Abstract :
[en] Forage quality or nutritive value is related to chemical composition, which can be determinated by laboratory methods. The NIR technique in comparison with classical methods is non-destructive, non-polluting, fast and relatively inexpensive per analysis. Investigations on nutritional quality of Carpathians Apuseni Mountains (Romania) grasslands are rarely performed with NIR technique. Therefore, the objective of the thesis was to develop non-destructive methods for evaluating the quality of feed originating from the Gârda area of the Carpathians Apuseni Mountains (Romania) potentially and to similar grassland arround the world. The first task was to study the potential of NIR spectroscopy for building a spectral database for forage quality based on a large collection of semi-natural grassland samples, using a ‘local’ calibration model built by the Walloon Agricultural Research Centre (CRA-W), in Belgium, to determine various parameters (e.g., protein, dry matter, ash, fibre, fat, aNDFom, ADF, lignin, digestibility, crude energy) from samples collected worldwide, outside Romania. The second task was to develop calibration models for an NIR-HSI system, which involved larger spectral data registration as an image. Until now, analyses to determine plant species were based on botanical composition evaluation, including visual observation, which is a subjective method involving identifying plants directly in the field. Distinguishing samples of pure grassland species can be time consuming, and it was therefore decided to build a spectral database of pure samples and then discriminate these samples into binary and ternary artificial sample mixtures. The main objective of these tasks was to identify the botanical families to which the samples belonged (Poaceae, Fabaceae and Other Botanical Families [OBF]). The focus was not on quantity monitoring, but rather on determining forage quality from stationary experiments in the grasslands. To conclude, this research has shown that it is possible to develop calibration models not only for quality assessment, but also for sample discrimination in dry powder samples. It was intended, that the mathematical models constructed and the database obtained, would be used for future research.
