Prediction of metabolisable energy concentrations from nutrient digestibility and chemical composition in grass silages offered to sheep at maintenance

Abstract

A total of 174 perennial ryegrass silages were evaluated in nine studies for chemical composition, nutrient digestibility and urinary energy output with sheep (four sheep/silage) offered the silage as a sole diet at maintenance energy feeding level. Silage metabolisable energy (ME) concentration was estimated from measured energy intake and outputs from faeces and urine and predicted methane energy. All silage data were expressed on an alcohol corrected toluene dry matter (DM) basis. The objectives were to use these silage data to develop prediction equations for ME concentration and ME/GE (gross energy) and then validate these equations using published grass silage data. There was a large range in ME concentration (7.7–13.6 MJ/kg DM), ME/GE (0.432–0.668) and digestible organic matter in DM (DOMD; 0.530–0.769). The ME concentration and ME/GE were positively related to GE and DE concentration, DOMD and digestibility of DM, organic matter (OM), GE, crude protein (CP) and neutral detergent fibre ( P < 0.05 or less). Prediction of ME concentration using digestible energy (DE) or GE digestibility plus residual GE concentration (GE—mean GE) produced a strong relationship ( R 2 = 0.98), while using DM or OM digestibility or DOMD as a sole predictor reduced R 2 values to approximately 0.73. The latter R 2 values were marginally increased when CP concentration was added as a secondary predictor, and substantially increased to over 0.90 when residual GE was added. Prediction of ME/GE using GE digestibility produced a higher R 2 value (0.97) than those (0.88–0.90) using DM or OM digestibility or DOMD. The R 2 values were marginally increased when adding CP or residual GE concentration as a secondary predictor to the latter relationships. These equations were validated using published grass silage data since 1977 ( n = 21) and the mean-square-prediction error. Prediction of ME using DE or GE digestibility with residual GE had the lowest mean prediction error (MPE), with the predicted ME close to observed ME, and the majority of error derived from random variation. Using DM or OM digestibility or DOMD as a sole predictor for ME concentration produced a relatively large MPE, while adding CP or residual GE generally reduced the MPE and errors derived from both bias (predicted−actual) and line (slope). Similar results also occurred for prediction of ME/GE. DE concentration is the most accurate predictor for ME concentration. Prediction of ME using OM digestibility or DOMD as a sole predictor can result in error, but the prediction accuracy can be improved by adding GE concentration as a secondary predictor.