Determination of Fenitrothion in Water by Near Infrared Spectroscopy and Chemometric Analysis

Abstract

The feasibility of the determination of fenitrothion in water is reported using two-dimensional correlation spectroscopy to select the modeled spectral regions for near-infrared spectroscopy. Partial least squares discriminant analysis (PLSDA) and least squares support vector machines discriminant analysis (LSSVM) were used to model and predict the spectra of fenitrothion aqueous solutions. Spectral regions from 5365 to 6757 per centimeter and 6944 to 7800 per centimeter were selected according to the dynamic spectra and the two-dimensional correlation synchronous spectra. Four classification methods (random division 1, random division 2, Kennard-Stone 1, and Kennard-Stone 2) were used to divide the samples into calibration and validation sets. The classification methods impacted both PLSDA and LSSVM, but the impact was larger on PLSDA than on LSSVM. In addition, different pretreatment methods (mean center, standard normal variate transformation, and orthogonal signal correction) were applied to PLSDA; the pretreatments also affected the predictions. Both PLSDA and LSSVM models were used to characterize real water samples and yielded satisfactory results. These methods allowed the direct and rapid determination of fenitrothion in water.