Analysis of nutritional content in 13 forage crop varieties in the Loess Plateau based on visible/near infrared reflectance spectroscopy

A visible/near-infrared reflectance spectroscopy (visible/NIRS) model was developed to determine the forage cultivar nutritional composition of 780 hay samples under simulated rotational grazing. Hay samples (n=780) from 13 forage crop varieties under simulated grazing in Loess Plateau during the 2012 to 2013 growing season were evaluated using calibration methods for prediction of nutrient contents using NIRS. The following results were obtained. The optimal calibrations in Leguminosae were ether extract (EE), acid detergent fiber (ADF), and crude ash(Ash). The multiple correlation coefficients(RSQ) and 1-variance ratio (1-VR) were 0.94 and 0.7, and standard error of calibration (SEC) and standard error of cross validation (SECV) were 0.071~0.713 and 0.160~2.751, respectively. Optimal calibrations in Gramineae were EE and water-soluble carbohydrate content(WSC). RSQ were 0.916 and 0.859 and 1-VR were 0.609 and 0.810 for EE and WSC, respectively, and SEC was 0.250 and 1.488 and SECV was 0.505 and 3.172, respectively. For the other species, the results for nutrient predication were reasonably good, with the exception of ADF. RSQ and 1-VR were 0.85 and 0.70, and SEC and SECV were 0.361～3.557 and 0.495～4.602, respectively. These results indicate that the accuracy of prediction using NIRS was acceptable for 13 forage crop nutrients, although the crude protein(CP), ADF, neutral detergent fibre(NDF), Ash of Gramineae and the ADF of others species (RSQ 0.7) may require further calibration. The accuracies of the predictions for CP and WSC in Leguminosae (RSQ 0.55) were not acceptable and thus more samples and greater precision during measurement will be required in further investigations.