Oil spill forensics - Identification of Sources for Illegal Oil Spills. A New Approach Combining Multivariate Statistics and Gas Chromatography and Mass-Spectrometry (GC-MS)
MetadataVis full innførsel
- Institutt for kjemi 
The CEN/TR 15522-2:2012 standard is most widely used in Europe for oil spill fingerprinting purposes, which is based on comparison of specific source biomarkers by GC MS. This thesis was mainly aimed at establishing and proposing alternative chemometric methods for oil spill fingerprinting, i.e., principal component analysis (PCA) and hierarchical cluster analysis (HCA), which could perform better or at least similar with the univariate CEN-method. This is important because CEN-method often depends on the skill and expertise of the analyst (i.e., person who perform the analysis). The method is also slow and this lead to corresponding increase in man-hour costs. Oil spill fingerprinting using CEN-method performed on 45 samples (oil and bird samples) identified them as match (26), probable match (7), non-match (7) and 4 were used as references. Compared to SINTEF s result, the results obtained in this thesis deviated to some extent, which shows the significance of the analyst s skill and experience when applying this standard methodology. PCA and HCA were applied to an established dataset from a previous oil spill, the MS Server, obtained from SINTEF. PCA performed very well in identifying spilled oil samples that were not from the suspected source as outliers; however, it failed in explicitly classifying spilled samples that were identified to be similar with the suspected sources. HCA, on the other hand, was more powerful not only in classifying spilled oil samples as similar and not similar (when compared to candidate sources), but also by explicitly grouping those similar spilled oil samples according to the relative degree of weathering they experienced. In general, of the 19 spilled oil samples studied using PCA and HCA, 6 samples using PCA and 5 samples using HCA were identified as not similar to the candidate source samples. Compared to the CEN-methodology, the overall performance of both chemometric methods investigated in this thesis is found to be promising. Both methods demonstrated to be faster and also merely dependent on the skill level and experience of the analyst. This thesis has demonstrated that PCA and HCA can be used as complementary methods to boost the defensibility of the univariate CEN-method. Often, establishing a conclusion based on a study carried out on real oil spill incident, where everything tends to be uncontrollable, is not sufficient. In order to develop these chemometric methods as alternative to CEN-method, further research using (more) controllable samples is recommended.