Near infrared spectroscopy predictions on heterogeneous databases : an example of plants from sub-alpine meadows and shrublands

Lembke Manuel, Bastianelli Denis, Dobremez Jean-François, Hervouet Catherine, Bonnal Laurent, Maillard Daniel. 2007. Near infrared spectroscopy predictions on heterogeneous databases : an example of plants from sub-alpine meadows and shrublands. In : Near infrared spectroscopy : Proceedings of the 12th International Conference, Auckland, New Zealand, 9th - 15th April 2005. Burling-Claridge G.R. (ed.), Holroyd S.E. (ed.), Sumner R.M.W. (ed.). NIRCE. Chichester : IM Publications, 99-100. ISBN 978-0-473-11646-0|978-0-473-11746-7 International Conference on Near Infrared Spectroscopy. 12, Auckland, Nouvelle-Zélande, 9 Avril 2005/15 Avril 2005.

Résumé : The aim of this study was to describe the diet habitat of a mountain ungulate species: Capra i. ibex through the vegetation characteristics. The sub-alpine vegetation is composed by a great number of species that forms a heterogeneous cover. This cover constitutes a great variable diet resource for ibex. An analysis for all species forming the plant cover was impossible, as there are 91 determined species. Moreover there is only a relative knowledge of the diet selection of the ibex in the study site, so the approach is to assess nutritional quality of the plant cover as a whole through it chemical composition. Plant samples from various places throughout the study zone were collected during 2 years, resulting in nearly 500 samples. Samples were dried, ground and spectra in diffuse reflectance were measured on a FOSS NIRSystem 6500 spectrometer in small ring cups. Chemical analysis of samples was performed on 150-200 samples selected for their spectral representativeness of the whole database: the parameters measured were Crude Protein (CP), Acid Detergent Lignin (ADL) and in vitro organic matter digestibility (OMD). Calibration equations were built after mathematical pre-processing of data (SNV and detrend with 2nd derivative of spectra). Partial Least Squares regression (modified PLS in WINISI software) was found to be the most efficient method for calibrations. General equations had SECV values of 0.52% for CP, 1.37% for ADL and 2.73% for OMD, leading to RPD (=SD/SECV) values of 7.7, 6.1 and 5.2 respectively. However the choice of computing general equations can be questioned since plant covers can be roughly split into 2 categories: herbaceous (HB) or shrubby (SH), which differ strongly in term of chemical composition and nutritional value. Therefore it was decided to test different equations for each of these categories. Samples were affected to a category on the basis of their known botanical composition. However a part of samples could not be classified a priori. Therefore a Factorial Discriminant Analysis was performed on spectral data for samples of known category, and the equation was applied to unknown samples to classify them. Then calibrations were run for HB and SH subsets. SECV values for CP, ADL and OMD were respectively 0.50%, 1.43%, 3.43% for HB and 0.44%, 1.30%, 2.42% for SH. The conclusion of this study is that NIRS analysis was an efficient tool to predict the composition of our heterogeneous plant databases. NIR spectra were also useful for sample classification. Moreover the splitting of databases into different categories did not improve predictions, which suggests to keep the use of a general database.

Mots-clés Agrovoc : habitat, régime alimentaire, chèvre, végétation alpine, composition chimique

Classification Agris : L20 - Écologie animale

Auteurs et affiliations

• Lembke Manuel, ONCFS (FRA)
• Bastianelli Denis, CIRAD-EMVT-UPR Systèmes d'élevage (FRA) ORCID: 0000-0002-6394-5920
• Dobremez Jean-François, Université de Savoie (FRA)
• Hervouet Catherine, CIRAD-EMVT-UPR Systèmes d'élevage (FRA)
• Bonnal Laurent, CIRAD-EMVT-UPR Systèmes d'élevage (FRA) ORCID: 0000-0001-5038-7432
• Maillard Daniel, ONCFS (FRA)

Source : Cirad - Agritrop (https://agritrop.cirad.fr/531012/)

Métriques

[ Page générée et mise en cache le 2024-04-01 ]