Prediction performances of portable near infrared instruments for at farm forage analysis

The objective of this study was to evaluate the use of Near Infrared Spectroscopy (NIRS) to analyze maize silage with a portable instrument. The instrument was a Zeiss Corona 45 working between 960 and 1700 nm which was used in Italy, Czech Republic and Poland. Best prediction performances were obtained using the Italian data set. Prediction error were 1.0, 0.16 and 0.4 respectively for DM, CP and NDF on a as is basis. With the instrument from Poland and Czech Republic there were lower accuracy of prediction compared to the Italian dataset, probably for their limited (less than 100 samples) calibration data set. Merging all the data set improved prediction accuracy for CP but not DM. It would appear that some form of instrument standardization is needed before merging data set