The quantification of pollutants in drinking water by use of artificial neural networks

Drinking water attained from aquifers (ground water) is susceptible to contamination from a wide variety of sources. The importance of ensuring that the water is of high quality is paramount. Multivariate calibration in conjunction with analytical techniques can assist in qualifying and quantifying a wide range of pollutants. These can be divided into two types: inorganic and organic. The former typically includes heavy metals such as cadmium and lead; the latter includes a range of compounds such as pesticides and by-products of industrial processes such as oil refining. This article presents the application of the well known nature-inspired paradigm of artificial neural networks (ANNs) for the quantitative determination of inorganic pollutants (namely cadmium, lead and copper) and organic pollutants (namely anthracene, phenanthrene and naphthalene) from multivariate analytical data acquired from the samples. The success of the determination of the pollutants via ANNs is reported in terms of the overall root mean square error of prediction (RMSEP) which is an accepted measure of the difference between the predicted concentrations and the actual concentrations. The work represents a good example of nature-inspired methods being used to solve a genuine environmental problem.