Prediction of Chemical Composition in Brachiariagrass by Near Infrared Reflectance Spectroscopy

Abstract

Near Infrared Reflectance Spectroscopy (NIRS) was used to predict chemical composition of Brachiariagrass (192 samples from eight genotypes). 192 samples were divided two groups, one group (n＝128) was selected for calibration development, and the remaining group (n＝64) was used for equation validation. The accuracy of calibration from partial least square regression (PLSR) and multiple linear regression (MLR) was evaluated by the correlation coefficients (r) and the standard deviation of prediction (SDP). Furthermore, the accuracy of prediction was evaluated by the evaluation index (EI). The ‘r’ and SDP of moisture, crude protein (CP), crude fiber (CF), neutral detergent fiber (NDFom), acid detergent fiber (ADFom), acid detergent lignin (ADL) and in vitro dry matter digestibility (IVDMD) from PLSR method were 0.91-0.99 and 0.25-2.38 respectively. “EI” of moisture and CP exhibited “A rank”, however the other chemical compounds were “B rank”. These results indicated that moisture, CP, CF, NDFom, ADFom, ADL and IVDMD contents in Brachiariagrass can be predicted by using NIRS techniques with an acceptable accuracy.