digestibility]). The validation indicated that the equations developed in the present experiment could accurately predict DE and ME concentrations and DE/GE and ME/GE in grass silages.