| Literature DB >> 18998694 |
Tony Woodcock1, Gerard Downey, Colm P O'Donnell.
Abstract
The potential of near-infrared transflectance spectroscopy (1100-2498 nm) combined with chemometric techniques to confirm the geographical origin of European olive oil samples was evaluated. In total, 913 extra virgin olive oil samples (210 Ligurian and 703 non-Ligurian) were collected over three consecutive harvests (2005, 2006, and 2007). A multivariate spectral fingerprint for Ligurian olive oil was developed and deployed to confirm or refute a claim that any given sample was Ligurian. Samples were pseudorandomly split into calibration (n = 280) and validation sets (n = 633); the only selection constraint applied was to insist on equal numbers of Ligurian and non-Ligurian samples in the calibration set. Following preliminary examination by principal component analysis, the full spectrum modeling method applied to the spectral data set was discriminant partial least-squares regression; various data pretreatments were also investigated. The best models correctly predicted the origins of samples in the prediction set up to 92.8 and 81.5% for Ligurian and non-Ligurian olive oil samples, respectively, using a first-derivative data pretreatment. The potential of this approach in commercial traceability and quality assurance schemes is noted.Entities:
Mesh:
Substances:
Year: 2008 PMID: 18998694 DOI: 10.1021/jf802792d
Source DB: PubMed Journal: J Agric Food Chem ISSN: 0021-8561 Impact factor: 5.279