A study was conducted to identify chemical parameters and/or models for assessing Medicago sativa L. (L) hay quality, using near infrared reflectance spectroscopy (NIRS) analysis and Cornell Net Carbohydrate and Protein System (CNCPS) milk prediction as a criterion of accuracy. Milk yield (MY) derived from the CNCPS model, by replacing the average L hay in a complete diet with 168 representative South African L hay samples, was used as a criterion to evaluate and/or develop models for L hay quality grading. The best single predictor of MY was the acid detergent fibre (ADF) content of L hay, which explained 67% of the measured variation. A multiple linear equation (Y = 64.18 – b1ADF – b2ash – b3lignin) explains 96% of the measured variation in MY. The relatively poor performance of crude protein (CP) (r² = 0.04) and other protein related parameters (r²< 0.25; adjusted-crude protein, ADF-CP, neutral detergent fibre-CP and soluble protein) in predicting MY suggests that protein content of L hay is an unreliable indicator of L hay quality. It is clear that MY derived from the CNCPS model by replacing L hay in a basal diet with others in the South African L hay population can be significantly predicted with high accuracy by the developed empirical model named lucerne milk value (LMV) consisting of only ADF, ash and lignin.
"Experientia docet" - Experience is the best teacher