Development of NIR calibration valid for two different grass sample collections

V. Míka, J. Pozdíšek, Peter Tillmann, P. Nerušil, K. Buchgraber, L. Gruber

2003 · 18 citations · 2 references

Concepts

Abstract

e prediction equation of parameters of forage nutritive value, derived entirely from Czech sam- ples, was not suitable for sufficiently accurate evaluation of the nutritive quality of Austrian samples. However, when it was extended by randomly selected 171 Austrian samples, it became more robust and the standard error of prediction decreased significantly (e.g. from 0.826 to 0.321 MJ NEL/kg DM). It can be used for samples from larger geographical and climatic areas or under different grassland management practices with various proportions of grasses, legumes and other herbs. ere was not any consistent improvement when NEL values were calculated from partial prediction of constituents in comparison with direct prediction of NEL by NIRS. e r values were 0.866 vs. 0.876 in the case of Czech validation set and 0.852 vs. 0.811 for the Austrian validation set.

References

2