Predicting Forage Nutritional Quality With Near-Infrared Spectroscopy

Abstract

The quality of green forage is crucial in pasture grazing, influencing both animal welfare, environmental sustainability, and production yield. Traditionally, the evaluation of forage composition requires time-consuming and costly chemical analysis. In this context, near-infrared spectroscopy (NIR) emerges as a promising alternative. This study adopted Fourier transform NIR (FT-NIR) spectroscopy to predict nutritional characteristics of green forages. A total of 324 samples were collected from pastures in central Italy. Partial least squares (PLS) regression models were then developed, applying variable selection methods to improve PLS model accuracy. The interval PLS (iPLS) variable selection method gave the best results for fresh forage, while the genetic algorithm (GA) performed best for dried samples. The best results from the PLS models were obtained for dry matter (DM) and crude protein (CP). The DM model for fresh forage yielded an R²_P of 0.96 and an RMSEP of 2.95 g 100 g⁻¹ FW, while the CP model for dried forage yielded an R²_P of 0.94 and an RMSEP of 1.84 g 100 g⁻¹ DW, with a normalised root-mean-square error of cross-validation (NRMSECV) of 3.8% and 5.6%, respectively. The results for the neutral detergent fibre (aNDF) were acceptable. NIR spectroscopy has proven to be a useful tool for assessing forage nutritional quality. Variable selection through iPLS also enabled the identification of “core” spectral regions for the development of compact and portable NIR sensors. Future research should further investigate sample preparation and moisture content effects and expand sampling to different geographical areas to enhance model robustness.