Prediction of organic acid concentration in BRS Capiaçu elephant grass silage

Abstract This study aimed to develop prediction models for organic acid concentrations in Capiaçu grass ( Pennisetum purpureum Schum) silage, based on fermentation parameters and chemical composition. Data were obtained from an experiment that evaluated BRS (Brazil Seeds) Capiaçu grass silage with the inclusion of dehydrated cashew pseudofruit ( Anacardium occidentale ) at levels of 0%, 10%, 20% and 30% (on a fresh matter basis). The most accurate models for predicting lactic acid used dry matter as the only predictor variable, reaching an R 2 of 0.98 and root mean square prediction error (RMSPE) of 1.52. In the case of acetic acid, the best prediction model combined pH and the quadratic terms of pH and buffering capacity (BC), with R 2 of 0.73 and RMSPE of 0.98. The most accurate prediction model for butyric acid included crude protein and BC, with R 2 of 0.51. The prediction model with the greatest predictive capacity for propionic acid combined organic matter, BC and the quadratic term of soluble carbohydrates, reaching R 2 of 0.75. Thus, prediction models for lactic acid ( R 2 = 0.98) are suitable for field applications, whereas those for butyric acid ( R 2 = 0.51) require caution. In the validation of the selected prediction models, all showed accuracy and precision with correlation and agreement coefficients ≥ 0.70 for most models. Prediction errors were low, and the null hypothesis ( β ₀ = 0 and β ₁ = 1) was accepted for all prediction models. The organic acids in Capiaçu‐grass silage can be estimated based on its pH and chemical composition, particularly dry matter, buffering capacity, crude protein, organic matter and soluble carbohydrates.