ABSTRACT
Due to the recurrent small spills, oil pollution along coastal regions is still a major environmental issue. Standardized oil fingerprinting techniques are useful for oil spill identifications, but time- and resource-consuming. There have been ongoing needs for simple yet rapid approach for field screening of oil spill. Laser induced fluorescence (LIF) technology can be incorporated into a spectrometer, and with the integration of chemometrics can be consolidated as a potentially useful portable oil type classification device. Using a LIF spectrometer, 775 oil spectra were calibrated into supervised classification models and validated with 162 oil spectra. Reliability of the device to accurately remove background emission from fluorescence spectra was verified. Prediction performance and model robustness were further validated by comparison between commonly used classification models such as partial least square discriminant analysis (PLS-DA) and support vector machine-discriminant analysis (SVM-DA). Robustness in both models were comparable with PLS-DA having a lower number of misclassification (PLS-DA: 5.50%, SVM-DA: 13.8%) while SVM-DA having a lower number of unassigned samples (PLS-DA: 10.9%; SVM-DA: 16 1.39%). This study explicitly demonstrated the development of a new convenient and handy device which can be used as part of the screening process for oil spill fingerprinting.
