Objective chemical fingerprinting of oil spills by partial least-squares discriminant analysis

An objective method based on partial least-squares discriminant analysis (PLS-DA) was used to assign an oil lump collected on the coastline to a suspected source. The approach is an add-on to current US and European oil fingerprinting standard procedures that are based on lengthy and rather subjecti...

Full description

Saved in:
Bibliographic Details
Published inAnalytical and bioanalytical chemistry Vol. 403; no. 7; pp. 2027 - 2037
Main Authors Gómez-Carracedo, M. P., Ferré, J., Andrade, J. M., Fernández-Varela, R., Boqué, R.
Format Journal Article
LanguageEnglish
Published Berlin/Heidelberg Springer-Verlag 01.06.2012
Springer
Subjects
Analysis
Analytical Chemistry
Biochemistry
Characterization and Evaluation of Materials
Chemistry
Chemistry and Materials Science
Chromatography, Gas
Coastlines
Discriminant Analysis
Fingerprinting
Food Science
Gas Chromatography-Mass Spectrometry
Indexes (ratios)
Laboratory Medicine
Least squares method
Least-Squares Analysis
Mathematical analysis
Mathematical models
Monitoring/Environmental Analysis
Oil spills
Original Paper
Petroleum - analysis
Petroleum Pollution
ANE, Spain, Galicia
Hydrocarbons
Discriminant PLS
Oil spills
Variable selection
Online AccessGet full text

Cover

More Information
Summary:An objective method based on partial least-squares discriminant analysis (PLS-DA) was used to assign an oil lump collected on the coastline to a suspected source. The approach is an add-on to current US and European oil fingerprinting standard procedures that are based on lengthy and rather subjective visual comparison of chromatograms. The procedure required an initial variable selection step using the selectivity ratio index (SRI) followed by a PLS-DA model. From the model, a “matching decision diagram” was established that yielded the four possible decisions that may arise from standard procedures (i.e., match, non-match, probable match, and inconclusive). The decision diagram included two limits, one derived from the Q-residuals of the samples of the target class and the other derived from the predicted y of the PLS model. The method was used classify 45 oil lumps collected on the Galician coast after the Prestige wreckage. The results compared satisfactorily with those from the standard methods.
Bibliography:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 23
ISSN:1618-2642
1618-2650
DOI:10.1007/s00216-012-6008-5