Abstract
This thesis investigated the potential application of Raman and IR vibrational spectroscopic techniques combined with multivariate analysis approaches for the detection and quantification of fatty acid, astaxanthin and adulterant content in marine derived lipids (krill oil, cod liver oil, salmon oil), and descrimination of lake (Lindavia intermedia) snow from other invasive algal species (Nostoc sp., Spirogyra, sp., Zygnema sp. and Didymosphenia geminata).
The potential of data-fused Raman and IR spectral models were explored in the first experiment (Chapter 3) to evaluate whether data-fused Raman and IR spectral models provided any advantage over the individual techniques for multi-component (omega-3 polyunsaturated fatty acids and astaxanthin) quantitation in commercial krill oil capsules. The spectral measurements were carried out for predicting EPA+DHA and astaxanthin content directly through krill oil capsules. Reference data for EPA+DHA and astaxanthin PLSR models were compared against the data obtained through GC-MS and UV-Vis spectrophotometry, respectively. Raman spectroscopy was capable of quantitating astaxanthin in krill oil and encapsulated krill oil, whereas IR spectroscopy was not. Raman spectroscopy also out-performed IR spectroscopy for quantitation of EPA+DHA in commercial krill oils, and had the added benefit of directly analysing encapsulated oils. Low- and mid-level fused Raman and IR spectroscopic data models (r2 = 0.99; RMSEP = 1.4%; r2 = 0.99; RMSEP = 1.2%) resulted in more accurate EPA+DHA concentration predictions than Raman (r2 = 0.90; RMSEP = 4.5%) or IR (r2 = 0.73; RMSEP = 7.3%) spectroscopy alone. However, data fusion did not improve the quantitative astaxanthin models over Raman alone.
In Chapter 4, two complementary vibrational spectroscopic methods (Raman and IR) and the low-level fusion of these datasets were used to identify the presence of adulterants (palm oil, PO; ω-3 concentrates in ethyl ester, O3C and fish oil, FO) in krill oil. This chapter successfully quantified the amount of adulterant present and identified the most promising spectroscopic technique or combination thereof for the quantification of adulterant(s) in krill oil. IR spectroscopy performed better (r2 = 0.96; RMSEP = 3.8%) than Raman spectroscopy (r2 = 0.85; RMSEP = 7.6%) for quantification of PO adulteration fortified at different concertation range (1-50%) in krill oil. Raman and IR spectroscopy showed satisfactory prediction of O3C adulterants in krill oil whereas poor correlation was obtained for detecting fish oil adulteration. The low-level fused data model outperformed individual techniques for the concentration of PO (r2 = 0.96, RMSEP = 3.7%), O3C (r2 = 0.96, RMSEP = 3.4%) and FO (r2 = 0.83, RMSEP = 7.7%) as an adulterant in krill oil.
This approach was extended for the detection of the adulterants (PO, O3C, FO) in other valuable marine oil samples (cod liver oil, CLO and salmon oil, SO) in Chapter 5. A global model was developed using multiple oil samples (CLO and SO together) to detect and quantify adulterants (vegetable oil or cheap fish oil) in the cod liver oil and salmon oil samples compared to individual models developed using each individual valuable marine oil type. The global model with low-level fusion detected and quantified all adulterants with the RMSEP value less than 4% (PO: r2 = 0.96, RMSEP = 3.9%; O3C: r2 = 0.99, RMSEP = 2.4%), which is comparable with the individual model developed using cod liver oil (PO: r2 = 0.98, RMSEP = 2.5%; O3C: r2 = 0.99, RMSEP = 1.5%) or salmon oil (PO: r2 = 0.95, RMSEP = 4.1%; O3C: r2 = 0.99, RMSEP = 1.9%). However, the accuracy of quantifying FO adulteration was not as good as PO and O3C adulteration regardless of whether the global (r2 = 0.77, RMSEP = 8.6%) or individual models (CLO: r2 = 0.77, RMSEP = 8.5%; SO: r2 = 0.79, RMSEP = 8.7%) were used. Overall, this study demonstrated that the global model with data fusion would be useful for the detection of PO and O3C adulterant in CLO and SO samples.
Chapter 6 investigated the ability of Raman and IR to quantify fatty acid concentrations in a wide range of unknown oil samples, where the Raman and IR data was compared and validated against the data obtained from GC-MS. Raman (r2 = 0.94%; RMSEP = 2.4%) and IR (r2 = 0.95%; RMSEP = 2.3%) showed comparable results for predicting ω-3 fatty acids, whereas low-level fusion significantly improved the model performance and predictive accuracy for quantifying ω-3 fatty acids (r2 =0.96%; RMSEP = 1.9%), PUFA (r2 =0.83%; RMSEP = 4.0% and SFA (r2 =0.79%; RMSEP = 4.1%) content. IR and low-level fusion also gave comparable prediction results for EPA+DPA+DHA, DHA and MUFA quantification. The EPA and DHA are similar in structural and provide spectral homology, which makes it difficult to quantitate independently in oils containing both compounds. This study concludes that the data fusion of IR and Raman data could be a viable way to overcome the analytical challenge.
The last experimental chapter of this thesis investigated the potential application of vibrational spectroscopy (Raman and IR) to differentiate lake snow from other New Zealand algal species found in New Zealand lakes based on spectral signature difference due to the compositional changes of lake snow growing at different sampling locations and depth. Raman spectroscopy in combination with two different chemometric techniques (support vector machine classification, SVM and principal component analysis-linear discriminate analysis, PCA-LDA) were successfully utilised for rapid discrimination of five algal taxa (L. intermedia, D. geminata, Nostoc, Spirogyra, Zygnema). In addition, Raman spectroscopy showed a linear relationship (r2 =0.93%; RMSEP = 0.002% and r2 =0.70%; RMSEP = 0.005%) against reference data obtained from different loading concentration of suspended lake snow. This demonstrated the potential for future work to develop a Raman system for continuous monitoring of water bodies for both detection of algal blooms and EPS production events with discrimination of the algae responsible.
Overall, this thesis contributes a clear insight about the potential application of Raman and IR spectroscopy for the qualitative and quantitative assessment of marine derived lipid and invasive algal species.