Autoria: CAMPOS, R. L. C. de; OLIVEIRA, R. H. M. de; MACEDO, M. C. M.; GOMES, R. da C.

Resumo: Abstract – The aim was to improve multivariate calibration curves for near- infrared refl ectance spectroscopy (NIRS) for the measurement and prediction of nutritive value of tropical forages. Samples from diff erent species, years of harvest and management techniques were used for spectral readings and chemical reference analyses. For the development of partial least squares (PLS) regressions, between 1,435 and 2,080 samples were used as the calibration set and between 713 and 1,080 samples as the external validation set per property. The calibrations were chosen based on the best Q-Value (Büchi, Switzerland) which encompasses, among other factors, the coeffi cients of determination, standard errors of calibration, standard error of prediction, consistency, validation bias and correlation between predicted and reference data. The predicted and reference results had similar means and standard deviations for the characteristics studied. The Q-Value obtained varied between 0.57 and 0.82. The magnitude of these values suggests that the models can be used for routine analysis of forages nutritional value, replacing or complementing chemical analysis, in plant breeding programs and research projects on animal nutrition, pasture management and production systems.

Ano de publicação: 2024

Tipo de publicação: Folhetos

Unidade: Embrapa Gado de Corte

Palavras-chave: Análise, Nutrição Animal