NIRS Prediction of Ingredient Composition in Compound Feedingstuffs
Exchange visit from Silvia Ampuero to University of Cordoba

The STSM of Silvia Ampuero from ALP in Switzerland was done in October 2011. Host was Anna Garrido-Varo,University of Córdoba in Spain. The topic was a collaborative study on NIRS Prediction of Ingredient Composition in Compound Feedingstuffs

Purpose of the visit
The goal of the present STSM was to explore the possibility to build classification models with NIR spectra for the prediction of ingredient composition in compound feedingstuffs via the combination of typical data bases from Spain and Zwitzerland. This project was motivated by the need of a fast and reliable method to ensure the compliance of a product declaration. Indeed, since 2005 semi-open declarations of compound feedingstuffs are obligatory in Switzerland and open declarations are required by European Commission since 2003.

Description of the work carried out during the visit and main results obtained
a) Standardization of NIR spectrometers. In order to combine spectra from the two institu-tions the first task was to standardize the instruments: Foss 6500 from the University of Cordoba (UCO) and Büchi NIRFlex N-500 from Agroscope Liebefeld-Posieux (ALP). Hence, a set of 35 Swiss samples of compound feed for swine was analyzed at ALP prior its transfer to Cordoba. These samples were also analyzed at UCO for the standardization of the Büchi spectra to the Foss (master). Following the visit to Cordoba the Foss spectra will also be standardized to the Büchi (master) for further modeling. For this purpose, the 35 samples were carefully packed at UCO under vacuum in aluminum bags to keep them as much as possible in the same state (humidity and particle size distribu-tion) during both measurements (at UCO and at ALP)
b) Harmonization and simplification of the reference data. The list of corresponding ingre-dients was simplified and harmonized according to a common definition.
c) Exchange of spectra and modeling on mixed data bases. Local algorithms were tested on spectra from around 450 Spanish and 35 Swiss samples using the existing UCO models. Further work is to be performed after the visit to improve classification rates.

Future collaboration with the host institution
A common interest on NIRS as a fast and reliable technique was evident. Hopefully, this short project will also help in revealing the importance of using combined databases, especially in today's global market situation where more and more often feed ingredients come from abroad.

Projected publications/articles resulting or to result from the STSM
A collaborative joint publication is foreseen in the near future on the outcomes of the project.

Other comments
The visit was greatly interesting due to the extended experience of UCO on NIRS predictions and especially of Profs Ana Garrido-Varo and Dolores Pérez Marín. Thus, different applications of NIRS could be discussed and many helpful practical techniques could be demonstrated fulfilling one of the principal goals of the Feed for Health COST Action STSM.

Silvia Ampuero STSM
Design & Programming: Zenlex