Quantitative multicomponent analysis of polycyclic aromatic hydrocarbons in water samples

A comparative study between three different multivariate calibration methods, classical least squares (CLS), principal component regression (PCR) and partial least squares regression (PLSR) was carried out. The calibration models for all three methods were obtained from a combination of two synchronous fluorescence spectra (recorded at 50 and 100 nm wavelength increments) for each standard of a calibration set of 70 standards, each containing ten polycyclic aromatic hydrocarbons (anthracene, benz[a]anthracene, benzo[a]pyrene, chrysene, fluoranthene, fluorene, naphthalene, perylene, phenanthrene and pyrene). The predictions of the model were compared with the relative root mean squared difference (RRMSD) obtained from the results of an external validation set, formed by 15 independent mixtures. Finally, the PLSR and PCR models were used for the determination of the above mentioned PAHs in spiked natural water samples at concentration levels between 4 and 20 ng ml(-1). Recoveries ranged from 80 to 120% in most cases, although fluorene gave significantly lower results. (C) 1998 Elsevier Science B.V.