Development of a fast analytical tool based on NIR spectroscopy and pattern recognition techniques for enviromental forensic studies

YPF has developed substantial expertise in the use of NIR technology for many applications of petroleum products. The purpose of this study is to take advantage of this successful experience and extend its field of application to environmental forensic studies.
The presented work is an attempt to apply an alternative fast method for identification of potential oil spillages based on NIR Spectroscopy and pattern recognition techniques.
We collect 275 samples of potential sources from typical petrochemical streams. The NIR spectra were acquired using a NIR spectrophotometer (Brimrose AOTF Luminar 5030) equipped with a fiber optic sampling probe for transmittance measurements within 1050 and 1650 nm and 1 nm wavelength accuracy. The raw spectra are converted to their first derivative spectra and the chemometrics analysis was performed using Unscrambler X and SIMCA P+.
The analysis by PCA shows a clear separation into two clusters with four sub clusters. The first component discriminates based on aromatic content of samples, and the second component allows clustering based on distillation range. This discrimination was confirmed by the analysis for hierarchic grouping and PLS-DA, the model was validated with weathered references samples.
The assessment with NIR in combination with chemometrics techniques helps to reveal in a more time-effective manner, similarities between unknown samples and potencial sources. The conventional approach by physicochemical and gas chromatography analysis is labor and time consuming. However, the fingerprint of some samples with strongs weathered effects were not sufficient close to spectra source assignment to permit conclusive identification. This particulate case would require additional analysis or evaluate the predicted ability of a new model with weathered representative sample.