Calibraciones de espectrofotometría de reflectancia en el infrarrojo cercano para carbohidratos en pasto Urochloa

The goal was to develop near infrared reflectance spectroscopy (NIRS) calibration models to estimate fiber and carbohydrate fractions content in Urochloa grasses grown in tropical Mexico. One hundred and eighty-nine samples of three Urochloa hybrids were cut at 35 days of regrowth for a year. Grasses were wet-chemistry analyzed for neutral detergent fiber (NDF), neutral detergent insoluble protein (NDIP), acid detergent fiber (ADF), acid detergent insoluble crude protein (ADIP), and lignin (LIG). Also, estimations of total carbohydrates (TC), fibrous carbohydrates (FC), non-fibrous carbohydrates (NFC), and undigestible carbohydrates (UC) were computed. Calibration and validation models were evaluated by their corresponding determination coefficients (R2); cross validation standard errors (SECV); and the residual prediction of the deviation (RPD). Besides, with an external set of Urochloa samples, a sensitivity analysis was performed using the mean square prediction error (MSPE) to evaluate the best calibration models. The NIRS calibration models developed achieved adequate adjustment and predictive efficiency for NDF (RPD 2.4) and TC (RPD 2.5) showing reliability to predict the fiber and carbohydrate fractions of Urochloa grasses. However, LIG, FC, NFC and UC calibration models require a larger number of samples to increase the prediction accuracy.