Identification of sources of diesel oil spills using parallel factor analysis: A bridge between American society for testing and materials and Nordtest methods

Abstract

American Standards for Testing and Materials method (ASTM5739-00) and Nordtest methodology, as the two major approaches for identifying the source of spilled oils using gas chromatography-mass spectrometry (GC-MS) data, are critically compared and a new method based on multi-way parallel factor analysis (PARAFAC2) is proposed. The new approach exploits both ASTM and Nordtest methodologies by using the entire extracted ion chromatogram (EIC) and taking into account the concentration diversities of different compound classes, respectively. A multi-way data preprocessing is proposed to preserve the diagnostic properties of the original GC-MS data, which are destroyed in the ASTM method by normalizing the EICs individually. Petroleum oils, in particular diesel oils, that are difficult to classify using current methods are shown to be excellent candidates for PARAFAC2 in which EIC matrices of different sizes can be analyzed simultaneously. A diesel oil sample from an oil spill and seven very similar suspect diesel source oils, which had undergone controlled weathering for 2-15 days, were compared by this method. 79% of pairwise group comparisons were separated, in contrast to the method in which EICs were each normalized to 100, which gave 32% separation of the comparisons. (c) 2008 Elsevier B.V. All rights reserved.