Data analysis tools for safe drinking water production

Abstract

Providing safe and high quality drinking water is essential for a high quality of life. However, the water resources in Europe are threatened by various sources of contamination. This has led to the development of concepts and technologies to create a basis for provision of safe and high quality drinking water, which had thus resulted in the formation of the Artificial Recharge Demonstration project (ARTDEMO). The overall aim of this thesis in relation to the ARTDEMO project was to develop a realtime automated water monitoring system, capable of using data from various complementary sources to determine the amounts of inorganic and organic pollutants. The application of multivariate calibration to differential pulse anodic stripping voltammograms and fluorescence spectra (emission and excitation-emission matrix) is presented. The quantitative determination of cadmium, lead and copper acquired on carbon-ink screen-printed electrodes, arsenic and mercury acquired on gold-ink screen-printed electrodes, in addition to the quantitative determination of anthracene, phenanthrene and naphthalene have been realised. The statistically inspired modification of partial least squares (SIMPLS) algorithm has been shown to be the better modelling tool, in terms of the root mean square error of prediction (RMSEP), in conjunction with application of data pre-treatment techniques involving rangescaling, filtering and weighting of variables. The % recoveries of cadmium, lead and copper in a certified reference material by graphite furnace atomic absorption spectrometry (GF-AAS) and multivariate calibration are in good agreement. The development of a prototype application on a personal digital assistant (PDA) device is described. At-line analysis at potential contamination sites in which an instant response is required is thus possible. This provides quantitative screening of target metal ions. The application imports the acquired voltammograms, standardises them against the laboratory-acquired voltammograms (using piecewise direct standardisation), and predicts the concentrations of the target metal ions using previously trained SIMPLS models. This work represents significant progress in the development of analytical techniques for water quality determination, in line with the ARTDEMO project's aim of maintaining a high quality of drinking water.