Prediction of protein in fresh leaves of alfalfa by NIRS with an interactance-reflectance probe

Abstract

The objetive of this work was to estimate the protein content of fresh leaves of alfalfa (Medicago sativa L.), using near infrared reflectance spectroscopy (NIRS) technology. The sample set included 33 varieties grown under irrigation in the in the province of Salamanca (Spain). Two cutting dates were considered, and samples of leaves were taken from three plant positions (apical, middle and basal). A total of 190 samples were obtained. The protein content ranged between 9.92-45.32% of dry matter. NIRS calibrations were developed by two regression methods (multiple linear regression-MLR and partial least squares regression-PLSR) and three mathematical treatments (log l/R, first and second derivatives). The best regression model reached a coefficient of multiple determination (r2) of 0.68 and a standard error of prediction (SEP) of 3.27% in validation. This study found that NIRS calibrations based on spectra of fresh leaves have potential for the rapid screening of crude protein in forage breeding programs.