Application of multivariate analysis to the turbidimetric determination of sulphate in seawater

Abstract

Spectroscopic techniques are widely used in the field of analytical chemistry for the determination of a huge number of analytes. The use of diode array spectrophotometers involves the possibility of collecting the whole UV-Vis spectrum, which provides an opportunity to perform a multivariate analysis of the samples. The effectiveness of the use of multivariate data analysis against the univariate counterpart (which is the most common for classic spectroscopic methods) is demonstrated through application in the sulphate determination in seawater samples using a modified turbidimetric method. The original method recommends performing the analysis at 420 nm, this being a univariate analysis. The modification presented uses multivariate methods for the calibration with the whole UV-Vis spectrum from 200 to 800 nm instead of the single measurement at 420 nm. The external calibration shows that there is a matrix effect which can compromise the accuracy of the measurements, so that an in-house Iterative Multivariate Standard Addition Method (IMSAM) was successfully tested in order to avoid this effect. As the spectrum interval was too wide, both Genetic Algorithm (GA) and interval Partial Least Square (iPLS) regression methods were used in order to select the best wavelengths for the regression and back prediction of the samples. The results obtained in this work show that the performance of the method at 420 nm, as the original method recommends, leads to a considerable error and also that the best segment of the spectra for the determination of the sulphate concentration in seawater is from 600 to 800 nm.