Abstract
The mineral composition of corn silage is of paramount importance for precision nutrition in dairy cattle and other ruminants, but its determination depends on laborious, time-consuming and expensive reference methods. This work evaluated, for the first time, the potential of Vis-NIR spectroscopy for the quantification and discrimination of 17 mineral elements (Li, Na, Mg, P, S, K, Ca, Mn, Fe, Co, Cu, Zn, As, Mo, Cd, I, and Pb) in both fresh and dried–ground corn silage samples. Partial least squares (PLS) regression was used for quantification models, while partial least squares–discriminant analysis (PLS-DA) was applied to discriminate samples according to their mineral content. Drying and grounding improved the performance of the quantification models but showed a limited effect on the discrimination models. The best PLS models were developed using the processed samples (dried and ground) for the quantification of Li, P and Fe, with RER and RPD values of 10.1 and 2.2 for Li, 12.0 and 2.3 for P and 10.0 and 1.9 for Fe, respectively, demonstrating an accurate prediction. Most of the remaining PLS models can be used for screening purposes, as the R2P and RER were higher than 0.65 and 4, respectively. The PLS-DA results further demonstrated the suitability of Vis-NIR spectroscopy for screening mineral content in corn silage, with most models achieving prediction accuracies above 75% for processed samples and above 70% for fresh samples. These results demonstrate that Vis-NIR spectroscopy provides a rapid, less labor-intensive, environmentally friendly, and cost-effective approach for mineral screening in corn silage. This methodology enables precision mineral nutrition through the optimization of feed formulation, ensuring healthy animal growth and reducing the environmental footprint connected with mineral supplementation.