Near infrared spectroscopy (NIRS) calibration equations to predict nutritional quality parameters of tropical forages

摘要

The objective of this study was to develop NIRS calibration curves to predict nutritional quality parameters of tropical forage species. For this a total of 1,991 samples of tropical forages (grasses and legumes) were employed. These samples showed a high heterogeneity in regrowth age, vegetative state and parts of the plants collected, sampling period and sample origin. Chemical analysis was performed to determine crude protein, neutral detergent fiber, acid detergent fiber (ADF), in vitro digestibility of dry matter and digestibility of neutral detergent fiber. A group of samples was used for each chemical parameter to develop the calibration curves. The curves were chosen taking into account the standard errors of: calibration; cross validation (SECV); and prediction. To evaluate the accuracy of each equation the coefficient of determination (R 2 ) and the RPD (ratio performance deviation) index (relation between standard deviation and SECV), which assesses the predictive power of the equations, were calculated. Validation was performed with external samples and results showed correlation coefficients (r) of 0.90 for all parameters except for ADF (r=0.72), demonstrating the precision of the predictive equations. However, the determination of ADF by this analytical method requires further work to increase its reliability. Keywords: Analytical methods, animal nutrition, forage quality, grasses, legumes. DOI: 10.17138/TGFT(4)139-145